MULTIHOP CELLULAR NETWORKS TOWARD LTE-ADVANCED. Its Architecture and Performance

Transcription

1 PHOTO F/X2 MULTIHOP CELLULAR NETWORKS TOWARD LTE-ADVANCED Its Architecture and Performance Kan Zheng, Bin Fan, Zhangchao Ma, Guangyi Liu, Xiaodong Shen, and Wenbo Wang The multihop cellular architecture with relaying is expected to be a cost-effective solution to reduce the transmission distance and increase the amount of users under more favorable channel conditions with better channel quality and higher throughput. In this Digital Object Identifier /MVT article, we present the architecture and evaluate the performances of multihop cellular networks (MCNs) for thirdgeneration (3G) long-term evolution (LTE) toward its advanced networks. First, the network architecture is introduced with the specific features of 3G LTE. Then, the frame structure based on time-division duplex (TDD) for relaying is proposed, and the relay nodes (RNs) are /09/$ IEEE IEEE VEHICULAR TECHNOLOGY MAGAZINE SEPTEMBER 2009

2 classified into different types according to the layer that is used to forward the user-plane traffic data. Next, the essential radio resource management (RRM) functions in medium-access control (MAC) layer, which handle the radio resources efficiently, are described in detail. Finally, the dynamic system-level simulations are carried out to demonstrate the effectiveness of relaying techniques in MCN toward 3G LTE-Advanced. Future broadband wireless systems, announced by International Telecommunication Union Radiocommunications Sector (ITU-R) officially as International Mobile Telecommunications (IMT)-Advanced, are expected to support enhanced peak data rates in the order of 100 Mb/s for high-mobility and 1 Gb/s for low-mobility environments, respectively. Thus, more spectrum bands are needed to meet these requirements. In addition to the existing spectrum for the 3G mobile communication systems, more spectrum bands located at MHz, MHz, GHz, and GHz have been identified for 3G and IMT-Advanced systems. Most of these spectrum bands are above 2-GHz band, where the radio propagation is more vulnerable to nonfavorable channel conditions. Under the condition of traditional cellular architectures, the density of base station has to be significantly increased to meet the service-coverage requirements, offering high-data rates at these high-frequency bands. Obviously, the increasing density of base stations means high-deployment costs. Instead, a cost-effective solution would be the multihop cellular architecture with relaying, which reduces the transmission distance and increases the amount of users under more favorable channel conditions, allowing for better channel quality and higher throughput [1]. Standardization efforts integrating multihop relaying technologies to future cellular networks like LTE- Advanced have commenced recently [2]. Several contributions have discussed relaying techniques and performance evaluations [3]. Meanwhile, many research focused on RRM schemes, such as relay deployment, radio resource partition, and link adaptation [4], [5]. However, most of these studies do not solve the problems from the systems point of view so that somehow lack practical feasibility. Architecture Overview MCN As shown in Figure 1, a typical multihop relaying network toward LTE-Advanced consists of mobility management entity (MME)/system architecture evolution (SAE) gateway, evolutional NodeB (enb), RN, and user equipment (UE). The enbs, which provide the evolved universal terrestrial radio-access network (E-UTRAN) user-plane and control-plane terminations toward the UE, connect to the MME/SAE gateway by the S1 interface through a manyto-many relationship. A new interface called X2 logically RELAYING TECHNIQUES ARE EXPECTED TO IMPROVE THE COVERAGE OF HIGH DATA RATES, GROUP MOBILITY, TEMPORARY NETWORK DEPLOYMENT, THE CELL-EDGE THROUGHPUT, AND/OR TO PROVIDE THE COVERAGE IN NEW AREAS. connects the neighboring enbs, enabling direct communication between the elements and eliminating the need to tunnel data back and forth through the radio network controller (RNC). The MME handles control signaling, e.g., for mobility. User data are forwarded between enbs and gateway nodes over an IP-based transport infrastructure. There are usually three radio links that are involved in the end-to-end (e2e) communication in multihop relaying cellular networks, i.e., between enb and RN (enb M RN), RN and UE (RN M UE), and enb and UE (enb M UE). Furthermore, for the