Вы находитесь на странице: 1из 5

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts

for publication in the IEEE Globecom 2011 proceedings.

Relay Selection for Cooperative MAC Considering Retransmission Overhead


Key Laboratory of Science and Technology on Communications University of Electronic Science and Technology of China, Chengdu 611731, China Key Lab of Mobile Communication Technology Chongqing University of Post and Telecommunications, Chongqing 400065, China
National

Bin Cao , Gang Feng , Yun Li

AbstractRelay node (RN) plays a key role in cooperative communications and RN selection may substantially affects the performance gain. In this paper we address the issue of RN selection while taking into account Medium Access Control (MAC) overhead, which is incurred by not only handshake signaling but also frame retransmissions due to transmission error. We use a theoretical model to analyze the cooperation performance gains of cooperative MAC mechanism, and are thus able to select the optimal relay node. We derive the network saturation throughput of the designed MAC with our RN selection algorithm. Numerical results validate the effectiveness of our analytical model and show that our designed MAC signicantly outperforms existing cooperative MAC mechanisms which do not consider retransmission MAC overhead.

cooperative communications, in order to achieve a better performance gain, a good tradeoff between transmission delay and error probability should be made. This is the key issue in relay selection we would address in this paper. The remainder of this paper is organized as follows. Section II describes the motivation and states the problem. Section III presents our proposed cooperative MAC mechanism, relay selection algorithms in this paper. Section IV shows our analytical model for performance evaluation. We present the simulation results based on NS-2 to validate our cooperative MAC and analytical model in Section V and nally, conclude the paper in Section VI. II. M OTIVATION AND P ROBLEM D ESCRIPTION In direct transmission mode, when Signal-to-Noise Ratio (SNR) of the link between source and destination is low, and the probability of retransmission is high. For ease of presentation, suppose the modulation is BPSK, and the BER, D , can be calculated as [6] denoted by Pb
D = Q[ Pb

I. I NTRODUCTION A novel transmission mode which could provide spatial diversity gain, called cooperative communications[1][2], has emerged as one of the most promising techniques in wireless networks. Most of the existing research work[2][3] on cooperative communications is focused on cooperative method in physical layer, such as how to decrease Bit Error Rate (BER) and outage probability, and improve network capacity and energy efcient. In recent years, researchers have also been investigating cooperative MAC mechanisms to exploit the benets of cooperative communications as an implementation in practical wireless networks. Some cooperative MAC mechanisms are proposed and signicant performance gain has been observed [4][5]. Although existing cooperative MAC mechanisms can effectively improve the performance in terms of throughput and service delay compared to conventional (non-cooperative) MAC mechanism, all of them have not taken into account MAC overhead caused by retransmission due to impaired wireless channel. In this paper, we address the relay section and associated cooperative MAC design in a more realistic situation where retransmission due to not only failure of contending but also transmission error is considered. In
1 This

2D ],

(1)

where D is the received SNR combined at destination in direct transmission mode. In this mode, D is equal to the received SNR at destination from source denoted by sd . Q2 t2 function is dened as Q(x) = x 1 e dt. 2 In Decode and Forward (DF) cooperation mode, the BER, C , can be calculated as denoted by Pb
C = Q[ Pb

2C ],

(2)

work was supported by the National Natural Science Fund of China (61071098, 61071118 and 60702055), the National Major Projects on Nextgeneration Broadband Wireless Communication Networks (2010ZX03002008-03),the Natural Science Foundation Project of CQ CSTC (CSTC2010BA4010) and Chongqing Municipal Education Commissions Science and Technology Research Project (KJ111506).

where C = rd + sd is the combined SNR at destination in DF cooperation mode, rd is the received SNR of the signal sent from the relay node at destination. Let the frame loss probability for direct transmission and cooperative transmission be denoted by PD and PC respectively. We dene the duration from the time instance at which a frame is sent by the source to that the frame is received by the destination as f rame transmission time (FTT). Let the FTT for direct transmission and cooperative transmission be denoted by tD and tC respectively. Considering transmission error, the throughput for direct and cooperative communications, denoted by TD and TC respectively, can be computed as Ld TD = (1 PD ), (3) tD

