US 2006O167784A1 (19) United States (12) Patent Application Publication (10) Pub. No.: US 2006/0167784 A1 Hoffberg (43) Pub. Date: Jul. 27, 2006
(54) GAME THEORETIC PRIORITIZATION Related U.S. Application Data SCHEME FOR MOBILEAD HOC NETWORKS PERMITTING HERARCHAL (60) Provisional application No. 60/609,070, filed on Sep. DEFERENCE 10, 2004. (76) Inventor: Steven M. Hoffberg, West Harrison, Publication Classification NY (US) (51) Int. Cl. G06O 40/00 (2006.01) Correspondence Address: (52) U.S. Cl...... T05/37 Steven M. Hoffberg, Esq. (57) ABSTRACT MILDE & HOFFBERG, LLP Suite 460 A method for providing unequal allocation of rights among 10 Bank Street agents while operating according to fair principles, com White Plains, NY 10606 (US) prising assigning a hierarchal rank to each agent; providing a synthetic economic value to a first set of agents at the a high level of the hierarchy; allocating portions of the Syn (21) Appl. No.: 11/005,460 thetic economic value by the first set of agents to a second set of agents at respectively different hierarchal rank than the first set of agents; and conducting an auction amongst agents (22) Filed: Dec. 6, 2004 using the synthetic economic value as the currency.
Time Update ("Predict") Measurement Update ("Correct")
(1) Project the state ahead (1) Compute the Kalman Gain
x = AX-1 + Buk K = P, HT (HP, H+R) (2) Project the error covariance ahead (2) Update estimate with
P = AP A+ Q measurement I k Xk = x + K. (Ik- HS)
(3) Update the error covariance
P (I- KH) P.
Initial estimates for X-1 and P-1 Patent Application Publication Jul. 27, 2006 Sheet 1 of 6 US 2006/0167784 A1
Patent Application Publication Jul. 27, 2006 Sheet 2 of 6 US 2006/0167784 A1
O1 O2
HMM with mixture of Gaussians output Fig. 3
A1 A2
O1 O2 B1 B2
input-output HMM Factorial HMM Coupled HMM Fig. 4A Fig. 4B Fig. 4C Patent Application Publication Jul. 27, 2006 Sheet 3 of 6 US 2006/0167784 A1 u-N Time Update Measurement Update ("Predict”) ("Correct")
Fig. 5
Time Update ("Predict") Measurement Update ("Correct") (1) Project the state ahead (1) Compute the Kalman Gain * = Axk-1 + Buk K = P, HT (HP, H+R) (2) Project the error covariance ahead (2) Update estimate with P AP. A+ Q measurement / S = X + K. (lk - HX)
(3) Update the error covariance
P = (1 - KH) P.
Initial estimates for X-1 and P-1 Fig. 6 Patent Application Publication Jul. 27, 2006 Sheet 4 of 6 US 2006/0167784 A1
Time Update ("Predict") Measurement Update ("Correct") (1) Project the state ahead (1) Compute the Kalman Gain x = fix. , uk, O) K = P. H. (H.P.H + W.R.V.) (2) Update estimate with (2) Project the error covariance ahead measurement Ik P = AP. A + W.Qk-W". Xk Xk H K (Ik- h(xk y O))
(3) Update the error Covariance
P = (1 - KH) P.
Initial estimates for X-1 and P
Fig. 7 Patent Application Publication Jul. 27, 2006 Sheet 5 of 6
„?Ilossebold„Ilosseoold;
DE!¡EE|--|--
Josseoold 9 Patent Application Publication Jul. 27, 2006 Sheet 6 of 6 US 2006/0167784 A1
Proximity 101 (x) w
Prospective Conjunction 102 (x) W2
Type of event 103 (x) W3
Type of event and sensed Condition 104. (x) W4 Fig. 9
High Memory dump Unit D Location Codes priority Itinerary string 201 202 203 messages 205 206 204
Event 301 Location 302 Time 303 Source 304 Expiration 305 Reliability 306 Message 307 Fig. 10 US 2006/0167784 A1 Jul. 27, 2006
GAME THEORETIC PRIORITIZATION SCHEME cations and display rendering considered normal. Typically, FOR MOBILEAD HOCNETWORKS PERMITTING both client and server are connected to the Internet through HERARCHAL DEFERENCE Internet Service providers, each having its own router. 0006. It is also known to provide so-called proxy servers FIELD OF THE INVENTION and firewalls, which are automated systems that insulate the 0001. The present invention relates to the field of ad hoc client system from the Internet. Further, so-called Internet network protocols and control architectures. applications and applets are known which provide local intelligence at the client system. Further, it is known to BACKGROUND OF THE INVENTION provide a local server within the client system for locally processing a portion of the information. These local servers, 0002. A number of fields of endeavor are relevant to the applications and applets are non-standard, and thus require present invention, and exemplary prior art, incorporated special software to be available locally for execution. herein by reference, are disclosed below. The references disclosed provide a skilled artisan with embodiments of 0007 Thus, the Internet poses a number of advantages elements of the present invention, and the teachings therein for commercial use, including low cost and ubiquitous may be combined and Subcombined in various manners in connectivity. Therefore, it is desirable to employ standard accordance with the present teachings. The topical headings Internet technologies while achieving Sufficient quality com are advisory only, and are not intended to limit the appli munications to effect an efficient transaction. cability of any reference. While some embodiments are 0008. A widely dispersed network of access points may discussed as being preferred, it should be understood that all implement a mobile telecommunications protocol. Such as embodiments discussed, in any portion of this documents, IETF RFC 3344 (Mobile IP, IPv4), or various mobile ad hoc whether stated as having advantages or not, form a part of network (MANET) protocols, 2.5G or 3G cellular, or other the invention and may be combined and/or subcombined in types of protocols. Preferably, the protocol allows the client a consistent manner in accordance with the teachings hereof. to maintain a remote connection while traversing between 0003) Internet various access points. See, U.S. Pub. application No. 20040073642, expressly incorporated herein by reference. 0004 The Internet is structured such various networks Mobile Internet Protocol (Mobile IP or MIP, in this case, v4) are interconnected, with communications effected by is an Internet Engineering Task Force (IETF) network layer addressed packets conforming to a common protocol. Based protocol, specified in RFC-3344. It is designed to allow on the packet addressing, information is routed from Source seamless connectivity session maintenance under TCP to destination, often through a set of networks having (Transmission Control Protocol) or other connection ori multiple potential pathways. The communications medium ented transport protocols when a mobile node moves from is shared between all users. Statistically, some proportion of one IP subnet to another. MIPv4 uses two network infra the packets are extraordinarily delayed, or simply lost. structure entities, a Home Agent (HA) and an optional Therefore, protocols involving communications using these Foreign Agent (FA), to deliver packets to the mobile node packets include error detection schemes that request a when it has left its home network. MIPv4 also supports retransmit of required data not received within a time point-of-attachment Care-of Addresses (CoA) if a FA is window. In the even that the network nears capacity or is unavailable. Mobile IP is increasingly being deployed for otherwise Subject to limiting constraint, the incidence of 2.5/3G (2.5 or third generation wireless) provider networks delayed or lost packets increases, thereby increasing and may be deployed in medium and large Enterprise IEEE requests for retransmission and retransmission. Therefore, as 802.11 -based LANs (Local Area Networks) with multiple the network approaches available bandwidth, the load subnets. MIPv4 relies on the use of permanently assigned increases, ultimately leading to failure. In instances where a “home’ IP addresses to help maintain connectivity when a minimum quality of service must be guaranteed, special mobile device connects to a foreign network. On the other Internet technologies are required, to reserve bandwidth or hand, IPsec-based (Internet Protocol Security, a security to specify network pathways. End-to-end quality of service protocol from IETF) VPNs (Virtual Private Networks) use a guarantees, however, may exceed the cost of circuit tunneling scheme in which the outer source IP address is Switched technologies, such as dialup modems, especially based on a CoA at the point-of-attachment and an inner where the high quality needs are intermittent. source IP address assigned for the “home' domain. In 0005 Internet usage typically involves an Internet server, general if either address is changed, such as when the mobile an automated system capable of responding to communica node Switches IP subnets, then a new tunnel is negotiated tions received through the Internet, and often communicat with new keys and several round trip message exchanges. ing with other systems not directly connected to the Internet. The renegotiation of the tunnel interferes with seamless The server typically has relatively large bandwidth to the mobility across wired and wireless IP networks spanning Internet, allowing multiple simultaneous communications multiple IP subnets. sessions, and usually Supports the hypertext transport pro tocol (HTTP), which provides, in conjunction with a so 0009 Market Economy Systems called web browser on a remote client system, a human 0010. In modem retail transactions, predetermined price readable interface which facilitates navigation of various transactions are common, with market transactions, i.e., resources available in the Internet. The client systems are commerce conducted in a setting which allows the transac typically human user interfaces, which employ a browser to tion price to float based on the respective valuation allocated display HTTP “web pages”. The browser typically does not by the buyer(s) and seller(s), often left to specialized fields. provide intelligence. Bandwidth between the client and While interpersonal negotiation is often used to set a transfer Internet is typically relatively small, and various communi price, this price is often different from a transfer price that US 2006/0167784 A1 Jul. 27, 2006 might result from a best-efforts attempt at establishing a live Dutch auction, each participant is provided with the market price. Assuming that the market price is optimal, it current price, the quantity on hand and the time remaining is therefore assumed that alternatives are sub optimal. There in the auction. This type of auction, typically takes place fore, the establishment of a market price is desirable over over a very short period of time and there is a flurry of simple negotiations. activity in the last portion of the auction process. The actual auction terminates when there is no more product to be sold 0011. One particular problem with market-based com or the time period expires. merce is that both seller optimization and market efficiency depend on the fact that representative participants of a 0015. In selecting the optimal type of auction, a number preselected class are invited to participate, and are able to of factors are considered. In order to sell large quantities of promptly communicate, on a relevant timescale, in order to a perishable commodity in a short period of time, the accurately value the goods or services and make an offer. descending price auctions are often preferred. For example, Thus, in traditional market-based system, all participants are the produce and flower markets in Holland routinely use the in the same room, or connected by a high quality telecom Dutch auction (hence the derivation of the name), while the munications link. Alternately, the market valuation process U.S. Government uses this form to sell its financial instru is prolonged over an extended period, allowing non-real ments. The format of a traditional Dutch auction encourages time communications of market information and bids. Thus, early bidders to bid up to their “private value', hoping to pay attempts at ascertaining a market price for non-commodity some price below the “private value”. In making a bid, the goods can be subject to Substantial inefficiencies, which “private value” becomes known, helping to establish a reduce any potential gains by market pricing. Further, while published market value and demand curve for the goods, market pricing might be considered “fair, it also imposes an thus allowing both buyers and sellers to define strategies for element of risk, reducing the ability of parties to predict future auctions. future pricing and revenues. Addressing this risk may also reduce efficiency of a market-based system. 0016. In an auction, typically a seller retains an auction eer to conduct an auction with multiple buyers. (In a reverse 0012 Auction Systems auction, a buyer solicits the lowest price from multiple competing vendors for a desired purchase). Since the seller 0013 When a single party seeks to sell goods to the retains the auctioneer, the seller essentially defines the rules highest valued purchaser(s), to establish a market price, the of the auction. These rules are typically defined to maximize rules of conduct typically define an auction. Typically, the revenues or profit to the seller, while providing an known auctions provide an ascending price or descending inviting forum to encourage a maximum number of high price over time, with bidders making offers or ceasing to valued buyers. If the rules discourage high valuations of the make offers, in the descending price or ascending price goods or services, or discourage participation by an impor models, respectively, to define the market price. After deter tant set of potential bidders, then the rules are not optimum. mining the winner of the auction, the pricing rules define A rule may also be imposed to account for the valuation of uniform price auctions, wherein all Successful bidders pay the good or service applied by the seller, in the form of a the lowest successful bid, second price auctions wherein the reserve price. It is noted that these rules typically seek to winning bidder pays the amount bid by the next highest allocate to the seller a portion of the economic benefit that bidder, and pay-what-you-bid auctions. The pay-what-you would normally inure to the buyer, creating an economic bid auction is also known as a discriminative auction while inefficiency. However, since the auction is to benefit the the uniform price auction is known as a non-discriminative seller, not society as a whole, this potential inefficiency is auction. In a second-price auction, also known as a Vickrey tolerated. An optimum auction thus seeks to produce a auction, the policy seeks to create a disincentive for specu maximum profit (or net revenues) for the seller. An efficient lation and to encourage bidders to Submit bids reflecting auction, on the other hand, maximizes the Sum of he utilities their true value for the good. In the uniform price and second for the buyer and seller. It remains a subject of academic price schemes, the bidder is encourages to disclose the actual debate as to which auction rules are most optimum in given private value to the bidder of the good or service, since at circumstances; however, in practice, simplicity of imple any price below this amount, there is an excess gain to the mentation may be a paramount concern, and simple auctions buyer, whereas by withholding this amount the bid may be may result in highest revenues; complex auctions, while unsuccessful, resulting in a loss of the presumably desirable theoretically more optimal, may discourage bidders from opportunity. In the pay-what-you-bid auction, on the other participating or from applying their true and full private hand, the buyer need not disclose the maximum private valuation in the auction process. valuation, and those bidders with lower risk tolerance will bid higher prices. See, www.isoc.org/inet 98/proceedings/3b/ 0017. Typically, the rules of the auction are predefined 3b 3.html: www.ibm.com/iac/reports-technical/reports and invariant. Further, for a number of reasons, auctions bus-neg-internet.html. typically apply the same rules to all bidders, even though, with a priori knowledge of the private values assigned by 0014 Two common types of auction are the English each bidder to the goods, or a prediction of the private value, auction, which sells a single good to the highest bidder in an an optimization rule may be applied to extract the full value ascending price auction, and the Dutch auction, in which assigned by each bidder, while selling above the sellers multiple units are available for sale, and in which a starting price is selected by the auctioneer, which is successively reServe. reduced, until the supply is exhausted by bidders (or the 0018. In a known ascending price auction, each partici minimum price/final time is reached), with the buyer(s) pant must be made aware of the status of the auction, e.g., paying the lowest successful bid. The term Dutch auction is open, closed, and the contemporaneous price. A bid is also applied to a type of sealed bid auction. In a multi-unit indicated by the identification of the bidder at the contem US 2006/0167784 A1 Jul. 27, 2006 poraneous price, or occasionally at any price above the to those with the highest value. Therefore, if post-auction minimum bid increment plus the previous price. The bids are trading is permitted, the seller will not benefit from these asynchronous, and therefore each bidder must be immedi later gains, and the seller will obtain Sub optimal revenues. ately informed of the particulars of each bid by other 0022. These studies, however, typically do not consider bidders. transaction costs and internal inefficiencies of the resellers, 0019. In a known descending price auction, the process as well as the possibility of multiple classes of purchasers, traditionally entails a common clock, which corresponds to or even multiple channels of distribution, which may be a decrementing price at each decrement interval, with an Subject to varying controls or restrictions, and thus in a real ending time (and price). Therefore, once each participant is market, such theoretical optimal allocation is unlikely. In made aware of the auction parameters, e.g., starting price, fact, in real markets the transaction costs involved in transfer price decrement, ending price/time, before the start of the of ownership are often critical in determining a method of auction, the only information that must be transmitted is sale and distribution of goods. For example, it is the effi auction status (e.g., inventory remaining). ciency of sale that motivates the auction in the first place. Yet, the auction process itself may consume a substantial 0020. As stated above, an auction is traditionally consid margin, for example 1-15% of the transaction value. To ered an efficient manner of liquidating goods at a market presume, even without externally imposed restrictions on price. The theory of an auction is that either the buyer will resale, that all of the efficiencies of the market may be not resell, and thus has an internal or private valuation of the extracted by free reallocation, ignores that the motivation of goods regardless of other's perceived values, or that the the buyer is a profitable transaction, and the buyer may have winner will resell, either to gain economic efficiency or as a fixed and variable costs on the order of magnitude of the part of the buyers regular business. In the later case, it is a margin. Thus, there are substantial opportunities for the general presumption that the resale buyers are not in atten seller to gain enhanced revenues by defining rules of the dance at the auction or are otherwise precluded from bid auction, strategically allocating inventory amount and set ding, and therefore that, after the auction, there will remain demand for the goods at a price in excess of the price paid ting reserve pricing. during the auction. Extinction of this residual demand 0023 Therefore, perfect resale is but a fiction created in results in the so-called “winner's curse', in which the buyer auction (game) theory. Given this deviation from the ideal can make no profit from the transaction during the auction. presumptions, auction theory may be interpreted to provide Since this detracts from the value of the auction as a means the seller with a motivation to misallocate or withhold based of conducting profitable commerce, it is of concern to both on the deviation of practice from theory, likely based on the buyer and seller. In fact, experience with initial public respective transaction costs, seller's utility of the goods, and offerings (IPOs) of stock through various means has dem other factors not considered by the simple analyses. onstrated that by making stock available directly to all 0024. A number of proposals have been made for effect classes of potential purchasers, latent demand for a new ing auction systems using the Internet. These systems issue is extinguished, and the stock price is likely to decline include consumer-to-consumer, business-to-consumer, and after issuance, resulting in an IPO which is characterized as business-to-business types. Generally, these auctions, of “unsuccessful. This potential for post IPO decline tempers various types and implementations discussed further below, even initial interest in the issue, resulting in a paradoxical are conducted through Internet browsers using hypertext decline in revenues from the vehicle. In other words, the markup language (HTML) “web pages, using HTTP. In “money on the table resulting from immediate retrading of some instances, such as BIDWATCH, discussed further IPO shares is deemed a required aspect of the IPO process. below, an application with associated applets is provided to Thus, methods that retain latent demand after IPO shares define a user interface instead of HTML. result in post IPO increases, and therefore a “successful IPO. Therefore, where the transaction scheme anticipates 0025. As stated above, the information packets from the demand for resale after the initial distribution, it is often transaction server to client systems associated with respec important to assure a reasonable margin for resellers and tive bidders communicate various information regarding the limitations on direct sale to ultimate consumers. status of an interactive auction during the progress thereof. The network traffic from the client systems to the transaction 0021 Research into auction theory (game theory) shows server is often limited to the placement of bids; however, the that in an auction, the goal of the seller is to optimize the amount of information required to be transmitted can vary auction by allocating the goods inefficiently, and thus to greatly, and may involve a complex dialogue of communi appropriate to himself an excess gain. This inefficiency cations to complete the auction offer. Typically, Internet manifests itself by either withholding goods from the market based auction systems have scalability issues, wherein or placing the goods in the wrong hands. In order to assure economies of Scale are not completely apparent, leading to for the seller a maximum gain from a misallocation of the goods, restrictions on resale are imposed; otherwise, post implementation of relatively large transaction server sys auction trading will tend to undue the misallocation, and the tems to handle peak loads. When the processing power of the anticipation of this trading will tend to control the auction transaction server system is exceeded, entire system outages pricing. The misallocation of goods imposed by the seller may occur, resulting in lost sales or diminished profits, and through restrictions allow the seller to achieve greater rev diminished goodwill. enues than if free resale were permitted. It is believed that 0026. In most Internet auction system implementations, in an auction followed by perfect resale, that any mis there are a large quantity of simultaneous auctions, with assignment of the goods lowers the seller's revenues below each auction accepting tens or hundreds of bids over a the optimum and likewise, in an auction market followed by timescale of hours to days. In systems where the transaction perfect resale, it is optimal for the seller to allocate the goods Volume exceeds these scales, for example in Stock and US 2006/0167784 A1 Jul. 27, 2006 commodity exchanges, which can accommodate large num auction system, including non-real time updating of bidding bers of transactions per second involving the same issue, a information, especially in the final stages of an auction. private network, or even a local area network, is employed, and the public Internet is not used as a direct communica 0032 Because of existing bandwidth and technological tions system with the transaction server. Thus, while infra hurdles, Internet auctions are quite different from live auc structures are available to allow Successful handling of tions with respect to psychological factors. Live auctions are massive transaction per second Volumes, these systems often monitored closely by bidders, who strategically make typically avoid direct public Internet communications or use bids, based not only on the “value of the goods, but also on of some of its limiting technologies. The transaction pro an assessment of the competition, timing, psychology, and progress of the auction. It is for this reason that so-called cessing limitations are often due to the finite time required proxy bidding, wherein the bidder creates a preprogrammed to handle, e.g., open, update, and close, database records. “strategy”, usually limited to a maximum price, are disfa 0027. In business-to-business auctions, buyers seek to Vored. A maximum price proxy bidding system is somewhat ensure that the population of ultimate consumers for the inefficient, in that other bidders may test the proxy, seeking good or services are not present at the auction, in order to to increase the bid price, without actually intending to avoid the “winner's curse', where the highest bidder in the purchase, or contrarily, after testing the proxy, a bidder auction cannot liquidate or work the asset at a profit. Thus, might give up, even below a price he might have been business-to-business auctions are distinct from business-to willing to pay. Thus, the proxy imposes inefficiency in the consumer auctions. In the former, the optimization by the system that effectively increases the transaction cost. seller must account for the desire or directive of the seller to avoid direct retail distribution, and instead to rely on a 0033. In order to address a flurry of activity that often distribution tier represented in the auction. In the latter, the occurs at the end of an auction, an auction may be held open seller seeks maximum revenues and to exhaust the possi until no further bids are cleared for a period of time, even if bilities for downstream trade in the goods or services. In advertised to end at a certain time. This is common to both fact, these types of auctions may be distinguished by various live and automated auctions. However, this lack of deter implementing rules, such as requiring sales tax resale cer minism may upset coordinated Schedules, thus impairing tificates, minimum lot size quantities, preregistration or efficient business use of the auction system. qualification, Support or associated services, or limitations 0034. In order to facilitate management of bids and on the title to the goods themselves. The conduct of these bidding, Some of the Internet auction sites have provided auctions may also differ, in that consumer involvement non-Hypertext Markup Language (HTML) browser based typically is permissive of mistake or indecision, while in a software “applet” to track auctions. For example, ONSA pure business environment professionalism and decisiveness LE.COM has made available a Marimba Castanet(R) applet are mandated. called Bidwatch to track auction progress for particular 0028. In many instances, psychology plays an important items or classes of items, and to facilitate bidding thereon. role in the conduct of the auction. In a live auction, bidders This system, however, lacks real-time performance under can see each other, and judge the tempo of the auction. In many circumstances, having a stated refresh period of 10 addition, multiple auctions are often conducted sequentially, seconds, with a long latency for confirmation of a bid, due so that each bidder can begin to understand the other to constraints on Software execution, quality of Service in bidder's patterns, including hesitation, bluffing, facial ges communications streams, and bid confirmation dialogue. tures or mannerisms. Thus, bidders often prefer live auctions Thus, it is possible to lose a bid even if an attempt was made to remote or automated auctions if the bidding is to be prior to another bidder. The need to quickly enter the bid, at conducted Strategically. risk of being too late, makes the process potentially error prone. 0029 Internet Auctions 0035) Proxy bidding, as discussed above, is a known 0030. On-line electronic auction systems which allow technique for overcoming the constraints of Internet com efficient sales of products and services are well known, for munications and client processing limitations, since it example, EBAY.COM, ONSALE.COM, UBID.COM, and bypasses the client and telecommunications links and may the like. Inverse auctions that allow efficient purchases of execute solely on the host system or local thereto. However, product are also known, establishing a market price by proxy bidding undermines some of the efficiencies gained by competition between sellers. The Internet holds the promise a live market. of further improving efficiency of auctions by reducing transaction costs and freeing the “same time-same place' 0.036 U.S. Pat. No. 5,890,138 to Godin, et al. (Mar. 30, limitations of traditional auctions. This is especially appro 1999), expressly incorporated herein by reference in its priate where the goods may be adequately described by text entirety, relates to an Internet auction system. The system or images, and thus a physical examination of the goods is implements a declining price auction process, removing a not required prior to bidding. user from the auction process once an indication to purchase 0031. In existing Internet systems, the technological has been received. See, Rockoff, T. E., Groves, M.: “Design focus has been in providing an auction system that, over the of an Internet-based System for Remote Dutch Auctions'. course of hours to days, allow a large number of simulta Internet Research, V 5, n 4, pp. 10-16, MCB University neous auctions, between a large number of bidders to occur. Press, Jan. 01, 1995. These systems must be scalable and have high transaction 0037. A known computer site for auctioning a product throughput, while assuring database consistency and overall on-line comprises at least one web server computer designed system reliability. Even so, certain users may selectively for serving a host of computer browsers and providing the exploit known technological limitations and artifacts of the browsers with the capability to participate in various auc US 2006/0167784 A1 Jul. 27, 2006
tions, where each auction is of a single product, at a specified Dutch type auction, the price markdown feature may be time, with a specified number of the product available for responsive to bidding activity over time, amount of bids sale. The web server cooperates with a separate database received, and number of items bid for. Likewise, in the computer, separated from the web server computer by a progressive auction, the award price may be dependent on firewall. The database computer is accessible to the web the quantity desired, and typically implements a lowest computer server computer to allow selective retrieval of successful bid price rule. Bids that are below a preset product information, which includes a product description, maximum posted selling price are maintained in reserve by the quantity of the product to be auctioned, a start price of the system. If a certain sales Volume is not achieved in a the product, and an image of the product. The web server specified period of time, the price is reduced to liquidate computer displays, updated during an auction, the current demand above the price point, with the new price becoming price of the product, the quantity of the product remaining the posted price. On the other hand, if a certain sales volume available for purchase and the measure of the time remain is exceeded in a specified period of time, the system may ing in the auction. The current price is decreased in a automatically increase the price. These automatic price predetermined manner during the auction. Each user is changes allow the seller to respond quickly to market provided with an input instructing the system to purchase the conditions while keeping the price of the merchandise as product at a displayed current price, transmitting an identi high as possible, to the seller's benefit. A “Proxy Bidding fication and required financial authorization for the purchase feature allows a bidder to place a bid for the maximum of the product, which must be confirmed within a predeter amount they are willing to pay, keeping this value a secret, mined time. In the known system, a certain fall-out rate in displaying only the amount necessary to win the item up to the actual purchase confirmation may be assumed, and the amount of the currently high bids or proxy bids of other therefore some overselling allowed. Further, after a purchase bidders. This feature allows bidders to participate in the is indicate, the user's screen is not updated, obscuring the electronic auction without revealing to the other bidders the ultimate lowest selling price from the user. However, if the extent to which they are willing to increase their bids, while user maintains a second browser, he can continue to monitor maintaining control of their maximum bid without closely the auction to determine whether the product could have monitoring the bidding. The feature assures proxy bidders been purchased at a lower price, and if so, fail to confirm the the lowest possible price up to a specified maximum without committed purchase and purchase the same goods at a lower requiring frequent inquiries as to the State of the bidding. price while reserving the goods to avoid risk of loss. Thus, the system is flawed, and may fail to produce an efficient 0041. A "Floating Closing Time’ feature may also be transaction or optimal price. implemented whereby the auction for a particular item is automatically closed if no new bids are received within a 0038 An Internet declining price auction system may predetermined time interval, assuming an increasing price provide the ability to track the price demand curve, provid auction. Bidders thus have an incentive to place bids expe ing valuable marketing information. For example, in trying ditiously, rather than waiting until near the anticipated close to determine the response at different prices, companies of the auction. normally have to conduct market Surveys. In contrast, with a declining price auction, Substantial information regarding 0.042 U.S. Pat. No.5,905,975, Ausubel, issued May 18, price and demand is immediately known. The relationship 1999, expressly incorporated herein by reference in its between participating bidders and average purchasers can entirety, relates to computer implemented methods and then be applied to provide a conventional price demand apparatus for auctions. The proposed system provides intel curve for the particular product. ligent systems for the auctioneer and for the user. The auctioneer's system contains information from a user system 0039 U.S. Pat. No. 5,835,896, Fisher, et al., issued Nov. based on bid information entered by the user. With this 10, 1998, expressly incorporated herein by reference in its information, the auctioneer's system determines whether the entirety, provides method and system for processing and auction can be concluded or not and appropriate messages transmitting electronic auction information over the Internet, are transmitted. At any point in the auction, bidders are between a central transaction server system and remote provided the opportunity to submit not only their current bidder terminals. Those bids are recorded by the system and bids, but also to enter future bids, or bidding rules which the bidders are updated with the current auction status may have the opportunity to become relevant at future times information. When appropriate, the system closes the auc or prices, into the auction system's database. Participants tion from further bidding and notifies the winning bidders may revise their executory bids, by entering updated bids. and losers as to the auction outcome. The transaction server Thus, at one extreme, a bidder who wishes to economize on posts information from a database describing a lot available his time may choose to enter his entire set of bidding rules for purchase, receives a plurality of bids, stored in a bid into the computerized system at the start of the auction, database, in response to the information, and automatically effectively treating this as a sealed-bid auction. At the categorizes the bids as Successful or unsuccessful. Each bid opposite extreme, a bidder who wishes to closely participate is validated, and an electronic mail message is sent inform in the auction may choose to constantly monitor the auc ing the bidder of the bid status. This system employs HTTP, tions progress and to submit all of his bids in real time. See and thus does not automatically update remote terminal also, U.S. patent application Ser. No.08/582.901 filed Jan. 4, screens, requiring the e-mail notification feature. 1996, which provides a method for auctioning multiple, 0040. The auction rules may be flexible, for example identical objects and close Substitutes. including Dutch-type auctions, for example by implement ing a price markdown feature with Scheduled price adjust 0043. Secure Networks ments, and English-type (progressive) auctions, with price 0044) A number of references relate to secure networks, increases corresponding to Successively higher bids. In the which are an aspect of various embodiments of the present US 2006/0167784 A1 Jul. 27, 2006
invention. These references are incorporated herein by ref alized case of registering a trusted device with a trusted third erence in their entirety, including U.S. Pat. No. 5,933,498 party and receiving authorization from that party enabling (Schneck, et al., Aug. 3, 1999); U.S. Pat. No. 5,978,918 the device to communicate with other trusted devices. Fur (Scholnick, et al., Nov. 2, 1999); U.S. Pat. No. 6,005,943 ther preferred embodiments provide for rekeying and (Cohen, et al., Dec. 21, 1999); U.S. Pat. No. 6,009,526 upgrading of device firmware using a certificate system, and (Choi, Dec. 28, 1999); U.S. Pat. No. 6,021.202 (Anderson, encryption of stream-oriented data. et al., Feb. 1, 2000); U.S. Pat. No. 6,021,491 (Renaud, Feb. 0049 U.S. Pat. No. 6,052,467 (Brands, Apr. 18, 2000), 1, 2000); U.S. Pat. No. 6,021,497 (Bouthillier, et al., Feb. 1, expressly incorporated herein by reference, relates to a 2000); U.S. Pat. No. 6,023,762 (Dean, et al., Feb. 8, 2000); system for ensuring that the blinding of Secret-key certifi U.S. Pat. No. 6,029.245 (Scanlan, Feb. 22, 2000); U.S. Pat. cates is restricted, even if the issuing protocol is performed No. 6,049,875 (Suzuki, et al., Apr. 11, 2000); U.S. Pat. No. in parallel mode. A cryptographic method is disclosed that 6,055,508 (Naor, et al., Apr. 25, 2000); U.S. Pat. No. enables the issuer in a secret-key certificate issuing protocol 6,065,119 (Sandford, II, et al., May 16, 2000); U.S. Pat. No. to issue triples consisting of a secret key, a corresponding 6,073,240 (Kurtzberg, et al., Jun. 6, 2000); U.S. Pat. No. public key, and a secret-key certificate of the issuer on the 6,075,860 (Ketcham, Jun. 13, 2000); and U.S. Pat. No. public key, in Such a way that receiving parties can blind the 6,075.861 (Miller, II, Jun. 13, 2000). public key and the certificate, but cannot blind a predeter 0045 Cryptographic Technology mined non-trivial predicate of the secret key even when executions of the issuing protocol are performed in parallel. 0046 U.S. Pat. No. 5,956,408 (Arnold, Sep. 21, 1999), expressly incorporated herein by reference, relates to an 0050 U.S. Pat. No. 6,052,780 (Glover, Apr. 18, 2000), apparatus and method for secure distribution of data. Data, expressly incorporated herein by reference, relates to a including program and Software updates, is encrypted by a computer system and process for accessing an encrypted and public key encryption system using the private key of the self-decrypting digital information product while restricting data sender. The sender also digitally signs the data. The access to decrypted digital information. Some of these receiver decrypts the encrypted data, using the public key of problems with digital information protection systems may the sender, and Verifies the digital signature on the trans be overcome by providing a mechanism that allows a mitted data. The program interacts with basic information content provider to encrypt digital information without stored within the confines of the receiver. As result of the requiring either a hardware or platform manufacturer or a interaction, the software updates are installed within the content consumer to provide support for the specific form of confines of the user, and the basic information stored within corresponding decryption. This mechanism can be provided the confines of the user is changed. in a manner that allows the digital information to be copied easily for back-up purposes and to be transferred easily for 0047 U.S. Pat. No. 5,982,891 (Ginter, et al., Nov. 9, distribution, but which should not permit copying of the 1999); U.S. Pat. No. 5,949,876 (Ginter, et al., Sep. 7, 1999); digital information in decrypted form. In particular, the and U.S. Pat. No. 5,892,900 (Ginter, et al., April 6, 1999), encrypted digital information is stored as an executable expressly incorporated herein by reference, relate to systems computer program that includes a decryption program that and methods for secure transaction management and elec decrypts the encrypted information to provide the desired tronic rights protection. Electronic appliances, such as com digital information, upon Successful completion of an autho puters, help to ensure that information is accessed and used rization procedure by the user. In combination with other only in authorized ways, and maintain the integrity, avail mechanisms that track distribution, enforce royalty pay ability, and/or confidentiality of the information. Such elec ments and control access to decryption keys, an improved tronic appliances provide a distributed virtual distribution method is provided for identifying and detecting sources of environment (VDE) that may enforce a secure chain of unauthorized copies. Suitable authorization procedures also handling and control, for example, to control and/or meter or enable the digital information to be distributed for a limited otherwise monitor use of electronically stored or dissemi number of uses and/or users, thus enabling per-use fees to be nated information. Such a virtual distribution environment charged for the digital information. may be used to protect rights of various participants in electronic commerce and other electronic or electronic 0051) See also, U.S. Pat. No. 4,200,770 (Cryptographic facilitated transactions. Distributed and other operating sys apparatus and method); U.S. Pat. No. 4,218,582 (Public key tems, environments and architectures, such as, for example, cryptographic apparatus and method); U.S. Pat. No. 4,264. those using tamper-resistant hardware-based processors, 782 (Method and apparatus for transaction and identity may establish security at each node. These techniques may verification): U.S. Pat. No. 4,306,111 (Simple and effective be used to Support an all-electronic information distribution, public-key cryptosystem); U.S. Pat. No. 4,309,569 (Method for example, utilizing the “electronic highway.” of providing digital signatures); U.S. Pat. No. 4.326,098 (High security system for electronic signature verification); 0.048 U.S. Pat. No. 6,009, 177 (Sudia, Dec. 28, 1999), U.S. Pat. No. 4,351,982 (RSA Public-key data encryption expressly incorporated herein by reference, relates to a system having large random prime number generating cryptographic system and method with a key escrow feature microprocessor or the like); U.S. Pat. No. 4.365,110 (Mul that uses a method for verifiably splitting users’ private tiple-destinational cryptosystem for broadcast networks); encryption keys into components and for sending those U.S. Pat. No. 4,386.233 (Crytographic key notarization components to trusted agents chosen by the particular users, methods and apparatus); U.S. Pat. No. 4.393.269 (Method and provides a system that uses modem public key certifi and apparatus incorporating a one-way sequence for trans cate management, enforced by a chip device that also action and identity verification): U.S. Pat. No. 4.399,323 self-certifies. The methods for key escrow and receiving an (Fast real-time public key cryptography); U.S. Pat. No. escrow certificate are also applied herein to a more gener 4.405,829 (Cryptographic communications system and US 2006/0167784 A1 Jul. 27, 2006
method); U.S. Pat. No. 4,438,824 (Apparatus and method using a personal identification device); U.S. Pat. No. 6,041, for cryptographic identity verification); U.S. Pat. No. 4,453, 122 (Method and apparatus for hiding cryptographic keys 074 (Protection system for intelligent cards); U.S. Pat. No. utilizing autocorrelation timing encoding and computation); 4.458,109 (Method and apparatus providing registered mail U.S. Pat. No. 6,041,123 (Centralized secure communica features in an electronic communication system); U.S. Pat. tions system); U.S. Pat. No. 6,041.357 (Common session No. 4,471,164 (Stream cipher operation using public key token system and protocol); U.S. Pat. No. 6,041,408 (Key cryptosystem); U.S. Pat. No. 4,514,592 (Cryptosystem); distribution method and system in secure broadcast com U.S. Pat. No. 4,528,588 (Method and apparatus for marking munication); U.S. Pat. No. 6,041,410 (Personal identifica the information content of an information carrying signal); tion fob); U.S. Pat. No. 6,044,131 (Secure digital X-ray U.S. Pat. No. 4,529,870 (Cryptographic identification, finan image authentication method); U.S. Pat. No. 6,044,155 cial transaction, and credential device); U.S. Pat. No. 4,558, (Method and system for securely archiving core data 176 (Computer systems to inhibit unauthorized copying, secrets); U.S. Pat. No. 6,044,157 (Microprocessor suitable unauthorized usage, and automated cracking of protected for reproducing AV data while protecting the AV data from software); U.S. Pat. No. 4,567,600 (Method and apparatus illegal copy and image information processing system using for maintaining the privacy of digital messages conveyed by the microprocessor); U.S. Pat. No. 6,044,205 (Communica public transmission); U.S. Pat. No. 4.575,621 (Portable tions system for transferring information between memories electronic transaction device and system therefor); U.S. Pat. according to processes transferred with the information); No. 4,578,531 (Encryption system key distribution method U.S. Pat. No. 6,044.349 (Secure and convenient information and apparatus); U.S. Pat. No. 4,590,470 (User authentication storage and retrieval method and apparatus); U.S. Pat. No. system employing encryption functions); U.S. Pat. No. 6,044.350 (Certificate meter with selectable indemnification 4.595,950 (Method and apparatus for marking the informa provisions); U.S. Pat. No. 6,044.388 (Pseudorandom num tion content of an information carrying signal); U.S. Pat. No. ber generator); U.S. Pat. No. 6,044,462 (Method and appa 4,625,076 (Signed document transmission system); U.S. Pat. ratus for managing key revocation): U.S. Pat. No. 6,044,463 No. 4,633,036 (Method and apparatus for use in public-key (Method and system for message delivery utilizing Zero data encryption system); U.S. Pat. No. 5,991,406 (System knowledge interactive proof protocol); U.S. Pat. No. 6,044. and method for data recovery); U.S. Pat. No. 6,026,379 464 (Method of protecting broadcast data by fingerprinting (System, method and article of manufacture for managing a common decryption function); U.S. Pat. No. 6,044,466 transactions in a high availability system); U.S. Pat. No. (Flexible and dynamic derivation of permissions); U.S. Pat. 6,026,490 (Configurable cryptographic processing engine No. 6,044,468 (Secure transmission using an ordinarily and method); U.S. Pat. No. 6,028,932 (Copy prevention insecure network communication protocol such as SNMP); method and apparatus for digital video system); U.S. Pat. U.S. Pat. No. 6,047,051 (Implementation of charging in a No. 6,028,933 (Encrypting method and apparatus enabling telecommunications system); U.S. Pat. No. 6,047,066 multiple access for multiple services and multiple transmis (Communication method and device); U.S. Pat. No. 6,047. sion modes over a broadband communication network); U.S. 067 (Electronic-monetary system); U.S. Pat. No. 6,047,072 Pat. No. 6,028,936 (Method and apparatus for authenticating (Method for secure key distribution over a nonsecure com recorded media); U.S. Pat. No. 6,028,937 (Communication munications network); U.S. Pat. No. 6,047,242 (Computer device which performs two-way encryption authentication system for protecting Software and a method for protecting in challenge response format); U.S. Pat. No. 6,028,939 (Data software); U.S. Pat. No. 6,047.268 (Method and apparatus security system and method); U.S. Pat. No. 6,029,150 (Pay for billing for transactions conducted over the internet); U.S. ment and transactions in electronic commerce system); U.S. Pat. No. 6,047.269 (Self-contained payment system with Pat. No. 6,029,195 (System for customized electronic iden circulating digital vouchers); U.S. Pat. No. 6,047.374 tification of desirable objects); U.S. Pat. No. 6,029.247 (Method and apparatus for embedding authentication infor (Method and apparatus for transmitting secured data); U.S. mation within digital data); U.S. Pat. No. 6,047,887 (System Pat. No. 6,031,913 (Apparatus and method for secure com and method for connecting money modules); U.S. Pat. No. munication based on channel characteristics); U.S. Pat. No. 6,049,610 (Method and apparatus for digital signature 6,031,914 (Method and apparatus for embedding data, authentication); U.S. Pat. No. 6,049,612 (File encryption including watermarks, in human perceptible images); U.S. method and system); U.S. Pat. No. 6,049,613 (Method and Pat. No. 6,034,618 (Device authentication system which apparatus for encrypting, decrypting, and providing privacy allows the authentication function to be changed); U.S. Pat. for data values); U.S. Pat. No. 6,049,671 (Method for No. 6,035,041 (Optimal-resilience, proactive, public-key identifying and obtaining computer software from a network cryptographic system and method); U.S. Pat. No. 6,035.398 computer); U.S. Pat. No. 6,049,785 (Open network payment (Cryptographic key generation using biometric data); U.S. system for providing for authentication of payment orders Pat. No. 6,035,402 (Virtual certificate authority); U.S. Pat. based on a confirmation electronic mail message); U.S. Pat. No. 6,038.315 (Method and system for normalizing biomet No. 6,049,786 (Electronic bill presentment and payment ric variations to authenticate users from a public database system which deters cheating by employing hashes and and that ensures individual biometric data privacy); U.S. digital signatures); U.S. Pat. No. 6,049,787 (Electronic Pat. No. 6,038.316 (Method and system for protection of business transaction system with notarization database and digital information); U.S. Pat. No. 6,038.322 (Group key means for conducting a notarization procedure); U.S. Pat. distribution); U.S. Pat. No. 6,038,581 (Scheme for arith No. 6,049,838 (Persistent distributed capabilities); U.S. Pat. metic operations in finite field and group operations over No. 6,049,872 (Method for authenticating a channel in elliptic curves realizing improved computational speed); large-scale distributed systems); U.S. Pat. No. 6,049,874 U.S. Pat. No. 6,038,665 (System and method for backing up (System and method for backing up computer files over a computer files over a wide area computer network); U.S. wide area computer network); U.S. Pat. No. 6,052,466 Pat. No. 6,038,666 (Remote identity verification technique (Encryption of data packets using a sequence of private keys US 2006/0167784 A1 Jul. 27, 2006 generated from a public key exchange); U.S. Pat. No. ment system); U.S. Pat. No. 6,069,954 (Cryptographic data 6,052,467 (System for ensuring that the blinding of secret integrity with serial bit processing and pseudo-random gen key certificates is restricted, even if the issuing protocol is erators); U.S. Pat. No. 6,069,955 (System for protection of performed in parallel mode); U.S. Pat. No. 6,052,469 goods against counterfeiting); U.S. Pat. No. 6,069,969 (Interoperable cryptographic key recovery system with veri (Apparatus and method for electronically acquiring finger fication by comparison); U.S. Pat. No. 6,055.314 (System print images); U.S. Pat. No. 6,069,970 (Fingerprint sensor and method for secure purchase and delivery of video and token reader and associated methods); U.S. Pat. No. content programs); U.S. Pat. No. 6,055.321 (System and 6,070,239 (System and method for executing verifiable method for hiding and extracting message data in multime programs with facility for using non-verifiable programs from trusted sources); U.S. Pat. No. 6,072,870 (System, dia data): U.S. Pat. No. 6,055,508 (Method for secure method and article of manufacture for a gateway payment accounting and auditing on a communications network); architecture utilizing a multichannel, extensible, flexible U.S. Pat. No. 6,055,512 (Networked personal customized architecture); U.S. Pat. No. 6,072,874 (Signing method and information and facility services); U.S. Pat. No. 6,055,636 apparatus using the same); U.S. Pat. No. 6,072,876 (Method (Method and apparatus for centralizing processing of key and system for depositing private key used in RSA crypto and certificate life cycle management); U.S. Pat. No. 6,055, system): U.S. Pat. No. 6,073,125 (Token key distribution 639 (Synchronous message control system in a Kerberos system controlled acceptance mail payment and evidencing domain); U.S. Pat. No. 6,056,199 (Method and apparatus for system); U.S. Pat. No. 6,073,160 (Document communica storing and reading data); U.S. Pat. No. 6,057.872 (Digital tions controller); U.S. Pat. No. 6,073,172 (Initializing and coupons for pay televisions); U.S. Pat. No. 6,058,187 reconfiguring a secure network interface); U.S. Pat. No. (Secure telecommunications data transmission); U.S. Pat. 6,073,234 (Device for authenticating user's access rights to No. 6,058,188 (Method and apparatus for interoperable resources and method); U.S. Pat. No. 6,073,236 (Authenti validation of key recovery information in a cryptographic cation method, communication method, and information system): U.S. Pat. No. 6,058,189 (Method and system for processing apparatus); U.S. Pat. No. 6,073,237 (Tamper performing secure electronic monetary transactions); U.S. resistant method and apparatus); U.S. Pat. No. 6,073.238 Pat. No. 6,058,193 (System and method of verifying cryp (Method of securely loading commands in a Smart card); tographic postage evidencing using a fixed key set); U.S. U.S. Pat. No. 6,073,242 (Electronic authority server); U.S. Pat. No. 6,058.381 (Many-to-many payments system for Pat. No. 6,075.864 (Method of establishing secure, digitally network content materials); U.S. Pat. No. 6,058,383 (Com signed communications using an encryption key based on a putationally efficient method for trusted and dynamic digital blocking set cryptosystem); U.S. Pat. No. 6,075.865 (Cryp objects dissemination); U.S. Pat. No. 6,061.448 (Method tographic communication process and apparatus); U.S. Pat. and system for dynamic server document encryption); U.S. No. 6,076,078 (Anonymous certified delivery); U.S. Pat. Pat. No. 6,061.454 (System, method, and computer program No. 6,076.162 (Certification of cryptographic keys for chip for communicating a key recovery block to enable third cards); U.S. Pat. No. 6,076,163 (Secure user identification party monitoring without modification to the intended based on constrained polynomials); U.S. Pat. No. 6,076, 164 receiver); U.S. Pat. No. 6,061,692 (System and method for (Authentication method and system using IC card); U.S. Pat. administering a meta database as an integral component of No. 6,076,167 (Method and system for improving security an information server); U.S. Pat. No. 6,061,789 (Secure in network applications); U.S. Pat. No. 6,078,663 (Commu anonymous information exchange in a network); U.S. Pat. nication apparatus and a communication system); U.S. Pat. No. 6,061,790 (Network computer system with remote user No. 6,078,665 (Electronic encryption device and method); data encipher methodology); U.S. Pat. No. 6,061,791 (Initial U.S. Pat. No. 6,078,667 (Generating unique and unpredict secret key establishment including facilities for verification able values); U.S. Pat. No. 6,078,909 (Method and apparatus of identity); U.S. Pat. No. 6,061,792 (System and method for for licensing computer programs using a DSA signature); fair exchange of time-independent information goods over a U.S. Pat. No. 6,079,018 (System and method for generating network); U.S. Pat. No. 6,061,794 (System and method for unique secure values for digitally signing documents); U.S. performing secure device communications in a peer-to-peer Pat. No. 6,079,047 (Unwrapping system and method for bus architecture); U.S. Pat. No. 6,061,796 (Multi-access multiple files of a container); U.S. Pat. No. 6,081,597 virtual private network); U.S. Pat. No. 6,061,799 (Remov (Public key cryptosystem method and apparatus); U.S. Pat. able media for password based authentication in a distrib No. 6,081,598 (Cryptographic system and method with fast uted system); U.S. Pat. No. 6,064,723 (Network-based mul decryption); U.S. Pat. No. 6,081.610 (System and method timedia communications and directory system and method for verifying signatures on documents); U.S. Pat. No. 6,081, of operation): U.S. Pat. No. 6,064,738 (Method for encrypt 790 (System and method for secure presentment and pay ing and decrypting data using chaotic maps); U.S. Pat. No. ment over open networks); U.S. Pat. No. 6,081.893 (System 6,064,740 (Method and apparatus for masking modulo expo for Supporting secured log-in of multiple users into a plu nentiation calculations in an integrated circuit); U.S. Pat. rality of computers using combined presentation of memo No. 6,064,741 (Method for the computer-aided exchange of rized password and transportable passport record), U.S. Pat. cryptographic keys between a user computer unit U and a No. 6,192.473 (System and method for mutual authentica network computer unit N); U.S. Pat. No. 6,064,764 (Fragile tion and secure communications between a postage security watermarks for detecting tampering in images); U.S. Pat. device and a meter server), each of which is expressly No. 6,064,878 (Method for separately permissioned com incorporated herein by reference. munication); U.S. Pat. No. 6,065,008 (System and method for secure font subset distribution): U.S. Pat. No. 6,067,620 0.052 See, also, U.S. Pat. No. 6,028,937 (Tatebayashi et (Stand alone security device for computer networks); U.S. al.), U.S. Pat. No. 6,026,167 (Aziz), U.S. Pat. No. 6,009,171 Pat. No. 6,069,647 (Conditional access and content security (Ciacelli et al.) (Content Scrambling System, or “CSS), method); U.S. Pat. No. 6,069,952 (Data copyright manage U.S. Pat. No. 5,991,399 (Graunke et al.), U.S. Pat. No. US 2006/0167784 A1 Jul. 27, 2006
