PAPER Adaptive Modulation in Coded Cooperation under Rayleigh Fading Channels

Transcription

1 82 IEICE TRANS. COMMUN., VOL.E93 B, NO.1 JANUARY 2010 PAPER Adaptive Modulatio i Coded Cooperatio uder Rayleigh Fadig Chaels Ka ZHENG a), Member, Lijie HU, Lig WANG, Webo WANG, ad Li HUANG, Nomembers SUMMARY Cooperative commuicatio provides a ew way of itroducig spatial diversity to wireless systems. I order to icrease the spectral efficiecy of coded cooperative relayig system, the adaptive modulatio techique is preseted uder Rayleigh fadig chael i this paper. The source ad relay adapt their modulatio schemes based o the chael coditio of all three liks, i.e. source to relay, source to destiatio ad relay to destiatio. Furthermore, sice the available chael kowledge of the source to relay lik is usually o-ideal at the destiatio i practice, a simplified estimatio of this lik quality is also give. Simulatio demostrates the effectiveess of the proposed techique i improvig the data throughput. key words: adaptive modulatio, coded cooperatio 1. Itroductio Cooperative commuicatio provides a ew way of itroducig spatial diversity i wireless systems where the mobile statios may ot be able to support multiple ateas due to size or other costraits [1]. By creatig a virtual array through cooperatio, cooperative diversity emerges as a special form of spatial diversity. Recetly some repetitiobased cooperative diversity algorithms such as amplify-adforward (AF), decode-ad-forward (DF) are developed to fully exploit the spectral diversity for reducig the outage probability [2]. However, cooperative diversity schemes such as coded cooperatio usually decrease the spectral efficiecy of the system because of their repetitio based structure [3], [4]. O the other had, by adjustig the trasmissio parameters to the istataeous lik quality, lik adaptatio mechaisms aim at improvig both spectral efficiecy ad lik reliability. So far, the ivestigatio o adaptive techology i cooperative etwork have t bee fully carried o [5] [7]. Most of them focus o the adaptive modulatio cocept applied i amplify-ad-forward scearios i [6], [7]. Also, the authors i [5] study the performace of adaptive modulatio with retrasmissio to maximize the data throughput i coded cooperatio systems. However, they did t cosider much about the practical issues. For example, sice um- Mauscript received February 10, Mauscript revised August 11, The authors are with Key Lab of Uiversal Wireless Commuicatios, Miistry of Educatio, Beijig Uiversity of Posts ad Telecommuicatios, Chia. The author is with Orage R&D, Beijig, Chia. This work was supported i part by Chia NSFC uder Grat ad Research Fud for the Doctoral Program of Higher Educatio uder Grat a) zka@bupt.edu.c DOI: /trascom.E93.B.82 bers of modulated symbols trasmitted i two hops are differet i their scheme, it is hard to directly apply it i practical systems with the pre-defied frame structure. Moreover, the chael iformatio of all the liks has to be kow whe choosig the modulatio scheme by the scheme proposed i [5], which is quite difficult to meet i practical systems. This paper presets adaptive modulatio schemes for the coded cooperatio system uder Rayleigh fadig chaels, which belog to throughput-orieted adaptive modulatio mechaisms. Such mechaisms select the modulatio scheme which maximizes the istataeous system throughput. I coded cooperatio systems, cooperatig source ad relay adapt their modulatio schemes based o the chael coditio of all three liks, i.e. source to relay, source to destiatio ad relay to destiatio. Also, cosiderig the istataeous chael iformatio betwee the source ad the relay is usually ot kow perfectly available at the destiatio, the we proposed a adaptive modulatio scheme which does t require the accuracy chael state iformatio istataeously. It ca estimate the chael quality through collectig the previous