Relay-Based Deployment Concepts for Wireless and Mobile Broadband Radio

WIRELESS WORLD RESEARCH FORUM

Ralf Pabst, Bernhard H. Walke, and Daniel C. Schultz, RWTH Aachen University Patrick Herhold, Technical University of Dresden; Halim Yanikomeroglu, Carleton University Sayandev Mukherjee and Harish Viswanathan, Lucent Technologies Matthias Lott and Wolfgang Zirwas, SIEMENS ICM; Mischa Dohler and Hamid Aghvami, Kings College David D. Falconer, Carleton University; Gerhard P. Fettweis, Technical University of Dresden

ABSTRACT located well above the 2 GHz band used by the 3G systems. The radio propagation in these In recent years, there has been an upsurge of bands is significantly more vulnerable to non- interest in multihop-augmented infrastructure- line-of-sight conditions, which is the typical based networks in both industry and academia, mode of operation in today’s urban cellular such as the seed concept in 3GPP, mesh net- communications. works in IEEE 802.16, and coverage extension The brute force solution to these two prob- of HiperLAN/2 through relays or user-coopera- lems is to significantly increase the density of tive diversity mesh networks. This article, a syn- base stations, resulting in considerably higher opsis of numerous contributions to Working deployment costs that would only be feasible if Group 4 of the Wireless World Research Forum the number of subscribers also increased at the and other research work, presents an overview same rate. This seems unlikely, with the penetra- of important topics and applications in the con- tion of cellular phones already high in developed text of relaying. It covers different approaches countries. On the other hand, the same number to exploiting the benefits of multihop communi- of subscribers will have a much higher demand cations via relays, such as solutions for radio in transmission rates, making the aggregate range extension in mobile and wireless broad- throughput rate the bottleneck in future wireless band cellular networks (trading range for capac- systems. Under the working assumption that ity), and solutions to combat shadowing at high subscribers will not be willing to pay the same radio frequencies. Furthermore, relaying is pre- amount per data bit as for voice bits, a drastic sented as a means to reduce infrastructure increase in the number of base stations does not deployment costs. It is also shown that through seem economically justifiable. the exploitation of spatial diversity, multihop It is obvious from the above discussion that relaying can enhance capacity in cellular net- more fundamental enhancements are necessary works. We wish to emphasize that while this for the very ambitious throughput and coverage article focuses on fixed relays, many of the con- requirements of future systems. Toward this end, cepts presented can also be applied to systems in addition to advanced transmission techniques with moving relays. and collocated antenna technologies, some major modifications in the wireless network INTRODUCTION architecture itself that will enable effective distri- bution and collection of signals to and from The very high data rates envisioned for fourth- wireless users are required. The integration of generation (4G) wireless systems in reasonably multihop capability into conventional wireless large areas do not appear to be feasible with networks is perhaps the most promising architec- the conventional cellular architecture due to tural upgrade. In the following, the terms multi- two basic reasons. First, the transmission rates hop and relaying will be used to refer to the same envisioned for 4G systems are two orders of concept; see also [1]. magnitude higher than those of 3G systems. Multihop wireless networking traditionally This demand creates serious power concerns has been studied in the context of ad hoc and since it is well known that for a given transmit peer-to-peer networks. The application of multi- power level, the symbol (and thus bit) energy hop networking in wide-area cellular systems has decreases linearly with the increasing transmis- the following benefits. sion rate. Second, the spectrum that will be While conventional cellular networks are released for 4G systems will almost certainly be assumed to have cells of diameter 2–5 km, a