5,948,136 (Smyers) (IEEE 1394-1995), and U.S. Pat. No. 6,058.303 (System and method for subscriberactivity super 5,915.018 (Aucsmith), expressly incorporated herein by vision): U.S. Pat. No. 6,055,575 (Virtual private network reference, and Jim Wright and Jeff Robillard (Philsar Semi system and method); U.S. Pat. No. 6,052,788 (Firewall conductor), “Adding Security to Portable Designs, Portable providing enhanced network security and user transpar Design, March 2000, pp. 16-20. ency); U.S. Pat. No. 6,047.325 (Network device for sup porting construction of virtual local area networks on arbi 0053) See also, Stefik, U.S. Pat. No. 5,715,403 (System trary local and wide area computer networks); U.S. Pat. No. for controlling the distribution and use of digital works 6,032,118 (Virtual private network service provider for having attached usage rights where the usage rights are asynchronous transfer mode network); U.S. Pat. No. 6,029. defined by a usage rights grammar); U.S. Pat. No. 5,638.443 067 (Virtual private network for mobile subscribers); U.S. (System for controlling the distribution and use of composite Pat. No. 6,016,318 (Virtual private network system over digital works); U.S. Pat. No. 5,634,012 (System for control public mobile data network and virtual LAN); U.S. Pat. No. ling the distribution and use of digital works having a fee 6,009.430 (Method and system for provisioning databases in reporting mechanism); and U.S. Pat. No. 5,629,980 (System an advanced intelligent network); U.S. Pat. No. 6,005,859 for controlling the distribution and use of digital works), (Proxy VAT-PSTN origination): U.S. Pat. No. 6,002,767 expressly incorporated herein by reference. (System, method and article of manufacture for a modular 0054 Computer Security and Devices gateway server architecture); U.S. Pat. No. 6,002.756 (Method and system for implementing intelligent telecom 0055. A number of references relate to computer system munication services utilizing self-sustaining, fault-tolerant security, which is a part of various embodiment of the object oriented architecture), each of which is expressly invention. The following references relevant to this issue are incorporated herein by reference. incorporated herein by reference: U.S. Pat. No. 5,881,225 (Worth, Mar. 9, 1999); U.S. Pat. No. 5,937,068 (Audebert, 0059) See also, U.S. Pat. No. 6,081,900 (Secure intranet Aug. 10, 1999); U.S. Pat. No. 5,949,882 (Angelo, Sep. 7, access); U.S. Pat. No. 6,081,750 (Ergonomic man-machine 1999); U.S. Pat. No. 5,953,419 (Lohstroh, et al., Sep. 14, interface incorporating adaptive pattern recognition based 1999); U.S. Pat. No. 5,956,400 (Chaum, et al., Sep. 21, control system); U.S. Pat. No. 6,081,199 (Locking device 1999); U.S. Pat. No. 5,958,050 (Griffin, et al., Sep. 28, for systems access to which is time-restricted); U.S. Pat. No. 1999); U.S. Pat. No. 5,978.475 (Schneier, et al., Nov. 2, 6,079,621 (Secure card for E-commerce and identification); 1999); U.S. Pat. No. 5,991,878 (McDonough, et al., Nov. 23, U.S. Pat. No. 6,078.265 (Fingerprint identification security 1999); U.S. Pat. No. 6,070,239 (McManis, May 30, 2000); system): U.S. Pat. No. 6,076.167 (Method and system for and U.S. Pat. No. 6,079,021 (Abadi, et al., Jun. 20, 2000). improving security in network applications); U.S. Pat. No. 6,075,455 (Biometric time and attendance system with epi 0056. A number of references relate to computer security dermal topographical updating capability); U.S. Pat. No. devices, which is a part of various embodiment of the 6,072,894 (Biometric face recognition for applicant screen invention. The following references relevant to this issue are ing); U.S. Pat. No. 6,070,141 (System and method of incorporated herein by reference: U.S. Pat. No. 5,982,520 assessing the quality of an identification transaction using an (Weiser, et al., Nov. 9, 1999); U.S. Pat. No. 5,991,519 identification quality score); U.S. Pat. No. 6,068.184 (Secu (Benhammou, et al., Nov. 23, 1999); U.S. Pat. No. 5,999,629 rity card and system for use thereof): U.S. Pat. No. 6,064. (Heer, et al., Dec. 7, 1999); U.S. Pat. No. 6,034,618 (Tate 751 (Document and signature data capture system and bayashi, et al., Mar. 7, 2000); U.S. Pat. No. 6,041,412 method); U.S. Pat. No. 6,056,197 (Information recording (Timson, et al., Mar. 21, 2000): U.S. Pat. No. 6,061.451 method for preventing alteration, information recording (Muratani, et al., May 9, 2000); and U.S. Pat. No. 6,069,647 apparatus, and information recording medium); U.S. Pat. (Sullivan, et al., May 30, 2000). No. 6,052,468 (Method of securing a cryptographic key); U.S. Pat. No. 6,045,039 (Cardless automated teller transac 0057 Virtual Private Network tions); U.S. Pat. No. 6,044,349 (Secure and convenient 0.058 A number of references relate to virtual private information storage and retrieval method and apparatus); networks, which is a part of various embodiment of the U.S. Pat. No. 6,044,155 (Method and system for securely invention. The following references relevant to this issue are archiving core data secrets); U.S. Pat. No. 6,041,410 (Per incorporated herein by reference: U.S. Pat. No. 6,079,020 sonal identification fob); U.S. Pat. No. 6,040,783 (System (Liu, Jun. 20, 2000); U.S. Pat. No. 6,081,900 (Secure and method for remote, wireless positive identity verifica intranet access); U.S. Pat. No. 6,081,533 (Method and tion): U.S. Pat. No. 6,038,666 (Remote identity verification apparatus for an application interface module in a Subscriber technique using a personal identification device); U.S. Pat. terminal unit); U.S. Pat. No. 6,079,020 (Method and appa No. 6,038.337 (Method and apparatus for object recogni ratus for managing a virtual private network); U.S. Pat. No. tion): U.S. Pat. No. 6,038,315 (Method and system for 6,078,946 (System and method for management of connec normalizing biometric variations to authenticate users from tion oriented networks); U.S. Pat. No. 6,078,586 (ATM a public database and that ensures individual biometric data virtual private networks); U.S. Pat. No. 6,075,854 (Fully privacy); U.S. Pat. No. 6,037,870 (Detector system for flexible routing service for an advanced intelligent network); access control, and a detector assembly for implementing U.S. Pat. No. 6,075,852 (Telecommunications system and such a system): U.S. Pat. No. 6,035,406 (Plurality-factor method for processing call-independent signalling transac security system); U.S. Pat. No. 6,035,402 (Virtual certificate tions); U.S. Pat. No. 6,073,172 (Initializing and reconfigur authority); U.S. Pat. No. 6,035.398 (Cryptographic key ing a secure network interface); U.S. Pat. No. 6,061,796 generation using biometric data); U.S. Pat. No. 6,031,910 (Multi-access virtual private network); U.S. Pat. No. 6,061, (Method and system for the secure transmission and storage 729 (Method and system for communicating service infor of protectable information); U.S. Pat. No. 6,026,166 (Digi mation in an advanced intelligent network); U.S. Pat. No. tally certifying a user identity and a computer system in US 2006/0167784 A1 Jul. 27, 2006
combination): U.S. Pat. No. 6,018,739 (Biometric personnel cal imaging device); U.S. Pat. No. 5,920,058 (Holographic identification system); U.S. Pat. No. 6,016,476 (Portable labeling and reading machine for authentication and security information and transaction processing system and method applications); U.S. Pat. No. 5,915,973 (System for admin utilizing biometric authorization and digital certificate Secu istration of remotely-proctored, secure examinations and rity); U.S. Pat. No. 6,012,049 (System for performing finan methods therefor); U.S. Pat. No. 5,913,196 (System and cial transactions using a smartcard); U.S. Pat. No. 6,012,039 method for establishing identity of a speaker); U.S. Pat. No. (Tokenless biometric electronic rewards system); U.S. Pat. 5.913,025 (Method and apparatus for proxy authentication); No. 6,011,858 (Memory card having a biometric template U.S. Pat. No. 5,912,974 (Apparatus and method for authen stored thereon and system for using same); U.S. Pat. No. tication of printed documents); U.S. Pat. No. 5,912.818 6,009, 177 (Enhanced cryptographic system and method (System for tracking and dispensing medical items); U.S. with key escrow feature); U.S. Pat. No. 6,006,328 (Com Pat. No. 5,910,988 (Remote image capture with centralized puter Software authentication, protection, and security sys processing and storage); U.S. Pat. No. 5,907,149 (Identifi tem); U.S. Pat. No. 6,003,135 (Modular security device); cation card with delimited usage); U.S. Pat. No. 5,901.246 U.S. Pat. No. 6,002,770 (Method for secure data transmis (Ergonomic man-machine interface incorporating adaptive sion between remote stations); U.S. Pat. No. 5,999,637 pattern recognition based control system); U.S. Pat. No. (Individual identification apparatus for selectively recording 5,898,154 (System and method for updating security infor a reference pattern based on a correlation with comparative mation in a time-based electronic monetary system); U.S. patterns); U.S. Pat. No. 5,999,095 (Electronic security sys Pat. No. 5,897,616 (Apparatus and methods for speaker tem); U.S. Pat. No. 5,995,630 (Biometric input with encryp verification/identification/classification employing non tion); U.S. Pat. No. 5,991,431 (Mouse adapted to scan acoustic and/or acoustic models and databases); U.S. Pat. biometric data); U.S. Pat. No. 5,991,429 (Facial recognition No. 5,892.902 (Intelligent token protected system with system for security access and identification); U.S. Pat. No. network authentication); U.S. Pat. No. 5,892,838 (Biometric 5.991,408 (Identification and security using biometric mea recognition using a classification neural network); U.S. Pat. surements); U.S. Pat. No. 5,987,155 (Biometric input device No. 5,892.824 (Signature capture/verification systems and with peripheral port); U.S. Pat. No. 5,987,153 (Automated methods); U.S. Pat. No. 5,890,152 (Personal feedback verification and prevention of spoofing for biometric data); browser for obtaining media files); U.S. Pat. No. 5,889,474 U.S. Pat. No. 5,986,746 (Topographical object detection (Method and apparatus for transmitting Subject status infor system): U.S. Pat. No. 5,984,366 (Unalterable self-verifying mation over a wireless communications network); U.S. Pat. articles); U.S. Pat. No. 5,982,894 (System including sepa No. 5,881,226 (Computer security system); U.S. Pat. No. rable protected components and associated methods); U.S. 5,878,144 (Digital certificates containing multimedia data Pat. No. 5,979,773 (Dual smart card access control elec extensions); U.S. Pat. No. 5,876,926 (Method, apparatus and tronic data storage and retrieval system and methods); U.S. system for verification of human medical data); U.S. Pat. Pat. No. 5,978.494 (Method of selecting the best enroll No. 5,875,108 (Ergonomic man-machine interface incorpo image for personal identification); U.S. Pat. No. 5,974,146 rating adaptive pattern recognition based control system); (Real time bank-centric universal payment system); U.S. U.S. Pat. No. 5,872,849 (Enhanced cryptographic system Pat. No. 5,970,143 (Remote-auditing of computer generated and method with key escrow feature); U.S. Pat. No. 5,872, outcomes, authenticated billing and access control, and 848 (Method and apparatus for witnessed authentication of Software metering system using cryptographic and other electronic documents); U.S. Pat. No. 5,872,834 (Telephone protocols); U.S. Pat. No. 5,966,446 (Time-bracketing infra with biometric sensing device); U.S. Pat. No. 5,870,723 structure implementation): U.S. Pat. No. 5,963,908 (Secure (Tokenless biometric transaction authorization method and logon to notebook or desktop computers); U.S. Pat. No. system); U.S. Pat. No. 5,869,822 (Automated fingerprint 5.963,657 (Economical skin-pattern-acquisition and analy identification system); U.S. Pat. No. 5,867,802 (Biometri sis apparatus for access control; systems controlled thereby); cally secured control system for preventing the unauthorized U.S. Pat. No. 5,954,583 (Secure access control system); U.S. use of a vehicle); U.S. Pat. No. 5,867,795 (Portable elec Pat. No. 5,952,641 (Security device for controlling the tronic device with transceiver and visual image display); access to a personal computer or to a computer terminal); U.S. Pat. No. 5,867.578 (Adaptive multi-step digital signa U.S. Pat. No. 5,951,055 (Security document containing ture system and method of operation thereof): U.S. Pat. No. encoded data block); U.S. Pat. No. 5,949.881 (Apparatus 5,862.260 (Methods for surveying dissemination of propri and method for cryptographic companion imprinting); U.S. etary empirical data); U.S. Pat. No. 5,862.246 (Knuckle Pat. No. 5,949.879 (Auditable security system for the gen profile identity verification system); U.S. Pat. No. 5,862.223 eration of cryptographically protected digital data); U.S. Pat. (Method and apparatus for a cryptographically-assisted No. 5,949,046 (Apparatus for issuing integrated circuit commercial network system designed to facilitate and Sup cards); U.S. Pat. No. 5,943,423 (Smart token system for port expert-based commerce); U.S. Pat. No. 5,857,022 secure electronic transactions and identification); U.S. Pat. (Enhanced cryptographic system and method with key No. 5,935,071 (Ultrasonic biometric imaging and identity escrow feature); U.S. Pat. No. 5,850,451 (Enhanced cryp verification system): U.S. Pat. No. 5,933,515 (User identi tographic system and method with key escrow feature); U.S. fication through sequential input of fingerprints); U.S. Pat. Pat. No. 5,850.442 (Secure worldwide electronic commerce No. 5,933,498 (System for controlling access and distribu over an open network); U.S. Pat. No. 5,848.231 (System tion of digital property); U.S. Pat. No. 5,930,804 (Web configuration contingent upon secure input); U.S. Pat. No. based biometric authentication system and method); U.S. 5,844,244 (Portable identification carrier); U.S. Pat. No. Pat. No. 5,923,763 (Method and apparatus for secure docu 5.841,907 (Spatial integrating optical correlator for verify ment timestamping); U.S. Pat. No. 5,920,477 (Human fac ing the authenticity of a person, product or thing); U.S. Pat. tored interface incorporating adaptive pattern recognition No. 5,841,886 (Security system for photographic identifi based controller apparatus); U.S. Pat. No. 5,920,384 (Opti cation); U.S. Pat. No. 5,841,865 (Enhanced cryptographic US 2006/0167784 A1 Jul. 27, 2006
system and method with key escrow feature); U.S. Pat. No. a security device and unauthorized access to a resource); 5.841,122 (Security structure with electronic smart card U.S. Pat. No. 5,742,685 (Method for verifying an identifi access thereto with transmission of power and data between cation card and recording verification of same); U.S. Pat. the Smart card and the Smart card reader performed capaci No. 5,742,683 (System and method for managing multiple tively or inductively): U.S. Pat. No. 5,838,812 (Tokenless users with different privileges in an open metering system); biometric transaction authorization system); U.S. Pat. No. U.S. Pat. No. 5,737,420 (Method for secure data transmis 5,832,464 (System and method for efficiently processing sion between remote stations); U.S. Pat. No. 5,734,154 payments via check and electronic funds transfer); U.S. Pat. (Smart card with integrated reader and visual image dis No. 5,832,119 (Methods for controlling systems using con play); U.S. Pat. No. 5,719,950 (Biometric, personal authen trol signals embedded in empirical data); U.S. Pat. No. tication system); U.S. Pat. No. 5,712,914 (Digital certificates 5,828,751 (Method and apparatus for secure measurement containing multimedia data extensions); U.S. Pat. No. 5,712, certification); U.S. Pat. No. 5,825,880 (Multi-step digital 912 (Method and apparatus for securely handling a personal signature method and system); U.S. Pat. No. 5,825,871 identification number or cryptographic key using biometric (Information storage device for storing personal identifica techniques); U.S. Pat. No. 5,706,427 (Authentication tion information): U.S. Pat. No. 5,815,577 (Methods and method for networks); U.S. Pat. No. 5,703,562 (Method for apparatus for securely encrypting data in conjunction with a transferring data from an unsecured computer to a secured personal computer); U.S. Pat. No. 5,815.252 (Biometric computer); U.S. Pat. No. 5,696,827 (Secure cryptographic identification process and system utilizing multiple param methods for electronic transfer of information); U.S. Pat. eters scans for reduction of false negatives); U.S. Pat. No. No. 5,682,142 (Electronic control system/network); U.S. 5,805,719 (Tokenless identification of individuals); U.S. Pat. Pat. No. 5,682,032 (Capacitively coupled identity verifica No. 5,802,199 (Use sensitive identification system); U.S. tion and escort memory apparatus); U.S. Pat. No. 5,680,460 Pat. No. 5,799,088 (Non-deterministic public key encryp (Biometric controlled key generation); U.S. Pat. No. 5,668, tion system): U.S. Pat. No. 5,799,086 (Enhanced crypto 878 (Secure cryptographic methods for electronic transfer of graphic system and method with key escrow feature); U.S. information); U.S. Pat. No. 5,666,400 (Intelligent recogni Pat. No. 5,799,083 (Event verification system); U.S. Pat. No. tion): U.S. Pat. No. 5,659,616 (Method for securely using 5,790,674 (System and method of providing system integ digital signatures in a commercial cryptographic system); rity and positive audit capabilities to a positive identification U.S. Pat. No. 5,647.364 (Ultrasonic biometric imaging and system); U.S. Pat. No. 5,790,668 (Method and apparatus for identity verification system); U.S. Pat. No. 5,647,017 securely handling data in a database of biometrics and (Method and system for the verification of handwritten associated data): U.S. Pat. No. 5,789,733 (Smart card with signatures); U.S. Pat. No. 5,646,839 (Telephone-based per contactless optical interface); U.S. Pat. No. 5,787,187 (Sys sonnel tracking system); U.S. Pat. No. 5,636,282 (Method tems and methods for biometric identification using the for dial-in access security using a multimedia modem); U.S. acoustic properties of the ear canal); U.S. Pat. No. 5,784,566 Pat. No. 5,633,932 (Apparatus and method for preventing (System and method for negotiating security services and disclosure through user-authentication at a printing node); algorithms for communication across a computer network); U.S. Pat. No. 5,615,277 (Tokenless security system for U.S. Pat. No. 5,784,461 (Security system for controlling authorizing access to a secured computer system); U.S. Pat. access to images and image related services); U.S. Pat. No. No. 5,613,012 (Tokenless identification system for authori 5,774,551 (Pluggable account management interface with Zation of electronic transactions and electronic transmis unified login and logout and multiple user authentication sions); U.S. Pat. No. 5,608.387 (Personal identification services); U.S. Pat. No. 5,771,071 (Apparatus for coupling devices and access control systems); U.S. Pat. No. 5,594,806 multiple data sources onto a printed document); U.S. Pat. (Knuckle profile identity verification system); U.S. Pat. No. No. 5,770,849 (Smart card device with pager and visual 5,592.408 (Identification card and access control device); image display); U.S. Pat. No. 5,768,382 (Remote-auditing U.S. Pat. No. 5,588,059 (Computer system and method for of computer generated outcomes and authenticated billing secure remote communication sessions); U.S. Pat. No. and access control system using cryptographic and other 5,586,171 (Selection of a voice recognition data base protocols); U.S. Pat. No. 5,767,496 (Apparatus for process responsive to video data): U.S. Pat. No. 5,583,950 (Method ing symbol-encoded credit card information); U.S. Pat. No. and apparatus for flash correlation); U.S. Pat. No. 5,583.933 5,764,789 (Tokenless biometric ATM access system); U.S. (Method and apparatus for the secure communication of Pat. No. 5,763,862 (Dual card Smart card reader); U.S. Pat. data): U.S. Pat. No. 5,578,808 (Data card that can be used No. 5,761,298 (Communications headset with universally for transactions involving separate card issuers); U.S. Pat. adaptable receiver and voice transmitter); U.S. Pat. No. No. 5,572.596 (Automated, non-invasive iris recognition 5,757,916 (Method and apparatus for authenticating the system and method); U.S. Pat. No. 5,561.718 (Classifying location of remote users of networked computing systems); faces); U.S. Pat. No. 5,559,885 (Two stage read-write U.S. Pat. No. 5,757,431 (Apparatus for coupling multiple method for transaction cards); U.S. Pat. No. 5,557,765 data sources onto a printed document); U.S. Pat. No. 5,751, (System and method for data recovery); U.S. Pat. No. 836 (Automated, non-invasive iris recognition system and 5,553,155 (Low cost method employing time slots for method); U.S. Pat. No. 5,751,809 (Apparatus and method thwarting fraud in the periodic issuance of food stamps, for securing captured data transmitted between two sources); unemployment benefits or other governmental human Ser U.S. Pat. No. 5,748,738 (System and method for electronic vices); U.S. Pat. No. 5,544,255 (Method and system for the transmission, storage and retrieval of authenticated docu capture, storage, transport and authentication of handwritten ments); U.S. Pat. No. 5,745,573 (System and method for signatures); U.S. Pat. No. 5,534.855 (Method and system for controlling access to a user secret); U.S. Pat. No. 5,745,555 certificate based alias detection); U.S. Pat. No. 5,533,123 (System and method using personal identification numbers (Programmable distributed personal security); U.S. Pat. No. and associated prompts for controlling unauthorized use of 5,526,428 (Access control apparatus and method); U.S. Pat. US 2006/0167784 A1 Jul. 27, 2006
No. 5,523,739 (Metal detector for control of access com specific plug-in application modules in multi-issued trans bined in an integrated form with a transponder detector); action device); U.S. Pat. No. 5,020,105 (Field initialized U.S. Pat. No. 5,497.430 (Method and apparatus for image authentication system for protective security of electronic recognition using invariant feature signals); U.S. Pat. No. information networks); U.S. Pat. No. 4,993,068 (Unforget 5,485,519 (Enhanced security for a secure token code); U.S. table personal identification system); U.S. Pat. No. 4,972, Pat. No. 5,485,312 (Optical pattern recognition system and 476 (Counterfeit proof ID card having a scrambled facial method for verifying the authenticity of a person, product or image); U.S. Pat. No. 4,961,142 (Multi-issuer transaction thing); U.S. Pat. No. 5,483.601 (Apparatus and method for device with individual identification verification plug-in biometric identification using silhouette and displacement application modules for each issuer); U.S. Pat. No. 4,952, images of a portion of a person's hand); U.S. Pat. No. 928 (Adaptable electronic monitoring and identification 5,478,993 (Process as safety concept against unauthorized system); U.S. Pat. No. 4.941,173 (Device and method to use of a payment instrument in cashless payment at payment render secure the transfer of data between a videotex ter sites); U.S. Pat. No. 5,475.839 (Method and structure for minal and a server); U.S. Pat. No. 4,926,480 (Card-computer securing access to a computer system); U.S. Pat. No. 5,469, moderated systems); U.S. Pat. No. 4,896,363 (Apparatus 506 (Apparatus for verifying an identification card and and method for matching image characteristics such as identifying a person by means of a biometric characteristic); fingerprint minutiae): U.S. Pat. No. 4,890.323 (Data com U.S. Pat. No. 5.457,747 (Anti-fraud verification system munication systems and methods); U.S. Pat. No. 4,868,376 using a data card); U.S. Pat. No. 5,455,407 (Electronic (Intelligent portable interactive personal data system); U.S. monetary system); U.S. Pat. No. 5,453,601 (Electronic Pat. No. 4,827,518 (Speaker verification system using inte monetary system); U.S. Pat. No. 5,448,045 (System for grated circuit cards); U.S. Pat. No. 4,819,267 (Solid state protecting computers via intelligent tokens or Smart cards); key for controlling access to computer systems and to U.S. Pat. No. 5,432,864 (Identification card verification computer Software and/or for secure communications); U.S. system); U.S. Pat. No. 5,414.755 (System and method for Pat. No. 4,752,676 (Reliable secure, updatable “cash” card passive voice verification in a telephone network); U.S. Pat. system): U.S. Pat. No. 4,736,203 (3D hand profile identifi No. 5,412,727 (Anti-fraud voter registration and voting cation apparatus); U.S. Pat. No. 4,731.841 (Field initialized system using a data card); U.S. Pat. No. 5.363,453 (Non authentication system for protective security of electronic minutiae automatic fingerprint identification system and information networks); U.S. Pat. No. 4,564,018 (Ultrasonic methods); U.S. Pat. No. 5,347,580 (Authentication method system for obtaining ocular measurements), each of which is and system with a smartcard); U.S. Pat. No. 5,345,549 expressly incorporated herein by reference. (Multimedia based security systems); U.S. Pat. No. 5,341, 428 (Multiple cross-check document verification system); 0060) E-Commerce Systems U.S. Pat. No. 5,335,288 (Apparatus and method for biomet 0061 U.S. Pat. No. 5,946,669 (Polk, Aug. 31, 1999), ric identification); U.S. Pat. No. 5,291,560 (Biometric per expressly incorporated herein by reference, relates to a sonal identification system based on iris analysis); U.S. Pat. method and apparatus for payment processing using debit No. 5.283,431 (Optical key security access system); U.S. based electronic funds transfer and disbursement processing Pat. No. 5,280,527 (Biometric token for authorizing access using addendum-based electronic data interchange. This to a host system); U.S. Pat. No. 5.272,754 (Secure computer disclosure describes a payment and disbursement system, interface); U.S. Pat. No. 5.245,329 (Access control system wherein an initiator authorizes a payment and disbursement with mechanical keys which store data); U.S. Pat. No. to a collector and the collector processes the payment and 5.229,764 (Continuous biometric authentication matrix): disbursement through an accumulator agency. The accumu U.S. Pat. No. 5.228,094 (Process of identifying and authen lator agency processes the payment as a debit-based trans ticating data characterizing an individual); U.S. Pat. No. action and processes the disbursement as an addendum 5,224,173 (Method of reducing fraud in connection with based transaction. The processing of a debit-based employment, public license applications, social security, transaction generally occurs by electronic funds transfer food stamps, welfare or other government benefits); U.S. (EFT) or by financial electronic data interchange (FEDI). Pat. No. 5.208.858 (Method for allocating useful data to a The processing of an addendum-based transaction generally specific originator); U.S. Pat. No. 5,204.670 (Adaptable occurs by electronic data interchange (EDI). electric monitoring and identification system); U.S. Pat. No. 5,191,611 (Method and apparatus for protecting material on 0062 U.S. Pat. No. 6,005,939 (Fortenberry, et al., Dec. storage media and for transferring material on storage media 21, 1999), expressly incorporated herein by reference, to various recipients); U.S. Pat. No. 5,163,094 (Method for relates to a method and apparatus for storing an Internet identifying individuals from analysis of elemental shapes user's identity and access rights to World Wide Web derived from biosensor data); U.S. Pat. No. 5,155,680 resources. A method and apparatus for obtaining user infor (Billing system for computing software); U.S. Pat. No. mation to conduct secure transactions on the Internet with 5,131,038 (Portable authentification system): U.S. Pat. No. out having to re-enter the information multiple times is 5,073.950 (Finger profile identification system): U.S. Pat. described. The method and apparatus can also provide a No. 5,067,162 (Method and apparatus for verifying identity technique by which secured access to the data can be using image correlation); U.S. Pat. No. 5,065,429 (Method achieved over the Internet. A passport containing user and apparatus for protecting material on storage media); defined information at various security levels is stored in a U.S. Pat. No. 5,056,147 (Recognition procedure and an secure server apparatus, or passport agent, connected to apparatus for carrying out the recognition procedure); U.S. computer network. A user process instructs the passport Pat. No. 5,056,141 (Method and apparatus for the identifi agent to release all or portions of the passport to a recipient cation of personnel); U.S. Pat. No. 5,036,461 (Two-way node and forwards a key to the recipient node to unlock the authentication system between user's Smart card and issuer passport information. US 2006/0167784 A1 Jul. 27, 2006
0063 U.S. Pat. No. 6,016,484 (Williams, et al., Jan. 18, networks. Vouchers can also be transferred from one remote 2000), expressly incorporated herein by reference, relates to device to another remote device. These remote devices can a system, method and apparatus for network electronic communicate through a number of methods with each other. payment instrument and certification of payment and credit For example, for a non-face-to-face transaction the Internet collection utilizing a payment. An electronic monetary sys is a choice, for a face-to-face or close proximity transactions tem provides for transactions utilizing an electronic-mon tone signals or light signals are likely methods. In addition, etary system that emulates a wallet or a purse that is at the time of a transaction a digital receipt can be created customarily used for keeping money, credit cards and other which will facilitate a fast replacement of vouchers stored in forms of payment organized. Access to the instruments in a lost remote device. the wallet or purse is restricted by a password to avoid unauthorized payments. A certificate form must be com 0066 Micropayments pleted in order to obtain an instrument. The certificate form 0067 U.S. Pat. No. 5,999,919 (Jarecki, et al., Dec. 7, obtains the information necessary for creating a certificate 1999), expressly incorporated herein by reference, relates to granting authority to utilize an instrument, a payment holder an efficient micropayment system. Existing Software pro and a complete electronic wallet. Electronic approval results posals for electronic payments can be divided into “on-line' in the generation of an electronic transaction to complete the schemes which require participation of a trusted party (the order. If a user selects a particular certificate, a particular bank) in every transaction and are secure against overspend payment instrument holder will be generated based on the ing, and “off-line' schemes which do not require a third selected certificate. In addition, the issuing agent for the party and guarantee only that overspending is detected when certificate defines a default bitmap for the instrument asso vendors submit their transaction records to the bank (usually ciated with a particular certificate, and the default bitmap at the end of the day). A new “hybrid’ scheme is proposed will be displayed when the certificate definition is com which combines the advantages of both “on-line' and “off pleted. Finally, the number associated with a particular line' electronic payment schemes. It allows for control of certificate will be utilized to determine if a particular party overspending at a cost of only a modest increase in com can issue a certificate. munication compared to the off-line schemes. The protocol 0064 U.S. Pat. No. 6,029,150 (Kravitz, Feb. 22, 2000), is based on probabilistic polling. During each transaction, expressly incorporated herein by reference, relates to a with some small probability, the vendor forwards informa system and method of payment in an electronic payment tion about this transaction to the bank. This enables the bank system wherein a plurality of customers have accounts with to maintain an accurate approximation of a customers an agent. A customer obtains an authenticated quote from a spending. The frequency of polling messages is related to specific merchant, the quote including a specification of the monetary value of transactions and the amount of goods and a payment amount for those goods. The customer overspending the bank is willing to risk. For transactions of sends to the agent a single communication including a high monetary value, the cost of polling approaches that of request for payment of the payment amount to the specific the on-line schemes, but for micropayments, the cost of merchant and a unique identification of the customer. The polling is a small increase over the traffic incurred by the agent issues to the customer an authenticated payment off-line schemes. advice based only on the single communication and secret shared between the customer and the agent and status 0068 Micropayments are often preferred where the information, which the agent knows about the merchant, amount of the transaction does not justify the costs of and/or the customer. The customer forwards a portion of the complete financial security. In the micropayment Scheme, payment advice to the specific merchant. The specific mer typically a direct communication between creditor and chant provides the goods to the customer in response to debtor is not required; rather, the transaction produces a receiving the portion of the payment advice. result which eventually results in an economic transfer, but which may remain outstanding Subsequent to transfer of the 0065 U.S. Pat. No. 6,047,269 (Biffar, Apr. 4, 2000), underlying goods or services. The theory underlying this expressly incorporated herein by reference, relates to a micropayment scheme is that the monetary units are small self-contained payment system with creating and facilitating enough Such that risks of failure in transaction closure is transfer of circulating digital vouchers representing value. A relatively insignificant for both parties, but that a user gets digital voucher has an identifying element and a dynamic few chances to default before credit is withdrawn. On the log. The identifying element includes information Such as other hand, the transaction costs of a non-real time transac the transferable value, a serial number and a digital signa tions of Small monetary units are Substantially less than ture. The dynamic log records the movement of the voucher those of secure, unlimited or potentially high value, real time through the system and accordingly grows over time. This verified transactions, allowing and facilitating Such types of allows the system operator to not only reconcile the vouch commerce. Thus, the rights management system may ers before redeeming them, but also to recreate the history employ applets local to the client system, which communi of movement of a voucher should an irregularity like a cate with other applets and/or the server and/or a vendorf duplicate voucher be detected. These vouchers are used rights-holder to validate a transaction, at low transactional within a self-contained system including a large number of COStS. remote devices that are linked to a central system. The central system can e linked to an external system. The 0069. The following U.S. Patents, expressly incorporated external system, as well as the remote devices, is connected herein by reference, define aspects of micropayment, digital to the central system by any one or a combination of certificate, and on-line payment systems: U.S. Pat. No. networks. The networks must be able to transport digital 5,930,777 (Barber, Jul. 27, 1999, Method of charging for information, for example the Internet, cellular networks, pay-per-access information over a network); U.S. Pat. No. telecommunication networks, cable networks or proprietary 5,857,023 (Jan. 5, 1999, Demers et al., Space efficient US 2006/0167784 A1 Jul. 27, 2006
method of redeeming electronic payments); U.S. Pat. No. and method); U.S. Pat. No. 5,754,939 (May 1998, Herz et 5,815,657 (Sep. 29, 1998, Williams, System, method and al., System for generation of user profiles for a system for article of manufacture for network electronic authorization customized electronic identification of desirable objects); utilizing an authorization instrument); U.S. Pat. No. 5,793, U.S. Pat. No. 5,768,385 (Jun. 16, 1998, Untraceable elec 868 (Aug. 11, 1998, Micali, Certificate revocation system), tronic cash); U.S. Pat. No. 5,799,087 (Aug. 25, 1998, U.S. Pat. No. 5,717,757 (Feb. 10, 1998, Micali, Certificate Electronic-monetary system); U.S. Pat. No 5,812,668 issue lists); U.S. Pat. No. 5,666,416 (Sep. 9, 1997, Micali, (Sep.22, 1998, System, method and article of manufacture Certificate revocation system); U.S. Pat. No. 5,677,955 for verifying the operation of a remote transaction clearance (Doggett et al., Electronic funds transfer instruments); U.S. system utilizing a multichannel, extensible, flexible archi Pat. No. 5,839,119 (Nov. 17, 1998, Krsul; et al., Method of electronic payments that prevents double-spending); U.S. tecture); U.S. Pat. No. 5,828,840 (Oct. 27, 1998, Server for Pat. No. 5,915,093 (Berlin et al.); 5,937.394 (Wong, et al.); starting client application on client if client is network U.S. Pat. No. 5,933,498 (Schneck et al.); U.S. Pat. No. terminal and initiating client application on server if client is 5,903,880 (Biffar); U.S. Pat. No. 5,903,651 (Kocher); 5,884, non network terminal); U.S. Pat. No. 5,832,089 (Nov. 3, 277 (Khosla); U.S. Pat. No. 5,960,083 (Sep. 28, 1999, 1998, Off-line compatible electronic cash method and sys Micali, Certificate revocation system); U.S. Pat. No. 5,963, tem); U.S. Pat. No. 5,850,446 (Dec. 15, 1998, System, 924 (Oct. 5, 1999, Williams et al., System, method and method and article of manufacture for virtual point of sale article of manufacture for the use of payment instrument processing utilizing an extensible, flexible architecture); holders and payment instruments in network electronic U.S. Pat. No. 5,889,862 (Mar. 30, 1999, Method and appa commerce); U.S. Pat. No. 5,996,076 (Rowney et al., System, ratus for implementing traceable electronic cash); U.S. Pat. method and article of manufacture for secure digital certi No. 5,889,863 (Mar. 30, 1999, System, method and article of fication of electronic commerce); U.S. Pat. No. 6,016,484 manufacture for remote virtual point of sale processing (Jan 18, 2000, Williams et al., System, method and article of utilizing a multichannel, extensible, flexible architecture); manufacture for network electronic payment instrument and U.S. Pat. No. 5,898,154 (Apr. 27, 1999, System and method certification of payment and credit collection utilizing a for updating security information in a time-based electronic payment); U.S. Pat. No. 6,018,724 (Arent); U.S. Pat. No. monetary system); U.S. Pat. No. 5,901.229 (May 4, 1999, 6,021.202 (Anderson et al., Method and system for process Electronic cash implementing method using a trustee); U.S. ing electronic documents); U.S. Pat. No. 6,035,402 (Vaeth et Pat. No. 5,920,629 (Jul. 6, 1999, Electronic-monetary sys al.); U.S. Pat. No. 6,049,786 (Smorodinsky); U.S. Pat. No. tem); U.S. Pat. No. 5,926,548 (Jul. 20, 1999, Method and 6,049,787 (Takahashi, et al.); U.S. Pat. No. 6,058,381 (Nel apparatus for implementing hierarchical electronic cash); son, Many-to-many payments system for network content U.S. Pat. No. 5,943,424 (Aug. 24, 1999, System, method materials); U.S. Pat. No. 6,061.448 (Smith, et al.); U.S. Pat. and article of manufacture for processing a plurality of No. 5,987,132 (Nov. 16, 1999, Rowney, System, method and transactions from a single initiation point on a multichannel, article of manufacture for conditionally accepting a payment extensible, flexible architecture); U.S. Pat. No. 5,949,045 method utilizing an extensible, flexible architecture); U.S. (Sep. 7, 1999, Micro-dynamic simulation of electronic cash Pat. No. 6,057,872 (Candelore); and U.S. Pat. No. 6,061,665 transactions); U.S. Pat. No. 5,952,638 (Sep. 14, 1999, Space (May 9, 2000, Bahreman, System, method and article of efficient method of electronic payments); U.S. Pat. No. manufacture for dynamic negotiation of a network payment 5,963,648 (Oct. 5, 1999, Electronic-monetary system): U.S. framework). See also, Rivest and Shamir, “PayWord and Pat. No. 5,978,840 (System, method and article of manu MicroMint: Two Simple Micropayment Schemes' (May 7, facture for a payment gateway system architecture for 1996); Micro PAYMENT transfer Protocol (MPTP) Version processing encrypted payment transactions utilizing a mul 0.1 (22 Nov. 1995) et seq., http://www.w3.org/pub/WWW/ tichannel, extensible, flexible architecture); U.S. Pat. No. TR/WD-mptp: Common Markup for web Micropayment 5,983.208 (Nov. 9, 1999, System, method and article of Systems, http://www.w3.org/TR/WD-Micropayment manufacture for handling transaction results in a gateway Markup (09Jun. 1999); “Distributing Intellectual Property: a payment architecture utilizing a multichannel, extensible, Model of Microtransaction Based Upon Metadata and Digi flexible architecture); U.S. Pat. No. 5,987,140 (Nov. 16, tal Signatures. Olivia, Maurizio, http://olivia.modlang.den 1999, System, method and article of manufacture for secure ison.edu/-olivia/RFC/09/, all of which are expressly incor network electronic payment and credit collection); U.S. Pat. porated herein by reference. No. 6,002,767 (Dec. 14, 1999, System, method and article of manufacture for a modular gateway server architecture); 0070) See, also: U.S. Pat. No. 4,977,595 (Dec. 11, 1990, U.S. Pat. No. 6,003,765 (Dec. 21, 1999, Electronic cash Method and apparatus for implementing electronic cash); implementing method with a Surveillance institution, and U.S. Pat. No. 5.224,162 (Jun. 29, 1993, Electronic cash user apparatus and Surveillance institution apparatus for system): U.S. Pat. No. 5.237,159 (Aug. 17, 1993, Electronic implementing the same); U.S. Pat. No. 6,021,399 (Feb. 1, check presentment system); U.S. Pat. No. 5,392,353 (Feb 2000, Space efficient method of verifying electronic pay ruary 1995, Morales, TV Answer, Inc. Interactive satellite ments); U.S. Pat. No. 6,026,379 (Feb. 15, 2000, System, broadcast network); U.S. Pat. No. 5,511,121 (Apr. 23, 1996, method and article of manufacture for managing transac Efficient electronic money); U.S. Pat. No. 5,621.201 (April tions in a high availability system); U.S. Pat. No. 6,029,150 1997, Langhans et al., Visa International Automated pur (Feb. 22, 2000, Payment and transactions in electronic chasing control system); U.S. Pat. No. 5,623.547 (Apr. 22, commerce system): U.S. Pat. No. 6,029,151 (Feb. 22, 2000, 1997, Value transfer system): U.S. Pat. No. 5,679,940 (Octo Method and system for performing electronic money trans ber 1997, Templeton et al., TeleCheck International, Inc. actions); U.S. Pat. No. 6,047,067 (Apr. 4, 2000, Electronic Transaction system with on/off line risk assessment); U.S. monetary system); U.S. Pat. No. 6,047,887 (Apr. 11, 2000, Pat. No. 5,696,908 (December 1997, Muehlberger et al., System and method for connecting money modules); U.S. Southeast Phonecard, Inc. Telephone debit card dispenser Pat. No. 6,055,508 (Apr. 25, 2000, Method for secure US 2006/0167784 A1 Jul. 27, 2006 accounting and auditing on a communications network); encompassed, may be Supplemental and provide an under U.S. Pat. No. 6,065,675 (May 23, 2000, Processing system standing for one who is not expert in all areas encompassed and method for a heterogeneous electronic cash environ by the present disclosure. ment); U.S. Pat. No. 6,072,870 (Jun. 6, 2000, System, method and article of manufacture for a gateway payment TELEMATICS REFERENCES architecture utilizing a multichannel, extensible, flexible architecture), each of which is expressly incorporated herein 0082 On the Propagation of Long-Range Dependence in by reference. the Internet Veres, Kenesi, Molnár. (2000) galahad.elte.hu/ -Vattay/cikkeim/vmkv.pdf 0071 Neural Networks 0.083 Anthill: A Framework for the Development of 0072 The resources relating to Neural Networks, listed Agent-Based.—Babaoglu, Meling. (2002) www.cs.unibo.it/ in the Neural Networks References Appendix, each of which babaoglu/papers/icacs02.pdf is expressly incorporated herein by reference, provides a sound basis for understanding the field of neural networks 0084 Automatic Web Page Categorization by Link and (and the subset called artificial neural networks, which Context.—Attardi, Gulli. (1999) aure.iei.pi.cnr.it/-fabrizio/ distinguish biological systems) and how these might be used Publications/THAI99/THAI99.ps to solve problems. A review of these references will provide 0085 Distributed Interactive Media-Mauve (2000) a state of knowledge appropriate for an understanding of www.informatik.uni-mannheim.de/informatik/pi4/publica aspects of the invention which rely on Neural Networks, and tions/library/Mauve2000dps.gZ to avoid a prolix discussion of no benefit to those already possessing an appropriate state of knowledge. 0.086 Weak Bisimulation for Fully Probabilistic Pro cesses—Baier, Hermanns (1999) web.informatik.uni 0073 Telematics bonn.de/I/papers/acta infps 0074 The resources relating to telematics listed in the 0087 Structured Multimedia Authoring Hardman, van Telematics Appendix, each of which is expressly incorpo Rossum, Bulterman (1993) www.cwi.nl/ftp/CWIreports/ rated herein by reference, provides a background in the CST/CS-R9304ps.Z theory and practice of telematics, as well as some of the underlying technologies. A review of these references is 0088 Solving the modeling problems of object-oriented therefore useful in inderstanding practical issues and the languages.—Aksit, Tekinerdogan (1998) trese.cs.utwen context of functions and technologies which may be used in te.nl/aop-ecoop98/papers/Aksit.pdf conjunction with the advances set forth herein. 0089 Group Communication in Differentiated Services 0075 Game Theory Networks—Bless, Wehrle (2001) www.tm.uka.de/-bless/ IQ2001-ds multicast.pdf 0.076 The following resources listed in the Game Theory References Appendix, relating to Game Theory, each of 0090 Ad Hoc Relay Wireless Networks over Moving which is expressly incorporated herein by reference, pro Vehicles on Highways—Chen, Kung, Vlah (2001) vides a basis for understanding Game Theory and its impli www.eecs.harvard.edu/~htk/publication/2001-mobihoc-ck cations for the design, control, and analysis of systems and v.pdf networks. A review of these references will assure a back 0091) A Framework for Generating Network-Based ground in this field for an understanding of aspects of the Moving Objects—Brinkhoff (2002) www.fh-wilhelmshav invention which relate to game Theory. en.de/oow/institute/iapg/personen/brinkhoff/paper/ 0077 Use of Game Theory to Control Ad Hoc Networks GeoInformatica2002.pdf 0078. The resources relating to ad hoc networks and 0092. The Amsterdam Hypermedia Model: extending game theory listed in the Game Theory and Ad Hoc Netow hypertext.—Hardman, Bulterman. (1993) www.cwi.nl/ftp/ rks References Appendix, each of which is expressly incor CWIreports/CST/CS-R9306.ps.Z. porated herein by reference, provides a sound basis for 0093. On functional equivalence of certain fuzzy control understanding the implications of game theory for the lers and. Koczy, Tikk, Gedeon (2000) www.mft.hu/publi design, control and analysis of communications networks, cations/tikk/tikk3.pdf and in particular, ad hoc networks. A review of these references will assure a background in this field for an 0094 Specification of a Service Management Architec understanding of aspects of the invention which rely on ture. Mayerl, Nochta. (2000) www.cooperation-manage these topics. ment.de/publikationen/paper/usm2000 mayerl-nochta.pdf 0079) Other Patents by Inventor Hereof 0095 Agent-mediated Electronic Commerce: Scientific and Technological. Sierra (2001) www.iiia.csic.es/-sierra/ 0080. The following patents are expressly incorporated articles/2001/Roadmap AMEC/Roadmap.ps herein by reference: U.S. Patent Nos. 6,640,145, 6,418,424, 6,400,996, 6,081,750, 5,920,477, 5,903,454, 5,901,246, 0096. On the Integration of IR and Databases—de Vries, 5,875,108, 5,867,386, 5,774,357, 6,429,812, and 6,252,544. Wilschut (1999) www.cs.utwente.nl/~aren/Pics/ 0081. The following references are provided as enhance ds8 short.ps.gZ ments to the present disclosure and to provide a service to 0097 Towards a Harmonization of UML-Sequence Dia the art in consolidating these citations in a single location. A grams and MSC Rudolph, Grabowski, Graubmann (1999) review of all of these references is not believed necessary for www.itm.mu-luebeck.de/publications/SDL99-Harmoniza an understanding of the invention, but given the range of arts tion.ps.gZ US 2006/0167784 A1 Jul. 27, 2006
0.098 Logarithmic Time Cost Optimal Parallel Sorting is 0116. The Trust Factor in the Virtual Enterprise. Mobile Not Yet Fast in. Natvig (1996) www.idi.ntinu.no/-lasse? Broadband Service paula.oulu.fi/Publications/Submited/ publics/SC90.ps ICE2000fpdf 0099 CMIFed: A Presentation Environment for Por 0.117 User and Session Mobility in a Plug-and-Play table.—van Rossum. (1993) www.cwi.nl/ftp/CWIreports/ Network. Mazen Malek Shiaa www.item.ntinu.no/ CST/CS-R9305.ps.Z -plugandplay/publications/eunice2002.pdf 0100 Requirements of Traffic Telematics to Spatial Data 0118 Dynamic Aggregation of Reservations for Internet bases—Brinkhoff (1999) www.fh-oldenburg.de/iapg/per Services—Bless www.tm.uka.de/-bless/ictsm 10-daris.pdf sonen/brinkhof/SSD99.pdf 0119) Advanced Service Creation Using Distributed 0101. A Task-Specific Ontology for the Application Object Technology www.ee. Surrey.ac.uk/Personal/G.Pav and.—Shahar, Miksch, Johnson www-Smi. Stanford.edu/ lou/Publications/Journal-papers/Adam-02b.pdf pubs/SMI Reports/SMI-96-0649.pdf 0120 Semantics and Verification of UML Activity Dia 0102 Practical Considerations in Building a Multi-Lin grams for Workflow. Eshuis (2002) www.home.cs.utwen gual. Business Letters John acl.ldc.upenn.edu/W/W97/ te.nl/~eshuis/thesis-hyperlinks.pdf W97-0906.pdf 0121 be cited as a National Research Council report.— 0103) BPP: A Protocol for Exchanging Pricing Informa The Four Way www7.nationalacademies.org/CSTB/wp tion. Oberle, Ritter, Wehrle www.tm.uka.de/-wehrle/pubs/ geo armstrong.pdf 2001 07 HPSR01 BPP.pdf 0.122 Support Specification and Selection in TAPAS Aagesen, Anutariya, Shiaa, Helvik www.item.ntinu.no/ 0104. An Open Architecture for Evaluating Arbitrary -plugandplay/publications/euniceCap2002.pdf Quality of Service. Wehrle (2001) www.tm.uka.de/-we hirle/pubs/2001 06 ICNO1 KIDS.pdf 0123 Profiling and Internet Connectivity in Automo tive.—Cilia, Hasselmeyer. www.cs.ust.hk/vldb2002/NV 0105 Improving the performance of TCP on guaranteed LDB2002-papers/S33P08.pdf bandwidth. Ritter, Wehrle, Wolf www.tm.uka.de/-wehrle? pubs/2001 02 KIVS01 GR TCP.pdf 0.124 MIETTA A Framework for Uniform and Multi lingual Access to—Structured Database www.xtramind 0106 Efficient Image Compression of Medical Images .com/-holger/paper/ira 100.pdf Using the Wavelet Transform And Fuzzy rtsimage.di.uo a.gr/publications/euro00-1.pdf 0.125 Definition And Utilisation Of OMG IDL TO TTCN-3 MAPPINGS Ebner, Yin, Li www.itm.mu-lue 0107 Advanced Mechanisms for Available Rate Usage beck.de/publications/idltottcn3mapping.pdf in ATM.—Bless, Holzhausen. www.tm.uka.de/-wehrle? pubs/2001 04 ICATM01 AvRateUsage.pdf 0.126 Campiello New user interface approaches for community.—Antonietta Grasso Michael (1998) 0108) An Extensible Agent Architecture for a Competi www11.informatik.tu-muenchen.de/publications/pdf/ tive. Hoen, Bohte: www.cwi.nl/~hoen/publications/SEN Grasso 1998a.pdf R0217.ps 0127 SDL and MSC Based Test Generation for Distrib 0109 Elements of an Open Framework for Pricing in the uted.—Grabowski, Koch.(1998) www.itm.mu-luebeck.de/ Future. Gerke, Ritter. www.tm.uka.de/-wehrle/pubs/ publications/SDL99-DistributedTesting.ps.gZ 2000 09 Qofis00 Pricing.pdf 0.128 Autolink—A Tool for Automatic Test Generation 0110 Traffic Priorization and Differentiation with Active from. Koch, Grabowski. (1998) www.itm.mu-luebeck.de/ Queue.—Walter, Wehrle (2002) www.tm.uka.de/-wehrle/ publications/A-98-05Report-A-98-05.ps.gZ pubs/2002 09 ICTSM02 AF AR.pdf 0129 PATROCLOS: A Flexible and High-Performance 0111 A Simulation Suite for Internet Nodes with the Transport Subsystem—Braun (1994) www.iam.unibe.ch/ Ability to. Wehrle, Reber, Kahmann www.tm.uka.de/ -braun/lit/vanc p.ps.gZ wehrle/pubs/2001 01 CNDS01 Simpdf 0.130 Multilingual Generation and Summarization of Job. Somers, Black. (1997) sskkili.gu.se/Tree/Reports/ 0112 Image Compression Using the Wavelet Trans anlp97/ANLP paperps form. On Textural Regions of Interest rtsimage.di.uoa.gr/ publications/euro98.pdf 0131 Real-time TTCN for testing real-time and multi media systems Walter, Grabowski (1997) www.itm.mu 0113 Towards Better Support of Transaction Oriented.— luebeck.de/publications/IWTCS98RT/IWTCS98-RTTTC Bless, Holzhausen. www.tm.uka.de/ bless/Qof IS2001-qf N.ps.gZ online.pdf 0132) Applying SAMSTAG to the B-ISDN protocol 0114. User feedback on the use of public key certifi SSCOP Grabowski, Scheurer, Dai. (1997) www.itm.mu cates—Klobucar (2001) www.e5.js.si/staff/tomaZ/ luebeck.de/publications/IWTCS97SSCOP/IWTCS97-SS erk2001.pdf COPps.gZ 0115 Using Realistic Internet Topology Data for Large 0.133 A Continuously Available and Highly Scalable.— Scale Network. Bless www.trn.uka.de/ blessfom Hvasshovd. (1991) www.idi.ntinu.no/IDT/grupper/DB-grp/ net ws 2002-1.pdf tech paperS/hpts91.ps US 2006/0167784 A1 Jul. 27, 2006
0134) A Modular VLSI Implementation Architecture 0152 Verifying Business Processes using SPIN Jans for. Braun, Schiller. (1994) www.iam.unibe.ch/-braun/lit/ sen, Mateescu, Mauw. (1998) ftp.win.tue.nl/pub/techre Vanc bSZ.ps.gZ ports/sjouke/spin98.ps.Z 0135 Transportation modelling methods and advanced 0153. Implementation of a Parallel Transport Subsystem transport telematics. Toint (1993) thales...math.fundp.ac.be/ on a.—Braun, Schmidt (1993) www.telematik.informa pub/reports/TR92-24.ps tik.uni-karlsruhe.def-Schmidt/schmidt hpdc93.ps.gZ 0.136 Impact of ATM ABR Control on the Performance 0154 Uniform Versioning: The Change-Oriented of TCP-Tahoe and. Feng, Ghosal networks.cs.uc Model Munch, Larsen, Gulla. (1993) www.idt.unit.no/ davis.edu/~ghosal/globecom/globecom paperps ~bjornmu?cov-Scm4.ps 0137) Detecting Opportunities for Parallel Observations O155 Architectural Considerations for Personal Mobility on the. Michael Lucks Space (1992) www.stsci.edu/pub In The Wireless Internet www.item.ntinu.no/ plugand licfsst/poms/poms.ps play/publications/PWC2002.pdf 0138 Expression Of Interest Network Of Excellence 0156 Synchronization. In Ad Hoc Networks Based On www.ctitutwente.nl/interna/info-programmes/IST/Eol UTRA TDD Ebner, Rohling, Halfmann, Lott (2002) NoE/eHealth.pdf www.et2...tu-harburg.de/Mitarbeiter/Ebner/ 0139 Optimisation of trapezoidal membership functions PIMRC02 Ebner.pdf in a fuzzy rule System. By The www.univ-savoie.fr/labos/ O157 Pervasive Privacy with Identity Management lamii/Busefal/Documents/85-05.pdf Jendricke, Kreutzer, Zugenmaier www.teco.edu/ philip/ 0140) A Framework For Video Modelling Petkovic, ubicomp2002ws/organize/identity.pdf Jonker (2000) www.cwi.nl/-acoi/DMW/documents/./world/ 0158 Formalising Hinting in Tutorial Dialogues Tsov innsb.ps altzi, Matheson (2002) www.ltg.ed.ac.uk/edilog/papers/ 0141 An Architecture For Reverse Charging. In The 185.pdf Internet Sprenkels, Parhonyi, Pras... (2000) wwwSnmp.c- 0159 TimedTTCN-3 A Real-Time Extension For S.utwente.nl/nm/research/results/publications/sprenkels/ TTCN-3—Dai, Grabowski, Neukirchen www.itm.mu-lue Sprenkelsram-reverse-charging.pdf beck.de/publications/timedttcn3.pdf 0142. Behavior of a Bayesian Adaptation Method for.— 0160 Mapping CORBAIDL to TTCN-3 based on IDL to Fredouille, MariéthoZ. (2000) ftp.idiap.ch/pub/marietho/ TTCN-2 mappings—Ebner (2001) www.itm.mu-lue publications/AdaptICASSP2000 IRR02ps.gZ beck.de/publications/FBT2001/fbt2001 ebner.pdf 0143 Software Process Modeling and Evolution in 0.161 Mobile Identity Management—Jendricke, EPOS Jaccheri, Larsen, Conradi (1992) www.idt.unit.no/ Kreutzer, Zugenmaier ftp.informatik.uni-freiburg.de/docu ~epos/Papers/capri-final.ps ments/reports/report178/report00178.ps.gZ 0144) Individual Management of Personal Reachability 0162 Causal Reasoning. In A Medical Knowledge in. Reichenbach. (1997) www.inftu-dresden.de/-hf2/ Based System www.newcastle.edu.au/school/design-comm publ/1997/RDFR 97IFIPSec.pdf info/staff/gibbon/ecis98.ps 0145 A Generic Software Component Framework for Distributed.—Batteram, Idzenga (1997) www.trc.nl/events/ 0163 A Secure Web-based Medical Digital Library– ecscw97oogp/papers/batteram.pdf Architecture Based On thalis.cs.unipi.gr/-pap/medical.pdf 0146 Efficient Partitioning of Sequences—Olstad, 0.164 IPv6 Autoconfiguration in Large Scale Mobile Ad Manne (1995) www.ii.uib.no/-fredrikm/fredrik/papers/effi Hoc Networks—Kilian Weniger Martina www.iponair.de/ cient.ps publications/Weniger-EuroWireless02.pdf 0147 Integrated Video Archive Tools—Hjelsvold, 0.165 Serviceware Framework For Developing 3g Langorgen, Midtstraum. (1995) www.idt.ntinu.no/IDT/grup Mobile Services Sahin Albayrak (2002) www.ssgrrit/en/ per/DB-grp/tech papers/ACM-MM95.ps ssgrr2002w?papers/37.pdf 0148 MULINEX: Multilingual Web Search and Naviga 0166 Traffic Controlmethods For High Speed Packet tion Joanne Capstick Abdel (1998) speech.ftw.at/~gor/ Switched Networks (2002) hsnlab.ttt.bme.hu/~molnar/files/ pub/twilt 98/mulinex-twilt 98.pdf pch2002.pdf 0149 A Non-Interleaving Semantics for MSC Heymer 0.167 The Impact Of Filtering On Spatial Continuous (1998) www.itm.mu-luebeck.de/publications/ANIS/ Queries—Brinkhoff (2002) www.fh-wilhelmshaven.de/ ANIS.ps.gZ oow/institute/iapg/personen/brinkhoff/paper/SDH2002.pdf 0150. An Overview Of The Cave Project Research. 0.168. On the Generation of Time-Evolving Regional Bimbot, Hutter. (1998) www.ubilab.org/publications/print Data Tzouramanis. (2002) delab.csd.auth.gr/papers/ versions/pdf/bim98.pdf GEO02tvm.pdf 0151 Cross Language Retrieval with the Twenty-One 0169. A dynamic service delivery framework based on system Wessel Kraaij (1998) trec.nist.gov/pubs/trec6/pa the OSGi model Vos, Buytaert, Buytaert (2002) www.ss pers/twentyone.ps grrit/en/ssgrr2002w?papers/165.pdf US 2006/0167784 A1 Jul. 27, 2006
0170 Applying bacterial algorithm to optimise trapezoi 0188 Towards Adaptive, Resilient and Self-Organiz dal. Botzheim, Hamori, Kóczy (2001) www.mft.hu/hallg/ ing. Montresor, Meling. www.elet-polimi.it/Users/DEI/ 200106.pdf Sections/CompEng/GianPietro. Picco?ntwo2-p2p/papers/ 18.pdf 0171 Deadlock Probability in Unrestricted Wormhole Routing Networks—Folkestad, Roche (1997) ftp.cs.u- 0189 Stability of interpolative fuzzy KH controllers— cla.edu/tech-report/1997-reports/970008.ps.Z. Tikk, Joo. (2002) www.mft.hu/publications/tikk/tikk4.pdf 0172 The Infostations Challenge: Balancing Cost and— 0.190 Comprehensive analysis of a new fuzzy rule inter Ubiquity In Delivering www.winlab.rutgers.edu/~ryates/pa polation method Tikk, Baranyi (2000) www.mft.hu/publi persfieeepc6b.ps cations/tikk/tikk2.pdf 0173 A Proposal for a Combination of Compression and 0191) Design and Use of Clinical Ontologies:—Curricu Encryption Lutz Vorwerk Thomas www.informatik.uni lar Goals For www-smi. Stanford.edu/pubs/SMI Reports/ trier.de/-Vorwerk/publics/perth.ps SMI-1999-0767.pdf 0.174 Mobility Management In Plug And Play Net 0.192 Toward Standardization of Electronic Guideline.— work Architecture Mazen Malek (2002) www.item.nt Elkin, Peleg, Lacson, ... (2000) www-smi. Stanford.edu/pubs/ nu.no/-plugandplay/publications/Smartnet2002.pdf SMI Reports/SMI-2001-0865.pdf 0175 Adaptive Importance Sampling Simulation Of 0193 Extracting Information for Automatic Indexing of Queueing Networks—de Boer www.informs-cs.org/ Multimedia Material—Saggion. (2002) parlevink.cs.utwen wsc00papers/086.PDF te.nl/projects/mumis/documents/mumis-lrec2002.pdf 0176 Ics-Forth Erich Leisch Stelios (1999) www.ics 0.194 Investigation of a new alpha-cut based fuZZy.— ..forth.gr/ICS/acti/cmi hta/publications/papers/1999/hector Baranyi, Tikk, Yam. www.mft.hu/publications/tikk/ Solutions/hector Solutions.pdf tikk11.pdf 0177 Co-Evolution of Bargaining Strategies in a Decen 0.195 The Evolution of Jini" Technology. In Telematics tralized.—Eymann (2001) www.iig.uni-freiburg.de/ By www.s. sun.com/software/iini/whitepapers/PsiNapticT ~eymann/publications/TEymann01.pdf elematics.pdf 0178 VBR Video Source Characterization And A Prac 0196. Desktop Synchronous Distance Learning Applica tical. Cselényi, Molnár hsnlab.ttt.bme.hu/-molnar/files/ tion—Enhanced With Efficient ru6.cti.gr/Publications/ VBRtelsys.ps.gZ 645.pdf, ru6.cti.gr/Publications/904.pdf 0179 An approach to dynamic reconfiguration of— 0.197 Adaptive compression of DICOM-image data— Almeida, Wegdam. (2001) www.cs.utwente.nl/~alme/cvi tae/sbrc2001 -final-footnote.pdf www.ub.utwente.nl/web Hludov, Engel, Meinel www.informatik.uni-trier.de/ docs/ctit/1/0000004f.pdf ~meinel/papers/tes 2.ps 0180 A GSM/GPS Receiver With a Bandpass Sigma 0198 Telematic Tools to Support Group Projects in Delta Analog. Müller, Boehm, Hentschel (1999) Higher Education Jan Van Der www.ub.utwente.nl/web www.ifnet.tu-dresden.de/MNS/veroeffentlichungen/1999/ docs/ctit/1/00000004.pdf Mueller T. EUW 99.pdf 0199 A Receiver-initiated WDM Multicast Tree Con 0181 An Intelligent Educational Metadata Repository— struction Protocol. Ip Dense Mode www.ub.utwente.nl/ Bassiliades, Kokkoras. (2002) www.cscl.auth.gr/%7Elpis/ webdocs/ctit/1/00000032.pdf publications/crc-chapter1.pdf 0200 AUTOLINK-A Tool for the Automatic and— 0182 Messor: Load-Balancing through a Swarm of Schmitt, Koch. (1997) 141.83.21.121 /publications/ Autonomous Agents—Montresor, Meling, Babaoglu (2002) ALASA/SchmKochGraHog.ps.gZ www.cs.unibo.it/babaoglu/papers/ap2p02.pdf 0201 The Term Processor Generator Kimwitu van 0183 Exact trade-off between approximation accuracy Eijk, Belinfante, Eertink. (1997) wwwtios.cs.utwente.nl/ and. Tikk, Baranyi www.mft.hu/publications/baranyi/ kimwitu?tacas 97.ps.gZ baranyi2.pdf 0202 Security Engineering of Lattice-Based Policies— 0184 On the Queue Tail Asymptotics for General Mul Bryce (1997) set.gmd.de/-kuehnhsr/CWASAR/ tifractal Traffic Molnar, Dang, Maricza hsnlab.ttt.bme.hu/ D06.1.anniS.ps.gZ ~molnar/files/ifipnw.02.pdf 0203) A scheme for adaptive biasing in importance sam 0185. Feature Ranking Based On Interclass Separability pling Heegaard (1997) www.idt.unit.no/-poulh/publica For Fuzzy. Tikk, Gedeon (2000) www.mft.hu/publica tions/IS-adaptive.ps www.item.ntinu.no/-poulh/publica tions/tikk/tikk8.pdf tions/adapt-scheme-abstract.ps 0186 Public Key Certificate Revocation Schemes—Ar 0204 Towards the Industrial Use of Validation.—Ek, nes (2000) www.pVV.ntinu.no/-andrearn/certrev/thesis/Cer Grabowski. www.itm.mu-luebeck.de/publications/A-97-03/ tRevThesis 29Feb2000.ps.gZ SDL-Forum-97.ps.gZ 0187 Competitive market-based allocation of consumer 0205) A Browser for a Versioned Entity-Relationship attention.—Bohte, Gerding. (2001) www.cwi.nl/ftp/CWIre Database Gulla (1992) www.idt.unit.no/-epos/Papers/ ports/SEN/SEN-R0131ps.Z. browser.ps US 2006/0167784 A1 Jul. 27, 2006
