Chapter 7

Relaying for IMT-Advanced

Afif Osseiran, Ahmed Saadani, Peter Rost and Alexandre Gouraud

This chapter overviews relaying techniques in general, and describes them within the International Mobile Telecommunications Advanced framework in particular. The most common relaying deployment scenarios are analyzed and the efficient combination of Coordinated MultiPoint transmission or reception and relaying for an International Mobile Telecommunications Advanced cellular networks is evaluated. Relaying for indoors scenario, is compared to femtocells. Finally an overview of cooperative relaying is given.

7.1 An Overview of Relaying

The main driving force in the development of wireless communication networks and systems is to provide, among other aspects, increased coverage and support for higher data rates. At the same time, the cost of building and maintaining the system is of great importance and is expected to become even more so in the future. Moreover, higher demand on data rates and on longer battery life are requested from the end-user perspective. Until recently the main topology of wireless communication systems has been fairly unchanged, for the first three generations of cellular networks. The topology of existing wireless communication systems has been hitherto characterized by the cellular architecture, which consists of fixed radio BS and UEs as the only transmitting and receiving entities in the network. Alternatively a radio device node called a Relay Node (RN) can be deployed in order to help wirelessly conveying the information between the BS and the UE. These RNs may consist mainly of passive repeaters that forward analog signals in scenarios where a wired backhaul is not feasible or is very costly. Alternatively active and smart Relay Nodes (RNs) can be deployed. In that case the RNs may communicate with other network elements (e.g. Base Station (BS), another RN or a User Equipment (UE)).

Although some cellular operators have been deploying RNs in form of repeaters in order to extend the coverage of the networks, it was only from 2005 that RNs were specified in major cellular standards. IEEE was quick to adopt it in 2007, in its specification (IEEE 2007), later in 2010, 3GGP followed suit (3GPP 2010).

7.1.1 Relay Evolution

Relaying was a common technique used to convey messages over large distances, in ancient empires such as Egypt, Babylon, China, Greece, Persia, Rome and the Omeyyad. The messages were in various form such as beacon fire relayed by tower or mountain peak. A more common method was sending messengers on horseback between RNs until the final destination is reached. With the advent of science, communication techniques improved. In 1793, the Chappe brothers proposed a telegraph system relying on RNs equipped with telescopes and lighted by lamps. An example of the Chappe brothers' RN is shown in Figure 7.1.

In modern time, RNs were initially simple devices, which amplify a signal and forward it immediately, and were mainly intended to extend the coverage of the wireless system. These were low-cost devices, compared to BSs, and didn't include any base band processing and hence no network protocol operation was possible. The backhaul connection was usually implemented with microwave links in own frequency bands to avoid interference with the access link, or in-band with appropriate (receive and transmit) antenna isolation avoiding Larsen-like effects. However, this flexible approach revealed significant drawbacks after its deployment, such as difficult monitoring of a RN's operation as a physical site visit was required, which significantly increased the operational costs. Operators were therefore reluctant to deploy relay-based solutions and used them on a case-by-case basis when no other alternative was suitable.

With the growth of data traffic and the emergence of new services, network densification is one solution to increase the network capacity. Increasing the density of deployed BSs implies additional backhaul that corresponds to increased deployment costs. One solution to this problem is to use RNs that are smarter than a simple repeater. This encourages the definition of new RN families that are monitored by the network and could have different levels of intelligence. The simplest RN case is called “Layer 1” RN where the RN is able to receive, Amplify-and-Forward (AF) the received signal and also to exchange some control information with the network. “Layer 1” RN can be seen as a repeater with control signalings. “Layer 2” RNs are able to Decode-and-Forward (DF) the successfully decoded signal. It also has its own Layer 2 protocols such as Medium Access Control (MAC) and Radio Link Control (RLC) layers. Finally, “Layer 3” RNs are comparable to BSs with wireless backhaul (i.e. the three layers are present in that type of RN). The “Layer 3” RN has been standardized and will be described in the following. It should be noted that a Layer-2 relaying solution (IEEE 2007; Pabst et al. 2004; Döttling et al. 2009, Chapter 8) was developed on how to integrate into the LTE standard, by the Wireless World Initiative New Radio (WINNER) project.

7.1.2 Relaying Deployment Scenarios

Relay nodes, by construction, are flexible devices in terms of deployment positions and transmit power. Those inherent qualities make them very useful in several deployment scenarios. In particular, three main scenarios, which are depicted in Figure 7.2, can be identified:

• Network densification where the aim is to increase the network capacity with the introduction of new BS sites. However in urban and suburban zones it is difficult and very time-consuming to negotiate contracts for new sites. In addition, the installation of backhaul is quite costly, which makes RNs a cost-efficient solution as they are exploiting wireless backhaul. Further, the RNs' positioning is more flexible. In fact, they can be deployed on street lamps, building walls, or even buses. Hence, RNs represent an opportunity to ubiquitously provide high data rates.
• Coverage extension relates to dead zones not covered by the network in urban and suburban environments. Using RNs will extend the coverage to these zones without the need for a wired backhaul. On the other hand, the lack of a dense network can cause bad coverage in indoor environments. For instance, rural buildings without a broadband connection can suffer from a lack of multimedia services. The use of indoor RNs is one solution to cover them with the required data rates.
• Fast roll-out where the fast penetration with new wireless services requires a very dynamic and fast change of an operator's infrastructure. Relay nodes can be deployed much faster and support operators to be the first to offer new services.

7.1.3 Relaying Protocol Strategies

This section gives a brief overview of Layer 1 and Layer 2 relaying. A more detailed overview is given in Kramer et al. (2005) where DF- and CF-based protocols are analyzed.