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. relay will only be expected to cover a region of new radio access network architecture based on diameter 200–500 m. This means that the trans- fixed relays, and the conclusion gives a summary TDMA-based systems mit power requirements for such a relay are of the key research topics identified in the vari- significantly reduced compared to those for a ous contributions to this article. are especially well- base station (BS). This in turn permits eco- suited to introduce nomical design of the amplifier used in the HE TATE OF THE RT relaying, as this relay. Furthermore, the mast on which the T S A relay is placed does not need to be as high as No smart relaying concept has been adopted in scheme allows for for a BS, reducing operating expenses such as existing cellular systems so far. Solely bidirec- an easy allocation of tower leasing and maintenance costs for the tional amplifiers have been used in 2G systems service provider. and will be introduced for 3G systems. However, resources to the Relays do not have a wired connection to the these analog repeaters increase the noise level mobile-to-relay and backhaul. Instead, they store the data received and suffer from the danger of instability due to relay-to-BS links. wirelessly from the BS and forward to the user their fixed gain, which has limited their applica- terminals, and vice versa. Thus, the costs of the tion to specific scenarios. The first system backplane that serves as the interface between Most existing and standardized systems were based on Time the BS and the wired backhaul network can be designed for bidirectional communication eliminated for a relay. between a central BS and mobile stations direct- Division Multiplex If the density of relays in a cell is moderately ly linked to them. The additional communication (TDM) and Relays high, most terminals are significantly closer to traffic between a mobile terminal and a relay connecting mobiles one or more relays than to the BS. This means intermediately inserted into a link between that the propagation loss from the BS to such a mobile terminal and BS requires additional to the fixed network terminal is larger than from a nearby relay to the radio resources to be allocated — one of the fac- was proposed terminal. This results in higher data rates on the tors that hitherto have hindered the deployment links between the relays and the terminals, there- of smart relay concepts. in 1985. by potentially solving the coverage problem for Time-division multiple access (TDMA)-based high data rates in larger cells. systems are especially well suited to introduce Since it is possible to have simultaneous relaying, as this scheme allows for easy alloca- transmission by both the BS and relays, capacity tion of resources to the mobile-to-relay and gains may also be achieved by either exploiting relay-to-BS links. The first system based on reuse efficiency or spatial diversity. time-division multiplex (TDM) and relays con- The relay-to-user links could use a different necting mobiles to the fixed network was pro- (unlicensed) spectrum (e.g., IEEE 802.11x) than posed in 1985 [3]. Another method proposed for the BS-to-user links (the licensed spectrum), F/TDMA (F: frequency) systems is to reuse a yielding significant gains from load balancing frequency channel from neighboring cells [4]. through relays [2]. The European Telecommunications Standards Compared to ad hoc networks , networks Institute/Digital Enhanced Cordless Telephony applying relaying via fixed infrastructure do not (ETSI/DECT) standard in 1998 was the first need complicated distributed routing algorithms, specifying fixed relays (called wireless BSs) for while retaining the flexibility of being able to cordless systems using TDM channels for voice move the relays as the traffic patterns change and data communications. The ETSI/TETRA over time. standard specifies a dual-watch function allowing It is worth emphasizing the basic difference the aggregate traffic of a number of mobile ter- between the fundamental goals of conventional minals connected in direct mode to be relayed ad hoc networks and the described multihop- by TDM channel-switching to a dispatcher panel augmented infrastructure-based networks. While connected to a BS. Relaying in cellular code- the defining goal of ad hoc networks is the abili- division multiple access (CDMA)-based systems ty to function without any infrastructure, the has been investigated by Zadeh et al. [5]. Uplink goal in the latter types is the almost ubiquitous and downlink are separated using frequency- provision of very-high-data-rate coverage and division duplex (FDD), as is done in IS-95 and throughput. Universal Mobile Telecommunications System In this article, BS is used for traditional cellu- (UMTS) terrestrial radio access (UTRA) FDD. lar systems, while the term access point (AP) is All these concepts can easily be extended to used for wireless local area network (WLAN)- packet-based systems [6, 7]. type systems. Both have in common the fact that A completely different approach is consid- they are directly connected to their backbone ered in [8], incorporating an additional ad hoc network, which distinguishes them from relay mode into the Global System for Mobile Com- stations (RSs). However, it should be noted that munications (GSM) protocol stack to enable future air interfaces and deployments could blur relaying. Similarly, [2] employs relaying stations the distinction between BSs and APs and even- to divert traffic from possibly congested areas of tually lead to their convergence. a cellular system to cells with a lower traffic This article is organized as follows. First, an load. These relaying stations utilize a different overview of the state of the art in current air interface for communication among them- research on relaying topics is given. The next selves and with mobile stations, which could, for sections deal with performance benefits gained example, be provided by a WLAN standard. from relaying when combating heavy path loss ETSI-broadband radio access network and exploiting spatial diversity. Implications of (BRAN), high-performance LAN (HiperLAN/1), relaying on routing and radio resource manage- and IEEE 802.11x contain the elements to oper- ment are discussed next. Then we introduce the ate ad hoc networks. ETSI HiperLAN/2 in the wireless media system (WMS), an example of a home extension contains an ad hoc mode of

IEEE Communications Magazine • September 2004 81

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. ETSI-BRAN/HiperLAN/1 Two-hop cell 2. AP: Area deployment of two-hop cells and IEEE 802.11x R1 S1 S2 S3 S4 S5

contain the ele- 1. S S S S S 6 7 8 9 10Fixed ments to operate ad core R AP S11 S12 S13 S14 network hoc networks. 4 R2 AP ETSI/HiperLAN/2 in S15 S16 S17 S18 S19 R3 the Home Extension S20 S21 S22 S23 S24

contains an ad hoc R: Tx/Rx range AP: Access point mode of operation Figure 1. Left: a city scenario with one AP (serving the white area) and four RSs covering the shadowed that allows the areas around the corners shown in grey. Middle: a schematic of the scenario. Right: wide-area coverage nodes to agree on a using the basic element (left) and two groups of frequencies (source: [10]). Central Controller (CC) to take the role operation that allows the nodes to agree on a repeaters, bridges, or routers, the relay nodes of an AP in a cluster central controller (CC) to take the role of an AP regenerate the signal by fully decoding and re- in a cluster of nodes, but no multihop functions encoding the signals prior to retransmission. By of nodes, but no are specified so far in any WLAN standard. contrast, in amplify-and-forward systems the multihop functions Multihop operation based on wireless relays that relays essentially act as analog repeaters, thereby operate alternately on different frequency chan- increasing the systems noise level. Unless other- are specified so far nels to connect neighboring clusters has been wise stated, we consider decode-and-forward sys- in any WLAN proven workable in [9]. In HiperLAN/2 basic tems as the majority of proposed concepts are of standard. mode (using an AP) it has been shown that mul- this class, and they are generally considered to tihop operation via forwarding mobile terminals be more viable with respect to implementation. is easy to perform within the framework of the standard [6, 7]. ULTIHOP PERATION In general, relaying systems can be classified M O as either decode-and-forward or amplify-and-for- This section presents examples of relay imple- ward systems. In decode-and-forward schemes, mentations and aims to point out the perfor- where relays are also referred to as digital mance benefits multihop relaying can provide in

Fixed relay: Access point: Receiving from AP with antenna gain Carrier frequency: f1 Carrier frequency: f1

f1 f1

Internet LOS Fixed relay: Receiving from AP with antenna gain: (a) Carrier frequency: f1 Access point: Transmission to terminals: carrier frequency f2 Carrier frequency: f1 (with 1 or 2 Tx)

f1 f2

Internet

(b) LOS Fixed relay: Access point: Receiving from AP via forwarder Carrier frequency: f1 carrier frequency: f2 f1 f2

f1 f2 Forwarding terminal: Internet no link with mobile terminals

(c)

Figure 2. Examples of relay concepts (source: [10]).

