Mobile Handsets Come E911 and the Promise of Loca- up Signal Acquisition, and Especially Tion-Based Services (LBS) Are Reduce the Time to First Fix

M obileRTK Using Low-Cost GPS a n dI n t e r n e t- E nabledWireless Phones

Kimmo Alanen, LauriW irola, JaniKä ppi,a nd JariSy rjärinne NokiaTec hnologyPla tforms ©iStockPhoto.com/WilliamFawcett

Anever-increasing number overnment regulation such as receiver to improve its sensitivity, speed of mobile handsets come E911 and the promise of loca- up signal acquisition, and especially tion-based services (LBS) are reduce the time to first fix. However, equipped with GPS and the biggest drivers for integrat- these approved standards do not contain Ging positioning capability into mobile sufficient information for the receiver to some with inertial sensors. However, these single- phones. The increasing sophistication do carrier phase positioning. frequency units do not of applications and refinement of map Until now, no compelling reason databases are continually tightening the existed for adding carrier phase posi- exploit the higher accuracy accuracy requirements for GNSS positioning related features into cellular possible with real-time tioning. In particular, location-based standards so that they could employ kinematic(RTK) techniques. games and features such as “friend find- real-time kinematic (RTK) techniques. Now a group of Nokia er” sometimes require better accuracy Generally, RTK-enabled devices on the than what is achievable with state-of- market are expensive and intended pri- researchersare developing a the-art network-assisted GPS (A-GPS) marily for geodetic and survey appli- software-onlyRTK solution platforms. cations. Also, there has been no real using the hardware and Cellular standards for GPS assistance need in the cellular world for the accu- wirelessconnections already data exist for both control plane and user racy RTK provides. With evolving LBS plane protocols. These protocols carry applications, however, this situation is existing in mobile phones. information that help the integrated GPS changing.