Layer 1 Relaying

Layer 1 relaying, also called Amplify-and-Forward (AF), forwards a linear function . More specifically, the transmitted RN signal is an amplified version of its own received signal, ie. . Amplify-and-Forward is a very simple and low-complexity protocol, which requires no digital processing. However, it also does not separate the useful source signal and noise signal at the RN. Hence, it also forwards an amplified version of the receiver noise at the RN. An overview on AF is given in Hammerström et al. (2004), where AF is applied to a block-fading multiple-relay scenario. In particular, each RN forwards a phase-rotated version of its received signal, which introduces additional time diversity due to the time-variance of the effective channel. However, in Laneman et al. (2004) it was shown that the achievable Diversity-Multiplexing-Tradeoff (DMT) of digital relaying (for instance incremental relaying) always dominates AF.

Decode-and-Forward

By contrast to AF relaying, DF and CF apply nonlinear functions on the received signal at the RN. Figure 7.3(b) illustrates the DF strategy for a single communication pair, where the RN r decodes the source message submitted in block t using its own receive signal . Based on the decoded source message a RN message is selected and transmitted in the next block , which does not necessarily have the same codebook (i.e. the set of all possible codewords) size as the source message. More specifically, the RN can assign multiple source messages to the same RN message. The destination node then decodes first the RN message , which helps to reduce the number of possible source messages to the set of messages, which is assigned to the decoded RN message. Afterwards, the destination may decode the source message using this prior knowledge. In the case of cellular systems, the main benefit of relaying is the significantly reduced path loss. Hence, the destination solely exploits the RN signals, which is called in the following noncooperative relaying. By contrast, if the destination experiences similar channel gains towards source and relay, it might be beneficial to exploit the diversity provided by combining both paths, which is referred to as cooperative relaying.

The achievable rates of the DF strategy are given in Cover and Gamal (1979, Theorem 1). It is relevant to mention that after decoding at the RN only correctly decoded message should be re-encoded as otherwise the system diversity is decreased (Laneman et al. 2004). A practical implementation of DF is to determine parity bits at the RN and forward those to the destination. Finally, we distinguish DF with different codebook sizes at source and relay, which is also referred to as irregular coding while using the same codebook size at each terminal is referred to as regular coding (Xie and Kumar 2004, 2005).

Compress-and-Forward

While DF provides transmit diversity, CF provides receive diversity and is illustrated in Figure 7.3(c). Instead of decoding the source message, the RN quantizes its received signal using , which then represents a digital version of the analog received signal . Using the RN message , the quantized signal is communicated to the destination. At the destination, this quantized signal is restored using the destination's own receive signal as side information. Once the quantized RN signal is known at the destination, the destination can use both and to decode the source message. The achievable rates of the CF strategy are given in Cover and Gamal (1979, Theorem 6).

The major benefit of CF over DF is the ability to provide a second (quantized) received signal to the destination. This is of particular interest if the RN is placed close to the destination node and experiences similar channel conditions. In this case, DF cannot improve the data rates as the link between source and RN is the bottleneck and provides the same data rates as a direct transmission from source to RN. In Rost and Fettweis (in press), a regular encoding approach for CF has been presented, which is able to improve data rates in networks with multiple RNs.

It should be mentioned that another relaying decoding strategy, called Demodulate-and-Forward (DmF) relaying, or uncoded DF (Chen and Laneman 2006), exists. In DmF, the RN simply demodulates the received signals and retransmits it. For instance, assume a Binary Phase Shift Keying (BPSK) modulation is used, then applying DmF is equivalent to use CF with one bit of quantization.

7.1.4 Half Duplex and Full Duplex Relaying

Relay Nodes can be further differentiated into those operating in half-duplex and those operating in full-duplex. Half-duplex RNs are subject to an orthogonality constraint, which implies that they either transmit or receive on a time-frequency resource and therefore must operate in a half-duplex mode. This requires that the RN coordinates its resource allocation with UEs in the uplink and the assigned BS in the downlink. Such a coordination can be done using a predefined RN-zone resulting in a static centralized solution or using more dynamic solutions, where the RN is allowed to assign resources (distributed coordination). In addition, half-duplex RNs have less stringent requirements on the deployed hardware compared to full-duplex RNs.

In the case of full duplex RNs, they receive the signal on one carrier frequency, process it and transmit it on the same frequency with a small delay compared to the received frame duration. However, RNs continue simultaneously receiving from the source. This assumes that there is good isolation between the receiver antennas and the transmit antennas of RNs. There is only a minor rate loss caused by the delay. There is hence a tradeoff between performance and RN cost (due to the antenna isolation).

7.1.5 Numerical Example

In order to illustrate the operating regions of the discussed protocols (in Section 7.1.3), Figure 7.5 shows the achievable end-to-end rates for the line network illustrated in Figure 7.4 and for the case of full- and half-duplex RNs. For these results, we use the same transmit power at source and RN () as well as AWGN with the same noise power at RN and destination (). Furthermore, the Signal to Noise Ratio (SNR) of the source-destination channel is given by and we use the path loss exponent . Figure 7.5(a) shows the achievable rates for full-duplex relaying and for the previously discussed protocols as well as an upper bound on the capacity and a combined approach Cover and Gamal (1979, Theorem 6), which integrates both DF and CF. The shown upper bound uses the max-flow min-cut theorem and is achievable in very specific cases such as the degraded RN channel. In addition, Figure 7.5(a) shows the performance of direct communication with power normalization () and without power normalization ().

We can observe in Figure 7.5(a) that full-duplex relaying has the potential to significantly increase the achievable rate particularly for . Some interesting points can be singled out such as when . In that case CF provides the same performance as direct communication with power normalization. This shows that CF implements a distributed form of receive diversity. At , DF provides the same performance as direct communication without power normalization as both RN and destination experience the same channel.

Figure 7.5(b) shows DF and a max-flow min-cut upperbound for the same scenario but using half-duplex RN nodes. It draws a similar picture as for full-duplex relays, that is, instead of a maximum performance of about 100% provided by DF and full-duplex relays, half-duplex relaying only improves the maximum by about 50%. While full-duplex DF achieves capacity for a wide range of parameters, half-duplex DF only achieves performance close to capacity at when the time slot for communicating the data to the RN is very short. A detailed discussion of achievable rates in half-duplex multiple-relay channels can be found in Rost and Fettweis (in press).