trasmissio status averagely. This scheme ca provide better performace tha direct trasmissio ad demostrate oly a slight drop agaist idealistic performace. This paper is orgaized as follows. Sectio 2 gives the brief descriptio of a system model ad procedure of coded cooperatio. The, the throughput of the differet systems are aalyzed i Sect. 3. Next, the adaptive modulatio schemes for the idealistic case ad i absece of the istataeous chael iformatio betwee the source ad relay i coded cooperatio systems are described i Sect. 4. I Sect. 5, the simulatio results are preseted ad discussed. Fially, Sect. 6 gives the coclusio. Notatios: A circularly symmetric complex Gaussia variace x with mea m ad covariace R is deoted x CN(m, R). 2. System Overview 2.1 System Model As show i Fig. 1(a), the source (S) trasmits the sigals directly to the destiatio (D) i the systemswith directtrasmissio. I cooperative etworks, besides the direct lik betwee the S ad the D, iformatio also flows through relay (R) odes, which help S to commuicate with D as illus- Copyright c 2010 The Istitute of Electroics, Iformatio ad Commuicatio Egieers

2 ZHENG et al.: ADAPTIVE MODULATION IN CODED COOPERATION UNDER RAYLEIGH FADING CHANNELS 83 Fig. 2 Procedure of coded cooperatio trasmissio. Fig. 1 Illustratio of wireless commuicatio systems. trated i Fig. 1(b). All the chaels icludig source to the destiatio (S-to-D), relay to the destiatio (R-to-D) ad source to relay (S-to-R) are idepedet of each other ad are assumed to be flat fadig for the sake of simplificatio. However, the proposed techique i this paper could be easily exteded to widebad systems that experiece frequecy selective fadig chael. The trasmitted data is segmeted i packets ad protected agaist trasmissio errors with a chael ecoder before beig mapped to a particular sigal costellatio. At the ith frame with the duratio of T s,theith output of the modulator for S is s (i) adŝ (i) forr. We assume that E{ s (i) 2 } = E{ ŝ (i) 2 } = P s,wherep s is the average trasmitted power. The correspodig sigal received by D ad R is y SR (i) = α SR y RD y SD (i) = α RD (i) = α SD s (i) + z SR (i) ŝ (i) + z RD (i) s (i) + z SD (i) (1) where α SR, α RD ad α SD deote the idepedet complex fadig chael gai from S to R, from R to D ad from S to D, modeled as α SR CN(0,σ 2 SR ), αrd CN(0,σ 2 RD )ad α SD CN(0,σ 2 SD ) with σ2 SR = E{ αsr 2 }, σ 2 RD = E{ αrd 2 } ad σ 2 SD = E{ αsd 2 } respectively. For slow fadig (quasistatic) cases, the fadig coefficiets are costat withi the trasmissio of the etire frame. The, the idex i of chael fadig coefficiets ca be omitted. Without loss of geerality, we assume that the oise terms z SR (i), z RD (i) ad z SD (i) have equal variaces σ 2 N ad are modeled as zsr (i) CN(0,σ 2 N ), zrd (i) CN(0,σ 2 N )adzsd N (i) CN(0,σ2 N ) respectively. Defie the istataeous sigal-to-oise ratio (SNR) at lik S-R, R-D ad S-D as γ SR = α SR 2 γ, γ RD = α RD 2 γ,adγ SD = α SD 2 γ, respectively, with γ = P s /σ 2 N. 2.2 Procedure of Coded Cooperatio Trasmissio The framework of coded cooperatio i [2] is used i our paper. I a cooperative etwork, two or more odes share their iformatio ad trasmit joitly as a virtual atea array. This eables them to achieve higher data rates ad diversity tha they could have idividually. Coded cooperatio provides sigificat performace gais for a variety of chael coditio. I additio, by allowig differet code rates ad partitios, it also has a great degree of flexibility to adapt to chael coditio. Figure 2 shows the procedure of coded cooperatio trasmissio. The source bits of the source are segmeted ito blocks of K bits, icludig cyclic redudacy check (CRC) bits. Each iformatio source block is ecoded by a rate R c = K/N code geeratig N = K/R c bits. Let us assume that each codeword of legth N ca be divided ito two parts of legth N 1 ad N 2 with N 1 + N 2 = N. ThefirstN 1 bits are