Adaptive Modulation and Coding with Cooperative Transmission in MIMO fading Channels Yuling Zhang1, a, Qiuming Ma2, b

Transcription

1 4th atonal Conference on Electrcal, Electronc and Computer Engneerng (CEECE 05) Adaptve Modulaton and Codng wth Cooperatve Tranmon n MIMO fadng Channel Yulng Zhang, a, Qumng Ma, b School of Informaton and Electrcal Engneerng, Ludong Unverty, Yanta, Shandong, 6405, P.. Chna School of Informaton and Electrcal Engneerng, Ludong Unverty, Yanta, Shandong, 6405, P.. Chna a emal: zhang-yulng@hotmal.com, bemal: mqml@na.com Keyword: Cooperatve tranmon, adaptve modulaton and codng (AMC), multple-nput multple-output (MIMO) channel. Abtract. In th paper, we propoe a cooperatve tranmon framework baed on adaptve modulaton and codng (AMC) ytem n multple-nput multple-output (MIMO) fadng channel. We conder the cenaro that cont of one ource node, one relay node and one detnaton node, they all equpped wth multple antenna and employed the adaptve modulaton and codng cheme. Through computer mulaton, we got the effectve capacty of the adaptve cooperatve tranmon ytem accordng to dfferent QoS requrement. Introducton The ncreang development of wrele applcaton, epecally real-tme meda traffc wth trngent QoS contrant requre a hgh effcent utlzaton of the carce rado reource. Many technque are propoed to mprove the throughput of tme-varyng fadng channel whle mantanng a atfed QoS, uch a lnk adaptaton baed on adaptve modulaton and codng (AMC) and automatc repeat requet (AQ) [][][3], multple-nput and multple-output ytem and cooperatve tranmon through node cooperaton. There are many lterature about the adaptve cooperatve tranmon, Jall Sefal nvetgate the effectve capacty for mult-rate relay channel wth delay contrant explotng adaptve cooperatve dverty [4], and author n [5] degn a decon-makng algorthm on cooperatve tranmon by ung a partally obervable Markov decon proce framework. Shan Chu explot the ue of cooperatve relay tranmon n a MIMO-baed ad hoc network to cope wth harh channel condton [6]. In th paper, we propoe a framework of adaptve cooperatve tranmon baed on adaptve modulaton and Low-denty party-check (LDPC) code n mult-antenna ytem. The ret of th paper organzed a follow. In Secton II, we preent the ytem model and llutrate the prncple of the cooperaton tranmon. How to calculate the effectve capacty are explaned n Secton III. In Secton IV, we get numercal reult through mulaton. Fnally, ome concluon are drawn. Sytem Model The ytem model of cooperatve tranmon baed on LDPC code n MIMO channel hown n Fg.. The ytem cont of ource node S, detnaton node D and relay node. Aumng that there are T tranmt antenna, relay antenna and D receve antenna. The adaptve tranmon acheved by two mean, at the phycal layer, there are multple modulaton and codng cheme (MCS) avalable, not only n ource node but alo n relay node. Coded ymbol are tranmtted to the relay node and the detnaton node multaneouly on a frame-by-frame ba through MIMO fadng channel after pace-tme block codng. The CSI etmated at the recever and then ent back through a feedback channel to the AMC controller for both the ource node and delay node, whch chooe the approprate MCS n the next tranmon 06. The author - Publhed by Atlant Pre 363