978-1-4244-9268-8/11/$26.00 2011 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

Ld (1 PC ), (4) tC where Ld is the length of data frame. From (3) and (4), we know that cooperation could be adopted when TC > TD as the frame loss probability is decreased in cooperative communication mode. When tC > tD , conventional relay selection MAC mechanisms[4][5] will choose non-cooperation. On the other hand, in order to increase the combined SNR at destination, a cooperative relay could still be selected considering retransmission caused by transmission error. In summary, retransmission due to transmission error is overlooked in existing cooperative MAC mechanisms. This may cause the following subsequences: (1) the optimal relay node could be missed; (2) the cooperative opportunity could be wasted; and thus (3) the performance gain of cooperative communication is not fully exploited. In this paper, we propose an effective solution for the relay selection when MAC overhead is taken into account, especial the probability of retransmission due to error. TC = III. C OOPERATIVE MAC M ECHANISM AND R ELAY S ELECTION A. Cooperative MAC Mechanism In conventional wireless network MAC, such as IEEE 802.11[7], when a source has data to transmit, it sends a Ready-To-Send (RTS) if the source contends the channel successfully. The destination replies Clear-To-Send (CTS) to the source upon receiving the RTS. After receiving CTS, the source begins sending data to the destination, which will feedback ACK to the source when the data is received correctly. Otherwise, a retransmission is performed to recover the lost data. Cooperative communication technique can be applied in MAC with some modications to facilitate cooperation. In our proposed cooperative MAC, a triangular handshake is used for cooperative communication. To this end, we introduce three new types of frames, namely Cooperative-Ready-To-Send (CRTS), Ready-To-Relay (RTR) and Cooperative-Clear-To-Send (C-CTS), in our cooperative MAC. The format of C-RTS and C-CTS are similar to that of RTS and CTS respectively dened in IEEE 8021.11 standard. But they include additional identier information of the selected relay node. In addition, RTR is introduced to reserve the channel resources for the relay in cooperative MAC mechanism. The data transmission procedure in our cooperative MAC is shown in Fig.1. The handshaking mechanism is based on conventional 802.11 handshaking [7] to coordinate the transmission of nodes involved in cooperative communications. The details of the triangular handshaking procedure are as follows. Step 1: The source decides whether to use cooperative communication or not according to the performance gain comparison. If the answer is yes, which means cooperative gain can be achieved, the source will select the best relay based on our proposed relay selection algorithm (which will be elaborated in next sub-section). Otherwise,

C-RTS SIFS SIFS SIFS C-CTS

DATA SIFS SIFS ACK

D RTR

DATA

Fig. 1.

Data transmission procedure in cooperative communications

the source uses direct transmission, similar to that in IEEE 802.11. Step 2: The source sends a C-RTS when the channel is idle, and waits for RTR and C-CTS. Step 3: All nodes overhearing the C-RTS interpret the received C-RTS and the one which has been selected as the relay sends RTR. Other nodes will set Network Allocation Vector (NAV). Step 4: Upon receiving RTR, the destination replies CCTS to the source. Step 5: After receiving the C-CTS, the source transmits data to relay and destination by exploiting broadcasting nature of wireless communication. Step 6: The destination receives the data from the source and stores it. At the same time, the relay receives the data and forwards it to destination. Step 7: The destination sends an ACK if the decoding of combined data from the source and the relay respectively is successful. Otherwise, retransmission is performed.