0206. The Eternal Resource Locator: An Alternative 0224) A Hyperlink-Proposal Mechanism to Exemplify Means of Vaclav www.usenix.org/publications/library/ Cognitive. Haffner, Roth, Meinel www.informatik.uni proceedingSfec98/full paperSfanderson/anderson.pdf trier.de/-meinel/papers/PaperO2.ps 0207. A Continuously Available and Highly Scalable.— 0225. Digital Signatures For Automobiles? —Gollan, Hvasshovd. (1991) www.idintnu.no/IDT/grupper/DB-grp/ Meinel www.informatik.uni-trier.de/-meinel/papers/ tech paperS/hpts91.ps DigitalSignatures.Auto02.pdf 0208 A Parallel Implementation of XTP on Transput 0226 Designing Safe Smart Home Systems for Vulner ers—Braun, Zitterbart (1991) www.iam.unibe.ch/-braun/lit/ able People—Dewsbury, Taylor www.smartthinkin lcn 16 Xtp.ps.gZ g.ukideas.com/ DIRC.pdf 0209 Stublets: A Notion for Mobility-Aware Application 0227 Providing X.509-based user Access Control To Adaption Dietmar Kottmann Christian (1996) ftp.diku.dk/ ftp.polito.it/pub/security/papers/sec98/sec98.ps.gZ diku/distlab/wmr96/sommerps.gZ 0228 Seeing Speech. In Space And Time: Psychological 0210 ATM Traffic Measurements and Analysis on a Real And. Ruth Campbell Department www.asel.udel.edu/ic Testbed Molnár, Cselényi. (1996) hsnlab.ttt.bme.hu/ slip/cdrom?vol3/1008/a1008.pdf ~molnar/files/itc2.ps.gZ 0229. The Process of Designing Appropriate Smart 0211) A CORBA Platform for Component Groupware Homes: Including.—Guy Dewsbury Bruce www.smart Hofte, van der Lugt, Bakker (1996) www.telin.nl/publi thinking.ukideas.com/Dewsbury et al Appropriate desi caties/1996/ozchi96.pdf gn of Smart homes.pdf 0212. The European Web Index: An Internet Search Ser 0230 Transparent Dynamic Reconfiguration for vice for.—Lundberg, Ardó. (1996) www.lub.lu.se/desire/ CORBA Almeida, Wegdam, van. (2001) www.cs.utwen radar/reports/D3.12/D3.12.v1.0.ps te.nl/~alme/cvitae/doa01.pdf 0213 Results Of The CEO Project WWW Manage 0231. Dissemination of Mutable Sets of Web Objects ment Hazewinkel, Van Hengstum, Pras (1996) www.sn Buchholz, Goebel, Schill, Ziegert (2001) wwwrn.inf.tu mp.cs.utwente.nl/nm/research/results/publications/pras/ dresden.de/-buchholz/askom/PDCS2001.pdf WWW.pdf 0232. How to Support the Negotiation of Service Level 0214) Towards Integrated QoS Management Schmidt, Agreements.—Köppel, Böning, Abeck (1999) www.coop Zitterbart (1995) www.telematik.informatik.uni-karlsru eration-management.de/publikationen/paper?isas'99 Abeck he.def-Schmidt/schmidt hipp94.ps.gZ Boening-Koeppel.pdf 0215. On engineering support for business process mod 0233 PAMINA: A Certificate Based Privilege Manage elling.—Franken, de Weger. (1996) www.home.cs.utwen ment System Nochta, Ebinger, Abeck (2002) www.iso te.nl/~pires/publications/bpr96.pdf c.org/isoc/conferences/indsS/02/proceedingS/papers/ 0216) Optimization of Spatial Joins Using Filters Veen nochta.pdf hof, Apers, Houtsma (1995) wwwis.cs.utwente.nl:8080/is 0234) Modeling IT Operations to Derive Provider doc/confpaper/veenhof.BNCOD95.accepted ps.gZ Accepted Management.—Abeck, Mayerl (1999) www.co 0217 Software Architecture of Ubiquitous Scientific operation-management.de/publikationen/paper/ Computing. Tzvetan Drashansky (1995) www.cs.pur im99 abeck-mayerl.pdf due.edu/homes/saw/publications/95/ubiq-pses.ps 0235. On deriving rules for nativised pronunciation in.— 0218 Wavefront implementation of Self Organizing Trancoso, Viana. (1999) www.12finesc.pt/documents/pa Maps on RENNS Gaute Myklebust (1995) www.idt.u- pers/Trancoso99b.pdf nit.no/~gautemyk/icasp95.ps 0236) A CANDLE to light the way? Sebastian Abeck 0219. On the Enterprise Modelling of an Educational Jodok www.cooperation-management.de/publikationen/pa Information. www.ub.utwente.nl/webdocs/ctit/17 per/ACandletoLightTheWay.pdf 0000002fpdf 0237 Internet Agents for Telemedicine Services—Mea, 0220 VBR Video Source Characterization And A Prac Roberto, Conti, Di. (1999) www.telemed.uniud.it/papers/ tical. Cselényi, Molnár hsnlab.ttt.bme.hu/-molnar/files/ VDM-MI99.pdf VBRtelsys.ps.gZ 0238 Supporting Secure and Transparent Delegation in 0221) Use of the CANTOR system for collaborative the CORBA.—Zoltn Nochta Taufiq (2001) www.coopera learning in. Hans Andersen Verner newmedia.colorad tion-management.de/publikationen/paper/ o.edu/cscl/270.pdf pwc2001 nrr cr.pdf 0222 Jgroup/ARM: A Distributed Object Group Plat 0239). The adaptation and use of a WWW-based course form. Meling, Montresor... www.CS.UniBO.it/babaoglu/ management system. De Boer, Collis (2000) www.ub.ut papers/group-arm.pdf wente.nl/webdocs/ctit/1/00000059.pdf 0223) Validation of the Open Service Access API for 0240 Evaluation of Mobile Agent Systems with Respect UMTS. Maarten Wegdam Dirk-Jaap arch.cs.utwente.nl/ to Failure.—Otto Wittner October www.item.ntinu.nof-ot publications/papers/proms01-osa-lincs22 130210.pdf tow?papers/failsemMASreport.pdf US 2006/0167784 A1 Jul. 27, 2006 20
0241 Proactive Services in a Distributed Traffic Telemat 0259. The HyperMuseum Theme Generator System: ics.—Gura, Held, Kaiser www.informatik.uni-ulm.defrs/ Ontology based. Stuer, Meersman, De. (2001) www.archi projekte/core/ProctiveServ.pdf muse.com/mw2001/papers/stuer/stuer.html http://wise.Vu b.ac.be/Download/Papers/stuerMW2001.pdf 0242 TCP over GPRS Performance Analysis Manner (1999) www.cs.helsinki.fi/u/manner/papers/Thesis-Man 0260 Failure Semantics of Mobile Agent Systems ner-ps Involved in. Otto Wittner Carsten (1999) www.item.nt nu.no/-ottow/papers/failsemNIK99.pdf 0243 Ip Over Wavelength-Routed Access Networks— Marcos Rogrio Salvador wwwctit.cs.utwente.nl/~salvador/ 0261 Cross-Entropy Guided Mobile Agents Finding Eunice99.pdf Cyclic Paths in. Wittner, Helvik (2002) www.item.nt nu.no/-wittner/aamas2002 submitted.pdf 0244 Increasing Retrievability and Reusability of Learn ing.—Hiddink Van Der www.ub.utwente.nl/webdocs/ctit/1/ 0262 Cross Entropy Guided Ant-like Agents Finding 00000056.pdf Dependable. Wittner, Helvik (2002) www.item.ntinu.no/ -wittner/cec2002.pdf 0245 Experiences from Development of Home Health Care Applications.—Leili Lind Erik ftp.limt.liu.se/pub/bildb/ 0263. Using Information Flow Control to Evaluate MIpapers/524 LIND.PDF Access. Mantel, Schairer... www.dfki.de/~schairer/publi cations/report00159.ps.gZ 0246 A Review of Parallel Implementations of Back propagation. Torresen, Tomita www.ifi.uio.no/-jimtoer/ 0264 Network Architecture of a Packet-switched WDM chp2.ps LAN/MAN Dey Koonen And (2000) wwwctit.cs.utwen te.nl/-salvador/LEOS2000.pdf 0247 ESCORT: Towards Integration in Intersection Con trol—Andrea Savigni Filippo (2000) www.cs.ucl.ac.uk/ 0265 Specification and Validation of a Real-Time Paral staff/A.Savigni/papers/2000 jubilee escort roma.pdf lel.—de Farias, Pires. (1997) www.ub.utwente.nl/webdocs/ ctit/1/00000066.pdf 0248 Usability Field-Test Of A Spoken Data-Entry Sys tem—Marcello Federico And poseidon.itc.it:7117/~ssi/ 0266 Mobile Ip: Security Application Tuquerres, Sal DITELO/papers/ICASSP99 1ps vador, Sprenkels intrg.cs.tcd.ie/htewari/papers/MobileIP Sec.pdf 0249 FLAMINGO: A Packet-switched IP over WDM Metro Optical Network Dey Koonen Geuzebroek (2001) 0267 M3POC: a multimedia multicast transport protocol wwwctit.cs.utwente.nl/~salvador/NOC2001.pdf for cooperative. Owezarski www.laas.fr/-owe/PUBLIS/ 99525.ps.gZ 0250 Real-time test specification with TTCN-3 Dai, Grabowski, Neukirchen (2001) www.itm.mu-luebeck.de/ 0268 User Interfaces for All Kobsa, (eds.) (1999) publications/FBT2001/Abstract fbtO1 neukdai.pdf www.gmd.de/publications/report/0074/Text.pdf 0269 Supporting PIM-SM in All-Optical Lambda 0251 A System for Uniform and Multilingual Access to Switched Networks—Marcos Rogrio Salvador (2001) Structured. Xu, Netter, Stenzhorn (2000) www.cs.ust.hk/ wwwctit.cs.utwente.nl/~salvador/SBRC2001.pdf acl2000/Demo/04 Xu.pdf 0270. Managing Distributed Personal Firewalls with 0252 Corpus-driven learning of Event Recognition Smart.—Haffner, Roth, Heuer, www.informatik.uni-trier.de/ Rules—Roberto Basili Maria www.dcs.shef ac.uk/-fabiof ~meinel/papers/Managing01.ps ML4IE/2.PS.gz 0271 FINAL REPORT: LAURIN http://laurinuibka 0253) Web-Support for Activating Use of Theory in c.at/-Version November Author germanistik.uibk.ac.at/lau Group-Based Learning Jan Van Der www.ub.utwente.nl/ rin/reports/finalrep01.pdf webdocs/ctit/1/0000005a.pdf 0272 Some Implications of MSC, SDL and TTCN Time 0254 Automated Generation of Category-Specific The Extensions.—Hogrefe, Koch. (2001) www.itm.mu-lue sauri for.—Attardi, Di Marco. (1998) faureiei.pi.cnrit/-fab beck.de/publications/DH BK HN 2001 SDLForum/ rizio/Publications/TRO698.ps sdlforum2001.pdf 0255 The use of CMC in applied social science train 0273 Supporting IP Dense Mode Multicast Routing Pro ing Interim Report Merja www.stir.ac.uk/schema/deliver tocols in. Marcos Rogerio Salvador (2000) wwwctit.cs.ut ables/D5.3.pdf wente.nl/~salvador/OPTICOMM2000.pdf 0256 The IT-Potential Of Haptics Touch access for 0274 Report on the course for Technology Teachers people with. Sjöström (1999) www.certec.lth.se/doc/ WWW Course of. Jyrki Pulkkinen And telematics.ex touchaccess/TouchAccess.pdf .ac.uk/T3/0/downloads/d13-1.pdf 0257 An All-Optical WDM Packet-Switched Network 0275 A Multi-DSP Laboratory Course Rinner, Architecture. Marcos Rogrio Salvador (2001) wwwctit.c- Schneider, Steger, Weiss www.iti.tu-graz.ac.at/de?people/ S.utwente.nl/~salvador/ICN2001.pdf schneider/papers/rinner98.pdf 0258 An Adaptive, Collaborative Environment to 0276 Modularity A Concept For New Neural Network Develop. Vizcano. oreto.inf-cr.uclm.es/personas/ Architectures—Schmidt, Bandar (1998) www.comp.lanc avizcaino/itsenV.ps s.ac.uk/-albrecht/pubs/pdf/schmidt csa irbid 1998.pdf US 2006/0167784 A1 Jul. 27, 2006
0277 Supporting IP Dense Mode Multicast Routing in 0296 Use Of Real And Contaminated Speech For Train All-Optical. Marcos Rogrio Salvador (2001) wwwctit.c- ing Of A. Matassoni Omologo And (2001) posei s.utwente.nl/-salvador/ONM2001.pdf don.itc.it:7117/~ssi/SHINE/ps/eurospeech01.ps.gZ 0278 Dagstuhl Seminar on Ubiquitous Computing 0297 Extending the Data Storage Capabilities of a Java September The International (2001) www.inf.ethz.ch/VS/ based.—Clemens Cap Nico wwwiuk.informatik.uni-ros events/dag2001/intro/DagstuhlIntroductions.pdf tock.de/-maibaum/docs/maibtune.ps 0279) A Framework For Video Modelling Centre For 0298 Towards Precision Tools For ATM Network Telematics www.cs.utwente.nl/~milan/docs/innsb.ps Design, Dimensioning.—Molnár, al. hsnlab.ttt.bme.hu/ ~molnar/files/peripol.ps.gZ 0280 Conceptual Stage in Designing Multimedia for Tele Learning Kommers (2001) www.ub.utwente.nl/web 0299 Fair Bandwidth Allocation of a Wireless Base docs/ctit/1/00000060.pdf Station Gyorgy Miklos Traffic hsnlab.ttt.bme.hu/~mol nar/files/iqwim 99.ps.gZ 0281 Signed Preservation Of Online References— Heuer, Losemann, Meinel www.informatik.uni-trier.de/ 0300 Forecasting the Success of Telecommunication -meinel/papers/webnet00b.ps Services in. Detlef Schoder. (2000) www.iig.uni freiburg.de/telematik/forschung/publikationen/pubfiles/ 0282 Chapter 7 Implementation of Backpropagation Sc2000.pdf Neural. Jim Torresen Department www.ifi.uio.no/-jim toer/chp3.ps 0301 A General Fractal Model of Internet Traffic Molnar hsnlab.ttt.bme.hu/-molnar/files/multifractalLCN 0283 TIMe at a glance Braek, Gorman, Haugen, .pdf Melby. www.sintefno/time/report.pdf 0302) Correlations in ATM Cell Streams Exposed to Cell 0284 Executive Summary The Vision Of www.it.b- Delay Variation Molnár, Blaabjerg hsnlab.ttt.bme.hu/ ton.ac.uk/research/seake/knowledge.pdf ~molnar/files/hung1.ps.gZ 0285 Supporting the Travelling Tradition: A report on 0303) A General Traffic Control Framework in ATM the work. Ken Marks Department (2000) Networks Fodor, Marosits, Molnár (1996) hsnlab.ttt.b- ui4all.ics.forth.gr/i3SD2000/Marks.PDF me.hu/~molnar/files/gtfps.gZ 0286 Circuits and Systems—Benini, De Micheli, Macii, 0304) NAVIGATION IN CYBERSPACE. Using Multi Maloberti www.nd.edu/~stjoseph/newscas/ Dimensional Scaling. Schoder Institut Fur (1999) www.ii CASMagvoll no1.pdf g.uni-freiburg.de/telematik/forschung/publikationen/pub files/Sc1999a.pdf 0287 Enterprise Modelling For An Educational Informa tion. Ing Widya Cees (2001) www.home.cs.utwente.nl/ 0305 Performance Measurement Tool for Packet For -widya/webpapers/iceis2001. 205.pdf warding Devices—Tam AS Kovacshazy www.mit.bme.hu/ -khazy/publications/imtc2001 3472.pdf 0288 Telematics For Group-Based Learning: Simplicity 0306 Benefits of a Universal Security Framework Versus.—van der Veen, Collis. www.ub.utwente.nl/web Report By Arnd (2000) www.iig.uni-freiburg.de/telematik/ docs/ctit/1/00000057.pdf forschung/publikationen/pubfiles/We2000fpdf 0289 Research Report 1997–1999. Department Of 0307 Methods for Computing B-ISDN Link Blocking Computer (1997) www.cs.ucy.ac.cy/Research/archives/ Probabilities—Molnár, Blaabjerg hsnlab.ttt.bme.hu/~mol rr97-99.ps nar/files/link.ps.gZ 0290 Translation Resources, Merging Strategies and 0308 Inter-organizational Networking of Small and.— Relevance.—Djoerd Hiemstra Wessel janus.cs.utwente.nl/ Framework And... (1999) www.iig.uni-freiburg.de/telematik/ ~hiemstra/papers/clefl.pdf forschung/publikationen/pubfiles/EgEn1999.pdf 0291 An Information System for Long-distance Coop 0309 Using Objects and Patterns to Implement Domain eration in Medicine. Kosch, al. (2000) www.ii.uib.no/ Ontologies—Guizzardi, Falbo, Filho (2001) www.home.c- para2000/program/jacek.ps S.utwente.nl/~guizzard/bcs.pdf 0292 Next Generation Internet in Europe Published By 0310 Highly Secure Low-cost Computers—Arnd Weber The www.infowin.org/ACTS/ANALYSYS/PRODUCTS/ Today’s (2000) www.iig.uni-freiburg.de/telematik/fors THEMATIC/NGI/ngi in europe.pdf chung/publikationen/pubfiles/We2000e.pdf 0293 R. Mu-noz, M. Saiz-Noeda, A. Su arez and M. 0311. On Modeling and Shaping Self-Similar ATM Traf Palomar Grupo De Investigaci (2000) gplsi.dlsi.ua.es/ fic—Andor Molnár And hsnlab.ttt.bme.hu/~molnar/files/ gplsi/articulos/a2000/mt2000.ps itc97.ps.gZ 0294 Lazy Users and Automatic Video Retrieval Tools in 0312 Pitfalls in Long Range Dependence Testing and (the) Lowlands—The Lowlands Team carol wins.uva.nl/ Estimation Molnar, Dang hsnlab.ttt.bme.hu/-molnar/files/ ~cg.msnoek/pub/trec10video.pdf pitfalls.pdfgZ 0295) Remote MIB item look-up service Pras, Boros, 0313 Distributed Fair Bandwidth Allocation of a Wire Helthuis (2002) www.simpleweb.org/nm/research/results/ less Base. Gyorgy Miklos Traffic (2000) hsnlab.ttt.b- publications/pras/2002-04-noms.pdf me.hu/-molnar/files/netwo0.ps.gZ US 2006/0167784 A1 Jul. 27, 2006 22
0314 Highly Secure Low-cost PDA-phones—Weber 0332 Trends: Training Educators Through Networks (2000) www.iig.uni-freiburg.de/telematik/forschung/pub And..—Christos Bouras Computer (1996) ru6.cti.gr/Publica likationen/pubfiles/We2000d.pdf tions/439.pdf 0315 Sharing Telematics Courses. The CANDLE 0333) On-Demand Hypermedia/Multimedia Service project Aiko Pras Centre (2001) www.simpleweb.org/nm/ over.—Bouras, Kapoulas. (1996) ru6.cti.gr/Publications/ research/results/publications/pras/2001-09-04-eunice.pdf 358.pdf 0316 A Prototype for an Agent-based Secure Elec 0334 Tele-working services from the Greek PTT tronic.—Padovan, Sackmann. (2001) www.iig.uni Christos Bouras Vaggelis (1999) ru6.cti.gr/Publications/ freiburg.de/telematik/forschung/projekte?e sicherheit/ 322.pdf comet/publikationen/PaSaEyPi2000.pdf 0335 HIPPOCRATES: A Tool for Distance Education 0317. On Measurements of Multimedia Traffic in ATM Bouras Fotakis Kapoulas ru6.cti.gr/Publications/305.pdf Networks Cselényi. hsnlab.ttt.bme.hu/~-molnar/files/ 0336 A Platform for the Implementation of the Ser icomt.ps.gZ vices.—Bouras, Gkamas. (1998) ru6.cti.gr/Publications/ 0318 Advanced Generation Tool of Application's Net 259.pdf work Traffic—Petroczi, Molnar hsnlab.ttt.bme.hu/~molnar/ 0337 Training Centres : An Architecture for the Reali files/agentant.ps.gZ sation. Christos Bouras Computer ru6.cti.gr/Publications/ 0319. On Burst And Correlation Structure of Teletraffic 292.pdf Models—Molnár, Miklós hsnlab.ttt.bme.hu/~molnar/files/ 0338. In-Service Training through ODL Environments: ilkley97.ps.gZ From User. Bouras Lampsas Spirakis ru6.cti.gr/Publica 0320 Scaling Analysis of IP Traffic Components—Mol tions/275.pdf nar, Dang (2000) hsnlab.m.bme.hu/-molnar/files/ 0339 Distributed Learning Environment using Advanced itcssip00.ps.gZ Services. Ch Bouras Computer (1999) ru6.cti.gr/Publica 0321) A System for Supporting Cross-Lingual Informa tions/453.pdf tion Retrieval Capstick, al. (1999) speech.ftw.at/~gor/pub/ 0340 Routing Management Application Based On ipm/mulinex-ipm99.pdf Mobile Agents On.—Anglica Reyes Ernesto www.tgs.cs.ut 0322 Spatiotemporal Segmentation and Tracking of wente.nl/eunice/summerSchool/papers/paper9-2.pdf Objects for. Kompatsiaris, Strintzis (2000) egnatia.ee.au 0341 Quality Of Service Monitoring In Ip Networks By th.gr/-ikom/CSVT2000.pdf Using. Tams Varga Andrs www.tgs.cs.utwente.nl/eunicef SummerSchool/papers/paper4-3.pdf 0323 Region-Based Color Image Indexing And Retrieval—Ioannis Kompatsiaris Evagelia (2001) egnati 0342 Issues on QoS based Routing in the Integrated a.ee.auth.gr/-ikom/icip2001.pdf Services. Gbor Rtvri Department wwwtgs.cs.utwente.nl/ eunice/summerschool/papers/paper4-1.pdf 0324 Kompatsiaris—And Michael Strintzis egnatia.ee .auth.gr/-ikom/ICIP00.pdf 0343 Usability Research in a Housing Fair: Problems and—Sajaniemi, Tossavainen (1995) csjoensuu.fi/pub/Re 0325 th IFLA Council and General Conference Aug. ports/A-1995-7.ps.gZ 16-25, 2001- Code Number Division www.ifla.org/IV/ iflag7/papers/161-165e.pdf 0344) Tele-Education/-Co-Operation Pilot Pilot Study Plan www.cg...its.tudelft.nl/-charles/publications/ 0326 Agent-Mediators. In Media-On-Demand Eletronic MESH Report D213.pdf Commerce—Joo Paulo Andrade www.home.cs.utwente.nl/ ~guiZZard/mod-amec-cuba.pdf 0345 CAVE Speaker verification in bank and telecom services—Lindberg, Blomberg, Melin ftp.ling.umu.se/pub/ 0327) A Web-based Distributed Environment to Support phonum/phonum4/65.ps Teleteaching:..—Ch Bouras Computer ru6.cti.gr/Publica tions/261.pdf 0346) The Catallaxy as a new Paradigm for the Design of Eymann, Padovan, Schoder (2000) www.iig.uni 0328 Web-Enabled Distance Education Environment— freiburg.de/telematik/forschung/publikationen/pubfiles/ Bouras, Lampsas, Bazaios. (1998) ru6.cti.gr/Publications/ EyPaSc2000.pdf 296.pdf 0347 Internet Accounting Pras, van Beijnum, Spren 0329 ID-No. of presentation: t3a01391—Authors Chris kels. (2001) www.simpleweb.org/nm/research/results/publi tos Bouras ru6.cti.gr/Publications/279.pdf cations/pras/internet-accounting.pdf 0330. Usability meets Security. The Identity-Manager 0348 EWI Search and User Interface Functions Ardó, as your. Jendricke, Markotten (2000) www.acsac.org/ Cao, Lundberg, Roslund. www.lub.lu.se/combine/docs/ 2000/papers/90.pdf D34 search uips 0331 Deployment Scenarios of DVEs in Education— 0349 Algebras and Automata for Timed and Stochastic Bouras Computer Technology ru6.cti.gr/Publications/ Systems—D'Argenio www.home.cs.utwente.nl/~dargeniof 324.pdf dissertation/dissertation.ps.gZ US 2006/0167784 A1 Jul. 27, 2006
0350 Characterizing Video Coding Computing in Con 0369 Teaching and learning with the WWW in the ference Systems—By Tuquerres Tuquerre www.ub.utwen undergraduate.—Oliver, Omari, Cowan elrond. Scam.ec te.nl/webdocs/ctit/1/0000004d.pdf u.edu.au/oliver/docs/96/AUSWEB2d.pdf 0351 A Model to Evaluate Certificate Revocation Forn 0370 Toward a Standard for Guideline Representation: Castro Department (2000) www-mat.upc.es/-forne/f- an.—Domenico Pisanelli Aldo Saussure.irmkant.rm.cnr.it/ SCI2000 1.pdf onto/publ/amia99/amia99.pdf 0352 Dublin Bus Tracking Service Design and imple 0371 Time Domain MLE of the Parameters of FBM mentation of a. Fallon (2000) ftp.cs.tcd.ie/pub/tech-re Traffic Vidács, Virtamo (1999) keskus.tct.hut.fi/tutkimus/ ports/reports.00/TCD-CS-2000-47.pdf com2/publ/fbm2.pdf 0353 Integrating Different Strategies for Cross-Lan 0372 Artificial Coordination Simulating Organiza guage.—Buitelaar, Netter, Xu www.dfki.de/lt/mietta/mietta tional. Eymann, Padovan, Schoder (1998) www.iig.uni twilt.ps freiburg.de/-padovan/publications/cefes98.pdf 0354) Integrating Trading and Load Balancing for Effi 0373 REMOT-A Project to Remotely Monitor and Con cient. Thien, Neukirchen (2000) www.itm.mu-lue trol Scientific—Experiments Pucillo Oat www.aps...anl.gov/ beck.de/publications/DT HN 2000 ITalBfhEMoSiDS/ icalepcs97/paper97/p235.pdf USM2000.pdf 0374 Performability Analysis of Markov Reward Mod 0355 Junction Point Aspect: A Solution to Simplify els with Rate and—Andor Acz, And (1999) webspin.hit.b- Implementation of Berger (2000) micado.project.free.fr/ me.hu/-telek/cikkek/racz99aps.gZ Publi/ecoop2000.ps.gZ 0375 Analysis of the Completion Time of Markov 0356 Csaba Antal Jzsef Molnir Sindor www.cs.kau.se/ Reward Models. Miklos Telek Andras webspin.hit.b- ~Soren/dkdoc/documentstore/cc fp.ps me.hu/-telek/cikkek/tele98b.ps.gZ 0357 GRAVE Cave General Video Client Wasskog 0376 A New Method for Spectral Shaping Coding (1995) www.idi.ntinu.no/grupper/db/report dipomas/il Peter Amos Aszlo webspin.hit.bme.hu/-telek/cikkek/ myggo/diplom.ps. Z Vamo98aps.gZ 0358 Z39.50 Application programmer's Interface SYS 0377 MRMSolve: A Tool for Transient Analysis of Large TEM. Document No Document ftp.ddb.de/pub/dbvosi/ Markov. Rácz, Tóth, Telek webspin.hit.bme.hu/-telek/ ses V3.ps.gZ cikkek/racz00fps.gZ 0359 Hardware Implementation of a Secure Bridge in 0378. Managing Services in Distributed Systems by Inte Ethernet. Forn Soriano Mels www-mat.upc.es/-jforne/ grating. Thien, Neukirchen (2000) www.itm.mu-lue jf GLOBECOM93.pdf beck.de/publications/ISCC2000/ISCC2000.pdfgZ 0360 Performance Evaluation of Strategies for Integra 0379 Conformance Testing with TTCN-Schiefer tion. Queija, van den Berg. (1999) www.cwi.nl/ftp/ decker, Grabowski (2000) www.itm.mu-luebeck.de/publica CWIreports/PNAVPNA-R9903.ps.Z. tions/I J 2000 CTwT/Telektronnikka 2000 CTandTTC N.pdf 0361 Inverse Multiplexing for ATM. Operation.—Agui lar-Igartua. (1999) marley.upc.es/pub/articles/icatm99.pdf 0380 Analysis and Modelling of Collaborative Learning Interactions—Workshop Notes (2000) collide.informa 0362 An integrated solution for secure communications tik.uni-duisburg.de/-martin/publication/Muehlenbrock over B-ISDN. Forn Mels Department www-mat.upc.es/ ECAI-2000.pdf -jforne/if CMS96.pdf 0381 Aligning IT and Organization in the MediaSite 0363 Web Representation with Dynamic Thumbnails— project—Iacucci, Axelsson (2000) iris23.htu.se/proceed Schmid www.comp.lancs.ac.uk/computing/users/SSchmid/ ings/PDF/20final.PDF Yuforic/YuforicExtAbstrips 0382 Enriching Textual Documents with Time-codes 0364 Distributed educational multimedia databases: from Video. van der Sluis, de Jong (2000) 133.23.229.11/ design, production. Hiddink (1998) www.home.ctitutwen -ysuzuki/Proceedingsall/RIAO2000/Wednesday/37CP1.ps te.nl/-hiddinkg/professional/papers/romy/romy.ps 0383 Modular Automated Transport Frequently Asked 0365 Sojourn Times in Non-Homogeneous QBD Pro Questions (FAQ)—Schweizer (2000) circwww.epfl.ch/staff/ cesses with Processor. Queija (1999) www.cwi.nl/ftp/ joerg/mat/doc/faq/faqps CWIreports/PNAVPNA-R9901.ps.Z. 0384 The CIMI Profile Release 1.0H Profile For Cul 0366. A guided tour through LGM How to generate tural www.cimi.org/public docs/HarmonizedProfile/ spoken.—Krahmer, Landsbergen. www.ipo.tue.nl/homep CIMIProfile10H.pdf ages/ekrahmer/Pubs/lgmps 0385) Real-Time Traffic Simulation of the German Auto 0367 Verifying a Smart Design of TCAP Arts, van bahn Network—Rickert, Wagner, Gawron (1996) www.z- Langevelde (1999) www.cwi.nl/ftp/CWIreports/SEN/SEN pruni-koeln.de/-mr/documents/PASA 96.ps.gZ R9910.ps.Z. 0386 Analysis of a Distributed Wireless Fair Scheduling 0368 Securing Multimedia Applications over B-ISDN Scheme—Miklós hsnlab.ttt.bme.hu/~molnar/files/ Jordi Forn Mels www-mat.upc.es/-ifornelif PROMS96.pdf itcSSmob00.ps.gZ US 2006/0167784 A1 Jul. 27, 2006 24
0387 CAC Algorithm Based on Advanced Round Robin 04.05 Educational Multimedia Databases: Past and Method for QoS.—Marosits, Molnár hsnlab.ttt.bme.hu/ Present—Gerrit Hiddink Centre www.home.ctitutwente.nl/ ~molnar/files/isccO1.pdf.gZ -hiddinkg/professional/papers/systems.ps 0388 Link Capacity Sharing Between Guaranteed- and 0406 Resource-limited information retrieval in Web Best Effort. Rácz, Telek webspin.hit.bme.hu/-telek/cikkek/ based. Daan Velthausz And (1997) www.trc.nl/publicaties/ racZ01a.ps.gZ 1997/reslim/resource-limited.pdf 0389 Quality of Service on the Internet: Evaluation of 0407 CRL supported a smart redesign of a real-life the. Elisabete Reis Elreis (2001) dragao.co.it.pt/ protocol Thomas Arts Email extranet.telin.nl/dscgi/dspy/ conftele2001 proc/pap101.pdf Get/File-8309/fmics99.ps.Z. 0390 A MixDemonstrator for teaching Security in the 0408 Convergence In The Digital Age Table Of Con Virtual. Jendricke, Rannenberg www.scis.cowan.edu.au/ tent ftp.cordis.lu/pub/libraries/docs/proceedings.pdf research/wise/WISE1 Proceedings/pdf/jendricke.pdf 04.09. Using Automated Assistance Systems—Putting 0391) Fair Allocation Of Elastic Traffic For A Wireless Base Station—Gyorgy Miklos Traffic (1999) hsnlab.ttt.b- The Driver Into Focus—Reichardt (1998) www.daimler me.hu/-molnar/files/globe99.ps.gZ benz.com/research/events/pdf/IV980240.PDF 0410 The Cave-Wp4 Generic Speaker Verification Sys 0392 Cell Delay Variation in an ATM Multiplexer tem. Jaboulet, KOOLWAAIJ. (1998) www.ubilab.org/pub Molnár, Blaabjerg hsnlab.ttt.bme.hu/-molnar/files/cdvis lications/print versions/pdf/jab98.pdf new.ps.gZ 0393 Supporting All Service Classes in ATM: A 0411 Field Test Of A Calling Card Service Based On Novel. Marosits. (1999) hsnlab.ttt.bme.hu/-molnar/files/ den Os, Boves. www.ubilab.org/publications/print ver info99.ps.gZ sions/pdf/den97.pdf 0394 The Impact Of Long Range Dependence On Cell 0412 Distributed Electronic Commerce Systems—Sonja Loss In An. Vidacs, Molnar, Gordos (1998) hsnlab.ttt.b- Zwill Er ftp.cs.umass.edu/pub/net/pub/hgschulz/i96/ me.hu/-molnar/files/globe98.ps.gZ Zwisslerps.gZ 0395. Investigation of Fractal Properties in Data Traf 0413 Needed Services For Network Performance Evalu fic—Dinh, Molnár. hsnlab.ttt.bme.hu/~molnar/files/ ation—Dung Dinh Luong gollum.ttt.bme.hu/~luong/tools/ jc98.ps.gZ atmip.ps 0396 Performance Evaluation of a General Traffic Con 0414 Link Proposals with Case-Based Reasoning Tech trol.—Marosits. hsnlab.ttt.bme.hu/~molnar/files/ niques—Haffner, Roth, Heuer. www.ti.fhg.de/conferences/ ipccc99.ps.gZ 200011171414080ps 0397) The Demand for Stored Value Payment Instru 0415 Remote Access to Medical Records via the Inter ments—Ingo Pippow Detlef (2001) www.iig.uni net: Feasibility. P. Lees Ce (1999) www.ics.forth.gr/ICS/ freiburg.de/telematik/forschung/projekte?e sicherheit/ acti/cmi hta/publications/papers/1999/cic99/lees cic99.pdf comet/publikationen/PiScho2001.pdf 0416 Partial Methods Versus End-to-End Measure 0398. On The Effects Of Non-Stationarity In Long-Range ments—Dung Dinh Luong gollum.ttt.bme.hu/~luong/tools/ Dependence Tests Trang Dinh And hsnlab.ttt.bme.hu/ ifip.ps ~molnar/files/trendeffps.gZ 0399. Bottlenecks on the Way Towards Fractal Charac 0417. The Role of Packet-dropping Mechanisms in QoS terization of—Andor Molnar Attila hsnlab.ttt.bme.hu/ Differentiation Goncalo Quadros Antonio (2000) www ~molnar/files/pmccn97.ps.gZ .dei.uc.pt/-boavida/papers/2000icon.pdf 0400 Multimedia Databases in Education Gerrit Hid 0418 Component-Based Groupware Tailorability using dink Centre www.home.ctitutwente.nl/-hiddinkg/profes Monitoring.—de Farias, Diakov (2000) amidst.ctitutwen sional/papers/dolls97.ps te.nl/publications/cscw cbg2000.pdf 04.01 Content-based video retrieval Petkovic (2000) 0419 Modeling of Time and Document Aging for www.edbt2000.uni-konstanz.de/pha-workshop/papers/Pet Request. Haffner, Roth, Engel. (2000) www.ti.fhg.de/con kovic.pdf ferences/2000092.11730520.ps 0402 Adaptive Optimisation of Importance Sampling 0420 Mpeg-4 Authoring Tool For The Composition Of for. Heegaard (1996) www.idi.ntnu.no/-poulh/publica 3D.—Daras. (2001) egnatia.ee.auth.gr/-ikom/ tions/nts 13.a.ps ISCAS2001.pdf 0403 Factors of reuse of Units of Learning Material— 0421 Disambiguation Strategies for Cross-language Gerrit Hiddink Centre www.home.ctitutwente.nl/-hiddinkg/ Information. Djoerd Hiemstra And (1999) www.cs.utwen professional/papers/reuse.ps te.nl/-hiemstra/papers/ecd199.ps 0404 Multilateral Security A concept and examples for 0422 the IFLA Council and General Conference Aug. balanced security—Rannenberg (2000) cSrc.nist.gov/nissc/ 16-25, 2001—Code Number Division www.ifla.org/IV/ 2000/proceedings/papers/202ra.pdf iflag7/papers/161-165e.pdf US 2006/0167784 A1 Jul. 27, 2006
0423 CANDLE: an European E-Education project to . . 0442 Towards Dynamic Composition of Hybrid Com ... —Batlogg, al. (2000) www.ssgrrit/en/ssgrr2000/papers/ munication Services—Floch, Braek (2000) www.item.nt 164.pdf nu.no/-jacf/paper/smarnett2000.pdf 0424 State-dependent M/G/1 Type Queueing Analysis 0443) Optimising the Operation of the World Wide Web for. Altman. (2000) www.cwi.nl/ftp/CWIreports/PNA/ in.—Hadjiefthymiades. www.cs.auc.dk/~tryfona/papers/ PNA-R0005.ps.Z. cacherelps 0425 Continuous Queries within an Architecture for 0444 Generating Test Cases for Infinite System Specifi Querying.—Brinkhoff, Weitkämper (2001) www.fh-wil cations—Stefan Heymer And www.itm.mu-luebeck.de/pub helmshaven.de/oow/institute/iapg/personen/brinkhoff/pa lications/GTCfISS/HeymerGrabowskips.gZ per/SSTD2001.pdf 0445 Computational Perspectives on Discourse and Dia 0426) The Distribution And Partitioning Scheme Of logue—Bonnie Lynn Webber www.dai.ed.ac.uk/daidb/ The. Jiménez. ches.ing.ula.ve/INVESTIGACION/AR people/homes/bonnie/handbook.ps.gZ TICULOS/TANIA/SIM-71 ps.gz 0427. An Architecture For Video On Demand Agent 0446. Virtual Universities—Ebner, Hogrefe (1999) Mediated.—Almeida, Guizzardi. www.home.cs.utwente.nl/ www.itm.mu-luebeck.de/publications/VFH/ ~guizzard/vod-amec-workcomp99.pdf ebner hogrefe waki'99.ps.gZ 0428 Reusing Multi-Media Components: A Catalogue 0447 Compensation methods to support generic graph Implementation—Steinmann, Shearer www.fernuni editing: A case.—Even, Spelt www.home.cs.utwente.nl/ hagen.de/DVT/Publikationen/Papers/emmsec.pdf -seven/ECOOPWS.pdf 0429 Multiagent Systems—Instructor Prof Dh 0448 Towards the Generation of Distributed Test Cases 136.159.122.221/seminar/enmf619 02/enmfo2.pdf Using Petri. Heymer, Grabowski www.itm.mu-lue beck.de/publications/FBT99/fbt99.ps.gZ 0430) Twenty-One at CLEF-2000: Translation resources, merging.—Djoerd Hiemstra Wessel www.ieipi.cnrit/DE 0449 Test Case Specification with Real-Time TTCN LOS/CLEF/twentyon.pdf Walter, Grabowski www.itm.mu-luebeck.de/publications/ 0431 Papabiles Torday, Bierlaire (2001) rosowww.ep TCSwRTTTCN/WalterGrabowskips.gZ fl.ch/mbi/strc-papabiles.pdf 0450 Scientific Approaches and Techniques for Negotia 0432. Using Objects and Patterns to Implement Domain tion.—Gerding, van Bragt. www.cwi.nl/projects/TA/re Ontologies—Guizzardi, Filho (2001) www.home.cs.utwen ports/negotiation.ps te.nl/-guizzard/SBES2001 vf.pdf 0451 Towards an Integrated Test Methodology for 0433 Decision Support Systems from a Health Informat Advanced.—Grabowski, Walter www.itm.mu-luebeck.de/ ics Perspective Nykänen (2000) acta.uta.fi/pdf7951-44 publications/tcS99.ps.gZ 4897-9.pdf 0452 TTCN-3—A new Test Specification Language for 0434) NetTrouble: A TTS for Network Management Black-Box Testing.—Grabowski (2000) www.itm.mu-lue Lus Santos Pedro www.dei.uc.pt/-psimoes/papersfits98.pdf beck.de/publications/ttcn3/Grabowski.pdfigZ 0435 Results and experience from the application of a 0453 Test Architectures for Distributed Systems State common.—Antonis Bouras. (1998) ru6.cti.gr/Publications/ of the Art and. Walter (1998) www.itm.mu-luebeck.de/ 3.18.pdf publications/IWTCS98TA/ 0436 The Informationand Communication Technologies IWTCS98Testarchitectures.ps.gZ In Education Christos Bouras Computer ru6.cti.gr/Publi 0454) Asbru: A Task-Specific, Intention-Based, and.— cations/289.pdf Miksch, Shahar, Johnson ftp.ifs.tuwien.ac.at/pub/publica 0437 Internet Protocols for Synchronous Distance tions/mik kem197.pdf Learning Ch Bouras Computer (2000) ru6.cti.gr/Publica 0455 Protocol Specifications Written By Jacob tions/431.pdf cmc.dsv.su.se/select/SEL-prot-spec-v 11-p-99.1009.pdf 0438. The Euro in the Electronic Purse Allard, Aly 0456) Towards The Third Edition Of TTCN ankian, Ankri, Collin. (2000) www.eurosmart.com/down Grabowski, Hogrefe (1999) www.itm.mu-luebeck.de/publi load/WhitePaper.pdf cations/iwtcs99.ps.gZ 0439 Monitoring Extensions for Component-Based Dis tributed.—Diakov, van Sinderen. (2000) amidst.ctitutwen 0457. A Framework for the Specification of Test Cases te.nl/publications/proms2000.pdf for.—Walter, Grabowski (1999) www.itm.mu-luebeck.de/ publications/Walter-Grabowski-JIST99/ist.ps.gZ 0440 The Q-bit Scheme Congestion Avoidance Using (1992) gatekeeper.dec.com/pub/doc/sigcomm/ccr/archive/ 0458 Verification of Compensation Requirements for the 1992/apr2/qbitps.Z. SEPIA. Even, Spelt (1998) www.home.cs.utwente.nl/ -seven/CTIT-TR-98-25.pdf 0441 Performance of a Parallel Transport Subsystem Implementation Torsten Braun Institute www.iam.u- 0459 Cote de Resyste Conformance Testing Reactive nibe.ch/-braun/lit/hpcs93.ps.gZ wwwtios.cs.utwente.nl/Docs/projects/cote-de-resyste/stw.ps US 2006/0167784 A1 Jul. 27, 2006 26
0460 Long Cycles and Long Paths in the Kronecker 0479 diryahoo.com/Science/Engineering/Electri Product of a. Jha, Agnihotri, al. (1995) www.csjhu.edu/ cal Engineering/Neural Networks/ ~rajesh/ps/ctaps 0480) www.aist.go.jp/NIBH/-b0616/Links.html 0461 Tutorial on Message Sequence Charts (MSC96)— Rudolph, Grabowski, Graubmann (1996) www.itm.mu-lue 0481 www.creative.net.au/~adrian/mirrors/neural/ beck.de/publications/MSC96/dis-tutorial-ps.gZ 0482 www.fi.uib.no/Fysisk/Teori/NEURO/neurons.html 0462) The Standardization of Core INAP CS-2 by 0483 aass.oru.se/~tdt/ann/faq/FAQ.html ETSI Grabowski, Hogrefe (1999) www.itm.mu-lue beck.de/publications/CS2-Standardization.ps.gZ 0484 www.cis.hut.fi/-jari/research.html 0485) www.eg3.com/WebID/elect/neur-net/blank/over 0463. On The Design Of The New Testing Language view/a-Z.htm TTCN-3 Grabowski, Wiles, Willcock. (2000) www.it m.mu-luebeck.de/publications/New TTCN3/Grabowski 0486) directory.google.com/Top/Computers/Artificial EtAll.pdf.gZ Intelligence/Neural Networks/ 0464) omVR—A Safety Training System for a Virtual 0487 directory.google.com/Top/Computers/Artificial Refinery—Haller, Kurka, Volkert. www.gup.uni-linZ.a- Intelligence/Neural Networks/FAQs, Help, and Tuto c.at:8001/staff/kurka/docs/ismcr99.pdf rials/ 0465 Senior Online Telematics De Report cmc.dsv 0488 dmoz.org/Computers/Artificial Intelligence/Neu ..su.se/sol/sol-transfer-spec.pdf ral Networks/ 0466 Formal Methods and Conformance Testing or— 0489 dmoz.org/Computers/Artificial Intelligence/Neu What are we.—Heymer, Grabowski www.itm.mu-lue ral Networks/FAQs, Help, and Tutorials/ beck.de/publications/FBT98/FBT98.ps.gZ 0490 www.cs.cqub.ac.uk/~J.Campbell/my web/book/ 0467 A Theorem Prover-Based Analysis Tool for Object nn.html Oriented Databases—Spelt, Even (1999) www.home.cs.ut 0491 www.cere-pa.cnr.it/IDAschool/lectures/neu wente.nl/-seven/CTIT-TR-98-22.pdf ral.html 0468 Chemistry in Action: Discovering the Behaviour of 0492 cognet.mit.edu/MITECS/Entryfjordan2 a Network. Heymer, Grabowski (1998) www.itm.mu-lue beck.de/publications/A-98-18/Report-A-98-18.ps.gZ 0493 www.faqs.org/faqs/ai-faq/neural-nets/part1/pre amble.html 0469 ERP in the e-commerce era—Luttighuis, Biemans extranet.telin.nl/dscgi/dspy/Get/File-2092/baanUSP.pdf 0494 Zhanshou.hypermart.net/thesis.htm 0470 Business-Driven Design of Transaction Services— 0495) www.links999.net/hardware/neural.html Biemans, Janssen, Luttighuis... (1999) extranet.telin.nl/ 0496 www-ra.informatik.uni-tuebingen.de/links/neu dscgi/dspy/Get/File-664/ICE.pdf ronal/welcome e.html 0471 Modelling organisations—Wetering (1999) 0497 www.cogneuro.ox.ac.uk/links/ann.html extranet.telin.nl/dscgi/dspy/Get/File-2902/ modellingV2.pdf 0498 faculty.cs.tamu.edu/choe/resources/ 0499 www.galaxy.com/galaxy/Engineering-and-Tech 0472. On Wrapping Query Languages and Ecient XML nology/Electrical-Engineering/Neural-Networks/ Integration Vassilis Christophides Sophie (2000) www.oa sis-open.org/cover/vassilisOuery Wrap.pdf 0500 mu.dmt.ibaraki.ac.jp/yanai/neu/facq/ 0473 MESH Release 2 implementation at CTIT Dia 0501) bubl.ac.uk/link/n/neuralnetworks.htm kov, Van Sinderen, Koprinkov amidst.ctitutwente.nl/publi 0502 www.webopedia.com/TERM/n/neural net cations/ctit tr99-08.pdf work.html 0474 EUROgatherer: a Personalised Gathering and 0503 www.ie.ncsu.edu/fangroup/neural.dir/indexneu Delivery. Amato, Straccia, Thanos (2000) faure.iei.pi.cn rit/%7Estraccia/download/papers/SCI2000/SCI2000.pdf ral.html 0504 www.geneticprogramming.com/AI/nn.html 0475 Frameworks for protocol implementation Bar bosa, Pires, van Sinderen (1998) www.home.cs.utwente.nl/ 0505 www.cs.utk.edu/-yarkhan/neural networks.html ~sinderen/publications/sbrc98.pdf 0506 www.physiol.ox.ac.uk/~ket/nn.html NEURAL NETWORKS REFERENCES 0507 www.aaai.org/AITopics/html/neural.html 0476 www.inference.phy.cam.ac.uk/mackay/Bayes 0508 www.inference.phy.cam.ac.uk/mackay/itprinn/ FAQ.html book.html 0477 www-2.cs.cmu.edu/Groups/AI/html/faqs/ai/neu 0509) www.hh..se/staff/nicholas/NN Links.html ral/faq.html 0510 xpidea.com/products/neurovcl/neuroabout.htm 0478 www.cs.stir.ac.uk/~lss/NNIntro/InvSlides.html 0511. www.msm.ele.tue.nl/research/neural/ US 2006/0167784 A1 Jul. 27, 2006 27
0512) homepages.goldsmiths.ac.uk/nikolaev/Nnets.htm ference on Machine Learning, pp. 21-29. San Francisco: Morgan Kaufmann. Also NeuroCOLT Technical Report 0513) www.triumf.ca/project ETA/neural network.html NC-TR-96-034 (February 1996). personal.bgsu.edu/~Suny/nn.html 0514) 0536) Bartlett, P. L. & Williamson, R. C. (1996) The VC 0515) www.icmc.Sc.usp.br/~andre/ann links.html dimension and pseudodimension of two-layer neural net www. Stud.ntinu.nof-hirpa/links/AI links.htm works with discrete inputs. Neural Computation 8(3), 625 0516) 628. 0517 it.umary.edu/Library/research/www subjects/neu 0537 Benveniste, A., Metivier, M. & Priouret, P. (1990) ral networks.html Adaptive Algorithms and Stochastic Approximations. Ber 0518) cindy.cis.nctu.edu.tw/NN/NN5/www.html lin: Springer. 0519) www.public.iastate.edu/~acl/links/links.html 0538 Bratko, I. & Muggleton, S. (1995) Applications of 0520 www.cs.cf.ac.uk/User/O.F.Rana/neural.html inductive logic programming. Communications of the Asso ciation for Computing Machinery 38(11), 65-70. 0521) www.cs.unredu/-bebis/CS791S/ 0539 Breiman, L. (1994) Bagging predictors. Technical 0522) www.geocities.com/fastiland/NNwww.html Report 421, Department of Statistics, University of Califor 0523 cns-web.bu.edu/pub/snorrason bookmarks/neu nia at Berkeley. ral.html 0540 Breiman, L. (1996a) Bagging predictors. Machine 0524) www.open.brain.riken.go jp/-cao/index Learning. work.html 0541 Breiman, L. (1996b) The heuristics of instability in 0525 SVr-www.eng.cam.ac.uk/research/neural/other model selection. Annals of Statistics. neural net sites.html 0542 Breiman, L. (1996c) Bias, variance and arcing 0526. R. O. Duda and P. E. Hart. Pattern Classication and classifiers. Technical report, Department of Statistics, Uni Scene Analysis. Wiley, N.Y., 1973. versity of California at Berkeley. 0527 S. Ripley, “Pattern Recognition and Neural Net 0543 Brodley, C. E. & Utgoff, P. E. (1995) Multivariate works”, Statistics, 33, 1065-1076. Ripley, B. D., 1996: decision trees. Machine Learning 19, 45-77. Pattern Recognition and Neural Networks. Cambridge: Uni versity Press., 1996, citeseerist.psu.edu/ 0544 Buntine, W. & Niblett, T. (1992) A further com ripley96complements.html, and cited references: parison of splitting rules for decision-tree induction. Machine Learning 8, 75-86. 0528 Aha, D. W., Kibler, D. & Albert, M. K. (1991) Instance-based learning algorithms. Machine Learning 6(1), 0545 Burnell, L. & Horvitz, E. (1995) Structure and 37-66. chance: Melding logic and probability for software debug 0529) Ali, K. M. & Pazzani, M. J. (1995) On the link ging. Communications of the ACM 38(3), 31-41, 57. between error correlation and error reduction in decision tree 0546 Catlett, J. (1991) On changing continuous ensembles. Technical Report 95-38, Department of Infor attributes into ordered discrete attributes. In Proceedings of mation and Computer Science, University of California at the European Working Session on Learning EWSL-91, ed. Irvine. Y. Kodratoff, pp. 164-178. Berlin: Springer. 0530 Almond, R. G. (1995) Graphical Belief Modeling. 0547 Cesa-Bianchi, M., Freund, Y., Helmbold, D. P. London: Chapman & Hall. ISBN 0-412-06661-0. Despite Haussler, D., Schapire, R. E. & Warmuth, M. K. (1993) How the date, this book was actually published in May 1996). to use expert advice. In Proceedings of the Twenth-Fifth 0531 Angluin, D. & Valiant, L. G. (1979) Fast probabi ACM Symposium on the Theory of Computing (San Diego, listic algorithms for Hamiltonian circuits and matchings. Calif., 1993), pp.382-391. New York: ACM Press. Journal of Computer and System Sciences 18, 155-193. 0548 Cesa-Bianchi, M., Freund, Y., Helmbold, D. P. 0532 Anthony, M. & Shawe-Taylor, J. (1993) A result of Haussler, D., Schapire, R. E. & Warmuth, M. K. (1996) How Vapnik with applications. Discrete Applied Mathematics 47. to use expert advice. Journal of the ACM. 207-217. Erratum (1994) 52,211 (the proof of theorem 2.1 0549) Cortes, C. & Vapnik, V. (1995) Support-vector is corrected). networks. Machine Learning 20, 273–297. 0533 Arbib, M. A. (ed.) (1995) The Handbook of Brain Theory and Neural Networks. Cambridge, Mass.: MIT 0550 Cost, S. & Salzberg, S. (1993) A weighted nearest Press. ISBN 0-262-01148-4. neighbor algorithm for learning with symbolic features. Machine Learning 10, 57-78. 0534 Atkeson, C. G. (1991) Using locally weighted regression for robot learning. In Proceedings of the IEEE 0551 Cover, T. M. (1968) Capacity problems for linear Conference on Robotics and Automation (Sacremento, machines. In Pattern Recognition, ed. L. Kanal, pp.283-289. Calif., 1991), pp. 958-963. IEEE Press. Thompson. 0535 Auer, P., Holte, R. C. & Maass, W. (1995) Theory 0552) Craven, M. W. & Shavlik, J. W. (1996) Extracting and application of agnostic PAC-learning with Small deci tree-structured representations of trained networks. In sion trees. In Proceedings of the Twelfth International Con Touretzky et al. (1996), pp. 24-30. ISBN 0-262-20107-0. US 2006/0167784 A1 Jul. 27, 2006 28
0553 Dasarathy, B. V. (ed.) (1991) Nearest Neighbor 0568 Fukunaga, K. & Flick, T. E. (1984) An optimal (NN) Norms: NN Pattern Classification Techniques. Los global nearest neighbor metric. IEEE Transactions on Pat Alamitos, Calif.: IEEE Computer Society Press. tern Analysis and Machine Intelligence 6, 314-318. Re printed in Dasarathy (1991). 0554 Dougherty, J., Kohavi, R. & Sahami, M. (1995) Supervised and unsupervised discretization of continuous 0569. Fulton, T., Kasif, S. & Salzberg, S. (1995) Efficient features. In Proceedings of the Twelfth International Con algorithms for finding multi-way splits for decision trees. In ference on Machine Learning, eds A. Prieditis & S. Russell, Proceedings of the Twelfth International Conference on pp. 194-202. San Francisco: Morgan Kaufmann. Machine Learning, eds. A. Prieditis & S. Russell, pp. 244 251. San Francisco: Morgan Kaufmann. 0555 Drucker, H. (1996) Fast decision tree ensembles for optical character recognition. In Proceedings of the Fifth 0570) Fung, R. & Del Favarro, B. (1995) Applying Baye Annual Symposium on Document Analysis and Information sian networks to information retrival. Communications of Retrival. the ACM 38(3), 42-48, 57. 0556 Drucker, H. & Cortes, C. (1996) Boosting decision 0571 Hampson, S. E. & Volper, D. J. (1986) Linear trees. In Touretzky et al. (1996), pp. 479-485. ISBN 0-262 function neurons: structure and training. Biological Cyber 2O107-O. netics. 0572 Hastie, T. & Tibshirani, R. (1996a) Discriminant 0557 Drucker, H., Schapire, R. & Simard, P. (1993) adaptive nearest neighbor classification and regression. In Boosting performance in neural networks. International Journal of Pattern Recognition and Artificial Intelligence Touretzky et al. (1996), pp. 409–415. ISBN 0-262-20107-0. 7(4), 705-719. 0573 Hastie, T. & Tibshirani, R. (1996b) Discriminant adaptive nearest neighbor classification. IEEE Transactions 0558 Drucker, H., Cortes, C., Jaeckel, L. D., LeCun, Y. on Pattern Analysis and Machine Intelligence 18, 607-618. & Vapnik, V. (1994) Boosting and other ensemble methods. Neural Computation 6(6), 1289-1301. 0574 Heath, D., Kasif, S. & Salzberg, S. (1993) Learning oblique decision trees. In Proceedings of the the Thirteenth 0559) Edwards, D. (1995) Introduction to Graphical International Joint Conference on Artificial Intelligence Modelling. Springer. (Chambery, France, 1993), pp. 1002-1007. San Francisco: 0560 Elomaa, T. & Rousu, J. (1996) Finding optimal Morgan Kaufmann. multi-splits for numerical attributes in decision tree learning. 0575 Heckerman, D. & Wellman, M. P. (1995) Bayesian NeuroCOLT Technical Report Series NC-TR-96-04 1, networks. Communications of the ACM 38(3), 26-30. Department of Computer Science, University of Helsinki. 0576. Heckerman, D., Breese, J. S. & Rommelse, K. 0561 Fayyad, U. M. & Irani, K. B. (1993) Multi-interval (1995) Decision-theoretic troubleshooting. Communica discretization of continuous-valued attributes in decision tions of the ACM 38(3), 49-57. tree generation. In Proceedings of the the Thirteenth Inter 0577 Helmbold, D. P. & Schapire, R. E. (1995) Predict national Joint Conference on Artificial Intelligence (Cham ing nearly as well as the best pruning of a decision tree. In bery, France, 1993), pp. 1022-1027. San Francisco: Morgan Proceedings of the Eight Annual Conference on Computa Kaufmann. tional Learning Theory, pp. 61-68. New York: ACM Press. 0562 Freund, Y. (1990) Boosting a weak learning algo 0578 Helmbold, D. P. & Schapire, R. E. (1996) Predict rithm by majority. In Proceedings of the Third Workshop on ing nearly as well as the best pruning of a decision tree. Computational Learning Theory, pp. 202-216. Morgan Machine Learning. Kaufmann. 0579. Ho, T. K. (1995) Random decision forests. In 0563 Freund, Y. (1995) Boosting a weak learning algo Proceedings of the Third International Conference on Docu rithm by majority. Information and Computation 121(2), ment Analysis and Recognition, pp.278-282. IEEE Com 256-285. puter Society Press. 0564 Freund, Y. & Schapire, R. E. (1995) A decision 0580 Hochreiter, S. & Schmidhuber, J. (1995) Simpli theoretic generalization of on-line learning and an applica fying neural nets by discovering flat minima. In Tesauro et tion to boosting. In Proceedings of the Second European al. (1995), pp. 529-536. ISBN 0-262-20104-6. Conference on Computational Learning Theory, pp. 23-37. 0581 Hochreiter, S. & Schmidhuber, J. (1996) Flat Springer. minima. Neural Computation. 0565 Freund, Y. & Schapire, R. E. (1996a) Game theory, 0582 Holte, R. C. (1993) Very simple classification rules on-line prediction and boosting. In Proceedings of the Ninth perform well on most commonly used datasets. Machine Annual Conference on Computational Learning Theory. Learning 11, 63-91. 0566) Freund, Y. & Schapire, R. E. (1996b) Experiments 0583. Hoffgen, K.-U., Simon, H.-U. & Van Horn, K. S. with a new boosting algorithm. In Proceedings of the (1995) Robust trainability of single neurons. Journal of Thirteenth International Conference on Machine Learning. Computer and System Sciences 50(1), 114-125. 0567 Friedman, J. H. (1994) Flexible metric nearest 0584) Karpinski, M. & Macintyre, A. (1995a) Bounding neighbor classification. Technical report, Department of VC-dimension of neural networks: Progress and prospects. Statistics, Stanford University. In Proceedings of the Second European Conference on US 2006/0167784 A1 Jul. 27, 2006 29
Computational Learning Theory (Barcelona, Spain), ed. P. 0602 Parberry, I. (1994) Circuit Complexity and Neural Vitanyi, number 904 in Lecture Notes in Artificial Intelli Networks. Cambridge, MA: MIT Press. ISBN 0-262-16148 gence, pp. 337-341. Berlin: Springer. 6. 0585 Karpinski, M. & Macintyre, A. (1995b) Polyno 0603 Parrondo, J. M. R. & Van der Broeck, C. (1993) mial bounds for VC dimension of sigmoidal neural net Vapnik-Chervonenkis bounds for generalization. J. Phys. A works. In Proceedings of the Twenty-Seventh Annual ACM 26, 2211-2223. Symposium on Theory of Computing (Las Vegas), pp.200 208 ACM PreSS. 0604) Przytula, K. W. & Prasanna, V. K. (1993) Parallel Digital Implementation of Neural Networks. Englewood 0586 Koiran, P. & Sontag, E. D. (1996) Neural networks Cliffs, N.J.: Prentice Hall. with quadratic VC dimension. In Touretzky et al. (1996), pp. 197-2O3. ISBN 0-262-2O107-O. 0605 Quinlan, J. R. (1996a) Bagging, boosting, and C4.5. In Proceedings of the Fourteenth National Conference 0587 Kooperberg, C., Bose, S. & Stone, C. J. (1996) on Artificial Intelligence. Menlo Park, Calif: AAAI Press. Polychotomous regression. Journal of the American Statis tical Association. 0606 Quinlan, J. R. (1996b) Improved use of continuous 0588 Krogh, A. & Vedelsby, J. (1995) Neural network attributes in C4.5. Journal of Artificial Intelligence Research ensembles, cross validation, and active learning. In Tesauro 4, 77-90. et al. (1995), pp. 231-238. ISBN 0-262-20104-6. 0607 Rachlin, J., Kasif, S., Salzberg, S. & Aha, D. (1994) 0589 Langley, P. (1996) Elements of Machine Learning. Towards a better understanding of memory-based and Baye sian classifiers. In Proceedings of the Eleventh International San Francisco: Morgan Kaufmann. Conference on Machine Learning (New Brunswick, N.J.), 0590 Langley, P. & Simon, H. A. (1995) Applications of pp. 242-250. machine learning and rule induction. Communications of the ACM 38(11), 54-64. 0608) Sakurai, A. (1993) Tighter bounds of the VC dimension of three-layer networks. In Proceedings of the 0591 Lauritzen, S. L. (1996) Graphical Models. Oxford: 1993 World Congress on Neural Networks, volume 3, pp. Clarendon Press. ISBN 0-19-852219-3. 540-543. Hillsdale, N.J.: Erlbaum. 0592 Ljung, L., Pflug, H. & Walk, H. (1992) Stochastic Approximation and Optimization of Random Systems. Ber 0609 Schapire, R. E. (1990) The strength of weak learn lin: Birkhauser. ability. Machine Learning 5(2), 197-227. 0610 Shafer, G. (1996) Probabilistic Expert Systems. 0593. Lowe, D. G. (1995) Similarity metric learning for Number 67 in CBMS-NSF Regional Conference Series in a variable-kernel classifier. Neural Computation 7(1), 72-85. Applied Mathematics. Philadelphia, Pa.; SIAM. ISBN 0594 Maass, W. & Tur'an, G. (1994) How fast can a O-898.71-373-O. threshold gate learn? In Computational Learning Theory and Natural Learning Systems: Constraints and Prospects, eds S. 0611 Shlien, S. (1990) Multiple binary decision tree J. classifiers. Pattern Recognition 23(7), 757-763. 0595 Hanson, G. A. Drastal & R. L. Rivest, volume I, pp. 0612 Short, R. D. & Fukunaga, K. (1980) A new nearest 381-414. MIT Press. neighbor distance measure. In Proceedings of the Fifth IEEE International Conference on Pattern Recognition (Miami 0596) Marchand, M., Golea, M. & Rujan, P. (1990) A Beach, 1980), pp. 81-86. Los Alamitos, Calif.: IEEE Com convergence theorm for sequential learning in two-layer puter Society Press. perceptrons. Europhysics Letters 11, 487-492. 0613) Short, R. D. & Fukunaga, K. (1981) The optimal 0597 Muroga, S. (1965) Lower bounds of the number of distance measure for nearest neighbor classification. IEEE threshold functions and a maximum weight. IEEE Transac Transactions on Information Theory 27, 622-627. Re tions on Electronic Computers 14, 136-148. printed in Dasarathy (1991). 0598) Muroga, S., Toda, I. & Takasu, S. (1961) Theory of 0.614 Stone, C. J., Hansen, M., Kooperberg, C. & majority decision elements. Journal of the Franklin Institute Truong, Y. K. (1997) Polynomial splines and their tensor 271, 376-418. products in extended linear modeling. Annals of Statistics. 0599 Murthy, S. K. Kasif, S., Salzberg, S. & Beigel, R. (1993) OCI: randomized induction of oblique decision tress. 0.615 Tesauro, G., Touretzky, D. S. & Leen, T. K. (eds) In Proceedings of the Eleventh National Conference on (1995) Advances in Neural Information Processing Systems Artificial Intelligence (Washington, D.C., 1993), pp.322 7. Proceedings of the 1994 Conference. Cambridge, Mass.: 327. AAAI PreSS. ISBN 0-262-51O71-5. MIT PreSS. ISBN 0-262-2O104-6. 0600 Murthy, S. K. Kasif, S. & Salzberg, S. 0616) Touretzky, D. S., Moser, M. C. & Hasselmo, M. E. (1994)Asystem for the induction of oblique decision trees. (eds) (1996) Advances in Neural Information Processing Journal of Artificial Intelligence Research 2, 1-33. Systems 8. Proceedings of the 1995 Conference. Cambridge, MaSS: MIT PreSS. ISBN 0-262-20107-O. 0601 Oliver, J. J. & Hand, D. J. (1995) On pruning and averaging decision trees. In Machine Learning: Proceedings 0.617 Utgoff, P. E. (1989) Perceptron trees: a case study of the Twelfth International Conference, pp. 430-437. Mor in hybrid concept representations. Connection Science 1(4), gan Kaufmann. 377-391. US 2006/0167784 A1 Jul. 27, 2006 30
0618) Valentin, D., Abdi, H., O'Toole, A.J. & Cottrell, G. 064.0 G. Strang “Wavelets and dilation equations: a brief (1994) Connectionist models of face processing: A Survey. introduction.” Pattern Recognition 27, 1208-1230. 0641 C. Torrence and G. P. Compo, “A Practical Guide 0619 Vapnik, V. N. (1995) The Nature of Statistical to Wavelet Analysis, with Significance and Confidence Learning Theory. New York: Springer. Testing.” 0620. Vapnik, V. N. (1996) Statistical Learning Theory. 0642 B. Vidakovic, “Wavelets for Kids,” also part 2. New York: Wiley. 0643) E. Tolkova, “Orthogonal Wavelets Construction.” 0621) Wahba, G., Wang, Y. Gu, C., Klein, R. & Klein, B. (1996) Smoothing spline anova for exponential families, 0644) Y. Meyer, Wavelets: Algorithms and Applications, with application to the Wisconsin epidemiological study of Society for Industrial and Applied Mathematics, Phila diabetic retinopathy. Annals of Statistics 23 (6), 1865-1895. delphia, 1993, pp. 13–31, 101-105. 0622 Waltz, D. (1990) Memory-based reasoning. In 0.645 G. Kaiser, A Friendly Guide to Wavelets, Natural and Artificial Parallel Computation, eds M. Arbib & Birkhauser, Boston, 1994, pp. 44-45. J. Robinson, pp. 251-276. Cambridge, Mass.: MIT Press. General Theory 0623 Wang, L. & Oja, E. (1993) Image compression by 0646 L. Andersson, N. Hall, B. Jawerth and G. Peters, MLP and PCA neural networks. In Proceedings of the Eight “Wavelets on Closed Subsets of the Real Line Scandinavian Conference on Image Analysis (Tromsø, Nor way), pp. 1317-1324. 0647 P. Auscher, G. Weiss and M. V. Wickerhauser, “Local Sine and Cosine Bases of Coifman and Meyer and 0624 Werbos, P. (1987) Learning how the world works. the Construction of Smooth Wavelets' In Proceedings of the IEEE Conference on Systems, Man and Cybernetics, pp. 320-310. New York: IEEE Press. 0648 C. Basdevant and V. Perrier, “Besov Norms in 0625 Werbos, P. J. (1995) Backpropagation: Basics and Terms of Continuous Wavelet Transforms and Applica new developments. pp. 134-139 of Arbib (1995). tion to Structure Functions.” 0626) Willems, F. M. J., Shtarkov, Y. M. & Tjalkens, T. J. 0649 B. E. Bassil, G. J. Dickson and D. M. Monro, (1993) Context tree weighting: A sequential universal source “Orthonormal Wavelets With Balanced Uncertainty”. coding procedure for FSMX sources. In Proceedings of the 0650 G. Beylkin and N. Saito, “Multiresolution Repre 1993 IEEE International Symposium on Information sentations using the Autocorrelation Functions of Com Theory, p.59. IEEE Press. pactly Supported Wavelets.” 0627 Willems, F. M. J., Shtarkov, Y. M. & Tjalkens, T. J. 0651) G. Beylkin and N. Saito, “Wavelets, their Autocor (1995) The context-tree weighting method: Basic properties. relation Functions and Multiresolution Analysis of Sig IEEE Transactions on Information Theory pp. 653-664. nals.” WAVELETS REFERENCES 0652 G. Beylkin and B. Torresani, “Implementation of Operators via Filter Banks, Autocorrelation Shell and 0628 Introductions to Wavelets Hardy Wavelets.” 0629 G. Kaiser, “A Friendly Guide to Wavelets” 0653 A.G. Bruce, H. Gao and D. Ragozin, “Non-smooth 0630 C. S. Burrus and R. A. Gopinath, “A Tutorial Wavelets: Graphing Functions Unbounded on Every Overview of Wavelets, Filter Banks and Interrelation Interval.’ ships' 0654 C. Cabrelli, C. Heil, and U. Molter, “Accuracy of 0631) R. A. DeVore and B. J. Lucier, “Wavelets” Lattice Translates of Several Multidimensional Refine 0632) T. Edwards, “Discrete Wavelet Transforms: Theory able Functions.” and Application.” 0655 R. C. Calderbank, I. Daubechies, W. Sweldens and 0633 E. Gootman and M. Wickerhauser, “Elementary B. Yeo “Wavelet Transforms that Map Integers to Inte Wavelets. gers.” There is also an uncompressed version. 0634 A. Graps, “An Introduction To Wavelets.” 0656 D. Chen, “Extended Families of Cardinal Spline Wavelets. 0635 C. Heil and D. Walnut, “Continuous and Discrete Wavelet Transforms. 0657 D. Chen, “Spline Wavelets of Small Support.” 0636 B. Jawerth and W. Sweldens, “An Overview of 0658 D. Chen, “Characterization of Biorthogonal Cardi Wavelet Based Multiresolution Analysis”. An abstract is nal Spline Wavelet Bases.” also available. 0659) D. Chen, “Cardinal Spline Wavelets”, dissertation. 0637. J. Lewalle, “Tutorial on Wavelet Analysis of Also part 2, part 3, part 4, part 5 and part 6. Experimental Data' 0660 S. Chen and D. Donoho, “Atomic Decomposition 0638 P. Schröder and W. Sweldens, “Building Your Own by Basis Pursuit.” Wavelets at Home.” An abstract is also available. 0661 A. Cohen, W. Dahmen and R. DeVore, “Multiscale 0639 G. Strang, “Wavelets.” Decompositions on Bounded Domains.” US 2006/0167784 A1 Jul. 27, 2006
0662) J. Cohen, “The Foot Problem in Wavelet Packet 0686) S. Haykin and S. Mann, “The Chirplet Transform: Splitting.” A Mathematica notebook converted to Post A New Signal Analysis Technique Based on Affine Rela Script. tionships in the Time-Frequency Plane.” This about 3.5 0663 J. Cohen, “Schauder Basis for 0,1).” A Math MB. ematica notebook converted to Postscript. 0687 C. Heil, G. Strang and V. Strela, “Approximation 0664) J. Cohen, “The Littlewood-Paley-Stein Wavelet.” By Translates of Refinable Functions.” A Mathematica notebook converted to Postscript. 0688 C. Heil and G. Strang, “Continuity of the Joint Spectral Radius: Application to Wavelets.” 0665) J. Cohen, “Battle-Lemarie Wavelets.” A Math ematica notebook converted to Postscript. 0689 B. Jawerth and W. Sweldens, “Biorthogonal Smooth Local Trigonometric Bases.” An abstract is also 0666) J. Cohen, “The Daubechies Minimum Phase Wave available. lets.” A Mathematica notebook converted to Postscript. 0690 B. Jawerth and W. Sweldens, “Weighted Multi 0667. J. Cohen, “Meyer Wavelets.” A Mathematica note wavelets on General Domains.” book converted to Postscript. 0691 M. K. Kwong and P. T. Peter Tang, “W-Matrices, 0668) R. Coifman and M. V. Wickerhauser, “Entropy Nonorthogonal Multiresolution Analysis and Finite Sig Based Algorithms for Best Basis Selection' nals of Arbitrary Length.” 0669 R. Coifman and M. Wickerhauser, “Best-adapted 0692 G. Leaf, J. M. Restrepo and G. Schlossnagle, Wave Packet Bases.’ “Periodized Daubechies Wavelets. 0670) R. Coifman, “Numerical Harmonic Analysis.” 0693) J. Lippus, “Wavelet Coefficients of Functions of Generalized Lipschitz Classes.” 0671) R. Coifman, Y. Meyer and M. Wickerhauser, “Size Properties of Wavelets Packets.” 0694 S. Mallat and Z. Zhang, “Matching Pursuit with Time-Frequency Dictionaries.” 0672 R. Coifman and Y. Meyer, “Orthonormal Wave Packet Bases.’ 0695 E. J. McCoy, D. B. Percival and A. T. Walden, “On the Phase of Least-Asymmetric Scaling and Wavelet 0673) R. Coifman and M. V. Wickerhauser, “Wavelets Filters. and Adapted Waveform Analysis” 0696 R. Piessens and W. Sweldens, “Wavelet Sampling 0674) R. Coifman, Y. Meyer and M. Wickerhauser, Techniques.” An abstract is also available. “Adapted Wave Form Analysis, Wavelet Packets and 0697) J. Shen and G. Strang, “The Zeros of the Applications.” Daubechies polynomials.” 0675 D. Colella and C. Heil, “Matrix Refinement Equa 0.698) J. Shen and G. Strang, “Asymptotics of Daubechies tions: Existence and Accuracy.’ Filters, Scaling Functions and Wavelets.” 0676 S. Dahlke. W. Dahmen, E. Schmitt and I. Weinre ich, “Multiresolution Analysis and Wavelets on S 2 and S 0699 M. J. Shensa, “An Inverse DWT for Nonorthogo 3' nal Wavelets 0700 W. Sweldens, “Compactly Supported Wavelets 0677 W. Dahmen, “Stability of Multiscale Transforma which are Biorthogonal to a Weighted Inner Product. An tions.” abstract is also available. 0678 W. Dahmen and C. A. Micchelli, “Biorthogonal 0701 W. Sweldens, “The Lifting Scheme: A Custom Wavelet Expansions.” Design Construction of Biorthogonal Wavelets. An 0679) G. Davis, S. Mallat and Z. Zhang, “Adaptive abstract is also available. Nonlinear Approximations.” 0702 W. Sweldens, “The Lifting Scheme: A Construc 0680 G. Davis, “Adaptive Nonlinear Approximations.” tion of Second Generation Wavelets.” An abstract is also available. 0681 G. Davis, S. Mallat and Z. Zhang, "Adaptive Time-Frequency Approximations with Matching Pur 0703 B. Suter and X. Xia, “Vector Valued Wavelets and suits.” Vector Filter Banks. 0682 I. Daubechies and W. Sweldens. An abstract is also 0704. C. Taswell, “Wavelet Transform Algorithms for available. Finite Duration Discrete-Time Signals.” 0683 C. deBoor, R. DeVore and R. Amos, “On the 0705 C. Taswell, “Near-Best Basis Selection Algorithms Construction of Multivariate (Pre)Wavelets.” with Non-Additive Information Cost Functions.” 0684 R. L. deOueiroz, “On Lapped Transforms.' 0706 K. Urban, “On Divergence-Free Wavelets.” 0707 R. O. Wells Jr., “Recent Advances in Wavelet 0685 M. Girardi and W. Sweldens, “A New Class of Technology” Unbalanced Haar Wavelets That Form an Unconditional Basis for Lp on General Measure Spaces.” An abstract is 0708 M. V. Wickerhauser, “Entropy of a Vector Relative also available. to a Decomposition.” US 2006/0167784 A1 Jul. 27, 2006 32