2 accordngly. Input Buffer AMC Controller Encoder Decoder C-LDPC Decoder Output S D Feedback channel Channel Etmator Source ode Detnaton ode AMC Controller Encoder elay ode Fg.. Sytem model At the data lnk layer, adaptve cooperatve retranmon employed to mprove the throughput performance, epecally for the communcaton of delay-entve packet traffc. The packet and frame tructure ued n th paper are mlar to thoe a llutrated n []. Source nformaton are tranmtted frame by frame, each frame dvded nto two tme-lot, durng the frt tme-lot, node S elect an AMC mode baed on the S-D channel condton, meanwhle node and node D lten. When an error detected n a packet at node D, a retranmon requet generated and ent back to node S and node va a feed back channel. Then at the econd tme-lot, node forward the packet receved from node S to node D when t able to decode the ource packet correctly, otherwe, node S wll delver the packet agan. For mplcty, we aume that the feed back channel error free and ha zero delay. A block fadng channel model adopted for all S-D, S-,and S-D lnk. Here we take the S-D lnk for an example, S- and S-D lnk can be analyzed n a mlar way. The MIMO fadng channel S between S and can be expreed a a matrx, H [ h j ], j, where h j the channel coeffcent between the jth tranmt antenna and the th relay antenna. Under the aumpton of ndependent aylegh fadng, the channel coeffcent h j are modeled a ndependent and dentcally dtrbuted (..d.) complex crcular Gauan random varable wth zero mean and unt varance. Accordng to the effectve SISO channel model for decrbed n [7],the at the recever gven by α α Pd Pd T γ γ H H F H F () F Tc Tc where the codng rate of c / T, P the average tranmt power per tream/antenna, P T the total tranmt power tranmtted on T antenna per ymbol duraton and γ Pd α T / σ defned to be the average peudo, n whch d the Eucldean dtance between noe S and D, α the path lo exponent. Snce H the um of K..d. χ F random varable, we can get the probablty denty functon (PDF) of γ a follow [8] K K γ T c T c ( K) γ γ pγ ( γ ) exp γ, γ 0 Γ where Γ () the Gamma functon. In the ame way, we let γ and γ 3 denote the receved on the -D, and S- channel, the Eucldean dtance between noe and D, and noe S and are d and d 3 repectvely, the average are gven by Pd α γ r / σ, and γ3 Pd α T 3 / σ. The AMC cheme adopted at node S and node can be llutrated a follow: Packet receved 364 ()

3 ncorrectly after r retranmon wll be dropped. In order to meet the ytem delay contrant, for a gven packet lo probablty PE lnk at the data lnk layer, the packet error rate (PE) P target at the phycal layer hould be [] /( r + ) Ptarget PElnk (3) The AMC mplemented at the phycal layer accordng to the target PE. Suppoe that there are MCS at the phycal layer wth ncreang rate n (n,,, ) n term of nformaton bt per ymbol. We wll conder the modulaton method wth the MQAM gnal contellaton, where M denote the number of pont n each gnal contellaton. If the codng rate of a MCS L, we have n L (log M). A n [], the whole range of the γ, γ are dvded nto + and M + non-overlappng conecutve nterval, denoted by [ γn,, γ n +,), n 0,,..., and [ γm,, γ m +,), m 0,,... M. When γ fall nto the nterval [ γn,, γ n +,), n, node S elect the AMC mode n and end data wth tranmon rate n, (bt/ymbol). In a mlar way, when γ [ γm,, γ m +, ), m, node elect mode m and end data to node D wth rate m, (bt/ymbol). When γ [ γ0,, γ,),no data ent by node. But whenγ [ γ 0,, γ, ), node S tll tranmt packet from t buffer wth the frt AMC mode (n). The threhold γ n, and γ m, are determned n the ame way a n [], accordng to the AMC rule, each MCS n wll be choen wth the followng probablty [] γ n, Γ( K, λγ n,) Γ( K, λγ n,) pn, + pγ ( γ ) dγ +. (4) γ n, Γ( K) It can be hown that the average PE for MCS n gven by γ n+, PE n PE ( γ ) p ( γ ) dγ. (5), γ n γ n, Then, the total average PE can be wrtten a follow P n, n n n, PE p n, n, n,. (6) In real tme multmeda ervce uch a vdeo tranmon, tme delay an mportant QoS parameter. Effectve capacty (EC) propoed to decrbe the maxmum throughput of the ytem under delay contrant, and wdely ued to analyze the QoS performance of wrele multmeda network. In aylegh fadng channel, let the bnary random varable X ndcate the number of packet ervced by the queue ervce proce at node S n frame. If the packet n the th frame decoded correctly by node D then upper-bounded a follow X, otherwe X 0 (packet error). The EC functon can be EC( θ ) log E { exp( )} X log[ exp( ) ( exp( )) Pr( 0) ] bx b b X T θ θ f θt θ + θ (7) f Where θ 0 reflect the qualty requrement of the tranmon [9]. A maller θ repreent a looer QoS contrant, when θ tend to 0, an arbtrarly long delay can be tolerated, the EC converge to the maxmum throughput (ergodc capacty) of the ytem. In an AMC-baed tranmon ytem wth a fxed ymbol rate, the duraton of each tme lot n a frame depend on the employed AMC mode. If n the frt tme-lot, node S tranmt wth mode n, the duraton of th tme-lot denoted by ( n) b T,, n (8) n, Where b the packet length n bt, and the channel ymbol rate per econd. Smlarly, the retranmon tme duraton n the econd tme-lot for node can be expreed a 365