In fading environments the picture changes as the RN introduces a diversity gain (Laneman et al. 2004). Finally, we can see in Figure 7.5(a) that digital relaying always outperforms analog relaying, which shows that the lower complexity of analog RNs comes at the expense of performance. We compare the relaying protocols with power-normalized direct communication, which is not always possible in mobile communication systems as the maximum transmission power is limited by the hardware employed such as the power amplifier. In a relaying system the power is more homogeneously distributed in the network and power constraints do not need to be changed.

7.2 Relaying in the Standard Bodies

3GPP started to specify relaying functionalities in LTE-Advanced (LTE-A) Release 10 (Rel-10) (3GPP 2010), while in IEEE 802.16m the relaying definition started in 2009. In 3GPP, a donor cell is defined as a cell via which the RN is wirelessly connected to the radio access network. One donor cell can support more than one RN and each RN can be transparent or nontransparent towards the UE: the difference is whether the UE is aware of the RN (nontransparent case) or not (transparent case) when it communicates with the network via the RN. The backhaul wireless connection could be done in the system frequency band or in another available band. However, in the latter case, the operator needs to acquire additional bands hence increasing the total deployment cost. The use of free bands does not allow operators to control the backhaul quality. For this reason 3GPP focuses in its specification process on the International Mobile Telecommunications Advanced (IMT-Advanced) bands. One can distinguish two ways for relaying: inband and outband RNs, which are further detailed in the following.

7.2.1 Relay Types in LTE-Advanced Rel-10

Two categories of RNs are defined in LTE-A Rel-10. In the first category, Type 1, the RN has its own Identity (ID) and own reference signalling. In the second category, Type 2, the RN is transparent to the UE and does not have its own reference signalling.

In case the BS-RN link shares the same carrier frequency with the RN-UE links the RN is said to be “inband”. In fact, in order to be backward compatible with Long Term Evolution (LTE) Release 8 (Rel-8) where the UEs should be able to be connected to the donor cell, the backhaul and UEs share the same resources.

For “outband” RNs, the BS-RN link does not operate in the same carrier frequency as the RN-UE links. In this case, Rel-8 UEs should also be able to be connected to the donor cell.

In the following, the main features of the RN types in Rel-10 are described and summarized in Table 7.1.

Table 7.1 Relay classification in 3GPP Release 10.

Class	Cell ID	Inband/Outband
Type 1	Yes	Inband half duplex
Type 1.a	Yes	Outband full duplex
Type 1.b	Yes	Inband full duplex
Type 2	No	Inband full duplex

Resource Partitioning for Inband RNs

In an LTE-A system the backhaul link (Un) and access link (Uu) are multiplexed in time on a single carrier frequency (3GPP 2010).

• For Frequency Division Duplex (FDD) mode, in the Un link, the BS to RN transmissions are done in the Downlink (DL) frequency band, whereas the RN to BS transmissions are done in the Uplink (UL) frequency band.
• For the Time Division Duplex (TDD) mode, the BS to RN transmissions are done in the DL subframes of the BS and RN. The RN to UE transmissions are done in the UL subframes of the BS and RN. This resource partitioning does not impact the LTE Rel-8 UEs.

Backward Compatibility Issues

When the RN is inband it should operate in a half-duplex mode unless its antennas are well separated and isolated. Consequently, the RN can either transmit or receive. In order to make this “gap in time” unnoticeable to previous LTE releases and in particular Rel-8, the MBSFN feature is exploited (3GPP 2010) as shown in Figure 7.6. In fact the MBSFN subframes are configured as frames where the RN transmits to (resp. receives from) BS and no UE is allowed to transmit (resp. receive) to the BS. In LTE, it is possible to allocate up to six MBSFN subframes per frame for the backhaul link. The allocation is done in a semistatic way.

7.2.2 Relay Nodes in IEEE 802.16m

As in LTE-A Rel-10, two categories of RNs are defined in IEEE 802.16m (IEEE 2009, 2010), namely transparent and nontransparent. The nontransparent and transparent categories correspond to RN Type 1 and Type 2 in LTE-A Rel-10, respectively. A nontransparent RN can operate in both centralized and distributed scheduling mode, while a transparent RN can only operate in centralized scheduling mode. An illustration of the RNs types used in the standard bodies is shown in Figure 7.7. As it can be seen, RN3 is of Type 1 (or non-transparent) since it has its own cell ID. On the other hand, RN1 and RN2 are of Type 2 (or transparent) and both have the same cell ID as the donor BS.

7.3 Comparison of Relaying and CoMP

Although Coordinated Multipoint transmission or reception (CoMP), introduced in the previous chapter, offers remarkable improvements. It still faces the problem of shadowed areas where not necessarily intercell interference but strong path loss impairs the cell throughput. In those situations relaying is the means of choice to improve the quality of service. Hence, IMT-Advanced cellular networks will not face questions about whether either CoMP or relaying is used but rather about how both can be efficiently combined.

Figure 7.8 illustrates different interference situations in cellular networks, where the following interference regions can be defined:

In the following, we will explore the possibilities and challenges of integrating CoMP and relaying in a cellular network.

7.3.1 Protocols and Resource Management

The integration of CoMP and relaying can be well described by the relay-assisted interference channel (Rost et al. 2009), which is illustrated in Figure 7.9. It models the case where two communication pairs are supported by two RNs and both transmitters are connected by a finite capacity backhaul link. A link-level analysis of this channel has been conducted in Rost et al. (2009) and in Rost et al. (in press) system-level results have been presented. Furthermore, in Somekh et al. (2010) analytical results for a similar scenario based on a linear Wyner model and using DF RNs have been presented.

