Multi-Relay Selection Strategy for Device to Device Communication

Transcription

1 Internatonal Conference on Computer, Networks and Communcaton Engneerng (ICCNCE 3) Mult-elay Selecton Strategy for Devce to Devce Communcaton Wecheng Xa, Shxang Shao, Jun Sun Jangsu Provncal Key Laboratory of Wreless Communcaton, Nanjng Unversty of Posts and elecommuncatons Nanjng, 3, P Chna xawecheng@63com, {shaosx, sunjun }@njupteducn must be weghed aganst ts outperformance here are some oer factors to consder, A Bletsas, A Khst, D P eed, and A Lppman[8] take e network lfetme nto consderaton durng e relayng selecton We should notce at users n tradtonal cellular network wll not nterfere to each oer as ey use dfferent resources In our case, e DD users reuse e cellular resources durng e communcaton, so e nterference can t be gnored he selecton strategy we propose dffers from tradtonal relayng selecton algorm n two aspects: () he transmsson power of DD users should be confned he sgnal transmtted by DD user may produce nterference to base staton, whch may affect e normal cellular communcaton () he number of relayng nodes cannot be predcted, ere may be one or more relayng nodes n communcaton accordng to specfc context he rest of e paper s organzed as follows: In Secton, we descrbe e system model and analyze e nterference between base staton and DD users In Secton 3, we propose dfferent relyng strategy reusng e uplnk (UL) cellular resources Smulaton results are presented n Secton 4 and Secton 5 concludes e paper Abstract - Introducng relayng nodes to DD (Devce to Devce) communcaton system can effectvely mprove e system performance However, e qualty of drect DD communcaton cannot be guaranteed when e users are far from each oer, e effect of relayng node becomes partcularly mportant n s case In s paper, we studed dfferent relayng selecton strateges of DD communcaton n cellular network, n whch we take e mutual nterference between e DD pars and cellular users nto consderaton he selecton strateges we proposed can adaptvely determne e number of relayng nodes accordng to specfc context Fnally we compare e proposed relayng selecton strategy w smulaton results and analyze er performances Index erms - Devce to Devce (DD); elay Selecton; Network Lfetme Introducton W e rapd development of moble communcaton, bandwd below 5G s becomng more and more crowed and e reuse of band resources s partcularly urgent n recent years [] Introducng Devce to Devce (DD) communcaton nto cellular network can effectvely ncrease e utlzaton of band resources and reduce e load of base staton But e problem (e, poor relablty of e edge user n e cellular network) also occurs n e DD communcaton system So t s mportant to ntroduce relay scheme nto DD communcaton to enhance e coverage of e system and provde better servces here are a varety of relay selectons at present SENG V, YANIKOMEOGLU H, Falconer D D [] proposed ree relay strateges accordng to e dstance between two nodes It s smple to select e best relay node w shortest relay dstance; however e symbol error rate of e whole system wll dramatcally ncrease f e channel between two nodes experence deep fadng HASNA M O, ALOUINI M S [3] proposed a relay selecton scheme accordng to e harmonc mean of e SN of two channels and SU W, IBAHIM A S, SADEK A K [4] furer analyzed e problem of when to cooperate and whom to cooperate w [5]-[6] are focused on e optmzaton of above two selecton Alough e strategy (e, selectng e relay node w best e channel state nformaton) outperforms e relayng node selecton based solely on dstance, e former selecton has hgher sgnalng overhead hs dsadvantage System Model and Problem Formulaton Fg shows e scenaro we consder n s paper, ere are M cellular users (e, C, C,, CM ) and one DD par (e, a C C CM In rf er en ce D S gn al te N S Fg System model of DD communcaton hs paper s supported by Natonal Important specal project [ZX33-5]; NSF fundng of Chna [6793]; Jangsu Provnce Important specal project [BK7]; School project [NY33] 3 he auors - Publshed by Atlants Press 439