4 ( m) b T,, m (9) m, Pr( X 0) M (0) p0,i PE0, + pn,i npen, + p0,( I ) PE0, + pn,( In) PEn, p0, + pm,pem, n n m γ m, Γ( K, λγ m,) Γ( K, λγ m,) pm, + pγ ( γ ) dγ + () γ m, Γ( K) I ( ) ( ) n PEn γ p γ dγ () 0 3 γ for the relay-ated tranmon ytem, the expected value of the frame length () ( n) ( m) T f E ( Tf ) T, Pn, + T, pm, p0,( I ) PE0, + pn,( In) PEn, n 0 m n Smulaton eult T f gven by (3) In th ecton, numercal reult howng the effect of dfferent parameter on the EC of our cooperatve tranmon framework are provded. Frtly, the ytem parameter are et a follow: The S-D dtance d normalzed to one, the relay poton controlled by changng the -D dtance d, path lo exponentα 4, the packet length b 008bt. The frame tme duraton () n the AMC mode, T, m, relay poton d d3 d /, equal tranmt power for node S and. ode S and ue the ame AMC mode et, adoptng from the IEEE 80.a tandard [0], we ue C-LDPC code ntead of convolutonal code. Aumng that the performance contrant at the data lnk layer PE lnk 0.0. The varable node degree dtrbuton of rregular LDPC code a follow λ( x) λ x x x x x x (4) The EC for both cooperatve and drect tranmon ytem under everal channel cenaro gven n Fg.. In noncooperatve tranmon cheme, node S drectly tranmt t packet to node D wthout the atance of the relay node. We can ee n fg., compared wth the noncooperatve tranmon, the cooperatve protocol dramatcally mprove the EC. 7 x 05 6 Effectve Capacty (bt/ec) non-cooperatve tranmon M6 cooperatve tranmon M theta(/bt) Fg. EC for dfferent θ Concluon In th paper, we propoed an adaptve tranmon frame n cooperatve communcaton ytem under 366

5 MIMO channel. The AMC at the phycal layer and the adaptve cooperatve tranmon at the data lnk layer are combned to acheve a better EC. At the ource node and the relay node, relevant MCS choen baed on the threhold calculated accordng to the LDPC PE- relatonhp. umercal reult how that the cooperatve ytem can provde better EC than the non-cooperatve ytem. Acknowledgment Th work upported by atonal Scentfc Foundaton of Chna (o. 6735, o.64785) and Scentfc Foundaton of Shandong Provnce (o. Z03FL007). eference [] Q. Lu, S. Zhou, and G. B. Gannak. Cro-layer combnng of adaptve modulaton and codng wth truncated AQ over wrele lnk [J]. IEEE Tranacton on Wrele Communcaton, 004,3 (5): [] Yulng Zhang, Dongfegn Yuan, Chengxang Wang, Cro-Layer Degn Baed on C-LDPC Code n MIMO Channel wth Etmaton Error [J]. AEU - Internatonal Journal of Electronc and Communcaton,008, 6(9): [3] D. Wu and. eg, Effectve Capacty: a wrele lnk model for upport of qualty of ervce. IEEE tranacton on Wrele Communcaton [J]. July 003, (4): [4] Jall Sefal Harn and Mchele Zorz. Effectve Capacty for Mult-ate elay Channel wth Delay Contrant Explotng Adaptve Cooperatve Dverty [J]. IEEE Tranacton on Wrele Communcaton, 0, (9): [5] Dong Heon Lee, Wha Sook Jeon, Kae Won Cho etc. Adaptve tranmon polcy over aylegh fadng channel for cooperatve network wth lmted feedback [J]. IET Communcaton, 03, 7(7): [6] Shan Chu ; Xn Wang ; Yuanyuan Yang. Explotng Cooperatve elay for Hgh Performance Communcaton n MIMO Ad Hoc etwork [J]. IEEE Tranacton on Computer, 03, 6(4): [7] V. Tarokh, H. Jafarkhan, and A.. Calderbank. Space-tme block code from orthogonal degn [J]. IEEE Tran. Inform. Theory, 999, 45 (5): [8] D. Lu. Stochatc Proce and It Applcaton. n Chnee, Tnghua Unverty Pre, 986. [9] L.Lu, P.Parag, and J.-F Chamberland. Qualty of ervce analy for wrele uer-cooperaton network. IEEE Tran. Inform. Theory, October 007, 53(0): [0] A. Doufex, S. Armour, M. Butler, A. x, D. Bull, J. McGeehan, and P. Karlon. A comparon of the HIPELA/ and IEEE 80.a wrele LA tandard [J]. IEEE Communcaton Magazne, 00, 40 (5):