0709) M. V. Wickerhauser, “Lectures on Wavelet Packet 0731 C. Basdevant, V. Perrier and T. Philipovitch, Algorithms' “Wavelet Spectra Compared to Fourier Spectra.” 0710) M. V. Wickerhauser, “Smooth Localized Orthonor 0732) J. P. Bonnet, J. Lewalle and M. N. Glauser, “Coher mal Bases' ent Structures: Past, Present and Future.” 0711 C. Zarowski, “Notes on Orthogonal Wavelets and 0733 G. Buresti, J. Lewalle and P. Petagna, “Wavelet Wavelet Packets’ statistics and the near-field structure of coaxial jets” 0712 V. Zavadsky, “Multiresolution Approximations of 0734 C. S. Burrus and R. A. Gopinath, “Wavelet-Based Banach Spaces.” Lowpass/Bandpass Interpolation' 0713 V. Zavadsky, “Wavelet Approximation of Sampled 0735 C. S. Burrus, R. A. Gopinath and J. E. Odegard, Functions.” “Design of Linear Phase Cosine Modulated Filter Banks 0714 B. G. Sherlock and D. M. Monro, “On the Space of for Subband Image Compression' Orthonormal Wavelets. 0736 S. Cabrera, V. Krienovich and O. Sirisaengtaksin, 0715 M. Vetterli and C. Herley, “Wavelets and Filter “Wavelets Compress Better Than All Other Methods: A Banks: Theory and Design.” IEEE Transactions on Signal 1-D Theorem. Processing, Vol. 40, 1992, pp. 2207-2232. 0737. S. Cabrera, V. Krienovich and O. Sirisaengtaksin, Frame Decompositions “Wavelet Nerual Networks are Optimal Approximators 0716) J. Benedetto, C. Heil, and D. Walnut, “Differentia for Functions of One Variable. tion and the Balian-Low Theorem.” 0738 R. Carmona, W. L. Hwang and B. Torresani, “Char 0717 O. Christensen and C. Heil, “Perturbations of acterization of Signals by the Ridges of Their Wavelet Banach Frames and Atomic Decompositions.” Transforms.' 0718 D. M. Healy, Jr. and S. Li, “On Pseudo Frame 0739 R. Carmona, W. L. Hwang and B. Torresani, Decompositions and Discrete Gabor Expansions.” “Multi-Ridge Detection and Time-Frequency Reconstruc tion. 0719. S. Li, “General Frame Decompsotions, Pseudo Duals and Applications for Weyl-Heisenberg Frames.” 0740 R. Coifman, "Adapted Multiresolution Analysis, Computation, Signal Processing and Operator Theory’ 0720 S. Li, “On Dimension Invariance of Discrete Gabor Expansions.” 0741. R. Coifman, Y. Meyer, S. Quake and M. Wicker hauser, “Signal Processing and Compression with Wave M-Band Wavelets and Filter Banks Packets. 0721 C. S. Burrus and R. A. Gopinath, “On the Corre 0742 R. Coifman, Y. Meyer and M. V. Wickerhauser, lation Structure of Multiplicity M Scaling Functions' “Wavelet Analysis and Signal Processing 0722 C. S. Burrus and R. A. Gopinath, “Wavelets and Filter Banks’ 0743 P. Crane, H. Higuchi and J. Lewalle, “On the structure of two-dimensional wakes behind a pair of flat 0723 C. S. Burrus and R. A. Gopinath, “Unitary FIR plates' Filter Banks and Symmetry” 0744) M. Goldburg, “Applications of Wavelets to Quan 0724 C. S. Burrus and R. A. Gopinath, “Theory of tization and Random Process Representations. About 1.1 Modulated Filter Banks and Modulated Wavelet Tight MB. Frames’ 0745) D. M. Healy, Jr., J. Lu and J. B. Weaver, “Signal 0725 C. S. Burrus and R. A. Gopinath, “Factorization Recovery and Wavelet Reproducing Kernels.” Approach to Time-Varying Unitary Filter Bank Trees and Wavelets’ 0746 D. M. Healy, Jr., J. Lu, J. B. Weaver and Y. Xu, “Noise Reduction With Multiscale Edge Representation 0726 C. Herley, “Boundary Filters for Finite-Length and Perceptual Criteria.” Signals and Time-Varying Filter Banks.” 0747 D. M. Healy, Jr. and J. Lu, “Contrast Enhancement 0727 P. Steffen, P. Heller, R. A. Gopinath and C. S. via Multiscale Gradient Transformations.” Burrus, “The Theory of Regular M-Band Wavelets” 0748 W. Hwang and S. Mallat, “Singularity Detection Wavelets and General Signal Processing and Processing with Wavelets.”“BJawerth, Y. Liu and W. 0728 M. Vetterli and J. Kovacevic, “Wavelets and Sub Sweldens, “Signal Compression with Smooth Local band Coding, Prentice Hall, 1995. Trigonometric Bases.” An abstract is also available. 0729) D. E. Ashpis and J. Lewalle, “Transport in bypass 0749) M. M. Lankhorst and M. D. van der Laan, “Wave transition: mapping the active time scales using wavelet let-Based Signal Approximation with Genetic Algo techniques' rithms.” 0730) D. E. Ashpis and J. Lewalle, “Demonstration of 0750 J. Lewalle, K. Read and M. T. Schobeiri, “Effect of wavelet techniques in the spectral analysis of bypass unsteady wake-passing frequency on boundary layer tran transition data' sition—experimental investigation and wavelet analysis’ US 2006/0167784 A1 Jul. 27, 2006
0751 J. Lewalle, S.J. Murphy and F. W. Peek, “Wavelet 0773) F. Murtagh, J. L. Starck and M. W. Berry, “Over analysis of olfactory nerve response to stimulus' coming the Curse of Dimensionality in Clustering by 0752 J. Lewalle, “Wavelet analysis of experimental data: means of the Wavelet Transform.” Some methods and the underlying physics’ 0774 R. A. Carmona, R. D. Frostig and W. L. Hwang, 0753 G. Strang, “Eigenvalues of (2)H and convergence “Wavelet Analysis for Brain Function Imaging.” of the cascade algorithm.” 0775 A. Chambolle, R. A. DeVore, N. Lee, and B. J. 0754 G. Strang, “Creating and comparing wavelets.” Lucier, “Nonlinear Wavelet Image Processing: Variational Problems, Compression, and Noise Removal through 0755 A. R. Tate, “Pattern Recognition Analysis of in Wavelet Shrinkage.” vivo Magnetic Resonance Spectra’ 0776. H. Chao and P. Fisher, “An Approach of Fast 0756 D. Donoho, “Nonlinear Wavelet Methods for Integer Reversible Wavelet Transforms for Image Com Recovery of Signals, Densities, and Spectra from Indirect pression.” and Noisy Data. Different Perspectives on Wavelets, Proceeding of Symposia in Applied Mathematics, Vol 47. 0777 R. A. DeVore and B. J. Lucier, “Fast Wavelet 1. Daubechies ed. Amer. Math. Soc., Providence, R.I., Techniques for Near-Optimal Image Processing 1993, pp. 173 -205. 0778 B. Deng, B. D. Jawerth, G. Peters and W. Sweld Wavelets and Image Processing ens, “Wavelet Probing for Compression Based Segmen 0757 E. Adelson and E. Simoncelli, “Subband Image tation'. An abstract is also available. Coding with Three-tap Pyramids.” 0779) J. Fan and A. Laine, “An Adaptive Approach for 0758 E. H. Adelson, W. T. Freeman, D. J. Heeger and E. Texture Segmentation by Multi-Channel Wavelet P. Simoncelli, “Shiftable Multi-Scale Transforms.' Frames. 0759 E. H. Adelson and E. P. Simoncelli, “Subband 0780 W.T. Freeman and E. P. Simoncelli, “The Steerable Transforms.' Pyramid: A Flexible Architecture for Multi-Scale Deriva tive Computation./C Source Code (75 k) 0760 V. R. Algazi, R. R. Estes and J. Lu, “Comparison of wavelet image coders using the Picture Quality Scale 0781 A. Grzeszczak, M. K. Mandal, S. Panchanathan and T. Yeap, “VLSI Implementation of Discrete Wavelet Transform.’ 0761) M. Bhatia, W. C. Karl, and A. S. Willsky, “A Wavelet-Based Method for Multiscale Tomographic 0782. O. Guleryuz, M. T. Orchard and Z. Xiong, “A Reconstruction.” DCT-based Embedded Image Coder.” 0762) M. Bhatia, W. C. Karl, and A. S. Willsky, “Using 0783 D. M. Healy, Jr., J. Lu and J. B. Weaver, “Contrast Natural Wavelet Bases and Multiscale Stochastic Models Enhancement of Medical Images Using Multiscale Edge for Tomographic Reconstruction.” Representation.” 0763 M. Louys, J. L. Starck, S. Mei, F. Bonnarel, and F. 0784 C. Heil, P. N. Heller, G. Strang, V. Strela, and P. Murtagh, “Astronomical Image Compression.” Topiwala, “Accuracy of Lattice Translates of Several 0764 M. Louys, J. L. Starck and F. Murtagh, “Lossless Multidimensional Refinable Functions.” Compression of Astronomical Images.” 0785 C. Herley, M. T. Orchard, K. Ramchandran and Z. 0765) F. Murtagh and J. L. Starck, “Wavelets and Mul Xiong, "Flexible Tree-structured Signal Expansions for tiscale Transforms in Massive Data Sets.” Compression Using Time-Varying Filter Banks.” 0766 F. Murtagh and J. L. Starck, “Image Processing 0786 M. L. Hilton, B. D. Jawerth and A. Sengupta, through Multiscale Analysis and Measurement Noise “Compressing Still and Moving Images with Wavelets” Modeling.” with figure. 0767 J. L. Starck and F. Murtagh, “Multiscale Entropy 0787 P. Kovesi, “Image Features from Phase Congru Filtering.” ency’ 0768 J. L. Starck and F. Murtagh, “Image Filtering by 0788 B. Lin, “Wavelet Phase Filter for Denoising Tomo Combining Multiple Vision Models”. graphic Image Reconstruction' 0769) F. Murtagh, “Wedding the Wavelet Transform and 0789 M. K. Mandal, T. Aboulnasr and S. Panchanathan, Multivariate Data Analysis.” “Image Indexing. Using Moments and Wavelets.” 0770 F. Murtagh and J. L. Starck, “Pattern Clustering 0790) M. K. Mandal, E. Chan, X. Wang and S. Pancha based on Noise Modeling in Wavelet Space.” nathan, “Multiresolution Motion Estimation Techniques for Video Compression.” 0771) G. Zheng, J. L. Starck, J. G. Campbell and F. Murtagh, “Multiscale Transforms for Filtering Financial 0791 M. K. Mandal, S. Panchanathan and T. Aboulnasr, Data Streams. “Choice of Wavelets for Image Compression.” 0772) M. Morehart, F. Murtagh and J. L. Starck, “Mul 0792 D. M. Monro and B. G. Sherlock, “Psychovisually tiresolution Spatial Analysis.” Tuned Wavelet Fingerprint Compression'. US 2006/0167784 A1 Jul. 27, 2006 34
0793 D. M. Monro and B. G. Sherlock, “Optimised 0813 J. N. Bradley and C. M. Brislawn, “The Wavelet/ Wavelets for Fingerprint Compression'. Scalar Quantization Compression Standard for Digital Fingerprint Images.” 0794. P. Moulin, “A Multiscale Relaxation Algorithm for SNR Maximization in 2-D Nonorthogonal Subband Cod 0814) J. Bradley, C. Brislawn and T. Hopper, “WSQ 1ng. 99 Gray-Scale Fingerprint Image Compression Specifica tion. 0795 M. T. Orchard, Z. Xiong and Y. Zhang, “A Simple Deblocking Algorithm for JPEG Compressed Images 0815 J. Bradley, C. Brislawn and T. Hopper, “The FBI Using Overcomplete Wavelet Representations.” Wavelet/Scalar Quantization Standard for Gray-Scale Fingerprint Image Compression' with figures. 0796 M. T. Orchard, K. Ramchandran and Z. Xiong, “Wavelet Packets Image Coding Using Space-Frequency 0816 C. M. Brislawn, Classification of Nonexpansive Symmetric Extension Transforms for Multirate Filter Quantization.” Banks’ 0797 M. T. Orchard, K. Ramchandran and Z. Xiong, “Space-frequency Quantization for Wavelet Image Cod 0817) C. M. Brislawn, “Fingerprints Go Digital” ing.” 0818 C. M. Brislawn, “Preservation of Subband Sym 0798 H. Pan, “Uniform Full-Information Image Match metry in Multirate Signal Coding.”“The FBI Wavelet/ ing Using Complex Conjugate Wavelet Pyramids”, with Scalar Quantization Fingerprint Image Compression figures. Standard.” 0799 H. Pan, “General Stereo Image Matching Using Wavelets and Speech Processing Symmetric Complex Wavelets.” presented at SPIE Con 0819 E. Wesfreid and M. V. Wickerhauser, “Adapted ference: Wavelet Applications in Signal and Image Pro Local Trigonometric Transforms and Speech Processing cessing, VI. Denver, August 1996, Published in SPIE 0820) M. Wickerhauser, "Acoustic Signal Compression Proceedings, vol. 2825. with Wavelets Packets. 0800 P. Schröder and W. Sweldens, “Spherical wavelets: Wavelets and Ordinary Differential Equations Efficiently representing functions on the sphere. An abstract is also available. 0821 G. Beylkin, “On Wavelet-based Algorithms for Solving Differential Equations.” 0801) P. Schröder and W. Sweldens, “Spherical Wavelets: Texture Processing.” An abstract is also available. 0822 B Jawerth and W. Sweldens, “Wavelet Multireso lution Analyses Adapted for the Fast Solution of Bound 0802. J. A. Solomon, J. Villasenor, A. B. Watson and G. ary Value Ordinary Differential Equations.” An abstract is Y. Yang, “Visual Thresholds For Wavelet Quantization also available. Error.” 0823. P. Monasse and V. Perrier, “Ondelettes sur 0803 V. Strela, P. Heller, G. Strang, P. Topiwala and C. 1Intervalle pour la Prise en Compte de Conditions aux Heil, “The application of multiwavelet filter banks to Limites.” signal and image processing.” 0824 A. Rieder, “Semi-Algebraic Multi-level Methods 0804) Y. Wang, “Image representations using multiscale Based on Wavelet Decompositions I: Application to Two differential operators.” Point Boundary Problems” 0805) Y. Wang and S. L. Lee, “Scale-space derived from 0825 W. C. Shann and J. C. Xu, “Galerkin-wavelet B-splines.” Methods for Two Point Boundary Value Problems.” 0806 G. Weiss, “Time-Frequency and Time-Scaling Wavelets and Partial Differential Equations Methods in Signal and Image Processing 0826 G. Kaiser, “Complex-Distance Potential Theory and Hyperbolic Equations' 0807 M. Wickerhauser, “Picture Compression by Best Basis Subband Coding.” 0827. A. Averbuch, G. Beylkin R. R. Coifman and M. Israeli, “Multiscale Inversion of Elliptic Operators.” 0808 M. V. Wickerhauser, “High-Resolution Still Picture Compression” 0828 E. Bacry, S. Mallat and G. Papanicolaou, “A Wave let Based Space-Time Adaptive Numerical Method for 0809 Z. Xiong, “Representation and Coding of Images Partial Differential Equations” Using Wavelets.” 0829 G. Beylkin and N. Coult, “A Multiresolution Strat 0810) D. M. Monro and B. G. Sherlock, “Space-Fre egy for Reduction of Elliptic PDEs and Eigenvalue quency Balance in Biorthogonal Wavelets.” Problems.’ 0811 Xuejun Li, "Low Bit Rate Wavelet Image and 0830) G. Beylkin and J. H. Keiser, “On the Adaptive Video Coding Algorithm and Software.” Numerical Solution of Nonlinear Partial Differential Equations in Wavelet Bases.” The FBI Wavelet Fingerprint Compression Standard 0831. D. M. Bond and S. A. Vavasis, “Fast Wavelet 0812 J. N. Bradley and C. M. Brislawn, “Proposed Transforms for Matrices Arising From Boundary Element First-Generation WSQ Bit Allocation Procedure” Methods.’ US 2006/0167784 A1 Jul. 27, 2006
0832 T. Chan, W. Tang and W. Wan. 0854 P. Monasse and V. Perrier, “Orthonormal Wavelet Bases Adapted for Partial Differential Equations with 0833 P. Charton and V. Perrier, “Factorisation sur Bases dOndelettes du Noyeau de la Chaleur et Algorithmes Boundary Conditions.” Matriciels Rapides Associes.” 0855 A. Rieder and X. Zhou, “On the Robustness of the Damped V-Cycle of the Wavelet Frequency Decomposi 0834 P. Charton and V. Perrier, “Towards a Wavelet Based Numerical Scheme for the Two-Dimensional tions Multigrid Method Navier-Stokes Equations.” 0856 A. Rieder, R. O. Wells, Jr. and X. Zhou, “A Wavelet Approach to Robust Multilevel Solvers for Anisotropic 0835 P. Charton and V. Perrier, “A Pseudo-Wavelet Elliptic Problems.” Scheme for the Two-Dimensional Navier-Stokes Equa tions.” 0857 A. Rieder, R. O. Wells, Jr. and X. Zhou, “On the Wavelet Frequency Decomposition Method” 0836) S. Dahlke and A. Kunoth, “Biorthogonal Wavelets and Multigrid.” 0858 K. Urban, “A Wavelet-Galerkin Algorithm for the Driven-Cavity-Stokes-Problem in Two Space Dimen 0837 S. Dahlke and I. Weinreich, “Wavelet-Galerkin Methods: An Adapted Biorthogonal Wavelet Basis.” sions.” 0859 O. V. Vasilyev and S. Paolucci, “A Dynamically 0838 S. Dahlke and I. Weinreich, “Wavelet Bases Adaptive Multilevel Wavelet Collocation Method for Adapted to Pseudo-Differential Operators.” Solving Partial Differential Equations in a Finite 0839 W. Dahmen and A. Kunoth, “Multilevel Precondi Domain.” tioning.” 0860) O. V. Vasilyev and S. Paolucci, “Thermoacoustic 0840 W. Dahmen, A. Kunoth and K. Urban “A Wavelet Wave Propagation Modeling. Using a Dynamically Adap Galerkin Method for the Stokes-Equations, also full tive Wavelet Collocation Method. version with pictures. 0861 O. V. Vasilyev and S. Paolucci, “A Fast Adaptive 0841) R. Glowinski, T. Pan, R. O. Wells, Jr. and X. Zhou, Wavelet Collocation Algorithm for Multi-Dimensional “Wavelet and Finite Element Solutions for the Neumann PDEs.” with figures. Problem Using Fictitious Domains” 0862 O. V. Vasilyev, S. Paolucci and M. Sen, “A Mul 0842) R. Glowinski, A. Rieder, R. O. Wells, Jr. and X. tilevel Wavelet Collocation Method for Solving Partial Zhou, “A Wavelet Multigrid Preconditioner for Dirichlet Differential Equations in a Finite Domain.” Boundary Value Problems in General Domains.” 0863 O. V. Vasilyev, Y.Y. Podladchikov and D. A. Yuen, 0843. R. Glowinski, A. Rieder, R. O. Wells, Jr. and X. “Modeling of Compaction Driven Flow in Poro-Vis Zhou, “A Preconditioned CG-Method for Wavelet-Galer coelastic Medium Using Adaptive Wavelet Collocation kin Discretizations of Elliptic Problems’ Method.” with figures. 0844) F. Heurtaux, F. Planchon and M. V. Wickerhauser, 0864 O. V. Vasilyev, D. A. Yuen and S. Paolucci, “The “Scale Decomposition in Burgers' Equation” Solution of PDEs Using Wavelets.” with figures. 0845 A. Jiang. 0865 O. V. Vasilyev, D. A. Yuen and Y.Y. Podladchikov, “Applicability of Wavelet Algorithm for Geophysical 0846. J. H. Keiser, "On I. Wavelet Based Approach to Viscoelastic Flow.” with figures. Numerical Solution on Nonlinear Partial Differential Equations and II. Nonlinear Waves in Fully Discrete 0866 R. O. Wells, Jr. and X. Zhou, “Wavelet Solutions Dynamical Systems.” for the Dirichlet Problem 0847 A. Kunoth, “Multilevel Preconditioning Append 0867. R. O. Wells, Jr. and X. Zhou, “Wavelet Interpola ing Boundary Conditions by Lagrange Multipliers.” tion and Approximate Solution of Elliptic Partial Differ ential Equations' 0848. G. Leaf and J. M. Restrepo, “Wavelet-Galerkin Discretization of Hyperbolic Equations.” 0868 R. O. Wells, Jr. and X. Zhou, “Representing the Geometry of Domains by Wavelets with Applications to 0849 J. Lewalle, “Wavelet Transforms of some Equa Partial Differential Equations” tions of Fluid Mechanics’ 0869. R. O. Wells, Jr., “Multiscale Applications of Wave 0850 J. Lewalle, “Energy Dissipation in the Wavelet lets to Solutions of Partial Differential Equations' Transformed Navier-Stokes Equations' Wavelets and Numerical Analysis 0851) J. Lewalle, “On the effect of boundary conditions on the multifractal statistics of incompressible turbu 0870) G. Beylkin, R. Coifman and V. Rokhlin, “Fast lence Wavelet Transforms and Numerical Algorithms I.’” 0852. J. Lewalle, “Diffusion is Hamiltonian”. 0871 G. Beylkin, “On the Representation of Operators in Bases of Compactly Supported Wavelets.” 0853 D. Lu, T. Ohyoshi and L. Zhu, “Treatment of Boundary Conditions in the Application of Wavelet 0872 G. Beylkin, “On the Fast Algorithm for Multipli Galerkin Method to a SH Wave Problem cation of Functions in the Wavelet Bases.” US 2006/0167784 A1 Jul. 27, 2006 36
0873. G. Beylkin, “Wavelets and Fast Numerical Algo 0893 F. Abramovich and B. W. Silverman, “The rithms.” Lecture notes for an AMS short course, 1993. Vaguelette-Wavelet Decomposition Approach to Statisti 0874) G. Beylkin, “Wavelets, Multiresolution Analysis cal Inverse Problems.’ and Fast Numerical Algorithms. Draft of INRIA lectures, 0894 E. H. Adelson and E. P. Simoncelli, “Noise May 1991. Removal via Bayesian Wavelet Coring.” 0875 G. Beylkin and M. E. Brewster, “A Multiresolution 0895 A. Antoniadis, G. Gregoire and G. P. Nason, “Den Strategy for Numerical Homogenization.” sity and Hazard Rate Estimation for Right Censored Data using Wavelet Methods.” 0876 P. Charton and V. Perrier, “Produits Rapides Matri ces-Vecteur en Bases d’Ondelettes: Application a la Reso 0896) T. Bailey, T. Sapatinas, K. Powell and W. J. Krza lution Numericue d’Equation aux Derivees Partielles.” nowski, "Signal Detection in Underwater Sounds using Wavelets. 0877 P. Charton, “Produits de Matrices Rapides en Bases dOndelettes: Application a la Resolution Numericue 0897] A. G. Bruce, D. L. Donoho, H. Gao and R. D. d’Equation aux Derivees Partielles.” Martin, “Denoising and Robust Non-linear Wavelet Analysis.” 0878 N. H. Getz, “A Fast Discrete Periodic Wavelet Transform.” An associated toolbox of Matlab routines is 0898 A. G. Bruce and H. Gao, “WaveShrink: Shrinkage also available. Functions and Thresholds.” 0879 L. Jameson, “On the Spline-Based Wavelet Differ 0899 A. G. Bruce and H. Gao, “WaveShrink with Semi entiation Matrix.’ soft Shrinkage.” 0880 L. Jameson, “On the Differention Matrix for 0900 A. G. Bruce and H. Gao, “Understanding Wave Daubechies-Based Wavelets on an Interval.” Shrink: Variance and Bias Estimation.” 0881. L. Jameson, “On the Daubechies-Based Wavelet 0901 A. G. Bruce, H. Gao and D. Ragozin, “S+WAVE Differentiation Matrix. LETS: An Object-Oriented Toolkit for Wavelet Analysis.” 0882) L. Jameson, “On the Wavelet Optimized Finite 0902 A. G. Bruce and H. Gao, “S+WAVELETS: Algo Difference Method.” rithms and Technical Details.” 0883 E. Kolaczyk, “Wavelet Methods for the Inversion 0903) J. Buckheit and D. Donoho, “WaveLab and Repro of Certain Homogeneous Linear Operators in the Pres ducible Research.” ence of Noisy Data,” with FIG. 5.1, FIGS. 5.2, FIG. 5.4, 0904) J. F. Burn, A. M. Wilson and G. P. Nason, “Impact FIG. 5.5, FIGS. 5.6, FIGS. 5.8, FIG. 5.10, FIGS. 5.11, During Equine Locomotion: Techniques for Measurement FIG. 5.13, and FIGS. 5.14. and Analysis.” 0884 R. Piessens and W. Sweldens, “Quadrature Formu 0905 R. Coifman and N. Saito, “Local Discriminant lae and Asymptotic Error Expansion of Wavelet Approxi Bases.’ mations of Smooth Functions.” An abstract is also avail 0906 R. Coifman and F. Majid, “Adapted Waveform able. Analysis and Denoising.” 0885 R. Piessens and W. Sweldens, “Asymptotic Error 0907 R. Coifman and D. Donoho, “Translation-Invariant Expansion of Wavelet Approximations of Smooth Func De-Noising.” tions II. An abstract is also available. 0908) R. Dahlhaus, M. H. Neumann and R. von Sachs, 0886 W. C. Shann, “Quadratures Involving Polynomials “Non-linear Wavelet Estimation of Time Varying and Daubechies Wavelets. Autoregressive Processes.” 0887 W. Sweldens, “Construction and Application of 0909. A. Davis, A. Marshak and W. Wiscombe, “Wave Wavelets in Numerical Analysis.” let-Based Multifractal Analysis of Non-Stationary and/or 0888) M. Wickerhauser, “Nonstandard Matrix Multipli Intermittent Geophysical Signals.” with figures. cation.” 0910) B. Deylon and A. Juditsky, “Wavelet Estimators. 0889) M. V. Wickerhauser, “Computation with Adapted Global Error Mesures Revisited. Time-Frequency Atoms” 0911 D. Donoho, “Nonlinear Solution of Linear Inverse 0890) M. V. Wickerhauser, “Wavelet Approximations to Problems by Wavelet-Vaguelette Decomposition Jacobians and the Inversion of Complicated Maps' 0912 D. Donoho, “Smooth Wavelet Decompositions Wavelets and Statistics with Blocky Coefficient Kernels' 0891 F. Abramovich, T. Sapatinas and B. W. Silverman, 0913 D. Donoho, “De-noising by Soft Thresholding “Wavelet Thresholding via a Bayesian Approach.” 0914 D. Donoho, “Interpolating Wavelet Transforms” 0892 F. Abramovich, T. Sapatinas and B. W. Silverman, 0915) D. Donoho, “Unconditional Bases are Optimal “Stochastic Atomic Decompositions in a Wavelet Dictio Bases for Data Compression and for Statistical Estima nary. tion US 2006/0167784 A1 Jul. 27, 2006 37
0916 D. Donoho and I. Johnstone, “Adapting to 0941 M. H. Neumann and R. von Sachs, “Wavelet Unknown Smoothness by Wavelet Shrinkage' Thresholding in Anisotropic Function Classes and Appli 0917 D. Donoho and I. Johnstone, “Ideal Spatial Adap cation to Adaptive Estimation of Evolutionary Spectra.” tation via Wavelet Shrinkage' 0942 A. B. Owen, “Monte Carlo Variance of Scrambled Equidistribution Quadrature.” 0918 D. Donoho and I. Johnstone, “Minimax Estimation via Wavelet Shrinkage' 0943) D. B. Percival, “On the Estimation of the Wavelet Variance.” 0919) D. Donoho and I. Johnstone, “Minimax Risk over 1 p Balls” 0944 A. Pinheiro and B. Vidakovic, “Estimating the 0920 D. Donoho, I. Johnstone, G. Kerkyacharian and D. Square Root of a Density Via Compactly Supported Picard, “Density Estimation via Wavelet Shrinkage' Wavelets. 0921. D. Donoho, I. Johnstone, G. Kerkyacharian and D. 0945) J. Raz, L. Dickerson and B. Turetsky, “A Wavelet Picard, “Wavelet Shrinkage: Asymptopia?” Packet Model of Evoked Potentials. 0946 N. Saito, “Local Feature Extraction and Its App 0922 D. Donoho and I. Johnstone, “Ideal Denoising in plications. Using a Library of Bases.” an Orthonormal Basis Chosen From a Library of Bases.” 0923. D. Donoho, S. Mallat and R. von Sachs, “Estimat 0947 N. Saito, “Simultaneous Noise Supression and Sig ing Covariances of Locally Stationary Processes: Rates of nal Compression using a Library of Orthonormal Bases Convergence of Best Basis Methods.” and the Minimum Description Length Criterion.” 0924) T. Downie and B. W. Silverman, “The Discrete 0948 B. Vidakovic, “A Note on Random Densities via Multiple Wavelet Transform and Thresholding Methods.” Wavelets’ 0925 H. Y. Gao, “Choice of Thresholds for Wavelet 0949 B. Vidakovic, “Nonlinear Wavelet Shrinkage with Shrinkage Estimate of the Spectrum.” Bayes Rules and Bayes Factors.” 0950 R. von Sachs, G. P. Nason and G. Kroisandt, 0926 P. Goel and B. Vidakovic, “Wavelet Transforma “Adaptive Estimation of the Evolutionary Wavelet Spec tions as Diversity Enhancers' trum. 0927 P. Hall and G. P. Nason, “On Choosing a Non 0951 R. von Sachs and K. Schneider, “Smoothing of integer Resolution Level when Using Wavelet Methods.” Evolutionary Spectra by Non-linear Thresholding.” Also 0928 I. Johnstone, “Minimax Bayes, Asymptotic Mini available are the figures. max and Sparse Wavelet Priors’ 0952 R. von Sachs and M. H. Neumann, “A Wavelet 0929 I. M. Johnstone and B. W. Silverman, “Wavelet based Test for Stationarity.” Threshold Estimators for Data with Correlated Noise.” 0953 R. von Sachs and B. MacGibbon, “Non-parametric 0930 A. Juditsky, “Wavelet Estimators. Adapting To Curve Estimation by Wavelet Thresholding with Locally Unkown Smoothness.” Stationary Errors.” 0931 A. Juditsky and F. Leblanc, “Computing Wavelet 0954 A. T. Walden, D. B. Percival and E. J. McCoy, Density Estimators for Stochastic Processes.” “Spectrum Estimation by Wavelet Thresholding of Mul 0932 G. Katul and B. Vidakovic, “Partitioning eddy titaper Estimators.” motion using Lorentz wavelet filtering.” 0955) Yazhen Wang, “Jump and sharp cusp detection by wavelets.” 0933) R. Morgan and G. P. Nason, “Wavelet Shrinkage of Itch Response Sata.” 0956 Yazhen Wang, “Function estimation via wavelet shrinkage for long-memory data.” 0934 P. Moulin, “Wavelet Thresholding Techniques for Power Spectrum Estimation.” 0957 Yazhen Wang, “Small ball problems via wavelets for Gaussian processes.” 0935 G. P. Nason and B. W. Silverman, “The Discrete Wavelet Transform in S. 0958) Yazhen Wang, "Fractal function estimation via wavelet shrinkage.” 0936 G. P. Nason and B. W. Silverman, “The Stationary Wavelet Transform and some Statistical Applications.” 0959) Yazhen Wang, “Minimax estimation via wavelets for indirect long-memory data.” 0937 G. P. Nason and B. W. Silverman, “Wavelets for 0960 Yazhen Wang, “Change curve estimation via wave Regression and other Statistical Problems.” lets” (with an application to image processing); FIG. 4(a), 0938 G. P. Nason, T. Sapatinas and A. Sawczenko, FIG. 4(b). “Statistical Modelling of Time Series using Non-deci 0961 Yazhen Wang, “Change-point analysis via wavelets mated Wavelet Representations.” for indirect data.” 0939 G. P. Nason, “Wavelet Regression by Cross-Vali 0962 Yazhen Wang, “Self-similarity index estimation via dation' wavelets for locally self-similar processes'(with 094.0 G. P. Nason, “Functional Projection Pursuit.” Cavanaugh and Song); FIG. 1, FIG. 2, FIGS. 3 and 4. US 2006/0167784 A1 Jul. 27, 2006
0963 M. V. Wickerhauser, “Fast Approximate Factor 0980) M. H. Gross and R. Koch, “Visualization of Mul Analysis.” tidimensional Shape and Texture Features in Laser Range Wavelets and Econometrics Data using Complex-Valued Gabor Wavelets.” 0981 M. H. Gross and L. Lippert, “Ray-tracing of Mul 0964 S. A. Greenblatt, “Wavelets in Economics: An tiresolution B-Spline Volumes.” Also abstract available. Application to Outlier Testing.” Technical Report No. 239, Computer Science Depart 0965) M. J. Jensen, “Wavelet Analysis of Fractionally ment, ETH Zürich, 1996. Integrated Processes.” 0982) P. Hanrahan and P. Schröder, “Wavelet Methods for 0966 M. J. Jensen, “OLS Estimate of the Fractional Radiance Computations.” With FIG. 8, FIG. 9 left, FIG. Differencing Parameter Using Wavelets Derived From 9 right, FIG. 10, FIG. 10 top left and FIG. 10 top right. Smoothing Kernels.” 0983 C. Herley, “Exact Interpolation and Iterative Sub Wavelets and Fractals division Schemes.” 0967 A. Davis, A. Marshak and W. Wiscombe, “Wave 0984 P. Schröder, W. Sweldens and D. Zorin, “Interpo let-Based Multifractal Analysis of Non-Stationary and/or lating Subdivision for Meshes with Arbitrary Topology’. Intermittent Geophysical Signals.” with figures. An abstract is also available. 0968 C. Jones, 2-D Wavelet Packet Analysis of Struc Wavelets and Physics tural Self-Organization and Morphogenic Regulation in 0985 J. C. van den Berg, ed., “Wavelets in Physics”, Filamentous Fungal Colonies. Cambridge University Press, 1999 (a survey of many 0969) J. Lewalle, “Wavelet Transforms of the Navier applications). Stokes Equations and the Generalized Dimensions of 0986 G. Kaiser, “A Detailed Introduction to Mathemati Turbulence’ cal and Physical Wavelets” 0970 W. Hwang and S. Mallat, “Characterization of 0987) C. Best and A. Schäfer, “Variational Description of Self-Similar Multifractals with Wavelet Maxima. Statistical Field Theories using Daubechies Wavelets.” Wavelets and Communication Theory 0988 G. Beylkin, J. Dunn and D. Gines, “Order N Static 0971) J. Dill and A. R. Lindsey, “Wavelet Packet Modu and Quasi-Static Computations in Electrodynamics using lation: A Generalized Method for Orthogonally Multi Wavelets. plexed Communication.” 0989) J. C. Cohen and T. Chen, “Fundamentals of the 0972 R. Learned, H. Krim, B. Claus, A. S. Willsky, and Discrete Wavelet Transform for Seismic Data Process W. C. Karl, “Wavelet-Packet-Based Multiple Access ing.” Communication.” 0990 A. Fournier, “Wavelet Analysis of Observed Geo 0973 A. R. Lindsey, “Multidimensional Signaling via potential and Wind: Blocking and Local Energy Coupling Wavelet Packets. Across Scales.” 0974 A. R. Lindsey, Generalized Orthogonally Multi 0991 A. Fournier, “Wavelet Multiresolution Analysis of plexed Communication via Wavelet Packet Bases, chapter Numerically Simulated 3d Radiative Convection.” 1, chapter 2, chapter 3, chapter 4, chapter 5, chapter 6. 0992 A. Fournier, “Wavelet Representation of Lower Also with appendix and references. Atmospheric Long Nonlinear Wave Dynamics, Governed Wavelets and Computer Graphics by the Benjamin-Davis-Ono-Burgers Equation.” 0975) M. Cohen, S. Gortler, P. Hanrahan and P. Schröder, 0993 F. Herrmann, “A scaling medium representation, a “Wavelet Radiosity.” With FIG. 12 and FIG. 14. discussion on well-logs, fractals and waves.” An abstract is also available. 0976 M. Cohen, S. Gortler, P. Hanrahan and P. Schröder, “Wavelet Projections for Radiosity.” 0994 I. Pierce and L. Watkins, “Modelling optical pulse propagation in nonlinear media using wavelets.” 0977 A. Dreger, M. H. Gross R. Koch and L. Lippert, “A New Method to Approximate the Volume Rendering 0995 W. C. Shann, “Finite Element Methods for Max Equation using Wavelet Bases and Piecewise Polynomi well's Equations with Stationary Magnetic Fields and als, with FIGS. 5-6, FIGS. 7-10, and FIGS. 11-13. Also Galerkin-wavelet Methods for Two Point Boundary Value abstract available. Technical Report No. 220, Computer Problems, with separate abstract and table of contents. Science Department, ETH Zürich, 1994. 0996 L. R. Watkins and Y. R. Zhou, “Modelling Propa 0978 A. Fournier, “Wavelets and their Applications in gation in Optical Fibres using Wavelets.” Computer Graphics.” This is 2.5 MB compressed. 0997 R. O. Wells, “Adaptive Wave Propogation Model 0979 M. H. Gross and L. Lippert, “Fast Wavelet Based ling.” Volume Rendering by Accumulation of Tranparent Tex 0998 L. Zubair, “Studies in Turbulence using Wavelet ture Maps. With FIG. 6, FIG. 7, FIG. 8, FIG. 9, FIG. Transforms for Data Compression and Scale-Separation.” 10 and FIG. 11. Also abstract available. Technical Report No. 228, Computer Science Department, ETH Zürich, 0999 A. Fournier, “An introduction to orthonormal 1995. wavelet analysis with shift invariance: Application to US 2006/0167784 A1 Jul. 27, 2006 39
observed atmospheric-blocking spatial structure', to 1023 Basu and Weibul, strategy subsets closed under appear in J. Atmos. Sci., Dec. 1, 2000. rational behavior. E. Letters 1991 1000 A. Fournier, "Atmospheric energetics in the wave 1024 Larry Samuelson, Dominated strategies and com let domain I: Governing equations and interpretation for mon knowledge. GEB 1992 idealized flows', submitted to J. Atmos. Sci., 2000 1025 Marx and Swinkels, Order independence for iter (revised). ated weak dominance. GEB 1997 1001 A. Fournier, "Atmospheric energetics in the wave let domain II: Time-averaged observed atmospheric Equilibrium: Nash, Refinements, Correlated blocking, submitted to J. Atmos. Sci., 1999. 1026 Selten, Reexamination of the Perfectness concept for equilibrium points in extensive form games. IJGT 1002 A. Fournier, "Atmospheric energetics in the wave 1975. let domain III: Instantaneous transfers between block and local eddies', submitted to J. Atmos. Sci., 1999. 1027 Myerson, Refinements of the Nash equilibrium Hardware and Software Implementation of Wavelet Trans concept. IJGT 1975. forms 1028 Kalai and Samet, Persistent equilibria in strategic 1003 J. Fridman and E. S. Manolakos, “On Linear games. IJGT 1984. Space-Time Mapping for the 1-D Discrete Wavelet Trans 1029 Kohlberg and Mertens. On the strategic stability of form.’ Equilibria. Econometrica, 1986. 1004 J. Fridman and E. S. Manolakos, “Distributed 1030 Aumann, Correlated equilibria as an expression of Memory and Control VLSI Architectures for the 1-D baysian rationality. Econometrica, 1987. Discrete Wavelet Transform.’ 1031 Aumann and Brandenberger, Espitemic conditions 1005 J. Lu, “Computation of 2-D Wavelet Transform on for equilibrium. EMA 1995 the Massively Parallel Computer for Image Processing.” 1032 Hal Varian. A model of Sales. AER 1980 1006 MathSoft Engineering & Education, Inc., “Wave lets Extension Pack', 1999. The Extensive Form Games with Perfect Information 1007 O. Nielsen and M. Hegland, “A Scalable Parallel 1033 Rubinstein, On the interpretation of game theory. 2D Wavelet Transform Algorithm.” Econometrica 1991 1034 Reny, Common beleifs and the theory of games GAME THEORY REFERENCES with perfect information. JET 1993. 1008 GEB: Games and Economic Behavior 1035 Aumann Backward induction and common knowl 1009 EMA: Econometrica edge of rationality. GEB 1995 1036 Binmore. A note on backward induction: Aumann, 1010 JET: Journal of Economic Theory Reply to Binmore. GEB 1996 1011 IJGT: International Journal of Game Theory 1037) Selten, A Reexamination of the perfectness 1012 AER: American Economic Review Hyperbolic Discounting 1013 QJE: Quarterly Journal of Economics 1038 O'Donoghue and Rabin, Doing it now or doing it 1014) JPE: Journal of Political Economy later. AER 1999 1015 REStud: Review of Economic Studies 1039 David Laibson, Golden Eggs and Hyperbolic Dis counting. QJE 1997 Description of Games The Economics of Altruism 1016 Roger Myerson, Nash Equilibrium and the History of Economic Theory. JEL 1999 1040 Gary Becker. A theory of social interactions. JPE 1974 Rationality, Dominance, Weak Dominance etc 1041 Ted Bergstrom, A fresh look at the rotten kid 1017 Douglas Bernheim, Rationalizable strategic behav theorem and other household mysteries. JPE 1989 ior. EMA 1984 1042 Bernheim and Stark, Altruism within the family 1018) David Pearce, Rationalizable strategic behavior reconsidered: do nice guys finish last. AER 1988 and the problem of perfection. EMA 1984 1043 Lindbeck and Weibull. Altruism and time consis 1019 Douglas Bernheim, Axiomatic characterization of tency: the economics of fait accompli. JPE 1988 rational choice in Strategic enviroments. 1044 Bruce and Waldman, Transfers in kind: why they 1020 Scand. J. of E. 1986 can be efficient and non-paternalistic. AER 1991 1021 Shimoji and Watson, Conditional Dominance, 1045 Jack Robles, Paternal altruism or smart parent rationalizability and game forms. JET 1998 altruism? CU WP 98-10 1022 David Roth, Rationalizable predatory pricing. JET 1046 Mathew Rabin, Incorperating fairness into eco 1996 nomics and game theory. AER 1993 US 2006/0167784 A1 Jul. 27, 2006 40
1047 Ray and Ueda, Egalitarianism and incentives. JET 1069 Geir Asheim, Extending renegotiation proofness to 1996 infinite horizon games. GEB 1991 1048 Bernheim, Shleifer adn Summers. The strategic 1070 Bernheim and Ray, Collective dynamic consis bequest motive. JPE 1985 tency in repeated games. GEB 1989 Extensive Form Games without Perfect Information 1071 Farrel and Maskin, Renegotiation in Repeated 1049 Kreps and Wilson, Sequential Equilibrium. Econo Games. GEB 1989 metrica, 1983. Cooperative Game Theory 1050 Van Damme, Stable Equilibria and forward induc 1072 Freidman, Game theory with applications to eco tion. JET 1989. nomics chapter 6 and 7 Strategic Information Transmision 1073 Nash, The Bargaining problem. EMA 1950 1051 Crawford and Sobel, Strategic information trans 1074 Kalai and Smordinski. Other Solutions to Nash's mission. Econometrica 1982. problem. EMA 1975 1052 Cho and Kreps, Signalling games and stable equi Noncooperative Bargaining libria. QJE 1987 1075 Rubinstein, Perfect equilibrium in a bargaining 1053) Mailath, Okuno-Fujiwara and Postlewaite, Beleif model. EMA 1982 based refinements in signalling games. JET 1993 1076 Joel Watson, Alternating offer bargaining with two 1054 Milgrom and Roberts, Limit pricing and entry sided incomplete information. REStud 1999 under incomplete information: an equilibrium analysis. Reputation EMA 1982 (pages 443-459) 1077 Kreps, Milgrom, Roberts and Wilson, Reputation 1055 Cho and Sobel, Strategic Stability and uniqueness and imperfect information: predation, reputation and in signalling games. JET 1990. entry deterence: rational cooperation in the finitely 1056 Farrell, Meaning and credibility in cheap talk repeated prisoner's dilemma. JET 1981 games. GEB 1078 Aumann and Sorin, Cooperation and Bounded 1057 Milgrom and Roberts, Limit pricing and entry recal. GEB 1989 under incomplete information, an equilibrium analysis, 1079 Klaus Schmidt, Reputation and equilibrium char EMA 1982 acterization in repeated games with conflicting interests, 1058 Milgrom, Good news and bad news, representation Econometrica 1993, 325-352 and applications, Rand. 1080 Cripps and Thomas, Reputation and Commitment Folk Theorems for Repeated Games in Two person games without discounting, EMA, 1995, 1059. Dilip Abreu. On the theory of infinitely repeated 1401-1420 games with Discounting. Econometrica 1988 1081 Joel Watson. A reputation refinement withough 1060 Benoit and Krishna. Finitely Repeated games. equilibrium, EMA 1993, 199-206 Econometrica, 1985. 1082) Celentani, Fudenberg and Levine, Maintaining a 1061 James Friedman. A noncooperative equilibrium for reputation against a long lived opponent EMA 1996, supergames. REStud 1971. 691-704 1062) James Friedman. Cooperative equilibria in finite Evolutionary Game Theory horizon supergames. JET 1985. 1083 Vince Crawford, An Evolutionary interpretation of 1063 Fudenberg and Maskins. The Folk Theorem in VHBB’s experimental results on coordination. GEB 1991 repeated gmaes with discounting or with incomplete 1084 Gilboa and Matsui, Social Stability and Equilib information. Econometrica 1986. rium, EMA 1991 1064 Roy Radner. Collusive Behavior in non-coopera 1085 Kandori, Mailath and Rob, Learning, Mutation, tive epsilon equilibria in oligopolies with long but finite and Long Run Equilibria in games, EMA 1993. lives. JET 1980. 1086 Peyton Young. An Evolutionary Model of Bargain 1065 Ariel Rubinstein. Equilibrium in supergames with ing, JET 1993 the overtaking criterion. JET 1977. 1087 Peyton Young, The Evolution of Conventions, Renegotiation EMA 1993 1066 Benoit and Krisna, Renegotiation in finitely 1088 Larry Samuelson, Stochastic Stability with alter repeated games. EMA 1993 native best replies. JET 1067 Bergin and MacCleod, Efficiency adn renegotiation 1089) Noldeke and Samuelson, The Evolution of Back in repeated games. JET 1993 wards and Forwards Induction, GEB 1993 1068 Andreas Blume, Interplay communication in 1090) Jack Robles, An Evolutionary Folk Theorem For repeated games. GEB 1994 Finitely Repeated Games CU WP 99– US 2006/0167784 A1 Jul. 27, 2006
1091 Kim and Sobel, An Evolutionary Approach to 1114 Ulvila J. W. & R. Brown, Decision analysis comes Preplay Communication EMA 1995 of age, Harvard Buisness Review 1982. General Game Theory 1115 Howard R. A., Decision analysis: practice and promise, Management Science, 1988. 1092 Bierman H. S. & Fernandez L., GameTheory with Economic Applications. Addison-Wesley, 1993. 1116 Clemen R. T., Making Hard Decisions: An Intro duction to Decision Analysis, Belmont, Calif.: Duxbury, 1093) Dixit A., & Nalebuff B. Thinking Strategically: the 1996. Competitive Edge in Business, Politics, and Everyday Life, New York: Norton, 1991. 1117 Samson D., Chs 2-6, 11, Managerial Decision Analysis, Chicago: R. D. Irwin, 1988. 1094) McMillan J. Games, Strategies, and Managers, Oxford: OUP, 1992. Strategic Moves 1095) Baird D. G., Gertner R. H., and Picker R. C., Game 1118) Dixit & Nalebuff, Ch5 Strategic moves. Theory and the Law, Harvard U.P., 1994. 1119 Brams S. J. & J. M. Togman, Cooperation through 1096 Rasmusen E. Games and Information: An Intro threats: the Northern Ireland case, PS: Political Science & duction to Game Theory, Oxford: B. Blackwell, 2nd Politics, March 1998. edition, 1994. 1120 Gardner, Ch4 n-person games, Ch5 Non-coopera 1097 Ghemawat P. Games Businesses Play: Cases and tive games. Models, New York: Wiley, 1995. 1121 Colman A. M., Ch8 Multi-person games: social 1098 Gardner R. Games for Business and Economics, dilemmas, in his Game Theory and Experimental Games, New York: Wiley, 1995. Oxford: Pergamon, 1982. 1122 Kay J., Ch5 Co-operation and Co-ordination, in his Strategic Decision Making Foundations of Corporate Success: How Business Strat 1099 Dixit & Nalebuff, Intro; Ch2 Anticipating your egies Add Value, Oxford: OUP, 1993. rival's response; 1123 Brams S. J., Chl International relations games, in 1100 Ch;3 Seeing through your rival's response. Game Theory and Politics, NY: Macmillan, 1975. 1101 Barnett, F. W. Making game theory work in prac Credible Commitment tice, Wall Street Journal, 1995. 1124) Dixit & Nalebuff, Chô Credible commitments. 1102 Bierman & Fernandez, Ch5 Nash equilibrium I, 1125 Bierman & Fernandez, Ch23 Subgame-perfect Ch11 Nash equilibrium II equilibrium 1103 O'Neill B., International escalation and the dollar 1126 Rasmusen, Ch4.1 Subgame perfection. auction, Journal of Conflict Resolution, 1986. 1127 Gardner, Chó Credibility and subgame perfection. 1104 Schelling T. C., Ch7 Hockey helmets, daylight saving, and other binary choices, in his Micromotives and 1128 Ghemawat, Ch3 Preemptive capacity expansion in Macrobehavior, NY: Norton, 1978. the titanium dioxide industry. 1105 Marks R. E. Competition and common property, Repetition and Reputation 1998. 1129 Dixit & Nalebuff, Ch4 Resolving the Prisoner's 1106 McMillan J., Ch3 Understanding cooperation and Dilemma; Ch9 Cooperation and coordination. conflict. 1130 Nowak, M., R. May, & K. Sigmund, The arithmetic 1107 McAfee R. P. & J. McMillan, Competition and of mutual help, Scientific American, 1995 game theory, Journal of Marketing Research, 1996. 1131 Hofstadter D., Ch29 The Prisoner's Dilemma com puter tournaments and the evolution of cooperation, in his 1108 Baird, Gertner, & Picker, Ch1 Simultaneous deci Metamagical Themas, Penguin, 1985. sion-making and the normal form game. 1132 Marks R. E., Midgley F D. F., & Cooper L. G., 1109 Gardner, Chl Introduction, Ch2 Two-person Adaptive behaviour in an oligopoly, in Evolutionary games, Ch16 Voting games. Algorithms in Management Applications, ed. by J. 1110 Rasmusen, Chl The rules of the game. Biethahn & V. Nissen, (Berlin: Springer-Verlag), 1995. 1111 Schelling T. C., What is game theory? in his Choice 1133 Baird Gertner & Picker, Ch2 Dynamic interaction and Consequence: Perspectives of an Errant Economist, and the extensive-form game, Ch5 Reputation and Camb.: Harvard UP, 1980. repeated games. Decision Analysis—Games Against Nature 1134 Gardner, Ch7 Repeated games, Ch8 Evolutionary stability and bounded rationality. 1112 Apocalpse maybe, and An insurer’s worst night 1135 Rasmusen, Ch4 Dynamic games and symmetric mare, The Economist, 1995/96 information, Ch5 Reputation and repeated games with 1113 Bierman & Fernandez, Chs 1-3. symmetric information. US 2006/0167784 A1 Jul. 27, 2006 42
Unpredictability 1162 McAfee R. P. & J. McMillan, Analyzing the air 1136) Dixit & Nalebuff, Ch7 Unpredictability; Ch8 waves auction, Journal of Economic Perspectives, 1996 Brinkmanship. 1163) R. Marks, Closed tender vs. open bidding auctions, 22 Dec. 1994. 1137 Bierman & Fernandez, Ch11.9 1164. The Economist, Secrets and the prize, 12 Oct. 1138 Gardner, Ch3 Mixed strategies. 1996, p.98. 1139 Rasmusen, Ch3 Mixed and continuous strategies. 1165 Scientific American, Making honesty pay, January Bargaining 1997, p. 13. 1140 Dixit & Nalebuff, Ch10 The voting strategy; Chl 1 1166 Gardner, Ch11 Auctions. Bargaining. 1167 Brams S. J. & A. D. Taylor, Fair division by 1141. McMillan, Ch5 Gaining bargaining power; Chó auctions, Ch9 of their Fair Division: From Cake-Cutting Using information Strategically. to Dispute Resolution, Cambridge: CUP, 1996. 1142 Elster J., Ch14 Bargaining, in Nuts and Bolts for 1168 Rasmusen, Ch12 Auctions. the Social Sciences, Camb.: CUP, 1989 Contracting, or the Rules of the Game 1143) Murnighan J. K. Games End, Chapter 15 in his: 1169 Kay, Ch4 Relationships and contracts. Bargaining Games: A New Approach to Strategic Think ing in Negotiations, NY: William Morrow, 1992. 1170 Dixit & Nalebuff, Ch12 Incentives. 1171 McMillan, Ch8 Creating incentives; Ch9 Design 1144 Bierman & Fernandez, Chó Bargaining. ing contracts; Ch10 Setting executives’ salaries. 1145 Schelling T. C., Ch2 Essay on bargaining, in The 1172 Williamson O. E., Strategizing, economizing, and Strategy of Conflict, Camb.: Harvard UP, 1980. economic organization, Strategic Management Journal, 1146 Baird Gertner & Picker, Ch7 Noncooperative bar 1991. gaining 1173 Bierman & Fernandez, Ch7 Involuntary unemploy 1147 Gardner, Ch12 Two-person bargains. Ch14 n-per ment. Son bargaining and the core. 1174) Gardner, Ch10 Games between a principal and an 1148 Rasmusen, Ch11 Bargaining. agent. 1149 Brams S. J., Negotiation Games: Applying Game 1175 Milgrom P. & Roberts J., Ch5 Bounded rationality Theory to Bargaining and Arbitration, NY: Routledge, and private information; Chó Moral hazard and perfor 1990. mance incentives. Economics, Organization and Manage ment, Englewood Cliffs: Prentice-Hall, 1992. Using Information Strategically Choosing the Right Game: Co-opetition 1150 McMillan, Ché Using information strategically 1176 Brandenburger A. M. & B. J. Nalebuff, The right 1151 Bierman & Fernandez, Ch17 Bayesian equilibrium, game: using Game Theory to shape strategy, Harvard Ch19 Adverse selection and credit rationing Business Review, 1995 1152 Rasmusen, Ch2 Information P-13 mayet. Som.yale.edu/coopetition/index2.html 1153 Baird Gertner & Picker, Ch4 Signalling, screening, 1177 Koselka R., Businessman's dilemma, and Evolu and nonverifiable information tionary economics: nice guys don’t finish last, Forbes, 1154 Gardner, Ch9 Signaling games. Oct. 11, 1993. 1178 Brandenburger A. M. & B. J. Nalebuff, Co-opeti Bidding in Competition tion: 1. A revolutionary mindset that combines competi 1155 Revenge of the nerds. It’s only a game, and Learn tion and cooperation; 2. The Game Theory Strategy that’s ing to play the game. The Economist, 1994 changing the game of business. 1156 Landsburg S. E. Cursed winners and glum losers, 1179 New York: Currency Doubleday, 1996. Ch18 of his The Armchair Economist: 1180 Brandenburger A. M. & Harborne W. S. Jr., Value 1157) Economics and Everyday Life, New York: The based business strategy, Journal of Economics and Man Free Press, 1993. agement Strategy, 5(1), 1996. 1158 Norton, R., Winning the game of business, Fortune, 1181 Baird Gertner & Picker, Ché Collective action, 1995, embedded games, and the limits of simple models. 1159 Koselka, R., Playing poker with Craig McCaw, 1182 Morrow J. D., GameTheory for Political Scientists, Forbes, 1995, Princeton: P.U.P., 1994. 1183 Casson M., The Economics of Business Culture: 1160 Dixit & Nalebuff, Ch12 Incentives. Game Theory, Transaction Costs and Economic Perfor 1161) McMillan, Ch11 Bidding in competition mance, Oxford: OUP, 1991. US 2006/0167784 A1 Jul. 27, 2006
1184 Schelling T. C., Altruism, meanness, and other 1200 R. Buyya, H. Stockinger, J. Giddy, and D. Abram potentially strategic behaviors, American Economic son. Economic Models for Management of Resources in Review, 68(2): 229-231, May 1978. Peer-to-Peer and Grid Computing. In Proceedings of the SPIE International Symposium on The Convergence of 1185 Crawford V. P., Thomas Schelling and the analysis Information Technologies and Communications (ITCOM), of strategic behavior, in Strategy and Choice, ed. by R. J. August 2001. Zeckhauser, MIT Press, 1991. 1201 M. Castro, P. Druschel, A. Ganesh, A. Rowstron, 1186 For a history of game theory since Old Testament and D. S. Wallach. Security for structured peer-to-peer times, point your browser at the following URL: www overlay networks. In Proc. 5th Symposium on Operating ..canterbury.ac.nz/econ/hist.htm Systems Design and Implementation, Boston, Mass., 1187 For further surfing on the Net about game theory, December 2002. start at the following URLs: www.pitt.edu/~alroth/al 1202 Clarke, E. H. (1971). Multipart pricing of public roth.html goods, Public Choice 11, 17-33. 1188 Eddie Dekel, Drew Fudenberg and David K. 1203 T. Clausen and P. Jacquet. Optimized link state Levine, Learning to Play Bayesian Games (Jun. 20, routing protocol. October 2003. http://hipercom.inria.fr/olsr/ 2001). 1204 S. Corson and J. Macker. Mobile ad hoc network 1189 www.gametheory.net/html/lectures.html ing (MANET): Routing protocol performance issues and evaluation considerations. Request for comments 2501, 1190 Drew Fudenberg and David K. Levine, The Nash Internet Engineering Task Force, 1999. http://www.rfe-edi Threats Folk Theorem With Communication and tor.org/ Approximate Common Knowledge in Two Player Games (Jun. 10, 2002). 1205 J. Crowcroft, R. Gibbens, F. Kelly, and S. Ostring. Modelling incentives for collaboration in mobile ad hoc GAME THEORY AND AD HOCNETWORKS networks. In Proceedings of WiOptO3, 2003. REFERENCES 1206 Dasgupta, P. P. Hammond, and E. Maskin (1979). 1191 E. Altman, R. El Azouzi, and T. Jimenez. Slotted The implementation of social choice rules, Review of Eco aloha as a stochastic game with partial information. In nomic Studies 46, 185-216. Proceedings of WiOptO3, 2003. 1207 Deering, S., D. Estrin, D. Farinacci, V. Jacobson, C. 1192 A Archer and E Tardos. Truthful mechanisms for Liu, and L. Wei (1996). The PIM architecture for wide-area one-parameter agents. In Proc. 42nd IEEE Symp. On Foun multicast routing, ACM/IEEE Transactions on Networking dations of Computer Science, 2001. 54, 153-162. 1193 N. Ben Salem, L. Buttyan, J. P. Hubaux, and 1208 A. Demers, S. Keshav, and S. Shenker. Analysis Jakobsson M. A charging and rewarding scheme for packet and simulation of a fair queueing algorithm. In SIGCOMM forwarding. In Proceeding of Mobihoc, June 2003. 89, Proceedings of the ACM Symposium on Communica tions Architectures & Protocols, pages 1-12, 1989. 1194 L. Blazevic, L. Buttyan, S. Capkun, S. Giordiano, J.-P. Hubaux, and J.-Y. Le Boudec. Self-organization in 1209 C. Douligeris and R. Mazumdar. User optimal flow mobile ad-hoc networks: the approach of terminodes. IEEE control in an integrated environment. In Proceedings of the Communications Magazine, 39(6):166-174, June 2001. Indo-US Workshop on Signals and Systems, 1988. 1195 G. E. Bolton and A. Ockenfels. ERC: A theory of 1210 Joan Feigenbaum, Christos H Papadimitriou, and equity, reciprocity and competition. The American Eco Scott Shenker. Sharing the cost of multicast transmissions. nomic Review, 90:166-193, March 2000. Journal of Computer and System Sciences, 63:21-41, 2001. 1196 A. Bovopoulos and A. Lazar. Asynchronous itera 1211 Feigenbaum, Joan, Christos Papadimitriou, Rahul tive algorithms for optimal load balancing. In Proceedings Sami, and Scott Shenker (2002). “A BGP-based Mechanism of the 22nd Annual Conference on Information Sciences and for Lowest-Cost Routing.” In Proc. 21 st Symposium on Systems, pages 1051-1057, 1988. Principles of Distributed Computing, ACM Press, 173-182. 1197 Felix Brandt and Gerhard Wei?. Antisocial Agents 1212 J. Feigenbaum and S. Shenker. Distributed algo and Vickrey Auctions. In Pre-proceedings of the Eighth rithmic mechanism design: Recent results and future direc International Workshop on Agent Theories, Architectures, tions. In Proc. 6th Int’l Workshop on Discrete Algorithms and Languages (ATAL-2001), pages 120-132, 2001. and Methods for Mobile Computing and Communications, pages 1-13, Atlanta, Ga., September 2002. 1198 Felix Brandt. A Verifiable, Bidder-Resolved Auc tion Protocol. In Proceedings of the 5th International Work 1213 T. Fent, G. Feichtinger, and G. Tragler. A dynamic shop on Deception, Fraud and Trust in Agent Societies, game of offending and law enforcement. International Game pages 18-25, 2002. Theory Review, 4(1):71-89, 2002. 1199 S. Buchegger and J. Le Boudec. Performance 1214 Ferguson, D., C. Nikolaou, and Y. Yemini (1989). analysis of the CONFIDANT protocol. In Proceedings of An economy for flow control in computer networks, in IEEE/ACMWorkshop on Mobile Ad Hoc Networking and “Proceedings of the 8th Infocom, pp. 100-118, IEEE Com Computing (MobiFIOC), June 2002. puter Society Press, Los Alamitos. US 2006/0167784 A1 Jul. 27, 2006 44
1215 S. Floyd and K. Fall. Promoting the use of end-to 1231 Johnson, D. S., M. Minko , and S. Phillips (2000). end congestion control. IEEE/ACM Transactions on Net The prize collecting Steiner tree problem: theory and prac working, 7(4):458-472, August 1999. tice, in “Proceedings of the 11th Symposium on Discrete Algorithms, pp. 760-769, ACM Press/SIAM, New York/ 1216 E. Friedman and D. Parkes, “Pricing WiFi at Star Philadelphia. bucks—Issues in Online Mechanism Design”. In Proc. Fourth ACM Conf. on Elec. Commerce (EC03), 2003. 1232 Y. Korilis and A. Lazar. On the existence of equi Extended version at www.eecs.harvard.edu/econcs/pubs/on libria in noncooperative optimal flow control. Journal of the line.pdf. citeseer.nj.nec.com/article/friedman03pricing.html ACM, 42(3):584-613, 1995. 1217) Friedman, E., and S. Shenker (1997). “Learning 1233 Korilis, Y., A. A. Lazar, and A. Orda (1995). and Implementation in the Internet,” preprint.www.aciri.org/ Architecting noncooperative networks, Journal on Selected shenker/decent.ps Areas in Communications 13, 1241-1251. 1218 A. C. Fuqua, T-W Ngan, and D. S. Wallach, “Eco 1234 R. La and V. Anantharam. Charge-sensitive tcp and nomic Behavior of Peer-to-Peer Storage Networks, Work rate control in the internet. In Proceedings of INFOCOM shop on Economics of Peer-to-Peer Systems, Berkeley, Calif., 2003, citeseer.nj.nec.com/fuquao3economic.html 2000, 2000. 1219 Goldberg, A. V., J. D. Hartline, and A. Wright 1235 R. La and V. Anantharam. Optimal routing control: (1999). “Competitive Auctions and Digital Goods.” Inter Repeated game approach. IEEE Transactions on Automatic Trust Technical Report 99-01. Available at http://www.in Control, 47(3):437-450, 2002. tertrust.com/star/tr/tr-99-01.html 1236 Ron Lavi and Noam Nisan. Competitive analysis 1220 A. Goldsmith and S. Wicker. Design challenges for of incentive compatible on-line auctions. In Proc. 2nd ACM energy-constrained ad hoc wireless networks. IEEE Wire Conf. on Electronic Commerce (EC-00), 2000. less Communications, 9(4):8-27, 2002. 1237. S.-J. Lee, W. Su, J. Hsu, M. Gerla, and R. Bagrodia. 1221 P. Golle, K. Leyton-Brown, I. Mironov, and M. A performance comparison study of ad hoc wireless multi Lillibridge. Incentives for sharing in peer-to-peer networks. cast protocols. In Proceedings of IEEE INFOCOM 2000, In Proc. 3rd ACM Conf. on Electronic Commerce, Tampa, pages 565-574, March 2000. Fla., October 2001. 1238 A. Legout and E. W. Biersack. Revisiting the fair 1222 Z. Haas. A new routing protocol for reconfigurable queueing paradigm for end-to-end congestion control. IEEE wireless networks. In IEEE 6th International Conference on Network Magazine, 16(5):3846, September 2002. Universal Communications Record, Volume 2, pages 562 1239 A. B. MacKenzie and S. B. Wicker. Selfish users in 566, October 1997. aloha: A game-theoretic approach. In Vehicular Technology 1223 Garrett Hardin. The Tragedy of the Commons. Conference, 2001. VTC 2001 Fall. IEEE VTS 54th, volume Science, 162:1243-1248, 1968. Alternate Location: http:// 3, October 2001. diedff.com/page95.htm. 1240 J. MacKie-Mason and H. Varian. Pricing congest 1224 Michael Harkavy, J. D. Tygar, and Hiroaki Kiku ible network resources. IEEE Journal on Selected Areas in chi. Electronic Auctions with Private Bids. In 3rd USENIX Communications, 13(7): 1141-1149, 1995. Workshop on Electronic Commerce, pages 61-74, Septem ber 1998. 1241 P. Marbach. Priority service and max-min fairness. In Proceedings of IEEE INFOCOM 2002, volume 1, pages 1225 X. Hong, K. Xu, and M. Gerla. Scalable routing 266-275, 2002. protocols for mobile ad hoc networks. IEEE Networks, 16(4):11-21, July 2002. 1242 S. Marti, T. J. Giuli, K. Lai, and M. Baker. Miti 1226 M.-T. Hsiao and A. Lazar. Optimal decentralized gating routing misbehavior in mobile ad hoc networks. In flow control of markovian queueing networks with multiple Proceedings of MOBICOM 2000, August 2000. controllers. Performance Evaluation, 13(3):181-204, 1991. 1243 K. Mase, Y. Wada, N. Mori, K. Nakano, M. Sen 1227 Jackson, Matthew O., and Asher Wolinsky (1996). goku, and S. Shinoda. Flooding schemes for a universal ad “A Strategic Model of Social and Economic Networks.’ hoc network. In Industrial Electronics Society, 2000. Journal of Economic Theory 71, 44-74. IECON 2000., volume 2, pages 1129-1134, 2000. 1228 M. Jakobsson, J. P. Hubaux, and L. Huttyan. A 1244 R. Mazumdar, L. Mason, and C. Douligeris. Fair micro-payment Scheme encouraging collaboration in multi ness in network optimal flow control: Optimality of product hop cellular networks. In Proceedings of Financial Crypto forms. IEEE Transactions on Communications, 39(5): 775 2003, January 2003. 782, 1991. 1229) Y. Jin and G. Kesidis. Equilibria of a noncoopera 1245 P. Michiardi and R. Molva. Core: A collaborative tive game for heterogenous users of an ALOHA network. reputation mechanism to enforce node cooperation in mobile IEEE Communications Letters, 6(7):282-284, July 2002. ad hoc networks. In Communication and Multimedia Secu 1230 D. Johnson, D. Maltz, and Y-C. Hu. The dynamic rity 2002 Conference, 2002. Source routing protocol for mobile ad hoc networks, April 1246 P. Michiardi and R. Molva. Game theoretic analy 2003. http://www.ietforg/internet-drafts/draft-ietf-manet sis of security in mobile ad hoc networks. Technical Report dSr-09.txt RR-02-070, Institut Eurécom, 2002. US 2006/0167784 A1 Jul. 27, 2006
1247 P. Michiardi and R. Molva. A game theoretical 1264 C. Papadimitriou. Algorithms, games, and the approach to evaluate cooperation enforcement mechanisms internet. In Proc. 33rd ACM Symposium on Theory of in mobile ad hoc networks. In Proceedings of WiOptO3, Computing, pages 1-5. Hersonissos, Crete, Greece, July March 2003. 2001. 1248 Dov Monderer and Moshe Tennenholtz. Distrib 1265 D. C. Parkes. Iterative Combinatorial Auctions: uted Games: From Mechanisms to Protocols. In Proceedings Achieving Economic and Computational Efficiency (Chap of the 16th National Conference on Artificial Intelligence ter 2). PhD thesis, Univesity of Pennsylvania, May 2001. (AAAI), pages 32-37, 1999. http://www.eecs.harvard.edu/-parkes/pubs/ch2.ps. 1249 Moulin, H. (1999). Incremental cost sharing; char 1266 W. Peng, X.-C. Lu. On the reduction of broadcast acterization by strategy proofness, Social Choice and Wel redundancy in mobile ad hoc networks. In Mobile and Ad fare 16, 279-320. Hoc Networking and Computing, 2000. MobiFIOC., pages 1250 Moulin, H. and S. Shenker (1997). Strategyproof 129-130, 2000. Sharing of Submodular Costs: Budget Balance Versus Effi 1267 C. Perkins, E. Belding-Royer, and S. Das. Ad hoc ciency, to appear in Economic Theory. www.aciri.org/shen on-demand distance vector (AODV) routing. Request for ker/cost.ps comments 3561, Internet Engineering Task Force, 2003. 1251) Moulin, Hervé, and Scott Shenker (2001). “Strat 1268 C. E. Perkins, editor. Ad Hoc Networking. Addi egyproof Sharing of Submodular Costs: Budget Balance son-Wesley, Boston, 2001. versus Efficiency.” Economic Theory 18, 511-533. 1269 Adrian Perrig, Sean Smith, Dawn Song, and J. 1252) J. B. Nagle. On packet switches with infinite stor Doug Tygar. SAM: A Flexible and Secure Auction Archi age. IEEE Transactions on Communications, 35(4):435-438, tecture Using Trusted Hardware, 1991. citeseer.nj.nec.com/ April 1987. perrig91 Sam.html 1253 J. F. Nash. Non-cooperative games. Annals of 1270 V. Rodoplu and H.-Y. Meng. Minimum energy Mathematics, 54(2):286-295, September 1951. mobile wireless networks. IEEE Journal on Selected Areas 1254 T.-W. J. Ngan, D. S. Wallach, and P. Druschel. in Communications, 17(8): 1333-1344, August 1999. Enforcing fair sharing of peer-to-peer resources. In Proc. 1271) A. Rowstron and P. Druschel. Pastry: Scalable, 2nd Int’l Workshop on Peer-to-Peer Systems, Berkeley, distributed object address and routing for large-scale peer Calif., February 2003. to-peer systems. In Proc. IFIP/ACM Int’l Conf. on Distrib 1255 Noam Nisan and Amir Ronen. Algorithmic mecha uted Systems Platforms, pages 329-350, Heidelberg, Ger nism design. Games and Economic Behavior, 35:166-196, many, November 2001. 2001. 1272 A. Rowstron and P. Druschel. Storage management 1256 Nisan, N. (1999). Algorithms for selfish agents, in and caching in PAST, a large-scale, persistent peer-to-peer “Proceedings of the 16th Symposium on Theoretical Aspects storage utility. In Proc. 18th ACM Symposium on Operating of Computer Science, pp. 1-17, Springer-Verlag, Berlin. Systems Principles, pages 188-201, Chateau Lake Louise, Lecture Notes in Computer Science, Vol.1563. Banff, Canada, October 2001. 1257 Nisan, N. and A. Ronen (2000). Computationally 1273 E. Royer and C.-K. Toh. A review of current Feasible VCG Mechanisms, to be presented at “Games routing protocols for ad hoc mobile wireless networks. IEEE 2000.’ http:/www.cs.huji.ac.il/-noam/vcgbased.ps Personal Communications, 6(2):46-55, April 1999. 1258 Noam Nisan and Amir Ronen. Algorithmic Mecha 1274 T. Sandholm. Distributed rational decision making. nism Design. In Proceedings of the 31st ACM Symposium In G. Wei?, editor, Multiagent Systems: A Modern Approach on Theory of Computing, pages 129-140, 1999. to Distributed Artificial Intelligence, chapter 5. The MIT 1259) Osborne, M.-J. and A. Rubinstein (1994). “A Press, 1999. Course in Game Theory, MIT Press, Cambridge Mass. 1275 Tuomas Sandholm. Limitations of the Vickrey 1260 R. Ogier, F. Templin, and M. Lewis. Topology Auction in Computational Multiagent Systems. In Proceed dissemination based on reversepath forwarding, October ings of the 2nd International Conference on Multi-Agent 2003. Vesuvio.ipv6.cselt.it/internet-drafts/draft-ietf-manet Systems (ICMAS). AAAI Press, 1996. Menlo Park, Calif. thrpf-11.txt 1276 S. Shenker. Making greed work in networks: A 1261 E. Ogston and S. Vassiliadis. A Peer-to-Peer Agent game-theoretic analysis of switch service disciplines. IEEE/ Auction. In Proceedings of the First International Joint ACM Transactions on Networking, 3(6):819-831, Decem Conference on Autonomous Agents and Multi-Agent Sys ber 1995. tems (AAMAS), 2002. 1277 S. Singh, M. Woo, and C. S. Raghavendra. Power 1262 A. Orda, R. Rom, and N. Shimkin. Competitive aware routing in mobile ad hoc networks. In Proceeding of routing in multi-user communication networks. IEEE/ACM MOBICOM 1998, pages 181-190, 1998. Transactions on Networking, 1(5):510-521, October 1993. 1278 Shapley, L. S. (1953). A value for n-person games, 1263 M. J. Osborne and A. Rubinstein. A Course in in “Contributions to the Theory of Games, pp. 31-40, Game Theory. MIT Press, Cambridge, 1994. Princeton University Press, Princeton. US 2006/0167784 A1 Jul. 27, 2006 46