2 DD transmtter, S, and a DD recever, D) and N dle users (e,,,, N ) exst n a sngle cell[7] DD users can to relay and destnaton, whch are modeled as aylegh reuse bo e uplnk and downlnk resources of e cellular network However, ey have dfferent system performance If DD users reuse e uplnk resources, e cellular user wll nterfere to e DD recever and e base staton wll be affected by DD users at e same tme, whch are reflected n Fg- and Fg- On e contrary, f DD users reuse e downlnk resources, e DD user wll affect e recevng sgnal of e cellular user and e base staton wll also nterfere to e DD recever, whch s reflected n Fg-3 and Fg-4 From e above analyss, we know at reusng e uplnk resources outperforms e later w e followng reasons: If we reuse e uplnk resource, e nterference to DD user comes from cellular user However, e nterference manly comes from e base staton f we reuse e downlnk resource We also know at e base staton always has a larger transmt power an e moble user, us t wll cause more nterference to DD user So we reuse e uplnk resource n s paper What s more, to prevent e accumulated nterference, we assume DD pars reuse dfferent cellular resources random varables w varance and d SD respectvely d S s e pa-loss constant and d AB denotes e dstance between A and B In e second stage, e relayng nodes forward er receved sgnals to e destnaton node D he sgnal forwarded by relayng node can be wrtten as: YD P hd x I D' n (3) P s e transmtted power at relayng node and I D' s e nterference from cellular to D Assumng DF (Decode and Forward) mode s adopted n transmsson From equaton (), we can get e receved SN at destnaton when S drectly transmts sgnal to D wout relayng nodes: SNSD PS hsd I Dj N (4) We also assume e decode SN reshold s set to be If SNSD, we know at e channel between S and D s DD UE DD UE asssted by sutable relays to guarantee e servce qualty he relay selecton strategy s dscussed n e next secton DD UE DD UE Fg- Interference from cellular user to DD user Fg- Interference from DD user to base staton 3 elay Selecton Strategy In s secton, we propose several relay selecton strateges for e scenaro shown n Fg We take bo e pa-loss and nterference from cellular user nto consderaton here may be one or more relayng nodes accordng to specfc context he selecton process s descrbed as follows: () Frst, source S sends a flag sgnal (e, S) to base staton and destnaton W e receved sgnal, base staton and D can evaluate e channel coeffcent between S and em, whch can be wrtten as: hsb, hsd hsb and hsd are DD UE good enough to do e drect communcaton wout relyng nodes However, f SNSD, e communcaton must be DD UE DD UE DD UE Fg-3 Interference from DD user to cellular user Fg-4 Interference from base to DD user Each communcaton n DD system contans two stages In e frst stage, e source node S transmts ts sgnals (denoted by x ) to relayng nodes he receved sgnal at relayng node and destnaton node can be denoted by: YS PS hs x I j n S () YSD PS hsd x I Dj ' n () addtve whte nose,, d SD d SB d SB ( d SD ) s e dstance between S and base staton (D) From secton, we know at base staton bears nterference from DD user f we reusng e uplnk resource o guarantee e normal cellular communcaton, e nterference must be restrcted, whch means we must restrct e transmt power of source node If e nterfere reshold at base s, e I j and I Dj ' are e nterference from e cellular user to e - relay and destnaton respectvely node aylegh random varable w varance nterfere power receved at base s n s e PS, t be smaller an nterfere reshold So we can get PS s e transmtted power at source mean e largest transmt power of S s hs and hsd are e channel coeffcents from e source 44 PS, whch

3 P S (5) SB W equaton (5), e judgment condton of wheer a relay s needed can be rewrtten as: If SD j ( ID N) SD j ( ID N) (6) succeeds, S communcates w D wout any relayng nodes, f equaton (6) sn t satsfed, S broadcasts a flag O (e, equest of elay) () Upon recevng e sgnal O, each relay estmate er channel coeffcent to S, whch can be denoted by h S h s aylegh random varable w varance d S, f P (7) S S S e - node s not sutable to act as relay and backs off, oerwse e we add e nto canddate relay set C (3) Sort e canddate relays n C accordng to some crteron (whch wll be dscussed later) (4) Assume e fnal relay set s S C (ntalzed to be frst) Sequentally addng e nodes n C to C untl e accumulated SN at D s larger an decode reshold We dscuss e crteron n step (3) n detal here Sort accordng to e receved SN at D Each node n set C sends a flag to D and base staton, w s flag ey can estmate er channel coeffcent to, whch can be denoted by h B and h D Smlar to step (), we can estmate e maxmum transmt power of s: SN, at s P B So e receved SN at D h D (8) hb hen we can sort relays n C accordng to SN o smplfy e SN, we gnore e nterference from cellular user to D, but we take e nterference nto consderaton when e relays are determned Here we use a dstrbuted meod to sort ese nodes For each node n C, we set up a tmer w ntal value: D (9) B s a const w unt of tme [8] From (7), we know at e smaller s, e earler wll be selected out Sort accordng to e dstance between and D o sort nodes w meod above, we need to exchange a lot of flag sgnals between and base staton or D, whch brngs extra overhead to e system o smplfy e selecton process, we can sort nodes just accordng to er dstance to D We set a tmer w ntal value: he smaller d () D dd s, e earler wll be selected out Sort accordng to bo SN and network lfetme If a node s selected as relay at last round, ere s a bg possblty at t wll stll be a relay at next round he relay wll be prematurely depleted out due to excessvely act as relay So we need to take bo e SN and network lfetme nto consderaton Here we defne network lfetme as e duraton at all nodes can work normally he weghted metrc s set to: E E / h ( ) ( )( D ) () E s e remanng energy of, E s e ntal energy of each node s e weghed coeffcent Each node sets ts tmer w ntal value: () After step (3), we can get e fnal relay nodes set C o furer decrease e nterference between cellular user and D n e second stage, t s necessary to optmze e resources schedulng he meod s dscussed as follows: Select out a node n C, en s node sends a flag to all M cellular users, w s flag sgnal we can get e channel coeffcent between and all ese M cellular users esource of cellular user whose channel coeffcent s e worst to wll be reused by, us we can guarantee e nterference to be e smallest Assume e transmt power of cellular user s P, e nterference t produces to D can be wrtten as: I e Pe hed (3) 44