32 InsideGNSS may/june 2006 www.insidegnss.com This article describes a solution mance dual-frequency receivers. called mobile RTK (mRTK), a system We are designing the mRTK solu- specifically designed and implemented tion to work with low-cost, off-the-shelf for the cellular terminal use. Its design GPS receivers with certain requirements incorporates low-cost single-frequency (for example, the ability to report carrier A-GPS receivers, Bluetooth (BT) com- phase measurements and data polarity). munications, and inertial sensors. Basi- Therefore, performance degradations cally, the technique involves exchanging are expected in terms of time to ambigu- measurements in real-time between two ity resolution, accuracy, and achievable units — one designated as the reference baseline length. and the other as the user terminal — and Double-Difference Solutions. The producing the best possible estimate of mRTK solution is based on double-dif- the baseline between the terminals using ference measurements to resolve double- RTK techniques. We are developing the difference integer ambiguities, similar to solution so that in the future it will be traditional RTK methods that are cal- Assisted Bluetooth GPS (BAG) possible to add any other Global Navi- culated either from carrier phase mea- demonstration platform developed by gation Satellite System (GNSS) measure- surements alone or from both carrier Nokia for R&D purposes only. ments in addition to GPS measurements phase and code phase measurements. — or even instead of GPS measure- The formulation of the single-frequency The mRTK solution is also designed ments. double-difference ambiguity resolution to use inertial sensor measurements Using a simulator, we shall provide problem is well documented in the lit- to detect receiver movement. If both data that show it is possible to enable erature and hence, is not summarized receivers are completely stationary high-precision, carrier phase-based posi- here (See “Additional Resources” section for the entire initialization period, the tioning in handsets with minimal addi- at the end of this article for references of ambiguity resolving algorithm can tional hardware costs. Further, we shall appropriate articles on traditional RTK assume that the positions of the receiv- describe some of the protocol aspects techniques.) ers have not changed between the first and especially the aspects of adding The integer ambiguity resolution in and the last epoch and therefore it can support for mRTK messaging to already mRTK is based on the Least-Squares reduce the number of unknowns from existing cellular standards — GSM and Ambiguity Decorrelation Adjustment, the equations. This leads to a situation UMTS. We believe that the mRTK solu- or LAMBDA method, developed by where the ambiguities can be resolved tion will bring high performance to the Prof. J.G.P Teunisson and colleagues with fewer measurements and therefore mass market. at Delft Technical University, in The also faster. Moreover, additional GPS signals, Netherlands. The LAMBDA method is Two-Step Process. Our design target such as L2C and L5, and other GNS- well established both theoretically and is to provide the baseline solution as Ses such as Galileo will become opera- experimentally, which makes it suitable quickly as possible even though some tional in the near future. Consequently, for the current study. Moreover, a refer- accuracy penalties will occur as a result. it would be very beneficial to begin ence implementation is easily available The developed mRTK solution works in incorporating mRTK into the pertinent from the developers. (Delft University of two phases: the initialization phase for wireless standards now so that the infra- Technology, Netherlands, http://www. solving the ambiguities in real time and structure and the service providers will lr.tudelft.nl.) the maintenance phase for baseline esti- be ready when business opportunities The validation of the integer ambi- mation using the ambiguities resolved in present themselves guities is performed by calculating the the initialization phase. discrimination ratio, which can readily The speed of the mRTK solution mRTK Solution Overview be calculated based on the results pro- originates from a design that contains A plethora of RTK surveying solutions duced by the LAMBDA algorithm. The several different levels of the initializa- is available on the market today. Gen- discrimination ratio is a statistical quan- tion phase. The very first baseline esti- erally, they are characterized by the use tity that describes the relative power of mate is determined simply by calculating of both GPS frequencies, L1 and L2, the best and the second-best ambiguity the position difference of the two receiv- enabling ambiguity resolution in sec- candidate vectors. If the discrimination ers. The uncertainty of the estimated onds over baselines of up to 20 kilome- ratio exceeds a certain threshold, K, the baseline cannot be any better than the ters, or even 100 kilometers with more best integer ambiguity candidate vector uncertainty of either receiver’s position time and under good conditions. We is validated and the fixed baseline solu- solution, but the baseline estimate is must emphasize that this article does tion may be calculated using the ambi- made available to the user instantly. not claim to demonstrate similar per- guities. The threshold K is commonly set After that, the mRTK solution calcu- formance and reliability as high-perfor- to 2.0 or above. lates a baseline estimate using measure- www.insidegnss.com may/june 2006 InsideGNSS 33

