advertisement

Energy and Overhead Aware Adaptive Forwarding Strategy for Multi-Hop Device-to-Device Communications Bleron Klaiqi, Xiaoli Chu, and Jie Zhang Dept. of Electronic and Electrical Engineering, University of Sheffield, Sheffield, S1 3JD, UK Email: {b.klaiqi, x.chu, jie.zhang}@sheffield.ac.uk Abstract—Device-to-device (D2D) communications in cellular networks enable direct transmissions between user equipments (UEs). If the source UE (SUE) and the destination UE (DUE) are far away from each other or the channel between them is too weak for direct transmission, then multi-hop D2D communications, where relay UEs (RUEs) forward the SUE’s data packets to the DUE, can be used. In this paper, we propose an energy-efficient adaptive forwarding strategy for multi-hop D2D communications to optimally choose between the best RUE forwarding (BRF) mode and the cooperative RUEs beamforming (CRB) mode, depending on which of them provides the higher instantaneous energy efficiency. We analyse the associated average energy efficiency considering the overhead for obtaining channel state information, forwarding mode selection and cooperative beamforming. Simulation results show that multi-hop D2D communications with the proposed adaptive forwarding strategy exhibit a significantly higher energy efficiency than the BRF, CRB, direct D2D communications and conventional cellular communications. Index Terms—D2D communications, energy efficiency, multihop, cooperative beamforming, overhead, cellular networks. I. I NTRODUCTION Different from conventional cellular communications, where user equipments (UEs) communicate via the base station (BS), device-to-device (D2D) communications enable UEs to communicate directly with other UEs in its vicinity using cellular resources [1]-[3]. D2D communications show potential for three types of gains: proximity gain, reuse gain and hop gain [4]. In practice, D2D UEs might not be close to each other or the channel conditions between them could be so poor that direct D2D communications would be impossible. Under these circumstances, relays could assist the communication between D2D UEs [5]-[8]. In [5], a distributed best relay selection method for D2D communications underlaying cellular networks was proposed, where the best relay among the ones that will not cause much interference to the cellular network was selected. Multi-hop UE relaying for sending emergency messages from disconnected areas was studied in [6]. For L3 relay assisted D2D communications underlaying LTE-A cellular networks, a gradient-based distributed resource This work was partially supported by the EC H2020 Project DECADE (MSCA-RISE-2014-645705). allocation scheme was proposed in [7]. In [8], energy and spectral efficiency of multi-hop D2D communication using single relay is analysed. In the works mentioned above, the overhead for obtaining channel state information (CSI) and for performing relay selection in multi-hop D2D communications has been neglected. D2D communications have not been considered in the existing works that analyze the overhead costs for implementing cooperative relaying in practical systems and the related energy consumption [9]-[13]. Nevertheless, these schemes select the number of relays based on the size of decoding relay set, which requires the knowledge of decoding set size and the availability of a lookup table (containing the optimal number of selected relays for any possible size of decoding set and location of cooperating relays) at the source [9]-[11][13] or at the destination [12]. In this paper, we propose a distributed adaptive forwarding strategy that optimally switches between the best relay UE (RUE) forwarding (BRF) and cooperative RUEs beamforming (CRB). The adaptive