82 IEEE Communications Magazine • September 2004

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. broadband networks when applied in certain scenarios. Figure 1 (left) shows a city scenario with Throughput with fixed relays at different distances one AP (providing radio coverage to the areas 45 marked white) and four fixed-mounted RSs to TP at AP provide radio coverage to areas around the cor- 40 TP at relay 1 at 100 m (20 dB receive antenna gain) ner shadowed from the AP, shown in beige. TP at relay 2 at 200 m (20 dB receive antenna gain) TP at relay 3 at 300 m (20 dB receive antenna gain) While the intersection can be covered well by 35 Maximum acheivable TP the AP, nearby streets can only be served if line- of-sight (LOS) connectivity is available between 30 a mobile terminal and its serving station, due to the difficult radio propagation conditions known, 25 for example, in the 5–6 GHz frequency band. RSs allow extending radio coverage to these a High bit rate due streets. 20 to receive antenna gain at fixed relays A schematic of this scenario (a multihop network with AP) is illustrated in Fig. 1 (middle) Throughput (Mb/s) 15 where the transmit/receive radius R is shown to be the parameter determining the connectivity of 10 b the nodes shown. RS S1 would have to route the traffic of the wireless terminals (not shown) it is 5 serving via intermediate RS S8 using a low-rate but robust combination of modulation and cod- 0 0 50 100 150 200 250 300 350 400 450 500 ing (PHY mode), or via RSs S2 and S8 using a faster PHY mode and thus higher link capacity, Distance (m) and so forth from S8 either via S9 or directly to the AP. The interpretation of the multihop net- AP Relay 1 Relay 2 Relay 3 work in the middle of Fig. 1 is that all the nodes shown are either RSs or APs, and the mobile Figure 3. Analytical estimation of the extension of the radio range of an AP user terminals roaming in the area (not shown) by relays with receive antenna gain (source: [10]). are served by the nodes shown. The basic element of Fig. 1 (left) can be repeated to cover a wide area, as shown in Fig. 1 (right). tion (ISO/OSI) reference model in either layer 1 Figure 2 shows three examples of concepts (physical) “repeater,” 2 (link) “bridge,” or 3 for fixed or mobile relaying, see [10]: (network) “router.” Figures 3 and 4 show ana- a) Relaying in the time domain with the AP lytical and simulation results according to the and RSs operating at the same carrier fre- solution of a layer 2 relay as described in [6, 7] quency, f1, while accessing the physical for the HiperLAN/2 standard. medium in time multiplex. Figure 4 (left) shows for the concept intro- b)Relaying in the frequency domain with the duced in Fig. 2a the simulated end-to-end AP and RSs operating at two different car- throughput between an AP and a mobile termi- rier frequencies, one of them regularly nal located at distance d for different modula- switching between the two frequencies or tion and coding schemes known from the air having two transceivers. interface of HiperLAN/2 or IEEE 802.11a [11]. c) Two-stage relaying in the frequency The terminal is shadowed by a building at the domain, where a fixed mounted RS that is street corner and is therefore connected around in the range of both the AP and a second the corner with the help of an RS. The shaded RS connects the AP and the second RS by area under the curves shows the gain in terms of store-and-forward operation and dynami- throughput possible from the use of the RS cally switches between frequency carriers f1 without which the mobile terminal could not be and f2. Unlike the second RS, which serves connected to the AP. It can be seen that the mobile terminals, the first RS’s only role is range extension resulting from using the RS is to bridge the distance between the AP and substantial. It is worth noting that smart antenna the second RS where direct communication technology at the AP or mobile terminal cannot between the AP and the first RS is not pos- provide radio coverage around an obstacle. A sible due to lack of LOS (see also [9]). relay is currently the only way to achieve this. In a and b the radio link is based on LOS It has been stated that the capacity of an AP radio with transmit and receive gain antennas at when using a modem as standardized for Hiper- the AP and RS. LAN/2 and IEEE 802.11a might be excessive An analytical estimation of the bit rate over compared to the rate requirements of mobile distance from an AP that is supported by RSs to terminals roaming in its picocell area formed by extend the radio range is shown schematically in omnidirectional or sectorized antennas and a Fig. 3. This assumes a TDM approach according maximum transmit power of 1 W equivalent to Fig. 2a. Without receive antenna gain, RS 1 isotropically radiated power (EIRP) [12]. would have available only a bit rate equivalent to The fixed relay concept can also be used to value b (Mb/s), while with receive antenna gain extend the range of an AP to nonshadowed it achieves value a (Mb/s). A similar considera- areas beyond the regulatory EIRP limits as tion applies to RSs 2 and 3. shown from the simulation results for Hiper- In general, the relaying function could be LAN/2 (Fig. 4, right) according to the scenario performed according to the International Stan- in Fig. 2a. It can be seen that the radio range dards Organization open systems interconnec- can be dramatically increased, especially when

IEEE Communications Magazine • September 2004 83

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. 50 50 64QAM3/4 one-hop 64QAM3/4 one-hop 64QAM3/4 two-hop 45 64QAM3/4 two-hop 16QAM3/4 one-hop 45 16QAM3/4 one-hop 16QAM3/4 two-hop 16QAM3/4 two-hop BPSK3/4 one-hop 40 40 BPSK3/4 one-hop BPSK3/4 two-hop BPSK3/4 two-hop AP 35 35

30 30 AP d d/√2 d 25 25 AP Relay Mobile terminal 20 20 d/2 d/2 √ Mobile Relay d/ 2 terminal 15 15 LOS: gain using relay with directed receive antennas 10 10 NLOS: gain using relay around a corner Maximum end-to-end user throughput (Mb/s) 5 Maximum end-to-end user throughput (Mb/s) 5

0 0 0 100 200 300 400 500 600 700 0 200 400 600 800 1000 Distance d (m) Distance d (m)

Figure 4. Left: a fixed wireless router at an intersection to extend the radio range of an AP around the corner into a shadowed area to serve a remote mobile terminal. Right: maximum end-to-end throughput vs. distance for forwarding under LOS conditions with directed receive antennas having a gain of 11 dB (source: [11]).