CoMP Transmission and Detection

As was mentioned in Chapter 6, CoMP transmission exploits the possibility of cooperatively transmitting from multiple BSs and therefore improving the Signal to Interference plus Noise Ratio (SINR) particularly of interference-limited UEs. One prominent approach is to apply the Wiener transmit filter (Joham et al. 2005) based on the knowledge of the compound channel and the transmit data of the cooperating BSs. This implies strong requirements on the backhaul as a significant amount of Channel State Information (CSI) and data must be exchanged between the individual BSs. In addition, particularly nonlinear approaches such as Dirty Paper Coding (DPC) (Costa 1983) are very sensitive to imperfect CSI caused by low-rate feedback and imperfect channel state estimation. Hence, it is more likely to apply CoMP at colocated BSs, which share the same cabinet at one site. This allows for an efficient exchange of data between the individual BSs and counteracts intersector interference, which poses one of the main challenges in cellular networks. Like CoMP transmission, CoMP detection can exploit the availability of compound CSI and data knowledge in order to improve the array gain (centralized approach) and to improve the multiplexing gain (Marsch and Fettweis 2007).

Coordinated Relaying with CoMP feeder link

Hence, the required CSI updates are less frequent and the overhead, which needs to be exchanged between cooperating BSs, can be reduced. This allows for more efficient backhaul usage and much improved performance on the link between BSs and RNs, which usually represents the bottleneck in a relay-assisted cellular network. In addition, the signaling overhead can be reduced as less RNs than UEs must be served and therefore the scheduling and the Radio Resource Management (RRM) signaling require less overhead.

On the link between RNs and UEs we need to apply an interference mitigation scheme (Han and Kobayashi 1981), which does not require common CSI and data knowledge at the transmitters. One such approach is Han-Kobayashi (HK) coding where both transmitters divide their messages into a common and private part. The common part is decoded by both receivers in order to reduce the interference for the private part, which is only decoded by the respective receiver. Although the common part is decoded by both receivers, it is only exploited by the actual destination while the other receiver only uses it for interference reduction. However, this approach requires a very complex and resource-intensive optimization over all possible power allocations, which makes the approach less suitable for practical implementations. Etkin et al. introduced, in Etkin et al. (2008), an approach that always achieves rates within 1 bit per channel use (bpcu) of capacity¹ and deterministically assigns the individual power levels. Furthermore, empirical observations in Rost (2009) showed that it is sufficient to apply the described scheme based on the long-term SINR and only using two possible power assignments: one where only a private message and one where only a common message is transmitted. In this way, the receiver either ignores interference (only private message) or jointly decodes interference and useful signal (common message).

Integrated Approach

User Equipments that are located in the main lobe direction of a BS or have a Line of Sight (LoS), are likely to experience very high SINR towards the BS. In those situations it is suitable not to serve UEs using relaying but using either a conventional Multiple-Input Multiple-Output (MIMO) transmission or CoMP transmission in the case that the UE experiences strong interference. On the other hand, if UEs are located in shadowed areas or have a very weak non-LoS connection, it is preferable to use relaying. It is difficult to predict in real time the optimal choice with regard to whether relaying or CoMP should be used. Alternatively, it is much simpler to assign UEs to the radio access points with the smallest effective path loss towards the UE, that is, considering the different transmission power at RNs and BSs. In the following analysis, a snapshot-based simulation has been employed and UEs are assigned to the RN or BS, which has the smallest effective path loss towards the UE. Such an approach is also used in systems with femtocells, which face similar problems, and it provides remarkable performance gains (Rost 2009; Rost et al. in press).

The channel as illustrated in Figure 7.9 as well as the protocols applied to this channel can be categorized as shown in Table 7.2. In order to categorize the protocols, we group them in terms of their ability to cooperate or to coordinate across different communication pairs (multiple paths) and across different hops (multiple subsystems, that is, one between BSs and RNs, and one between RNs and UEs). In the case that neither BSs nor RNs share common CSI among each other, we apply intercell TDMA or noncooperative relaying with full reuse of resources. If we allow for interpath cooperation we obtain the previously discussed integrated approach where BSs jointly transmit and receive while RNs apply TDMA between both communication pairs or HK coding. However, we do not consider the case of joint RN-transmission and detection as it is unlikely that RNs obtain the required common CSI and data knowledge in order to coherently transmit. We can further consider protocols where the individual subsystems are cooperating. An example is cooperative relaying, which is detailed later in this chapter.

Table 7.2 The individual protocols classified by their ability to cooperate.

7.3.2 Simulation Results

Finally, we examine results for a system-level simulation using the channel model defined by the European research project WINNER (WINNER-II 2006). In particular, we apply Orthogonal Frequency Division Multiplexing (OFDM) with 100 MHz bandwidth and 2048 subcarriers. The applied simulator uses the channel fading model defined by WINNER for the Urban Macro scenario. Furthermore, in each sector two RNs are deployed at a radius of about 333 m, which corresponds to one-third of the inter-site distance of 1000 m. User Equipment is randomly and uniformly distributed such that on average about 29 UEs are assigned to each cell. We apply a fair scheduler for the assignment of UE resources based on history acquired for each UE. The results presented are obtained for one central site surrounded by two tiers of interferers which totals 54 interfering BSs. All RNs are modeled as half-duplex in-band RNs. All further parameters and a simulator description are given in Rost (2009) and Rost et al. (in press). The previous assumptions differ slightly from the IMT-Advanced as well as 3GPP assumptions where an intersite distance of 500 m and a fixed UE density of 10 UEs per cell is assumed. However, this worst case assumptions illustrates the ability of relaying to significantly improve performance even in cases of low-density deployment and high user density.