forwarding strategy considers the overhead energy consumption for obtaining CSI, forwarding mode selection and cooperative beamforming as well as the maximum transmission power constraint and channel coherence time. The proposed forwarding strategy is realized in two main steps. In the first step, the best RUE with the strongest secondhop channel in the main-cluster that contains all correctly decoding RUEs is selected using timers at RUEs. In the second step, if the sub-cluster formed by RUEs with first-hop channels at least as strong as that of the best RUE in the main-cluster is not empty, then the best RUE in the sub-cluster is selected to perform cooperative beamforming with the best RUE in the main-cluster; otherwise, BRF is performed. In other words, in the second step CBR is performed only if the instantaneous energy efficiency is improved compared to BRF, where only the best RUE in the main-cluster forwards the data. The remainder of the paper is organized as follows. The system model is presented in Section II. The proposed adaptive forwarding strategy for multi-hop D2D communications is described in Section III. Section IV analyses the energy efficiency for multi-hop D2D communications utilizing the proposed forwarding strategy, conventional cellular communi- cations, and direct D2D communications. The simulation results are presented in Section V. Finally, the paper is concluded in Section VI. II. S YSTEM M ODEL We consider D2D communications overlaying a cellular network as depicted in Fig. 1. The source UE (SUE) intends to transmit data packets to the destination UE (DUE). The data transmission from SUE to DUE can be realized in three different ways: conventional cellular communications via the BS, direct D2D communications between SUE and DUE, and multi-hop D2D communications through half-duplex decodeand-forward (DF) relay UEs (RUEs). The channel power gains between any two nodes are exponentially distributed and are represented as follows: hB is the channel power gain between SUE and BS; h0 is the channel power gain between SUE and DUE; hi (i=1,. . .,N ) denotes the channel power gain from SUE to RU Ei ; gB is the channel power gain between BS and DUE; and gi (i=1,. . .,N ) denotes the channel power gain from RU Ei to DUE. We assume reciprocal channels and singlecellular direct D2D multi-hop D2D ℎ1 SUE ℎ |D| ℎ0 RUE2 r ℎ𝐵 RUE1 𝑔2 RUE |D| RUE𝑁 𝑔|D| 𝑇𝐶𝑂𝐻 𝑇𝑂𝑀 Training (SUE->RUE) 𝑡0 Training (DUE->RUE) 𝑡1 𝑇𝐸𝑀 Forwarding Strategy Selection 𝑡2 Data Transmission (SUE->RUE) 𝑡3 Data Transmission (RUE->DUE) 𝑡3 + 𝑇𝐸𝑀 /2 𝑡3 + 𝑇𝐸𝑀 Fig. 2: Timing diagram for multi-hop D2D communications with the proposed forwarding strategy selection. A. Training ℎ2 ℎ𝑁 We propose an adaptive forwarding strategy for multihop D2D communications to dynamically switch between the best RUE forwarding (BRF) mode and the cooperative RUEs beamforming (CRB) mode in a distributed manner depending on which of them provides the higher instantaneous EE. The proposed forwarding strategy is summarized in Algorithm 1 and is explained in the following. As shown in Fig.2, multi-hop D2D communications with the proposed forwarding strategy have three main activities: training for acquisition of CSIs for both hops at each RUE, forwarding mode selection, and data transmission. 