Joint Cooperative Shared Relaying and Multipoint Coordination for Network MIMO in 3GPP LTE-Advanced Multihop Cellular Networks

LTE-Advanced (Parkvall et al., 2011; 3GPP TR36.814) is the successor of LTE (Long Term Evolution), which is specified by the Third Generation Partnership Project (3GPP). LTEAdvanced can provide downlink and uplink peak rates up to 1 Gb/s and 500 Mb/s, respectively, in 100 MHz of bandwidth. Similar to its predecessor, LTE-Advanced is an Orthogonal Frequency Division Multiplexing (OFDM)-based radio access technology, with conventional OFDM on the downlink and Discrete Fourier Transform Spread OFDM (DFTSOFDM) in the uplink. In addition, LTE-Advanced includes several new key technological components, namely carrier aggregation, enhanced MIMO (Multiple-Input Multiple Output), Coordinated MultiPoint transmission and reception (CoMP), and relaying. In this chapter, we focus on the CoMP component and the relaying component. CoMP is a means of coordinating the transmission and reception of data from/to a single mobile terminal using several geographically distributed base stations. Essentially, CoMP eliminates inter-cell interference by effectively having the multiple base stations act as a single transceiver. CoMP is especially effective for improving data rates of cell-edge mobile terminals where performance is degraded due to inter-cell interference since LTEAdvanced uses full-frequency reuse. Relaying is employed as a low-cost solution to enhance cell-coverage and -capacity. With relaying, the mobile terminal communicates with the base station via a relay node that is wirelessly connected to the base station using the same radio resources as for the mobile terminal that is directly connected to the base station. An LTE-Advanced relay node is divided into a transparent type and a nontransparent type. The main difference between the two types is in the amount of functionality and intelligence included in the relay node. A non-transparent relay node has more functionality and intelligence than a transparent one. This means it is also costlier. The simplest and the cheapest transparent relay is known as Amplify and Forward (AF). The aim of this chapter is to leverage and combine the benefits of CoMP and relaying in order to improve the performance of cell-edge users, which is severely degraded due intercell interference. The joint relaying and CoMP technique can yield performance gains

Joint Cooperative Shared Relaying and Multipoint Coordination for Network MIMO in 3GPP LTE-Advanced Multihop Cellular Networks 219 out-band RN while the latter as in-band RN. An in-band RN can have higher spectral efficiency than out-band RNs. However, it suffers from self-interference. Therefore, adequate isolation must be achieved between the transmitting part of one link and the receiving part of the other link. A half-duplex RN always transmits and receives on orthogonal channels typically in the time domain.

Coordinated Multipoint transmission and reception (CoMP)
LTE-Advanced uses full frequency reuse, which in turn leads to inter-cell interference. CoMP aims at mitigating the inter-cell interference and hence improves spectral efficiency of cell-edge users. Fig. 2 shows the CoMP architecture. The same spectrum resources are used in all cells, leading to interference for UEs at the edge between the cells, where signals from multiple eNBs are received with similar signal power in the downlink. Multiple eNBs can cooperate to mitigate the inter-cell interference. The eNBs are interconnected by the interface X2. Physically, this could be a direct fast fibre link. CoMP can be applied both in the uplink and the downlink. In the downlink, CoMP can be divided into two schemes:  Joint Processing: user data to be transmitted to a single UE is available at each transmission, i.e., eNB.  Coordinated scheduling and beamforming: user data is always transmitted from one eNB only, but user scheduling and beamforming decisions are made with coordination among cells. In the CoMP uplink, multipoint reception implies coordination among multiple, geographically distributed eNBs. Uplink CoMP reception involves joint reception of the transmitted signal at multiple reception points and/or coordinated scheduling decisions among cells to control interference.