Figure 7.10 shows the UL and DL performance for the previously introduced approaches. Consider at first the DL performance shown in Figure 7.10(a). It can be deduced that the integrated approach and CoMP cases (denoted respectively in the figure caption by “2-RN integrated” and “CoMP”) outperform the direct transmission and the relay-only cases (denoted respectively by “Conv.” and “2-RN only”). The relay-only approach is outperformed due to the significant performance loss of UEs close to the BS, which experience a very high path loss towards the RNs. The integrated approach slightly improves the fairness (higher fifth percentile throughput) but does not significantly improve the average throughput. The slightly improved fairness results from the improved channel conditions at the cell edge and the almost equal average throughput follows from the interference cancellation ability of CoMP. However, compare the performance of CoMP with unlimited intersite cooperation (dashed line) and the performance of CoMP with intrasite cooperation without backhaul between individual sites (solid line). While the integrated approach maintains its performance as most of the interference originates from the same site, the CoMP performance significantly drops due to the worse SINR at intersite cell borders.

Now consider Figure 7.10(b), which shows the uplink performance and further emphasizes the conclusions from the discussed downlink results. By contrast to the downlink performance figures, the performance difference between the integrated approach and CoMP is in the range of a factor of 10. This enormous performance difference is a result of the significantly improved path loss between UEs and radio access points, which did not have the same effect in the DL due to the high transmit power at BSs (46 dBm) compared to the transmit power at RNs (37 dBm) and UE (24 dBm). These results demonstrate that beyond IMT-Advanced networks should rather integrate both CoMP and relaying than choosing only one particular technology.

7.4 In-band RNs versus Femtocells

The previous section discussed in-band RNs, which share the same spectral resources for UE-RN and RN-BS links. By contrast, as shown in Chapter 4, femtocells (Chandrasekhar et al. 2008; Knisely et al. 2009) are usually deployed indoors and exploit existing broadband connections as backhaul links towards the core network. This additional deployment of small indoor cells is a cost-efficient alternative to significantly improve the indoor coverage, data rates, and to allow for traffic offloading. Femtocells may also be deployed outdoors (in which case they are called pico cells) and use either wireless out-of-band backhaul towards the assigned BS or a wired backhaul connection towards the core network. Relay Nodes provide more flexibility than indoor and outdoor femtocells, but they suffer from a performance loss as the same spectral resources are used for BS-RN links and RN-UE links.

Nonetheless, RNs and femtocells face the same challenges, that is, intercell and intracell interference coordination and avoidance. In order to assess the performance loss due to the in-band operation of RNs and the intra/intercell interference, Tables 7.3 and 7.4 provide numerical results for the already introduced wide-area deployment and a typical Manhattan street-grid deployment, where we again apply a worst-case scenario, that is, high shadowing between the deployed outdoor-BSs and indoor-UEs due to the high carrier frequency. Consider at first Table 7.3, which shows the mean and the 5th Throughput (TP) for the same deployment as illustrated in Figure 7.8. We compare a deployment of two additional RNs (inband and half-duplex) per macrocell and a deployment of two additional (outdoor) femtocells per macrocell with an unlimited backhaul towards the core network. The same table shows that (outdoor) femtocells improve the average downlink throughput by about 60% and the uplink performance by about 35% compared to inband RNs. This improvement is caused by the increased number of resources available to RNs and BSs for serving UEs. While in the downlink the worst UE performance (represented by the 5th TP) is improved by about 60%, it remains the same in the uplink as the worst UEs suffer from the same insufficient channel conditions.

Table 7.3 System level throughput for wide-area scenario comparing CoMP, relaying and femtocell deployments. The stated number of RNs and femtocells is per macrocell.

Table 7.4 considers a typical Manhattan street-grid using the channel models described in WINNER-II (2006) and UEs placed uniformly indoors and outdoors. Furthermore, we consider two setups. The first one places two outdoor RNs/femtos per macrocell at the adjacent crossings. In the second setup, two additional (indoor) RNs/femtos are placed in the two adjacent buildings. The resulting setups are shown in Figure 7.11. We can see in Table 7.4 that femtocells with a backhaul connection do not significantly improve the performance for the setup of two additional outdoor access points. However, in the case of additional indoor femtocells the performance is improved by a factor of about four in the case of DL and a factor of about two in the case of UL. Furthermore, only a setup with indoor RNs or femtocells is able to provide high data rates due to the very high path loss from outdoor RNs or femtocells towards indoor UEs. These results are supported by recent field trial results (Fettweis et al. 2010) for relay-based communication in a LTE environment. Femtocells provide significant performance benefits, particularly for indoor UEs, but they also require a broadband backhaul connection, which makes the deployment less flexible. Hence, while femtocells appear to be the right choice for indoor UEs, RNs seem to be the better option for outdoor deployments due to their flexibility and cost-efficient deployment.

Table 7.4 System level throughput for Manhattan-area scenario comparing CoMP, relaying and femtocell deployments. The stated number of RNs and femtocells is per macrocell.

7.5 Cooperative Relaying for Beyond IMT-Advanced

One way to introduce diversity in the received signal is to utilize multiple antennas at the transmitter and possibly also at the receiver. An alternative approach is to introduce macro-diversity utilizing cooperative relaying.

The introduction of cooperative communication systems will increase macro- and/or multipath diversity gains. Such systems offer the possibility of reducing the effective path loss between communicating RN entities, which may benefit the end UE. Cooperative relaying systems are typically limited to only two (or a few) hops. In the literature, several names are in use, such as cooperative diversity (Laneman 2002), cooperative coding (Stefanov and Erkip 2004), and virtual antenna arrays (Döhler et al. 2002).

Cooperative communication can be divided into several categories: cooperative diversity, distributed space-time coding (Laneman and Wornell 2002) and cooperative (channel) coding.

There has been a plethora of cooperative diversity, and distributed space-time coding schemes (Anghel and Kaveh 2006; Ben Slimane and Osseiran 2006; Larsson 2003; Osseiran et al. 2007; Scaglione and Hong 2003; Wei et al. 2004). Just to cite a few: Alamouti diversity based cooperative relaying (Anghel and Kaveh 2006), coherent combining based relaying (Larsson 2003), Relay Selection Diversity (RSD), and Relay Cyclic Delay Diversity (RCDD).