𝑔1 𝑔𝑁 𝑔𝐵 DUE BS Fig. 1: Different communication modes between SUE and DUE. antenna nodes that are subject to the additive white Gaussian noise (AWGN) with power spectral density of N0 . Perfect channel estimation at each node is assumed. The communication between each pair of nodes is performed with fixed rate R (bits/symbol) and bandwidth B (Hz). We consider the scenario with orthogonal channel allocation between cellular and D2D communications [8]. All UEs and the BS are subject UE to the maximum transmission power constraints of PM AX and BS PM AX , respectively. III. M ULTI -H OP D2D C OMMUNICATIONS WITH THE P ROPOSED A DAPTIVE F ORWARDING S TRATEGY In multi-hop D2D communications, SUE transmits its data to DUE with the help of DF RUEs that forward the decoded data to DUE. During the training stage, NT training symbols are transmitted from SUE to RUEs and from DUE to RUEs at time instants t0 and t1 , respectively. The N available RUEs estimate the corresponding channels. It is assumed that RU Ei (i=1,. . .,N ) are relatively close to each other resulting in approximately the same distance to SUE (dSR ) and to DUE (dRD ), respectively. The energy consumption for the training stage can be calculated as follows ETM = PTS,M + PTD,M NT TS , (1) R/B d where PTS,M = h̄M1−2 P , h̄M = 1/ PLD dξSR , and ln(1−δout ) N D,M ξd 1−2R/B = ḡM ln(1−δout ) PN , ḡM = 1/ PLD dRD ; h̄M and PT ḡM denote the corresponding mean channel power gains; PLD is a path loss constant for D2D communications and ξd is the path loss exponent; All RU Ei (i=1,. . .,N ) with the channel power gains hi no less than the threshold for successful decoding UE θth = (2R/B − 1)PN /PM AX become part of the main-cluster D = {RU E1≤i≤N : hi ≥ θth }. B. Adaptive Forwarding Strategy At time t2 , the procedure for forwarding strategy selection is initiated, and each UE belonging to the main-cluster D starts a timer τj = λ/gj , where λ is a constant parameter in unit of time [14]. The RU E1:|D| with the shortest timer τ1:|D| , i.e., the strongest channel to DUE, becomes part of the forwarding set F = {RU E1:|D| } and transmits NT training symbol to SUE with transmission power PTR,M = PTS,M . SUE performs channel estimation to obtain the first-hop CSI of RU E1:|D| and calculates the minimum transmit power to I reach RU E1:|D| , PD,1:|D| = 2R/B − 1 PN /h1 . Due to the broadcast property of wireless channels, other RU Ej ∈ D \ {RU E1:|D| } may still correctly decode data transmitI ted with power PD,1:|D| and can potentially improve EE through CRB. Therefore, all RU Ej ∈ D \ {RU E1:|D| } put their timers on hold when they overhear the transmission from RU E1:|D| . Since RU Ej ∈ D \ {RU E1:|D| } do not I know PD,1:|D| and hence do not know whether they can improve EE or not, SUE broadcasts a triggering symbol with I power PD,1:|D| . All RU Ej ∈ D \ {RU E1:|D| } that can correctly decode this symbol constitute the RUE sub-cluster R/B S = RU Ej ∈ D \ {RU E1:|D| } : hj ≥ (2 P I −1)PN and D,1:|D| resume their timers. The best RUE in the sub-cluster S, RU E1:|S| , with the shortest timer τ1:|S| becomes part of F if it improves the instantaneous EE and the time consumed for overhead to perform relay selection and cooperative beamforming is not bigger than t3 − t2 . In this case, CRB is selected as the forwarding strategy and RU E1:|S| transmits a notification ξ 1−2R/B PLD (2r) d PN to satsymbol with power PNR,M = ln(1−δ out ) isfy target outage probability δout at the maximum distance 2r, where r is the radius of main cluster D. As soon as receiving the notification symbol from RU E1:|S| , the other RU Ej ∈ S \ {RU E1:|S| } with unexpired timers reset their timers and do not participate in the cooperative data transmission. If RU E1:|S| cannot improve the instantaneous EE or the time consumed for related overhead exceeds t3 − t2 , then BRF is chosen as the forwarding strategy. The energy consumption for the proposed adaptive forwarding strategy is given by R,M R,M M I ES,F = NT PT + PD,1:|D| + (|F| − 1)PN TS . (2) Algorithm 1: Multi-hop D2D communications utilizing the proposed adaptive forwarding strategy. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 C. Data Transmission At time t3 , the data transmission stage starts. The effective transmission time for multi-hop D2D communications is TEM = TCOH − TOM , where TOM denotes the corresponding time consumption for overhead and is given by ( λ/g1:|D| , BRF M TO = (3NT + |F|) TS + , (3) λ/g1:|S| , CRB In the first TEM /2 time interval, SUE transmits data packets I with transmission power PD,1:|D| . These data packets are decoded only by RU Ej ∈ F. In the second TEM /2 time interval, RU Ej ∈ F forward the decoded data packets to DUE. The overall energy consumed for data transmission in multihop D2D communications using adaptive forwarding strategy is given by ! TEM M I II II ED,F = PD,1:|D| + PD,1:|D| + PD,1:|S| , (4) 2 25 26 27 28 29 30 31 32 33 34 35 36 37 i = 1, l = 1, D = ∅, S = ∅; SUE and DUE transmit NT training symbols with powers PTS,M and PTD,M , respectively. Each RU E1≤i≤N , estimates the corresponding hi and gi ; UE θth = (2R/B − 1)PN /PM AX ; while i ≤ N do if hi ≥ θth then D = D ∪ {RU Ei }; end i = i + 1; end All RU Ej ∈ D, start timers τj = λ/gj ; RU E1:|D| transmits NT symbols to SUE with power PTR,M = PTS,M ; DRES = D \ {RU E1:|D| }; Each RU El ∈ DRES , puts its timer on hold once it overhears the transmission from RU E1:|D| ; SUE transmits a triggering symbol with minimum power I to reach RU E1:|D| , PD,1:|D| ; while l ≤ |D| do I if RU El ∈ DRES && hl ≥ (2R/B − 1)PN /PD,1:|D| then S = S ∪ {RU El }; end l = l + 1; end F = {RU E1:|D| }; if |S| > 0 then All RU Ei ∈ S resume their timers; τ1:|S| = min τi ; i τO = τ1:|S| + (NT + 2)TS ; M M if EEF ∪{RU E1:|S| } ≥ EEF && τO ≤ t3 − t2 then F = F ∪ {RU E1:|S| }; RU E1:|S| transmits a notification symbol with power PNR,M ; All RU Ei ∈ S \ {RU E1:|S| } reset their timers; end end I ; SUE transmits data with power PD,1:|D| if |F| == 1 then RU E1:|D| forwards data to DUE with power II PD,1:|D| ; else RU E1:|D| and RU E1:|S| cooperatively beamform II II data towards DUE with powers PD,1:|D| and PD,1:|S| , respectively; end where PTS,C = where II PD,1:|D| 1 R/B g1:|D| (2 g1:|D| − 1)PN , BRF = 2R/B − 1 PN , CRB 2 (g1:|D| +g1:|S| ) II = PD,1:|S| 0, BRF g1:|S| (g1:|D| +g1:|S| ) 2 2R/B − 1 PN , CRB , (5) , (6) are the optimal transmission powers in the second-hop. For CRB, RUEs transmit with optimal powers obtained in [15]. From (5) and (6) is seen that RU E1:|D| and RU E1:|S| need to know each other’s channel power gains in order to calculate the optimal transmission powers for CRB. RU E1:|D| and RU E1:|S| can obtain each other’s second-hop channel power gains in a distributed manner through overhearing the notification symbols sent upon expiration of their timers. Assume that at time ts , RU E1:|S| overhears the notification symbol sent from RU E1:|D| , then RU E1:|S| acquires g1:|D| = λ/(ts − t2 ) [11], where t2 is the time instant when all RU Ej ∈ D start their timers. Similarly, g1:|S| can be calculated. Propagation delays within the main cluster D are negligible compared to RUE selection time. IV. E NERGY E FFICIENCY A NALYSIS A. Multi-Hop D2D Communications Based on Proposed Adaptive Forwarding Strategy Without loss of generality we assume that |S| > 0. In order to ensure a minimum effective transmission time, overhead time consumption for multi-hop D2D communications needs to be bounded, i.e., TOM ≤ TM AX . The outage in multihop D2D communications occurs when the second-hop link cannot support target rate R with maximum transmission UE power PM AX or the time consumed for overhead exceeds TM AX . The instantaneous EE for multi-hop D2D communications with the proposed adaptive forwarding strategy is given by M EEF = RTEM M + EM 2 ETM + ES,F D,F M 1, BRF & g1:|D| ≥ θth & TO ≤ TM AX 1, CRB& g1:|D| + g1:|S| ≥ θth & TOM ≤ TM AX 0, otherwise PTD,C = c h̄B = 1/ PLC dξSB and ξc S,C = 