sake of clarity, we refer to enb M RN link as relay link, while both RN M UE and enb M UE links are called access links. The communication between UE and enb can be established either directly (i.e., one hop) or over two hop via RN. Relaying should happen only when it can improve the e2e throughput or enhance the coverage. Classification of RN Relaying techniques are expected to improve the coverage of high data rates, group mobility, temporary network deployment, the cell-edge throughput, and/or to provide the coverage in new areas. The radio protocol architecture of E-UTRAN is specified for the user and control planes. The user plane comprises the packet data-convergence protocol (PDCP), radio link control (RLC), MAC, and physical (PHY) layer; the control plane performs the radio resource control (RRC). Both the user and control planes are terminated in the enb. The relays can be categorized RN MME/SAE Gateway S1 X2 RN S1 enb X2 enb enb X2 RN S1 S1 RN FIGURE 1 Architecture of multihop cooperative networks. MME/SAE Gateway SEPTEMBER 2009 IEEE VEHICULAR TECHNOLOGY MAGAZINE 41

3 LINK ADAPTATION IS A PROMISING METHOD TO IMPROVE THE SYSTEM THROUGHPUT PERFORMANCE BY MAKING USE OF THE INSTANTANEOUS CHANNEL QUALITY IN VARYING CHANNEL CONDITIONS. into different types depending on which layer is used to forward the user-plane traffic data [3]. L1 RN A layer-1 (L1) RN, also called advanced repeater, amplifies and forwards the received signal from the source at the PHY layer (i.e., L1). It typically introduces very little delay compared with other multihop solutions operating on higher layers. However, L1 RN cannot differentiate between received desired signals and noise/interference, so that the noise/interference may be over-amplified through forwarding. L2 RN A layer-2 (L2) RN is characterized by forwarding its userplane traffic on L2. Different cooperative protocols, such as decode and forward (DF), can be implemented at L2 RN. The noise is not forwarded by the RN, and rate of adaptation may be performed individually for the relay and access links. The multiplex/demultiplex of MAC service data unit (SDU) and priority handling between RN and UE can be carried out for further transmission. However, a larger delay is introduced compared with L1 RN. The radio resource allocation/scheduling between UE and RN can be performed in coordination with enb, enabling the possibility of taking into account other RNs. In addition, the outer automatic repeat request (ARQ) and RLC protocol data unit (PDU) segmentation/concatenation might be resided in RN. L3 RN The layer-3 (L3) RN forwards the user-plane traffic data packet on the IP layer. It does not require significant modification to the radio protocol, allowing for reuse of existing equipments. This is because L3 RN has its own unique physical-layer cell ID, and there is no difference in accessing it from a UE perspective. However, some enhancements on the air interface have to be adopted to achieve higher throughput. Frame Structure In Figure 2, a frame structure is presented as one possible solution to enable relaying technology in TDD orthogonal frequency division multiplexing (OFDM)-based cellular systems toward LTE-Advanced. Each radio frame of length T f ¼ 10 ms consists of two half frames with length T half ¼ 5 ms each. In TDD systems, downlink and uplink communication alternates periodically at the same frequency band, requiring a guard period (GP) between them. Usually, each half frame includes four common subframes of length T sub ¼ 1 ms and three special fields, i.e., downlink pilot time slot (DwPTS), uplink pilot time slot (UpPTS), and GP. Each subframe comprises two slots with length One Radio Frame, T f = 307,200T s = 10 ms One Half - Frame, 153,600T s = 5 ms One Slot, T slot = 15,360T s 30,720T s Subframe #0 Subframe #2 Subframe #3 Subframe #4 Subframe #5 Subframe # DL Relay UL Access UL Relay DL Access DL Relay Zone DwPTS GP Zone UpPTS Zone Zone Zone DwPTS GP UpPTS DL Access Zone 7 OFDM Symbols (0.5 ms) FIGURE 2 TDD frame structure of multihop cellular networks. 42 IEEE VEHICULAR TECHNOLOGY MAGAZINE SEPTEMBER 2009