The photograph on page 33 shows this hardware platform (BAG). The BAG and the mRTK application inside the mobile terminal communicate with each other by means of a proprietary low- level GNSS control interface protocol, which is shown in Figure 2 with red arrows. This protocol contains the previously mentioned aid- ing information for the GPS receiver and the GPS measurements for the mRTK application. These measurements contain carrier phase measurements, code phase measurements, encoded FIGURE 1 Diagram of the demonstration GPS data bits, and data bit platform. Two A-GPS-enabled handsets polarity information. The a r e p o s i t i o n e d w i t h r e s p e c t t o e a c h o t h e r . control protocol, of course, A - G P S i s c o n n e c t e d t o t h e c e l l u l a r t e r m i n a l v i a B l u e t o o t h . T h e t e r m i n a l c o n n e c t s t o t h e also contains the means to assistance server over any given wire- control the receiver. less standard. The server also relays mRTK The mRTK applications measurements FIGURE 2 Block diagram of the mRTK testing system communicate with each other via a server in the ments from a single epoch, incorporat- Testing the System transmission control protocol/Inter- ing both carrier phase and code phase The mRTK performance testing was net protocol (TCP/IP) network (Inter- measurements. However, the uncertain- accomplished using two identical hard- net). The mobile terminal has a general ty of this baseline estimate is still in the ware platforms containing 12-channel packet radio service (GPRS) connection range of meters due to the noise in the off-the-shelf high-sensitivity OEM GPS that enables TCP/IP communication. code phase measurements. Usually after receiver modules and a 3-axis acceler- The server, shown in Figure 1, is needed 30 seconds, the mRTK solution tries to ometer. We constructed this test system because mobile terminal and network initialize the baseline by using only to determine the physical limitations implementations currently do not allow carrier phase measurements from two and requirements for the protocol and direct connections between two ter- epochs. If the carrier phase data produc- messaging aspects. minals. It also provides, for instance, es as inadequate number of measure- The hardware platforms have inte- ephemeris and almanac assistance to ments, the mRTK algorithm will also grated Bluetooth (BT) transceivers and the position calculation software run- include the code phase measurements were connected to the mobile terminals ning in the terminal. and thus achieve a better estimate than via BT connection. The actual position Communication between the mRTK the previous baseline. calculation and the mRTK calculation application and the server is accom- The final level of initialization is, of (mRTK application) are performed plished with a protocol that was specifi- course, when the baseline is calculated inside the mobile terminal. The mRTK cally designed for mRTK use. This com- by using only carrier phase measure- application is designed to be run in any munication is shown with blue arrows in ments. Once the mRTK solution is able Symbian Series 60 terminal. Figure 1 Figure 2. The server also acts as a source to solve the ambiguities using only the shows a diagram of the demonstration for the GPS assistance information pro- carrier phase measurements and validate platform. viding ephemeris, almanac, reference the ambiguities using the discrimination The mRTK application aids the GPS time, and ionospheric corrections. ratio, the ambiguities are said to be ini- receiver with expected signal and time tialized. The mRTK solution then moves information, and reference frequency Performance to the maintaining phase and starts to information. Therefore, the hardware We conducted several experiments using update the baseline using the ambigui- platform is considered to be a BT A-GPS the testing system and a GPS simulator. ties resolved in the initialization phase. receiver and hence the name “BAG”. The simulator was configured to output

34 InsideGNSS may/june 2006 www.insidegnss.com 400 3 Sensors ON data from the same eight satellites for shows clearly the benefit of 350 Sensors OFF both receivers with using several dif- the stationary information ferent baseline lengths varying from 0 when the baseline length is 300 meters to approximately 5 kilometers , longer than one kilometer. 250 and using scenarios for different GPS The accuracy of the 200 weeks. mRTK solution was evaluat- We chose to characterize the system ed using the same measure- 150 alidated Initialization (s) performance without modeling iono- ment set and by calculating 100 sphere, ephemeris, or satellite clock dis- the three-dimensional error turbances. The goal of the simulator tests vector norm (“raw error”) Time to V 50 was to provide information on the best from the mRTK solution 0 0 1000 2000 3000 4000 5000 obtainable performance (i.e., under ideal compared against the true Baseline Length (m) conditions). The future field tests will positions configured into the reveal the real-world performance. simulator. Figure 4 shows the FIGURE 3 Timerequired for validating the mRTK initialization as From the perspective of algorithm amount of errors as a func- a function of baseline length development, determining the effect tion of time spent in process- of using sensors and the stationary ing several different baseline lengths. information provides a clear benefit and information bit was one of the goals. More detailed results from this therefore its inclusion in the protocol is We tested this effect by making several mRTK experiment, especially from the justified. Second, due to the use of single measurements with different baseline algorithm point-of-view, can be found frequency receivers the performance will lengths and calculating two mRTK solu- in the paper by Wirola et al listed in deteriorate quite rapidly with increas- tions from the same measurements; one the Additional Resources. However, ing baseline length. Therefore, we need solution in which the sensor information some conclusions can already be drawn to exploit all means to keep the baseline is available and one where it is not. Figure from these results. First, the stationary length as short as possible. New horizons

NovAtel’s precise thinking makes it possible.

Dedicated to advancing innovative and emerging technologies, the Canadian Space Agency is investing in NovAtel’s precise thinking to turn an early concept into a concrete solution. The result will help the CSA to stay ahead of the ﬁeld as a new age in precise positioning emerges.