B. Relay Selection in Cooperative MAC In our cooperative MAC, we need to determine whether to use cooperative communication and which node is selected as the relay node. This is done by examining the performance gain brought by cooperative communication when different relay node is selected. To this end, we need to derive the MAC overhead in terms of time. For transmitting a data frame, two steps of handshaking and frame transmission are performed. According to the 802.11 MAC protocol, the frame transmission time (FTT) in direct communication mode is computed as tD = tP HY hdr + tM AChdr + trts + tcts + td D+ tack + 3tSIF S + tDIF S + 4 + tSLOT , (5)

Ld where td D = Rsd is the data frame transmission time in direct communication mode;trts , tcts , tack , tmachdr , tphyhdr , tSIF S , tDIF S , tSLOT and are the duration of RTS, CTS, ACK, MAC header, PHY header, Short Inter Frame Spacing (SIFS), DCF Inter Frame Spacing (DIFS), slot time and the propagation delay,respectively. In cooperative communication mode, due to the extended handshaking including C-RTS/RTR/C-CTS and data forwarding via relay, the FTT can be derived as [4][5]

tC =

2tP HY hdr + 2tM AChdr + tcrts + trtr + tccts + td C + tack + 5tSIF S + tDIF S + 6 + 6tSLOT ,

(6)

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

Ld Ld where td C = Rsr + Rrd is data frame transmission time cooperative communication mode; tcrts , trtr , and tccts are the duration of C-RTS, RTR and C-CTS, respectively. When the cooperative performance gain is higher than the overhead caused by cooperation, cooperative communication is adopted. The best relay is the node which could provide the most performance improvement. Considering retransmission caused by transmission error, the exact throughput could be calculated according to TD and TC in (3), (4) respectively. In our cooperative MAC, relay is selected by comparing the throughput for direct and cooperative communications, i.e., TD and TC .The details of the relay selection algorithm is shown as follows, where R[i] denotes the set of candidate relay nodes and T [i] denotes the set of throughput provided by different relay nodes.

In DF cooperation mode, if any frame is decoded incorrectly, the data could not be recovered at relay. In this case, no signal is transmitted from the relay to destination and thus the SNR at destination is the same as that of direct transmission. According to (5), (6), (7) and (8), we can calculate throughput in (3) and (4), and perform the optimal relay selection. IV. P ERFORMANCE A NALYSIS In this section, we analyze the saturation throughput of the proposed relay selection in cooperative MAC based on IEEE 802.11b physical layer specications. The result could be extended to other physical layer specications of wireless networks. The failure probability pf and the probability of a station transmitting in a randomly chosen slot time can be expressed as [8]: (9) pf = 1 (1 pcol )(1 pe ), 2(1 2pf ) , (1 2pf )(CWmin + 1) + pf CWmin (1 (2pf )m ) (10) where pcol = 1 (1 )n1 is the collision probability of frames, pe is the frame error probability, which is equal to PC in (8), CWmin is the minimum value of congestion window, m is the maximum backoff stage and n is the number of contending nodes. (9) and (10) represent a non-linear system and has been proven to have a single solution [10][11]. We can solve it using a cyclic iterative method. When only one node transmits in a time slot and the CRTS/RTR/C-CTS/data/ACK frame is corrupted, a transmission error occurs. Therefore, The probability that a C-RTS/RTR/Ccrts , CTS/data/ACK frame is lost in a slot time, denoted by PE ccts rtr d ack , PE and PE , respectively, can be computed as PE , PE =
crts rts PE = n (1 )n1 pc , e rtr rts rtr PE = n (1 )n1 (1 pc )pe , e ccts rts ccts PE = n (1 )n1 (1 pc )(1 prtr , e e )pe d rts ccts C PE = n (1 )n1 (1 pc )(1 prtr )pe , e e )(1 pe ack rts PE = n (1 )n1 (1 pc )(1 prtr e e ) ccts C ack )(1 pe )pe . (1 pe

