Cooperative Wireless Multicast: Performance Analysis and Power/Location Optimization

Transcription

1 88 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 6, JUNE Cooperatve Wreess Mutcast: Performance Anayss and Power/Locaton Optmzaton H. Vcky Zhao, Member, IEEE, and Wefeng Su, Member, IEEE Abstract The popuarty of mutmeda mutcast/broadcast appcatons over wreess networks makes t crtca to address the error-prone, heterogeneous and dynamcay changng nature of wreess channes. A promsng souton to combat channe fadng s to expore the cooperatve dversty n whch users may hep each other forward packets. Ths paper nvestgates cooperatve mutcast schemes that use a maxma rato combner to enhance the receved sgna-to-nose rato SNR, and provdes a thorough performance anayss. Two reay seecton schemes are consdered: the dstrbuted and the gene-aded cooperaton schemes. We derve the cosed-form formuaton and the approxmatons of ther average outage probabtes. We aso anayze the optma power aocaton and reay ocaton strateges, and show that aocatng haf of the tota transmsson power to the source mnmzes the average outage probabty. Our anayss and smuaton resuts show that cooperatve mutcast gves better performance when more reays hep forward sgnas. Cooperatve mutcast heps acheve dversty order, and user cooperaton can sgnfcanty reduce the outage probabty, especay n the hgh SNR regon. Fnay, we compare the two cooperaton strateges, and show that dstrbuted cooperatve mutcast s preferred snce t acheves a ower outage probabty wthout ntroducng extra overhead for contro messages. Index Terms Cooperatve wreess mutcast, cooperatve reayng, optmum power aocaton, reay ocaton optmzaton, outage probabty. I. INTRODUCTION WITH recent advance n communcatons, networkng and sgna processng technooges, we wtness the emergence of mutmeda broadcast and mutcast appcatons over wreess networks, where mutmeda data are devered to a group of users smutaneousy. Exampes ncude the Internet otoco teevson IPTV over WMax ] and mutmeda broadcast/mutcast servce MBMS wthn 3GPP ]. However, due to the error-prone and dynamcay changng characterstcs of wreess fadng channes, mutmeda mutcast over wreess medum s very chaengng. To further proferate mutmeda appcatons over wreess networks, t s of cruca mportance to combat nherent channe fadng, path oss, and shadowng effects n wreess channes n order to provde reabe and satsfactory servce. Varous spata/tempora/frequency dversty technques provde effectve soutons to enhance the reabty of wreess Manuscrpt receved September 3, 9; revsed January, ; accepted Apr 3,. The assocate edtor coordnatng the revew of ths paper and approvng t for pubcaton was G. Abreu. H. V. Zhao s wth the Department of Eectrca and Computer Engneerng, Unversty of Aberta, Edmonton, AB T6G V4 Canada e-ma: vzhao@ece.uaberta.ca. W. Su s wth the Department of Eectrca Engneerng, State Unversty of New York SUNY at Buffao, Buffao, NY 46 USA e-ma: wefeng@buffao.edu. Dgta Object Identfer.9/TWC /$5. c IEEE nks, n whch a destnaton receves mutpe dstorted versons of the orgna sgna and uses these sgnas coectvey to reduce detecton error rate and to mprove system performance. A we-known approach s to use mutpenput-mutpe-output MIMO technques to expot the spata dversty 3]. Recenty, an emergng concept - cooperatve communcaton - provdes a new communcaton vew, where users n a wreess network hep each other forward packets to mprove system performance 4]. Severa recent works 5], 6] addressed the nformaton-theoretc aspects of the cooperatve communcatons, whe many efforts have aso been focused on the desgn of cooperatve communcaton protocos to combat severe fadng n wreess channes. Specfcay, varous cooperaton protocos were proposed for wreess networks 7], 8] n whch a user/node may hep others to forward nformaton servng as a reay. As users or nodes n a wreess network cooperate wth each other by recevng and forwardng nformaton, each user may decode the receved nformaton and forward the decoded symbo, whch resuts n a decode-and-forward DF cooperaton protoco. Or t may smpy ampfy and forward the message, whch resuts n an ampfy-and-forward AF cooperaton protoco. The concept of user cooperaton was aso studed n 9], ] where a specfc two-user cooperaton scheme was nvestgated for CDMA systems and substanta performance gan was demonstrated wth comparson to the non-cooperatve approach. More recenty, the cooperatve communcaton protocos were further anayzed and generazed to mut-node scenaro see ] 3] and the references theren. The above pror works focused on optmzaton of cooperaton schemes for pont-to-pont communcatons where the prmary consderaton s for one ntended recever/user. There have been works on cooperatve mutcast/broadcast over wreess networks, where a group of ntended users receve the same data/vdeo servce from the source 4], 5]. To acheve user cooperaton n mutcast/broadcast appcatons, users who correcty decode the message sent by the source serve as reays and forward the message to others 6]. Same as the pont-to-pont cooperatve communcatons, focusng on the -hop transmsson of data, there are two stages phases n cooperatve mutcast/broadcast: the source e.g., the base staton transmts the message n the frst stage, whe the seected reays forward the message n stage. The work n 5] assumed that the seected reays transmt n orthogona channes, e.g., TDMA, FDMA or CDMA, and nvestgated the optma reay schedung and power aocaton strateges to mnmze the tota power consumpton. The authors n 7] consdered cooperatve mutcast of vdeos over wreess