Through its Space Technology Development Program, the CSA aims to be one of the ﬁrst to study the Galileo signals in space. NovAtel’s precise thinking has provided the means, with a dual-mode Galileo L1/E5a and GPS L1/L5 receiver, built with the ﬂexibility to adapt as system parameters, such as signal structure, change. The receiver not only enables the CSA to monitor the latest Galileo developments, but also helps ensure NovAtel will be ready to meet the precise positioning demands of the future.

To learn more about the Galileo receiver project and how NovAtel’s precise thinking can help you explore new horizons, visit our website.

Galileo satellite image courtesy ESA - J. Huart Precise thinking

1-800-NOVATEL (U.S. & Canada) or 403-295-4512 | Europe +44 1993 852-436 | E-mail [email protected] www.novatel.com

www.insidegnss.com 1817 10th Avenue SW Calgary, AB T3C 0K2 Telephone may/ju (403) 266-4094ne 2006 Fax (403) 269-1140 InsideGNSS 35 Docket: 26174 Date: Jan.19.06 Client: NovAtel Description: New Horizons Ad Size: 7”x4.875” Publication: Inside GNSS Colour: cmyk Account Ex: A.Ennis Designer: G.Crossley Production: B.Pfleger Mobile rtk

101 1 km baseline 2 km baseline of the properties of TCP/IP. environment. In other words, we deter- 3 km baseline 4 km baseline As TCP/IP already guaran- mined that all message fields larger than 100 5 km baseline tees that transmitted data eight bits start from the even offset. are error-free and also pre- This way the field in the message can ector Norm (m) 10-1 serves the order of the data, be directly read with any processor. For our protocol did not need to instance, an ARM processor is unable to include extensive error cor- access 32-bit fields from an address that -2 10 rections and packet order is not dividable by 4. The field-aligned counts. property is something that cannot be 3D Baseline Error V 10-3 In developing our pro- accomplished in cellular standards, but tocol, we considered the that was not seen as a major issue. 050 100 150 200 250 300 Time (s) traditional RTK protocol of The protocol needed to enable the the Radio Technical Com- use of multiple simultaneous mRTK FIGURE 4 Performance of the mRTK solution. mission for Maritime Ser- sessions. It also needed to contain sim- vices (RTCM); however, it ple authentication because the server appeared too inefficient for is located in an open Internet environ- Sizebits Unit Description Explanation (Type) a mobile terminal environment. Time and Position information (once per message) ment using TCP/IP. Even After authentication, the user termi- though an RTCM proto- nal requests a binding ID from the serv- 32 s UTC time UTC time in seconds. col specification exists for er. The binding ID given by the server is 32 (Q8) m Position X Receiver position in the ECEF transmitting RTCM data then sent to the other mRTK terminal system over TCP/IP (the Networked via short message service (SMS). Both 32 (Q8) m Position Y Transport of RTCM via terminals then bind to the server with 32 (Q8) m Position Z Internet Protocol), it did not the same ID. The binding ID itself is 1 - Stationary Stationary indicator. contain all the parameters just one way of linking the two termi- 31 (Q8) m Position Position uncertainty (CEP50) that we believe are needed to nals together. Of course, a lot of other uncertainty be transmitted between the methods can do the same job. Measurement information (once per signal) terminals. It also appeared After the binding is complete, the complicated to add new reference terminal starts to send mRTK 16 - SS ID Signal ID and SpaceVehicle ID. (Table II) forthcoming GNSS systems measurements back to the user terminal. into RTCM format. Table 1 presents the measurement mes- 8 - Polarity Carrier phase polarity flags: unknown and inverted. We also considered sage used. In the testing, measurements using the Receiver Inde- were sent at the rate of one message per 8 - Cycle Slip Cumulative loss of continuity pendent Exchange Format second. However, the rate of measure- indicator indicator. (RINEX) protocol, but ments does not have to be fixed. It can 32 (Q25) ms Code phase Code phase measurement. RINEX is text-based and, vary either way. The mRTK application 32 (Q32) ms Codephase ST D Code phase standard therefore, requires a lot of on the user side then incorporates the deviation. processing and would create received measurements and its own 32 (Q10) m Carrier Phase Accumulated carrier phase a lot of overhead to the wire- measurements to begin initializing the measurement. less connection. Because the baseline. After initialization, the mRTK 16 (Q16) m Carrier Phase Accumulated carrier phase main goal — and largest updates the baseline at the rate at which STD standard deviation. challenge — of this experi- measurements arrive from the reference 32 (Q10) m/s Doppler Doppler frequency for car- ment was to demonstrate terminal. rier phase extrapolation / that the necessary informa- The mRTK measurement message interpolation. tion can be also included in is comprised of two blocks: a time and Table 1. Measurement Message Content cellular standards, we decid- position information block that is present ed it would be beneficial to only once per message and measurement Testing Protocol develop a new protocol from scratch. information blocks that are included The testing protocol used in the mRTK From the processing point-of-view, once for every measured signal. The time solution was designed specifically for the protocol needed to be as efficient as is given as Universal Time Coordinated use in research and development and as possible, because the mobile terminal (UTC) and is therefore independent of a reference design for proposed changes environment has no extra processing any particular satellite system time. The to the pertinent cellular standards. The resources to waste. Therefore, our design time contains only the integer part of the protocol was designed to be as efficient as is binary and field-aligned in order to seconds; so, the measurements must be possible and especially to take advantage use the data directly in any processor either extrapolated to or actually mea-