Another way to achieve cooperative diversity is to use cooperative channel coding. Cooperative coding designates those schemes where the channel coding operations are distributed within the cooperative nodes instead of being conducted at a single node. The idea of extending convolutional codes, turbo coding and Low-Density Parity-Check (LDPC) to a cooperative manner has been proposed since the early 2000s. For instance, the distributed Rate Compatible Convolutional Codes (RCPC) similar to incremental redundancy Automatic Repeat-reQuest (ARQ) was proposed in 2002 by (Hunter and Nosratinia 2002, 2003). Moreover, distributed turbo coding technique was suggested by (Zhao and Valenti 2003) and distributed Space-Time Trellis Codes (STTC) were analyzed in (Rost and Fettweis 2007). Finally distributed LDPC appeared in (Chakrabarti et al. 2007; Razaghi and Yu 2007). In the following we will describe briefly the most known cooperative relaying schemes. It should be noted that cooperative communications in the context of Network Coding (NC) is treated in Chapter 8.

Relay Selection Diversity (RSD)

Relay Selection Diversity is similar to antenna selection diversity. In RSD the BS will select one RN out of a set of RNs belonging to the BS. The BS transmits in the first transmission phase. In the second transmission phase, the selected RN will forward the information from the BS to the UE. Contrary to classical antenna selection where the selected antenna for transmission is based on short-term statistics (e.g. fast fading), the selection criterion for RSD may be done on a slow basis and may consist of the distance and shadow fading gain.

Relay Coherent Combining

Relay Coherent Combining (RCC) was first proposed in Larsson (2003). It involves multiplying the transmitted signal at each RN by a phase that compensates the one introduced by the channel. In fact the effective channel at the UE will be a constructive summation of all the RN signals transmitting to the desired UE. However, this method requires very detailed phase-CSI at the RN.

Distributed Space Time Block Coding

Distributed Space Time Block Code (STBC) is another way to get the spatial diversity by reusing in a distributed manner the STBC initially designed for MIMO schemes. In the first step the BS transmits the information symbols to the RNs. In the second step each RN encodes the symbols linearly into new symbols corresponding to one line of a STBC matrix. Figure 7.12 illustrates this process in the case of two RNs where and () are linear combinations of symbols (). When multi-hop RNs are used, the synchronization between RNs can be lost. Note that the STBC properties can be destroyed due to imperfect synchronization between the transmitting antennas. Some recent works propose the use of new distributed STBC that are delay tolerant (Damen and Hammons 2007; Nahas et al. 2010).

A particular case of these distributed STBC is the Relay Alamouti code. Relay Alamouti diversity consists of two distributed RNs that are used to mimic conventional Space Time Transmit Diversity (STTD). For instance, the BS will transmit an even number of OFDM symbols in the first transmission phase then the RN will retransmit these symbols during the second transmission phase. For simplicity let us assume that the duration of the transmission phases is two OFDM symbols, for example, and . For the second transmission phase (of equal duration to the first transmission phase), two RNs attached to the BS are selected. Each of the two RNs is equipped with a single antenna. The two antennas of the RNs will act jointly as in the case of two transmitting antennas for a conventional STTD. The only difference is that each of the antennas is attached to a different antenna system instead of being controlled by the same radio unit. After demodulating the two received symbols the RNs will act jointly as a STTD encoder. While the first RN will resend the same symbols that is, , the second RN will swap the order of the received symbols in addition to conjugating them as is done at the second antenna of a conventional STTD encoder, that is, .

Distributed Relay Cyclic Delay Diversity

Distributed Cyclic Delay Diversity (CDD) is also known as RCDD (Osseiran et al. 2007) as it mimics the conventional CDD. The idea is to transmit the same signals from a set of distributed RNs each applying a different time delay (i.e. cyclic shift) and thereby causing multipath fading even when the actual channel is flat. Besides the improved link gain usually obtained from a cooperative relaying system, RCDD introduces frequency selectivity and macro-diversity in OFDM based systems.

Cooperative LDPC coding

Cooperative or distributed LDPC is commonly implemented in a two-hop DF relaying scenario. The encoding is executed in two steps. At first the information bits are encoded at the source (by a generator matrix). Then the RN decodes the received signal. Afterward it computes a new packet of the parity bits (based on the parity-check matrix of the information bit at the source). The new packet is forwarded to the destination. An illustration of a distributed LDPC is shown in Figure 7.13. There has been a large amount of work on the application of LDPC codes to full duplex and half duplex RN (Chakrabarti et al. 2007; Ezri and Gastpar 2006; Hu and Duman 2006; Razaghi and Yu 2007; Wu et al. 2010). In Chakrabarti et al. (2007) and Razaghi and Yu (2007) code designs for irregular coding are proposed. Moreover Ezri and Gastpar (2006) consider the use of independent source and RN codebooks. In Hu and Duman (2006) LDPC codes are applied to the half-duplex RN channel and a random puncturing scheme for LDPC codes was proposed.

In WINNER+ (2010) a new code design for distributed LDPC coding was devised. In fact, most of the distributed LDPC coding schemes proposed in the literature are based either on serial or parallel code concatenation. From the code design point of view, the serial or parallel concatenation of LDPC codes has intrinsic limitations, mainly because parity-check matrices used for decoding at the RN and the destination are included one in the other, resulting in inappropriate matrix topologies (density on nonzero entries, column and row weight distributions, cycles, etc.). The proposed code design in WINNER+ (2010) aims to create incremental redundancy for LDPC codes, while avoiding both serial and parallel concatenation. It is based on a Split-and-Extend (SpE) approach, and allows the construction of codes with enhanced correction capacity and low integration cost. After decoding the received signal, the RN computes extra parity bits by splitting parity checks corresponding to rows of the parity-check matrix. Then the RN transmits these new parity bits towards the destination. The whole process aims to create a new matrix, whose rows correspond to parity checks involving both old and new parity bits. This new matrix can be used at the destination to jointly decode the received signals from both the source and relay.

It was shown in WINNER+ (2010) that the distributed LDPC based on the SpE approach yields up to 4 dB improvement in terms of SNR compared to repetition coding schemes (i.e. the case where the RN forwards simply the information bits to the destination). The detailed results as well the simulation assumption can be found in WINNER+ (2010, App. C).