Joint cooperative shared relaying and Coordinated Multipoint transmission/reception
In this section, we leverage the benefits of the CoMP and the relaying techniques. The aim of the combined technique is to increase the cell-edge user data rates beyond what can be achieved by using either one of the techniques alone. Unlike users that are close to the eNB, cell-edge users experience lower signal strength because of the distance from the eNB and higher interference levels due to neighbouring eNBs. Furthermore, increasing transmission power does not necessarily lead to higher data rates due to an increase in inter-cell interference level. With the combined technique, an RN is employed to enhance the signal strength and CoMP to mitigate inter-cell interference due to neighbouring eNBs. The combined technique places an RN at the intersection of two or more cells. We propose a fullduplex AF RN for its high spectral efficiency. In the downlink, the RN amplifies and retransmits the received signals from the intersecting eNBs to multiple users. In the uplink, the RN amplifies and retransmits the received signals from multiple users to all intersecting eNBs. In either uplink or downlink, the destination (namely, UE in the downlink and eNB in the uplink) combines the relayed signal and the direct signals from all the sources (namely, eNB in the downlink and UE in the uplink). The RN and the sources form a network MIMO system. Such an RN is referred to as cooperative relay.

System model
Our system model considers an arbitrary hexagonal cellular network. The eNbs are located in the centre of each cell. The eNBs are grouped into L clusters. A cluster is composed of a single RN that is shared by M eNBs and K UEs. The eNB, the RN and the UE are equipped with one antenna element. In the downlink, each of the M eNBs transmits data streams of K UEs at the same time-frequency resources. In the uplink, the K UEs transmit to the M eNBs using the same time-frequency resources.  Fig. 3 shows a typical cluster for joint cooperative shared relay and CoMP under the hexagonal cellular model. The relay RN which belongs to the lth cluster is placed at the corner of three adjacent cells. Thus, for this configuration, M = 3 and K = 1. In subsections 3.1.1 and 3.1.2, we derive the capacity equations for the downlink and uplink, respectively. The rest of the chapter uses the following notation. Bold uppercase letters A denote matrices, bold lowercase letters a denote column vectors, and italic lowercase letters a denote scalars. I denotes the identity matrix, A T is the transpose of matrix A, A* is the Hermitian transpose of matrix A, ε is expectation operator, min{x, y} is the minimum of x and y, |z| is the modulus of complex number z, and ||h|| is the Euclidean norm of vector h.

Downlink capacity
The signal received by RN is given by ,, , where eNB x is similar to the transmitted signal vector of Equation (1), RL h is the RL channel vector as in Equation (1) For multiuser transmission and reception, the transmitted signal vector eNB x in Equation (5) is generated by a weighted linear combination of data symbols contained in a K × 1 vector where d k is the specific data symbol intended for the kth UE. Thus, the kth UE receives its own symbols as well as the other users' symbols. If the channel state information is perfectly known by each eNB then the zero-forcing method can be employed to nullify the undesired symbols. Using the zero-forcing method, the vector eNB x can be generated at the eNB by precoding d with an M × K weight matrix as shown in Equation (6).
where 11 12 1 The weight mk  represents the precoding coefficient allocated by the mth eNB for the kth UE. The precoding matrix W can be simply obtained by inverting the channel matrix H . The capacity of the network MIMO system of K UEs is the sum of the capacity of each UE which can be obtained from Equation (5) where k w denotes the kth column of W for the kth UE. The zero-forcing method ensures that interference due to multiuser is cancelled, i.e., ,,

Uplink capacity
The signal received by the RN is given by ,, , is the additive white Gaussian noise observed at RN which is identical to that in Equation (1). The signal received at the mth eNB is a superposition of all the signals transmitted by all UEs and the RN, which is given by ,, where UE x is similar to the transmitted signal vector of Equation (10), AL h is the AL channel vector as in Equation (10) where eNB x is given by Equation (6)

Numerical results
In this section, the performance of the proposed joint scheme is evaluated by using Monte Carlo simulations. The parameters of the simulation are given in Table 1. The simulated cluster is composed of two cells. Only one UE was located in each cell. For the AL and DL links, we used the WINNER II C2 -Typical Urban Macro-cell Environment with Non Lineof-Sight (NLOS) channel model (Kyosti et al., 2007) Feeder model (Kyosti et al., 2007) was used for the RL link. For the latter, a strong LOS signal is assumed. The assumption is valid because the position of the RN is fixed and it can be placed in such a way that a strong LOS signal is achieved. The RL channel is almost like in free space. Thus, the path loss does not depend on the antenna heights. The height of the eNB is similar to the RN is similar, which was set to 25 m in the simulation. The height of the UEs was set to 1.5 m above the ground. The multi-path fading is a Rayleigh distribution. The carrier frequency was set to 2 GHz. The total maximum transmit power of eNB was set to 17 dBW. For simplicity, we assume the power is equally distributed between the two eNBs. The total maximum UE transmit power was set to 5 dBW. Similar to eNB, the power is equally distributed between the two UEs. The noise power ( 22 2 ,, , ,a n d UE k RN l eNB m