2 ZHAO and SU: COOPERATIVE WIRELESS MULTICAST: PERFORMANCE ANALYSIS AND POWER/LOCATION OPTIMIZATION 89 networks. Same as n 5], 6], the reays used TDMA and took turns to forward packets, and ayered vdeo codng was used to provde users wth dfferent vdeo quaty dependng on ther channe condtons. The work n 7] studed the optma rate adaptaton and reay seecton strateges. Usng orthogona channes for dfferent reays enabes users n the mutcast/broadcast group to combne the sgnas receved from a nodes and to maxmze the receved sgnato-nose rato SNR, but t reduces the spectra effcency. To address the tradeoff between receved SNR and spectra effcency and to mnmze the contro overhead for reay coordnaton, the work n 8] adopted the dstrbuted cooperaton strateges proposed n 9]. In 8], a users who correcty decode the message transmtted by the source serve as reays, and a reays forward packets smutaneousy usng randomzed dstrbuted space-tme code RDSTC. In 7] 9], f an end user does not decode the message correcty n stage, t uses ony data receved n stage to decode the message. Note that utzng a receved sgnas, ncudng the one receved n stage, can hep further ncrease the sgna-tonose rato and mprove the system performance. In ths paper, we ncorporate the maxma rato combnng MRC technque ] nto cooperatve wreess mutcast schemes, and combne the sgnas receved n both stages to jonty detect the message at an ntended recever. We provde a thorough performance anayss for such MRC-based cooperatve mutcast schemes and optmze power aocaton between source and dynamc reays. Same as n 8], 9], we assume that a the seected reays transmt smutaneousy n stage. We consder two dfferent reay seecton schemes: the dstrbuted cooperaton strategy n 8], 9], and the one used n 7] wth a fxed number of reays at predetermned ocatons. For each scheme, we anayze ts average outage probabty and derve the cosed-form formuaton. We aso obtan an asymptotcay tght approxmaton for the average outage probabty to study the asymptotc behavor of cooperatve mutcast schemes n the hgh SNR regme. Based on the tght outage probabty approxmaton, we are abe to determne the optma power aocaton and reay poston strateges for cooperatve mutcast to mnmze the average outage probabty. Fnay, we compare the performance of dfferent mutcast schemes, ncudng the drect mutcast wthout cooperaton, the dstrbuted and the fxed cooperatve mutcast schemes, examne when users shoud cooperate wth each other, and study ther performance tradeoff. The rest of the paper s organzed as foows. Secton II ntroduces the drect mutcast scheme as we as the cooperatve mutcast schemes that we consder n ths paper. Secton III anayzes the average outage probabty of the mutcast schemes and derves the cosed-form formuaton. In Secton IV, we nvestgate the asymptotc behavor of the cooperatve mutcast schemes n the hgh SNR regon, and anayze the optma power aocaton and reay poston strateges. In Secton V, we compare the performance of dfferent mutcast schemes, examne ther tradeoff, and use smuaton resuts to vadate our theoretca anayss. Fnay, concusons are drawn n Secton VI. II. SYSTEM MODEL In ths paper, we consder a wreess network wth a crcuar ce of radus R. The base staton/access pont BS/AP s ocated at the center of the ce and mutcasts to M users who are unformy dstrbuted n the ce. For user M, the jont probabty densty functon of the user s dstance r from the BS/AP and the ange θ s fr,θ r /πr, wth r R and θ π. Themargna dstrbuton of r s fr r /R wth r R, θ s unformy dstrbuted n, π], andr and θ are ndependent of each other. We assume that {r,θ } M for dfferent users are ndependenty and dentcay dstrbuted..d.. We aso assume that the wreess nk between any two nodes s subject to ndependent narrow-band Rayegh fadng, propagaton path oss, and addtve whte Gaussan nose AWGN. A nodes n the network work n the haf-dupex mode, that s, they cannot transmt and receve n the same frequency band at the same tme. A. Drect Mutcast In the drect mutcast scheme, the BS/AP broadcasts a sgna x wth unt power. For user ocated at r,θ, ts receved sgna can be wrtten as y nc P h x n, where P s the transmsson power n the drect mutcast mode, h s the channe gan between the BS/AP and user, andη s the path oss parameter. Here, the subscrpt s the user ndex, and the superscrpt nc means no cooperaton drect mutcast. h s modeed as a zero-mean crcuary symmetrc compex Gaussan random varabe wth unt varance, and n s addtve whte Gaussan nose wth varance N. Therefore, the receved sgna-to-nose rato SNR n s γ nc P h /N, whch s an exponenta random N r η /P for a gven r. varabe wth parameter λ nc B. Cooperatve Mutcast In -hop cooperatve mutcast, users receve more than one copy of the message and expore cooperatve dversty to mprove system performance. It conssts of two stages: n the frst stage, the BS/AP broadcasts the message; and n the second phase, those reays who decode the message correcty n stage forward the message to other users. In ths work, we consder a repetton code n stage,.e., a reays forward the same message comng from the BS/AP smutaneousy n stage. Fnay, for those users who decode the message erroneousy n stage, they use MRC to combne the sgna from the BS/AP n stage and that from reays n stage, and jonty decode the message. We aso assume that a users are wng to cooperate, and there s no sefsh free rdng or macous behavor. We assume that a reays are synchronzed and the deay spread of arrvng sgnas s neggbe, whch s vad n narrow-band wreess communcatons.

3 9 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 6, JUNE Dstrbuted Cooperatve Mutcast Scheme: In the dstrbuted cooperatve mutcast scheme, the BS/AP frst broadcasts a unt-power sgna x. For user ocated at r,θ, ts receved sgna s y sd,d h x n wth SNR h, N where s the transmsson power used by the BS/AP n the dstrbuted cooperaton scheme, and other parameters are the same as n. Here, the subscrpt s the user ndex and the superscrpt sd,d means the source-destnaton channe n the dstrbuted cooperaton scheme. Snce h CN,, for any gven r, foows an exponenta dstrbuton wth parameter λ sd,d N r η /. Assume that N N M out of M users decode the message correcty n stage and C s {,, N } denotes the set ncudng ther ndces. C f {,,,M} C s { N,, M } contans the ndces of those who decode ncorrecty n stage. In stage, a users n C s serve as reays and broadcast the message x smutaneousy, usng the same power P r,d.forauser j C f, ts receved sgna y rd,d j n stage s the superposton of a the N sgnas broadcasted by the reays, subject to narrow-band Rayegh channe fadng, propagaton path oss and addtve whte Gaussan nose. So y rd,d j can be wrtten as y rd,d j C s P r,d r j η h j x n j, where r j r j r r j r cosθ j θ. 3 In 3, r j s the dstance