36 InsideGNSS may/june 2006 www.insidegnss.com Signal ID Value System sured at an even second. (It doesn’t have the terminal can calculate the baseline Any 0 - to be this way; the resolution may be to the serving BS (Mobile Station Based GPS L1 C/A 1 GPS chosen quite freely. However, this choice (MS-Based) method) and, of course, the was made for initial testing.) The posi- reverse: for the network to calculate the GPS L2C (data) 2 GPS tion is given with 1/256 meter resolution baseline from the BS to the terminal GPS L2C (pilot) 3 GPS in order to reduce quantization errors in (MS-Assisted method). GPS L5 (data) 4 GPS accurate absolute positioning. As baseline length has a huge effect GPS L5 (pilot) 5 GPS The measurement information block on the performance of the mRTK (as Reserved for future use 6-7 - always starts with a signal and space seen, for example, in Figure 4), the length GALILEO L1-B (data) 8 Galileo vehicle (SV) identification (ID) field shouldn’t exceed two to three kilometers. that is basically 12 bits long, even though The cell size, for instance, in the global GALILEO L1-C (pilot) 9 Galileo the testing protocol uses 16 bits due to system for mobile communications GALILEO E5A (data) 10 Galileo alignment. The field is a bit mask of two (GSM) can be as long as 35 kilometers. GALILEO E5A (pilot) 11 Galileo components: the signal ID and SV ID. Therefore, it doesn’t make any sense to GALILEO E5B (data) 12 Galileo The different signal ID values are listed survey all the BS locations, because the GALILEO E5B (pilot) 13 Galileo in Table 2 and include the satellite sys- baseline lengths inside the cells could tem, which is automatically determined exceed the performance limits. Reserved for future use 14-15 - from the signal ID. The SV ID portion is, However, the use of a virtual refer- GLONASS L1 16 GLONASS for instance, in the GPS system case the ence station (VRS) service, for instance, GLONASS L2C 17 GLONASS pseudorandom noise (PRN) number. could solve this issue. The VRS system Reserved for future use 18-19 - The pseudorange measurement in can be used to calculate a virtual RTK QZSS L1 C/A 20 QZSS the measurement information block reference station anywhere within the is given with seven bits for the integer VRS service area, and that “anywhere” Reserved for future use 21-24 - part in order to avoid ambiguities in the can always be close to the user terminal. SBAS L1 C/A 25 SBAS signal’s time of flight. The number of bits In this way, the baseline length never LAAS L1 C/A 26 LAAS reserved for the carrier phase measure- becomes too long, and availability of Reserved for future use 27-31 - ment field ensures that the field would the VRS service eliminates the need to Table 2. Signal ID not roll over more than once between survey all the BS locations. the first and the last epoch in the mRTK Current versions of the control plane initialization. The Doppler field would, protocols lack the capability of transmit- broadcast channels to simultaneously however, also work with as few as 22 ting the required carrier phase measure- serve multiple terminals, the cell size in bits, but, due to alignment, 32 bits were ments. Also, the accurate BS position GSM is so large that the single BS would used. cannot be transmitted to the terminal have to broadcast more than one item with the current standard. However, of reference information in order to get Cellular Protocol Aspects activity in the 3GPP standardization the required performance. As Figure During the testing protocol design and process is now under way to include 4 showed, the performance degrades implementation, several issues emerged assisted-GNSS data formats in GSM rather rapidly if the baseline length gets concerning the addition of the mRTK standards. Therefore, the opportunity long. A GSM cell radius can be as long feature into cellular protocols. This sec- now exists to include new features such as 35 kilometers; therefore, the BS would tion lists the considerations and conclu- as mRTK with rather minimal effort have to broadcast more than 100 refer- sions from those findings. for improved positioning performance ence measurements in order to keep the User-to-user relative positioning compared to the existing A-GPS imple- baseline length less than 3 kilometer. is not recommended for control plane mentation. Therefore, in the GSM case, the broad- systems because it would require a lot Control Plane Broadcasts. GSM stan- cast solution is most likely not feasible. of protocol and implementation work dards make it possible to deliver GPS For other cellular standards, for to get the binding of two terminals and assistance from the network to the ter- instance the Universal Mobile Telecom- relaying measurements between two terminal via broadcast channels. These munications System (UMTS), the cell minals to actually work. same channels could be used to serve sizes are relatively small. The radius of Control Plane. The control plane level mRTK reference measurements for one cell is usually less than three kilo- would still benefit from the mRTK fea- accurate absolute positioning of multiple meters, even though the maximum in ture as an accurate absolute position- user terminals. UMTS is 6 kilometers. The bandwidth ing method. If the binding between the The rate of data that can be trans- of the broadcast channels is also much mobile units is not implemented but the mitted via GSM broadcast channels, higher than in GSM. Therefore, the use serving base station (BS) is surveyed and however, is rather restricted. So, even of a broadcast solution in the UMTS case paired as a stationary reference to them, though it would be possible to use shows high potential. www.insidegnss.com may/june 2006 InsideGNSS 37 Mobile rtk