7.6 Relaying for beyond IMT-Advanced

Relaying is still an active and promising research area. A few of the promising ideas for future directions are summarized below. Among those ideas is the daunting task of multihop relaying (i.e. more than two hops), which has already been considered in an IEEE standard and is expected to be standardized stin the 3GPP in near future. Moreover the use of relaying in a mobile framework such as high-speed trains is economically very attractive.

7.6.1 Multihop RNs

Hitherto we focused on two-hop RNs. In some scenarios the coverage extension provided by a single RN does not suffice but additional RNs are required. There exist two solutions to control multihop RNs: the centralized and the distributed case. The centralized case implies that the BS manages all resource allocations for all involved RNs while in the distributed case RNs independently assign their resources. One problem in multihop relaying is the difficulty of synchronization due to the unavailability of wired backhaul by construction.

7.6.2 Mobile Relay

Mobile Relay is a generic term indicating that the RN is moving. These RNs can be mounted on top of a car, a bus or a train, where a large number of UEs are located and where a wireless high-data rate network access is required. In this case, the backhaul link must cope with extreme conditions such as frequent handovers and therefore the predictability of the trajectory is a good means to reduce overhead, facilitate handovers and improve the backhaul link.

Mobile Relaying represents an interesting area where real-time positioning information will greatly improve the communication link. On the other hand, these RNs introduce new problems for the owners of cars. On top of a public transportation vehicle one can easily imagine who will own, maintain and operate the RN. However, this is less evident if the RNs are targeted for private vehicles, which gives rise to difficulties similar to those problems experienced with femtocell, that is, such as the question which set of UEs is eligible to access the mobile RN. As long as the vehicle is large enough, it is likely that the transmit power of these RNs will be sufficiently high to support a large number of UEs. On the other hand, mobile RNs complicate the cell planning and long-term interference mitigation in networks.

7.6.3 Network Coding

The RN deployment in the network implies that several points share the same information. Moreover, these points are able to support a high processing complexity. In order to increase further the network gains offered by cooperative relaying schemes, network coding can be used on top of cooperative communication. Hence the naming of cooperative networks. These networks will be presented in the following chapter.

Note

1. Bits per channel use (i.e. bits per transmission) is a generic unit for measuring channel capacity.

References

3GPP 2010 Further Advancements for E-UTRA Physical Layer Aspects. Technical Report 36.814 v9.0.0, 3rd Generation Partnership Project (3GPP), http://www.3gpp.org/ftp/Specs/2010-03/Rel-9/36_series/36814-900.zip.

Anghel P and Kaveh M 2006 On the performance of distributed space-time coding systems with one and two non-regenerative relays. IEEE Trans. Wireless Commun. 5(2), 682–692.

Ben Slimane S and Osseiran A 2006 Relay communication and delay diversity for future communication systems. Proc. VTC 2006 Fall – IEEE 64th Vehicular Technology Conf., pp. 1– 5.

Chakrabarti A, Baynast AD, Sabharwal A and Aazhang B 2007 Low density parity check codes for the relay channel. IEEE J. Selected Areas in Commun. 25(2), 280–291.

Chandrasekhar, V. Andrews J and Gatherer A 2008 Femtocell networks: a survey. IEEE Commun. Mag. 46(9), 59–67.

Chen D and Laneman J 2006 Modulation and Demodulation for Cooperative Diversity in Wireless Systems. IEEE Trans. Wireless Commun. 5(7), 1785–1794.

Costa M 1983 Writing on dirty paper. IEEE Trans. Inform. Theory IT-29(3), 439–441.

Cover T and Gamal AE 1979 Capacity theorems for the relay channel. IEEE Trans. Inform. Theory 25(5), 572–584.

Damen M and Hammons A 2007 Delay-tolerant distributed tast codes for cooperative diversity. IEEE Trans. Inform. Theory 53, 3755–3773.

Dohler M, Lefranc E and Aghvami H 2002 Virtual antenna arrays for future wireless mobile communication systems ICT 2002, Beijing, China.

Döttling M, Mohr W and Osseiran A 2009 Radio Technologies and Concepts for IMT-Advanced. John Wiley & Sons, Ltd, Chichester.

Etkin R, Tse D and Wang H 2008 Gaussian interference channel capacity to within one bit. IEEE Trans. Inform. Theory 54(12), 5534–5562.

Ezri J and Gastpar M 2006 On the performance of independently designed LDPC codes for the relay channel, 2006 IEEE International Symposium on Information Theory, pp. 977– 981.

Fettweis G, Holfeld J, Kotzsch V, Marsch P, Ohlmer E, Rong Z and Rost P 2010 Field trial results for LTE-Advanced concepts. Proc. ICASSP 2010 – IEEE Int. Conf. Acoust. Speech and Signal Processing, Dallas, TX.

Figuier L 1868 Les Merveilles de la science ou description populaire des inventions modernes vol. 2. Jouvet Furne, Paris.

Hammerström I, Kuhn M and Wittneben A 2004 Cooperative diversity by relay phase rotations in block fading environments, Proc. SPAWC 2004 – Sig. Proc. Advances in Wireless Commun., pp. 293– 297.

Han T and Kobayashi K 1981 A new achievable rate region for the interference channel. IEEE Trans. Inform. Theory IT-27(1), 49–60.

Hu J and Duman T 2006 Low density parity check codes over half-duplex relay channels, 2006 IEEE International Symposium on Information Theory pp. 972– 976.

Hunter T and Nosratinia A 2002 Cooperation diversity through coding, 2002 International Symposium on Information Theory, p. 220.

Hunter T and Nosratinia A 2003 Performance analysis of coded cooperation diversity, Proc. ICC 2003 – IEEE Int. Conf. Commun. vol. 4, pp. 2688– 2692.