between the reay C s and the user j C f. Here, the superscrpt rd,d means the reaydestnaton channe n the dstrbuted cooperaton scheme. We assume that the channe gans {h j } j C f, C s are..d. foowng crcuary symmetrc compex Gaussan dstrbuton CN, and n j are addtve whte Gaussan nose wth zero mean and varance N. Therefore, y rd,d j P r,d h j x n j, 4 where h N N j r j η h j CN, j. Gven {r,θ } and C S {,,, N }, the SNR of y rd,d j s γ rd,d j P r,d h /N j, whch s an exponenta random varabe wth parameter j ] N N r j η /P r,d. We assume that the channe gans h j and h j are known to user j C f, and user j uses MRC to combne y sd,d j n and y rd,d j n 3 and jonty decodes the message. Gven r,θ and C s, the SNR of the combned sgna s γ d j j γ rd,d j h j j P r,d h j N. 5 For far comparson, we et the average tota transmsson power used by the BS/AP and by a reays n the cooperatve mutcast be the same as the tota power used n the drect mutcast. That s, we seect and P r,d such that EN]P r,d P 3], n whch EN] s the expected number of users who decode the message correcty n stage and serve as reays. Gene-Aded Cooperatve Mutcast Scheme: The authors n 7] used a fxed number of reays at fxed postons to hep forward packets. Foowng the work n 7], we consder a gene-aded cooperatve mutcast scheme n ths paper, where we can put any number of reays to hep forward the message and the reays can be paced at optma ocatons to mnmze the outage probabty. Assume that there are N dedcated reays ocated on a fxed crce of radus R wth equa separaton ange, and the th N reay s ocaton s R, π/n.instage, the BS/AP broadcasts a unt-power sgna x, andforuser ocated at r,θ, ts receved sgna s wth SNR y sd,g h x n γ sd,g h. 6 N In 6, s the transmsson power used by the BS/AP n the gene-aded cooperatve mutcast scheme, and the rest terms are the same as n. Here, the subscrpt s the user ndex, and the superscrpt sd,g means the source-destnaton channe n the gene-aded cooperaton scheme. Gven r, γ sd,g s an exponenta random varabe wth parameter λ sd,g y sr,g N r η /. Fortheth reay N, gven R, ts receved sgna s R η h x n wth SNR h R η. 7 N γ sr,g For each reay, f t correcty decodes the transmtted sgna x n 7, then t w forward the decoded message n stage. Otherwse, the reay remans de n stage. Assume that N out of N reays decode the message correcty n stage, and RC s {,,, N } s the set ncudng ther ndces. In stage, reays,,, N transmt the message x smutaneousy usng the same power P r,g. Smar to the anayss of the dstrbuted cooperatve mutcast, for user at r,θ, n stage, ts receved sgna y rd,g s the superposton of a N reay messages and can be wrtten as y rd,g j RC s P r,g j h j x n, 8 where r j r R r R cos θ j π. N In 8, r j and h j are the dstance and the channe gan between user and the j th reay, respectvey. Here, the superscrpt rd,g means the reay-destnaton channe n the geneaded cooperaton scheme. Denote h j RC s j h j. We assume that a channe gans {h j } are..d. crcuary symmetrc compex Gaussan CN,, andn are addtve whte Compex Gaussan CN,N. Therefore, gven r,θ and the ndces of the reays who decode correcty n stage {,,, N }, h s aso a crcuary symmetrc compex

4 ZHAO and SU: COOPERATIVE WIRELESS MULTICAST: PERFORMANCE ANALYSIS AND POWER/LOCATION OPTIMIZATION 9 Gaussan wth zero mean and varance j RC s j. Thus, we have γ rd,g wth SNR y rd,g P r,g h x n γ rd,g P r,g h /N. 9 n 9 foows the exponenta dstrbuton wth parameter ] λ rd,g N j RC s j /P r,g. Assume that user has perfect knowedge of the channe gans h and h.gvenysd,g and y rd,g as n 6 and n 8, respectvey, user uses MRC to combne these two sgnas, and the SNR of the combned sgna s γ g γsd,g γ rd,g h P r,g h. N Smar to the dstrbuted cooperatve mutcast, gven R and N, we seect and P r,g such that EN]P r,g P, where EN] s the expected number of reays who decode correcty n stage. III. OUTAGE PROBABILITY ANALYSIS In ths secton, we anayze average outage probabtes for the two cooperatve mutcast schemes. The outage probabty s defned as the probabty that the maxmum mutua nformaton between the message transmtted by the BS/AP and the sgna receved by a user s smaer than a predetermned threshod T R, or equvaenty, the probabty that the receved SNR s beow a threshod γ 3]. If the receved SNR s hgher than the threshod γ, the user s assumed to be abe to decode the receved message wth a neggbe probabty of error. A. Outage obabty Anayss for Drect Mutcast For comparson, we frst derve outage probabty for the drect mutcast scheme. In the drect mutcast mode, for user at ocaton r,θ, the maxmum mutua nformaton between the nput x and the output y nc wth..d. crcuary symmetrc compex Gaussan nputs s I nc og γ nc, where γ nc foows the exponenta dstrbuton wth parameter λ nc N r η /P. Thus, gven a threshod T R on the maxmum mutua nformaton and user s ocaton r,user s outage probabty s ] I nc <T R r ] og h P <T R r exp N { TR N r η P }. For a network wth M unformy dstrbuted users, under the assumpton that a channe gans {h } are..d., the average outage probabty of drect mutcast s M P nc Inc <T R r ] M fr fr fr M dr dr M R I nc <T R r ] fr dr exp { TR N r η P } r dr R /η P ηr Γ N γ d η, Rη N γ d, P where γ d T R and Γa, x x e t t a dt s the ncompete Gamma functon. The second equaty n s due to the..d. assumpton of users ocatons {r }. B. Outage obabty Anayss for the Dstrbuted Cooperatve Mutcast Decodng Resuts n Stage : In dstrbuted cooperatve mutcast, for user ocated at r,θ, the SNR of ts receved sgna n stage,, s an exponenta random varabe wth parameter λ sd,d N r η /. Wth..d. crcuary symmetrc compex Gaussan nput x, the maxmum mutua nformaton between the nput and the output y sd,d n s I sd,d og γsd,d, where the normazaton factor / s due to the fact that the cooperatve mutcast scheme uses two tme sots to transmt one symbo ]. If I sr,d s arger than the threshod T R, we assume that user can decode the message wth a neggbe probabty of error n stage. Thus, gven r, the probabty that user decodes ncorrecty n stage s P,d { I sd,d <T R r exp γ mn r η }, 3 where γ m T R. Here, the subscrpt s the user ndex, and the superscrpt,d means the stage- decodng n the dstrbuted cooperatve scheme. We assume that a channe gans {h } are..d., so gven the ocatons of the M users {r,θ },,M, the probabty that users n C s {,, N } decode correcty and users n C f {,,,M} C s decode ncorrecty s P,d C s,c f r,,r M P,d C s { γ mn r η exp C s { exp j C f P,d j j C f } γ mn r η j }]. 4 Condtona Outage obabty: For user j C f who decodes ncorrecty n stage, gven ts ocaton {r,θ } and C s, the SNR of the maxmum-rato combned sgna s γ d j are exponenta random varabes wth parameters λ sd,d j N r η j / and j N C s r j η /,d, respectvey. For user j, the maxmum mutua nformaton wth..d. compex Gaussan nputs s ] j γ rd,d j n 5, where j I d j og γ d j og and γ rd,d j j γ rd,d j. 5 Thus, gven {r,θ } and C s, for user j C f, ts condtona outage probabty s P d j I d j <T R {r,θ },C s j γ rd,d j <γ m {r,θ },C s