User Plane Aspects. The testing pro- work capabilities. In the future, more forthcoming satellite systems (e.g., Gali- tocol used during this experiment was satellite based navigation systems and leo and modernized GPS), the solution already implemented on the user plane civil GNSS signals will become avail- will significantly improve the accuracy level and therefore serves very well, able. Most likely some terminals will of positioning in the mobile terminal. almost as is, in the secure user plane not contain the ability to measure all Nonetheless, the standardization of (SUPL) protocol. The same features that possible signals. Therefore, regardless the mRTK features will require a lot of were available in the testing protocol can of the carrier, the class marking of the joint effort among terminal and network also be included directly in the SUPL terminal’s measuring capability must manufacturers and cellular operators. protocol, with some modifications. For be solved somehow. This also applies to example, in using VRS services in the class marking for VRS service capability Acknowledgments user plane, the rough position of the user and the signals in that service. This article is based in part on two terminal must be somehow transmitted papers, “Bringing RTK to Cellular Ter- from the terminal to the VRS service Future Work minals Using a Low-Cost Single-Fre- provider at the beginning of the session. This article has introduced a new con- quency AGPS Receiver and Inertial Sen- Still, implementing this is quite trivial. cept called mobile Real-Time Kinemat- sors,” by L. Wirola, K. Alanen, J. Käppi, Other Aspects. When specifying the ics and shows that RTK-like features are and J. Syrjärinne, and “Inertial Sensor mRTK protocol for any carrier specifica- possible using low-cost components and Enhanced Mobile RTK Solution Using tion, several things must be considered. existing cellular communication carri- Low-Cost Assisted GPS Receivers and Firstly, the bandwidth requirement was ers. Even though a lot of development Internet-Enabled Cellular Phones,” by K. calculated to be roughly 2.3 kbs for 12 work remains on the mRTK algorithm Alanen, L. Wirola, J. Käppi, J. Syrjärinne, signals. In the future, however, the num- side, the biggest challenge still involves presented at the IEEE/ION PLANS 2006 ber of available signals will most likely cellular carriers and their standardiza- conference, © 2006 IEEE. triple due to the forthcoming GNSS tion. Of course, even after standardiza- satellite systems and modernization of tion, the development of the infrastruc- Manufacturers GPS. However, the bandwidth calcula- ture would require a huge effort. The mRTK prototype platform uses the tion assumes a message rate of 1 Hz and, Future work with the existing testing iTrax03/16 GPS OEM receiver manufac- as was already mentioned, the rate can protocol includes more testing, especial- tured by Fastrax Ltd., Vantaa, Finland. be less. When compared to the RTCM ly field testing, and testing with different The accelerometer is an LIS3L02DQ protocol, which requires (with 12 sig- signal conditions and satellite constella- from STMicroelectronics, Geneva, nals) 1.6 kbs, the bandwidth require- tions. The testing protocol itself should Switzerland. A GSS7700 GPS/SBAS sim- ment isn’t significantly bigger. be modified with new features such as ulator from Spirent Communications, The second aspect that we should the VRS service. Using VRS, the base- Paignton, Devon, United Kingdom. consider is the real-time requirement line can always be kept very short, and Additional Resources of the mRTK. Actually, there aren’t any accurate absolute positioning is available strict real-time requirements. The user everywhere using mRTK. Protocols & Standards [1] 3GPP TS 04.31 and 44.031, Location Services application just has to buffer its own One of the ideas that also need to be (LCS); Mobile Station (MS) - Serving Mobile Loca- measurements for the delay that is caused further developed is peer-to-peer proto- tion Centre (SMLC) Radio Resource LCS Protocol by the carrier of the reference measure- cols. In those protocols the mRTK mea- (RRLP), http://www.3gpp.org/ ments and that delay can be several surements would be transmitted directly [2] 3GPP TS 04.35 and 44.035, Location Services seconds. Even tens of seconds shouldn’t from one terminal to another without (LCS); Broadcast network assistance for Enhanced cause any major problems technically. the use of a server in between. As an Observed Time Difference (E-OTD) and Global The only considerable effect is on the example, this kind of protocol could be Positioning System (GPS) positioning methods, user who directly experiences the delay. embedded into voice-over-IP (VoIP), http://www. 3gpp.org/ Thirdly, the testing protocol assumed in which the data channel for the voice [3] 3GPPTS 25.331, Radio Resource Control (RRC) that the carrier (in the testing case, encoding is already open and could eas- protocol specification, http://www.3gpp.org/ TCP/IP) guaranteed that the absence of ily accommodate other data transmis- [4] 3GPP2 TSG-C C.S0022-0, Location Ser- transmission errors and stability of the sions that do not have strict real-time vices (Position Determination Service), http:// message order. Of course, these assump- requirements, such as mRTK. Other www.3gpp2.org/Public_html/specs/tsgc.cfm tions do not apply in all possible wireless peer-to-peer protocol means would [5] 3GPP2 TSG-X X.P0024-0 V0.9, IP-Based Loca- carriers. Therefore, when specifying this exist, for instance, in WLAN, where the tion Services protocol in such a carrier that does not terminals are connected to the same [6] OMA-TS-ULP-V1_0-20050719-C, User Plane guarantee these assumptions, they must subnet and would be able to open direct Location Protocol, http://www.openmobileal- be addressed. connections to each other. liance.org/release_program/supl_v1_0.html Finally, the biggest issue is the class The solution we have presented holds [7] NTRIP, NetworkedTransport of RTCM via Inter- marking of user equipment and net- a lot of potential. Especially with the net Protocol, http://www.rtcm.org/