1/ PLC dBD ; PT and 1−2R/B P , h̄B ln(1−δout ) N 1−2R/B ḡB ln(1−δout ) PN , ḡB PTD,C are the required training transmit power levels from BS to SUE and from BS to DUE, respectively, to satisfy target rate R with outage probability δout ; TS = 1/B is symbol duration; PN = N0 B denotes the noise power; h̄B and ḡB are the mean channel power gains from SUE to BS and from BS to DUE, respectively; PLC is a path loss constant for cellular communications and ξc is the corresponding path loss exponent; dSB and dBD denote the distances from SUE to BS and from BS to DUE, respectively. Once DUE has estimated its channel to the BS, it feeds back the estimated CSI to BS using NF B symbols with the minimum transmission power that supports target rate R, R/B PFD,C = 2 − 1 P /g N B. B The energy consumption for the CSI feedback is given by EFCB = PFD,C B NF B TS . (9) During data transmission SUE transmits data to BS with the S,C adaptive power, PD = 2R/B − 1 PN /hB . BS forwards BS the received data to DUE with transmission power PD = R/B 2 − 1 PN /gB . The overall energy consumption for the data transmission is given by C S,C C BS TE , (10) ED = PD + PD 2 where TEC = TCOH − TOC is the effective transmission time for cellular communications; TCOH denotes the channel coherence time; TOC = (NT + NF B )TS is the time consumed for overhead in the cellular mode. The factor 1/2 comes from the two-hop half-duplex transmission. The outage in the cellular communications mode occurs when one of the links cannot support target rate R with UE maximum transmission power PM AX . The instantaneous EE for conventional cellular communications is given by C RTE C +E C +E C , {hB ≥ θth } & {gB ≥ θth } C 2 E ( T FB D) EE = , 0, otherwise . (11) C. Direct D2D Communications (7) The factor 1/2 is due to the two-hop half-duplex transmission. B. Cellular Communications In conventional cellular communications, SUE transmits data to DUE via the BS. Prior to data transmission, NT training symbols are broadcast from the BS to enable SUE and DUE to estimate their channels to the BS. The training broadcasting energy for reaching both SUE and DUE is given by n o ETC = max PTS,C , PTD,C NT TS , (8) In direct D2D communications, SUE directly transmits data to DUE. First, SUE transmits NT training symbols R/B PN . h̄0 = to DUE with the power PTS,D = h̄0 1−2 ln(1−δ out ) ξd 1/ PLD dSD is the mean channel power gain between SUE and DUE; dSD denotes the distance from SUE to DUE. The energy consumption for training can be calculated as ETD = PTS,D NT TS , (12) Then, DUE performs channel estimation and uses NF B symbols to feed back CSI to SUE with power PFD,D = B 2R/B − 1 PN /h0 . The energy consumption for the CSI feedback is given by EFDB = PFD,D B NF B TS , (13) After reception of CSI, SUE is able to adapt its transmission power to the minimum level required to support target rate R, S,D PD = PFD,D B , leading to the following energy consumption for data transmission: S,D D D ED = PD TE , TED (14) TOD where = TCOH − is the effective transmission time for direct D2D communications; TOD = (NT + NF B )TS is the time consumed for overhead in direct D2D mode. The outage in direct D2D communications occurs when the channel between SUE and DUE cannot satisfy target rate R UE with PM AX . The instantaneous EE for this mode is given by ( D RTE D +E D +E D , h0 ≥ θth D E T FB D . (15) EE = 0, otherwise Fig. 3: Average EE versus dSR for different communication modes and the proposed forwarding strategy with dSD = 200m, dSB = dBD = 300m, |D| = 10, and |S| = 5. V. S IMULATION R ESULTS The performance of the proposed adaptive forwarding strategy for multi-hop D2D communications in terms of average EE is evaluated through simulation. Average EE is obtained