Algorithm 1 Relay Selection Algorithm When S needs to transmit data to D if R[i] = and TD < TC then TC = max TCi , (i) R = arg max TC
RR[i] TCi T [i]

node R is selected as the relay, and cooperation mode is adopted else direct transmission mode is adopted end if To compute the throughput in (3) and (4), we must derive PD and PC rst. For simplicity, let us again suppose the modulation is BPSK. According to the corresponding frame length L and BER pb , we can now calculate the Frame Error crts ), RTR (prtr Ratio (FER) of RTS (prts e ), C-RTS (pe e ), CTS cts ccts ack ) and ACK (pe ). (pe ), C-CTS (pe The error probability of data frame (with length Ld ) at destination in direct transmission, denoted by pD e , can be D Ld = 1 (1 p ) , and then P calculated as pD D is given e b by
cts D ack PD = 1 (1 prts e )(1 pe )(1 pe )(1 pe ),

(7)

The error probability of data frame at destination in cooperative transmission, denoted by pC e , is given by
sr sd sr DF pC e = pe pe + (1 pe )pe ,

where
sr Ld psr e = 1 (1 pb ) , psr b = Q[ 2sr ], D Ld psd e = 1 (1 pb ) , Ld pDF = 1 (1 pC e b ) ,

Let tI ,tcol ,tS be average time duration that the channel is sensed idle, sensed busy during a collision, and sensed busy by each station because of a successful transmission, respectively. rts , Let tEIF S be during of Error Inter Frame Spacing, tc E ccts rtr d ack tE , tE , tE and tE be the average time the channel is sensed busy due to a transmission error of C-RTS/RTR/CCTS/data/ACK frame, respectively, which can be computed as tI = tSLOT , tcol = tcrts + + tSLOT + tEIF S , tS = 2tP HY hdr + 2tM AChdr + tcrts + trtr + tccts + td C + tack + 5tSIF S + tDIF S + 6 + tSLOT ,
rts tc = tcrts + + tSLOT + tEIF S , E

and sr is the SNR of the received signal sent from the source node at relay. As a result, PC can be computed as PC =
rts 1 (1 pc )(1 prtr e e ) cts C )(1 p )(1 pack (1 pc e e e ).

(8)

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

800

800

700

700

600 Throughput in Kbps Throughput in Kbps analytic result for proposed cooperative MAC simulation result for proposed cooperative MAC analytic result for CoopMAC simulation result for CoopMAC 9 9.5 10 10.5 11 11.5 12 12.5 13 SNRsr=SNRrd in dB 13.5 14 14.5 15

600

500

500

400

400

300

300 analytic result for proposed cooperative MAC simulation result for proposed cooperative MAC analytic result for CoopMAC simulation result for CoopMAC 8 8.5 9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15 SNRsd in dB

200

200

100

100

Fig. 2. Comparison between our proposed cooperative MAC and CoopMAC with vary sr and rd