4 T slot ¼ 0:5 ms. One slot includes seven OFDM symbols. Within each slot, reference pilot symbols are located in the first and fifth OFDM symbols. This structure allows low-complexity and high-performance channel-estimation techniques, such as discrete Fourier transform (DFT)- based channel estimators. Downlink control signaling can be put in the first three OFDM symbols. The basic radio resource unit for OFDM transmission can be described as a two-dimensional (2-D) time frequency grid that corresponds to a set of OFDM symbols and subcarriers in the time and frequency domains. In LTE-Advanced, the basic unit for data transmission is a pair of resource blocks (RBs) that correspond to a 180-kHz bandwidth during a 1-ms subframe. Therefore, by aggregating frequency resources and adjusting transmission parameters, such as modulation order and channel code rate, one can flexibly support a wide range of data rates. TDD relaying is assumed in this article, which means that the relay link is scheduled in a different subframe from the access link. Then, in this frame structure, there are four kinds of subframes for transmission, i.e., relay and access zone in the downlink and uplink transmission, respectively. One subframe is used as the basic transmission granularity in the time domain. RRM in MCN All the radio resources in the networks with L1 RN are centrally managed at the enb. L2 RN and L3 RN have their own RRM functions so that the UEs located within the RN coverage could be managed by the RN itself. In this section, several important RRM functions in the MCN are described in detail, which can be implemented at the enb or RN with RRM functions. Downlink transmission is assumed in this article for the sake of description. Routing In the MCN, the purpose of routing is to find a proper access method for each user. Similar to the traditional single-hop cellular networks, the access of each user is based on the criterion of maximizing received power or signal-tonoise and interference ratio (SINR). However, both the routing criterions mentioned earlier have not taken into account the characteristics of relay transmission. Since two-hop transmission may cost more radio resources, it is more reasonable to use the transmission efficiency as the routing criterion. The transmission efficiency of using one-hop and two-hop transmission can be estimated from the mean SINR values of the possible transmission links. Then, each UE will connect to the enb directly or the aid of RN, depending on the estimated transmission efficiency. Resource Partition Schemes In the LTE-Advanced cellular networks, the whole radio resource is usually divided into the RBs along the time IN LTE-ADVANCED, THE BASIC UNIT FOR DATA TRANSMISSION IS A PAIR OF RESOURCE BLOCKS THAT CORRESPOND TO A 180-KHZ BANDWIDTH DURING A 1-MS SUBFRAME. and/or frequency domain. Different users share all the radio resources by the scheduling algorithms in the networks. Moreover, by introducing the RNs, the resources have to be shared or reused between the different kinds of links. Therefore, it is quite essential to design efficient resource partition schemes in the MCN. In addition, a good resource partition scheme may facilitate the scheduling algorithm and interference management. Here, two schemes are described and explained as follows. Orthogonal Resource Partition In the resource partition scheme [shown in Figure 3(a)], the radio resources allocated to the enb fi RN and enb fi UE are orthogonal in relay zone; thus, no intracell interference occurred. Moreover, in access zone, there is also no overlap between the resources allocated to the enb fi UE and RN fi UE links. With orthogonal radio resource partition, the division of the radio resources to different links is critical, which confines the efficiency of the resource usage. To ensure the fairness between the one-hop and two-hop users, the resource allocation should take into account their traffic load and distribution. On the assumption of uniform traffic load and even user distribution over the coverage, the ratio of the resources allocated to one-hop and two-hop users equals to that of their user number. Moreover, the Frequency Access Zone Relay Zone enb S UE enb S UE RN S UE enb S RN Frequency Access Zone (a) (b) Relay Zone enb S UE enb S UE RN S UE enb S RN FIGURE 3 Illustration of resource partition schemes in MCN: (a) orthogonal and (b) reuse. Time Time SEPTEMBER 2009 IEEE VEHICULAR TECHNOLOGY MAGAZINE 43