by averaging of instantaneous EE over 104 different channel realizations. Main system parameters are listed in Table I. During training, NT = 1 symbol is transmitted at the target rate R with outage probability δout = 0.1. We consider 16QAM modulation (R = 4) and, the channel coherence time of TCOH = 120 symbols, and TM AX = 0.25TCOH . DUE uses NF B = 2 symbols to feedback CSI to BS or SUE. The radius of main-cluster D is set to r = 5m. and increases with increasing dSR . Cellular communication is more energy-efficient than direct D2D communication due to the lower path-loss that results in lower transmission power required to satisfy target rate R and lower outage probability. Fig. 4 plots the average EE versus dSD for dSR = 0.1dSD and dSB = dBD = 300m. We can observe that the proposed forwarding strategy achieves the highest average EE. The EE for direct D2D and multi-hop D2D communications decreases with increasing dSD due to increasing transmission power and higher outage probability. For dSD ≥ 135m, direct D2D communication is even less energy-efficient than cellular communication. TABLE I: System parameters Bandwidth, B 10 MHz Noise power spectral density, N0 -174 dBm/Hz BS Maximum BS Tx power, PM AX 43 dBm UE Maximum UE Tx power, PM AX 23 dBm Path-loss for cellular communications 128.1 + 37.6 log10 [d(km)] dB Path-loss for D2D communications 148 + 40 log10 [d(km)] dB Fig. 3 shows the comparison of the average EE versus dSR among the proposed adaptive forwarding strategy, conventional cellular communications, direct D2D communications, BRF, and CRB for dSD = 200m and dSB = dBD = 300m. The proposed adaptive forwarding strategy achieves the highest average EE and outperforms both conventional forwarding strategies (BRF and CRB) for dSR < 100m, and is as energy efficient as BRF for larger dSR . This is because the proposed forwarding strategy selects optimally between BRF and CRB based on their instantaneous EE. For all the three forwarding strategies, the average EE initially increases with increasing dSR due to reduction of transmission power and outage probability in the second hop; after reaching the maximum, the average EE decreases because the energy consumption in the first hop dominates the overall energy consumption Fig. 4: Average EE versus dSD for different communication modes and the proposed forwarding strategy with dSB = dBD = 300m, dSR = 0.1dSD , |D| = 10, and |S| = 5. Fig. 5 plots average EE versus SUE or DUE to BS distance (dSB or dBD ), where dSB = dBD [8] and dSD = 150m. For dSB = dBD < 190m, it can be seen that cellular communica- tion is more energy-efficient than other communication modes due to lower transmission power and reduced outage probability. However, the average EE of cellular communications decreases quickly with increasing dSB = dBD . For dSB = dBD > 190m, the proposed forwarding strategy is more energy-efficient than cellular communications and becomes the most energy-efficient among all communication modes under comparison. Furthermore, for dSB = dBD > 350m, direct D2D communications outperforms cellular communications. Fig. 5: Average EE versus dSB or dBD for different communication modes and the proposed forwarding strategy with dSD = 150m, dSR = 0.4dSD , |D| = 10, and |S| = 5. VI. C ONCLUSIONS In this paper, we have proposed an adaptive forwarding strategy for multi-hop D2D communications. In the proposed strategy, the best RUE in the main-cluster, formed by all successful decoding RUEs, either forwards the data alone, i.e., via BRF or cooperatively through CRB with the best RUE in the sub-cluster, consisting of RUEs with first-hop channel not weaker than that of best RUE in the main-cluster. The decision to switch between BRF and CRB is made from