1279 Sharman Networks. Kazaa Guide: Supernode 1293 Aslan Tchamkerten Tchamkerten: 03 On the Dis FAQ, 2003. http://www.kazaa.com/us/help/facq/supern creteness of Capacity-Achieving Distributions IEEE Trans odes.htm. actions on Information Theory 1280 J. Shneidman and D. Parkes, “Rationality and 1294 Stephan ten Brink, Aslan Tchamkerten BrinkT:03 Self-Interest in Peer to Peer Networks'. In Proc. 2nd Int. Capacity of the Binary Input Rayleigh Channel with Perfect Workshop on Peer-to-Peer Systems (IPTPS'03), 2003, cite Side Information IEEE Transactions on Communications seer.nj.nec.com/shneidman03rationality.html 1295 L. Buttyan, J.-P. Hubaux and S. Capkun Butty 1281 Jeffrey Shneidman and David C. Parkes. Using anHC: 03 A Formal Model of Rational Exchange and its Redundancy to Improve Robustness of Distributed Mecha Application to the Analysis of Syverson’s Protocol Journal nism Implementations, 2003. Working Paper. Poster version of Computer Security to appear at ACM Conference on Electronic Commerce 1296 Marc Heissenblittel, Torsten Braun, Thomas Ber ECO3. noulli, and Markus Waelchli HeissenbuettelB:04a). BLR: 1282 V. Srinivasan, P. Nuggehalli, C. Chiasserini, and R. Beacon-Less Routing Algorithm for Mobile Ad-Hoc Net Rao. Cooperation in wireless ad hoc networks. In Proceed works Elsevier's Computer Communications Journal ings of INFOCOM 2003, volume 2, pages 808–817, 2003. 1297 Olivier Dousse, Francois Baccelli, Patrick Thiran 1283 A. Urpi, M. Bonuccelli, and S. Giordano. Model DousseBT:03a Impact of Interferences on Connectivity in ling cooperation in mobile ad hoc networks: a formal Ad Hoc Networks IEEE/ACM Transactions on Networking description of selfishness. In Proceedings ofWiOptO3, 1298 Mario Cagal, Jean-Pierre Hubaux and Christian March 2003. Enz, Cagal HE:03 Energy-efficient Broadcasting in All 1284 A. van den Nouweland, P. Borm, W. van Golstein wireless Networks ACM Mobile Networks and Applications Brouwers, R. Groot Bruinderink, and S. Tijs. A game (MONET) theoretic approach to problems in telecommunication. Man 1299) Irena Maravic, Martin Vetterli MaravicV:04a) agement Science, 42(2):294-303, February 1996. Exact Sampling Results for Some Classes of Parametric Non-Bandlimited 2-D Signals IEEE Transactions on Signal 1285 H. R. Varian. Mechanism design for computerized Processing, Volume 52, Issue 1 January 2004 agents. In Proc. First Usenix Workshop on Electronic Com merce, New York, N.Y. 1995 1300 Christian Plessl, Rolf Enzler, Herbert Walder, Jan Beutel, Marco Platzner, Lothar Thiele and Gerhard Troester 1286 Vickrey, W. (1961). Counterspeculation, auctions, PlesslEWBPTT:03) The Case for Reconfigurable Hardware and competitive sealed tenders, Journal of Finance 16, 8-37. in Wearable Computing Personal and Ubiquitous Comput 1287 Walsh, W. and M. Wellman (1998). A market ing, Springer-Verlag, Vol. 7, No. 5, pages 299-308, October protocol for decentralized task allocation, in “Proceedings of 2003 the Third International Conference on Multi-Agent Sys 1301 L. Buttyan, J P Hubaux ButtyanH:03a Stimulat tems, pp. 325-332, IEEE Computer Society Press, Los ing Cooperation in Self-Organizing Mobile Ad Hoc Net Alamitos. works ACM/Kluwer Mobile Networks and Applications, 1288) J. E. Wieselthier, G. D. Nguyen, and A. Ephre Vol. 8, No. 5, October 2003 mides. Algorithms for energy-efficient multicasting in static 1302 Karl Aberer, Philippe Cudré-Mauroux, Anwitaman ad hoc wireless networks. Mobile Networks and Applica Datta, Zoran Despotovic, Manfred Hauswirth, Magdalena tions, 6:251-263, 2001. Punceva, Roman Schmidt AbererCDDHPS:03 P-Grid: A Self-organizing Structured P2P System ACM SIGMOD 1289 C. Jason Woodard and David C. Parkes, 1st Work shop on the Economics of P2P systems, Strategyproof Record, 32(3), September 2003 Mechanisms for Ad Hoc Network Formation, 2003, citese 1303 Olivier Dousse, Patrick Thiran DousseT:03 er.nj.nec.com/woodard03strategy proof.html, www.sims Physical connectivity of self-organized ad hoc wireless .berkeley.edu/research/conferences/p2pecon/paperS/S6 networks IEEE Intelligent systems, Vol. 18, Iss. 4, July woodard.pdf August 2003 1290 Z. Zhang and C. Doulligeris. Convergence of syn 1304 Frank Siegemund, Michael Rohs SiegemundR:03 chronous and asynchronous greedy algorithms in a multi Rendezvous Layer Protocols for Bluetooth-Enabled Smart class telecommunications environment. IEEE Transactions Devices (extended version) Personal and Ubiquitous Com on Communications, 40(8): 1277-1281, 1992. puting Journal (PUC), vol. 7, nr. 2, July 2003 1291 Urs Anliker, Jan Beutel, Matthias Dyer, Rolf 1305 Benoit Garbinato, Philippe Rupp GarbinatoR:03b Enzler, Paul Lukowicz, Lothar Thiele, Gerhard Troester From Ad Hoc Networks to Ad Hoc Applications ERCIM AnlikerBDELTT:03 A Systematic Approach to the Design News, no 54, July 2003 of Distributed Wearable Systems IEEE Transactions on 1306 M. Gastpar, B. Rimoldi and M. Vetterli Gast Computers parrV:03 To code or not to code: Lossy source-channel 1292 Samuel Bendahan, Giovanni Camponovo, Yves communication revisited IEEE Transactions on Information Pigneur BendahanCP:03 Multi-issue actor analysis: tools Theory, 49(5): 1147-1158, May 2003 and models for assessing technology environments Journal 1307 Karl Aberer, Philippe Cudre-Mauroux, Anwitaman of Decision Systems Datta, Zoran Despotovic, Manfred Hauswirth, Magdalena US 2006/0167784 A1 Jul. 27, 2006 47
Punceva, Roman Schmidt, Jie Wu AbererCDDHPSW:03 1320) M. Vetterli, P. Marziliano, T. Blu VetterliMB:02 Advanced Peer-to-Peer Networking: The P-Grid System and Sampling Signals with Finite Rate of Innovation IEEE its Applications PIK Journal 2/2003: Special Issue on Peer Transactions on Signal Processing, 50(6), 1417-1428, June to-Peer Systems, April-June 2003 2002 1308 Samaijit Chakraborty, Simon Künzli, Lothar 1321 Q. Li, B. Rimoldi, and M. K. Simon LiRS:02 Thiele, Andreas Herkersdorf, Patricia Sagmeister Bandwidth-Efficient Constant-Energy Trellis-Coded Modu ChakrabortyKTHS:03 Performance Evaluation of Net lation Schemes with Prescribed Decoding Delay IEEE work Processor Architectures: Combining Simulation with Trans. on Information Theory, Vol. 48, Number 5, May 2002 Analytical Estimation Computer Networks, special issue on 1322 L. Blazevic, S. Giordano, J.-Y. Le Boudec Network Processors, Volume 41, Issue 5, pages 641-665, BlazevicOL:02a Self Organized Terminode Routing Clus Elsevier Science, April 2003 ter Computing Journal, Vol.5, No.2, April 2002 1309 Srdjan Capkun, Levente Buttyan and Jean-Pierre 1323 Mark Heitmann, Peter Aschmoneit Hubaux CapkunBH:03. Self-Organized Public-Key Man HeitmannA:02a Customer Centred Community Applica agement for Mobile Ad Hoc Networks IEEE Transactions on tion Design International Journal on Media Management Mobile Computing, Vol. 2, No. 1, January-March 2003 (4:1. Spring 2002) 1310) D. Tuninetti and G. Caire TuninettiG:03) The Long-term average capacity region per unit cost with appli 1324 S. Capkun, M. Hamdi, J. P. Hubaux Cap cation to protocol for sensor networks European Transac kunHH:02 GPS-free Positioning in Mobile Ad-Hoc Net tions on Telecommunications. Special Issue on Selected works In Cluster Computing Journal, April 2002, Vol. 5, No. Papers from the Conference European Wireless 2002. ETT 2 Vol 14, No. 1, January-February 2003 1325 Magaly Dubosson, Alexander Osterwalder, Yves 1311 L. Buttyan and J.-P. Hubaux (eds.), G. Avoine, S. Pigneur DubossonOP:02 eBusiness Model Design, Clas Buchegger, S. Capkun, J. Y. Le Boudec, S. Vaudenay et al. sification and Measurements Thunderbird International Buttyan H:03b Report on a Working Session on Security in Business Review, January 2002, vol. 44, no. 1:5-23 Wireless Ad Hoc Networks ACM Mobile Computing and 1326 K. Aberer, M. Punceva, M. Hauswirth, R. Schmid Communications Review. Vol. 7, No. 1, January 2003 AbererPHS:02 Improving Data Access in P2P Systems 1312 Karl Aberer, Anwitaman Datta, Zoran Despotovic, IEEE Internet Computing, January/February 2002, pp. Andreas Wombacher AbererDDW:03 Separating Business 58-67 Process from User Interaction in Web-Based Information 1327 L. Blazevic, L. Buttyan, S. Capkun, S. Giordano, J. Commerce Electronic Commerce Research, 3 (1-2): 83-111, P. Hubaux, J. Y. Le Boudec BlazevicBCGHL:01 Self January–April, 2003, Kluwer Academic Publishers Organization in Mobile Ad-Hoc Networks: the Approach of 1313 G. Camponovo, Y. Pigneur CamponovoP:02a) Terminodes IEEE Communications Magazine, June 2001. Analyzing the m-business landscape Annals of Telecommu 1328 S. Servetto, K. Nahrstedt ServettoN:01 Broad nications, Hermes, January-February 2003, vol. 58, no. 1-2 cast-Quality Video over IPIEEE Transactions on Multime 1314) Jeremy Elson, Kay Romer ElsonR:03 Wireless dia, 3(1):162-173, March 2001 Sensor Networks: A New Regime for Time Synchronization 1329 J. P. Hubaux, Th. Gross, J. Y. Le Boudec, M. ACM SIGCOMM Computer Communication Review Vetterli HubauxGLV:01 Towards self-organized mobile ad (CCR), January 2003 hoc networks: the Terminodes project IEEE Communica 1315) Karl Aberer, Philippe Cudre-Mauroux, Manfred tions Magazine, January 2001. Hauswirth AbererCH:02a) A Framework for Semantic Gos 1330] Evaluation of the Zaurus SL5600 PDA as a plat siping ACM SIGMOD Record, December 2002 form for ad-hoc networking L. Previtali http://www.termi 1316 G. Avoine and S. Vaudenay AvoineV:03a Cryp nodes.org/MV2003-Present/Lu13/Zaurus Previtali.pdf tography with Guardian Angels: Bringing civilization to 1331 Simulating large ad-hoc networks with ns-2 V. pirates—Abstract In L. Buttyan and J.-P. Hubaux (Eds.), Naoumov, http://www.terminodes.org/MV2003-Present/ Report on a Working Session on Security in Wireless Ad Lu 13/Simulating-Naoumov.pdf Hoc Networks, ACM Mobile Computing and Communica tions Review (MC2R), Vol. 7. No. 1., 2003, pp.74-94 1332 Security of emergent properties in ad-hoc networks Prof. V. Gligor, Univ. of Maryland www.terminodes.org/ 1317 Kay Römer, Oliver Kasten, Friedemann Mattern MV2003-Present/Ma14/Gligor-Monteverita.pdf RomerKM:02 Middleware Challenges for Wireless Sensor Networks ACM SIGMOBILE Mobile Computing and Com 1333) On the self-organization of security in multi-hop munication Review (MC2R), Fall 2002 networks Prof. J. P. Hubaux www.terminodes.org/MV2003 Present/MalA/Multi-Hop-JPH.pdf 1318. P. L. Dragotti, S. Servetto and M. Vetterli Drag ottiSV:02 Optimal filter banks for multiple description 1334 Immune Networking Systems Prof. J.-Y. Le Bou coding: analysis and synthesis IEEE Transactions on Infor dec www.terminodes.org/MV2003-Present/Ma14/Immu mation Theory, 48(7):2036-2052, July 2002 neLeBoudec.pdf 1319 R. Karrer and T. Gross KarrerG:02 Location 1335 Fair exchange with guardian angels G. Avoine Selection for Active Services Cluster Computing 5(3): 265 www.terminodes.org/MV2003-Present/MalA/FairEx 275; July 2002 change-Avoine.pdf US 2006/0167784 A1 Jul. 27, 2006 48
1336 Spatial analysis of mobile ad-hoc networks under First Embodiment the Signal to Interference Ratio connectivity model Prof. F. Baccelli, Ecole Normale Supérieure and INRIA, Paris www 1352. In a typical auction, each player is treated fairly; ...terminodes.org/MV2003-Present/Me15/Spacial-Baccel that is, the same rules apply to each player, and therefore a li.pdf single economy describes the process. The fair auction therefore poses challenges for an inherently hierarchal set of 1337 Connectivity and interferences in wireless ad-hoc users, such as a military organization, where rank is accom networks, a percolation approach 0. Dousse www.termin panied by privilege. The net result, however, is a decided odes.org/MV2003-Present/Me15/OlivierDousseMonteveri disadvantage to lower ranking agents. In a mobile ad hoc ta.pdf network, a real issue is user defection or non-compliance. For example, where a cost is imposed on a user for partici 1338 Ad-hoc networks: the worst and the average case pating in the ad hoc network, e.g., battery power consump Prof. R. Wattenhofer www.terminodes.org/MV2003 tion, if the anticipated benefit does not exceed the cost, the Present/Me15/MV-Wattenhofer.pdf www.terminodes.org user will simply turn off the device until actually needed. www.google.com/search?&q=terminodes www.ietforg/ht The result of mass defection will of course be the instability ml.charters/manet-charter.html. citeseer.nj.nec.com/cs?cs= and failure of the ad hoc network itself, leading to decreased 1&q=manet&Submit=Documents www.google.com/ utility, even for those who gain an unfair advantage under search?&q=%22ad+hoc%22+and+game the system. Thus, perceived fairness and net benefit is citeseer.nj.nec.com/cs?cs=1&q=game--and--%22ad+ required to network Success, assuming that defection or hoc'/622 carmen.cselt.it/idXwg/manet.html non-compliance are possible. BRIEF DESCRIPTION OF THE DRAWINGS 1353. On the other hand, in military systems, the asser tion of rank as a basis for priority is not itself arbitrary and 1339. The drawings show: capricious. Orders and communications from a central com mand are critical for the organization itself, and thus the 1340 FIG. 1 shows a Bayesian Network: lower ranking agents gain at least a peripheral benefit as their own chain of command employs their resources. There 1341 FIG. 2 shows a Markov chain; fore, the difficulty in analyzing the application of a fair game 1342 FIG. 3 shows a model of the output of a Markov to a hierarchal organization is principally a result of con chain as a mixture of Gaussians; ceptualizing and aligning the individual incentives with those of the organization as a whole and the relationship 1343 FIGS. 4A-4C show an input-output, a factorial, between branches. and a coupled Hidden Markov Model (HMM), respectively; 1354. In establishing an economic system, a preliminary 1344 FIG. 5 shows a predictor corrector algorithm of the question is whether the system is microeconomic or mac discrete Kalman filter cycle: roeconomic; that is, whether the economy is linked to a real economy or insulated from it. One disadvantage of a real 1345 FIG. 6 shows aspects of the discrete Kalman filter economy with respect to a peer relationship is that external cycle algorithm; wealth can override internal dynamics, thus diminishing the 1346) FIG. 7 shows aspects of the extended Kalman filter advantages to be gained by optimization, and potentially cycle; creating a perception of unfairness for less wealthy agents, at least until the system accomplishes a wealth redistribu 1347 FIG. 8 shows a block diagram of a preferred tion. An artificial economy provides a solution for a peer embodiment of a communications system according to the network in which each node has an equal opportunity to gain present invention; control over the ad hoc network, independent of outside influences. 1348 FIG. 9 is a schematic diagram showing the priori tization scheme; and 1355. In this proposed artificial economy, each node has a generator function for generating economic units, which 1349) FIG. 10 is a block diagram representing a message are then used in an auction with other nodes to create a format. market economy. The economic units have a temporally and spatially declining value, so that wealth does not accumulate DESCRIPTION OF THE INVENTION over long periods and cannot be transferred over large distances. This creates a set of microeconomies insulated 1350. The present invention seeks to apply aspects of from each other. Where distant microeconomies must deal game theory to the enhancement or optimization of com with each other, there is a discount. munities. These communities may themselves have various rules, arrangements or cultures, which can be respected or 1356) I propose that this scheme may be extended to the programmed as a part of the system operation. These com hierarchal case by treating branch within the chain of munities may be formed or employed for various purposes, command as an economic unit with respect to the generator and a preferred embodiment optimizes wireless communi function. At any level of the hierarchy, the commander cations in an open access band. retains a portion of the wealth generation capacity, and delegates the remainder to its subordinates. Therefore, the 1351. By optimizing communications, a greater commu rank and hierarchal considerations are translated to an nications bandwidth will generally be available, which will economic wealth (or wealth generation) distribution. One allow richer communications. This, in turn, permits new aspect of this system allows wealth transfer or redistribution, applications which depend on communications. although in a real system, a time delay is imposed, and in the US 2006/0167784 A1 Jul. 27, 2006 49 event of a temporally and/or spatially declining value, the has insufficient initial wealth (or wealth generating function) transfer will impose a cost. Thus, an initial misallocation is allocation, it may still participate, but it must expend its undesired, and there will be an incentive to optimally internal resources to obtain wealth for participation in its distribute the wealth initially. own benefit. This, in turn, leads to a potential exhaustion of 1357. In this system, there may be an economic competi resources, and the unavailability of the node for ad hoc tive distortion, under which a node's subjective value of a intermediary use, even for the benefit of the hierarchy. An resource is influenced by its then subjective wealth. If a node initial surplus allocation will lead to overbidding for is supplied with wealth beyond its needs, the wealth is resources, and thus inefficient resource allocation, potential wasted, since it declines in value and cannot be hoarded waste of allocation, and a disincentive to act as an interme indefinitely. If a node is supplied with insufficient wealth, diary in the ad hoc network. economic Surplus through transactional gains are lost. Thus, 1364. In a military system, it is clearly possible to for each node must analyze its expected circumstances to retain mulate an “engineered solution which forces participation or delegate the generator function, and to optimally allocate and eliminates defection; however, it is clear that such wealth between competing Subordinates. Likewise, there Solutions forfeit the potential gains of optimality, and incen may be a plurality of quasi-optimal states. tivizes circumvention. Further, because Such a system is not 1358. In any transaction, there will be a component “cost sensitive' (however the appropriate cost function which represents the competitive “cost, and a possible might be expressed), it fails to respond to “market” forces. redistribution among nodes within a hierarchal chain. This 1365 Accordingly, a peer to peer mobile ad hoc network redistribution may be of accumulated wealth, or of the Suitable for respecting hierarchal organization structures is generation function portion. Trading among hierarchally delegation is provided. related parties is preferred, since the perceived cost is low, and the wealth can be repeatedly redistributed. In fact, it is Second Embodiment because of the possibility of wealth oscillation and teaming that the declining wealth function is preferred, since this will 1366 Multihop Ad Hoc Networks require cooperation of tend to defeat closely related party control over the network nodes which are relatively disinterested in the content being for extended periods. conveyed. Typically, Such disinterested intermediaries incur a cost for participation, for example, power consumption or 1359. It is noted that, in a multihop mobile ad hoc opportunity cost. Economic incentives may be used to network, if a communication path fails, no further transfers promote cooperation of disinterested intermediaries. An are possible, potentially resulting in stalled or corrupt sys economic optimization may be achieved using a market tem configuration. It is possible to transfer an expiring or finding process, such as an auction. In many scenarios, the declining portion of the generating function; however, this desire for the fairness of an auction is tempered by other might lead a node which is out of range to have no ability concerns, i.e., there are constraints on the optimization to rejoin the network upon return, and thus act as an which influence price and parties of a transaction. For impediment to efficient network operation. example, in military communication systems, rank may be 1360. In practice, the bulk of the wealth generating deemed an important factor in access to, and control over, function will be distributed to the lowest ranks with the the communications medium. A simple process of rank highest numbers. Thus, under normal circumstances, the based preemption, without regard for subjective or objective network will appear to operate according to a non-hierarchal importance, will result in an economic distortion. In order to model, with the distortion that not all nodes have a common normalize the application of rank, one is presented with two generator function. On the other hand, hierarchically Supe options: imposing a normalization scheme with respect to rior nodes either retain, or more likely, can quickly recruit rank to create a unified economy, or providing considering Surrounding Subordinates to allocate their wealth generating rank using a set of rules outside of the economy. One way function and accumulated wealth to pass urgent or valuable to normalize rank, and the implicit hierarchy underlying the messages. rank, is by treating the economy as an object oriented 1361) One way that this allocation of wealth may be hierarchy, in which each individual inherits or is allocated a apparent is through the use of expensive assets. Thus, a high subset of the rights of a parent, with peers within the level node might have access to a high power broadcast hierarchy operating in a purely economic manner. The system, while low level nodes might ordinarily be limited to extrinsic consideration of rank, outside of an economy, can cellular wireless communications. For a low level node to be denominated “respect', which corresponds to the societal generate a broadcast using an expensive asset (or to allocate treatment of the issue, rather than normalizing this factor a massive amount of space.bandwidth product, it must pass within the economy, in order to avoid unintended secondary the request up through the chain of command, until Sufficient economic distortion. Each system has its merits and limita wealth (i.e., authority) is available to implement the broad tions. CaSt. 1367. An economic optimization is one involving a trans action in which all benefits and detriments can be expressed 1362. In fact, such communications and authorizations in normalized terms, and therefore by balancing all factors, are quite consistent with the expectations within a hierarchal including Supply and demand, at a price, an optimum is organization, and this construct is likely to be accepted. achieved. Auctions are well known means to achieve an 1363 Under normal circumstances, a superior would economic optimization between distinct interests, to transfer have an incentive to assure that each Subordinate node a good or right in exchange for a market price. While there possesses sufficient wealth to carry out its function and be are different types of auctions, each having their limitations incentivized to participate in the network. If a subordinate and attributes, as a class these are well accepted as a means US 2006/0167784 A1 Jul. 27, 2006 50 for transfer of goods or rights at an optimum price. Where factor, and thus permits each bidder weight its own consid multiple goods or rights are required in a Sufficient combi erations. It is further noted that Buttyan et al. have discussed nation to achieve a requirement, a so-called Vickrey–Clarke this factor as a part of an automated means for ensuring Groves (VCG) auction may be employed. In such an auc compliance with network rules, in the absence of a hierar tion, each Supplier asserts a desired price for his component. chy. The bias introduced in the manner is created by an The various combinations which meet the requirement are assertion by one claiming privilege, and deference by one then compared, and the lowest selected. In a combinatorial respecting privilege. One way to avoid substantial economic Supply auction, a plurality of buyers seek a divisible com distortions is to require that the payment made be based on modity, and each bids its best price. The bidders with the a purely economic optimization, while selecting the winner combination of prices which is maximum is selected. In a based on other factors. In this way, the perturbations of the commodity market, there are a plurality of buyers and auction process itself is Subtle, that is, since bidders realize sellers, so the auction is more complex. In a market that the winning bid may not result in the corresponding economy, the redistribution of goods or services are typi benefit, but incurs the publication of private values and cally transferred between those whoe value them least to potential bidding costs, there may be perturbation of the those who value them most. The transaction price depends bidding strategy from optimal. Likewise, since the privilege on the balance between Supply and demand; with the Surplus is itself unfair and predictable, those with lower privilege being allocated to the limiting factor. ratings will have greater incentive to defect from, or act 1368. A previous scheme proposes the application of against, the network. Therefore, it is important that either the game theory in the control of mulotihop mobile ad hoc assertion of privilege be subjectively reasonable to those networks according to “fair principles. In this prior Scheme, who must defer to it, or the incidence or impact of the nodes seeking to control the network (i.e., are “buyers' of assertions be uncommon or have low anticipated impact on bandwidth), conduct an auction for the resources desired. the whole. Likewise, potential intermediate nodes conduct an auction to 1371. In the extreme case, the assertion of privilege will Supply the resources. The set of winning bidders and win completely undermine the auction optimization, and the ning sellers is optimized to achieve the maximum economic system will be prioritized on purely hierarchal grounds, and Surplus. Winning bidders pay the maximum bid price or the pricing Suboptimal or unpredictable. This condition may second price, while winning sellers receive their winning be acceptable or even efficient in military systems, but may ask or second price. The remaining Surplus is redistributed be unacceptable where the deference is voluntary and choice among losing bidders, whose cooperation and non-interfer of network protocol is available. ence with the winning bidders is required for network operation, in accordance with their proportionate bid for 1372. It is noted that those seeking access based on contested resources. The winning bids are determined by a respect, must still make an economic bid. This bid, for VCG combinatorial process. The result is an optimum example, should be sufficient in the case that respect is not network topology with a reasonable, but by no means the afforded, for example, from those of equal rank or above, or only, fairness criterion, while promoting network Stability those who for various reasons have other factors that over and utility. ride the assertion of respect. Therefore, one way to deter mine the amount of respect to be afforded is the self-worth 1369. One issue with such a fair allocation of resources is advertised for the resources requested. This process there that it does not readily permit intentional distortions. How fore minimizes the deviation from optimal and therefore ever, in some instances, a relatively extrinsic consideration promotes stability of the network. It is further noted that to supply and Subjective demand may be a core requirement those who assert respect based on hierarchy typically have of a system. For example, in military systems, it is tradi available Substantial economic resources, and therefore it is tional and expected that higher military rank will provide largely a desire to avoid economic redistribution rather than access to and control over resources on a favored basis. In an inability to effect such a redistribution, that compels a civilian systems, emergency and police use may also be consideration of respect. considered privileged. However, by seeking to apply eco nomic rules to this access, a number of issues arise. Most 1373. In a combinatorial auction, each leg of a multihop significantly, as a privileged user disburses currency, this is link is separately acquired and accounted. Therefore, admin distributed to unprivileged users, leading to an inflationary istration of the process is quite involved. That is, each bidder broadcasts a bid, and an optimal network with maximum effect and comparative dilution of the intended privilege. If surplus is defined. Each leg of each path is therefore the economy is real, that is the currency is linked to a real allocated a value. Accordingly, if a bidder seeks to acquire economy, this grant of privilege will incur real costs, which the route, even though it was an insufficient economic is also not an intended effect. If the economy is synthetic, bidder, but awarded the route based on respect, then those that is, it is unlinked to external economies, then the redis who must defer or accept reduced compensation must acqui tribution of wealth within the system can grant dramatic and esce based on deference, which is neither mandated nor undesired control to a few nodes, potentially conveying the uniform. If pricing is defined by the economic optimization, privilege to those undeserving, except perhaps due to for then the respect consideration requires that a Subsidy be tuitous circumstances. applied, either as an excess payment up to the amount of the 1370. Two different schemes may be used to address this winning bid, or as a discount provided by the sellers, down desire for both economic optimality and hierarchal consid to the actually bid value. Since we presume that a surplus erations. One scheme maintain optimality and fairness exists, this value may be applied to meet the gap, while within the economic structure, but applies a generally maintaining optimal valuation. The node demanding respect orthogonal consideration of “respect as a separate factor may have an impact on path segments outside the required within the operation of the protocol. Respect is a subjective route; and thus the required payment to meet the differential US 2006/0167784 A1 Jul. 27, 2006 between the optimum network and the resulting network thus leading to a disincentive to cooperate. Meanwhile, may this be significant. If there is insufficient Surplus, then bystander nodes must defer their own communications in a different strategy may be applied. order to avoid interference, especially in highly loaded networks. By understanding the decision analysis of the 1374 Since the allocation of respect is subjective, each various nodes in a network, it is possible to optimize a bidder Supplies a bid, as well as an assertion of respect. Each system which, in accordance with game theory, provides Supplier receives the bids, and applies a weighting or dis benefits or incentives, to promote network reliability and count based on its Subjective analysis of the respect asser stability. The incentive, in economic form, may be charged tion. In this case, the same bid is interpreted differently by to those benefiting from the communication, and is prefer each Supplier, and the Subjective analysis must be performed ably related to the value of the benefit received. The pro by or for each Supplier. By converting the respect assertion posed network optimization scheme employs a modified into a subjective weighting or discount, a pure economic combinatorial (VCG) auction, which optimally compensates optimization may then be performed. those involved in the communication, with the benefiting 1375 An alternate scheme for hierarchal deference is to party paying the second highest bid price (second price). The organize the economy itself into a hierarchy. In a hierarchy, surplus between the second price and VCG price is distrib a node has one parent and possibly multiple children. At uted among those who defer to the winning bidder according each level, a node receives an allocation of wealth from its to respective bid value. Equilibrium usage and headroom parent, and distributes all or a portion of its wealth to may be influenced by deviating from a Zero-sum condition. children. A parent is presumed to control its children, and The mechanism seeks to define fairness in terms of market therefore can allocate their wealth or subjective valuations to value, providing probable participation benefit for all nodes, its own ends. When nodes representing different lineages leading to network Stability. must be reconciled, one may refer to the common ancestor for arbitration, or a set of inherited rules to define the 1382) Ad Hoc Networks hierarchal relationships. 1383 An ad hoc network is a wireless network which does not require fixed infrastructure or centralized control. 1376. In this system, the resources available for reallo The terminals in the network cooperate and communicate cation between branches of the hierarchy depends on the with each other, in a self organizing network. In a multihop allocation by the common grandparent, as well as competing network, communications can extend beyond the scope of a allocations within the branch. This system presumes that single node, employing neighboring nodes to forward mes children communicate with their parents and are obedient. In sages to their destination. In a mobile ad hoc network, fact, if the communication presumption is violated, one must constraints are not placed on the mobility of nodes, that is, then rely on a priori instructions, which may not be Sufi they can relocate within a time scale which is short with ciently adaptive to achieve an optimal result. If the obedi respect to the communications, thus requiring consideration ence presumption is violated, then the hierarchal deference of dynamic changes in network architecture. requires an enforcement mechanism within the hierarchy. If both presumptions are simultaneously violated, then the 1384 Ad hoc networks pose control issues with respect system will likely fail, except on a voluntary basis, with to contention, routing and information conveyance. There results similar to the “reputation scheme described above. are typically tradeoffs involving equipment size, cost and complexity, protocol complexity, throughput efficiency, 1377 Thus, it is possible to include hierarchal deference energy consumption, and “fairness of access arbitration. as a factor in optimization of a multihop mobile ad hoc network, leading to compatibility with tiered organizations, Other factors may also come into play 4-10). as well as with shared resources. 1385 Game theory studies the interactions of multiple independent decision makers, each seeking to fulfill their 1378 Game Theory own objectives. Game theory encompasses, for example, 1379 Use of Game Theory to control arbitration of ad auction theory and strategic decision-making. By providing hoc networks is well known 1.2). Some prior studies have appropriate incentives, a group of independent actors may focused on the incremental cost to each node for participa be persuaded, according to self-interest, to act toward the tion in the network, without addressing the opportunity cost benefit of the group. That is, the selfish individual interests of a node foregoing control over the communication are aligned with the community interests. In this way, the medium 3. The present paper adds a new dimension to this community will be both efficient and the network of actors analysis. stable and predictable. Of course, any systems wherein the “incentives' impose too high a cost, themselves encourage 1380 A game theoretic approach addresses the situation circumvention. In this case, game theory also addresses this where the operation of an agent which has freedom of 1SSC. choice, allowing optimization on a high level, considering the possibility of alternatives to a well designed system. 1386. In computer networks, issues arise as the demand According to game theory, the best way to ensure that a for communications bandwidth approaches the theoretical system retains compliant agents, is to provide the greatest limit. Under such circumstances, the behavior of nodes will anticipated benefit, at the least anticipated cost, compared to affect how close to the theoretical limit the system comes, and also which communications are permitted. The well the alternates. known collision sense, multiple access (CSMA) protocol 1381 Game Theory provides a basis for understanding allows each node to request access to the network, essen the actions of Ad hoc network nodes. A multihop ad hoc tially without cost or penalty, and regardless of the impor network requires a communication to be passed through a tance of the communication. While the protocol incurs disinterested node. The disinterested node incurs some cost, relatively low overhead and may provide fully decentralized US 2006/0167784 A1 Jul. 27, 2006 52 control, under congested network conditions, the system the problem of determining the network topology, and the may exhibit instability, that is, a decline in throughput as communications themselves, are ancillary, though real, demand increases, resulting in ever increasing demand on applications of game theory. Since the communications the system resources and decreasing throughput 11. incidental to the arbitration require consideration of some of According to game theory, the deficit of the CSMA protocol the same issues as the underlying communications, corre is that it is a dominant strategy to be selfish and hog sponding elements of game theory may apply at both levels resources, regardless of the cost to Society, resulting in “the of analysis. Due to various uncertainties, the operation of the tragedy of the commons. 12 system is stochastic. This presumption, in turn, allows estimation of optimality within a margin of error, simplify 1387. In an ad hoc network used for conveying real-time ing implementation as compared to a rigorous analysis information, as might be the case in a telematics system, there are potentially unlimited data communication require without regard to statistical significance. ments (e.g., video data), and network congestion is almost 1392 There are a number of known and proven routing guaranteed. Therefore, using a CSMA protocol as the para models proposed for forwarding of packets in ad hoc net digm for basic information conveyance is destined for works. These include Ad Hoc On-Demand Distance Vector failure, unless there is a disincentive to network use. (In (AODV) Routing, Optimized Link State Routing Protocol power constrained circumstances, this cost may itself pro (OLSR), Dynamic Source Routing Protocol (DSR), and vide such a disincentive). On the other hand, a system which Topology Dissemination Based on Reverse-Path Forward provides more graceful degradation under high load, sensi ing (TBRPF) 14-26). In most systems analyzed to date, the tivity to the importance of information to be communicated, performance metrics studied were power consumption, end and efficient utilization of the communications medium to-end data throughput and delay, route acquisition time, would appear more optimal. percentage out-of-order delivery, and efficiency. A critical 1388 One way to impose a cost which varies in depen variable considered in many prior studies is power cost, dence on the Societal value of the good or service, is to presuming a battery operated transceiver with limited power conduct an auction, which is a mechanism to determine the availability 27-32). There can be significant differences in market value of the good or service, at least between the optimum routing depending on whether a node can modulate auction participants 13. In an auction, the bidder seeks to its transmit power, which in turn controls range, and pro bid the lowest value, up to a value less than or equal to his vides a further control over network topology. Likewise, own private value (the actual value which the bidder steerable antennas, antenna arrays, and other forms of mul appraises the good or service, and above which there is no tiplexing provide further degrees of control over network surplus), that will win the auction. Since competitive bidders topology. Note that the protocol-level communications are can minimize the gains of another bidder by exploiting preferably broadcasts, while information conveyance com knowledge of the private value attached to the good or munications are typically point-to-point. Prior studies typi service by the bidder, it is generally a dominant strategy for cally presume a single transceiver, with a single omnidirec the bidder to attempt to keep its private value a secret, at tional antenna, operating according to in-band protocol data, least until the auction is concluded, thus yielding strategies for all communications. The tradeoff made in limiting sys that result in the largest potential gain. On the other hand, in tem designs according to these presumptions should be certain situations, release or publication of the private value clear. is a dominant strategy, and can result in Substantial effi 1393. It is the general self-interest of a node to conserve ciency, that is, honesty in reporting the private value results its own resources, maintain an opportunity to access net in the maximum likelihood of prospective gain. work resources, while consuming whatever resource of other nodes as it desires. Clearly, this presents a significant 1389 Application of Game Theory to Ad Hoc Networks risk of the “tragedy of the commons', in which selfish 1390 There are a number of aspects of ad hoc network individuals fail to respect the very basis for the community control which may be adjusted in accordance with game they enjoy, and a network of rational nodes operating theoretic approaches. An example of the application of game without significant incentives to cooperate would likely fail. theory to influence system architecture arises when commu On the other hand, if donating a node's resources generated nications latency is an issue. A significant factor in latency a Sufficient associated benefit to that node, while consuming is the node hop count. Therefore, a system may seek to network resources imposed a sufficient cost, stability and reduce node hop count by using an algorithm other than a reliability can be achieved. So long as the functionality is nearest neighbor algorithm, bypassing some nodes with Sufficient to meet the need, and the economic Surplus is longer distance communications. In analyzing this possibil “fairly allocated, that is, the cost incurred is less than the ity, one must not only look at the cost to the nodes involved private value of the benefit, and that cost is transferred as in the communication, but also the cost to nodes which are compensation to those burdened in an amount in excess of prevented from simultaneously accessing the network dude their incremental cost, adoption of the system should to interfering uses of network resources. As a general increase stability. In fact, even outside of these bounds, the proposition, the analysis of the network must include the system may be more stable than one which neither taxes impact of each action, or network State, on every node in the system use nor rewards altruistic behavior. While the basic system, although simplifying presumptions may be appro system is a Zero Sum system, and over time, the economic priate where information is unavailable, or the anticipated effects will likely average out (assuming symmetric nodes), impact is trivial. in any particular instance, the incentive for selfish behavior 1391 Game theory is readily applied in the optimization by a node will be diminished. of communications routes through a defined network, to 1394 One way to remedy selfish behavior is to increase achieve the best economic Surplus allocation. In addition, the cost of acting this way, that is, to impose a cost or tax for US 2006/0167784 A1 Jul. 27, 2006
access to the network. In a practical implementation, how tion. On the other hand, a node may have a selfish motiva ever, this is problematic, since under lightly loaded condi tion in failing to reward behavior with a good reputation. tions, the “value' of the communications may not justify a fixed cost which might be reasonable under other conditions, 1397 Economics and reputation may be considered and likewise, under heavier loads, critical communications orthogonal, since the status of a node's currency account may still be delayed or impeded. A variable cost, dependent provides no information about the status of its reputation. on relative “importance, may be imposed, and indeed, as 1398. This reputation parameter may be extended to alluded to above, this cost may be market based, in the encompass respect, that is, a subjective deference to another manner of an auction. In a multihop network, Such an based on an asserted or imputed entitlement. While the prior auction is complicated by the requirement for a distribution system uses reputation as a factor to ensure compliance with of payments within the chain of nodes, with each node system rules, this can be extended to provided deferential having potential alternate demands for its cooperation. The preferences either within or extrinsic to an economy. Thus, market-based price-finding mechanism excludes nodes in a military hierarchy, a relatively higher ranking official which ask a price not supported by its market position, and can assert rank, and if accepted, override a relatively lower the auction itself may comprise a value function encom ranking bidder at the same economic bid. For each node, an passing reliability, latency, quality of service, or other non algorithm is provided to translate a particular assertion of economic parameters, in economic terms. The network may respect (i.e., rank and chain of command) into an economic further require compensation to nodes which must defer perturbation. For example, in the same chain of command, communications because of inconsistent states, such as in each difference in rank might be associated with a 25% order to avoid interference or duplicative use of an inter compounded discount, when compared with other bids, i.e. mediary node, and which take no direct part in the commu nication. It is noted that the concept of the winner of an B=Box10(1+0.25xAR), auction paying the losers is not well known, and indeed somewhat counterintuitive. Indeed, the effect of this rule while outside the chain of command, a different, generally perturbs the traditional analysis framework, since the pos lower, discount may be applied, possibly with a base dis sibility of a payment from the winner to the loser alters the count as compared to all bids within the chain of command, allocation of economic surplus between the bidder, seller, i.e., and others. Likewise, while the cost to the involved nodes B=Box10(1+dCOC+dNCOCXAR). may be real, the cost to the uninvolved nodes may be The discount is applied so that higher ranking officers pay subjective. Clearly, it would appear that involved nodes less, while lower ranking officers pay more. Clearly, there is should generally be better compensated than uninvolved a high incentive for each bid to originate from the highest nodes, although a rigorous analysis remains to be performed. available commander within the chain of command, and 1395 The network provides competitive access to the given the effect of the perturbation, for ranking officers to physical transport medium, and cooperation with the proto “pull rank” judiciously. col provides significant advantages over competition with it. Under normal circumstances, a well developed ad hoc 1399) The Modified VCG Auction network system can present as a formidable coordinated competitor for access to contested bandwidth by other 1400. A so-called Vickrey–Clarke-Groves, or VCG, auc systems, while within the network, economic Surplus is tion, is a type of auction Suitable for bidding, in a single optimized. Thus, a node presented with a communications auction, for the goods or services of a plurality of offerors, requirement is presented not with the simple choice to as a unit 34-47). In the classic case, each bidder bids a value participate or abstain, but rather whether to participate in an vector for each available combination of goods or services. ad hoc network with predicted stability and mutual benefit, The various components and associated ask price are evalu or one with the possibility of failure due to selfish behavior, ated combinatorially to achieve the minimum sum to meet and non-cooperation. Even in the absence of a present the requirement. The winning bid set is that which produces communication requirement, a network which rewards the maximum value of the accepted bids, although the cooperative behavior may be preferable to one which simply second (Vickrey) price is paid. In the present context, each expects altruism. offeror submits an ask price (reserve) or evaluatable value function for a component of the combination. If the mini 1396 The protocol may also encompass the concept of mum aggregate to meet the bid requirement is not met, the node reputation, that is, a positive or negative statement by auction fails. If the auction is successful, then the set of others regarding the node in question 33). This reputation offerors selected is that with the lowest aggregate bid, and may be evaluated as a parameter in an economic analysis, or they are compensated that amount. applied separately, and may be anecdotal or statistical. In any case, if access to resources and payments are made 1401 The VCG auction is postulated as being optimal for dependent on reputation, nodes will be incentivized to allocation of multiple resources between agents. It is “strat maintain a good reputation, and avoid generating a bad egyproof and efficient, meaning that it is a dominant reputation. Therefore, by maintaining and applying the strategy for agents to report their true valuation for a reputation in a manner consistent with the community goals, resource, and the result of the optimization is a network the nodes are compelled to advance those goals in order to which maximizes the value of the system to the agents. benefit from the community. Game theory distinguishes Game theory also allows an allocation of cost between between good reputation and bad reputation. Nodes may various recipients of a broadcast or multicast. That is, the have a selfish motivation to assert that another node has a communication is of value to a plurality of nodes, and a large bad reputation, while it would have little selfish motivation, set of recipient nodes may efficiently receive the same absent collusion, for undeservedly asserting a good reputa information. This allocation from multiple bidders to mul US 2006/0167784 A1 Jul. 27, 2006 54 tiple sellers is a direct extension of VCG theory, and a with a Substantially common estimation of network topol similar algorithm may be used to optimize allocation of ogy, only deviations from previously propagated informa costs and benefit. tion need be propagated. 1402. The principal issue involved in VCG auctions is 1409 CSMA is proposed for the protocol-related com that the computational complexity of the optimization grows munications because it is relatively simple and robust, and with the number of buyers and their different value functions well Suited for ad hoc communications in lightly loaded and allocations. While various simplifying presumptions networks. An initial node transmits using an adaptive power may be applied, studies reveal that these simplifications may protocol, to achieve an effective transmit range of somewhat undermine the VCG premise, and therefore do not promote less than about two times the estimated average inter-nodal honesty in reporting the buyer's valuation, and are thus not distance. This distance therefore promotes propagation to a “strategyproof, which is a principal advantage of the VCG set of neighboring nodes, without unnecessarily interfering process. with communications of non-neighboring nodes and there fore allowing this task to be performed in parallel. Neigh 1403. The surplus, i.e., gap between bid and ask, is then boring nodes also transmit in Succession, providing sequen available to compensate the deferred bidders. This surplus is tial and complete protocol information propagation over a distributed proportionately to original the bid value for the relevance range. bidder, thus further encouraging an honest valuation of control over the resource. 1410) If we presume that there is a spatial limit to relevance, for example, 5 miles or 10 hops, then the network 1404) The optimization is such that, if any offeror asks an state propagation may be so limited. Extending the network amount that is too high, it will be bypassed in favor of more to encompass a large number of nodes will necessarily “reasonable' offerors. Since the bidder pays the second reduce the tractability of the optimization. Each node has a highest price, honesty in bidding the full private value is local estimate of relevance. This consideration is accommo encouraged. The distribution of the Surplus to losing bidders, dated, along with a desire to prevent exponential growth in which exercise deference to the winner, is proportional to the protocol-related data traffic, by receiving an update from all amount bid, that is, the reported value. nodes within a node's network relevance boundary, and a 1405. In a scenario involving a request for information state variable which represents an estimate of relevant status meeting specified criteria, the auction is complicated by the beyond the arbitrarily defined boundary. The propagation of fact that the information resource content is unknown to the network State may thus conveniently occur over a finite recipient, and therefore the bid is blind, that is, the value of number of hops, for example 5-10. the information to the recipient is indeterminate. However, 1411 Under conditions of relatively high nodal densities, game theory Supports the communication of a value function the system may employ a Zone strategy, that is, proximate or utility function, which can then be evaluated at each node groups of nodes are is treated as an entity for purposes of possessing information to be communicated, to normalize its external state estimation, especially with respect to distant value. Fortunately, it is a dominant strategy in a VCG nodes or Zones. Such a presumption is realistic, since at auction to communicate a truthful value, and therefore extended distances, geographically proximate nodes may be broadcasting the private value function, to be evaluated by modeled as being similar or inter-related, while at close a recipient, is not untenable. In a mere request for informa distances, and particularly within a Zone in which all nodes tion conveyance, such as the transport nodes in a multihop are in direct communication, inter-node communications network, or in a cellular network infrastructure extension may be subject to mutual interference, and can occur without model, the bid may be a true (resolved) value, since the substantial external influence. Alternately, it is clear that to information content is not the subject of the bidding; rather limit latencies and communication risks, it may be prudent it is the value of the communications perse, and the bidding to bypass neighboring nodes, thus trading latency for power node can reasonably value its bid. consumption and overall network capacity. Therefore, a hierarchal Scheme may be implemented to geographically 1406 Game theory also allows an allocation of cost organize the network at higher analytical levels, and geo between various recipients of a broadcast or multicast. That graphic cells may cooperate to appear externally as a single is, in many instances, information which is of value to a plurality of nodes, and a large set of recipient nodes may entity. efficiently receive the same information. This allocation is a 1412. A supernode within a Zone may be selected for its direct extension of VCG theory. Superior capability, or perhaps a central location. The Zone is defined by a communication range of the basic data 1407 Operation of Protocol interface for communications, with the control channel 1408. The preferred method for acquiring an estimate of having a longer range, for example at least double the the state of the network is through use of a proactive routing normal data communications range. Communications con protocol. Thus, in order to determine the network architec trol channel transmitters operate on a number of channels, ture state, each node must broadcast its existence, and, for for example at least 7, allowing neighboring Zones in a example, a payload of information including its identity, hexagonal tiled array to communicate simultaneously with location, itinerary (navigation vector) and “information out interference. In a geographic Zone system, alternate value function'. Typically, the system operates in a continu Zones which would otherwise be interfering may use an ous state, so that it is reasonable to commence the process adaptive multiplexing scheme to avoid interference. All with an estimate of the state based on prior information. nodes may listen on all control channels, permitting rapid Using an in-band or out-of-band propagation mechanism, propagation of control information. this information must propagate to a network edge, which 1413. In order to effective provide decentralized control, may be physically or artificially defined. If all nodes operate either each node must have a common set of information to US 2006/0167784 A1 Jul. 27, 2006 allow execution of an identical control algorithm, or nodes 1417. After the network architecture is defined, compen defer to the control signals of other nodes without internal sation is paid to those nodes providing value or Subjected to analysis for optimality. A model of semi-decentralized con a burden (including foregoing communication opportunity) trol is also known, in which dispersed Supernodes are by those gaining a benefit. The payment may be a virtual nominated as master, with other topologically nearby nodes currency, with no specific true value, although the virtual remaining as slave nodes. In the pure peer network, com currency system provides a convenient method to tax, Sub plete information conveyance to each node is required, sidize, or control the system, and thus apply a normalized imposing a relatively high overhead. In a master-slave (or extrinsic value. Supernode) architecture, increased reliance on a single node 1418 Using the protocol communication system, each trades-off reliability and robustness (and other advantages of node transmits its value function (or change thereof), passes pure peer-to-peer networks) for efficiency. A Supernode through communications from neighboring nodes, and may, within a cellular Zone may be selected for its superior for example transmit payment information for the immedi capability, or perhaps is at a central location or is immobile. ate-past bid for incoming communications. 