using receive antenna gain at both ends. The use nature and its ability to achieve diversity through of smart antennas and beamforming to improve independent channels. While its broadcast the propagation characteristics between an AP nature is frequently considered a drawback as it and RSs and to connect multiple relays at the leads to mutual interference in a wireless net- same time has been known for a long time [3]. work, the concepts discussed in this subsection Other benefits of relaying are the possibility exploit the fact that a signal, once transmitted, of radio resource reuse within the area served by can be received (and usefully forwarded) by mul- an AP and its related relays (e.g., frequency tiple terminals. reuse within the area served by one AP), and the In general, cooperative relaying systems have cost advantage. The cost to connect the radio a source node multicasting a message to a num- access network to the infrastructure (core) net- ber of cooperative (i.e., helping) relays, which in work is reduced substantially by reducing the turn resend a processed version to the intended number of points of presence required, for exam- destination node (Fig. 5a). The destination node ple, to 20 percent for the example shown in Fig. combines the signals received from the relays, 1 (left). In a Manhattan grid according to Fig. 1, possibly also taking into account the source’s a second tier of relays would enlarge the number original signal. By performing this combining, of relays to 12, thereby covering 13 intersections. the inherent diversity of the relay channel is use- The infrastructure connectivity cost would then fully exploited. More advanced concepts addi- be reduced to about 8 percent of the cost of a tionally include successive interference purely AP-based system. The number of tiers is cancellation to maximize throughput. In this bounded by the capacity available from the AP, context, such cooperative multihop scenarios can the capacity per area element required by the be regarded as virtual antenna arrays: each relay terminals roaming in the service area, and the becomes part of a larger distributed array. This delay requirements. It seems advisable not to distributed nature naturally leads to a number of exceed two tiers of relays in such networks. new challenges, among them synchronization, availability of channel state information, and appropriate cluster formation. COOPERATIVE RELAYING AND Like conventional relaying systems, coopera- VIRTUAL ANTENNA ARRAYS tive schemes benefit from path loss savings; moreover, the following gains can be expected: CONCEPTS • Power gain due to the fact that each of the Common to all relaying techniques discussed so relays may add additional transmit power far is the store-and-forward operation of the that is combined in the destination terminal nodes in a relay chain. Each receiver along this • Macrodiversity gain that allows combating chain exploits solely the copy of the information shadowing that has been sent by its respective transmitter, • Diversity gain in the presence of fading while it discards emissions from other transmit- Finally, it is worth noting that an integration ters in the chain. of multiple-input multiple-output (MIMO) and The basic idea of cooperative relaying is to go so-called dirty paper coding techniques may lead one step further by exploiting two fundamental to advanced multihop networks with high spec- features of the wireless medium: its broadcast tral efficiency.

84 IEEE Communications Magazine • September 2004

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. Cooperative protocols can be static or adaptive; the latter can be enhanced by allowing Each stage of cooperating relays feedback and/or signaling between forwarding nodes. While in static protocols the relay nodes Relays constantly retransmit a processed version of First stage their received signals, in adaptive versions the relays resend signals only when they believe it to be helpful for the destination node. The Source Destination adaptation may be done by each relay indepen- dently or jointly for all together if information is exchanged between the relays. Theoretically, (a) Single-stage concept (b) Multistage concept the transmission in up- and downlink directions is fully symmetric, which follows from reversing Figure 5. The concept of cooperative relaying. the direction of the arrows in Fig. 5. In prac- tice, however, the boundary conditions for up- /downlink will often be quite different, as the depicted in Fig. 5b. More specifically, the source radio links between AP and RSs will exhibit node transmits to the relays in the first stage, typically good LOS conditions and may be which then may use space-time coding tech- enhanced by directional antennas in contrast to niques in a distributed manner to retransmit the the relay-MS links, which are quite unpre- signal to the next stage of relays. For example, dictable. As a result, the optimum algorithms to each of the nodes in a stage can transmit a col- be used in so-called single-frequency networks umn of an orthogonal space-time code matrix, (SFNs) for up- and downlink may be different where the individual column indices have been as well, and the benefits may not be entirely chosen by the stage’s nodes via a local exchange symmetric. of information. This cooperation could possibly be done over an alternative air interface such as EXAMPLES AND RESEARCH AREAS Bluetooth. Eventually, the destination terminal Of the various approaches to cooperative relay- combines the signals transmitted by the last ing that have been investigated, we focus here stage. It can be shown that up to 5 dB power on the concepts of SFNs and virtual antenna savings are achieved with four distributed relays arrays. per stage compared to a conventional relay pro- First, results for amplify-and-forward net- tocol. works indicate that such systems indeed provide Among others, important research areas are: a degree of spatial diversity that is proportional • The design of flexible protocols and for- to the number of distributed antenna elements, warding strategies that allow integrating or, in our context, the number of involved relay cooperative approaches in conventional nodes [13]. Even under strict power and band- relaying systems width constraints, systems with one or two relays • The question of availability of CSI at the can achieve diversity gains over single-input sin- transmitters and relays gle-output (SISO) and conventional relaying • The trade-off between macrodiversity gains transmission for communication over a Rayleigh and usage of additional radio resources fading channel [13]. required by cooperative relaying A mature concept of an SFN as well as MIMO concepts are presented in [14]. Fixed relays provide power gains and macrodiversity by ROUTING AND RADIO forwarding signals from an AP to the destination RESOURCE MANAGEMENT terminal. Various protocol versions can be implemented based on the underlying orthogo- ROUTING nal frequency-division multiplexing (OFDM) air We consider routing in a multihop network sup- interface. The potential of SFNs is illustrated in ported by an infrastructure and communication [14], where achievable performance gains of up relations limited to a few hops only. Multiple to 9 dB for the ideal simultaneous operation of simultaneous routes become possible, making eight fixed relays are achieved. Specifically, three the choice of routing algorithm important. An cases are compared: algorithm ensuring that no queue at a relay node • Random superposition of the individual explodes for the largest possible set of packet subcarriers (classical SFN) arrival rates is called throughput-optimal [16]. • Subcarrier selection based on strongest Routing becomes more challenging when received power considering mobile relays. In the Mobile Ad Hoc • Ideal phase predistortion leading to con- Networks (MANET) subgroup of the Internet structive superposition at the terminal Engineering Task Force (IETF), several routing Note that predistortion techniques require algorithms for MANETs have been investigated. channel state information (CSI) at the relays, Studies of these algorithms have shown high which leads to comparatively more complex sys- routing overhead and low efficiency in network tem implementations. throughput. Based on this observation, it is pro- While the former concept is a two-hop con- posed that routing in the multihop network be cept with one stage of intermediate relays, this supported by an area-wide cellular overlay net- can be generalized to a multistage concept as work [17]. There, a hybrid routing scheme called suggested in [15]. The idea is to group together Cellular Based Multihop (CBM) routing has closely spaced relays, thereby forming a stage been studied, where route requests are sent to that implements a virtual antenna array as the BS of the overlaying cellular network. This