best RUE in the sub-cluster and based on instantaneous energy efficiency. We performed EE analysis under maximum transmission power and channel coherence time constraints, considering the related overhead. The simulation results revealed that multi-hop D2D communications with the proposed adaptive forwarding strategy is more energy-efficient than best RUE forwarding, cooperative RUE beamforming, direct D2D communications and conventional cellular communications. R EFERENCES [1] L. Wei, R. Q. Hu, Y. Qian, and G. Wu, “Enable device-to-device communications underlaying cellular networks: challenges and research aspects,” IEEE Communications Magazine, vol. 52, no. 6, pp. 90–96, June 2014. [2] A. Asadi, Q. Wang, and V. Mancuso, “A survey on device-to-device communication in cellular networks,” IEEE Communications Surveys Tutorials, vol. 16, no. 4, pp. 1801–1819, Fourthquarter 2014. [3] P. Mach, Z. Becvar, and T. Vanek, “In-band device-to-device communication in ofdma cellular networks: A survey and challenges,” IEEE Communications Surveys Tutorials, vol. 17, no. 4, pp. 1885–1922, Fourthquarter 2015. [4] G. Fodor, E. Dahlman, G. Mildh, S. Parkvall, N. Reider, G. Mikls, and Z. Turnyi, “Design aspects of network assisted device-to-device communications,” IEEE Communications Magazine. [5] X. Ma, R. Yin, G. Yu, and Z. Zhang, “A distributed relay selection method for relay assisted device-to-device communication system,” in 2012 IEEE 23rd International Symposium on Personal, Indoor and Mobile Radio Communications - (PIMRC), Sept 2012, pp. 1020–1024. [6] H. Nishiyama, M. Ito, and N. Kato, “Relay-by-smartphone: realizing multihop device-to-device communications,” IEEE Communications Magazine, vol. 52, no. 4, pp. 56–65, April 2014. [7] M. Hasan, E. Hossain, and D. I. Kim, “Resource allocation under channel uncertainties for relay-aided device-to-device communication underlaying lte-a cellular networks,” IEEE Transactions on Wireless Communications, vol. 13, no. 4, pp. 2322–2338, April 2014. [8] L. Wei, R. Q. Hu, Y. Qian, and G. Wu, “Energy efficiency and spectrum efficiency of multihop device-to-device communications underlaying cellular networks,” IEEE Transactions on Vehicular Technology, vol. 65, no. 1, pp. 367–380, Jan 2016. [9] B. Klaiqi, X. Chu, and J. Zhang, “Energy-efficient cooperative beamforming using timer based relay subset selection,” in 2015 IEEE Wireless Communications and Networking Conference (WCNC), March 2015, pp. 369–374. [10] G. Lim and J. Cimini, L.J., “Energy-efficient cooperative beamforming in clustered wireless networks,” Wireless Communications, IEEE Transactions on, vol. 12, no. 3, pp. 1376–1385, March 2013. [11] B. Klaiqi, X. Chu, and J. Zhang, “Energy efficiency of location-aware clustered cooperative beamforming without destination feedback,” in 2015 IEEE International Conference on Communications (ICC), June 2015, pp. 2295–2300. [12] R. Madan, N. Mehta, A. Molisch, and J. Zhang, “Energy-efficient cooperative relaying over fading channels with simple relay selection,” Wireless Communications, IEEE Transactions on, vol. 7, no. 8, pp. 3013–3025, August 2008. [13] B. Klaiqi, X. Chu, and J. Zhang, “Energy-efficient and low signaling overhead cooperative relaying with proactive relay subset selection,” IEEE Transactions on Communications, vol. 64, no. 3, pp. 1001–1015, March 2016. [14] A. Bletsas, A. Khisti, D. Reed, and A. Lippman, “A simple cooperative diversity method based on network path selection,” Selected Areas in Communications, IEEE Journal on, vol. 24, no. 3, pp. 659–672, March 2006. [15] A. E. Khandani, J. Abounadi, E. Modiano, and L. Zheng, “Cooperative routing in static wireless networks,” IEEE Transactions on Communications, vol. 55, no. 11, pp. 2185–2192, Nov 2007.