1414. Once each control node (node or Supernode) has an 1419 Messages are forwarded outward (avoiding redun estimate of network topology, the next step is to optimize dant propagation back to the source), with messages network channels. According to VCG theory, each agent has appended from the series of nodes. Propagation continues an incentive to broadcast its truthful value or value function for a finite number of hops, until the entire community has for the scarce resource, which in this case, is control over an estimate of the state and value function of each node in communications physical layer, and or access to informa the community. Advantageously, the network beyond a tion. This communication can be consolidated with the respective community may be modeled in simplified form, network discovery transmission. Each control node then to provide a better estimate of the network as a whole. performs a combinatorial solution for the set of simulta 1420. After propagation, each node evaluates the set of neous equations according to VCG theory (or extensions value functions for its community, with respect to its own thereof). This solution should be consistent between all information and ability to forward packets. Each node may nodes, and the effects of inconsistent solutions may be then make an offer to supply or forward information, based resolved by collision sensing, and possibly an error/incon on the provided information. In the case of multihop com sistency detection and correction algorithm specifically munications, the offers are propagated to the remainder of applied to this type of information. the community, for the maximum number of hops, including 1415. As part of the network mapping, communications the originating node. At this point, each node has a repre impairment and interference sources are also mapped. GPS sentation of the state of its community, with community assistance may be particularly useful in this aspect. Where edge estimates providing consistency for nodes with differ interference is caused by interfering communications, the ing community Scopes, the valuation function each node issue is a determination of a strategy of deference or assigns to control over portions of the network, as well as a competition. If the interfering communication is continuous resolved valuation of each node for Supplying the need. or unresponsive, then the only available strategy is compe Under these circumstances, each node may then evaluate an tition. On the other hand, when the competing system uses, optimization for the network architecture, and come to a for example, a CSMA system, such as 802.11, competition conclusion consistent with that of other members of its with Such a communication simply leads to retransmission, community. If Supported, node reputation may be updated and therefore ultimately increased network load, and defer based on past performance, and the reputation applied as a ence strategy may be more optimal, at least and until it is factor in the optimization and/or externally to the optimiza determined that the competing communication is incessant. tion. As discussed above, a VCG-type auction is employed Other communications protocols, however may have a more as a basis for optimization. Since each node receives bid or less aggressive strategy. By observation of a system over information from all other nodes within the maximum node time, its strategies may be revealed, and game theory count, the VCG auction produces an optimized result. permits composition of an optimal strategy to deal with 1421 Transmissions are made in frames, with a single interference or coexistence. bidding process controlling multiple frames, for example a 1416) The optimization process produces a representation multiple of the maximum number of hops. Therefore, the bid of optimal network architecture during the Succeeding encompasses a frame's-worth of control over the modalities. period. That is, value functions representing bids are broad In the event that the simultaneous use of, or control over, a cast, with the system then being permitted to determine an modality by various nodes is not inconsistent, then the value optimal real valuation and distribution of that value. Thus, of the respective nodes may be summed, with the resulting prior to completion of the optimization, potentially incon allocation based on, for example, a ratio of the respective sistent allocations must be prevented, and each node must value functions. As a part of the optimization, nodes are communicate its evaluation of other node's value functions, rewarded not only for Supporting the communication, but so that the optimization is performed on a normalized also for deferring their own respective needs. As a result, economic basis. This step may substantially increase the after controlling the resources, a node will be relatively less system overhead, and is generally required for completion of wealthy and less able to Subsequently control the resources, the auction. This valuation may be inferred, however, for while other nodes will be more able to control the resources. intermediate nodes in a multihop network path, since there The distribution to deferred nodes also serves to prevent is little subjectivity for nodes solely in this role, and the pure reciprocal communications, since the proposed mecha respective value functions may be persistent. For example, nism distributes and dilutes the wealth to deferring nodes. the valuation applied by a node to forward information is 1422 Because each node in the model presented above generally independent of content and involved party. has complete information, for a range up to the maximum US 2006/0167784 A1 Jul. 27, 2006 56 node count, the wealth of each node can be estimated by its cized. The likelihood of this type of misbehavior is also neighbors, and payment inferred even if not actually con diminished by avoiding monetization of the virtual currency. Summated. (Failure of payment can occur for a number of reasons, including both malicious and accidental). Because 1427. This currency generation and allocation mecha each hop adds significant cost, the fact that nodes beyond the nism generally encourages equal consumption by the vari maximum hop distance are essentially incommunicado is ous nodes over the long term. In order to discourage typically of little consequence; since it is very unlikely that consumption of bandwidth, an external tax may be imposed a node more than 5 or 10 hops away will be efficiently on the system, that is, withdrawing value from the system included in any communication, due to the increasing cost base on usage. Clearly, the effects of Such a tax must be with distance, as well as reduction in reliability and increase carefully weighed, since this will also impose an impedi in latency. Thus, large area and Scalable networks may exist. ment to adoption as compared to an untaxed system. On the other hand, a similar effect use-disincentive may be obtained 1423. The Synthetic Economy by rewarding low consumption, for example by allocating an advertising Subsidy between nodes, or in reward of 1424 Exerting external economic influences on the sys deference. tem may have various effects on the optimization, and may exacerbate differences in subjective valuations. The appli 1428. A synthetic economy affords the opportunity to cation of a monetary value to the virtual currency Substan provide particular control over the generator function, which tially also increases the possibility of misbehavior and in turn Supports a hierarchy. In this scheme, each node external attacks. On the other hand, a virtual currency with controls the generator function at respectively lower nodes, no assessed real value is self-normalizing, while monetiza and thus can allocate wealth among Subordinates. If one tion leads to external and generally irrelevant influences as assumes real time communications, then it is clear that the well as possible arbitrage. External economic influences superordinate node can directly place bids on behalf of may also lead to benefits, which are discussed in various subordinates, thus effectively controlling its entire branch. In papers on non-Zero Sum games. the absence of real time communications, the Superordinate node must defer to the discretion of the subordinate, subject 1425. In order to provide fairness, the virtual currency to reallocation later if the subordinate defects. If communi (similar to the so-called “nuglets” or “nuggets’ proposed for cations are impaired, and a set of a priori instructions are use in the Terminodes project) 48-55) is self-generated at insufficient, then it is up to the Subjective response of a node each node according to a schedule, and itself may have a to provide deference. time dependent value. For example, the virtual currency may have a half-life or temporally declining value. On the other 1429. It is noted that when sets of nodes “play favorites’’, hand, the value may peak at a time after generation, which the VCG auction will no longer be considered “strat would encourage deference and short term savings, rather egyproof. The result is that bidders will assume bidding than immediate spending, and would allow a recipient node strategies that do not express their secret valuation, with the to benefit from virtual currency transferred before its peak result being likely Suboptimal market finding during the value. This also means that long term hoarding of the auction. This factor can be avoided if hierarchal overrides currency is of little value, since it will eventually decay in and group bidding play only a small role in the economy, and value, while the system presupposes a nominal rate of thus the expected benefits from shaded bidding are out spending, which is normalized among nodes. The variation weighed by the normal operation of the system. function may also be adaptive, but this poses a synchroni zation issue for the network. An external estimate of node 1430 Military Hierarchy wealth may be used to infer counterfeiting, theft and failure 1431. In a typical auction, each player is treated fairly; to pay debts, and to further effect remediation. that is, the same rules apply to each player, and therefore a single economy describes the process. The fair auction 1426. The currency is generated and verified in accor therefore poses challenges for an inherently hierarchal set of dance with micropayment theory 56-57. Micropayment users, such as a military organization. In the military, there theory generally encompasses the transfer of secure tokens is typically an expectation that rank has its privileges. The (e.g., cryptographically endorsed information) having pre net result, however, is a decided subjective unfairness to sumed value, which are intended for verification, if at all, in lower ranking nodes. In a mobile ad hoc network, a real a non-real time transaction, after the transfer to the recipient. issue is user defection or non-compliance. For example, The currency is circulated (until expiration) as a token, and where a cost is imposed on a user for participating in the ad therefore is not subject to immediate authentication by hoc network, e.g., battery power consumption, if the antici Source. Since these tokens may be communicated through an pated benefit does not exceed the cost, the user will simply insecure network, the issue of forcing allocation of payment turn off the device until actually needed. The result of mass to particular nodes may be dealt with by cryptographic defection will of course be the instability and failure of the techniques, in particular public key cryptography, in which ad hoc network itself. Thus, perceived fairness and net the currency is placed in a cryptographic “envelope” benefit is required to network Success, assuming that defec addressed to the intended recipient, e.g., is encrypted with tion or non-compliance are possible. the recipient’s public key, which must be broadcast and used as, or in conjunction with, a node identifier. This makes the 1432. On the other hand, in military systems, the asser payment unavailable to other than the intended recipient. tion of rank as a basis for priority is not arbitrary and The issue of holding the encrypted token hostage and capricious, and is generally not perceived subjectively as extorting a portion of the value to forward the packet can be Such. Orders and communications from a central command dealt with by community pressure, that is, any node pre are critical for the organization itself. Therefore, the diffi senting this (or other undesirable) behavior might be ostra culty in analyzing the application of a fair game to a US 2006/0167784 A1 Jul. 27, 2006 57 hierarchal organization is principally a result of conceptu the request up through the chain of command, until Sufficient alizing and aligning the individual incentives with those of wealth (i.e., authority) is available to implement the broad the organization as a whole. CaSt. 1433. An artificial economy provides a basis for an 1438. In fact, such communications and authorizations economically efficient solution. In this economy, each node are quite consistent with the expectations within a hierarchal has a generator function for generating economic units organization, and this likely to be accepted. which are used in a combinatorial auction with other nodes. The economic units have a declining value, so that wealth 1439 Under normal circumstances, a superior would does not accumulate over long periods, and by implication, have an incentive to assure that each Subordinate node wealth accumulated in one region is not available for possesses sufficient wealth to carry out its function and be transfer in a distant region. incentivized to participate in the network. If a subordinate has insufficient initial wealth (or wealth generating function) 1434. This scheme may be extended to the hierarchal allocation, it may still participate, but it must expend its case by treating each chain of command as an economic unit internal resources to obtain wealth for participation in its with respect to the generator function. At any level of the own benefit. This, in turn, leads to a potential exhaustion of hierarchy, the commander retains a portion of the wealth resources, and the unavailability of the node for ad hoc generation capacity, and delegates the remainder to its intermediary use, even for the benefit of the hierarchy. An Subordinates. In the case of real-time communications, a initial surplus allocation will lead to overbidding for commander may directly control allocation of the generator resources, and thus inefficient resource allocation, potential function at each time period. Typically, there is no real-time waste of allocation, and a disincentive to act as an interme communications capability, and the wealth generator func diary in the ad hoc network. tion must be allocated a priori. Likewise, wealth may also be reallocated, although a penalty is incurred in the event of an 1440. In a military system, it is clearly possible to for initial misallocation since the transfer itself incurs a cost, mulate an “engineered solution which forces participation and there will be an economic competitive distortion, under and eliminates defection; however, it is clear that such which a node's subjective value of a resource is influenced Solutions forfeit the potential gains of optimality, and incen by its subjective wealth. If a node is supplied with wealth tivized circumvention. beyond its needs, the wealth is wasted, since it declines in 1441 Cellular Network Extension value and cannot be hoarded indefinitely. If a node is Supplied with insufficient wealth, economic Surplus through 1442 Cellular Networks provide efficient coverage of transactional gains are lost. Thus, each node must analyze its large portions of the inhabited landmass. On the other hand, expected circumstances to retain or delegate the generator achieving complete coverage, including relatively uninhab function, and to optimally allocate wealth between compet ited areas, may be cost inefficient or infeasible. On the other ing Subordinates. hand, there remains significant unmet demand for coverage of certain areas. 1435. In any transaction, there will be a component which represents the competitive “cost, and a possible 1443 Generally, it is likely that a need for service arises redistribution among nodes within a hierarchal chain. This within a few miles from the edge of a cellular network. That redistribution may be of accumulated wealth, or of the is, the fixed infrastructure is almost in reach. On the other generation function portion. In the former case, if the hand, the infrastructure costs required to fill in gaps or communication path fails, no further transfers are possible, marginally extend the network may be inordinately high, for while in the later case, the result is persistent until the the direct economic benefits achieved. At present, there is no transfer function allocation is reversed. It is also possible to effective means for remediating these gaps. transfer an expiring or declining portion of the generating 1444 One problem arises in that the present networks function; however, this might lead a node which is out of generally have a threshold usage plan. All territory encom range to have no ability to rejoin the network upon return, passed by a network is treated as fungible, and incurs the and thus act as an impediment to efficient network operation. same cost. Likewise, usage of partner networks is also 1436. In practice, the bulk of the wealth generating treated as fungible. Therefore, the incentive to extend net function will be distributed to the lowest ranks with the work reach for any company is limited to the overall highest numbers. Thus, under normal circumstances, the incentive for customers to defect to different networks, network will appear to operate according to a non-hierarchal balanced against the increased cost of extending the net VCG model, with the distortion that not all nodes have a work. It is in the context of this economic problem that a common generator function. On the other hand, hierarchi Solution is proposed. Quite simply, in the same areas where cally Superior nodes either retain, or more likely, can quickly the cellular infrastructure is insufficient and there is demand recruit surrounding subordinates to allocate their wealth for service, it may be possible to implement a peer-to-peer generating function and accumulated wealth to pass urgent network or multihop network to extend a cellular network or valuable messages. system. In fact, if we presume that the coverage is absent, the 1437. One way that this allocation of wealth may be network extension function may make use of the licensed apparent is through the use of expensive assets. Thus, a high spectrum, thus making the transceiver design more efficient, level node might have access to a high power broadcast and eliminating extrinsic competing uses for the bandwidth. system, while low level nodes might ordinarily be limited to On the other hand, different spectrum may be employed, cellular wireless communications. For a low level node to which may be licensed or unlicensed. generate a broadcast using an expensive asset (or to allocate 1445 Various studies have shown that modeled multihop a massive amount of space.bandwidth product, it must pass mobile ad hoc network architectures tend to have low US 2006/0167784 A1 Jul. 27, 2006 efficiency over three to five hops. This is due to node detriment of others. Notably, the user typically has no mobility and the probability of finding an end-to-end con reasonable ability to reprogram the phone or alter its opera nection, mutual interference and competition for bandwidth tion in accordance with the protocol, unless granted this in shared channel protocols, and the overhead of maintain option by the network operator. The user cannot reasonably ing useful routing tables. If we take five hops as a reasonable compete or interfere with the licensed spectrum. While older maximum, and each transceiver has a 1000 meter range, analog cellular phones provided the user with an option to then a 5 km maximum range extension is possible. It is install power amplifiers and vehicle mount antennas, few believed that by extending the fringe of cellular networks by current users employ these options. 3-5 km, a significant portion of the unmet demand for 1450. If one limits the present system to a five hop cellular service will be satisfied. distance from fixed cellular infrastructure (or more accu 1446. If we assume that a significant portion of the rately, permits the system to deny service to nodes more than mobile nodes are power constrained (e.g., battery operated), five hops away) then the routing requirements and node that is, retransmission of packets imposes a power cost, then complexity may be substantially simplified. We also pre the stability of the mobile ad hoc network and cooperation Sume that each node has geolocation capability, and there with its requirements will depend on properly incentivizing fore can provide both its location and velocity vector. This intermediary nodes to allocate their resources to the net is reasonable, since the FCC E911 mandate provides for work. Since this incentive is provided in a commercial geolocation of handsets within range of the cellular infra context, that is, the cellular service is a commercial enter structure, and GPS is a one option to provide this feature. prise with Substantial cash flow, a real economy with mon 1451. The ad hoc communications can occur using a etary incentives may be provided. Under Such circum licensed or unlicensed band. For example, since we presume stances, it is relatively straightforward to allocate costs and that nodes are beyond range of a fixed cellular tower (except benefits between the competing interests to achieve consis the closest node), the ad hoc network may reuse licensed tent and apparent incentives. On the other hand, the cost of bandwidth in the uncovered region. The ad hoc communi this additional process must be commensurate with the cations may also occur in unlicensed spectrum, Such as the benefits provided, or else the ad hoc network will become 2.4 GHz ISM band. unreliable. 1452. In order to provide optimum compensation, two 1447 While the true economics of the costs and value issues are confronted. First, the total compensation paid; and functions for the participants are beyond the scope of this second, the distribution of payments between the interme paper, some estimates are available. The issue is: are users diaries. The VCG auction is a known means for optimizing willing to pay for extended cellular reach? If so, do they a payment which must be distributed between a number of value the benefits commensurate with the overall costs, participants. In this case, each potential intermediary places including service fees, hardware, and ad hoc cooperative a “bid”. A multi-factorial optimization is performed to burdens? As such, care must be exercised to define com determine the lowest cost set which provides sufficient petitive compensation or the business will be inefficient. services. Since this extension is driven by the cellular network 1453. In a cellular system, each subscriber typically pur operator, a suitable return on investment is mandated. chases a number of minute units on a monthly recurring 1448 Many analyses and studies have concluded that charge basis. Compensation might therefore be based on voluntary ad hoc networks are efficient when the incentives minutes or money. Since there is a Substantial disincentive to cooperate with the network goals are aligned and Sufi to exceed the number of committed minutes, providing a cient to incentivize users accordingly. If the reward for Surplus of minutes may not provide a significant incentive, cooperation is optimum, then the network will benefit by because the user will rarely exceed the committed amount. increased coverage and reliability, each node will benefit Monetary incentives, on the other hand, must be coupled to from increased utility, and intermediary nodes will specifi a higher monthly recurring fee, since the proposal would by cally benefit through compensation. Due to the technical unprofitable otherwise. possibility for potential intermediaries to fail to either par 1454. A more direct scheme provides an economy for ticipate in network administration or operation, while taking multihop networks somewhat independent from the cellular advantage of the network as a beneficiary, the promotion of system economy. That is, nodes that participate as interme network availability as an incentive for cooperation is typi diary, may also participate as a principal to the information cally itself insufficient incentive to assure cooperation. The communication, while those who abstain from intermediary particular cost of the limited power resource for potential activities are denied access to the network extension as a intermediaries makes non-cooperation a particularly impor principal. tant factor. On the other hand, the presumption of power cost 1455 While, on a theoretical basis, optimization of both as a critical factor may be accurate only in Some circum price and distribution would be considered useful, in a stances: In many cases, a cheap power source is available, practical system, it may be useful to make simplifying Such as in a home or office, or in a vehicle, making other presumptions and simplifications. For example, while a factors more important. VCG auction may provide an optimal cost and distribution 1449. It is noted that, in a cellular telephone system, the of compensation, in a commercial network, a degree of reasonable acts of a user which might undermine the net certainty may actually prove advantageous. For example, a work are limited. Clearly, the user can choose a different fixed compensation per hop or per milli Watt-Second may network or provider. The user may turn off his phone or prove both fair and reasonable. make it unavailable. The user may abuse the service con 1456. Likewise, a degree of certainty over cost would be tract, taking advantage of promotions or “free” access to the beneficial over an “optimal’ cost. On the other hand, fixed US 2006/0167784 A1 Jul. 27, 2006 59 cost and fixed compensation are inconsistent in a revenue principal issues impeding deployment are the inherent com neutral system. Even if the cellular carrier subsidizes the plexity of the system, as well as the overhead required to extension operation, there is little rationale for making the continuously optimize the system. Further work will allow usage at the fringe insensitive to cost, other than the relief a determination of a set of simplifying presumptions to from uncertainty, which will tend to increase fringe usage, reduce protocol overhead and reduce complexity. Hierarchal and the scope of the Subsidy cost. considerations can be imposed to alter the optimization of the system, which would be expected to provide only a small 1457. Therefore, it is realistic for a node requesting perturbation to the efficient and optimal operation of the extension service to apply a value function to define a system according to a pure VCG protocol. This paper maximum payment for service. The payment is therefore proposes a marketplace auction with competition between dependent on system cost, alleviating the requirement for potential buyers and potential sellers, with the economic Subsidy, but also dependent on need. surplus distributed between parties which must defer to 1458. In the typical case, the load on the extension active participants. network will be low, since if demand were high, the fixed 1465. The ad hoc network does not exist in a vacuum. infrastructure would likely be extended to this region. On the There are various competing interests seeking to use the other hand, there may be cases where demand is high, and same bandwidth, and technological Superiority alone does therefore there is competition for access to the network, not assure dominance and commercial Success. Game theory leading to a need to arbitrate access. may also be used as a tool to analyze the entities which seek 1459. In general, where economic demand is high, there to deploy ad hoc networks, especially where they compete. is a tendency to recruit new Sources of Supply. That is, the REFERENCES (SECOND EMBODIMENT) system may operate in two modes. In a first, low demand mode, costs are based on a relatively simple algorithm, with 1466 1. F. P. Kelly, A. Maulloo, and D. Tan. Rate control a monetary cap. In a second mode, costs are competitive in communication networks: shadow prices, proportional (and in excess of the algorithmic level), with compensation fairness and stability. Journal of the Operational Research also being competitive. The Surplus, in this case, would Society, 49, 1998. http://citeseerist.psu.edu/ generally be allocated to the cellular carrier, since this kelly98rate.html represents the commercial profit of the enterprise. The first 1467 2. J. MacKie-Mason and H. Varian. Pricing con mode also serves another purpose; under lightly loaded gestible network resources. IEEE Journal on Selected Areas conditions, the market may be thin, and therefore pricing in Communications, 13(7): 1141-1149, 1995. unstable. Therefore, the imposition of fixed pricing leads to 1468 3. Courcoubetis, C., Siris, V. A. and Stamoulis, G. reduced pricing risk. D. Integration of pricing and flow control for available bit 1460. In the second mode, an intended user specifies his rate services in ATM networks. In Proceedings IEEE Globe demand as a maximum price and demand function, that is, com '96, pp. 644-648. London, UK. http://citeseerist.ps a bid based on, for example, a value of the communication. u.edu/courcoubetis96integration.html Generally, this would be set by the user in advance as a static value or relatively simple function representing the contex 1469 4. L. Buttyan and J.-P. Hubaux. Rational tual value of the communication. The actual price may be, exchange—a formal model based on game theory. In Pro for example, the bid price less otherwise attributable dis ceedings of the 2nd International Workshop on Electronic count under the first mode based on the maximum number Commerce (WELCOM), November 2001. http://citese of hops, etc. The intermediate nodes set forth their bids in the erist.psu.edu/anO1 rational.html manner of a VCG auction, with each bid presumably 1470) 5. P. Michiardi and R. Molva. Game theoretic exceeding the first mode compensation. The VCG optimi analysis of security in mobile ad hoc networks. Technical Zation may be corrected for quality of service factors and Report RR-02-070. Institut Eurecom, 2002. anticipated network stability. 1471 6. P. Michiardi and R. Molva. A game theoretical 1461 Clearly, the economics of the system are substan approach to evaluate cooperation enforcement mechanisms tially under the control of the cellular carrier, who may offer in mobile ad hoc networks. In Proceedings of WiOptO3, “plans” and “services” for their customers, thus providing an March 2003. alternative to the usage-based bidding process, at least for 1472 7. Michiardi, P., Molva, R.: Making greed work in some users. The VCG process, however, remains useful for mobile ad hoc networks. Technical report, Institut Eur ecom compensating intermediaries. (2002) http://citeseer.ist.psu.edu/michiardiO2making.html 1462. The intermediary chooses its bid for providing 1473 8. S. Shenker. Making greed work in networks: A packet forwarding services based on a number of factors, game-theoretic analysis of switch service disciplines. IEEE/ Such as anticipated power cost, opportunity cost, etc. These ACM Transactions on Networking, 3(6):819-831, Decem bids are determined automatically, and therefore do not ber 1995. consume Substantial human efforts. 1474) 9. A. B. MacKenzie and S. B. Wicker. Selfish users 1463 Conclusion in aloha: A game-theoretic approach. In Vehicular Technol 1464 Game theory is a useful basis for analyzing ad hoc ogy Conference, 2001. VTC 2001 Fall. IEEE VTS 54th, networks, and understanding the behavior of complex net volume 3, October 2001. works of independent nodes. By presuming a degree of 1475) 10. J. Crowcroft, R. Gibbens, F. Kelly, and S. choice and decision-making by nodes, we obtain an analysis Ostring. Modelling incentives for collaboration in mobile ad that is robust with respect to such considerations. The hoc networks. In Proceedings of WiOptO3, 2003. US 2006/0167784 A1 Jul. 27, 2006 60
1476) 11. Durga P. Satapathy and Jon M. Peha, Perfor Wireless Ad Hoc Network Routing Protocols. In Proc. of the mance of Unlicensed Devices With a Spectrum Etiquette.” ACM/IEEE MobiCom, October 1998. http://citeseerist.ps Proceedings of IEEE Globecom, November 1997, pp. 414 u.edu/broch98performance.html 418. http://citeseerist.psu.edu/ satapathy.97 performance.html 1492) 27. Juha Leino, “Applications of Game Theory in Ad Hoc Network', Master's Thesis, Helsinki University Of 1477 12. Garrett Hardin. The Tragedy of the Commons. Technology (2003). Science, 162:1243-1248, 1968. Alternate Location: http:// diedff.com/page95.htm. 1493. 28. J. Shneidman and D. Parkes, “Rationality and Self-Interest in Peer to Peer Networks'. In Proc. 2nd Int. 1478 13. Walsh, W. and M. Wellman (1998). A market Workshop on Peer-to-Peer Systems (IPTPS03), 2003, cite protocol for decentralized task allocation, in “Proceedings of seer.nj.nec.com/shneidman03rationality.html the Third International Conference on Multi-Agent Sys tems, pp. 325-332, IEEE Computer Society Press, Los 1494. 29. V. Rodoplu and H.-Y. Meng. Minimum energy Alamitos. mobile wireless networks. IEEE Journal on Selected Areas in Communications, 17(8): 1333-1344, August 1999. 1479) 14. M. Mauve, J. Widmer, and H. Hartenstein. A Survey on position-based routing in mobile ad hoc networks. 1495 30. S. Singh, M. Woo, and C. S. Raghavendra. IEEE Network Magazine, 15(6):30-39, November 2001. Power-aware routing in mobile ad hoc networks. In Pro http://citeseerist.psu.edu/article/mauveO1 survey.html ceeding of MOBICOM 1998, pages 181-190, 1998. 1480 15. Z. Haas. A new routing protocol for reconfig 1496 31. A. Urpi, M. Bonuccelli, and S. Giordano. urable wireless networks. In IEEE 6th International Confer Modelling cooperation in mobile ad hoc networks: a formal ence on Universal Communications Record, Volume 2, description of selfishness. In Proceedings ofWiOptO3, pages 562-566, October 1997. March 2003. 1481 16.X. Hong, K. Xu, and M. Gerla. Scalable routing 1497. 32. A. van den Nouweland, P. Borm, W. van protocols for mobile ad hoc networks. IEEE Networks, Golstein Brouwers, R. Groot Bruinderink, and S. Tijs. A 16(4):11-21, July 2002. game theoretic approach to problems in telecommunication. 1482) 17. D. Johnson, D. Maltz, and Y-C. Hu. The Management Science, 42(2):294-303, February 1996. dynamic source routing protocol for mobile ad hoc net 1498 33. P. Michiardi and R. Molva. Core: A collabora works, April 2003. http://www.ietforg/internet-drafts/draft tive reputation mechanism to enforce node cooperation in ietf-manet-dsr-09.txt mobile ad hoc networks. In Communication and Multimedia 1483) 18. S.-J. Lee, W. Su, J. Hsu, M. Gerla, and R. Security 2002 Conference, 2002. Bagrodia. A performance comparison study of ad hoc wire 1499 34. Vickrey, W. (1961). Counterspeculation, auc less multicast protocols. In Proceedings of IEEE INFOCOM tions, and competitive sealed tenders, Journal of Finance 16, 2000, pages 565-574, March 2000. 8-37. 1484) 19. K. Mase, Y. Wada, N. Mori, K. Nakano, M. Sengoku, and S. Shinoda. Flooding schemes for a universal 1500 35. Clarke, E. H. (1971). Multipart pricing of ad hoc network. In Industrial Electronics Society, 2000. public goods, Public Choice 11, 17-33. ECON 2000., v.2, pp. 1129-1134, 2000. 1501 36. Felix Brandt and Gerhard Wei B. Antisocial 1485. 20. R. Ogier, F. Templin, and M. Lewis. Topology Agents and Vickrey Auctions. In Pre-proceedings of the dissemination based on reversepath forwarding, October Eighth International Workshop on Agent Theories, Archi 2003. Vesuvio.ipv6.cselt.it/internet-drafts/draft-ietf-manet tectures, and Languages (ATAL-2001), pages 120-132. thrpf-11.txt 2001. 1486 21. A. Orda, R. Rom, and N. Shimkin. Competitive 1502 37. Tuomas Sandholm. Limitations of the Vickrey routing in multi-user communication networks. IEEE/ACM Auction in Computational Multiagent Systems. In Proceed Transactions on Networking, 1 (5):510-521, October 1993. ings of the 2nd International Conference on Multi-Agent 1487. 22. C. Perkins, E. Belding-Royer, and S. Das. Ad Systems (ICMAS). AAAI Press, 1996. Menlo Park, Calif. hoc on-demand distance vector (AODV) routing. Request 1503) 38. Ron Lavi, Ahuva Mu'alem, and Noam Nisan, for comments 3561, Internet Engineering Task Force, 2003. “Towards a Characterization of Truthful Combinatorial Auc 1488) 23. C. E. Perkins, editor. Ad Hoc Networking. tions, citeseerist.psu.edu/lavi()3towards.html Addison-Wesley, Boston, 2001. 1504 39. Moulin, H. (1999). Incremental cost sharing: 1489 24. E. Royer and C.-K. Toh. A review of current characterization by strategy proofness, Social Choice and routing protocols for ad hoc mobile wireless networks. IEEE Welfare 16, 279-320. Personal Communications, 6(2):46-55, April 1999. 1505) 40. Moulin, H. and S. Shenker (1997). Strat 1490) 25. Holger Fü?sler, Hannes Hartenstein, Dieter egyproof Sharing of Submodular Costs: Budget Balance Vollmer, Martin Mauve, Michael Käsemann, Location Versus Efficiency, to appear in Economic Theory. Based Routing for Vehicular Ad-Hoc Networks, Reihe Infor www.aciri.org/shenker/cost.ps matik 3/2002, http://citeseerist.psu.edu/560036.html 1506) 41. Moulin, Hervé, and Scott Shenker (2001). 1491) 26. J. Broch, D. A. Maltz, D. B. Johnson, Y. C. Hu, “Strategyproof Sharing of Submodular Costs: Budget Bal and J. Jetcheva. A Performance Comparison of Multi-Hop ance versus Efficiency.” Economic Theory 18, 511-533. US 2006/0167784 A1 Jul. 27, 2006
1507 42. Feigenbaum, Joan, Christos Papadimitriou, 1520 55. N. Ben Salem, L. Buttyan, J. P. Hubaux, and Rahul Sami, and Scott Shenker (2002). “A BGP-based Jakobsson M. A charging and rewarding scheme for packet Mechanism for Lowest-Cost Routing.” In Proc. 21st Sym forwarding. In Proceeding of Mobihoc, June 2003. posium on Principles of Distributed Computing, ACM 1521 56. Rivest, R. L., A. Shamir, PayWord and Micro Press, 173-182. Mint: Two simple micropayment schemes, also presented at 1508) 43. J. Feigenbaum and S. Shenker. Distributed the RSA 96 conference, http://theory.lcs.mit.edu/rivest/ algorithmic mechanism design: Recent results and future RivestShamirmpay.ps, http://citeseerist.psu.edu/ directions. In Proc. 6th Int’l Workshop on Discrete Algo rivest96payword.html rithms and Methods for Mobile Computing and Communi 1522) 57. Silvio Micali and Ronald Rivest. Micropay cations, pages 1-13, Atlanta, Ga., September 2002. ments revisited. In Bart Preneel, editor, Progress in Cryp 1509 44. Nisan, N. and A. Ronen (2000). Computation tology—CT-RSA 2002, volume 2271 of Lecture Notes in ally Feasible VCG Mechanisms, to be presented at “Games Computer Science. Springer-Verlag, Feb. 18-22 2002. http:// 2000.’ http:/www.cs.huji.ac.il/-noam/vcgbased.ps citeseerist.psu.edu/micali02micropayments.html 1510 45. Tuomas Sandholm. Limitations of the Vickrey Auction in Computational Multiagent Systems. In Proceed Third Embodiment ings of the 2nd International Conference on Multi-Agent 1523) A third embodiment of the invention, described Systems (ICMAS). AAAI Press, 1996. Menlo Park, Calif. below, represents a system which may employ a self 1511) 46. C. Jason Woodard and David C. Parkes, 1 st organizing network to convey information between mobile Workshop on the Economics of P2P systems, Strategyproof nodes. It is expressly understood that the concepts set forth Mechanisms for Ad Hoc Network Formation, 2003, above in the first and second embodiments are directly www.sims.berkeley.edu/research/conferences/p2pecon/pa applicable, and each aspect of the third embodiment may be pers/so-woodard.pdf extended using the hierarchal principles and modifications, in a consistent manner, to achieve the advantages described 1512 47. D. C. Parkes. Iterative Combinatorial Auctions: herein. That is, while the third embodiment generally Achieving Economic and Computational Efficiency (Chap describes peer nodes, the extension of the systems and ter 2). PhD thesis, Univesity of Pennsylvania, May 2001. methods to non-peer nodes is specifically envisioned and http://www.eecs.harvard.edu/~parkes/pubs/ch2.ps. encompassed. 1513 48. L. Blazevic, L. Buttyan, S. Capkun, S. Giord 1524. This patent builds upon and extends aspects of U.S. iano, J.-P. Hubaux, and J.-Y. Le Boudec. Self-organization in Pat. No. 6.252,544 (Hoffberg), Jun. 26, 2001, and U.S. Pat. mobile ad-hoc networks: the approach of terminodes. IEEE No. 6,429,812, Aug. 6, 2002, which are expressly incorpo Communications Magazine, 39(6):166-174, June 2001. rated herein by reference in its entirety. See, also, U.S. Pat. No. 6,397,141 (Binnig, May 28, 2002, Method and device 1514. 49. M. Jakobsson, J. P. Hubaux, and L. Huttyan. A for signalling local traffic delays), expressly incorporated micro-payment Scheme encouraging collaboration in multi herein by reference, which relates to a method and an hop cellular networks. In Proceedings of Financial Crypto apparatus for signalling local traffic disturbances wherein a 2003, January 2003. decentralised communication between vehicles, which is 1515 50. J. P. Hubaux, et al., “Toward Self-Organized performed by exchanging their respective vehicle data. Mobile Ad Hoc Networks: The Terminodes Project', IEEE Through repeated evaluation of these individual vehicle Communications, 39(1), 2001. http://citeseerist.psu.edu/ data, each reference vehicle may determine a group of hubaux01 toward.html vehicles having relevance for itself from within a maximum group of vehicles and compare the group behavior of the 1516) 51. Buttyan, L., and Hubaux, J.-P. Stimulating relevant group with its own behavior. The results of this Cooperation in Self-Organizing Mobile Ad Hoc Networks. comparison are indicated in the reference vehicle, whereby Tech. Rep. DSC/http://citeseerist.psu.edu/ a homogeneous flow of traffic may be generated, and the buttyan01stimulating.html occurrence of accidents is reduced. See, also U.S. Pat. No. 4,706,086 (November, 1987 Panizza 340/902), and U.S. Pat. 1517 52. Levente Buttyan and Jean-Pierre Hubaux, No. 5,428,544 (June, 1995 Shyu 701/117), U.S. Pat. No. “Enforcing Service Availability in Mobile Ad-Hoc WANs, 6,473,688 (Kohno, et al., Oct. 29, 2002, Traffic information 1st IEEE/ACM Workshop on Mobile Ad Hoc Networking transmitting system, traffic information collecting and dis and Computing (MobiHOC http://citeseerist.psu.edu/ tributing system and traffic information collecting and dis buttyan00enforcing.html tributing method), U.S. Pat. No. 6,304,758 (October, 2001, 1518, 53. L. Buttyan and J.-P. Hubaux. Nuglets: a virtual Iierbig et al., 701/117): U.S. Pat. No. 6,411.221 (January, currency to stimulate cooperation in self-organized ad hoc 2002, Horber, 701/117): U.S. Pat. No. 6,384,739 (May, networks. Technical Report DSC/2001. http://citeseerist.p- 2002, Robert, Jr., 701/117): U.S. Pat. No. 6,401,027 (June, su.edu/article/buttyan01 nuglets.html 2002, Xa et al., 701/117): U.S. Pat. No. 6,411,889 (June, 2002, Mizunuma et al., 701/117), U.S. Pat. No. 6,359,571 1519 54. Mario Cagal, Jean-Pierre Hubaux, and Chris (Endo, et al., Mar. 19, 2002, Broadcasting type information tian Enz. Minimum-energy broadcast in all-wireless net providing system and travel environment information col works: Np-completeness and distribution issues. In The lecting device); U.S. Pat. No. 6,338,011 (Furst, et al., Jan. 8, Eighth ACM International Conference on Mobile Comput 2002, Method and apparatus for sharing vehicle telemetry ing and Networking (MobiCom 2002). http://citeseerist.p- data among a plurality of users over a communications su.edu/cagal.02minimumenergy.html network); U.S. Pat. No. 5,131,020 (July, 1992, Liebesny et US 2006/0167784 A1 Jul. 27, 2006 62 al., 455/422); U.S. Pat. No. 5,164,904 (November, 1992, database, to assist navigation functions. Systems which Sumner, 701/117): U.S. Pat. No. 5,539,645 (July, 1996, integrate GPS, GLONASS, LORAN or other positioning Mandhyan et al., 701/119): U.S. Pat. No. 5,594.779 (Janu systems into vehicular guidance systems are well known, ary, 1997, Goodman, 455/4); U.S. Pat. No. 5,689.252 and indeed navigational purposes were prime motivators for (November, 1997, Ayanoglu et al., 340/991): U.S. Pat. No. the creation of these systems. 5,699,056 (December, 1997, Yoshida, 340/905); U.S. Pat. 1529. Environmental sensors are well known. For No. 5,864,305 (January, 1999, Rosenquist, 340/905); U.S. example, sensing technologies for temperature, weather, Pat. No. 5,889,473 (March, 1999, Wicks, 340/825); U.S. Pat. No. 5,919.246 (July, 1999, Waizmann et al., 701/209); U.S. object proximity, location and identification, vehicular traffic Pat. No. 5,982.298 (November, 1999, Lappenbusch et al., and the like are well developed. In particular, known sys 340/905); U.S. Pat. No. 4,860,216 (August, 1989, Linsen tems for analyzing vehicular traffic patterns include both mayer, 342/159): U.S. Pat. No. 5,302,955 (April, 1994, stationary and mobile sensors, and networks thereof. Most Schutte et al., 342/59); U.S. Pat. No.5,809,437 (September, often, such networks provide a stationary or centralized 1998, Breed, 701/29); U.S. Pat. No. 6,115,654 (September, system for analyzing traffic information, which is then 2000, Eid et al., 701/34); U.S. Pat. No. 6,173,159 (January, broadcast to vehicles. 2001, Wright et al., 455/66); and Japanese Patent Document 1530 Encryption technologies are well known and Nos. JP 9-236650 (September, 1997); 10-84430 (March, highly developed. These are generally classified as being 1998); 5-151496 (June, 1993); and 11-183184 (July, 1999), symmetric key, for example the Data Encryption Standard each of which is expressly incorporated herein by reference. (DES), and the more recent Advanced Encryption Standard See also: Martin E. Liggins, II, et al., “Distributed Fusion (AES), in which the same key is used for encryption as Architectures and Algorithms for TargetTracking'. Proceed decryption, and asymmetric key cryptography, in which ings of the IEEE, vol. 85, No. 1, (XP-002166088) January, different and complementary keys are used to encrypt and 1997, pp. 95-106.: D. M. Hosmer, “Data-Linked Associate decrypt, in which the former and the latter are not derivable Systems, 1994 IEEE International Conference on Systems, from each other (or one from the other) and therefore can be Man, and Cybernetics. Humans, Information and Technol used for authentication and digital signatures. The use of ogy (Cat. No. 94CH3571-5), Proceedings of IEEE Interna asymmetric keys allows a so-called public key infrastruc tional Conference on Systems, Man and Cybernetics, San ture, in which one of the keys is published, to allow Antonio, Tex., vol.3, (XP-002166089) (1994), pp. 2075 communications to be directed to a possessor of a comple 2O79. mentary key, and/or the identity of the sender of a message to be verified. Typical asymmetric encryption systems 1525. One aspect of the invention provides a communi include the Rivest-Shamir-Adelman algorithm (RSA), the cations system, method and infrastructure. According to one Diffie-Hellman algorithm (DH), elliptic curve encryption preferred embodiment, an ad hoc, self organizing, cellular algorithms, and the so-called Pretty Good Privacy (PGP) radio system (sometimes known as a "mesh network”) is algorithm. provided. Advantageously, high gain antennas are employed, preferably electronically steerable antennas, to 1531. One embodiment of the invention provides a sys provide efficient communications and to increase commu tem that analyzes both a risk and an associated reliability. nications bandwidth, both between nodes and for the system Another embodiment of the invention communicates the risk comprising a plurality of nodes communicating with each and associated reliability in a manner for efficient human other. See, U.S. Pat. No. 6,507,739 (Gross, et al., Jan. 14, comprehension, especially in a distracting environment. See, 2003), expressly incorporated herein by reference. U.S. Pat. Nos. 6,201,493; 5,977,884; 6,118,403; 5,982,325: 5,485,161; WO0077539, each of which is expressly incor 1526 In general, time-critical, e.g., voice communica porated herein by reference, and the Uniden GPSRD (see tions require tight routing to control communications Uniden GPSRD User's Manual, expressly incorporated latency. On the other hand, non-time critical communica herein by reference). See, also U.S. Pat. Nos. 5,650,770; tions generally are afforded more leeway in terms of com 5,450,329; 5,504,482; 5,504,491; 5,539,645; 5,929,753; munications pathways, including a number of “hops'. 5,983,161; 6,084,510; 6,255,942; 6.225,901; 5,959,529; retransmission latency, and out-of-order packet communi 5,752,976; 5,748,103; 5,720,770; 6,005,517; 5,805,055; cation tolerance, between the source and destination or fixed 6,147,598; 5,687.215; 5,838,237; 6,044,257; 6,144,336; infrastructure, and quality of communication pathway. Fur 6,285,867; 6,340,928; 6,356,822; 6,353,679 each of which ther, it is possible to establish redundant pathways, espe is expressly incorporated herein by reference. cially where communications bandwidth is available, mul tiple paths possible, and no single available path meets the 1532 Statistical Analysis entire communications requirements or preferences. 1533. It is understood that the below analysis and ana 1527 Technologies for determining a position of a lytical tools, as well as those known in the art, may be used mobile device are also well known. Most popular are radio individually, in Sub-combination, or in appropriate combi triangulation techniques, including artificial satellite and nation, to achieve the goals of the invention. These tech terrestrial transmitters or receivers, dead reckoning and niques may be implemented in dedicated or reprogram inertial techniques. Advantageously, a satellite-based or aug mable/general purpose hardware, and may be employed for mented satellite system, although other Suitable geolocation low level processing of signals, such as in digital signal systems are applicable. processors, within an operating system or dynamic linked libraries, or within application software. Likewise, these 1528 Navigation systems are also well known. These techniques may be applicable, for example, to low level data systems generally combine a position sensing technology processing, system-level data processing, or user interface with a geographic information system (GIS), e.g., a mapping data processing. US 2006/0167784 A1 Jul. 27, 2006
1534. A risk and reliability communication system may probability distribution. Continuous variables, however, be useful, for example, to allow a user to evaluate a set of have an associated probability density function ("density'). events in statistical context. Most indicators present data by Where an event is a set of possible outcomes, the density means of a logical indicator or magnitude, as a single value. p(x) for a variable 'x' and events “a” and “b' is defined as: Scientific displays may provide a two-dimensional display of a distribution, but these typically require significant user focus to comprehend, especially where a multimodal distri - p(a six s b) bution is represented. User displays of a magnitude orbinary p(x) = Lift. value typically do not provide any information about a likelihood of error. Thus, while a recent positive warning of the existence of an event may be a reliable indicator of the 1538 where p(asxsb) is the probability that x lies actual existence of the event, the failure to warn of an event between a and b. Conventional systems for generating does not necessarily mean that the event does not exist. Bayesian networks cannot use continuous variables in their Further, as events age, their reliability often decreases. nodes. 1535 A Bayesian network is a representation of the 1539. There are two conventional approaches for con probabilistic relationships among distinctions about the structing Bayesian networks. Using the first approach (“the world. Each distinction, Sometimes called a variable, can knowledge-based approach'), first the distinctions of the take on one of a mutually exclusive and exhaustive set of world that are important for decision making are deter possible states. Associated with each variable in a Bayesian mined. These distinctions correspond to the variables of the network is a set of probability distributions. Using condi domain of the Bayesian network. The “domain of a Baye tional probability notation, the set of probability distribu sian network is the set of all variables in the Bayesian tions for a variable can be denoted by p(x,at, X), where “p” network. Next the dependencies among the variables (the refers to the probability distribution, where “t, denotes the arcs) and the probability distributions that quantify the parents of variable X, and where x denotes the knowledge strengths of the dependencies are determined. of the expert. The Greek letter “X indicates that the Baye sian network reflects the knowledge of an expert in a given 1540. In the second approach ("called the data-based field. Thus, this expression reads as follows: the probability approach'), the variables of the domain are first determined. distribution for variable X, given the parents of X, and the Next, data is accumulated for those variables, and an algo knowledge of the expert. For example, X is the parent of rithm is applied that creates a Bayesian network from this X. The probability distributions specify the strength of the data. The accumulated data comes from real world instances relationships between variables. For instance, if X has two of the domain. That is, real world instances of decision states (true and false), then associated with X is a single making in a given field. Conventionally, this second probability distribution p(x|y)p and associated with X are approach exists for domains containing only discrete vari two probability distributions p(X,X=ty) and p(XX=fy). ables. 1536 A Bayesian network is expressed as an acyclic 1541 U.S. application Ser. No. 08/240,019 filed May 9, directed graph where the variables correspond to nodes and 1994 entitled “Generating Improved Belief Networks” the relationships between the nodes correspond to arcs. The describes a system and method for generating Bayesian arcs in a Bayesian network convey dependence between networks (also known as “belief networks') that utilize both nodes. When there is an arc between two nodes, the prob expert data received from an expert ("expert knowledge’) ability distribution of the first node depends upon the value and data received from real world instances of decisions of the second node when the direction of the arc points from made (“empirical data'). By utilizing both expert knowledge the second node to the first node. Missing arcs in a Bayesian and empirical data, the network generator provides an network convey conditional independencies. However, two improved Bayesian network that may be more accurate than variables indirectly connected through intermediate vari conventional Bayesian networks or provide other advan ables are conditionally dependent given lack of knowledge tages, e.g., ease of implementation and lower reliance on of the values (“states') of the intermediate variables. In other “expert” estimations of probabilities. Likewise, it is known words, sets of variables X and Y are said to be conditionally to initiate a network using estimations of the probabilities independent, given a set of variables Z, if the probability (and often the relevant variables), and Subsequently use distribution for X given Z does not depend on Y. If Z is accumulated data to refine the network to increase its empty, however, X and Y are said to be “independent’ as accuracy and precision. opposed to conditionally independent. If X and Y are not 1542 Expert knowledge consists of two components: an conditionally independent, given Z, then X and Y are said to equivalent sample size or sizes (“sample size’), and the prior be conditionally dependent given Z. probabilities of all possible Bayesian-network structures 1537. The variables used for each node may be of dif (“priors on structures'). The effective sample size is the ferent types. Specifically, variables may be of two types: effective number of times that the expert has rendered a discrete or continuous. A discrete variable is a variable that specific decision. For example, a doctor with 20 years of has a finite or countable number of states, whereas a experience diagnosing a specific illness may have an effec continuous variable is a variable that has an effectively tive sample size in the hundreds. The priors on structures infinite number of states. An example of a discrete variable refers to the confidence of the expert that there is a rela is a Boolean variable. Such a variable can assume only one tionship between variables (e.g., the expert is 70% sure that of two states: “true' or “false.” An example of a continuous two variables are related). The priors on structures can be variable is a variable that may assume any real value decomposed for each variable-parent pair known as the between -1 and 1. Discrete variables have an associated “prior probability” of the variable-parent pair. Empirical US 2006/0167784 A1 Jul. 27, 2006 64 data is typically stored in a database. The database may 1548 However, this method is exponential in time, so the contain a list of the observed state of some or all of the more efficient forward-backward algorithm is used in prac variables in the Bayesian network. Each data entry consti tice. The following algorithm defines the forward variable a tutes a case. When one or more variables are unobserved in and uses it to generate Pr(O2)(at are the initial state prob a case, the case containing the unobserved variable is said to abilities, a are the state transition probabilities, and b are the have “missing data.” Thus, missing data refers to when there output probabilites). are cases in the empirical data database that contain no observed value for one or more of the variables in the 1549 a (text missing or illegible when filed)= domain. An assignment of one state to each variable in a set It b (O), for all states i of variables is called an “instance' of that set of variables. Thus, a “case' is an instance of the domain. The “database' is the collection of all cases. 1543. Therefore, it is seen that Bayesian networks can be used to probabilistically model a problem, in a mathematical form. This model may then be analyzed to produce one or otherwise L=text missing or illegible when filed) more outputs representative of the probability that a given fact is true, or a probability density distribution that a 1550 Calculating a? ) along the time axis, for t=2,.. variable is at a certain value. . I, and all states j, compute 1544. A review of certain statistical methods is provided below for the convenience of the reader, and is not intended to limit the scope of methods, of statistical of other type, at (i) = ai_1(i)ajibi (Ok) which may be employed in conjunction with the system and |X, a to method according to the present invention. It is understood that these mathematical models and methods may be imple mented in known manner on general purpose computing 1551 Final probability is given by platforms, for example as a compiled application in a real-time operating system such as RT Linux, QNX, Ver sions of Microsoft Windows, or the like. Further, these Pr(OIA) = X ar(i) techniques may be implemented as applets operating under ieSp Matlab or other scientific computing platform. Alternately, the functions may be implemented natively in an embedded control system or on a microcontroller. 1552) The first step initializes the forward variable with 1545. It is also understood that, while the mathematical the initial probability for all states, while the second step methods are capable of producing precise and accurate inductively steps the forward variable through time. The results, various simplyfying presumptions and truncations final step gives the desired result Pr(O2) and it can be shown may be employed to increase the tractability of the problem by constructing a lattice of states and transitions through to be solved. Further, the outputs generally provided accord ing to preferred embodiments of the present invention are time that the computation is only order O(NT). The back relatively low precision, and therefore higher order approxi ward algorithm, using a process similar to the above, can mation of the analytic Solution, in the case of a rapidly also be used to compute Pr(O2) and defines the convenience convergent calculation, will often be sufficient. variable f3. 1546 A time domain process demonstrates a Markov 1553. The estimation problem concerns how to adjust to property if the conditional probability density of the current maximize Pr(O2) given an observation sequence O. Given event, given all present and past events, depends only on the an initial model, which can have flat probabilities, the jth most recent events. If the current event depends solely on forward-backward algorithm allows us to evaluate this prob the most recent past event, then the process is a first order ability. All that remains is to find a method to improve the Markov process. There are three key problems in HMM use: initial model. Unfortunately, an analytical Solution is not evaluation, estimation, and decoding. The evaluation prob known, but an iterative technique can be employed. lem is that given an observation sequence and a model, what is the probability that the observed sequence was generated 1554 Using the actual evidence from the training data, a by the model (Pr(O2)) If this can be evaluated for all new estimate for the respective output probability can be competing models for an observation sequence, then the assigned: model with the highest probability can be chosen for rec ognition. 1547 Pr(O2) can be calculated several ways. The naive way is to sum the probability over all the possible state b;(k) = - sequences in a model for the observation sequence: 2. y, (i)
T Pr(OIA) =X abaab (O.) 1555 where text missing or illegible when filed is aiiS i-l defined as the posterior probability of being in state i at time t given the observation sequence and the model. Similarly, the evidence can be used to develop a new estimate of the US 2006/0167784 A1 Jul. 27, 2006 probability of a state transition (text missing or illegible specify the parameters of the probability density function. when filed) and initial state probabilities (text missing Usually the probability density is approximated by a or illegible when filed). weighted sum of M Gaussian distributions N. 1556. Thus all the components of model (w) can be re-estimated. Since either the forward or backward algo rithm can be used to evaluate versus the previous estimation, i the above technique can be used iteratively to converge the bi(a) = X. e-Mer X. a) model to some limit. While the technique described only n=1 fin handles a single observation sequence, it is easy to extend to a set of observation sequences. 1557. The Hidden Markov Model is a finite set of states, 1568 where, each of which is associated with a (generally multidimen c=weighting coefficients sional) probability distribution. Transitions among the states |lin-mean vectors are governed by a set of probabilities called transition XE =Covariance matrices probabilities. In a particular state an outcome or observation m can be generated, according to the associated probability 1569) c, should satisfy the stochastic constrains, distribution. It is only the outcome, not the state visible to an ce0, 1 sisN, 1sms M external observer and therefore states are "hidden' to the 1570) and outside; hence the name Hidden Markov Model. 1558. In order to define an HMM completely, following elements are needed. 1559 The number of states of the model, N. 1560. The number of observation symbols in the alpha bet, M. If the observations are continuous then M is infinite. 1571. The initial state distribution, (1561) A set of state transition probabilities A={a} a=p(q1=ilg-i, 1 sij sN, where, 1562 where q denotes the current state. JI;=p(q =i, 1 sisN 1563 Transition probabilities should satisfy the normal Therefore we can use the compact notation stochastic constraints, =9A, B, JI) to denote an HMM with discrete probability distributions, while aii > 0, 1 s i, is N and W=(A.Cim-lim in J)