IEEE Communications Magazine • September 2004 85

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. Compared to ad hoc networks , Point-to-point link networks applying Fixed relaying via fixed network (Internet) infrastructure do not need complicated distributed routing Point-to-multipoint link algorithms, while retaining the flexibility of being Access able to move the point relays as the traffic patterns change over time.

Multihop link Wide area cellular BS Fixed relays

Figure 6. Wireless media system: integration with mobile radio (source: [10]).

central entity determines the route and responds second hop. In [4] an aggressive strategy that with a packet comprising a series of mobile does not require any new channels for relaying is nodes willing to relay the data traffic between adopted: the relay channel is always selected the source and destination. Service and route among the already used channels in the adjacent discovery is performed in the overlay cellular cells. network and the packet data transmissions in the Various selection schemes for the relay and microrange multihop network. This exploits both relay channel, from random to smart selection, the ability of the macro network to communicate with and without power control, are considered with all of its nodes, and the throughput efficien- in [4]. It is observed that with the proper selec- cy of multihop transmission in the microrange tion of relay, relay channel, and relay power, a layer. Results have shown that CBM leads to low significant enhancement in high-data-rate cover- packet drops due to wrong route information age can be attained through two-hop mobile and adds little overhead to network traffic [17]. relaying. The observed trends and corresponding Moreover, it allows fast packet delivery because conclusions are as follows: of quick route establishments, and the routing •Performance gains due to relaying increase overhead increases only linearly with the number as the number of wireless terminals in the sys- of nodes, which indicates that CBM scales very tem increases. well with network size. •Employing power control in both hops fur- It is known that the protocol overhead associ- ther enhances performance, especially as the ated with establishing and maintaining the routes cells get small; the returns due to power control in multihop networks may become severe, thus become substantial for interference-limited cells. reducing the potential efficiency of multihop •The maximum relay transmit power level is techniques substantially. Limiting the number of an important factor only in large cells; in small hops (e.g., employing a single intermediate relay cells most of the benefits are gained with rela- station on any route) would greatly simplify tively small relay transmit power levels. There- routing complexity. fore, the impact of relaying on wireless terminals’ batteries may not be significant in microcellular RADIO RESOURCE MANAGEMENT systems. Radio resource management concerns the •Performance gains are quite sensitive to the assignment of BS, channel, and transmit power. relay selection criterion. If relays are chosen ran- In view of this, the sensitivity of radio coverage domly, performance gets worse than for the no- to the selection of relay, relay channel, and relay relaying case (this is analogous to the case where power control are investigated in [4] for a cellu- a user is connected to a wrong BS). However, lar TDMA system where two-hop mobile relay- highly suboptimal (with minimal intelligence) ing is employed whenever necessary. but implementation-wise feasible relay selection Whenever relaying is performed, an addition- schemes (e.g., relay selection based solely on al time or frequency channel is required for the proximity through the use of the Global Posi-