5 THE SYSTEM-LEVEL SIMULATIONS ARE CARRIED OUT TO EVALUATE THE DOWNLINK PERFORMANCES OF MCNS TOWARD LTE-ADVANCED. data rate of the enb fi RN and RN fi UE links should be equal to maintain efficient resource usage. Resource Reuse Partition In this scheme, the radio resources is reused by the enb fi UE and RN fi UE links in access zone, while the radio resource is still orthogonally allocated to the enb fi UE and enb fi RN links. An illustration of this scheme in the LTE-Advanced networks is shown in Figure 3(b). Similar to the orthogonal resource partition scheme, the available resource units are divided into different parts for three kinds of links. With full resource reuse in access zone, the whole band is allocated to both the enb fi UE and RN fi UE links. With the constraint that the transmission rate for the two-hop users in the access and relay zones should be equal, the resource allocated for the enb fi RN link can be determined. In addition, in some specified scenarios (e.g., one RN per sector), the number of one-hop users is larger than that of the two-hop users. Thus, more resources are required for the enb fi UE link. So, parts of the radio resources are reserved for enb fi UE link only in the access zone, which is referred to as partial resource reuse and not discussed in this article. Link Adaptation Link adaptation is a promising method to improve the system throughput performance by using instantaneous channel quality in varying channel conditions. Adaptive modulation and coding (AMC) and hybrid ARQ (HARQ) are two important link adaptation techniques in MCN. AMC In essence, the AMC technique selects the best modulation and coding scheme (MCS) according to current channel conditions. This, in turn, determines the data rate or error probabilities on each link. Depending on the network structure and the functions in RNs, AMC in MCN can be classified into two methods. 1) Common MCS scheme for both hops: forthel1rn with no or very limited RRM functions, this scheme should be adopted in different hops based on the knowledge of the channel qualities of both hops. 2) Individual MCS schemes for each hop: when the L2 or L3 RN with RRM functions is deployed, the resource usage of the enb fi RN and RN fi UE links can be decided independently. Then, different MCS schemes can be selected for each hop to achieve the highspectrum efficiency. Usually, the throughput-oriented principle can be applied in both of the aforementioned methods. This principle selects the MCS mode to maximize the instantaneous system throughput while maintaining the required block error rate (BLER). HARQ The HARQ technique gives the receiver redundancy information that enables it to avoid a certain amount of errors. According to the way in which they recover from errors, most of the existing HARQ protocols in the cooperative relaying networks can be categorized into e2e and hop-byhop protocol. With the e2e HARQ protocol, the RN just forwards the data and signaling feedback between the source and destination nodes without taking any additional action. All the transmission procedures are managed by the source and destination nodes without the aid of the relay. Let us take downlink transmission as an example. The RN normalizes only the received signal from the enb and retransmits the signal to the UE without any additional processing in the first transmission round. When retransmission occurred, the RN combines the current signal with those transmitted in the previous transmission rounds using maximal ratio combining (MRC) before forwarding. At the end of each transmission round, only the UE can decode the packet and detect the error through the cyclic redundancy check (CRC). Then, the transmission result is broadcasted from the UE and forwarded by the RN through a 1-b acknowledge (ACK) or non-ack (NACK) message. The NACK/ACK is assumed to be received error free and with negligible delay. As long as NACK is received after each HARQ round and the maximum number of HARQ rounds is not reached, the enb successively retransmits a packet corresponding to the same information bits. Otherwise, the subsequent packets are transmitted by the source. The e2e ARQ mechanism is very simple and deals with handover easily, because the source node knows the status of transmitted HARQ blocks. However, it also has many drawbacks such as low-transmission efficiency and long-transmission delay. With the hop-by-hop HARQ protocol in the downlink transmission, the RN not only forwards the data and feedback between the source and destination nodes but also generates its own feedback and transmits it to the source. For example, in downlink transmission, when the retransmission in each hop occurred, the receiver (at the RN or UE) combines the signals from the current and previous transmissions by MRC before decoding. At the end of transmission per hop, the receiver (i.e., at the RN or UE) decodes the packet and detects the error through the CRC. Then, the transmission result is feedback through a 44 IEEE VEHICULAR TECHNOLOGY MAGAZINE SEPTEMBER 2009