5 9 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 6, JUNE j j N j λ sd,d j λ sd,d j λ sd,d j exp λ sd,d j γ m exp j γ m. 6 3 Average Outage obabty: Gven the ocatons of the M users {r,θ }, wefrst average P d j n 6 over a users j C f and a possbe decodng resuts C s n stage, whch s shown n 7. Here, the superscrpt d means the condtona outage probabty of the dstrbuted cooperatve scheme. In 7, the fourth equaty s due to the nonnegatvty of γ rd,d j and the fact that j γ rd,d j <γ m mpes j < γ m. The ast equaty comes from the..d. assumpton for channe gans. We then ntegrate P d n 7 wth respect to user ocatons {r,θ }, and the average outage probabty of the dstrbuted cooperatve mutcast scheme s n 8. To anayze the average outage probabty, we cacuate the frst term A n 8 as foows π A k <γ m {r,θ } 9 fr k,θ k dθ k dr k R { exp γ mn r η } rk k R dr k /η Ps,d ηr Γ N γ m η, Rη N γ m, where Γa, x s the ncompete Gamma functon. Apparenty, A depends ony on the base staton s transmsson power. It s the same for a possbe decodng resuts n stage and for dfferent user ndex k C f. Then, for a gven decodng resut n stage C s and user j C f, we cacuate the second term n 8, whch s shown n. From, for a gven C s, f we consder two dfferent users j and k who decode ncorrecty n stage, we have B N C s, j B N C s, k. In addton, from, f we fx the number of users who decode correcty n stage as N, for two dfferent decodng resuts Cs,C f and C s,c f where Cs Cs N, we can show that k C B f N Cs, k j C B f N CS, j. Gven the tota number of users M, fora fxed number of users who serve as reays N, there are a tota of M N possbe decodng resuts n stage wth CS N. Wthout oss of generaty, we use C s N {,,,N} and j N C f N as an exampe, and defne B N BN C sn {,,,N}, j N. Therefore, we have C s N j C f B N C s, j M N M NBN. Based on the above anayss, P d n 8 can be wrtten as P d A ] M M M N M B N A ] M N M N N A ] M M M B N A ] M N. N The frst term n corresponds to the scenaro where a users decode ncorrecty n stage, and thus the condtona outage probabty P d j s for a users. The second term n cacuates the average outage probabty when N users decode correcty n stage for N M. Note that when N M, that s, a users decode correcty n stage, the condtona outage probabty s. Thus, does not ncude the term correspondng to N M. 4 Power Constrant: In dstrbuted cooperatve mutcast, the average transmsson power used by the base staton and a reays are EN]P r,d. Wth..d. channe gans, we have EN] M M π I sd >R r,θ ] fr {,θ dθ dr exp γ mn r η } r dr R M A ]. For far comparson, we seect and P r,d such that EN]P r,d P,whereP s the transmsson power used n drect mutcast to transmt one message. Therefore, gven the BS/AP s transmsson power and the tota number of users M, the transmsson power used by each reay s P r,d P M A ]. 3 Note that P r,d depends on the average not the nstantaneous number of users who receve correcty n stage. It can be precacuated and remans the same durng the transmssons. To summarze, for the dstrbuted cooperatve mutcast, gven the tota number of users M and the base staton s transmsson power, the transmsson power used by each reay shoud be determned by 3, and the average outage probabty can be cacuated usng. C. Outage obabty Anayss for the Gene-Aded Cooperatve Mutcast The anayss of the outage probabty for the gene-aded cooperatve mutcast s smar to that n the prevous secton, and we w present the man resuts by omttng some detas. Decodng Resuts n Stage : Smar to that n Secton III-B, n stage of the gene-aded cooperatve mutcast, for user ocated at r,θ, the SNR of ts receved sgna s γ sd,g, whch foows the exponenta dstrbuton wth parameter λ sd,g N r η /. For the N dedcated reays, from 7, gven R,theSNR of the receved sgna by the th reay n stage γ sr,g s an exponenta random varabe wth parameter λ sr,g N R η /. Same as dstrbuted cooperatve mutcast, we assume that reay can decode the message correcty when the maxmum mutua nformaton I sr,g og γsr,g s arger than the predetermned threshod T R.Otherwse,the reay s assumed to decode ncorrecty n stage. Thus, gven R, the probabty that reay decodes erroneousy n stage s P,g I sr,g <T R R ] ] og γ sr,g <T R R

6 ZHAO and SU: COOPERATIVE WIRELESS MULTICAST: PERFORMANCE ANALYSIS AND POWER/LOCATION OPTIMIZATION 93 P d P d M j P,d C s,c f {r,θ } C s j C f M j γ rd,d j <γ m Cs, {r,θ } M M M C s C s C s C s j C f j C f j C f j C f >γ m C s, k <γ m k C f {r,θ } j γ rd,d j <γ m,γ sd j <γ m, >γ m C s, γ sd j <γ m k C f and k j {r,θ } j γ rd,d j <γ m, >γ m C s, j <γ m k C f and k j {r,θ } j γ rd,d j <γ m Cs, {r,θ } >γ m {r,θ } C s k <γ m {r,θ }. 7 k C f, k j ] P d M M M P d fr,θ fr M,θ M dr dθ dr M dθ M N C s N j C f k C f, k j P d j fr j,θ j C s π k <γ m {r,θ } fr k,θ k dθ k dr k }{{} >γ m {r,θ } A ] ] fr,θ d dθ dr N dθ N dr j dθ j }{{} M N C s N B N C s, j j C f A ] M N B N C s, j. 8 B N C s, j C s γ rd,d j <γ m Cs, {r,θ } fr j,θ j >γ m {r,θ } fr,θ dr dθ dr N dθ N dr j dθ j. j { exp γ mn R η }. 4 }{{} CR Here, the subscrpt s the reay ndex, and the superscrpt,g means the stage- decodng n the gene-aded cooperaton scheme. Assume that N out of N reays decode correcty n stage, and RC s {,, N } contans ther ndces. Thus, the probabty that reays n RC s decode correcty n stage s P,g RC S P,g P,g j RC s j RC s CR ] N CR N N. 5 Outage obabty: After maxmum-rato combnng, for user at r,θ, the SNR of the combned sgna s γ sd,g γ rd,g,whereγ sd,g random varabes wth parameter λ sd,g γ g and γ rd,g are exponenta N r η / and ] N j RC s j /P r,g, respectvey. For user λ rd,g, the maxmum mutua nformaton wth..d. compex Gaussan nputs s I g og γ g, and the condtona outage probabty s P g I g <T R r,θ,rc s ] ] P γ sd,g γ rd,g <γ m r,θ,rc s λ rd,g exp λ sd,g γ m λ rd,g λ rd,g λ sd,g λ sd,g λ sd,g exp λ rd,g γ m. 6