86 IEEE Communications Magazine • September 2004

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. tioning System, GPS, data available at the BSs) Figure 6 also illustrates by means of an exam- still yield significant coverage enhancements. ple the discontinuous radio coverage available WMS is an example •Once a good relay is selected, performance from WMS in urban areas. The small picocells gains become fairly insensitive to the relay chan- highlighted around the MPs represent areas of a new radio nel selection criterion. Therefore, in systems where broadband radio coverage is available via access network with limited resources for monitoring and con- preplanned multihop communication. architecture for a trol purposes, priority should be given to proper From Fig. 4 it is apparent that the WMS relay selection rather than proper relay channel architecture scales quite well and is able to pro- mobile broadband selection. vide the traffic capacity of an AP to a small system using fixed For infrastructure-based solutions with fixed (pico) or much larger (micro) cell built up from relays, the selection of relays is much simpler many picocells defined by the AP and its related relays to provide and more or less predefined. For this case, possi- RSs, resulting in a very cost-efficient and flexible radio coverage to ble concepts can be based on centrally con- infrastructure. otherwise shadowed trolled heuristic methods for relay channel A mobile terminal must be able to support selection within a single multihop cell. Selection both mobile cellular radio and WMS air inter- areas. Its goal is to criteria involve the mutual interference between faces, but not at the same time. An operator of a ensure highest relay channels. WMS that is integrated into a cellular mobile radio system might apply a mixed cost calcula- spectrum efficiency tion to be able to trade the low cost for trans- and lowest possible WMS: A PROTOTYPICAL porting mass data via the WMS against the high transmit power. ELLULAR ELAY ETWORK cost of data transport across cellular systems. C R N This concept, of course, is also applicable to The wireless media system (WMS) is an example indoor service. Furthermore, this operator may of a new radio access network architecture for a end up with a very attractive tariff for all of its mobile broadband system using fixed relays to services compared to an operator purely relying provide radio coverage to otherwise shadowed on 3G technology. areas. Its goal is to ensure highest spectrum efficiency and lowest possible transmit power. WMS ONCLUSIONS architecture makes use of a wireless or mobile C broadband air interface to link low-power (1 W) Relaying is generally considered a method to pico BSs, APs, and RSs, trading the high capaci- ensure capacity and coverage in cellular systems. ty available at APs for radio coverage range. The This article gives an overview of relaying, pre- WMS, illustrated in Fig. 6, is embedded in a cel- senting two main concepts. The first is the coop- lular radio network providing medium-transmis- erative use of relays forming virtual antenna sion-rate communication to high-mobility arrays to exploit the spatial diversity inherent to terminals. The low power requirement for the multihop communications, leading to substantial broadband air interface leads to a picocellular increases in available capacity; the second is the concept relying essentially on multihop commu- wireless media system, a concept for a mobile nication across fixed relays and to some extent broadband system based on fixed relay stations. also on ad hoc networking, using mobile relays Examples of forwarding techniques as well as a at the periphery of the WMS [1, 6, 7]. pattern for wide-area urban coverage using clus- WMS is intended to have a very high multi- ters of access points and relays are proposed. It plexing data rate of between 100–1000 Mb/s at is shown for different application scenarios that the air interface for medium-velocity terminals multihop communications can provide a substan- and high deployment flexibility through the use tial increase in link and network capacity when of mass production building blocks for APs and applied in areas suffering from heavy path loss. relays. The candidate spectrum bands for opera- The benefit for broadband radio systems is that tion of the WMS (e.g., beyond 3 GHz or even the very high capacity that can be expected from beyond 5 GHz) will allow very small equipment these systems can be traded for radio range that size for picocellular BSs/APs and RSs, including would otherwise be limited due to high attenua- the antennas, so the infrastructure can be termed tion at high radio frequencies. This allows meet- more or less “invisible.” RSs might use either ing the expected capacity needs per unit area central or distributed control of the nodes without danger of capacity overprovisioning as involved, as described in [3]. WMS is integrated when relying on single-hop concepts only. into a 3G system, as shown in Fig. 6, by means A thorough investigation of the relaying con- of large hexagonal cells, sharing: cept is enormously complex due to the many • An IPv6 based fixed core network parameters involved, including propagation and • Functions of the cellular system like sub- physical layer challenges, system issues such as scriber identity module (SIM), authentica- medium access and radio resource management, tion, authorization, and accounting (AAA), networking and protocol design, and finally, and localization implementation-related issues. Moreover, a sur- As shown in Fig. 6, the feeder systems con- vey of the literature shows that analytical under- necting APs to the fixed network might be based standing of the subject is far from comprehensive on either wire/fiber or wireless, for example, at the present time. The overview of relaying point-to-multipoint (PMP) LOS radio systems. concepts and problems highlights the importance Both APs and RSs appear like a BS to a mobile of future research in the areas of virtual arrays terminal, and are henceforth referred to as and novel diversity schemes, multiple access and media points (MPs). In Fig. 6 the radio coverage multiplexing schemes, and the combination of areas served by RSs are shown in a color differ- medium access and radio resource management ent from the APs. protocols for multihop networks.