4 otal transmt power otal transmt power Network lfetme elay numbers It should be ponted out at all e channels are assumed to be symmetrcal epeatng s process untl all nodes n C are selected out 4 Smulaton and Analyss In s secton, we compare e performances w dfferent relay selecton strateges We assume e radus of e base staton s 5 meters and DD pars are located at half e radus to base staton Number of e dle users (e, ey are random dstrbuted at crcle doman) vares from 5 to 4, e decode reshold s set to be db Fg3 depcts e total power needed w dfferent relay selecton strateges From s pcture, we can see at to sort accordng to SN has e smallest transmt power, whch s followed by selecton accordng to dstance Here we assume ere s no deep fadng n e transmttng lnk (e, e paloss vares from 3 to 33); e requred total transmt power w consderaton of network lfetme s about % larger w respect to e "SN" selecton algorm We stll fnd e requred power of ese algorms decrease w e ncreasng of dle users; s s as possblty to select out a better relayng node ncrease w e ncreasng of dle user number Fg4 depcts e needed relay number for each selecton strategy Smlar to Fg3, selecton accordng to e receved sgnal needs e least relay numbers, e random selecton stll has e worst performance Fg5 depcts e network lfetme of each selecton strategy From s pcture, we know at SN-based selecton strategy has almost e same lfetme compared w random strategy; s s because e random strategy has e largest relay number, whch leads to a low lfetme of e whole network Lfetme of e strategy at take bo e remanng energy and SN s 5% larger an SN-based algorm Fg 6 shows e comparson of transmt power before and after resource schedulng We assume ere are 5 cellular users n e regon and e dstance between S and D s 4 meters, e teraton number s It s clear at after e resource schedulng e needed transmt power s reduced by % SN-based selecton dstance-based selecton random selecton SN+Lfetme selecton Number of dle users Fg4 elay number for each selecton strategy SN-based selecton SN+Lfetme selecton random selecton J 6J J 4J 8J Energy/J Fg5 Network lfetme of dfferent selecton strategy wout resource schedulng w resource schedulng 4 8 SN-based selecton dstance-based selecton random selecton SN+Lfetme selecton Number of dle users Number of dle users Fg3 ransmt power of dfferent selecton strateges Fg6 Comparson of transmt power w and wout resource schedulng 5 Conclusons In s paper, we propose ree relay selecton strateges and compare er system performance SN-based selecton strategy s better an dstance-based selecton strategy; however, e later outperforms e former one n terms of complexty herefore, n system w strngent energy 44

5 requrements, SN-based selecton strategy s superor Whle n e system at s more senstve to transmsson delay, e dstance-based selecton strategy s better; as e selecton process s very smple However, neer of e two algorms consders e lfetme of e system, s paper also propose a relay selecton, whch take bo SN and remanng energy nto consderaton; s meod can extend e lfetme of e entre system at e expense of e transmsson power, whch s very useful n energy-constraned termnals (e, moble phones) Fnally, we carred out resource schedulng for e above-descrbed relay algorms, whch can furer reduce e energy consumpton of e entre system Acknowledgements he auor would lke to ank Prof Shao and Prof Sun for many helpful dscussons and suggestons he correspondng auor of s paper s Shxang Shao eferences [] WANG Bn, CHEN Ln, ZHANG Xn, et al Devce-to- Devce communcaton as an underlay to LE advanced networks [J] Modern Scence & echnology Of elecommu ncatons,, (7) : 4 5 [] SENG V, YANIKOMEOGLU H, Falconer D D elayer selecton strateges n cellular networks w Peer -to-peer relayng[c] IEEE Vehcular echnology Conf, 3, (3) : [3] HASNA M O, ALOUINI M S Performance analyss of two-hop relayed transmssons over aylegh fadng cha nnels [C] IEEE Vehcular echnology Conf (VC),, (4): [4] SU W, IBAHIM A S, SADEK A K, et al Cooperatve communcatons w elay-selecton: when to cooperate and whom to cooperate w? [J] IEEE ransactons Wreless Communcatons, 8, 7(7): [5] CHEN Chun Kua, XIE Hong An mproved strategy of cooperatve relay node selecton [J] Journal of Harbn Engneerng Unversty,, () : 6-9 [6] FAEED M M, UYSAL M On relay selecton for decode and forward relayng [J] IEEE ransactons on Wreless Communcatons, 9, 8(7) : [7] Xran Ma, u Yn, Guandng Yu, Zhaoyang Zhang A Dstrbuted elay Selecton Meod for elay Asssted Devce-to-Devce Communcaton System [C] IEEE 3rd Internatonal Symposum on Personal Indoor and Moble ado Communcatons, : - 4 [8] A Bletsas, A Khst, D P eed, and A Lppman A Smple cooperatve dversty meod based on network pa selecton IEEE Journal on Selected Areas n Communcaton, 6, vol 4, no 3: [9] Dan Chen A novel mult-relay selecton and power allocaton optmzaton scheme n cooperatve networks[c] IEEE Wreless Communcatons and Networkng Conference, :