self-decodable while the secod N 2 bits are the parity bits that ca be efficietly puctured without failig to decode the origial message. I the first trasmissio phase, the source seds N 1 bits which are overhead by the relay ad destiatio. These N 1 bits are made up of K iformatio bits ad N 1 K parity bits. The, the codeword received at R is decoded ad checked by usig CRC. Whe the relay decodes the iformatio bit correctly or ot (which ca be checked by usig CRC), it will sed ACK or NACK to the source. If o error, the origial K bits ca be recovered ad the used to compute the remaiig N 2 bits of the codeword. Otherwise, the relay will keep silet i the secod trasmissio phase. The level of cooperatio is defied as ρ = N 2 /N. I the secod trasmissio phase, if cooperatio is happeed, the relay trasmits N 2 bits to the destiatio. Otherwise, the source will sed the remaiig N 2 bits by itself after beig iformed by NACK. After the destiatio decodes the K source bits from the received N 1 + N 2 bits, the ACK or NACK will be set to the source accordig to the result of CRC checkig. 3. Throughput Aalysis o Coded Cooperatio Systems 3.1 Uder Direct Trasmissio To better uderstad the data throughput for coded cooperatio, let us first take a close look at that for direct trasmissio, i.e. oly o-cooperative trasmissio. Figure 3(a) shows frame structure of the direct trasmissio, where the source trasmits each packet directly to the destiatio. The source bits are first segmeted ito blocks with size of K bits icludig cyclic redudacy check (CRC) bits. Each source block is ecoded ad puctured ito a Ñ-bit codeword with the code rate R c = K/Ñ. The,

3 84 IEICE TRANS. COMMUN., VOL.E93 B, NO.1 JANUARY 2010 Fig. 3 Frame structures for direct ad cooperative trasmissio schemes. these coded bits are modulated ito the M symbols with Q bits per symbol. So the data throughput from S to D ca be writte as ξ = M Q R c [1 P c (γ SD )]/T s (2) where P c (γ SD ) is the average packet-error rate (PER) for the quasi-chaels-to-d i o-cooperative trasmissio. Note that P c (γ SD ) depeds o the size Ñ, the chael quality γ SD of S-to-D ad code rate R c. For fair compariso, the same system parameters for chael codig for each modulatio ad codig scheme (MCS) are used for both direct trasmissio ad coded cooperative trasmissio, i.e. K = K, Ñ = N ad R c = R c. 3.2 Uder Coded Cooperatio Trasmissio Now let us cosider cooperative commuicatio i a dualhop wireless etwork where iformatio is trasmitted from the source to the destiatio with the frame structure as show i Fig. 3(b). As described i the previous sectio, the source seds M 1 modulated symbols with Q 1 bits per symbol to D ad R i the first time slot, cotaiig the first N 1 = M 1 Q 1 bits with code rate R 1 = K/N 1. Depedig o whether the relay ca decode the origial K bits from these N 1 bits, the secod N 2 bits are modulated as M 2 symbols with Q 2 bits per symbol ad trasmitted by the source or relay i the secod time slot. Sice the frame structure is usually pre-defied i practical wireless commuicatio systems, it is assumed that the duratio of the direct trasmissio phase (i.e. 1st time slot) ad that of cooperatio trasmissio phase (i.e. 2d time slot) are kept uchaged i this paper. Also, o the assumptio of cetralized cotrol strategy, the modulatio schemes for direct ad cooperatio trasmissio are adjusted together with the same order, i.e. Q = Q 1 = Q 2. I order to aalyze the data throughput, we let P 1 (γ ) deote the average PER of the sigals with codeword size N 1 = M 1 Q ad rate R 1 o the assumptio that the SNR of this lik is γ. Similarly, P c (γ ) deotes the average PER with codeword size N = (M 1 + M 2 )Q ad rate R c if the SNR of this lik is γ. There are two possible operatio modes i coded cooperatio systems, i.e. o-cooperative mode ad cooperative mode. I case of o-cooperatio, which happes with ), the destiatio