6 1-b ACK or NACK message to the RN or enb. It has a hightransmission efficiency and short round-trip delay, because any transmission failure can be recovered in each hop. However, the RN with RRM functions such as L2/L3 RN is necessary to implement the hop-by-hop HARQ protocol. Since the data transmission of different links is independent by each other, the RN should be able to buffer and queue the received data from the first-hop transmission. In addition, in case of UE handover, it must report additional information about its block status to the enb, because the HARQ block status in the enb and the RN are managed separately. Scheduling To simultaneously realize gains from both multiuser diversity and multihop relaying to enhance capacity and coverage, OFDM-based resource scheduling algorithms for cellular multihop networks have gained attention recently. Since the network architecture becomes more complex with the introduction of RNs, there are more than one control mechanisms of radio resources in MCN. Depending on the control mechanism, resource scheduling algorithms can be categorized as centralized and distributed ones. Centralized Scheduling With the centralized scheduling, the enb is responsible for the resource allocation of all the links for not only one-hop but also two-hop users and determines the transmission mode of all the resources, e.g., MCS selection, HARQ control, and power allocation. The RN only forwards the received data and signaling without any scheduling function. To perform the global management, the channel-state information (CSI) of all the links including the RN fi UE link are assumed to be known ideally at the enb. Besides, the list of all the one-hop and two-hop users is kept at the enb. Then, the optimal resource usage in MCNs is possibly achieved with the welldesigned centralized scheduling algorithms. However, it is necessary to feedback full or partial CSI of all the links to the enb, which causes huge overhead in the networks. Moreover, to satisfy the latency requirements, the fast backhaul transmission from RN to enb is required to forward the CSI of RN fi UE links to the enb. In practical, simplified centralized scheduling algorithms with limited CSI are feasible for L1 RN, L2 RN, or L3 RN. Distributed Scheduling The RRM of each hop is carried on at the enb and the RN independently when the distributed scheduling algorithms are applied. For the enb, it only schedules the radio resources for the one-hop users and the backhaul transmission between enb and RN. In this case, the powerful RN with its own RRM functions is required. Since the RN can buffer and rearrange the received A LAYER-1 RN AMPLIFIES AND FORWARDS THE RECEIVED SIGNAL FROM THE SOURCE AT THE PHY LAYER. packets, it makes decisions on the resource allocation of the two-hop users and generates its own signaling. In this way, the RN can decide how to transmit the signals in the RN fi UE links. In the networks with the distributed scheduling algorithms, the CSIs of the enb fi UE links and RN fi UE links are only feedback to enb and RN, respectively. There is no need for the RN to feedback the RN fi UE channel state to the enb, which means that the feedback overhead is greatly reduced. However, distributed scheduling algorithms cannot achieve the optimal resource allocation because no central node can control and coordinate the resource usage among different links efficiently. Instead, the suboptimal distributed resource allocation solution can be obtained with little overhead and infrequent control signaling exchange in relay-aided LTE- Advanced networks. Since there are requirements of RRM functions in the RN, only the networks with L2 or L3 RN can use the distributed scheduling algorithms. Simulation Results and Analysis In this section, the system-level simulations are carried out to evaluate the downlink performances of MCNs toward LTE-Advanced. Most of the simulation assumption follows the evaluation methodology as in [6] and [7]. The ideal hexagonal cell is assumed for each enb, and two tiers of cells are considered with respect to a reference cell in the center, i.e., a total of 19 hexagonal cells are assumed. Moreover, each cell is partitioned into three 120 sectors. For each sector, only one RN with RRM functions is deployed on the boresight