7 94 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 6, JUNE Note that when RC s and N,thats,aN reays decode ncorrecty n stage, then γ rd,g and user s outage probabty s γ sd,g <γ m r,θ { exp γ mn r η } A, 7 where A s defned n 9. Averagng P g over a possbe decodng resuts n stage RC s and user s ocaton r,θ, the average outage probabty of gene-aded cooperatve mutcast s n 8. In 8, CR and P g are defned n 4 and 6, respectvey. The superscrpt g means the gene-aded cooperaton scheme. In 8, the frst term corresponds to the scenaro where a reays decode ncorrecty n stage, and the second term consders the scenaro where there are N N reays that decode correcty n stage. 3 Power Constrant: Same as n Secton III-B4, n the gene-aded cooperatve mutcast, we adjust and P r,g such that EN]P r,g P,whereEN] s the average number of reays who decode correcty n stage. From 4, EN] N P,g N CR. Therefore, gven the base staton s transmsson power and R,wehave P r,g P N exp γ m N R η /. 9 Same as P r,d n dstrbuted cooperatve mutcast, P r,g depends on the average number of reays who decode correcty n stage, and can be precacuated. To summarze, for the gene-aded cooperatve mutcast, gven the tota number of reays N and the reay ocaton R, the transmsson power used by the base staton and that used by each reay shoud satsfy 9, and the average outage probabty s gven by 8. IV. OPTIMUM POWER ALLOCATION AND RELAY LOCATION Ths secton consders optma power aocaton for cooperatve mutcast schemes to mnmze ther outage probabty. For gene-aded cooperatve mutcast, we aso nvestgate the optma reay ocaton R to maxmze the system performance. A. Dstrbuted Cooperatve Mutcast In dstrbuted cooperatve mutcast, gven the power constrant n 3, we ook for the optma that mnmzes the average outage probabty n. Due to the compexty of, t s dffcut to fnd a cosed-form souton for the optma drecty from. In the foowng, we consder a hgh SNR scenaro. We obtan an asymptotcay tght approxmaton of to study ts asymptotc behavor n the hgh SNR regon, and derve an asymptotcay optma power aocaton scheme. Approxmaton of the Average Outage obabty: We frst anayze the term A n. From the defnton n 9, { A exp γ mn r η } rk k R dr k N γ m r η r k P k s,d R dr k N γ m R η η. 3 Here, we use the frst order Tayor seres approxmaton expx x for x cose to, whch s tght at hgh SNR and when the rato /N s arge. Smary, to smpfy the term B N n, gven C s {,,,N}, wth hgh SNR where /N and P r,d /N arearge,weusethefoowng approxmatons: N λsd,d N N λsd,d N λ sd,d N N λsd,d N ] N γ m {r,θ },C s exp λ sd,d N γ m exp N γ m λrd,d N λsd,d N γ m, 3 and λ sd,d >γ m r,θ { exp γ mn r η } γ mn r η 3 for N. Therefore, we have 33. Note that n 33, D N depends ony on N and R but not other parameters. Substtutng 3 and 33 nto, we have P d N γ m R η M η M M N γm N P r,d N N γ m R η M N D N. 34 η Note that n 34, the owest order of N / and N /P r,d s when N M. Therefore, for hgh SNR wth arge vaues of /N and P r,d /N, we can gnore the thrd and a other hgher order terms and further smpfy P d as P d P d N γm D M, 35 P r,d where D M s a constant and depends ony on the tota number of user M and the radus R. Optma Power Aocaton: Gven the above approxmaton of the average outage probabty, we can determne the optma power aocaton between the BS/AP and the reays. From 35, we have P d N γ D M / P r,d, where D M depends ony on the tota number of users M. Therefore, mnmzaton of P d s equvaent to maxmzaton of the product P r,d under the constrant that and P r,d satsfy 3.

8 ZHAO and SU: COOPERATIVE WIRELESS MULTICAST: PERFORMANCE ANALYSIS AND POWER/LOCATION OPTIMIZATION 95 P g π N P g P,gRC S fr,θ dθ dr RC s CR ] N CR ] N N N RC s N CR N A N N π P g r πr dθ dr π P g r RC πr s N CR ] N CR ] N N dθ dr, 8 B N λrd,d N λsd,d N γ m r r N πr dr dθ dr N dθ N N N N r η N N r N η γm r r N P r,d πr dr dθ dr N dθ N N N γ m D N, P r,d N where D N r rn r r N cosθ θ N η/ r η N r r N πr N dr dθ dr N dθ N. 33 ] To smpfy the power constrant 3, we need to smpfy the denomnator n 3. From 3, wth hgh SNR and arge vaues of /N, we can use the approxmaton A N γ m R η η where bn γ m R η /η. Therefore, we have b, 36 P P r,d M A ] P M b/ P. 37 M b Consequenty, to mnmze the outage probabty of dstrbuted cooperatve mutcast, we shoud seect to maxmze G d p p P p/mp b]. To fnd the optma,weet Gd p p p p P 3bp 4bP pp Mp b, s,d pp s,d Ps,d P 3b 4bP, Ps,d P 3b ± P 3b 3bP Wth hgh SNR and P b, the optma souton n 38 can be approxmated as Ps,d P, that s, the BS/AP uses haf of the tota transmsson power. The other haf s eveny dstrbuted among reays n the statstca sense, and we use the average number of reays EN] and 3 to cacuate the transmsson power used by each reay P r,d. B. Gene-Aded Cooperatve Mutcast To anayze the optma power aocaton and reay ocatons for the gene-aded cooperatve mutcast, smar to the prevous secton, we frst consder the hgh SNR scenaro wth arge transmsson power and fnd an approxmaton of the average outage probabty. Then, we use the approxmated outage probabty to fnd the optma power aocaton and reay ocaton strateges. Approxmaton of the Average Outage obabty: For agvenr, for hgh SNR wth arge /N, we start wth the frst term correspondng to N n 8, and use 3 to approxmate t as CR N A exp γ mn R η ] N A γm N R η N N γ m R η η N N γ m R N R η. 39 η To smpfy the second term n 8 correspondng to the scenaro where N N,wefrst need to fnd approxmatons for P g n 6. Smar to the approxmaton of 3 n the prevous Secton, for hgh SNRs, we can approxmate P g as ] P g γ sd,g γ rd,g<γm r,θ,rc s 4 λ rd,g exp λ sd,g γ m λ rd,g λ rd,g λ sd,g λ sd,g λ sd,g exp λ rd,g γ m λrd,g λ sd,g γm N γ m rη P r,g j j RC s In addton, wth arge /N, CR exp γmnrη γ m N R η / << and we can have.