receives the whole packet with the average PER of P c (γ SD ) from the source itself. Whe the source ad relay trasmit the data cooperatively, which happes with the probability of 1 P 1 (γ SR ), the N bits received at the destiatio iclude N 1 bits of the packet from the source with γ SD ad the remaiig N 2 bits from the relay with γ RD. Therefore, i order to estimate the average PER of the whole packet with these two istataeous γ SD ad γ RD, the method of expoetial effective SIR mappig (EESM) has to be used [8]. The basic idea of EESM is to map the multiple istataeous chael state ito a sigle scalar value, termed the possibility of P 1 (γ SR as a effective SNR, i.e. γ eq. The, this effective SNR is used to fid a estimate of the packet-error probability from basic additive white Gaussia oise (AWGN) lik-levelperformace. By EESM, the effective SNR i the cooperative mode ca be expressed as γ eq M 1 exp = β l ( γsd β ) + M 2 exp ) ( γrd β M 1 + M 2 (3) where β is a parameter that must be optimized from liklevel simulatio results for each combiatio of the modulatio ad codig rate. The, the data throughput from S to D with coded cooperatio ca be writte as ξ = (M 1 + M 2 )Q R c /T s {[1 P 1 (γ SR )][1 P c (γ eq )] + P 1 (γ SR )[1 P c (γ SD )]} (4) where P 1 (γ SR ) is the average PER of the first N 1 bits over the S-to-R chael, P c (γ SD ) deotes the average PER for S-to-D chael whe S trasmits the whole packet N bits by itself ad P c (γ eq ) represets the average PER for the cooperative fadig chael whe the destiatio receives first N 1 bits from S ad the remaiig N 2 bits from R. 4. Adaptive Modulatio Scheme for Coded Cooperatio Cooperative diversity schemes such as coded cooperatio usually decrease the spectral efficiecy of the system because of their repetitio-based structures. I this paper, we aim to icrease the spectral efficiecy of the coded cooperatio system by adaptive modulatio scheme. 4.1 SNR Estimatio Adaptive modulatio is typically based o the received SNR. I practice, the SNR is usually ukow at the trasmitter ad must be estimated by the receiver. So a dedicated pilot sequece p = {p(1), p(2),, p(l)} of legth L, set by the source or the relay, is time-multiplexed with the modulated data i each time slot as show i Fig. 4.

4 ZHENG et al.: ADAPTIVE MODULATION IN CODED COOPERATION UNDER RAYLEIGH FADING CHANNELS 85 Fig. 4 Frame structures with pilot sequece for cooperative trasmissio scheme. Let us take the lik betwee the source ad the destiatio as a example for illustratio. Accordig to (1), the received pilot sigal at the destiatio i the th frame ca be writte as y SD (i) = α SD p(i) + z SD (i), i {1, 2,, L} (5) where the chael fadig coefficiet α SD is assumed to be costat over the etire frame. The, the chael with least square (LS) priciple ad SNR estimate with miimum mea square error (MMSE) priciple ca be formulated as [9] α SD = γ SD = y SD (i)p (i) (6) p(i) 2 p(i) 2 L y SD (i)p (i) (i) 2 y SD 2 y SD (i)p (i) This method also ca be used to estimate the chael ad its SNR of other liks like S-to-R ad R-to-D. These estimates will guide the trasceiver to decide the proper modulatio scheme that fits the curret trasmissio characteristics. Usually, the better SNR, the higher modulatio scheme will be applied to the trasmissio. 4.2 Adaptive Modulatio Scheme I coded cooperatio systems, the source ad the relay should choose their modulatio schemes based ot oly o their ow chael quality to the destiatio but also o the chael of their parter to the destiatio. Accordig to (4), the data throughput ξ depeds o the qualities of all the chaels icludig the liks S to D, R to D, evethe lik S-to-R. For the adaptive modulatio, the data throughput will be computed based the SNRs at the destiatio o the assumptio of each possible modulatio scheme. The, the modulatio optio with the maximum data throughput will be selected ad the decisio is made