line of the directional antenna of the enb. The individual MCS selection of each hop is performed, and hop-by-hop HARQ is adopted in our simulations. Because of the low-cost requirements, only omnidirectional antenna is assumed at RN, and the relaying/ forwarding is performed in the time domain. The UEs are evenly distributed in circular areas around each enb. The accurate level of interference from all other cells cannot be captured in the simulation model because of the finite number of cells. To remove such boundary effects, the socalled wrap-around structure is used to generate the more accurate level of intercell interference. Figure 4 shows the cumulative distribution function (CDF) curves of the received SINR for different resource partition schemes. In the simulation, DF relaying strategy is assumed for the RN with RRM functions. For the UEs connected directly to the enb, i.e., one-hop communication, the SINR values of the enb fi UE links are directly SEPTEMBER 2009 IEEE VEHICULAR TECHNOLOGY MAGAZINE 45

7 IN TDD SYSTEMS, DOWNLINK AND UPLINK COMMUNICATIONS ALTERNATE PERIODICALLY AT THE SAME FREQUENCY BAND. collected for observation. When the UE is connected to the RN with two-hop communications, its measured SINR equals to the minimum of the SINRs of the RN fi UE and enb fi RN links. In case of orthogonal resource partition, no intracell interference between the one-hop and twohop users is induced. So, the SINR performances of the two-hop users are greatly improved without sacrificing the performances of the one-hop users. Therefore, the overall SINR performance of the networks with the orthogonal resource partition scheme quite outperforms that without RNs as shown in Figure 4. On the other hand, when the RNs are deployed with the nonorthogonal resource partition, intracell interference occurred, because the radio resources are reused between the onehop and two-hop users. Then, the improvement of the SINR performance is not as much as that with the orthogonal resource partition. However, it should be noted that the networks with resource reuse partition may achieve higher spectrum efficiency because of the higher frequency reuse factor. Table 1 compares the throughput performances of the MCN with different resource partition schemes. Compared with the network without RN deployment, about 22 and 60% throughput gain are obtained by deploying RNs with the orthogonal resource partition and resource reuse partition, respectively. Moreover, the resource reuse partition scheme can bring higher throughput not only for one-hop but also for two-hop users. In Figure 5, the corresponding CDFs of the per user throughput of these three cases are given. It can be seen that the networks can TABLE 1 Comparison of the throughput performances. Throughput (Mb/s) With RN (Orthogonal) With RN (Reuse) W/O RN One-hop users Two-hop users Total achieve the higher throughput for all the users by introducing RN. Furthermore, compared with the orthogonal resource partition, the network with the resource reuse partition obtains better throughput performance for the high-transmission rate users but worse performance for the users with low-transmission rate. It is because radio resources reuse between the one-hop and two-hop users by the resource reuse partition gives each user opportunities to use more resources but induces intracell interference. Next, Figure 6 shows the throughput performances of the networks with different transmission power at the RN, where only the resource reuse partition scheme is assumed. We can find that much more throughput gain is obtained by RN deployment when the transmission power at the RN is increased from 34 to 37 dbm. However, since the larger interference occurred with higher RN transmission power, the throughput gain becomes less when RN transmission power is further increased to 40 dbm. Meanwhile, with the increase of the RN power, the number of the two-hop users is increased whereas that of the one-hop users is decreased. On the other hand, most of the users with high throughput are through two-hop communication. So, under the resource reuse partition, each two-hop user is possibly to be allocated with less resource with the increase of RN power, which induces the performance degradation in the region of high throughput CDF Without RN With RN (Orthogonal Resource Partition) With RN (Resource Reuse Partition) SINR (db) CDF Without RN With RN (Orthogonal Resource Partition) 0.2 With RN (Resource Reuse Partition) ,000 1,200 Per User Throughput (kb/s) FIGURE 4 CDF of SINR with different resource partition schemes. FIGURE 5 CDF of throughput with different resource partition schemes. 46 IEEE VEHICULAR TECHNOLOGY MAGAZINE SEPTEMBER 2009