IEEE 2007 Draft Standard for Local and Metropolitan Area Networks-Part 16: Air Interface for Fixed and Mobile Broadband Wireless Access Systems – Multi-hop Relay Specification. Technical Specification P802.16j/D1, Task Group IEEE802.16j.

IEEE 2009 IEEE Standard for Local and metropolitan area networks Part 16: Air Interface for Broadband Wireless Access Systems Amendment 1: Multiple Relay Specification. Standard 802.16j-2009, Task Group IEEE802.16j, http://standards.ieee.org/getieee802/download/802.16j-2009.pdf.

IEEE 2010 IEEE Draft Amendment Standard for Local and Metropolitan Area Networks - Part 16: Air Interface for Broadband Wireless Access Systems – Advanced Air Interface. Standard P802.16m/D8.

Joham M, Utschik W and Nossek J 2005 Linear transmit processing in MIMO communications systems. IEEE Trans. Signal Processing (8), 2700–2712.

Knisely D, Yoshizawa T and Favichia F 2009 Standardization of femtocells in 3 GGP. IEEE Commun. Mag. IEEE 47(9), 68–75.

Kramer G, Gastpar M and Gupta P 2005 Cooperative strategies and capacity theorems for relay networks. IEEE Trans. Inform. Theory 51(9), 3037–3063.

Laneman JN 2002 Cooperative Diversity in Wireless Networks: Algorithms and Architectures PhD thesis Massachusetts Institute of Technology Cambridge, MA.

Laneman J, Tse D and Wornell G 2004 Cooperative diversity in wireless networks: Efficient protocols and outage behavior. IEEE Trans. Inform. Theory 50(12), 3062–3080.

Laneman J and Wornell G 2002 Distributed space-time coded protocols for exploiting cooperative diversity in wireless networks, vol. 1, pp. 77– 81.

Larsson P 2003 Large-scale cooperative relay network with optimal coherent combining under aggregate relay power constraints Proc.. Future Telecommunications Conference, pp. 166– 170, Beijing, China.

Marsch P and Fettweis G 2007 A framework for optimizing the uplink performance of distributed antenna systems under a constrained backhaul. Proc. ICC 2007 – IEEE Int. Conf. Commun. pp. 975– 979, Glasgow, Scotland.

Nahas M, Saadani A and Rekaya G 2010 New delay-tolerant code for distributed antenna IEEE Personal, Indoor and Mobile Radio Conferencee, Istanbul, Turkey.

Osseiran A, Logothetis A, Ben Slimane S and Larsson P 2007 Relay cyclic delay diversity: Modeling and system performance. IEEE International Conference on Signal Processing and Communication (ICSPC07), Dubai, UAE.

Pabst R, Walke B, Schultz D, Herhold P, Yanikomeroglu H, Mukherjee S, Viswanathan H, Lott M, Zirwas W, Dohler M, Aghvami H, Falconer D and Fettweis G 2004 Relay-Based Deployment Concepts for Wireless and Mobile Broadband Radio. Communications Magazine, IEEE 42(9), 80–89.

Razaghi P and Yu W 2007 Bilayer low-density parity-check codes for decode-and-forward in relay channels. Information Theory, IEEE Transactions on 53(10), 3723–3739.

Rost P 2009 Opportunities, Benefits, and Constraints of Relaying in Mobile Communication Systems PhD thesis Technische Universitt Dresden Dresden, Germany.

Rost P and Fettweis G 2007 Space-time trellis coding exploiting superimposed transmissions in half-duplex relay networks. VTC 2007 Fall – IEEE 66th Vehicular Technology Conf., Baltimore, USA.

Rost P and Fettweis G. In press. Protocols and performance limits for half-duplex relay networks submitted to IEEE Transactions on Communications.

Rost P, Fettweis G and Laneman J 2009 Opportunities, constraints, and benefits of relaying in the presence of interference, Proc. ICC 2009 – IEEE Int. Conf. Commun., Dresden, Germany.

Rost P, Fettweis G and Laneman J. In press. Energy and cost efficient mobile communication using multi-cell MIMO and relaying submitted to IEEE Transactions on Wireless Communications.

Scaglione A and Hong Y 2003 Opportunistic large arrays: cooperative transmission in wireless multihop ad hoc networks to reach far distances. IEEE Trans. SP '03 51(8), 2082–2092.

Somekh O, Simeone O, Poor V and Shamai S 2010 Cellular systems with non-regenerative relaying and cooperative base stations. IEEE Trans. Wireless Commun. 9(8), 2654–2663.

Stefanov A and E Erkip 2004 Cooperative Coding for Wireless Networks. IEEE Trans. Commun. 52(9), 1470–1476.

Wei S, Goeckel D and Valenti M 2004 Asynchronous cooperative diversity Conference on Information Sciences and Systems.

WINNER+ 2010 Celtic Project CP5-026 Final Innovation Report (ed. Svensson T. and Zinovief E.). Public Deliverable D1.9, Wireless World Initiative New Radio – WINNER+, April 2010, http://projects.celtic-initiative.org/winner+/index.html.

WINNER-II 2006 IST-4-027756 Test Scenarios and Calibration Issue 2 (ed. Döttling M.). Public Deliverable D6.13.7, Wireless World Initiative New Radio – WINNER, 2006, http://projects.celtic-initiative.org/winner+/index.html.

Wu M, Weitkemper P, Wubben D and Kammeyer KD 2010 Comparison of distributed LDPC coding schemes for decode-and-forward relay channels. International ITG Workshop on Smart Antennas (WSA), Bremen, Germany, pp. 127– 134.

Xie LL and Kumar P 2004 A network information theory for wireless communication: Scaling laws and optimal operation. IEEE Trans. Inform. Theory 50(5), 748–767.

Xie LL and Kumar P 2005 An achievable rate for the multiple-level relay channel. IEEE Trans. Inform. Theory 51(4), 1348–1358.

Zhao B and Valenti M 2003 Distributed turbo coded diversity for relay channel. Electronics Letters 39(10), 786–787.