38 InsideGNSS may/june 2006 www.insidegnss.com [8] RINEX, the Receiver Independent Exchange [15] Teunissen, P.J.G., and P.J. de Jonge and C.C.J.M in research on GNSS assistance protocol enhance- FormatVersion 2.10,Werner Gurtner, Astronomi- Tiberius,“The least-squares ambiguity decorre- ments. cal Institute, University of Berne, http://www.ngs. lation adjustment: its performance on short GPS Lauri Wirola received his M.Sc. degree from noaa. gov/CORS/Rinex2.html baselines and short observation spans”, Journal the Tampere University of Technology, with a [9] RTCM-104,The RadioTechnical Commission of Geodesy, number 71, pages 589-602. major in electrophysics. Shortly after completing for Maritime Services, http://www.rtcm.org/ [16] Teunissen, P., and P. de Jonge, C. Tiberius, his thesis, he started working in the area of positioning at Nokia Technology Platforms in 2005. Publications “On the Spectrum of GPS DD-Ambiguities”, Pro- His present research interests include RTK and [10] de Jonge, P., and C. Tiberius, The LAMBDA ceedings of ION GPS 1994, Salt Lake City, Utah, A-GNSS standardization. method for integer ambiguity resolution: imple- September 20-23 1994, pages 115-124. Jani Käppi received his M.Sc. degree from the mentation aspects, Publications of the Delft Geo- [17] Tiberius, C.C.J.M., and P.J. de Jonge, “Fast Tampere University ofTechnology. He worked at deticC omputingC enternumber 12, D elft:U niver- Positioning Using the LAMBDA-method”, Pro- the Institute of Computer and Digital Systems siteitsdrukkerij TU Delft, 1996. ceedings of the 4th International Symposium on at the Tampere University of Technology doing [11] Kim, D., and R.B. Langley, “GPS Ambiguity Differential Satellite Navigation SystemsD SNS’95, research in the area of personal positioning before Resolutionand Validation: M ethodologies,Trends Bergen, Norway, April 24-28 1995. joining Nokia Corporation in 2002. He is currently and Issues”, 7th GNSS Workshop – International [18] Wirola, L., and K. Alanen, J. Käppi, and J. researchingthe development of sensor-enhanced Symposium on GPS/GNSS, Seoul, Korea, Novem- Syrjärinne, Bringing RTK to Cellular Terminals positioning systems. ber 30 – December 2, 2000. Usinga L ow-CostSingle-Frequency A GPSReceiver Jari Syrjärinne received his M.Sc. degree and [12] Leick, A., GPS Satellite Surveying, 3rd edi- and Inertial Sensors, IEEE/ION PLANS 2006 Con- hisdoctoral degree, both from Tampere U niversity tion, Hoboken, NJ: John Wiley & Sons, 2004. ference, 24-27 April 2006, San Diego, CA ofTechnology, majoring in digital signal process- [13] Richert, T., and N. El-Sheimy, “Ionospheric ing and applied mathematics. Between 1996 and Modeling: The Key to GNSS Ambiguity Resolu- Authors 1998 he worked for Tampere University of Tech- tion”, GPS World, June 2005, pages 35-40. Kimmo Alanen received his M.Sc. degree from the nology Signal Processing Laboratory in the areas [14] Teunissen, P.J.G., “A new method for Fast Tampere University ofTechnology, Finland, with of data fusion and target tracking, and since 1999 CarrierPhase A mbiguityE stimation”,Proceedings a major in software engineering. He joined the he has been with the Nokia Corporation. He is cur- IEEEPosition, L ocationand Navigation Symposium Nokia Corporation in 1997 and has been working rently involved in work onA -GNSS and algorithms PLANS’94, LasVegas, NV, April 11-15 1994, pages with positioning research for the last eight years. for hybrid positioning. 562-573. He is currently undertaking postgraduate studies

USE IT OR LOSE IT! You are at risk of missing out on future issues of Inside GNSS unless you register for a FREE subscription at www.insidegnss.com