9 96 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 6, JUNE N N N N N CR ] N CR ] N N γm N R η N N RC s N γm N R η N N γ m N P r,g N π where Z N RC s N r η j j RC s π P g RC πr s N π Z N, r πr N γ m rη P r,g dθ dr j j RC s r πr dθ dr dθ dr. 4 CR. Therefore, the second term n P g can be approxmated as n 4. Note that Z N depends on the reay ocaton R and the reay s decodng resut n stage RC s, but not the transmsson powers. Combnng the resuts n 39 and 4, the average outage probabty of the gene-aded cooperatve mutcast scheme s N P g N γ m R N R η 4 N N γ m N P r,g η γm N R η N N Z N. Same as n 34, the owest order of N / and N /P r,g n 4 s when N N. Thus, for hgh SNR wth arge transmsson powers, we can further smpfy the expresson n 4 by omttng the thrd and hgher order terms, and we have P g P g N γm Z N. 43 P r,g Here, Z N depends on the tota number of reays N and ther ocatons R but not the transmsson powers. Optma Power Aocaton: To fnd the asymptotc optma power aocaton scheme, for a gven reay ocaton R, smar to the anayss n Secton IV-A, we fnd the optma that mnmzes the approxmated average outage probabty P g n 43. Note that n 43, Z N does not depend on. Thus, g to mnmze P N γ m Z N /P r,g, t s equvaent to maxmze the product P r,g under the constrant that and P r,g satsfy 9. In the hgh SNR regon wth arge /N, we can have the approxmaton exp γ m N R η / γ m N R η /. Therefore, foowng the same anayss as n Secton IV-A, to mnmze the outage probabty, we shoud seect to maxmze G g p p P p/n p b ], whereb γ m N R η.that s, the optma Ps,d shoud satsfy Gg p p,whch pp s,g gves Ps,g P 3b ± P 3b 3b P 4. In the hgh SNR regon where P b, we can further have the approxmaton that Ps,g P, that s, the BS/AP uses haf of the tota transmsson power. The rest haf s eveny dstrbuted among reays n the statstca sense, and 9 s used to cacuate the transmsson power used by each reay P r,g. TABLE I ASYMPTOTIC OPTIMAL RELAY LOCATIONS FOR THE GENIE-AIDED COOPERATIVE MULTICAST SCHEME. Optma Reay N N 3 N 4 N 5 N 6 Locaton u R /R Optma Reay Locaton: To fnd the asymptotc optma reay ocaton R, wth hgh SNR, foowng the above dscusson on the optma power aocaton, we et P, and P r,g P P N b N P b, 44 where b γ m N R η. Thus, gven the number of reays N, we seect R to mnmze P g γmn Z N γmn N P r,g P 3 P b Z N, 45 or equvaenty, to mnmze P b Z N. Wth hgh SNR and P γ m N R η >b, the probem can be smpfed to seect R to mnmze Z N. Wth r R and R R, et t r /R, ] and u R /R, ]. Thus, Z N n 4 can be rewrtten as n 46, and mnmzaton of Z N s equvaent to mnmzaton of z N. Note that z N s a functon of the number of reays N and the normazed reay poston u, but not other parameters. For each N, Monte Caro and numerca methods are used to fnd the optma u that mnmzes z N. Tabe I sts the optma u when N takes dfferent vaues from to 6. V. SIMULATION RESULTS AND PERFORMANCE COMPARISON A. Dstrbuted Cooperatve Mutcast We frst use Monte Caro smuatons to cacuate the exact and the approxmated average outage probabtes of the dstrbuted cooperatve mutcast, and compare them n Fg.. We assume that the radus of the crce s R, the path oss parameter s η.6, and the outage threshod s T R 4. We consder three cases wth M, M and M 5 users, respectvey, and et P/N vary from 75dB to 95dB. We et the transmsson power used by the BS/AP equa to the average transmsson power used by a reays. We use, Monte Caro smuatons to fnd B N

10 ZHAO and SU: COOPERATIVE WIRELESS MULTICAST: PERFORMANCE ANALYSIS AND POWER/LOCATION OPTIMIZATION 97 Z N π N r η π N t η R η r R r π R cosθ N η/ r πr dθ dr η/ t u π tu cosθ N t π dθ dt. 46 }{{} z N Average Outage obabty 3 4 M M Exact souton, M Approxmaton, M Exact souton, M Approxmaton, M Average Outage obabty 3 4 Exact souton, M5 Approxmaton, M P/N db P/N db Fg.. Comparson of the exact average outage probabty P d and the approxmaton P d for the dstrbuted cooperatve mutcast. R, η.6, and T R 4. P. Average Outage obabty 3 P/N 75 db P/N 85 db P/N 95 db Average Outage obabty 3 4 M M5 M Anaytca resuts, M Smuaton resuts, M Anaytca resuts, M Smuaton resuts, M Anaytca resuts, M5 Smuaton resuts, M α / P Fg.. Optma power aocaton n the dstrbuted cooperatve mutcast. R, η.6, T R 4,andM. P/N 75dB, 85dB, 95dB. The x axs s α /P. n P d for N M and D M n P d.fromfg., we can see that P d matches P d very we, especay when P/N s arge. We observe smar trend for other vaues of the parameters. Fgure pots the exact average outage probabty P d versus dfferent power aocaton strateges α /P P/N db Fg. 3. Smuaton resuts of the dstrbuted cooperatve mutcast schemes. R, η.6, T R 4,and P. n the dstrbuted cooperatve mutcast scheme. In Fg., we use M wth P/N 75dB, 85dB, and95db as an exampe, and we observe the same trend for other vaues of the parameters. From Fg., settng α to be around.5 can hep mnmze the average outage probabty, especay wth hgh SNR. Next, we use smuaton resuts to vadate our anayss

11 98 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 6, JUNE Exact souton, N Approxmaton, N Exact souton, N 4 Approxmaton, N 4 Average Outage obabty 3 4 Average Outage obabty P/N db a: N reays P/N db b: N 4reays Fg. 4. Comparson of the exact and the approxmated average outage probabtes for the gene-aded cooperatve mutcast scheme. R, η.6, and T R 4. P,andR R /5. Average Outage obabty 3 4 R /R. R /R.5 P/N 75dB R /R.8 P/N 85dB Average Outage obabty 3 4 Reays 4 Reays 6 Reays P/N 75dB P/N 85dB P/N 95dB α /P a 6 P/N 95dB ur /R b Fg. 5. a: Optma power aocaton n gene-aded cooperatve mutcast wth N 4. The x axs s α /P. b: Asymptotc optma reay postons for the gene-aded cooperatve mutcast scheme wth P and N, 4, 6. Inbothaandb,R, η.6, T R 4,and P/N 75dB, 85dB, 95dB. of the dstrbuted cooperatve scheme s outage probabty. In our smuatons, we randomy generate M user ocatons unformy dstrbuted nsde a crce wth radus R. A channe gans are generated ndependenty foowng the compex Gaussan dstrbuton CN,. Weet P and aocate haf of the tota transmsson power to the BS/AP. Fgure 3 compares the smuaton resuts based on 7 smuaton runs wth the anaytca resuts n Secton III. It can be seen that the smuaton resuts match our anaytca resuts very we. In addton, from Fg. 3, dstrbuted cooperatve mutcast acheves a smaer outage probabty when the tota number of users M s arger and when more users hep reay the message n stage. For exampe, there s a 3dB gan f M s ncreased from to, and another db gan f M s further ncreased to 5. Therefore, dstrbuted cooperatve mutcast performs better n denser networks wth more users. B. Gene-aded Cooperatve Mutcast For the gene-aded cooperatve mutcast scheme, Fg. 4 g compares the approxmated outage probabty P wth the exact P g, both cacuated usng Monte Caro smuatons. The system setup s smar to that n Fg.. We use P and R R / as an exampe, and we observe the same trend for other power aocaton and reay ocatons. We consder two scenaros wth N and N 4dedcated reays, respectvey, and other number of N gves the same trend. It can be seen that the approxmaton n 43 s very cose to the exact outage probabty, especay n the hgh SNR regon.