8 CDF ,000 Without RN With RN (Tx Power = 30 dbm) With RN (Tx Power = 34 dbm) With RN (Tx Power = 37 dbm) With RN (Tx Power = 40 dbm) 1,200 1,4001,600 Per User Throughput (kb/s) 1,800 2,000 FIGURE 6 Per user throughput with different transmission power. Conclusions This article presents the architecture of MCN toward 3G LTE-Advanced with TDD. On the assumption that L2 RN and L3 RN can provide their own RRM functions, several important RRM functions including routing, link adaptation, and scheduling in MAC layers at RNs can help the enb to handle the radio resources efficiently. In general, the RN with higher implementation complexity has the ability to provide more flexible RRM with more overhead. Our dynamic system-level simulations show that the MCN outperform the networks without relaying in terms of SINR and throughput. Acknowledgments This work was supported, in part, by China Mobile Research Institute, China NSFC under Grant and National Key Technology R&D Program of China under Grant 2008BAH30B11. Author Information Kan Zheng (kzheng@ieee.org) received his B.S., M.S., and Ph.D. degrees from Beijing University of Posts and Telecommunications (BUPT), China, in 1996, 2000, and 2005, respectively. He is a Senior Member of the IEEE. He is currently an associate professor at the BUPT, China. His research interests lie in the field of signal processing for digital communications, with emphasis on PHY/MAC algorithms in wireless cooperative networks. Bin Fan (fanbin84@gmail.com) received the B.S. degree from BUPT in 2005, where he is a Ph.D candidate. His research interests lie in the field of the radio resource management in wireless cooperative communication networks. THE PURPOSE OF ROUTING IS TO FIND A PROPER ACCESS METHOD FOR EACH USER. Zhangchao Ma (mzcroy@gmail.com) received the B.S. degree from BUPT in 2006, where he is a Ph.D candidate. His research interests lie in the field of the radio resource management in wireless cooperative communication networks. Guangyi Liu (liuguangyi@chinamobile.com) received the M.S. and Ph.D. degrees from BUPT in 2000 and 2006, respectively. From April to September 2000, he worked on wideband code division multiple access (WCDMA) in ShangHai Bell-Alcatel, Ltd. From October 2000 to September 2003, he worked on TD-SCDMA in Siemens Ltd. of China. Since October 2006, he has worked on LTE/IMT- Advanced research and standardization in China Mobile. His research interests include relay, CoMP, MIMO channel modeling, and system-level performance evaluation. Xiaodong Shen (shenxiaodong@chinamobile.com) received the M.S. degree from BUPT, Beijing, in He then joined Research Institute of China Mobile Communications Corporations. His research interests include wireless communications with emphasis on radio resource management, wireless relaying, and modeling on system-level simulation. He is now actively participating in LTE and LTE-Advanced standardization and research activities toward 3GPP. Wenbo Wang (wbwang@bupt.edu.cn) received his B.S., M.S., and Ph.D. degrees from BUPT in 1986, 1989, and 1992, respectively. He is currently a professor at the BUPT. His research interests include signal processing, mobile communications, and wireless networks. References [1] D. Schultz, L. Coletti, K. Navaie, M. Wodczak, and P. Rost. (2006, Oct.). WINNER D3.5.1, Relaying concepts and supporting actions in the context of CGs [Online]. Available: [2] Requirements for further advancements for evolved universal terrestrial radio access (E-UTRA) (LTE-advanced) (release 8), 3GPP TR , ( ). [3] Further details and considerations of different types of relays, Prague, Czech Republic, 3GPP TSG-RAN WG1 #54bis, R , Huawei, Sept [4] R. Schoenen, R. Halfmann, and B. H. Walke, MAC performance of a 3GPP-LTE multihop cellular network, in Proc. IEEE Int. Conf. Communications (ICC 08), May 2008, pp [5] F. Boye, P. Rost, and G. Fettweis, Adaptive radio resource management for a cellular system with fixed relay nodes, in Proc. IEEE Personal, Indoor and Mobile Radio Communications (PIMRC 08), Sept. 2008, pp [6] Further advancements for E-UTRA physical layer aspects (release 9), 3GPP TR V0.4.1 ( ). [7] Evaluation Methodology Document (EMD), IEEE Standard m-08/ 004r5, SEPTEMBER 2009 IEEE VEHICULAR TECHNOLOGY MAGAZINE 47