Fig. 3. Comparison between our proposed cooperative MAC and CoopMAC with vary sd

2 shows the network throughput as a function of sr and rd . We can see that the analytical results and simulation = tcrts + trtr + tccts +3 +2tSIF S + tSLOT + tEIF S , results match very well. When sr and rd are higher than d 9.5dB, the saturation throughput of proposed cooperative MAC td E = 2tP HY hdr + 2tM AChdr + tcrts + trtr + tccts + tC + is improved obviously. In contrast, the saturation throughput 5 + 4tSIF S + tSLOT + tEIF S , of CoopMAC is independent of sr and rd . The reason is tack E = 2tP HY hdr + 2tM AChdr + tcrts + trtr + tccts + that our proposed cooperative MAC takes into account the td C + tack + 6 + 5tSIF S + tSLOT + tEIF S . overhead caused by both transmission and retransmission, and We can then express the normalized saturation throughput the cooperation mode is adopted when the cooperative gain is higher than the overhead. In this case, cooperation mode with of cooperative communication in an error-prone channel as high sr and rd , can reduce the retransmission impact on crts crts S = PS Ld /(PI tI + PS tS + Pcol tcol + PE tE + (11)network performance due to reduced BER. CoopMAC selects ccts ccts rtr rtr d d ack ack P E t E + PE tE + PE tE + PE tE ). direct mode when the overhead caused by transmission only where PS = n (1 )n1 (1 pe ) is the probability of a is taken into account. On the other hand, when sr and rd successful transmission in a slot time; PI = (1 )n is the are lower than 9.5dB, the proposed cooperative MAC and probability that no transmission occurs in a slot time; Pcol = CoopMAC have very similar performance. The reason is that in this situation, cooperative communication cannot provide 1 (1 )n n (1 )n1 is the probability of collision. cooperative gain any more, and both mechanisms use the same V. S IMULATION R ESULTS direct transmission mode. To verify our analytical model and demonstrate the perIn the second experiment, let sr and rd be xed to 15dB, formance improvement of the proposed cooperative MAC Rsr and Rrd be set to 2Mbps, Rsd be set to 1Mbps, there be 10 mechanism, we use simulation experiments based on NS-2 contending nodes and data frame length be set to 1000 bytes. to evaluate the throughput performance. As CoopMAC[4] is We vary sd from 8dB to 15dB. Fig. 3 shows a comparison one of the most known cooperative MAC mechanism, we use of throughput of our proposed MAC and CoopMAC by both it as a comparison reference. analysis and simulation experiments. We can see the proposed The parameters used in our simulation are similar to those cooperative MAC outperforms CoopMAC in throughput when used in [4]. MAC header is set to 272bits, PHY header sd is lower than 10dB. In this situation, retransmission probais set to 192bits, tcrts =352s, tccts =304s, trtr =304s, bility due to the error-prone direct link is high and thus the pro =1s, tack =304s, tSLOT =10s, tDIF S =50s, CWmin =31 posed cooperative MAC chooses cooperation mode. Although and m=4. We use 802.11b MAC DCF model, UDP CBR trafc cooperative communication could improve the performance sources and the queue length is set to 50. In our experiments, when sd is lower than 10dB, cooperation opportunities are the simulation time is set to 100 seconds. We use saturation missed in CoopMAC, and thus lower throughput performance throughput as the performance metric in performance evalua- is resulted. When sd is higher than 10dB, both two MAC tion in all simulations with xed topology. For simplicity, we mechanisms choose direct transmission mode, because the set sr = rd and Rsr = Rrd . direct link is good enough. Furthermore, we can also see that In the rst experiment, sd is xed to 9dB, Rsr = Rrd analytical results match simulation results very well and this =2Mbps, Rsd =1Mbps, there are 10 contending nodes and supports the effectiveness of our analytical model. data frame length is set to 1000 bytes. In this wireless In the third experiment, let sr and rd be xed to 15dB, environment, we vary sr and rd from 9dB to 15dB. Fig. Rsr and Rrd be set to 2Mbps, Rsd be set to 1Mbps and there
cts tc E

trtr E = tcrts + trtr + 2 + tSIF S + tSLOT + tEIF S ,

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE Globecom 2011 proceedings.

4500 4000 3500 Data frame length in Byte 3000 2500 2000 1500 1000 500 0 proposed cooperative MAC CoopMAC

10

11 12 SNRsd in dB

13

14

15

Fig. 4.