12 ZHAO and SU: COOPERATIVE WIRELESS MULTICAST: PERFORMANCE ANALYSIS AND POWER/LOCATION OPTIMIZATION 99 Average Outage obabty 3 4 Anaytca, reays Smuaton, reays Anaytca, 4 reays Smuaton, 4 reays Anaytca, 6 reays Smuaton, 6 reays Average Outage obabty 3 4 O: Gene aded, reays dstrbuted, M5 x: Gene aded, 6 reays Drect mutcast Gene aded, reays Gene aded, 6 reays Dstrbted, M Dstrbuted, M Dstrbuted, M5 drect mutcast 5 5 dstrbuted, M dstrbuted, M P/N db Fg. 6. Smuaton resuts of the gene-aded cooperatve mutcast schemes. R, η.6, T R 4,and P. The radus R s seected accordng to Tabe I. For gene-aded cooperatve mutcast, Fg. 5a pots the exact average outage probabty P g n 8 for dfferent power aocaton strateges α /P. The system setup n Fg. 5a s smar to that n Fg. 4, and we use N 4reays wth R.R, R.5R,andR.8R as an exampe. We observe a smar trend for other vaues of the parameters. From Fg. 5a, aocatng haf of the transmsson power to the BS/AP heps mnmze the average outage probabty, especay when P/N s arge. Fgure 5b shows the exact average outage probabty P g for dfferent reay postons and dfferent SNRs. We can see from Fg. 5b that n the hgh SNR regon, the seected u n Tabe I heps mnmze the outage probabty. Note that we focus on the hgh SNR regon n ths work, and our anayss of optma power aocaton and reay poston s based on our approxmated outage probabty P g, whch matches the exact vaue P g very we ony n the hgh SNR regon e.g., when P/N 85dB. Therefore, the optma power aocaton and reay postons n the ow SNR regon e.g., when P/N 75dB may be dfferent from our anayss n Secton IV-B; whe our smuaton resuts are consstent wth our anayss n the hgh SNR regon e.g., when P/N 85dB and P/N 95dB. We then use smuatons resuts to vadate our anayss on the outage probabty of the gene-aded cooperatve mutcast scheme. Fgure 6 compares the smuaton resuts based on 7 smuaton runs wth the anaytca resuts n Secton III. The smuaton setup s the same as n Fg. 3, and we foow Tabe I to set the reay ocaton R. The smuaton resuts match our anaytca resuts very we, and the gene-aded cooperatve mutcast scheme heps reduce the outage probabty by a arge amount when the number of dedcated reays ncreases, especay n the hgh SNR regon. In Fg. 6, we observe a maxmum of db gan when N s ncreased from to 6. C. Performance Comparson We then compare the outage probabty of dfferent mutcast schemes wth dfferent SNRs, as shown n Fg. 7. The system setup n Fg. 7 s the same as that n Fg P/N db Fg. 7. Performance comparson of dfferent mutcast schemes. R, η.6, T R 4,and P. In the gene-aded cooperatve mutcast, the reay postons are seected based on the resuts n Tabe I. and 6. Comparng the performance of dstrbuted cooperatve mutcast wth that of drect mutcast, we observe that n the hgh SNR regon, user cooperaton can sgnfcanty hep reduce the outage probabty. For exampe, wth P/N 95dB, dstrbuted cooperatve mutcast can hep reduce the average outage probabty from O 4 to O 6 when there are hundreds of users n the network. We aso observe that the smuaton curves of dfferent cooperatve mutcast schemes go n parae n hgh-snr regons and they show a hgher dversty order than the drect mutcast scheme. Ths s consstent wth the theoretca anayss n, 35 and 43, whch show that the drect mutcast scheme provdes ony dversty order and the cooperatve mutcast schemes offer dversty order n the outage probabty performance. Furthermore, we observe that cooperaton does not aways gve the best performance. Wth the system setup as n Fg. 7, for the dstrbuted cooperaton strategy, when there are ten users n the network, drect mutcast s benefca when the SNR P/N s beow 8dB; and wth M users, user cooperaton gves a smaer outage probabty when SNR s arger or equa to 76dB. Smary, the gene-aded cooperaton scheme reduces the outage probabty ony when P/N s arger than 8dB. When comparng the two cooperatve mutcast schemes, for a sparse network wth fewer users, when the number of dedcated reays n gene-aded cooperatve mutcast N s sma, the two cooperatve mutcast schemes have smar performance. For denser networks wth arge number of users, the dstrbuted cooperatve mutcast scheme acheves a smaer outage probabty and has a db to 4dB gan compared wth gene-aded cooperatve mutcast. Note that the geneaded cooperatve mutcast scheme rees on the exstence of dedcated reays that we can put n any ocatons. Therefore, dstrbuted cooperatve mutcast s often preferred snce t gves a better performance wthout the hep of dedcated reays. Another advantage of the dstrbuted cooperaton scheme s that t s easy to mpement and does not ntroduce extra communcaton overhead for contro messages.