W to denote one with continuous densities. X ai = 1, 1 s is N 1572 For the sake of mathematical and computational i=l tractability, following assumptions are made in the theory of HMMS 1564) A probability distribution in each of the states, 1573 (1) The Markov Assumption B={b} 1574 As given in the definition of HMMs, transition b;(k)=p(o-j}, 1sjs N, 1sks M probabilities are defined as, 1565 where V denotes the k"observation symbol in the ag-p (4.1-i?q=i). alphabet, and o, the current parameter vector. 1575. In other words it is assumed that the next state is 1566 Following stochastic constraints must be satisfied. dependent only upon the current state. This is called the Markov assumption and the resulting model becomes actu ally a first order HMM. 1576. However generally the next state may depend on past k states and it is possible to obtain a such model, called and an k" order HMM by defining the transition probabilities as i Xb;(k) = 1, 1 s is N follows. k=1 lili... pig i?qi-ii, 4-1-i2, ..., q_k=ik, 1 sil, I2,..., is is N. 1577 But it is seen that a higher order HMM will have a 1567. If the observations are continuous then we will higher complexity. Even though the first order HMMs are have to use a continuous probability density function, the most common, Some attempts have been made to use the instead of a set of discrete probabilities. In this case we higher order HMMs too. US 2006/0167784 A1 Jul. 27, 2006 66
1578 (2) The Stationarity Assumption unknown, quantities, parameter estimation requires a sepa rate learning procedure (usually EM). In the latter case, we 1579 Here it is assumed that state transition probabilities typically do not represent the parameters in the graph; are independent of the actual time at which the transitions shared parameters (as in this example) are implemented by takes place. Mathematically, specifying that the corresponding CPDs are "tied’. 1590 An HMM is a hidden Markov model because we don’t see the states of the Markov chain, Q, but just a 1580 (3) The Output Independence Assumption function of them, namely Y. For example, if Y is a vector, we might define P(Y=y Q=i)=N(y: u. O.). A richer model, 1581. This is the assumption that current output(obser widely used in speech recognition, is to model the output Vation) is statistically independent of the previous output (conditioned on the hidden State) as a mixture of Gaussians. S(observations). We can formulate this assumption math This is shown in FIG. 3. ematically, by considering a sequence of observations, 1591 Some popular variations on the basic HMM theme O=O 1,02, . . . , OT. are illustrated in FIGS. 4A, 4B and 4C, which represent, 1582 Then according to the assumption for an HMM w. respectively, an input-output HMM, a factorial HMM, and a coupled HMM. (In the input-output model, the CPD P(QU) could be a softmax function, or a neural network.) Software T is available to handle inference and learning in general Bayesian networks, making all of these models trivial to t= implement. 1592. It is noted that the parameters may also vary with 1583. However unlike the other two, this assumption has time. This does not violate the presumptions inherent in an a very limited validity. In some cases this assumption may HMM, but rather merely complicates the analysis since a not be fair enough and therefore becomes a severe weakness static simplifying presumption may not be made. of the HMMs. 1593. A discrete-time, discrete-space dynamical system 1584) A Hidden Markov Model (HMM) is a Markov governed by a Markov chain emits a sequence of observable chain, where each state generates an observation. You only outputs: one output (observation) for each state in a trajec see the observations, and the goal is to infer the hidden state tory of such states. From the observable sequence of outputs, sequence. HMMs are very useful for time-series modeling, we may infer the most likely dynamical system. The result since the discrete state-space can be used to approximate is a model for the underlying process. Alternatively, given a many non-linear, non-Gaussian systems. sequence of outputs, we can infer the most likely sequence 1585 HMMs and some common variants (e.g., input of states. We might also use the model to predict the next output HMMs) can be concisely explained using the lan observation or more generally a continuation of the guage of Bayesian Networks, as we now demonstrate. sequence of observations. 1586 Consider the Bayesian network in FIG. 1, which 1594. The Evaluation Problem and the Forward Algo represents a hidden Markov model (HMM). (Circles denote rithm continuous-valued random variables, squares denote dis 1595. We have a model u-(A, B, JL) and a sequence of crete-valued, clear means hidden, shaded means observed.) observations O=oi, o, . . . , or, and p(O2) must be found. This encodes the joint distribution We can calculate this quantity using simple probabilistic arguments. But this calculation involves number of opera 1587 For a sequence of length T. we simply “unroll the tions in the order of N'. This is very large even if the length model for T time steps. In general, such a dynamic Bayesian of the sequence, T is moderate. Therefore we have to look network (DBN) can be specified by just drawing two time for an other method for this calculation. Fortunately there slices (this is sometimes called a 2TBN)—the structure (and exists one which has a considerably low complexity and parameters) are assumed to repeat. makes use an auxiliary variable, a (I) called forward vari able. 1588. The Markov property states that the future is inde pendent of the past given the present, i.e., Q(\indep 1596) The forward variable is defined as the probability Qt. Q. We can parameterize this Markov chain using a of the partial observation sequence oo. . . . . or, when it transition matrix, M-P(Q=jQ=i), and a prior distri terminates at the state i. Mathematically, bution, L=P(Q=i). a,(i)= i pol, O2, ..., o, q=ifi} (1.1) 1589. We have assumed that this is a homogeneous Markov chain, i.e., the parameters do not vary with time. 1597. Then it is easy to see that following recursive This assumption can be made explicit by representing the relationship holds. parameters as nodes: see FIG. 2: P1 represents L, P2 represents the transition matrix, and P3 represents the W (1.2) parameters for the observation model. If we think of these at-1 (i) = b;(0-1) a, (i)aii, 1 s is N, 1 s is T-1 parameters as random variables (as in the Bayesian i=1 approach), parameter estimation becomes equivalent to inference. If we think of the parameters as fixed, but US 2006/0167784 A1 Jul. 27, 2006 67
1598) where, 1610. Initialization. For all states i. 6 (i)= JL,b,(O)p(i)=text missing or illegible when a 1(j)=T;b;(o), 1sjsN filed 1599. Using this recursion we can calculate 1611 Recursion. From t=2 to T and for all states j. ar(i)=1 sisN 8, (i)=MaZö (i)a;b; (O), (G)=argmazö ()a; 1600 and then the required probability is given by, 1612) Termination. P=Mazö (text missing or illegible when filed) text missing or illegible when filed=argmazo (text missing or illegible when filed) 1613 Recovering the state sequence. From t=T-1 to 1, text missing or illegible when filed=9text missing or illegible when filed.) 1601 The complexity of this method, known as the forward algorithm is proportional to NT, which is linear 1614. In many HMM system implementations, the Vit with respect to T whereas the direct calculation mentioned erbi algorithm is used for evaluation at recognition time. earlier, had an exponential complexity. Note that since Viterbi only guarantees the maximum of Pr(O.S.D.) over all state sequences S (as a result of the first 1602. In a similar way we can define the backward order Markov assumption) instead of the sum over all variable B(i) as the probability of the partial observation possible state sequences, the resultant scores are only an sequence o O2 . . . . . or, given that the current state is approximation. i. Mathematically, 1615 So far the discussion has assumed some method of quantization of feature vectors into classes. However, 1603 As in the case of a (i) there is a recursive relation instead of using vector quantization, the actual probability ship which can be used to calculate B(i) efficiently. densities for the features may be used. Baum-Welch, Viterbi, and the forward-backward algorithms can be modified to handle a variety of characteristic densities. In this context, W (1.5) however, the densities will be assumed to be Gaussian. f(i) =X B (jabi (oil), 1 s is N, 1st s T-1 Specifically,
1 1604 where, B(i)=1, 1 sisN 1605) Further we can see that, 1616) Initial estimations of Land O may be calculated by dividing the evidence evenly among the states of the model 1606. Therefore this gives another way to calculate and calculating the mean and variance in the normal way. p{OW, by using both forward and backward variables as Whereas flat densities were used for the initialization step given in eqn. 1.7. See, http://jedlik.phy.bme.hu/~gerianos/ before, the evidence is used here. Now all that is needed is HMM/, expressly incorporated herein by reference. a way to provide new estimates for the output probability. We wish to weight the influence of a particular observation for each state based on the likelihood of that observation occurring in that state. Adapting the Solution from the discrete case yields
1607 Eqn. 1.7 is very useful, specially in deriving the formulas required for gradient based training. a = -i=1 1 - 1608. The Decoding Problem and the Viterbi Algorithm X, Y, (j) 1609 While the estimation and evaluation processes described above are sufficient for the development of an HMM system, the Viterbialgorithm provides a quick means of evaluating a set of HMM’s in practice as well as provid ing a solution for the decoding problem. In decoding, the goal is to recover the State sequence given an observation O = sequence. The Viterbialgorithm can be viewed as a special X,T Y, (j) form of the forward-backward algorithm where only the t= maximum path at each time step is taken instead of all paths. This optimization reduces computational load and allows the recovery of the most likely state sequence. The steps to the 1617 For convenience, u' is used to calculate o' instead Viterbi are of the re-estimated I'. While this is not strictly proper, the US 2006/0167784 A1 Jul. 27, 2006 values are approximately equal in contiguous iterations and other words there may be several optimization criteria for seem not to make an empirical difference. See, http:// learning, out of which a Suitable one is selected depending www.white.media.mit.edu/~testarnefasl/asl-tr375, expressly on the application. incorporated herein by reference. Since only one stream of data is being used and only one mixture (Gaussian density) 1627 There are two main optimization criteria found in is being assumed, the algorithms above can proceed nor ASR literature; Maximum Likelihood (ML) and Maximum mally, incorporating these changes for the continuous den Mutual Information (MMI). The solutions to the learning sity case. problem under each of those criteria is described below. 1618. We want to find the most likely state sequence for a given sequence of observations, O=oi, o, . . . . or and a 1628 Maximum Likelihood (ML) Criterion model, w=(A, B, JL). 1629. In ML we try to maximize the probability of a 1619. The solution to this problem depends upon the way given sequence of observations OY, belonging to a given “most likely state sequence' is defined. One approach is to class w, given the HMM of the class w, with respect to find the most likely state qat t=t and to concatenate all Such the parameters of the model W. This probability is the total q’s. But sometimes, this method does not give a physically likelihood of the observations and can be expressed math meaningful state sequence. Therefore we would seek ematically as another method which has no such problems. In this method, commonly known as Viterbi algorithm, the whole state Leipt OW...} sequence with the maximum likelihood is found. In order to facilitate the computation we define an auxiliary variable, 1630. However since we consider only one class w at a time we can drop the subscript and superscript 'w's. Then the ML criterion can be given as, Ö, (i) = max piq1, q2, ... , 4-1, q = i, O, O2, ... , O-1}, q92 ... it Le=p(OI). (1.9) 1631. However there is no known way to analytically 1620 which gives the highest probability that partial solve for the model w(A, B, JL), which maximize the observation sequence and state sequence up to t=t can have, quantity L. But we can choose model parameters such that when the current state is i. It is easy to observe that the it is locally maximized, using an iterative procedure, like following recursive relationship holds. Baum-Welch method or a gradient based method, which are described below. 1632 Baum-Welch Algorithm Ö, 1(j) = bio-imax-- 0.(i)a 1 < is N, 1 < t < T- 1 (1.8) 1633. This method can be derived using simple “occur rence counting” arguments or using calculus to maximize 1621 where, the auxiliary quantity 81 (j)=Tb (o), 1sjsN 1622) So the procedure to find the most likely state sequence starts from calculation of Ör(j), 1 sisN using recursion in 1.8, while always keeping a pointer to the “winning state' in the maximum finding operation. Finally the state j, is found where 1634 over I, p 344-346.). A special feature of the algorithm is the guaranteed convergence. To describe the Jf = arginas orT(j), (i Baum-Welch algorithm, (also known as Forward-Backward algorithm), we need to define two more auxiliary variables, in addition to the forward and backward variables defined in 1623 and starting from this state, the sequence of states a previous section. These variables can however be is back-tracked as the pointer in each state indicates.This expressed in terms of the forward and backward variables. gives the required set of States. 1635 First one of those variables is defined as the prob 1624. This whole algorithm can be interpreted as a search ability of being in state i at t=t and in state j at t=t--1. in a graph whose nodes are formed by the states of the HMM Formally, in each of the time instant t, 1sts.T. 1625) The Learning Problem 1636. This is the same as, 1626 Generally, the learning problem is how to adjust the HMM parameters, so that the given set of observations (called the training set) is represented by the model in the piq = i, q+1 = j. Oil: (1.11) best way for the intended application. Thus it would be clear (i, j) = — a that the "quantity” we wish to optimize during the learning process can be different from application to application. In US 2006/0167784 A1 Jul. 27, 2006 69
1637. Using forward and backward variables this can be 1643 Gradient Based Method expressed as, 1644. In the gradient based method, any parameter 0 of the HMM is updated according to the standard formula, a (i)aijf3-1 (j)bi (O-1) (1.12) (i, j) = W W I (1.19) a; (i)aiif3-1(j)bi (o, -1) Giel Gold -n.O G=G,
1645 where J is a quantity to be minimized. We define in 1638. The second variable is the a posteriori probability, this case, J-EMI=-logCp (OW)=-log(L) (1.20) 1646 Since the minimization of J=EM is equivalent to 1639 that is the probability of being in state i at t=t, given the maximization of L, eqn. 1.19 yields the required the observation sequence and the model. In forward and optimization criterion, ML. But the problem is to find the backward variables this can be expressed by, derivative
(1.14) I W O Xa, (if (i) for any parameter 0 of the model. This can be easily done by relating J to model parameters via L. As a key step to do so, using the eqns. 1.7 and 1.9 we can obtain, 1640. One can see that the relationship between Y, (i) and S(i, j) is given by, (1.21)
W y(i) =X 6, (i, j), 1 s is N, 1 s is M (1.15) i=l 1647 Differentiating the last equality in eqn. 1.20 with respect to an arbitrary parameter 0, 1641. Now it is possible to describe the Baum-Welch learning process, where parameters of the HMM is updated 8 J 1 Ó Lot (1.22) in such a way to maximize the quantity, p O2}. Assuming alo Lao a starting model W-(A, B, t), we calculate the C.'s and B's using the recursions 1.5 and 1.2, and then 'S's and 'Y's using 1.12 and 1.15. Next step is to update the HMM parameters 1648 Eqn. 1.22 gives according to eqns 1.16 to 1.18, known as re-estimation formulas.
7t; = y (i), 1 < is N (1.16) if we know X5T- (i, j) (1.17) ai = i , 1 s is N, 1 s is N 6 Lot y (i) O t= which can be found using eqn, 1.21. However this deriva tive is specific to the actual parameter concerned. Since (1.18) there are two main parameter sets in the HMM, namely transition probabilities a 1si,js N and observation probabilities b,(k), 1sis N, 1sks M, we can find the derivative
6 Lot O 1642. These reestimation formulas can easily be modified to deal with the continuous density case too. US 2006/0167784 A1 Jul. 27, 2006 70 for each of the parameter sets and hence the gradient, 1657 Finally we get the required probability, by substi tuting for
I 3G) 6 Lot ob (O.) 1649 Gradient With Respect to Transition Probabilities in eqn, 1.22 (keeping in mind that e=b,(o) in this case), 1650. Using the chain rule, which is obtained by substituting eqns.1.28 and 1.24 in eqn. 1.27. 0 Lo - 0 Lot do, (j) (1.23) 0 1 0, (h)f(i) (1.29) ob (a) Lot bi(a)
1651. By differentiating eqn. 1.21 with respect to C,G) we get, 1658. Usually this is given the following form, by first Substituting for L.tot from eqn.1.21 and then Substituting from eqn. 1.14. 6 Lot (1.24) da, (i) = B(i), 9 - ?) (1.30) ob (a) b; (a) 1652 and differentiating (a time shifted version of) eqn 1.2 with respect to a 1659. If the continuous densities are used then
'E'da, (i = b(o,)a, 1 (i) (1.25)1.25 Öc in optim ÖXin
1653 Eqns. 1.23, 1.24 and 1.25 give, can be found by further propagating the derivative
6 Lot I dai ob (O.) and Substituting this quantity in eqn. 1.22 (keeping in using the chain rule. The same method can be used to propagate the derivative (if necessary) to a front end pro mind that 0=a in this case), we get the required result, cessor of the HMM. This will be discussed in detail later. 1660. Maximum Mutual Information (MMI) Criterion I 1 (1.26) (i)bi (o)at-1 (i) 1661. In ML we optimize an HMM of only one class at Öaii fiX. Bobo a time, and do not touch the HMMs for other classes at that time. This procedure does not involve the concept “discrimi nation” which is of great interest in Pattern Recognition. 1654 Gradient With Respect to Observation Probabilities Thus the ML learning procedure gives a poor discrimination ability to the HMM system, specially when the estimated 1655. Using the chain rule, parameters (in the training phase) of the HMM system do not match with the inputs used in the recognition phase. This type of mismatches can arise due to two reasons. One is that 6 Lot 6 Lot 6a, (i) (1.27) the training and recognition data may have considerably ob (a) da, (i) ob (a) different statistical properties, and the other is the difficulties of obtaining reliable parameter estimates in the training. 1656. Differentiating (a time shifted version of) the 1662. The MMI criterion on the other hand consider eqn.1.2 with respect to b,(O) HMMs of all the classes simultaneously, during training. Parameters of the correct model are updated to enhance its contribution to the observations, while parameters of the do, (ii) or (i) (1.28) alternative models are updated to reduce their contributions. ob (a) b; (a) This procedure gives a high discriminative ability to the system and thus MMI belongs to the so called “discrimina tive training category. US 2006/0167784 A1 Jul. 27, 2006
1663. In order to have a closer look at the MMI criterion, 1674. As in the case of ML re-estimation I or gradient consider a set of HMMs methods can be used to minimize the quantity EMMI. In the following a gradient based method, which again makes use of the eqn.1.19, is described. 1664. The task is to minimize the conditional uncertainty of a class V of utterances given an observation sequence O 1675 Since EMM is to be minimized, in this case of that class. This is equivalent minimize the conditional information, 1676 and therefore J is directly given by eqn.1.37. The I(v/O, A)=-logp(v/O, A} (1.31) problem then simplifies to the calculation of gradients 1665) with respect to A. 1666. In an information theoretical frame work this leads I to the minimization of conditional entropy, defined as the 3G) expectation (E() of the conditional information I. where 0 is an arbitrary parameter of the whole set of H(VIO)=ELI(v/O) (1.32) HMMs, A. This can be done by differentiating 1.37 with 1667 where V represents all the classes and O represents respect to 0. all the observation sequences. Then the mutual information between the classes and observations, a J 1 a LE 1 a Limped (1.38) Limped do 1668 become maximized; provided H(V) is constant. This is the reason for calling it Maximum Mutual Informa tion (MMI) criterion. The other name of the method, Maxi 1677. The same technique, as in the case of ML, can be mum A Posteriori (MAP) has the roots in eqn. 1.31 where used to compute the gradients of the likelihoods with respect the a posteriori probability p{VO, A} is maximized. to the parameters. As a first step likelihoods from eqns. 1.35 1669) Even though the eqn, 1.31 defines the MMI crite and 1.36, are expressed in terms of forward and backward rion, it can be rearranged using the Bayes theorem to obtain variables using the form as in eqn. 1.7. a better insight, as in eqn.1.34. L." = X a,(i)6(i) (1.39) Effi = -log {VO, A (1.34) ieclass v LGF =X, X a,(if (i) (1.40) tu ie class w
1678. Then the required gradients can be found by dif ferentiating eqns. 1.39 and 1.40. But we consider two cases: one for the transition probabilities and another for the 1670 where w represents an arbitrary class. observation probabilities, similar to the case of ML. 1671) If we use an analogous notation as in eqn. 1.9, we 1679) Maximum Mutual Information (MMI) Criterion can write the likelihoods, 1680. The MMI criterion considers HMMs of all the classes simultaneously, during training. Parameters of the correct model are updated to enhance its contribution to the LE"" = p(y, OA (1.35) observations, while parameters of the alternative models are updated to reduce their contributions. This procedure gives LC" =Xpw, OIA) (1.36) a high discriminative ability to the system and thus MMI belongs to the so called “discriminative training category. 1681. In order to have a closer look at the MMI criterion, 1672. In the above equations the superscripts clamped consider a set of HMMs and free are used to imply the correct class and all the other classes respectively. 1682. The task is to minimize the conditional uncertainty of a class V of utterances given an observation sequence O 1673. If we substitute eqns. 1.35 and 1.36 in the eqn. 1.34, of that class. This is equivalent minimize the conditional We get, information,
clamped (1.37) 1683 with respect to A. In an information theoretical EMM = -log", frame work this leads to the minimization of conditional entropy, defined as the expectation (E()) of the conditional information I. US 2006/0167784 A1 Jul. 27, 2006 72
where 0 is an arbitrary parameter of the whole set of 1684 where V represents all the classes and O represents HMMs, A. This can be done by differentiating 1.37 with all the observation sequences. Then the mutual information respect to 0. between the classes and observations, a 1 a LE 1 a Limped (1.38) 1685 become maximized; provided H(V) is constant. 6 free 60 climped ao This is the reason for calling it Maximum Mutual Informa if if tion (MMI) criterion. The other name of the method, Maxi mum A Posteriori (MAP) has the roots in eqn. 1.31 where the a posteriori probability p{VO, A} is maximized. 1694. The same technique, as in the case of ML, can be used to compute the gradients of the likelihoods with respect 1686. Even though the eqn. 1.31 defines the MMI crite to the parameters. As a first step likelihoods from eqns. 1.35 rion, it can be rearranged using the Bayes theorem to obtain and 1.36, are expressed in terms of forward and backward a better insight, as in eqn.1.34. variables using the form as in eqn. 1.7.
Etti = -log {VO, A} (1.34) imped- X. a (i)f(i) (1.39) ieclass v LGF =X, X a,(if (i) (1.40) tu ie class w
1695. Then the required gradients can be found by dif 1687 where w represents an arbitrary class. ferentiating eqns. 1.39 and 1.40. But we consider two cases: one for the transition probabilities and another for the 1688 If we use an analogous notation as in eqn. 1.9, we can write the likelihoods, observation probabilities, similar to the case of ML. 1696 Gradient With Respect to Transition Probabilities Limped = p(y, O'A} (1.35) 1697) Using the chain rule for any of the likelihoods, free or clamped, LC" =Xpw, O"|a) (1.36)
dL' T dL' oa, (i) (1.41)1.41 1689. In the above equations the superscripts clamped daii da, (i) dai and free are used to imply the correct class and all the other classes respectively. 1690. If We Subst1tutebsti eqns.qns. 1.35l.35 and 1.36l.36 inin the eqn.eqn.l.34, 1.34 1698. Differentiating eqns. 1.39 and 1.40 with respect to We get, C(i), to get two results for free and clamped cases and using the common result in eqn, 1.25, we get Substitutions for both clamped (1.37) terms on the right hand side of eqn. 1.41. This substitution EMM = -log'. yields two separate results for free and clamped cases.
a Limped T (1.42) 1691. As in the case of ML re-estimation I or gradient Öaii = Ökyk 2. B(i)b(i)bi (a)a,( t) t-1 1 (i),(i methods can be used to minimize the quantity EMMI. In the ieclass k following a gradient based method, which again makes use of the eqn.1.19, is described. 1692. Since EMM is to be minimized, in this case 1699 where 8 is a Kronecker delta.
1693 and therefore J is directly given by eqn.1.37. The 0LE . (1.43) problem then simplifies to the calculation of gradients daii 2, f3(i)b (a)a. 1 (i)
I 3G) 1700 Substitution of eqns. 1.42 and 1.43 in the eqn. 1.38(keeping in mind that e=a in this case) gives the required result, US 2006/0167784 A1 Jul. 27, 2006
I 1 Öy la (1.44) I y(i) free - (i) kimped (1.51)1.51 tanpedk 2. f3, (i)b (a)a, 1 (i), ob (a) bi (a;) daii LE. L.if t= ieclass k
1709. This equation completely defines the update of 1701 Gradient With Respect to Observation Probabilities observation probabilities. If however continuous densities are used then we can further propagate this derivative using 1702. Using the chain rule for any of the likelihoods, free the chain rule, in exactly the same way as mentioned in the or clamped, case ML. A similar comments are valid also for preproces SOS. dL' 8 L 0a,(j) (1.45) 1710 Training ob (a, 6a, (i) db (a, 1711) We assume that the preprocessing part of the sys tem gives out a sequence of observation vectors 1703 Differentiating eqns. 1.39 and 1.40 with respect to C(i), to get two results for free and clamped cases, and using 1712 Starting from a certain set of values, parameters of the common result in eqn.1.28, we get Substitutions for both each of the HMMS terms on the right hand side of eqn. 1.45. This substitution yields two separate results for free and clamped cases. 1 sisN 1713 can be updated as given by the eqn.1.19, while the required gradients will be given by eqns. 1.44 and 1.48. C" a,(j)6(j) (1.46) However for this particular case, isolated recognition, like aba) - ok ba) - lihoods in the the last two equations are calculated in a jeclass k peculiar way. First consider the clamped case. Since we have an HMM for each class of units in isolated recognition, we can select the model of the class 1 to which the current 1704 where 8 is a Kronecker delta. And observation sequence O' belongs. Then starting from eqn. 1.39, 8 LC: a, (i)/3, (i) (1.47) ob (a) b (a) LC" = L =Xa,(i)6(i) (1.52)
1705 Substitution of eqns. 1.46 and 1.47 in eqn. 1.38 we get the required result, 1714 where the second line follows from eqn.1.3. aba)I is1 signedÖky "Elba) 6 C (1.48) 1715 Similarly for the fee case, starting from eqn. 1.40.
W W (1.53) 1706 This equation can be given a more aesthetic form by defining, L =XLn=1 =X Liean2. lift
a (i)f(i) . (1.49) y(i)cloped = 0, clamped , ie class k if 1716) where L, represents the likelihood of the current 1707 where 8 is a Kronecker delta, and observation sequence belonging to class 1, in the model W. With those likelihoods defined in eqns. 1.52 and 1.53, the gradient giving equations 1.44 and 1.48 will take the forms, a; (i)f(i) (1.50) clamped f Okl T (1.54) L X f(i)bj(a)a- (i), i, je As 1708. With these variables we express the eqn. 1.48 in the following form. US 2006/0167784 A1 Jul. 27, 2006 74
1735) Use of Fourier Transform in Pre-Processing -continued 1736. The Hartley Transform is an integral transform I 1 dia, (i)f(i) . (1.55) which shares some features with the Fourier Transform, but at \ct =| X Lh - - \ct -, je le which (in the discrete case), multiplies the kernel by n=1
1717 Now we can summarize the training procedure as cos(t) -sir." (1) follows. 1718 (1) Initialize the each HMM, =(A, B, JL), 1 sisN with values generated randomly or using an initial 1737 instead of ization algorithm like segmental K means. 1719 (2) Take an observation sequence and e-2?tikn/N - co- ag isin 2: } (2) 1720 Calculate the forward and backward probabili ties for each HMM, using the recursions 1.5 and 1.2. 1721. Using the equations 1.52 and 1.53 calculate the 1738. The Hartley transform produces real output for a likelihoods real input, and is its own inverse. It therefore can have 1722. Using the equations 1.54 and 1.55 calculate the computational advantages over the discrete Fourier trans gradients with respect to parameters for each model form, although analytic expressions are usually more com plicated for the Hartley transform. 1723 Update parameters in each of the models using the eqn.1.19. 1739 The discrete version of the Hartley transform can be written explicitly as 1724 (3) Go to step (2), unless all the observation sequences are considered. 1725 (4) Repeat step(2) to (3) until a convergence cri terion is satisfied. 7tal - Solo=0 ("")-ski") (3) 1726. This procedure can easily be modified if the con = RT al-J fa), (4) tinuous density HMMs are used, by propagating the gradi ents via chain rule to the parameters of the continuous probability distributions. Further it is worth to mention that preprocessors can also be trained simultaneously, with Such where F denotes the Fourier Transform. The Hartley trans a further back propagation. form obeys the convolution property 1727. Recognition 1728 Comparative to the training, recognition is much hta: b = (A.B. -A, B, + AB + AB), (5)5 simpler and the procedure is given below. 1729 (1) Take an observation sequence to be recognized and 1740 where 1730 Calculate the forward and backward probabili a=ao (6) ties for each HMM, using the recursions 1.5 and 1.2. an/2=an 2 (7) 1731. As in the equation 1.53 calculate the likelihoods, al-Fan-k (8) L', 1sms N (Arndt). Like the fast Fourier Transform algorithm, there is 1732. The recognized class 1*, to which the observa a “fast' version of the Hartley transform algorithm. A tion sequence belongs, is given by decimation in time algorithm makes use of Ha-Ha"+XHall (9) Higha-Haven-XHadd (10) (2) = argmax. L.2. 1< 1742 The discrete Fourier transform 1749 and error is denoted by J, then we can find I (14) 60 simply by using the chain rule, 1743 can be written T a a Xi (j) (2.2) 8X(j); 22. 80 Ak. |- XI,8 III.(in =0 F W- 27tkin Y 27tkn 1750 We assume that X: 1 -i 1 + i COS N S. N =0 21 + i 1 -i -S (2) COS (2) I T-1 N ; 22. H 8X(j) 1 1 + i 1 - i? a 21-i 1 + i is known and ÖX,r; 2 (i). SO 88: F=THT. 1744. See, http://mathworld.wolfram.com/HartleyTrans can simply be found by differentiating eqn.2.1 with respect form.html. to 0. Thus we get, 1745) A Hartley transform based fixed pre-processing may be considered, on Some bases, inferior to that based on (2.3) Fourier transform. One explanation for this is based on the i respective symmetries and shift invariance properties. Therefore we expect improved performances from Fourier transform even when the pre-processing is adaptive. How ever a training procedure which preserves the symmetries of weight distributions must be used. Main argument of the use of Hartley transform is to avoid the complex weights. A 1751 Eqns.2.2 and 2.3 define the backward pass. Note Fourier transform, however, can be implemented as a neural that 0, can be further back propagated as usual. network containing real weights, but with a slightly modi 1752. Training Procedure Which Preserves Symmetry fied network structure than the usual MLP. We can easily 1753 We can use a training procedure which preserves derive the equations which give the forward and backward symmetrical distribution of weights in the Hartley or Fourier pass. transform stages. In addition to the improved shift invari ance, this approach can lead to parameter reduction. The 1746 Forward pass is given by, procedure starts by noting the equal weights at initialization. Then the forward and backward passes are performed as usual. But in updating we use the same weight update for all ... 2 N-1 2 (2.1) the equal weights, namely the average value of all the weight updates corresponding to the equal weights. In this way we X. x, sin() can preserve any existing symmetry in the initial weight distributions. At the same time number of parameters is reduced because only one parameter is needed to represent 1747 where N denotes the window length, and the whole class of equal weights. 1754) See, “A Hybrid ANN-HMM ASR system with NN 1748 If we use the notation based adaptive preprocessing. Narada Dilp Warakagoda, M.Sc. thesis (Norges Tekniske Hogskole. Institutt for Teleteknikk Transmisjonsteknikk), http://jedlik.phy.bme.hu/ 6 = t2iti ~gerjanos/HMM/hoved.html. 1755 Asal alternate to the Hartley transform, a Wave let transform may be applied. US 2006/0167784 A1 Jul. 27, 2006 76 1756. The fast Fourier transform (FFT) and the discrete 1766. Within each family of wavelets (such as the wavelet transform (DWT) are both linear operations that Daubechies family) are wavelet subclasses distinguished by generate a data structure that contains segments of various the number of coefficients and by the level of iteration. lengths, usually filling and transforming it into a different Wavelets are classified within a family most often by the data vector of length. number of Vanishing moments. This is an extra set of 1757. The mathematical properties of the matrices mathematical relationships for the coefficients that must be involved in the transforms are similar as well. The inverse satisfied, and is directly related to the number of coefficients. transform matrix for both the FFT and the DWT is the For example, within the Coiflet wavelet family are Coiflets transpose of the original. As a result, both transforms can be with two vanishing moments, and Coiflets with three van viewed as a rotation in function space to a different domain. ishing moments. For the FFT, this new domain contains basis functions that 1767 The Discrete Wavelet Transform are sines and cosines. For the wavelet transform, this new domain contains more complicated basis functions called 1768 Dilations and translations of the “Mother func wavelets, mother wavelets, or analyzing wavelets. tion,” or “analyzing wavelet'd(x) define an orthogonal basis, our wavelet basis: 1758 Both transforms have another similarity. The basis functions are localized in frequency, making mathematical tools such as power spectra (how much power is contained -(2) (3) in a frequency interval) and scalegrams (to be defined later) (PG) (x) = 22 d(2 x - 1) useful at picking out frequencies and calculating power distributions. (2) indicates text missing or illegiblewhen filed 1759. The most interesting dissimilarity between these two kinds of transforms is that individual wavelet functions 1769. The variables s and 1 are integers that scale and are localized in space. Fourier sine and cosine functions are dilate the mother function d(x) to generate wavelets, such as not. This localization feature, along with wavelets local a Daubechies wavelet family. The scale index s indicates the ization of frequency, makes many functions and operators wavelets width, and the location index 1 gives its position. using wavelets “sparse' when transformed into the wavelet Notice that the mother functions are rescaled, or “dilated by domain. This sparseness, in turn, results in a number of powers of two, and translated by integers. What makes useful applications Such as data compression, detecting wavelet bases especially interesting is the self-similarity features in images, and removing noise from time series. caused by the scales and dilations. Once we know about the 1760. One way to see the time-frequency resolution dif mother functions, we know everything about the basis. Note ferences between the Fourier transform and the wavelet that the scaling-by-two is a feature of the Discrete Wavelet transform is to look at the basis function coverage of the Transform (DWT), and is not, itself, compelled by Wavelet time-frequency plane. theory. That is, while it is computationally convenient to employ a binary tree, in theory, if one could define a precise 1761. In a windowed Fourier transform, where the win wavelet that corresponds to a feature of a data set to be dow is simply a square wave, the square wave window processed, this wavelet could be directly extracted. Clearly, truncates the sine or cosine function to fit a window of a the utility of the DWT is its ability to handle general cases particular width. Because a single window is used for all without detailed pattern searching, and therefore the more frequencies in the WFT, the resolution of the analysis is the theoretical wavelet transform techniques based on precise same at all locations in the time-frequency plane. wavelet matching are often reserved for special cases. On 1762 An advantage of wavelet transforms is that the the other hand, by carefully selecting wavelet basis func windows vary. In order to isolate signal discontinuities, one tions, or combinations of basis functions, a very sparse would like to have some very short basis functions. At the representation of a complex and multidimensional data same time, in order to obtain detailed frequency analysis, space may be obtained. The utility, however, may depend on one would like to have some very long basis functions. A being able to operate in the wavelet transform domain (or way to achieve this is to have short high-frequency basis Subsequent transforms of the sparse representation coeffi functions and long low-frequency ones. This happy medium cients) for subsequent analysis. Note that, while wavelets are is exactly what you get with wavelet transforms. genrally represented as two dimensional functions of ampli 1763. One thing to remember is that wavelet transforms tude and time, it is clear that wavelet theory extends into do not have a single set of basis functions like the Fourier n-dimensional space. transform, which utilizes just the sine and cosine functions. 1770 Thus, the advantageous application of wavelet Instead, wavelet transforms have an infinite set of possible theory is in cases where a modest number of events, for basis functions. Thus wavelet analysis provides immediate example having associated limited time and space param access to information that can be obscured by other time eters, are represented in a large data space. If the events frequency methods such as Fourier analysis. could be extracted with fair accuracy, the data space could be replaced with a vector quantized model (VQM), wherein 1764 Wavelet transforms comprise an infinite set. The the extracted events correspond to real events, and wherein different wavelet families make different trade-offs between the VOM is highly compressed as compared to the raw data how compactly the basis functions are localized in space and space. Further, while there may be some data loss as a result how smooth they are. of the VOM expression, if the real data corresponds to the 1765 Some of the wavelet bases have fractal structure. wavelet used to model it, then the VOM may actually serve The Daubechies wavelet family is one example. as a form of error correction. Clearly, in Some cases, US 2006/0167784 A1 Jul. 27, 2006 77 especially where events are overlapping, the possibility for called a quadrature mirror filter pair in signal processing error occurs. Further, while the DWT is often useful in parlance. A more detailed description of the transformation denoising data, in Some cases, noise may be inaccurately matrix can be found in W. Press et al., Numerical Recipes in represented as an event, while in the raw data space, it might Fortran, Cambridge University Press, New York, 1992, pp. have been distinguished. Thus, one aspect of a denoised 498-499, 584–602. DWT representation is that there is an implicit presumption that all remaining elements of the representation matrix are 1777 To complete our discussion of the DWT, let's look signal. at how the wavelet coefficient matrix is applied to the data vector. The matrix is applied in a hierarchical algorithm, 1771 A particular advantage of a DWT approach is that Sometimes called a pyramidal algorithm. The wavelet coef it facilitates a multiresolution analysis of data sets. That is, ficients are arranged so that odd rows contain an ordering of if decomposition of the raw data set with the basis function, wavelet coefficients that act as the smoothing filter, and the transformed according to a regular progressions, e.g., pow even rows contain an ordering of wavelet coefficient with ers of 2, then at each level of decomnposition, a level of different signs that act to bring out the data's detail. The scale is revealed and presented. It is noted that the transform matrix is first applied to the original, full-length vector. Then need not be a simple power of two, and itself may be a the vector is smoothed and decimated by half and the matrix function or complex and/or multidimensional function. is applied again. Then the Smoothed, halved vector is Typically, non-standard analyses are reserved for instances Smoothed, and halved again, and the matrix applied once where there is, or is believed to be, a physical basis for the more. This process continues until a trivial number of application of Such functions instead of binary splitting of “Smooth-Smooth-Smooth . . . data remain. That is, each the data space. matrix application brings out a higher resolution of the data 1772 Proceeding with the DWT analysis, we span our while at the same time Smoothing the remaining data. The data domain at different resolutions, see www.eso.org/ output of the DWT consists of the remaining “smooth (etc.)” projects/esomidas/doc/user/98NOV/volb/node308.html, components, and all of the accumulated “detail” compo using the analyzing wavelet in a scaling equation: nentS. 1778. The Fast Wavelet Transform 1779. If the DWT matrix is not sparse, so we face the same complexity issues that we had previously faced for the discrete Fourier transform. Wickerhauser, Adapted Wavelet Analysis from Theory to Software, A K Peters, Boston, 1994, pp. 213-214,237,273-274,387. We solve it as we did for the 1773 where W(x) is the scaling function for the mother FFT, by factoring the DWT into a product of a few sparse function d(x), and care the wavelet coefficients. The wavelet matrices using self-similarity properties. The result is an coefficients must satisfy linear and quadratic constraints of algorithm that requires only order n operations to transform the form an n-sample vector. This is the “fast DWT of Mallat and Daubechies. 1780 Wavelet Packets 1781. The wavelet transform is actually a subset of a far more versatile transform, the wavelet packet transform. M. A. Cody, “The Wavelet Packet Transform,” Dr. Dobb's Jour 1774 where 8 is the delta function and 1 is the location nal, Vol 19, April 1994, pp. 44-46, 50-54. index. 1782 Wavelet packets are particular linear combinations 1775. One of the most useful features of wavelets is the of wavelets. V. Wickerhauser, Adapted Wavelet Analysis ease with which one can choose the defining coefficients for from Theory to Software, A K Peters, Boston, 1994, pp. a given wavelet system to be adapted for a given problem. 213-214, 237, 273-274, 387. They form bases which retain In Daubechies original paper, I. Daubechies, “Orthonormal many of the orthogonality, Smoothness, and localization Bases of Compactly Supported Wavelets.Comm. Pure properties of their parent wavelets. The coefficients in the Appl. Math., Vol 41, 1988, pp. 906-966, she developed linear combinations are computed by a recursive algorithm specific families of wavelet systems that were very good for making each newly computed wavelet packet coefficient representing polynomial behavior. The Haar wavelet is even sequence the root of its own analysis tree. simpler, and it is often used for educational purposes. (That is, while it may be limited to certain classes of problems, the 1783) Adapted Waveforms Haar wavelet often produces comprehensible output which can be generated into graphically pleasing results). 1784 Because we have a choice among an infinite set of basis functions, we may wish to find the best basis function 1776. It is helpful to think of the coefficients (c. . . . . c. for a given representation of a signal. Wickerhauser, Id. A as a filter. The filter or coefficients are placed in a transfor basis of adapted waveform is the best basis function for a mation matrix, which is applied to a raw data vector. The given signal representation. The chosen basis carries Sub coefficients are ordered using two dominant patterns, one stantial information about the signal, and if the basis descrip that works as a smoothing filter (like a moving average), and tion is efficient (that is, very few terms in the expansion are one pattern that works to bring out the data’s “detail needed to represent the signal), then that signal information information. These two orderings of the coefficients are has been compressed. US 2006/0167784 A1 Jul. 27, 2006 1785 According to Wickerhauser. Id., some desirable At each step, the number of scalar products is divided by 2. properties for adapted wavelet bases are Step by step the signal is Smoothed and information is lost. 1786 1. speedy computation of inner products with the The remaining information can be restored using the other basis functions; complementary subspace W of V in V. This subspace can be generated by a Suitable wavelet function p(x) 1787 2. speedy superposition of the basis functions: 1788. 3. good spatial localization, so researchers can 1797 with translation and dilation. identify the position of a signal that is contributing a large component; 1789 4. good frequency localization, so researchers can identify signal oscillations; and 1790) 5. independence, so that not too many basis O elements match the same portion of the signal. 1791. For adapted waveform analysis, researchers seek a basis in which the coefficients, when rearranged in decreas ing order, decrease as rapidly as possible. to measure rates of decrease, they use tools from classical harmonic analysis including calculation of information cost functions. This is We compute the scalar products with: defined as the expense of storing the chosen representation. Examples of such functions include the number above a threshold, concentration, entropy, logarithm of energy, Gauss-Markov calculations, and the theoretical dimension of a sequence. 1792 Multiresolution analysis results from the embedded Subsets generated by the interpolations at different scales. 1798. With this analysis, we have built the first part of a filter bank. In order to restore the original data, Mallat uses 1793. A function f(x) is projected at each step onto the the properties of orthogonal wavelets, but the theory has subset V This projection is defined by the scalar product been generalized to a large class of filters by introducing two c(k) of f(x) with the scaling function p(x) which is dilated and translated: other filters hand g named conjugated to hand g. 1799. The restoration, that is, the inverse transform after 1794. As p(x) is a scaling function which has the prop filtering in the transform domanin, is performed with: erty: le=X hor -n) O 1800. In order to get an exact restoration, two conditions are required for the conjugate filters: 1801) Dealiasing condition: 1795 where h(v) is the Fourier transform of the function X,h(n)ö(X-n). We get: ir -- ific)1Ya -- a -- igo)1Y a = 0 1802 Exact restoration: 1796. The property of the scaling function of p(x) is that it permits us to compute directly the set c(k) from c(k). 1803. In the decomposition, the function is successively If we start from the set co(k) we compute all the sets c(k), convolved with the two filters H (low frequencies) and G with j>0, without directly computing any other scalar prod (high frequencies). Each resulting function is decimated by uct: Suppression of one sample out of two. The high frequency signal is left, and we iterate with the low frequency signal. In the reconstruction, we restore the sampling by inserting a c; (k) = X h(n - 2k)c, (n) 0 between each sample, then we convolve with the conjugate filters H and G, we add the resulting functions and we multiply the result by 2. We iterate up to the smallest scale. US 2006/0167784 A1 Jul. 27, 2006 79 1804 Orthogonal wavelets correspond to the restricted 1810 which leads to three sub-images: case where: icy) = i (y) g(v) = g(v) and fity)M 12 + Micy + 12i = 1 We can easily see that this set satisfies the dealiasing TABLE 1. condition and exact restoration condition. Daubechies wave Wavelet transform representation of an image (two dimensional matrix) lets are the only compact solutions. For biorthogonal wave f? H.D. Horiz. Det. Horizontal Details lets we have the relations: = 2 j = 1 j = 0 W.D. D.D = 2 = 2 Vert. Det. Diag. Det. j = 1 j = 1 Vertical Details Diagonal Details j = 0 j = 0 and 1811. The wavelet transform can be interpreted as the 1 decomposition on frequency sets with a spatial orientation. 1812. The a trous Algorithm Which also satisfy the dealiasing condition and exact res 1813. The discrete approach of the wavelet transform can toration condition. A large class of compact wavelet func be done with the special version of the so-called a trous tions can be derived. Many sets of filters were proposed, algorithm (with holes). One assumes that the sampled data especially for coding. The choice of these filters must be {co(k)} are the scalar products at pixels k of the function guided by the regularity of the Scaling and the wavelet f(x) with a scaling function p(x) which corresponds to a low functions. The complexity is proportional to N. The algo pass filter. rithm provides a pyramid of N elements. 1814 The first filtering is then performed by a twice 1805. The 2D algorithm is based on separate variables magnified scale leading to the c(k) set. The signal dif leading to prioritizing of X and y directions. The scaling ference {co(k)}-c(k) contains the information between these two scales and is the discrete set associated with the function is defined by: wavelet transform corresponding to p(x). The associated wavelet is therefore p(x). The passage from a resolution to the next one is done by: le() = co-le) The distance between samples increasing by a factor 2 from the scale (i-1) (i>0) to the next one, c,(k) is given by: 1806. The detail signal is obtained from three wavelets: 1807) a vertical wavelet: 1808 a horizontal wavelet: 1815 and the discrete wavelet transform w(k) by: w; (k)=c(k)-c;(k) 1809 a diagonal wavelet: The coefficients {h(k) derive from the scaling function US 2006/0167784 A1 Jul. 27, 2006 1822) If we choose a B-spline for the scaling function, the coefficients of the convolution mask in one dimension 1 | . a ( 13 1. 1 1816. The algorithm allowing one to rebuild the data 1648 4 16 frame is evident: the last Smoothed array c, is added to all the differences w. and in two dimensions: np co(k) = c(k) cock) 1817) If we choose the linear interpolation for the scaling function (p. (p(x)=1-x if x ee-1, 1 (p(x)=0 if x gi-1, 1 1818 we have: 1823) The Wavelet Transform Using the Fourier Trans sh;)()=. = h(x + 1) + 3 (x)+1}(x-1)(x) + (x-1 form 1824 We start with the set of scalar products c(k)= c is obtained by: < f(x), p(X-k)2. If p(x) has a cut-off frequency C1 ( k)) = 1ico k - 1 ) + sco1 k) + aco1 k + 11) i 1819) and c is obtained from c. by: the data are correctly sampled. The data at the resolution j=1 a. 1 1 1 c -1(k) = ac (k - 2) + scick) -- ac (k + 2) The wavelet coefficients at the scale j are: 1825) and we can compute the set c(k) from c(k) with a discrete filter h(v): 1 . 1 1 C+1(k) -ic(k-2) + sc (k)- icy (k+2') 1820. The above a trous algorithm is easily extensible to the two dimensional space. This leads to a convolution with a mask of 3x3 pixels for the wavelet connected to linear O if v sv < 2 interpolation. The coefficents of the mask are: and Wv; W in h(y + n) = h(v) 1 1 1 16 8 16 1 1 1 8 4 8 1826 where n is an integer. So: 1 1 1 6, (v)=6;(v)h(2 v) 16 8 16 The cut-off frequency is reduced by a factor 2 at each step, allowing a reduction of the number of samples by this factor. 1821) At each scale j, we obtain a set {w (k, 1)} (we will call it wavelet plane in the following), which has the same 1827. The wavelet coefficients at the scale j+1 are: number of pixels as the image. US 2006/0167784 A1 Jul. 27, 2006 and they can be computed directly from c. (k) by: 8(u, v) = bar), with r = V(u? -- y?). 1828 where g is the following discrete filter: It is an isotropic function. 1830) The wavelet transform algorithm with n scales is the following one: 1831 1. We start with a B3-Spline scaling function and we derive up, h and g numerically. 1832 2. We compute the corresponding image FFT. We name To the resulting complex array; 1833 3. We set to 0. We iterate: The frequency band is also reduced by a factor 2 at each (1834) 4. We multiply T. by g(2u, 2v). We get the com step. Applying the sampling theorem, we can build a pyra plex array W. The inverse FFT gives the wavelet coeffi mid of cients at the scale 2: 1835) 5. We multiply T. by h(2u, 2v). We get the array T. Its inverse FFT gives the image at the scale 2. The frequency band is reduced by a factor 2. 1836 6. We increment elements. For an image analysis the number of elements is 1837 7. If is n, we go back to 4. 1838 8. The set {w, w, . . . . w np s en} describes the wavelet transform. 1839. If the wavelet is the difference between two reso i lutions, we have: The overdetermination is not very high. 1829. The B-spline functions are compact in this directe g(v)=1-h(v) space. They correspond to the autoconvolution of a square then the wavelet coefficients w(v) can be computed by function. In the Fourier space we have: c(v)-e, (v). 1840. The Reconstruction r sin ty'' 1841. If the wavelet is the difference between two reso 7 lutions, an evident reconstruction for a wavelet transform W={w, w, . . . . Wn en} is: B(x) is a set of 4 polynomials of degree 3. We choose the scaling function p(v) which has a B(x) profile in the Fourier Space: 3 But this is a particular case and other wavelet functions can d(v) = 5 Bs2(4) be chosen. The reconstruction can be done step by step, starting from the lowest resolution. At each scale, we have the relations: In the direct space we get: tx 4 1842) we look for c, knowing cl, w, h and g. We co- 3 Sin7. restore c, (v) with a least mean square estimator: 4. w(v)-g(2v)é, (v) 1843) is minimum. p(v) and p(v) are weight functions This function is quite similar to a Gaussian one and con which permit a general solution to the restoration of c(v). verges rapidly to 0. For 2-D the scaling function is defined By c(v) derivation we get: by US 2006/0167784 A1 Jul. 27, 2006 82 where the conjugate filters have the expression: wavelet analysis may be useful for characterizing and ana lyzing only a limited range of events. Advantageously, if an event is recognized with high reliability within a transform p, (v)h(v) domain, the event may be extracted from the data represen h(y) = tation and an inverse transform performed to provide the p, (v)h(v) + p(v) g(v) data set absent the recognized feature or event. This allows s p(v)g(v) a number of different feature-specific transforms to be g(v) = conducted, and analyzed. This analysis may be in series, that p(v)h(v) + p(v)lg(v) is, having a defined sequence of transforms, feature extrac tions, and inverse transforms. On the other hand, the process may be performed in parallel. That is, the data set is 1844. It is easy to see that these filters satisfy the exact subjected to various “tests”, which are conducted by opti reconstruction equation. In fact, above pair of equations give mally transforming the data to determine if a particular the general Solution to this equation. In this analysis, the feature (event) is present, determined with high reliability. Shannon sampling condition is always respected. No alias As each feature is identified, the base data set may be ing exists, so that the dealiasing condition is not necessary updated for the remaining “tests”, which will likely simplify (i.e., it is satisfied as a matter of course). the respective analysis, or improve the reliability of the 1845 The denominator is reduced if we choose: respective determination. As each event or feature is extracted, the data set becomes simpler and simpler, until only noise remains. This corresponds to the case where the wavelet is the 1856. It should be noted that, in some instances, a high difference between the square of two resolutions: reliability determination of the existence of an event cannot (2v) = p(v)°-6(2v)? be concluded. In those cases, it is also possible to perform a contingent analysis, leading to a plurality of possible The reconstruction algorithm is: results for each contingency. Thus, a putative feature is 1846 1. We compute the FFT of the image at the low extracted or not extracted from the data set and both results resolution. passed on for further analysis. Where one of the contingen cies is inconsistent with a Subsequent high reliability deter 1847) 2. We set to n. We iterate: mination, that entire branch of analysis may be truncated. 1848) 3. We compute the FFT of the wavelet coefficients Ideally, the output consists of a data representation with at the scale j. probabilistic representation of the existence of events or features represented within the data set. As discussed below, 1849 4. We multiply the wavelet coefficients w, by g. this may form the basis for a risk-reliability output space (i) 5. We multiply the image at the lower resolution c. representation of the data, useable directly by a human by h. (typically in the form of a visual output) and/or for further automated analysis. 1851) 6. The inverse Fourier Transform of the addition of wg and chgives the image c-1. 1857. It is also noted that the data set is not temporally static, and therefore the analysis may be conducted in real 1852) 7.j=-1 and we go back to 3. time based on a stream of data. 1853. The use of a scaling function with a cut-off fre 1858. The Process to Be Estimated quency allows a reduction of sampling at each scale, and limits the computing time and the memory size. 1859. The Kalman filter addresses the general problem of trying to estimate the state x 6. R" of a discrete-time 1854 Thus, it is seen that the DWT is in many respects controlled process that is governed by the linear stochastic comparable to the DFT, and, where convenient, may be difference equation employed in place thereof. While substantial work has been done in the application of wavelet analysis and filtering to (3.1) image data, it is noted that the wavelet transform analysis is 1860 with a measurement Ze R" that is not so limited. In particular, one embodiment of the present invention applies the transform to describe statistical events represented within a multidimensional data-space. By 1861. The random variables w and V represent the understanding the multi-resolution interrelationships of vari process and measurement noise (respectively). They are ous events and probabilities of events, in a time-space assumed to be independent (of each other), white, and with representation, a higher level analysis is possible than with normal probability distributions other common techniques. Likewise, because aspects of the p(w)-ND Q), (3.3) analysis are relatively content dependent, they may be accelerated by digital signal processing techniques or array p(v)-NO, R). (3.4) processors, without need to apply artificial intelligence. On 1862. In practice, the process noise covariance Q and the other hand, the transformed (and possibly filtered) data measurement noise covariance R matrices might change set, is advantageously suitable for intelligent analysis, either with each time step or measurement, however here we by machine or human. assume they are constant. 1855 Generally, there will be no need to perform an 1863 Kalman, Rudolph, Emil, “New Approach to Linear inverse transform on the data set. On the other hand, the Filtering and Prediction Problems”. Transactions of the US 2006/0167784 A1 Jul. 27, 2006 ASME Journal of Basic Engineering, 82D:35-45 (1960) (describes the namesake Kalman filter, which is a set of mathematical equations that provides an efficient computa K. = P, HT (HP, H+R) (3.8) tional (recursive) solution of the least-squares method. The filter is very powerful in several aspects: it supports estima P. H. tions of past, present, and even future states, and it can do HP, HT + R So even when the precise nature of the modeled system is unknown.) 1872 Looking at (3.8) we see that as the measurement 1864. The nxn matrix A in the difference equation (3.1) error covariance R approaches Zero, the gain K weights the relates the state at the previous time step k-1 to the state at residual more heavily. Specifically, the current step k, in the absence of either a driving function or process noise. Note that in practice A might change with lim Ka = H'. each time step, but here we assume it is constant. The nxl (2) - 0 (2) matrix B relates the optional control input u e R' to the state X. The mxn matrix H in the measurement equation (3.2) (2) indicates text missing or illegiblewhen filed relates the state to the measurement Zk. In practice H might change with each time step or measurement, but here we assume it is constant. 