Optimal data frame length with vary sd

Rsr and Rrd be set to 2Mbps, Rsd be set to 1Mbps and there be 10 nodes. We vary sd from 8dB to 15dB. In Fig.5, we compare the saturation throughput of our proposed cooperative MAC with that of CoopMAC with optimal data frame length. We can see that our proposed cooperative MAC outperforms CoopMAC in throughput when sd is lower than 11dB. According to Figs.4 and 5, we know when sd is lower than 11dB, our proposed cooperative MAC chooses cooperation mode while CoopMAC prefers direct transmission mode. The reason is the inuence of overhead incurred by retransmission, and then our proposed cooperative MAC can maximize the advantage of cooperation. When sd is higher than 11dB, both cooperative MAC mechanism choose direct transmission mode due to good condition of direct link. VI. C ONCLUSIONS In this paper, we have proposed a relay selection algorithm for cooperative MAC mechanism for cooperative communications, to exploit the potential benets by considering MAC overhead incurred by retransmission. Through careful protocol design, our proposed cooperative MAC can select the most suitable node as a relay thus alleviates the retransmission probability and enhances the cooperation opportunity. Simulation results have validated our analytical model and shown that our proposed cooperative MAC can signicantly improve network performance in terms of throughput when transmission error cannot be ignored.

1000 900 800 Throughput in Kbps 700 600 500 400 300 200

analytic result for proposed MAC simulation result for proposed MAC analytic result for CoopMAC simulation result for CoopMAC 8 9 10 11 12 SNRsd in dB 13 14 15

R EFERENCES
[1] A. Nosratinia, T. E. Hunter, and A. Hedayat, Cooperative Communication in Wireless Networks, IEEE Communications Magazine, vol. 42, no. 10, pp.74-80, October 2004. [2] J. Laneman, D. Tse, and G. Wornell, Cooperative Diversity in Wireless Networks: Efcient Protocols and Outage Behavior, IEEE Transactions on Information Theory, vol. 50, no. 12, pp. 3062-3080, December 2004. [3] B. Wang, Z. Han, and K.J.R. Liu, Distributed Relay Selection and Power Control for Multiuser Cooperative Communication Networks Using Buyer/Seller Game, Proceedings of IEEE INFOCOM 2007, Anchorage, AK, May 2007 [4] P. Liu, Z. Tao, S. Narayanan, T. Korakis, and S. S. Panwar, CoopMAC: A Cooperative MAC for Wireless LANs, IEEE Journal on Selected Areas in Communications, vol. 25, no. 2, pp.340-354, February, 2007. [5] Y Li, B Cao, CG Wang, et al. Dynamical cooperative MAC based on optimal selection of multiple helpers, Proceedings of IEEE GLOBECOM 2009, Hawaii, Nov 2009. [6] J. G. Proakis, Digital Communications, 2nd ed. New York: McGrawHill, 1989. [7] IEEE Standard for Information technology-Telecommunications and information exchange between systems-Local and metropolitan area networks-Specic requirements Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specications, IEEE Computer Society, 12 June 2007. [8] G. Bianchi, Performance analysis of the IEEE 802.11 distributed Coordination function, IEEE Journal on Selected Areas in Communications, vol. 18, no. 3, pp.535-547, March 2000.

Fig. 5.

Throughput obtained with the optimal data frame length

be 10 contending nodes. We vary sd from 8dB to 15dB. According to our analytical model, the optimal data frame length according to the channel conditions can be computed and shown in Fig.4. We increase the data frame length from 100 bytes to 4500 bytes with a step of 100 bytes to nd the optimal data frame length which provides the maximum throughput. In order to achieve the maximum throughput, the data frame length should be set based on the channel conditions. In Fig.4, we can know that the optimal data frame length of CoopMAC increased with sd , while it keeps the xed maximum value in our proposed cooperative MAC. The reason is that CoopMAC misses cooperative opportunity and the direct transmission mode is adopted, and thus the performance is inuenced by sd . On the other hand, our proposed cooperative MAC selects cooperation mode, the performance depends on sr and rd . When sr and rd are high enough, the probability of retransmission caused by error is very low, and thus the largest data frame length can be adopted. When sd reaches 11dB, this situation means that the channel condition between source and destination is good enough, and then both the two cooperative MAC mechanisms adopt direct transmission mode and choose 4500 bytes as the optimal data frame length. In the last experiment, let sr and rd be xed to 15dB,

Вам также может понравиться