IEEE Communications Magazine • September 2004 87

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. ACKNOWLEDGMENTS less World Research Forum (WWRF). His research interests include the performance of relay-based deployment con- The candidate cepts for next-generation wireless networks, associated The authors would like to thank Shalini Periyal- spectrum bands for resource management protocols, and spectral coexistence war of Nortel Networks, Ottawa and Bassam of wireless standards. operation of the Hashem of Saudi Telecom Company for support and inspiration for this article. Special thanks go BERNHARD H. WALKE ([email protected]) is an WMS (e.g., beyond internationally recognized expert (numerous publications, to all the unnamed participants in the WG4 ses- including several internationally published books) in the 3 GHz or even sions of the WWRF who have added their field of communication networks. His group’s research beyond 5 GHz) will thoughts and ideas to the discussion about relay- work focuses on formal specification, modeling, mathe- based systems. matical and simulative analysis, and pilot implementation allow very small of protocols and services for communication networks. Having worked at AEG-TELEFUNKEN Research Institute, and equipment size for REFERENCES later at Fern University of Hagen (Chair for Data Processing pico cellular BSs/APs [1] H. Yanikomeroglu, “Fixed and Mobile Relaying Tech- Techniques), he has held the chair of Communication Net- nologies for Cellular Networks,” 2nd Wksp. Apps and works at Aachen University since 1990. He is member of and RSs, including Svcs. in Wireless Networks, Paris, France, 3–5 July various national and international expert groups and coor- 2002, pp. 75–81. dination boards (e.g., ITG, AG Mobikom, WWRF) and has the antennas, so [2] H. Wu et al., “Integrated Cellular and Ad Hoc Relaying frequently served as an advisor to the German national regulation body (RegTP). that the infra- Systems: iCAR.” IEEE JSAC, vol. 19, no. 10, Oct. 2001, pp. 2105–15. structure can be [3] B. Walke and G. Briechle, “A Local Cellular Radio Network DANIEL C. SCHULTZ ([email protected]) received for Digital Voice and Data Transmission at 60 GHz,” Proc. his Dipl.-Ing. degree from Aachen University in 1999. He termed more or Int’l. Conf. Cellular and Mobile Commun., London, U.K., works for the chair of ComNets as research assistant since Nov. 1985, pp. 215–25; http://www.comnets.rwth- 2000. Before that he worked as a consultant for Vision less “invisible.” aachen.de/publications/~walke Unltd. GbR, Germany. At ComNets he has worked on IST [4] V. Sreng, H. Yanikomeroglu, and D. D. Falconer, “Relayer projects VIRTUOUS, FUTURE, and SAILOR. He was the Selection Strategies in Cellular Networks with Peer-to-Peer responsible researcher at ComNets for the German govern- Relaying,” IEEE VTC Fall ’03, Orlando, FL, Oct. 2003. ment funded miniWatt project. His current research activi- [5] A. N. Zadeh and B. Jabbari, “Performance Analysis of ties are in relaying technologies in a fixed and planned Multihop Packet CDMA Cellular Networks,” Proc. IEEE infrastructure. GLOBECOM 2001, vol. 5, San Antonio, TX, Nov. 2001, pp. 2875–79. PATRICK HERHOLD ([email protected]) (S02) [6] N. Esseling, H. S. Vandra, and B. Walke, “A Forwarding received his Dipl.-Ing. degree from Technische Universität Concept for HiperLAN/2,” Proc. Euro. Wireless 2000, Dresden in 2001, after having studied at TU Dresden and Sept. 2000, Dresden, Germany, pp. 13–18. Virginia Polytechnic Institute and State University. During [7] B. Walke et al., “IP over Wireless Mobile ATM Guaran- one of his internships he gained experience in radio net- teed Wireless QoS by HiperLAN2,” Proc. IEEE, vol. 89, work planning and optimization with Vodafone D2 GmbH Jan. 2001, pp. 21–40. in Munich. Since 2002 he is with the Vodafone Chair at [8] G. Neonakis Agglou and R. Tafazolli, “On the Relaying Technische Universitt Dresden, where he is working toward Capability of Next-Generation GSM Cellular Networks,” his Ph.D. degree. His current research interests include the IEEE Pers. Commun., vol. 8, no. 1, Feb. 2001, pp. 40–47. design of cooperative relaying networks and system-level [9] J. Habetha and B. Walke, “Fuzzy Rulebased Mobility performance evaluation of mobile communications sys- and Load Management for Self-Organizing Wireless tems. Networks,” J. Wireless Info. Networks, Special Issue on Mobile Ad Hoc Networks: Standards, Research, Applica- HALIM YANIKOMEROGLU ([email protected]) received a tions, vol. 9, Apr. 200, pp. 119–40. Ph.D. degree in electrical and computer engineering from [10] B. Walke, R. Pabst, and D. Schultz, “A Mobile Broadband the University of Toronto, Canada, in 1998. Since then he System Based on Fixed Wireless Routers,” Proc. Int’l. Conf. has been with the Department of Systems and Computer Commun. Tech., 2003, Beijing, China, pp. 1310–17. Engineering at Carleton University, Ottawa, where he is [11] N. Esseling et al., “A Multi Hop Concept for HiperLAN/2: now an associate professor. His research interests include Capacity and Interference,” Proc. Euro. Wireless 2002, vol. almost all aspects of wireless communications with special 1, Florence, Italy, Feb. 2002; http://www.comnets.rwth- emphasis on cellular multihop networks, radio resource aachen.de/cnroot_engl.html/~publication, pp. 1–7. management, and CDMA multi-antenna systems. He has [12] W. Mohr, R. Lüder, and K.-H. Möhrmann, “Data Rate been involved in the steering committees and technical Estimates, Range Calculations and Spectrum Demand program committees of numerous international confer- for New Elements of Systems Beyond IMT-2000,” Proc. ences in wireless communications; he has also given sever- 5th Int’l. Symp. Wireless Pers. Multimedia Commun., al tutorials in such conferences. He is Technical Program Honolulu, HI,, Oct. 2002. Co-Chair of the IEEE Wireless Communications and Net- [13] E. Zimmermann, P. Herhold, and G. Fettweis, “On the Per- working Conference 2004. He is an editor for IEEE Transac- formance of Cooperative Diversity Protocols in Practical tions on Wireless Communications, and a guest editor for Wireless Systems,” Proc. 58th VTC, Orlando, FL, Fall 2003. Wiley Journal on Wireless Communications & Mobile Com- [14] W. Zirwas et al., “Signaling in Distributed Smart puting. Currently he is serving as Vice-Chair of the IEEE Antennas,” Proc. 7th OFDM Wksp., Hamburg, Ger- Technical Committee on Personal Communications. He is a many, Sept. 200, pp. 87–91. registered Professional Engineer in the province of Ontario, [15] M. Dohler, A. Gkelias, and H. Aghvami, “2-Hop Dis- Canada. tributed MIMO Communication System,” IEE Elect. Lett., June 2003. SAYANDEV MUKHERJEE ([email protected]) (S’92-M’96) [16] H. Viswanathan and S. Mukherjee, “Performance of obtained a B.Tech. degree in electrical engineering from Cellular Networks with Relays and Centralized Schedul- the Indian Institute of Technology, Kanpur, in 1991, and ing,” Proc. IEEE VTC, Orlando, FL, Oct. 2003. M.S. and Ph.D. degrees in electrical engineering from Cor- [17] M. Lott et al., “Hierarchical Cellular Multihop Net- nell University, Ithaca, New York, in 1994 and 1997, works,” Proc. 5th Euro. Pers. Mobile Commun. Conf., respectively. Since 1996, he has been a member of techni- Glasgow, U.K., 22–25 Apr., 2003. cal staff in the Wireless Research Laboratory, Bell Laborato- riess, Lucent Technologies, New Jersey. His research BIOGRAPHIES interests include stochastic models, wireless system simula- tions, intelligent resource allocation in wireless systems, RALF PABST ([email protected]) served student and scheduling and routing in wireless networks. internships at SIEMENS ICN and D2 Vodafone before he received his Diploma in electrical engineering from RWTH HARISH VISWANATHAN ([email protected]) received his B. Aachen University in 2001 and has served since then as a Tech. degree from the Department of Electrical Engineer- research assistant to the chair of Communication Networks ing, Indian Institute of Technology, Madras, in 1992, and (ComNets) of Aachen University, where he is working M.S. and Ph.D. degrees from the School of Electrical Engi- toward his Ph.D. degree. He has worked on various IST neering, Cornell University, in 1995 and 1997, respectively. projects (WSI, WWRI, NEXWAY, ANWIRE, and WINNER) and He was awarded the Cornell Sage Fellowship during aca- is actively involved in the organization of WG4 in the Wire- demic year 1992–1993. He is presently with Bell Laborato-