 
) was set to -144 dBW. This noise power level corresponds to a 10-MHz channel. Fig. 4 shows the downlink sum-rate capacity for the joint scheme and CoMP. The sum-rate capacity is the rate of the two UEs in the cells averaged over 1000 iterations. The joint scheme outperforms the CoMP by more than 4 b/s/Hz. The sum-rates of both schemes increase with higher eNB transmit power because of the higher Signal-to-Noise Ratio (SNR). Uplink sum-rates are given in Fig. 5. As in the downlink case, the performance of the joint scheme is superior to CoMP. The achievable performance gain of the joint scheme is more than 7 b/s/Hz. The performance in the uplink is higher than in the downlink. This is because of the short-range connection between the UE and the RN. The signal received by the RN has very high SNR, which is then amplified and relayed to the eNB through a high quality channel. Fig. 6 shows the downlink sum-rate for the joint scheme and the CoMP as a function of the UE distance from the RN. Both eNBs were set to transmit at the maximum power. Clearly, joint scheme outperforms CoMP. An interesting observation is made when both UEs are distanced from the RN, the joint scheme delivers a significant performance gain as compared with CoMP. The effect of the RN on the SNR at the receiver is clearly evidenced. The joint scheme benefits from the RN amplification gain and the low noise level of the RL link. Fig. 7 shows the uplink sum-rate for the same techniques. Unlike in the downlink case, the performance of the joint scheme in the uplink degrades as both UEs are distanced from the RN which leads to low SNR of the AL link. The performance degradation of the joint scheme is attributed to the amplified noise received at the RN through the AL link. This noise amplification offsets any gain resulting from using the RN. Thus, CoMP yields superior performance. However, the performance of the joint scheme is still better when the UEs are near the RN. Figs. 8 and 9 show the downlink and uplink sum-rate capacity as a function of relay transmit power, respectively. The sum-rates of the joint scheme are higher than CoMP. The plots of CoMP are constant because no relays are included in its model. Increasing the transmission power of the RN, the sum-rate is increased by an approximately 0.5 b/s/Hz. The joint scheme gives a roughly 60% and 200% capacity increase relative to CoMP in the downlink and uplink, respectively. The capacity gain in the uplink is higher than in the downlink because of the short-range connection between the UE and the RN which leads to high SNR. Thus, the noise amplification by the RN is minimal as compared with the downlink transmission which has longer range. The joint scheme benefits from both spatial diversity gain and amplification, which are provided by the RN.

Conclusion
This chapter has proposed a joint cooperative shared relaying and CoMP technique to improve the performance of cell-edge users in LTE-Advanced multi-hop cellular networks. The performance of cell-edge users is severely degraded as a consequent of cochannel interference due to full-frequency reuse. The proposed joint technique considers a full-duplex amplify-and-forward relay due to its high spectral efficiency and low latency. The joint technique supports multi-user transmissions in order to increase system throughput. In the downlink, the relay node amplifies, combines and retransmits the received signals from the intersecting eNodeBs to multiple users. The zero-forcing method was used to nullify the interuser interference at the destination. In the uplink, the relay node amplifies combines and retransmits the received signals from multiple users to all intersecting eNodeBs. In both downlink and uplink, the destination combines the relayed signal and the direct signal from all the sources in order to attain cooperative diversity. We have derived the capacity equations for the joint technique and the CoMP as a baseline for comparison. Numerical results show that the performance of the joint technique is superior to CoMP in both uplink and downlink by at least a factor of three.
The results also indicate that the channel quality of the relay link has a strong impact on the downlink performance than in the uplink. Therefore, it is crucial to deploy the relay node in a location that gives the best relay link quality.