13 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 6, JUNE VI. CONCLUSIONS In ths paper, we nvestgated two MRC-based cooperatve mutcast schemes over wreess networks, where a group of users receve the same data from a BS/AP and they may cooperate wth each other to combat channe fadng and to acheve more reabe QoS. We anayzed the outage probabty performance of the two cooperatve mutcast schemes and optmzed ther power aocaton. We frst anayzed the outage probabtes of the dstrbuted and the gene-aded cooperatve mutcast schemes. We derved cosed-form formuatons for the outage probabtes, and provded approxmatons to show the asymptotc performance of the cooperatve mutcast schemes. Based on the asymptotcay tght outage probabty approxmatons, we obtaned the optma cooperaton strateges. It turns out that aocatng haf of the tota transmsson power to the BS/AP mnmzes the outage probabty of cooperatve mutcast schemes, and the other haf s eveny dstrbuted among reays n the statstca sense. We aso determned the optma reay ocatons for gene-aded cooperatve mutcast. We then compared the performance of dfferent mutcast schemes. For the dstrbuted cooperatve scheme, we observe a smaer outage probabty for denser networks wth more users and a arger average number of reays. Smary, the outage probabty of the gene-aded cooperaton scheme decreases as the number of dedcated reays ncreases. Compared to the drect mutcast scheme, the cooperatve mutcast schemes acheve dversty order, and user cooperaton can hep sgnfcanty mprove the performance especay when the sgna-to-nose rato s hgh. Compared wth the geneaded scheme, the dstrbuted cooperaton scheme gves a db to 4dB performance gan and, therefore, t s often preferred to maxmze the performance wthout the hep of dedcated reays and wthout extra overhead for contro sgnas. REFERENCES ] J. She, P. Ho, and L. Xe, IPTV over WMax: key success factors, chaenges, and soutns, IEEE Commun. Mag., vo. 45, no. 8, pp , Aug. 7. ] S. Kota, Y. Qan, E. Hossan, and R. Ganesh, Advances n mobe mutmeda networkng and QoS, IEEE Commun. Mag., vo. 45, no. 8, pp. 5 53, Aug. 7. 3] A. Godsmth, Wreess Communcatons. Cambrdge Unversty ess, st edton, 5. 4] K. J. R. Lu, A. Sadek, W. Su, and A. Kwasnsk, Cooperatve Communcatons and Networkng. Cambrdge Unversty ess, st edton, 9. 5] T. M. Cover and A. A. E Gama, Capacty theorems for the reay channe, IEEE Trans. Inf. Theory, vo. 5, pp , Sep ] A. A. E Gama and S. Zahed, Capacty of a cass of reay channes wth orthogona components, IEEE Trans. Inf. Theory, vo. 5, pp , May 5. 7] J. N. Laneman and G. W. Worne, Dstrbuted space-tme-coded protocos for expotng cooperatve dversty n wreess networks, IEEE Trans. Inf. Theory, vo. 49, pp , Oct. 3. 8] J. N. Laneman, D. N. C. Tse, and G. W. Worne, Cooperatve dversty n wreess networks: effcent protocos and outage behavor, IEEE Trans. Inf. Theory, vo. 5, pp , Dec. 4. 9] A. Sendonars, E. Erkp, and B. Aazhang, User cooperaton dversty part I: system descrpton, IEEE Trans. Commun., vo. 5, pp , Nov. 3. ] A. Sendonars, E. Erkp, and B. Aazhang, User cooperaton dversty part II: mpementaton aspects and performance anayss, IEEE Trans. Commun., vo. 5, pp , Nov. 3. ] W. Su, A. Sadek, and K. J. R. Lu, Cooperatve communcaton protocos n wreess networks: performance anayss and optmum power aocaton, Wreess Persona Commun., vo. 44, no., pp. 8 7, 8. ] K. G. Seddk, A. K. Sadek, W. Su, and K. J. R. Lu, Outage anayss and optma power aocaton for mut-node reay networks, IEEE Sgna ocess. Lett., vo. 4, no. 6, pp , June 7. 3] A. K. Sadek, W. Su, and K. J. R. Lu, Mut-node cooperatve communcatons n wreess networks, IEEE Trans. Sgna ocess., vo. 55, no. 6, pp , Jan. 7. 4] I. Marc and R. D. Yates, Cooperatve muthop broadcast for wreess networks, IEEE J. Se. Areas Commun., vo., no. 6, pp. 8 88, Aug. 4. 5] B. Mergen and A. Scagone, On the power effcency of cooperatve broadcast n dense wreess networks, IEEE J. Se. Areas Commun., vo. 5, no., pp , Feb. 7. 6] Q. Zhang and F. Ftzek, Cooperatve retransmsson for reabe wreess mutcast servce, n Cogntve Wreess Networks, F. Ftzek and M. Katz eds. Cambrdge Unversty ess, pp , 7. 7] O. Aay, T. Koraks, Y. Wang, E. Erkp, and S. Panwar, Layered wreess vdeo mutcast usng omn-drectona reays, IEEE Int. Conf. on Acoustcs, Speech and Sgna ocessng ICASSP, Apr. 8. 8] O. Aay, P. Lu, Z. Guo, L. Wang, Y. Wang, E. Erkp, and S. Panwar, Cooperatve ayered vdeo mutcast usng randomzed dstrbuted space-tme codes, n oc. IEEE Infocom MOVID Workshop, Apr. 9. 9] B. Mergen and A. Scagone, Randomzed spact-tme codng for dstrbuted cooperatve communcaton, IEEE Trans. Sgna ocess., vo. 55, no., pp , Oct. 7. ] D. G. Brennan, Lnear dversty combnng technques, oc. IEEE, vo. 9, pp , Feb. 3. ] T. Hunter and A. Nosratna, Dversty through coded cooperaton, IEEE Tran. Wreess Commun., vo. 5, no., pp , Feb. 6. H. Vcky Zhao M 5 receved the B.S. and M.S. degree from Tsnghua Unversty, Chna, n 997 and 999, respectvey, and the Ph. D degree from Unversty of Maryand, Coege Park, n 4, a n eectrca engneerng. She was a Research Assocate wth the Department of Eectrca and Computer Engneerng and the Insttute for Systems Research, Unversty of Maryand, Coege Park from Jan. 5 to Juy 6. Snce August 6, she has been an Assstant ofessor wth the Department of Eectrca and Computer Engneerng, Unversty of Aberta, Edmonton, Canada. Dr. Zhao s research nterests ncude nformaton securty and forenscs, mutmeda soca networks, dgta communcatons and sgna processng. Dr. Zhao receved the IEEE Sgna ocessng Socety SPS 8 Young Author Best Paper Award. She co-authored the book Mutmeda Fngerprntng Forenscs for Trator Tracng Hndaw, 5. She s the Assocate Edtor for IEEE SIGNAL PROCESSING LETTERS and Esever s JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION. Wefeng Su M 3 receved the Ph.D. degree n Eectrca Engneerng from the Unversty of Deaware, Newark n. He receved hs B.S. and Ph.D. degrees n Mathematcs from Nanka Unversty, Tanjn, Chna, n 994 and 999, respectvey. Hs research nterests span a broad range of areas from sgna processng to wreess communcatons and networkng, ncudng space-tme codng and moduaton for MIMO wreess communcatons, MIMO-OFDM systems, and cooperatve communcatons for wreess networks. Dr. Su s an Assstant ofessor at the Department of Eectrca Engneerng, the State Unversty of New York SUNY at Buffao. From June to March 5, he was a Postdoctora Research Assocate wth the Department of Eectrca and Computer Engneerng and the Insttute for Systems Research ISR, Unversty of Maryand, Coege Park. Dr. Su receved a Feowshp Award from U.S. Natona Research Counc NRC n. He was the recpent of IEEE Internatona Conference on Communcatons ICC Best Paper Award. Dr. Su receved the Inventon of the Year Award from the Unversty of Maryand n 5. He receved the Sgna ocessng and Communcatons Facuty Award from the Unversty of Deaware n. Dr. Su has served as Assocate Edtor for IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY and IEEE SIGNAL PROCESSING LETTERS. He has co-organzed two speca ssues for IEEE journas n the fed of cooperatve communcatons and networkng.