88 IEEE Communications Magazine • September 2004

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply. ries, Lucent Technologies, Murray Hill, New Jersey as a tative of the IEEE UKRI Section and a member of the Stu- member of technical staff. He has worked extensively on dent Activity Committee of IEEE Region 8. design and analysis of performance enhancement schemes The overview of for wireless systems and in particular on the application of HAMID AGHVAMI ([email protected]) obtained his relaying concepts multiple-antenna technology to third-generation wireless M.Sc. and Ph.D. degrees from King’s College, University of systems. His research interests include information theory, London, England, in 1978 and 1981, respectively. He is cur- and problems communication theory, wireless networks, and signal pro- rently a professor in telecommunications engineering at cessing. King’s College and director of research in the Department of highlights the Electronic and Electrical Engineering. He has published over MATTHIAS LOTT [M] ([email protected]) received a 150 research papers and has lectured on digital radio com- importance of Diploma in electrical engineering from the University of munications worldwide. He is a member of the U.K. IEE and future research in Braunschweig and a Ph.D. (Dr.-Ing.) degree from RWTH Chairman of the Communications Chapter of U.K. and R.I.. Aachen, Germany, in 1996 and 2001, respectively. Since He has served on the technical program and organizing the areas of virtual 2000 he is an employee at Siemens AG in the Department committees of many international and European confer- of Future Radio Concepts, where he is involved in Euro- ences, and founded the International Symposia on PIMRC. arrays and novel pean and German research projects. He has published over 50 papers and holds several patents in the areas of chan- DAVID D. FALCONER [F] ([email protected]) received a diversity schemes, nel modeling, modulation, and protocols for link and net- B.A.Sc. degree in engineering physics from the University of multiple access and work layers. His recent research has focused on the Toronto in 1962, and S.M. and Ph.D. degrees in electrical problems of WLAN, ad hoc networks, and interworking of engineering from Massachusetts Institute of Technology multiplexing IP and the link layer. He received ITG’s best paper award in (MIT) in 1963 and 1967, respectively. After a year as a post- 1999 and is a member of VDE. doctoral fellow at the Royal Institute of Technology, Stock- schemes, and the holm, Sweden, he was with Bell Laboratories, Holmdel, New WOLFGANG ZIRWAS ([email protected]) received Jersey from 1967 to 1980. During 1976–1977 he was a vis- combination of his Diploma in communication technologies in 1987 from iting professor at Linkping University, Sweden. Since 1980 medium access and the University of Munich. He started his work at Siemens he has been at Carleton University, Ottawa, Canada, where Munich in the central research department for communica- he is a professor in the Department of Systems and Com- radio resource tion technologies with a focus on high-frequency issues puter Engineering. Since 1990 he has been active in a and developed high-data-rate TDM fiber systems for data research project on broadband wireless communication at management rates up to 40 Gb/s. Further research topics were broad- EHF radio frequencies, involving several universities, spon- band access systems over coax cable, twisted pairs, and sored by the Canadian Institute for Telecommunications protocols for finally radio-in-the-loop systems. Besides research topics he Research. From 1990 to 1998 he led the project. multihop networks. developed hardware trial systems and conducted related field trials. In 1999 he joined the Mobile Network Division GERHARD P. FETTWEIS ([email protected]) earned at Siemens AG Munich, where he is project manager for his Dipl.-Ing. and Ph.D. degrees from Aachen University of the Bundesministerium fr Bildung und Forschung funded Technology (RWTH) in Germany. From 1990 to 1991 he was project COVERAGE in research of future radio systems a visiting scientist at the IBM Almaden Research Center, San based on multihop and OFDM. Jose, California, working on signal processing for disk drives. From 1991 to 1994 he was with TCSI Inc., Berkeley, MISCHA DOHLER [StM] ([email protected]) obtained his M.Sc. California, responsible for signal processor developments. degree in telecommunications from King’s College, Univer- Since September 1994 he has held the Vodafone Chair at sity of London, England, in 1999, and his Master’s and Technische Universität Dresden, Germany. He co-founded Diploma in electrical engineering from Dresden University Systemonic as CTO, a startup acquired successfully in 2002 of Technology, Germany, in 2000. He is currently lecturer by Philips Semiconductors. The main focus is on broadband at the Centre for Telecommunications Research, King’s Col- wireless chipsets. He has been an elected member of the lege London, where he works toward a PhD degree. Prior IEEE Solid State Circuits Society Board (Administrative Com- to telecommunications he studied physics in Moscow. He mittee) since 1999, and was associate editor for IEEE Trans- won various competitions in mathematics and physics ,and actions on Circuits and Systems II and the IEEE Journal on participated in the third round of the International Physics Selected Areas in Communications Wireless Series. He also Olympics for Germany. He has published various research serves on company supervisory boards, and on industrial as papers and holds four patents. He is the Student Represen- well as research institutes’ advisory committees.

IEEE Communications Magazine • September 2004 89

Authorized licensed use limited to: Stanford University. Downloaded on February 1, 2010 at 12:29 from IEEE Xplore. Restrictions apply.