1873. On the other hand, as the a priori estimate error covariance P approaches Zero, the gain K weights the 1865. The Computational Origins of the Filter residual less heavily. Specifically, 1866) We define xe R" (note the “super minus') to be our a priori State estimate at Step k given knowledge of the process prior to step k, and xe R to be our a posteriori (2)lim - 0 Ka(2) = 0. state estimate at step k given measurement Z. We can then define a priori and a posteriori estimate errors as (2) indicates text missing or illegiblewhen filed e =x-x and 1874. Another way of thinking about the weighting by K €15x-x. is that as the measurement error covariance R approaches 1867. The a priori estimate error covariance is then Zero, the actual measurement Z is “trusted more and more, while the predicted measurement HX is trusted less and P =Eee." (3.5) less. On the other hand, as the a priori estimate error 1868 and the a posteriori estimate error covariance is covariance P A-approaches Zero the actual measurement Z, is trusted less and less, while the predicted measurement P=Eleke.". (3.6) HX is trusted more and more. 1869. In deriving the equations for the Kalman filter, we 1875. The Probabilistic Origins of the Filter begin with the goal of finding an equation that computes an a posteriori state estimate x as a linear combination of an a 1876. The justification for (3.7) is rooted in the probabil priori estimate x and a weighted difference between an ity of the a priori estimate x conditioned on all prior actual measurement Z, and a measurement prediction HX. measurements Z (Bayes' rule). For now let it suffice to point as shown below in (3.7). Some justification for (3.7) is given out that the Kalman filter maintains the first two moments of in “The Probabilistic Origins of the Filter” found below. See, the state distribution, http://www.cs.unc.edu/-welch/kalman/kalman filter/kal man-1.htm, expressly incorporated herein by reference. 1877. The a posteriori state estimate (3.7) reflects the mean (the first moment) of the state distribution it is 1870. The difference (Z-Hx) in (3.7) is called the normally distributed if the conditions of (3.3) and (3.4) are measurement innovation, or the residual. The residual met. The a posteriori estimate error covariance (3.6) reflects reflects the discrepancy between the predicted measurement the variance of the state distribution (the second non-central He and the actual measurement L. A residual of Zero moment). In other words, means that the two are in complete agreement. 1871. The nxm matrix K in (3.7) is chosen to be the gain or blending factor that minimizes the a posteriori error covariance (3.6). This minimization can be accomplished by first substituting (3.7) into the above definition for e. Substituting that into (3.6), performing the indicated expec 1878 For more details on the probabilistic origins of the tations, taking the derivative of the trace of the result with Kalman filter, see Maybeck79; Brown92; Jacobs93). respect to K, setting that result equal to Zero, and then solving for K. For more details see Maybeck79; Brown92; 1879. The Discrete Kalman Filter Algorithm Jacobs93). One form of the resulting K that minimizes (3.6) 1880. The Kalman filter estimates a process by using a is given by form of feedback control: the filter estimates the process US 2006/0167784 A1 Jul. 27, 2006 state at some time and then obtains feedback in the form of Wiener filter Brown'92 which is designed to operate on all (noisy) measurements. As such, the equations for the Kal of the data directly for each estimate. The Kalman filter man filter fall into two groups: time update equations and instead recursively conditions the current estimate on all of measurement update equations. The time update equations the past measurements. FIG. 6 offers a complete picture of are responsible for projecting forward (in time) the current the operation of the filter, combining the high-level diagram state and error covariance estimates to obtain the a priori of FIG. 5 with the equations (3.9) to (3.13). estimates for the next time step. The measurement update equations are responsible for the feedback—i.e. for incor 1886 Filter Parameters and Tuning porating a new measurement into the a priori estimate to 1887. In the actual implementation of the filter, the mea obtain an improved a posteriori estimate. Surement noise covariance R is usually measured prior to 1881. The time update equations can also be thought of as operation of the filter. Measuring the measurement error predictor equations, while the measurement update equa covariance R is generally practical (possible) because we tions can be thought of as corrector equations. Indeed the need to be able to measure the process anyway (while final estimation algorithm resembles that of a predictor operating the filter) so we should generally be able to take corrector algorithm for solving numerical problems as Some off-line sample measurements in order to determine shown below in FIG. 5, which shows the ongoing discrete the variance of the measurement noise. Kalman filter cycle. The time update projects the current 1888. The determination of the process noise covariance state estimate ahead in time. The measurement update Q is generally more difficult as we typically do not have the adjusts the projected estimate by an actual measurement at ability to directly observe the process we are estimating. that time. Sometimes a relatively simple (poor) process model can produce acceptable results if one "injects’ enough uncer 1882. The specific equations for the time and measure tainty into the process via the selection of Q. Certainly in this ment updates are presented below: case one would hope that the process measurements are reliable. Discrete Kalman filter time update equations. 1889. In either case, whether or not we have a rational basis for choosing the parameters, often times Superior filter 3 = Aš 1 + But (3.9) performance (statistically speaking) can be obtained by P = AP. A + Q (3.10) tuning the filter parameters Q and R. The tuning is usually performed off-line, frequently with the help of another (distinct) Kalman filter in a process generally referred to as 1883. Again notice how the time update equations (3.9) system identification. and (3.10) project the state and covariance estimates forward 1890 Under conditions where Q and R are in fact from time step k-1 to step k. A and B are from (3.1), while constant, both the estimation error covariance P and the Q is from (3.3). Initial conditions for the filter are discussed Kalman gain K will stabilize quickly and then remain in the earlier references. constant (see the filter update equations in FIG. 6). If this is the case, these parameters can be pre-computed by either running the filter off-line, or for example by determining the Discrete Kalman filter measurement update equations. steady-state value of P as described in Grewal 93). K. = P, HT (HP, H+R) (3.11) 1891. It is frequently the case however that the measure ment error (in particular) does not remain constant. For i = x + Ki (zi - H3) (3.12) example, observing like transmitters, the noise in measure P = (1 - KH)P, (3.13) ments of nearby transmitters will generally be smaller than that in far-away transmitters. Also, the process noise Q is Sometimes changed dynamically during filter operation— 1884 The first task during the measurement update is to becoming Qi in order to adjust to different dynamics. For compute the Kalman gain, K. Notice that the equation example, in the case of tracking the head of a user of a 3D given here as (3.11 is the same as (3.8). The next step is to virtual environment we might reduce the magnitude of Q if actually measure the process to obtain Z, and then to the user seems to be moving slowly, and increase the generate an a posteriori State estimate by incorporating the magnitude if the dynamics start changing rapidly. In Such measurement as in (3.12). Again (3.12) is simply (3.7) cases Q might be chosen to account for both uncertainty repeated here for completeness. The final step is to obtain an about the users intentions and uncertainty in the model. a posteriori error covariance estimate via (3.13. All of the Kalman filter equations can be algebraically manipulated 1892) 2 The Extended Kalman Filter (EKF) into to several forms. Equation (3.8) represents the Kalman 1893. The Process to Be Estimated gain in one popular form. 1894 As described above, the Kalman filter addresses the 1885. After each time and measurement update pair, the general problem of trying to estimate the state x 6. R" of a process is repeated with the previous a posteriori estimates discrete-time controlled process that is governed by a linear used to project or predict the new a priori estimates. This stochastic difference equation. But what happens if the recursive nature is one of the very appealing features of the process to be estimated and (or) the measurement relation Kalman filter it makes practical implementations much ship to the process is non-linear? Some of the most inter more feasible than (for example) an implementation of a esting and Successful applications of Kalman filtering have US 2006/0167784 A1 Jul. 27, 2006 been such situations. A Kalman filter that linearizes about 1905 x is an a posteriori estimate of the state at step the current mean and covariance is referred to as an extended k, Kalman filter or EKF. 1906 the random variables w and V represent the 1895. In something akin to a Taylor series, we can process and measurement noise as in (3.3) and (4.4). linearize the estimation around the current estimate using the partial derivatives of the process and measurement functions 1907) A is the Jacobian matrix of partial derivatives of to compute estimates even in the face of non-linear rela f with respect to X, that is tionships. To do so, we must begin by modifying some of the analysis presented above. Let us assume that our process again has a state vector x 6 R", but that the process is now governed by the non-linear stochastic difference equation x-f(x-1, ul-wk), (4.1) 1896 with a measurement Ze R" that is 1908 W is the Jacobian matrix of partial derivatives of Zi-h(xlv), (4.2) f with respect to w, 1897 where the random variables w and V again rep resent the process and measurement noise as in (4.3) and (4.4). In this case the non-linear function f in the difference equation (4.1) relates the state at the previous time step k-1 to the state at the current time step k. It includes as parameters any driving function V and the Zero-mean pro 1909 H is the Jacobian matrix of partial derivatives of cess noise w. The non-linear function h in the measurement h with respect to X, equation (4.2) relates the state X to the measurement Z. See, http://www.cs.unc.edu/-welch/kalman/kalman filter/kal man-2.html, expressly incorporated herein by reference. 1898. In practice of course one does not know the indi (X, D), vidual values of the noise w and V at each time step. However, one can approximate the state and measurement vector without them as 1910 V is the Jacobian matrix of partial derivatives of x=f(x, u text missing or illegible h with respect to V, when filed) (4.3) and h z=h(xDtext missing or illegible when t (X, D), filed), (4.4) where x is some a posteriori estimate of the state (from a previous time step k). 1911 Note that for simplicity in the notation we do not use the time step subscript k with the Jacobians A. W. H. and 1899. It is important to note that a fundamental flaw of V, even though they are in fact different at each time step. the EKF is that the distributions (or densities in the con tinuous case) of the various random variables are no longer 1912. Now we define a new notation for the prediction normal after undergoing their respective nonlinear transfor error, mations. The EKF is simply an ad hoc state estimator that only approximates the optimality of Bayes' rule by linear ization. Some interesting work has been done by Julier et al. 1913 and the measurement residual, in developing a variation to the EKF, using methods that e=z-Z. (4.8) preserve the normal distributions throughout the non-linear 1914) Remember that in practice one does not have transformations Julier96. access to X in (4.7), it is the actual state vector, i.e. the 1900. The Computational Origins of the Filter quantity one is trying to estimate. On the other hand, one 1901) To estimate a process with non-linear difference does have access to Z in (4.8), it is the actual measurement and measurement relationships, we begin by writing new that one is using to estimate X. Using (4.7) and (4.8) we can governing equations that linearize an estimate about (4.3) write governing equations for an error process as and (4.4). 1915 where e and m represent new independent random variables having zero mean and covariance matrices WQW' and VRV", with Q and R as in (3.3) and (3.4) respectively. 1903 X and Z are the actual state and measurement 1916 Notice that the equations (4.9) and (4.10) are linear, Vectors, and that they closely resemble the difference and measure 1904 x and Z are the approximate state and measure ment equations (3.1) and (3.2) from the discrete Kalman ment vectors from (4.3) and (4.4). filter. This motivates us to use the actual measurement US 2006/0167784 A1 Jul. 27, 2006 residual ein (4.8) and a second (hypothetical) Kalman filter to estimate the prediction error egiven by (4.9). This esti mate, call ite, could then be used along with (4.7) to obtain EKF measurement update equations. the a posteriori state estimates for the original non-linear process as K. = P. H. (H. P. Hi + V. R. V.) (4.16) i = x + Ki (zi - h(, D)) (4.17) x=x+é (4.11) P = (1 - KH)P, (4.18) 1917. The random variables of (4.9) and (4.10) have approximately the following probability distributions (see the previous footnote): 1923. As with the basic discrete Kalman filter, the mea Surement update equations (4.16), (4.17) and (4.18) correct p(e)-Nctext missing or illegible when the state and covariance estimates with the measurement Z. filed), Eeet) Again h in (4.17) comes from (3.4), H and V are the p(e)-N(text missing or illegible measurement Jacobians at step k, and R is the measurement noise covariance (3.4) at step k. (Note we now subscript R when filed), WQWT) allowing it to change with each measurement.) p(n)-N(text missing or illegible 1924. The basic operation of the EKF is the same as the when filed), VRVT) linear discrete Kalman filter as shown in FIG. 5. FIG. 7 1918 Given these approximations and letting the pre offers a complete picture of the operation of the EKF, dicted value of el be zero, the Kalman filter equation used to combining the high-level diagram of FIG. 5 with the equa estimate e is tions (4.14) through (4.18). 1925. An important feature of the EKF is that the Jaco e=Ke. (4.12) bian H in the equation for the Kalman gain H. serves to 1919. By substituting (4.12) back into (4.11) and making correctly propagate or “magnify only the relevant compo use of (4.8) we see that we do not actually need the second nent of the measurement information. For example, if there (hypothetical) Kalman filter: is not a one-to-one mapping between the measurement Z. and the state via h, the Jacobian H affects the Kalman gain so that it only magnifies the portion of the residual Z-h(x Dtext missing or illegible when filed) that does affect it = k + Ke, (4.13) the state. Of course if over all measurements there is not a one-to-one mapping between the measurement Z and the = x + K. (2 - 3) state via h, then as you might expect the filter will quickly diverge. In this case the process is unobservable. 1926) The Process Model 1920 Equation (4.13) can now be used for the measure 1927. In a simple example we attempt to estimate a scalar ment update in the extended Kalman filter, with X and Z random constant, a Voltage for example. Let's assume that coming from (4.3) and (4.4), and the Kalman gain K we have the ability to take measurements of the constant, but coming from (3.11) with the appropriate substitution for the that the measurements are corrupted by a 0.1 volt RMS measurement error covariance. white measurement noise (e.g. our analog to digital con verter is not very accurate). In this example, our process is 1921. The complete set of EKF equations is shown governed by the linear difference equation below. Note that we have substituted X for x to remain consistent with the earlier “super minus’ a priori notation, and that we now attach the subscript k to the Jacobians A, xi = Ax1 + But + wik W. H. and V, to reinforce the notion that they are different at Xk- Wik, (and therefore must be recomputed at) each time step. EKF time update equations. 1928 with a measurement ZeR' that is 3 = f(x-1, ui, D) (4.14) P = A. P. A + W. Q.-W. (4.15) Xi V. 1922. As with the basic discrete Kalman filter, the time update equations (4.14) and (4.15) project the state and 1929. The state does not change from step to step so A=1. There is no control input so u=text missing or illegible covariance estimates from the previous time step k-1 to the When filed. Our noisy measurement is of the state directly current time step k. Again fin (4.14) comes from (4.3). A so H=1. (Notice that we dropped the subscript k in several and W are the process Jacobians at Step k, and Q is the places because the respective parameters remain constant in process noise covariance (3.3) at step k. our simple model.) US 2006/0167784 A1 Jul. 27, 2006 1930. The Filter Equations and Parameters 1941 Lewis86 Lewis, Richard. 1986. Optimal Estima tion with an Introduction to Stochastic Control Theory, John 1931 Our time update equations are Wiley & Sons, Inc. x=x-1, 1942 Maybeck79 Maybeck, Peter S. 1979. Stochastic P =P+Q, Models, Estimation, and Control, Volume 1, Academic 1932 and our measurement update equations are Press, Inc. 1943) Sorenson70 Sorenson, H. W. 1970. “Least-Squares estimation: from Gauss to Kalman.” IEEE Spectrum, Vol. 7, K = P (P + R) (5.1) pp. 63-68, July 1970. P. 1944) See, also: P + R 1945) “A New Approach for Filtering Nonlinear Sys i = x + K. (2 - ), tems’ by S. J. Julier, J. K. Uhlmann, and H. F. Durrant Whyte, Proceedings of the 1995 American Control Confer P = (1 - K)P. ence, Seattle, Wash., Pages: 1628-1632. Available from http://www.robots.ox.ac.uk/~siju/work/publications/ ACC95 przip Simon Julier's home page at http://www.ro 1933 Presuming a very small process variance, we let bots.ox.ac.uk/~siju/. Q=1e-5 (We could certainly let Q=text missing or illegible when filed but assuming a small but non-zero 1946 “Fuzzy Logic Simplifies Complex Control Prob value gives us more flexibility in “tuning the filter as we lems”. Tom Williams, Computer Design, Mar. 1, 1991. will demonstrate below.) Let's assume that from experience 1947. “Neural Network And Fuzzy Systems—A Dynami we know that the true value of the random constant has a standard normal probability distribution, so we will “seed' cal Systems Approach To Machine Intelligence', Bart our filter with the guess that the constant is 0. In other words, Kosko: Prentice Hall 1992: Englewood Cliffs, N.J.; pp. 13, before starting we let x=text missing or illegible 18, 19. when filed. 1948 B. Krogh et al., “Integrated Path Planning and 1934 Similarly we need to choose an initial value for Dynamic Steering Control for Autonomous Vehicles.” 1986. P. call it Po. If we were absolutely certain that our initial 1949 Brockstein, A., “GPS-Kalman-Augmented Inertial state estimatex-text missing or illegible when filed Navigation System Performance. Naecom 76 Record, pp. was correct, we would let P=text missing or illegible 864-868, 1976. When filed. However given the uncertainty in our initial 1950 Brooks, R. “Solving the Fine-Path Problem by estimate xc choosing P=Otext missing or illegible Good Representation of Free Space.” IEEE Transactions on when filed would cause the filter to initially and always Systems, Man, and Cybernetics, pp. 190-197, March-April, believex=text missing or illegible when filed. As it 1983. turns out, the alternative choice is not critical. We could choose almost any Pétext missing or illegible when 1951 Brown, R., “Kalman Filtering Study Guide A filed and the filter would eventually converge. It is conve Guided Tour.” Iowa State University, pp. 1-19, 1984. nient, for example, to start with Po-1. 1952 Brown, R., Random Signal Analysis & Kalman 1935 Brown'92 Brown, R. G. and P. Y. C. Hwang. 1992. Filtering, Chapter 5, pp. 181-209, no date. Introduction to Random Signals and Applied Kalman Fil 1953 D. Kuan et al., “Model-based Geometric Reason tering, Second Edition, John Wiley & Sons, Inc. ing for Autonomous Road Following, pp. 416–423, 1987. 1936 Gelb74 Gelb, A. 1974. Applied Optimal Estima 1954 D. Kuan, “Autonomous Robotic Vehicle Road Fol tion, MIT Press, Cambridge, Mass. lowing.” IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 647-658, 1988. 1937 Grewal.93 Grewal, Mohinder S., and Angus P. Andrews (1993). Kalman Filtering Theory and Practice. 1955 D. Touretzky et al., “What's Hidden in the Hidden Upper Saddle River, N.J. USA, Prentice Hall. Layers?.” Byte, pp. 227-233, August 1989. 1938) Jacobs93 Jacobs, O. L. R. 1993. Introduction to 1956) Data Fusion in Pathfinder and Travtek, Roy Sum Control Theory, 2nd Edition. Oxford University Press. ner, VNIS 91 conference, Oct. 20-23, Dearborn, Mich. 1939 Julier96 Julier, Simon and Jeffrey Uhlman. “A 1957 Database Accuracy Effects on Vehicle Positioning General Method of Approximating Nonlinear Transforma as Measured by the Certainty Factor, R. Borcherts, C. tions of Probability Distributions.” Robotics Research Collier, E. Koch, R. Bennet, VNIS 91 conference from Oct. Group, Department of Engineering Science, University of 20-23, Dearborn, Mich. Oxford cited 14 Nov. 1995). Available from http://www.ro 1958) Daum, F., et al., “Decoupled Kalman Filters for bots.ox.ac.uk/~siju/work/publications/Unscented.Zip. Phased Array Radar Tracking.” IEEE Transactions on Auto 1940 Kalman60 Kalman, R. E. 1960. “A New Approach matic Control, pp. 269-283, March 1983. to Linear Filtering and Prediction Problems.” Transaction of 1959) Denavit, J. et al., “A Kinematic Notation for the ASME Journal of Basic Engineering, pp. 35-45 (March Lower-Pair Mechanisms Bases on Matrices. pp. 215-221. 1960). June, 1955. US 2006/0167784 A1 Jul. 27, 2006 1960 Dickmanns, E. et al., “Guiding Land Vehicles 1978 Thomas Theu?B1 and Robert F. Tobler and Eduard Along Roadways by Computer Vision'. The Tools for Gröller, “The Multi-Dimensional Hartley Transform as a Tomorrow, Oct. 23, 1985. Basis for Volume Rendering', citeseer.nj.nec.com/ 450842.html. 1961 Edward J. Krakiwsky, “A Kalman Filter for Inte grating Dead Reckoning, Map Matching and GPS Position 1979) See also, U.S. Pat. Nos. (expressly incorporated ing, IEEE Plans x88 Position Location and Navigation herein by reference): U.S. Pat. Nos. 3,582,926; 4,291,749; Symposium Record, Kissemee, Fla. USA, Nov. 29-Dec. 2, 4,314,232; 4,337,821: 4,401,848; 4,407,564; 4,419,730; 1988, pp. 39-46. 4441405; 4,451,887; 4,477,874; 4,536,739; 4,582,389: 4,636,782; 4,653,003; 4,707,788; 4,731,769; 4,740,779; 1962 Fuzzy Systems and Applications, United Signals 4,740,780; 4,752,824; 4,787,039; 4,795,223; 4,809, 180; and Systems, Inc., Bart Kosko with Fred Watkins, Jun. 5-7, 4,818,048; 4,827,520; 4.837,551; 4.853,687; 4,876,594; 1991. 4,914,705; 4,967, 178; 4.988,976; 4,995,258; 4.996,959; 1963 IEEE Journal of Robotics & Automation, vol. 4, 5,006,829; 5,043,736; 5,051,735; 5,070,323; 5,070,931; No. 4, August 1988, IEEE (New York) J. LeM “Domain 5,119,504; 5,198.797; 5,203,499; 5,214,413: 5,214,707; dependent reasoning for visual navigation of roadways, pp. 5,235,633; 5,257, 190; 5,274,560; 5,278,532: 5,293,115; 419-427 (Nissan) Mar. 24, 1988. 5,299,132; 5,334,974; 5,335,276; 5,335,743; 5,345,817; 5,351,041: 5,361,165; 5,371,510; 5,400,045; 5,404,443; 1964) J. Crowley, “Part 3: Knowledge Based Supervision 5,414,439; 5,416,318; 5,422,565; 5,432,904; 5,440,428; of Robotics Systems.” 1989 IEEE Conference on Robotics 5,442,553: 5,450.321; 5.450,329; 5,450,613; 5,475,399; and Automation, pp. 37-42, 1989. 5,479,482; 5,483,632: 5,486,840; 5,493,658; 5,494,097; 5,497,271; 5,497.339: 5,504,622; 5,506,595; 5,511,724; 1965 Kaczmarek, K. W., “Cellular Networking: A Car 5,519,403; 5,519,410; 5,523,559; 5,525,977: 5,528,248; rier's Perspective', 39th IEEE Vehicular Technology Con 5,528,496; 5,534,888: 5,539,869; 5,547,125; 5,553,661; ference, May 1, 1989, vol. 1, pp. 1-6. 5,555,172; 5,555,286; 5,555,502; 5,559,520; 5,572.204; 1966. Knowledge Representation in Fuzzy Logic, Lotfi 5,576,724; 5,579,535; 5,627,547; 5,638,305; 5,648,769; A. Zadeh, IEEE Transactions on Knowledge and Data 5,650,929; 5,653.386; 5,654,715; 5,666,102; 5,670,953; Engineering, vol. 1, No. 1, March 1989. 5,689,252: 5,691,.695; 5,702,165; 5,712,625; 5,712,640; 5,714,852; 5,717,387; 5,732,368; 5,734,973; 5,742,226; 1967 Sennott, J. et al., “AQueuing Model for Analysis of 5,752,754; 5,758.311 5,777,394; 5,781872; 5,919,239; A Bursty Multiple-Access Communication Channel.” IEEE, 6,002,326; 6,013,956; 6,078,853; 6,104,101; and 6,449,535. pp. 317-321, 1981. 1980) M. Krebs, “Cars That Tell You Where To Go,” The 1968. Sheridan, T. “Three Models of Preview Control.” New York Times, Dec. 15, 1996, section 11, p. 1. IEEE Transactions on Human Factors in Electronics, pp. 91-102, June 1966. 1981 L. Kraar, “Knowledge Engineering.” Fortune, Oct. 28, 1996, pp. 163-164. 1969. Sheth, P., et al., “A Generalized Symbolic Notation 1982 S. Heuchert, “Eyes Forward: An ergonomic solu for Mechanism.” Transactions of the ASME, pp. 102-112, tion to driver information overload.” Society of Automobile February 1971. Engineering, September 1996, pp.27-31. 1970 Sorenson, W., "Least-Squares estimation: From 1983) J. Braunstein, “Airbag Technology Take Off,” Gauss to Kalman.” IEEE Spectrum, pp. 63-68, July 1970. Automotive & Transportation Interiors, August 1996, p. 16. 1971 “Automobile Navigation System Using Beacon 1984 I. Adcock, “No Longer Square. Automotive & Information' pp. 139-145. Transportation Interiors, August 1996, p. 38 1972 W. Uttal, “Teleoperators,” Scientific American, pp. 1985 One embodiment of the present invention advances 124-129, December 1989. the art by explicitly communicating reliability or risk infor mation to the user. Therefore, in addition to communicating 1973 Wareby, Jan, “Intelligent Signaling: FAR & SS7, an event or predicted event, the system also computes or Cellular Business, pp. 58, 60 and 62, July 1990. determines a reliability of the information and outputs this 1974 Wescon/87 Conference Record, vol. 31, 1987, (Los information. The reliability referred to herein generally is Angeles, US) M. T. Allison et al “The next generation unavailable to the original detection device, though Such navigation system, pp. 941-947. device may generate its own reliability information for a sensor reading. 1975 Ekaterina L.-Rundblad, Alexei Maidan, Peter Novak, Valeriy Labunets, Fast Color Wavelet-Haar Hartley 1986 Therefore, the user interface according to this Prometheus Transforms For Image Processing, www embodiment is improved by outputting information relating prometheus-inc.com/asi/algebra2003/papers/katya2.pdf. to both the event and a reliability or risk with respect to that information. 1976 Richard Tolimieri and Myoung An, Group Filters And Image Processing, www.prometheus-inc.com/asi/alge 1987. According to a preferred embodiment of the inven bra2003/papers/tolimieri.pdf. tion, a vehicle travel information system is provided, for example integrated with a vehicular navigation system. In a 1977 Daniel N. Rockmore, Recent Progress And Appli symmetric peer-to-peer model, each vehicle includes both cations. In Group FFTs. http://www.prometheus-inc.com/asi/ environmental event sensors and a user interface, but the algebra2003/papers/rockmore.pdf. present invention is not dependent on both aspects being US 2006/0167784 A1 Jul. 27, 2006 present in a device. As the vehicle travels, and as time counterpart, along with a failure of a detection of an event advances, its context sphere is altered. For any context triggering a warning. Alternately, Such communications may sphere, certain events or sensed conditions will be most be explicit. relevant. These most relevant events or sensed, to the extent known by the system, are then output through a user 1994 The present invention can provide a mobile warn interface. However, often, the nature or existence of relevant ing system having a user interface for conveying an event or potentially relevant event is unreliable, or reliance thereon warning and an associated reliability or risk of reliance on entails risk. the warning. 1988. In the case of a vehicle traveling along a roadway, 1995 Preferably, the reliability or risk of reliance is there are two particular risks to analyze: first, that the assessed based on a time between original sensing and recorded event may not exist (false positive), and second, proximity. The reliability may also be based on the nature of that an absence of indication of an event is in error (false the event or sensed condition. An intrinsic reliability of the negative). For example, the degree of risk may be indicated original sensed event or condition may also be relayed, as by an indication of color (e.g., red, yellow green) or mag distinct from the reliability or risk of reliance assuming the nitude (e.g., a bar graph or dial). event or condition to have been accurately sensed. 1989. In many cases, the degree of risk is calculable, and 1996. In order to determine risk, often statistical and thus may be readily available. For example, if the event probabilistic techniques may be used. Alternately, non-linear sensor is a detection of police radar, reliability may be techniques, such as neural networks, may be employed. In inferred from a time since last recording of an event. If a car employing a probabilistic scheme, a sensor reading at time is traveling along a highway, and receives a warning of Zero, and the associated intrinsic probability of error are traffic enforcement radar from a car one mile ahead, there is stored. A model is associated with the sensor reading to a high degree of certainty that the traffic enforcement radar determine a decay pattern. Thus, in the case of traffic will actually exist as the vehicle proceeds along the highway. enforcement radar, the half-life for a “radar trap' for K band Further, if the traffic radar is in fixed location, there is a high radar being fixed in one location is, for example, about 5 degree of certainty that there is no traffic enforcement radar minutes. Thereafter, the enforcement officer may give a closer than one mile. On the other hand, if a warning of ticket, and proceed up the road. Thus, for times less than traffic radar at a given location is two hours old, then the risk three minutes, the probability of the traffic enforcement of reliance on this information is high, and the warning radar remaining in fixed position is high. For this same should be deemed general and advisory of the nature of risks time-period, the probability that the traffic enforcement in the region. Preferably, as such a warning ages, the officer has moved up the road against the direction of traffic temporal proximity of the warning is spread from its original flow is low. A car following 3 miles behind a reliable sensor focus. at 60 mph would therefore have a highly reliable indication of prospective conditions. As the time increases, so does the 1990. On the contrary, if the warning relates to a pothole risk; a car following ten miles behind a sensor would only in a certain lane on the highway, the temporal range of risk have a general warning of hazards, and a general indication is much broader: even a week later, the reliability of the of the lack thereof. However, over time, a general (and continued existence at that location remains high. However, possibly diurnal or other cyclic time-sensitive variation) risk over the course of a year, the reliability wanes. On the other of travel within a region may be established, to provide a hand, while there may be a risk of other potholes nearby, the baseline. particular detected pothole would not normally move. 1997. It is noted that the risks are not limited to traffic 1991. The algorithm may also be more complex. For enforcement radar or laser. Rather, the scheme according to example, if a traffic accident occurs at a particular location, the present invention is generalized to all sorts of risks. For there are generally acceptable predictions of the effect of the example, a sensor may detect or predict Sun glare. In this accident on road traffic for many hours thereafter. These case, a model would be quite accurate for determining include rubbernecking, migrations of the traffic pattern, and changes over time, and assuming a reliable model is secondary accidents. These considerations may be pro employed, this condition could generally be accurately grammed, and the set of events and datapoints used to predicted. predict spatial and temporal effects, as well as the reliability of the existence of such effects. This, in turn, may be used 1998 Another example is road flooding. This may be to advise a traveler to take a certain route to a destination. detected, for example, through the use of optical sensors, tire drag sensors, “splash' sensors, or other known sensors. In 1992 Eventually, the reliability of the information is this case, the relevant time-constant for onset and decay will inferred to be so low as to cause an expiration of the event, be variable, although for a given location, the dynamics may although preferably a statistical database is maintained to be modeled with some accuracy, based on sensed actual indicate geographic regional issues broadly. conditions, regional rainfall, ground saturation, and particu lar storm pattern. Therefore, a puddle or hydroplaning risk 1993. Therefore, the system and method according to the may be communicated to the driver in terms of location, present invention provides an output that can be considered likely magnitude, and confidence. “two dimensional’ (or higher dimensional); the nature of the warning, and the reliability of the warning. In conjunction, 1999. It is noted that these three independent parameters the system may therefore output a reliability of an absence need not all be conveyed to the user. For example, the of warning. In order to conserve communications band geographic proximity to an event location may be used to width, it is preferred that an absence of warning is inferred trigger an output. Therefore, no independent output of from the existence of a communications channel with a location may be necessary in this case. In some cases, the US 2006/0167784 A1 Jul. 27, 2006 90 magnitude of the threat is relevant, in other cases it is not. about the device or its context. This sensor may populate a In many present systems (e.g., radar detection), threat mag map or mapping system with historical map data. nitude is used as a surrogate for risk. However, it is well 2005. During use, a receiving device seeks to output understood that there are high magnitude artifacts, and low location context-relevant information to the user, and there magnitude true threats, and thus this paradigm has limited fore in this embodiment includes a human user interface. basis for use. The use of risk or confidence as an independent Typically, in a vehicle having a general linear or highly factor may be express or intermediate. Thus, a confidence constrained type path, a position output is not a critical threshold may be internally applied before communicating feature, and may be Suppressed in order to simplify the an event to the user. In determining or predicting risk or interface. Rather, a relative position output is more appro confidence, it may be preferred to provide a central database. priate, indicating a relative position (distance, time, etc.) Therefore, generally more complex models may be with respect to a potential contextually relevant position. In employed, Supported by a richer data set derived from many addition, especially in systems where a plurality of different measurements over an extended period of time. The central types of sensors or sensed parameters are available, the database may either directly perform the necessary compu nature of the relevant context is also output. Further, as a tations, or convey an appropriate model, preferably limited particular feature of the present invention, a risk or reliabil to the context (e.g., geography, time, general environmental ity assessment is indicated to the user. This risk or reliability conditions), for local calculation of risk. assessment is preferably statistically derived, although it 2000. The incorporated references relate, for example, to may be derived through other known means, for example methods and apparatus which may be used as part of, or in Boolean analysis, fuZZy logic, or neural networks. conjunction with the present invention. Therefore, it is 2006 For example, the device may provide weather understood that the present invention may integrate other information to the user. Through one or more of meteoro systems, or be integrated in other systems, having comple logical data from standard reporting infrastructure (e.g., mentary, Synergistic or related in some way. For example, NOAA, Accuweather R, etc.), mobile reporting nodes (e.g., common sensors, antennas, processors, memory, communi mobiles devices having weather sensors), satellite data, and cations hardware, Subsystems and the like may provide a other weather data sources, a local weather map is created, basis for combination, even if the functions are separate. preferably limited to contextual relevance. In most cases, 2001 The techniques according to the present invention this weather map is stored locally; however, if the quality of may be applied to other circumstances. Therefore, it is service for a communications link may be assured, a remote understood that the present invention has, as an object to database system serving one or more devices may be pro provide a user interface harnessing the power of statistical vided. For example, a cellular data communications system methods. Therefore, it is seen that, as an aspect of the present may be used to communicate with the Internet or a service invention, a user interface, a method of providing a user provider. interface, computer Software for generating a human-com 2007. The mobile unit, in operation, determines its posi puter interface, and a system providing Such a user interface, tion, and, though explicit user input and/or inferential analy presents a prediction of a state as well as an indication of a sis, determines the itinerary or expected path of the device statistical reliability of the prediction. and time sequence. The device (or associated systems) then 2002 Within a vehicular environment, the statistical determines the available weather information for the route analysis according to the present invention may also be used and anticipated itinerary (which may itself be dependent on to improve performance and the user interface of other the weather information and/or reaction thereto). This avail systems. In particular, modern vehicles have a number of able information is then modeled, for example using a indicators and warnings. In most known systems, warnings statistical model as described hereinabove, to predict the are provided at pre-established thresholds. According to the forthcoming weather conditions for the device or transport present invention, a risk analysis may be performed on ing vehicle. sensor and other data to provide further information for the user, e.g., an indication of the reliability of the sensor data, 2008 The device then determines the anticipated condi or the reliability under the circumstances of the sensor data tions and relevance sorts them. In this case, both positive and as basis for decision. (For example, a temperature sensor negative information may be useful, i.e., a warning about alone does not indicate whether an engine is operating bad weather, ice, freezing road Surfaces, fog, sand-storms, normally.) rain, Snow, sleet, hail, Sun glare, etc., and an indication of dry, warm, well-illuminated road surfaces may both be Fourth Embodiment useful information. 2009. In addition, through the analysis, a number of 2003 The present example provides a mobile telecom presumptions and predictions are made, for example using a munications device having a position detector, which may chain. Therefore, while the system may predict a most likely be absolute, relative, hybrid, or other type, and preferably a state of affairs, this alone does not provide sufficient infor communications device for communicating information, mation for full reliance thereon. For example, the present typically location relevant information. The device may road Surface freezing conditions thirty miles ahead on a road serve as a transmitter, transmitting information relevant to may be a poor indicator of the road conditions when the the location (or prior locations) of the device, a receiver, device is at that position. In addition to changes in the receiving information relevant to the location (or prospec weather, human action may be taken, Such as road salt, sand, tive location) of the device, or a composite. traffic, etc., which would alter the conditions, especially in 2004. In the case of a transmitter device or stand-alone response to a warning. On the other hand, a report of device, a sensor is provided to determine a condition of or freezing road conditions one mile ahead would generally US 2006/0167784 A1 Jul. 27, 2006 have high predictive value for the actual road conditions tioning system used in conjunction need not be high, in order when the device is at that location, assuming that the vehicle to provide high reliability position information. For is traveling in that direction. example, where it is desired to map potholes, positional accuracy of 10 cm may be desired, far more precise than 2010. In many cases, there is too much raw information might be available from a normal GPS receiver mounted in to effectively display to the user all relevant factors in a moving automobile. Systems having such accuracy may making a reliability or risk determination. Thus, the device then be used as part of an automated repair system. How outputs a composite estimation of the reliability or risk, ever, when combined with other data, location and identi which may be a numeric or non-parametric value. This is fication of such events is possible. Further, while the system output in conjunction with the nature of the alert and its may include or tolerate inaccuracies, it is generally desired contextual proximity. that the system have high precision, as compensation for 2011. As stated above, there will generally be a plurality inaccuracies may be applied. of events, each with an associated risk or reliability and 2015. A typical implementation of the device provides a location. The relevance of an event may be predicted based memory for storing events and respective locations. Prefer on the dynamics of the vehicle in which the device is ably, further information is also stored, such as a time of the transported and the nature of the event. Thus, if the vehicle event, its character or nature, and other quantitative or requires 170 feet to stop from a speed of 60 MPH, a warning qualitative aspects of the information or its source and/or which might trigger a panic stop should be issued between conditions of acquisition. This memory may be a solid state 170-500 feet in advance. If the warning is triggered closer memory or module (e.g., 64-256 MB Flash memory), rotat than 170 feet, preferably the warning indicates that the ing magnetic and/or optical memory devices, or other evasive maneuver will be necessary. known types of memory. 2012. In this case, the risk indicator includes a number of 2016. The events to be stored may be detected locally, factors. First, there is the reliability of the data upon which Such as through a detector for radar and/or laser emission the warning is based. Second, there is the reliability of the Source, radio Scanner, traffic or road conditions (mechanical predictive model which extrapolates from, the time the raw vehicle sensors, visual and/or infrared imaging, radar or data is acquired to the conjunction of the device and the LIDAR analysis, acoustic sensors, or the like), places of location of the event. Third, there is an assessment of the interest which may be selectively identified, itinerary stops, relative risks of responding to a false positive versus failing and/or fixed locations. The events may also be provided by to respond to a false negative. Other risks may also be a remote transmitter, with no local event detection. There included in the analysis. Together, the composite risk is fore, while means for identifying events having associated output, for example as a color indicator. Using, for example, locations is a part of the system as a whole. Such means need a tricolor (red-green-blue) light emitting diode (LED) or not be included in every apparatus embodying the invention. bicolor LED (red-green), a range of colors may be presented 2017 Radar detectors typically are employed to detect to the user. Likewise, in an audio alert, the loudness or operating emitters of X (10.5 GHZ), K (25 GHz) and Ka (35 harmonic composition (e.g., harmonic distortion) of a tone GHz) radar emissions from traffic control devices or law or alert signal may indicate the risk or reliability. (In the case enforcement personnel for detecting vehicle speed by the of loudness, preferably a microphone measures ambient Doppler effect. These systems typically operate as Superhet noise to determine a minimum loudness necessary to indi erodyne receivers which Sweep one or more bands, and cate an alert). detect a wave having an energy significantly above back 2013 The position detector is preferably a GPS or com ground. As such, these types of devices are subject to bined GPS-GLONASS receiver, although a network posi numerous sources of interference, accidental, intentional, tion detection system (e.g., Enhanced 911 type system) may and incidental. A known system, Safety Warning System also be employed. Preferably, the position detector achieves (SWS) licensed by Safety Warning System L.C., Englewood an accuracy of +30 meters 95% of the time, and preferably Fla., makes use of Such radar detectors to specifically warn provides redundant sensors, e.g., GPS and inertial sensors, motorists of identified road hazards. In this case, one of a set in case of failure or error of one of the systems. However, for of particular signals is modulated within a radar band by a Such purposes as pothole reporting, positional accuracies of transmitter operated near the roadway. The receiver decodes 1 to 3 meters are preferred. These may be obtained through the transmission and warns the driver of the hazard. a combination of techniques, and therefore the inherent 2018. LIDAR devices emit an infrared laser signal, which accuracy of any one technique need not meet the overall is then reflected off a moving vehicle and analyzed for delay, system requirement. which relates to distance. Through Successive measure 2014 The position detector may also be linked to a ments, a sped can be calculated. A LIDAR detector therefore mapping system and possibly a dead reckoning system, in seeks to detect the characteristic pulsatile infrared energy. order to pinpoint a position with a geographic landmark. 2019 Police radios employ certain restricted frequencies, Thus, while precise absolute coordinate measurements of and in some cases, police vehicles continuously transmit a position may be used, it may also be possible to obtain useful signal. While certain laws restrict interception of messages data at reduced cost by applying certain presumptions to sent on police bands, it is believed that the mere detection available data. In an automotive system, steering angle, and localization of a carrier wave is not and may not be compass direction, and wheel revolution information may be legally restricted. These radios tend to operate below 800 available, thereby giving a rough indication of position from MHZ, and thus a receiver may employ standard radio a known starting point. When this information is applied to technologies. a mapping system, a relatively precise position may be 2020 Potholes and other road obstructions and defects estimated. Therefore, the required precision of another posi have two characteristics. First, they adversely effect vehicles US 2006/0167784 A1 Jul. 27, 2006 92 which encounter them. Second, they often cause a secondary ating efficiency may be measured during vehicle use, allow effect of motorists seeking to avoid a direct encounter or ing an accurate prediction of fuel efficiency under dynami damage, by slowing or executing an evasive maneuver. cally changing conditions, such as acceleration. Vehicle These obstructions may therefore be detected in three ways: sensors may assist in making a determination that optimum first, by analyzing the Suspension of the vehicle for unusual acceleration is safe; objects both in front and behind the shocks indicative of Such vents; second, by analyzing speed vehicle may be sensed. If an object is in front of the vehicle, and steering patterns of the Subject vehicle and possibly and the closing speed would predict a collision, then the Surrounding vehicles; and third, by a visual, ultrasonic, or acceleration is decreased, or even brakes applied. If an other direct sensor for detecting the pothole or other obstruc object is rapidly advancing from the rear, the acceleration tion. Such direct sensors are known; however, their effec may be increased in order to avoid impact or reduce speed tiveness is limited, and therefore an advance mapping of differential. See, U.S. Pat. No. 6,445,308 (Koike, Sep. 3, Such potholes and other road obstructions greatly facilitates 2002, Positional data utilizing inter-vehicle communication avoiding vehicle damage and executing unsafe or emer method and traveling control apparatus), U.S. Pat. No. gency evasive maneuvers. An advance mapping may also be 6,436,005 (Bellinger, Aug. 20, 2002, System for controlling useful in remediation of Such road hazards, as well. drivetrain components to achieve fuel efficiency goals), U.S. 2021 Traffic jams occur for a variety of reasons. Typi Pat. No. 6,418,367 (Toukura, et al., Jul. 9, 2002, Engine cally, the road carries traffic above a threshold, and for some transmission control system), expressly incorporated herein reason the normal traffic flow patterns are disrupted. There by reference. fore, there is a dramatic slowdown in the average vehicle speed, and a reduced throughput. Because of the reduced 2024 Likewise, the operation of a vehicle may be opti throughput, even after the cause of the disruption has abated, mized approaching a stop. Such as a stop sign, red light, or the roadways may take minutes to hours to return to normal. the like. In this case, the system optimization may be more Therefore, it is typically desired to have advance warnings complex. In addition to fuel economy, wear on brakes, of disruptions, which include accidents, icing, rain, Sun engine (especially if compression braking is employed), glare, lane closures, road debris, police action, exits and transmission, tires, Suspension, time, accident-related risks, entrances, and the like, in order to allow the driver to avoid and the like, may also be included. In the case of a stop sign, the involved region or plan accordingly. Abnormal traffic the issue also arises with respect to a so-called “rolling patterns may be detected by comparing a vehicle speed to stop'. Such a practice provides that the vehicle does not the speed limit or a historical average speed, by a visual actually stop, but reaches a sufficiently low speed that the evaluation of traffic conditions, or by broadcast road advi driver could stop if required by circumstances. While this sories. High traffic conditions are associated with braking of practice is technically considered a violation, in many traffic, which in turn results in deceleration and the illumi instances, it is both efficient and useful. For example, a stop nation of brake lights. Brake lights may be determined by line is often located behind an intersection, with impaired both the specific level of illumination and the center brake visibility. Thus, the vehicle might come to a complete stop, light, which is not normally illuminated. Deceleration may begin to accelerate, and then find that the intersection is not be detected by an optical, radar or LIDAR sensor for clear, and be forced to stop again. One particular reason for detecting the speed and/or acceleration state of nearby a rolling stop is the storage of energy in the vehicular vehicles. Suspension during acceleration and deceleration. As the vehicle comes to a stop, the springs and shock absorbers of 2022 While a preferred embodiment of the present the Suspension undergo a damped oscillation, which is invention employs one or more sensors, broadcast adviso relatively uncorfortable, and destabilizes the vehicle and its ries, including those from systems according to or compat COntentS. ible with the present invention, provide a valuable source of information relating to road conditions and information of 2025. According to one aspect of the present invention, interest at a particular location. Therefore, the sensors need the driver may locate a deceleration target and/or a target not form a part of the core system. Further, some or all of the speed. The vehicle navigation system may assist, recording required sensors may be integrated with the vehicle elec an exact location of a stop line, geographic (hills, curves, tronics ("vetronics’), and therefore the sensors may be lane marker locations, etc.), weather conditions (ice, sand, provided separately or as options. It is therefore an aspect of puddles, etc.) and other circumstances Surrounding the an embodiment of the invention to integrate the transceiver, vehicle. Other vehicles and obstructions or pedestrians, etc. and event database into a Vetronics system, preferably using may also be identified and modeled. Using models of the a digital Vetronics data bus to communicate with existing various components, as well as cost functions associated systems, such as speed sensors, antilock brake sensors, with each, as well as Subjective factors, which may include cruise control, automatic traction system, Suspension, vehicle occupant time-cost and comfort functions, an opti engine, transmission, and other vehicle systems. mal acceleration or deceleration profile may be calculated. The system may therefore express control over throttle, 2023. According to one aspect of the invention, an adap brakes, transmission shifts, clutch, valve timing, Suspension tive cuise control system is provided which, in at least one controls, etc., in order to optimize vehicle performance. mode of operation, seeks to optimize various factors of vehicle operation, Such as fuel efficiency, acceleration, com 2026 See U.S. Pat. Nos. (expressly incorporated herein fort, tire wear, etc. For example, an automatic acceleration by reference): U.S. Pat. Nos. 6,503,170; 6,470,265; 6,445, feature is provided which determines a most fuel-efficient 308; 6,292,743; 6,292,736; 6,233,520; 6,230,098: 6,220, acceleration for a vehicle. Too slow an acceleration will 986; 6,202,022: 6,199,001: 6,182,000; 6,178,377; 6,174, result in increased time at Suboptimal gear ratios, while too 262: 6,098,016; 6,092,014; 6,092,005; 6,091,956; 6,070, fast acceleration will waste considerable fuel. Actual oper 118; 6,061,003; 6,052,645; 6,034,626; 6,014,605; 5,990, US 2006/0167784 A1 Jul. 27, 2006 825; 5,983,154; 5,938,707; 5,931,890; 5,924,406; 5,835, 2033 Traffic congestion data can be exchanged between 881; 5,774,073; 6,442,473; 4,704,610; 5,712,632; 5,973, vehicles. On-coming traffic exchanges information on traffic 616; and 6,008,741. status ahead so that vehicle navigation systems can dynami 2027 The radio used for the communications subsystem cally provide the best route to a destination. can be radio frequency AM, FM, spread spectrum, micro 2034. An industry standard interoperable tolling platform wave, light (infrared, visible, UV) or laser or maser beam could expand the use of toll systems or processing payments (millimeter wave, infrared, visible), or for short distance at parking lots, drive-through establishments (food, gas), etc. communications, acoustic or other communications may be employed. The system preferably employs an intelligent 2035 Safety applications could benefit from use of transportation system (ITS) or Industrial, Scientific and DSRC. The DSRC automaker consortium (DaimlerChrysler, Medical (ISM) allocated band, such as the 915 MHz, 2.4 GM, Ford, Toyota, Nissan, & VW) are seeking ways to MHz or 5.8 GHz band. (The 2.350-2.450 GHz band corre enhance passenger safety with DSRC communications. For sponds to the emission of microwave ovens, and thus the example, in a typical collision, a car has only 10 millisec band suffers from potentially significant interference). The onds to tighten seatbelts, deploy airbags, etc. If an additional 24.125 GHz band, corresponding to K-band police radar, advance warning of 5 milliseconds was provided, one could may also be available; however, transmit power in this band tighten seatbelts, warm-up the airbags, etc. to prepare the car is restricted, e.g., less than about 9 mW. The signal may be for collision. Using radar, GPS data, etc. a car can determine transmitted through free space or in paths including fiber that a collision is imminent, and it can then notify the car optics, waveguides, cables or the like. The communication about to be hit to prepare for collision. may be short or medium range omnidirectional, line of sight, reflected (optical, radio frequency, retroreflector designs), 2036) See: satellite, secure or non-secure, or other modes of commu 2037 ASTM E2213-02 Standard Specification for nications between two points, that the application or state Telecommunications and Information Exchange Between of-the-art may allow. The particular communications meth Roadside and Vehicle Systems—5 GHZ Band Dedicated odology is not critical to the invention, although a preferred Short Range Communications (DSRC) Medium Access embodiment employs a spread spectrum microwave trans Control (MAC) and Physical Layer (PHY) Specifications mission. (This standard, ASTM E2213-02—Standard Specification 2028. A particularly preferred communications scheme for Telecommunications and Information Exchange employs steerable high gain antennas, for example a phased Between Roadside and Vehicle Systems 5 GHz Band array antenna, which allows a higher spatial reuse of com Dedicated Short Range Communications (DSRC) Medium munications bands and higher signal to noise ratio that an Access Control (MAC) and Physical Layer (PHY) Specifi omnidirectional antenna. cations, describes a medium access control layer (MAC) and physical layer (PHY) specification for wireless connectivity 2029. A number of Dedicated Short Range Communica using dedicated short-range communications (DSRC) ser tions (DSRC) systems have been proposed or implemented vices. This standard is based on and refers to the Institute of in order to provide communications between vehicles and Electrical and Electronics Engineers (IEEE) standard 802.11 roadside systems. These DSRC systems traditionally operate (Wireless LAN Medium Access Control and Physical Layer in the 900 MHz band for toll collection, while the FCC has specifications), and standard 802.11a (Wireless LAN recently made available 75 MHz in the 5.850-5.925 GHz Medium Access Control and Physical Layer specifications range for Such purposes, on a co-primary basis with micro High-Speed Physical Layer in the 5 GHz band). This stan wave communications, satellite uplinks, government radar, dard is an extension of IEEE 802.11 technology into the and other uses. However, spectrum is also available in the high-speed vehicle environment. It contains the information so-called U-NII band, which encompasses 5.15-5.25 GHZ necessary to explain the difference between IEEE 802.11 (indoors, 50 mW) and 5.25-5.35 (outdoors, 250 mW). A and IEEE 802.11a operating parameters required to imple Japanese ITS (“ETC) proposal provides a 5.8 GHz full ment a mostly high-speed data transfer service in the 5.9- duplex interrogation system with a half duplex transponder, GHz Intelligent Transportation Systems Radio Service (ITS operating at about 1 megabit per second transmission rates. RS) band or the Unlicensed National Information 2030. In August 2001, the DSRC standards committee Infrastructure (UNII) band, as appropriate). (ASTM 17.51) selected 802.11a as the underlying radio technology for DSRC applications within the 5.850 to 5.925 2038 ANSI X3.38-1988 (R1994) Codes Identifica GHz band. The IEEE 802.11a standard was modified, in a tion of States, the District of Columbia, and the Outlying and new standard referred to as 802.11a R/A (roadside applica Associated Areas of the United States for Information Inter tions) to meet DSRC deployment requirements, and includes change OFDM modulation with a lower data rate, 27 MBS for 2039. ASTM PS 111-98 Specification for Dedicated DSRC instead of 54 MBS for 802.11a. Short Range Communication (DSRC) Physical Layer Using 2031) Proposed DSRC applications include: Microwave in the 902 to 928 MHZ Band 2032 Emergency Vehicle Warning Currently, emer 2040 ASTM PS 105-99 Specification for Dedicated gency vehicles only have sirens and lights to notify of their Short Range Communication (DSRC) Data Link Layer: approach. With DSRC, the emergency vehicle can have the Medium Access and Logical Link Control traffic system change traffic lights to clear traffic along its intended route. Also, this route information can be broadcast 2041 CEN Draft Document: prENV278/9/#65 Dedi to other cars to provide user/vehicle specific directions to cated Short Range Communication (DSRC)—Application reduce collisions. Layer (Layer 7)