at the destiatio ad feedback to S ad R. At the destiatio, the SNRs of the liks icludig S to D ad R to D ca be estimated by usig the dedicated pilot sequece as itroduced before. However, it is quite difficult 2 (7) Fig. 5 Flowchart of adaptive modulatio scheme i coded cooperatio trasmissio. for D to kow the exact chael state iformatio of S-to-R lik, which is used to estimate P 1 (γ SR ). O the other had, oly whe the chael of S-to-R lik is reliable, i.e. the relay ca decode the first N 1 -bit codeword i the first phase, the cooperative trasmissio will be happeed i the secod phase. So the destiatio ca get the limited iformatio about S-to-R lik through the kowledge whether the cooperative trasmissio is happeed successfully or ot. Therefore, istead of P 1 (γ SR ), we propose to use the simplified factor P 1 () i (4) to estimate the throughout performace. This factor P 1 (), reflectig a quality estimate of S to D lik, is geerated through collectig the previous trasmissio status averagely as P 1 () = W 1 W P 1( 1) + 1 W f SR (), > 0 (8) P 1 (0) = 1 where W is the size of the average widow depedig o the coheret time of the fadig chael, ad f SR () istheflag of the trasmissio status of the th frame, i.e. f SR () = 0 i case of cooperative trasmissio, otherwise, f SR () = 1. Therefore, the proposed adaptive modulatio scheme i coded cooperatio trasmissio ca be summarized as show i Fig Simulatio Results The performaces of the proposed strategy for cooperatio commuicatio are evaluated by simulatios ad discussed i this sectio. Turbo code used for directio trasmissio ad coded cooperatio trasmissio is a Parallel Cocateated Covolutioal Code (PCCC) with two 8- state costituet ecoders, whose geerator polyomials is G(1, 13/11) ad its overall rate of 1/3, i.e. R c = R c = 1/3 [10]. The source block has M 1 + M 2 = 300 modulated symbols ad M 1, M 2, R 1 deped o the cooperatio level ρ (e.g. if ρ = 1/2, M 1 = 150, M 2 = 150, R 1 = 2/3; if

5 86 IEICE TRANS. COMMUN., VOL.E93 B, NO.1 JANUARY 2010 Table 1 MCS vs. β value. MCS Modulatio R c β 1 QPSK 1/ QAM 1/ QAM 1/ Fig. 7 Normalized throughput compariso of coded cooperatio with/without adaptive modulatio ( γ SR, γ RD, γ SD )=( γ + 10 db, γ, γ). Fig. 6 PER compariso of directio trasmissio ad coded cooperatio with ideal CSI/SNR. ρ = 33%, M 1 = 200, M 2 = 100, R 1 = 1/2). Three modulatio ad codig schemes (MCS) icludig QPSK, 16QAM ad 64QAM modulatio are applied i this sectio. For compariso, the same legth of codeword is used for directio trasmissio ad coded cooperatio trasmissio, i.e. Ñ = N = 600 with K = K = 200 for MCS1, Ñ = N = 1200 with K = K = 400 for MCS2 ad Ñ = N = 1800 with K = K = 600 for MCS3, respectively. The values of β for each MCS are geerated by our simulatios with the commo-used method i [11] ad show i Table 1. The idepedet fadig liks are geerated by usig Jakes Model havig a U-shape Doppler power spectrum with maximum doppler spreadig of 10Hz. For the chael ad SNR estimatio, the legth of pilot sequece is assumed to be 5, i.e. L = 5. With the referece to Fig. 1, we mostly cosider represetative scearios that correspodig to those i which R is located either close to S or i the middle of S ad D; the correspodig average output SNRs ( γ SR, γ RD, γ SD )i logarithmic-scale are ( γ+10db, γ, γ) ad( γ+5db, γ+5db, γ), respectively. 5.1 With Ideal Chael ad SNR Estimatio Here the istataeous chael state ad SNR are assumed to be kow ideally at the receiver. Let us first discuss the error rate advatages of coded cooperatio. O the assumptio of ideal chael quality of the lik from S to R, i.e.o error is happeed durig trasmissio, we compare the performaces betwee the direct trasmissio ad coded cooperatio i Fig. 6. I the direct trasmissio, all the N bits trasmitted by S are over the same chael fadig coefficiet (i.e. α SD ) of the slow fadig chael from S to D. O the other had, o the assumptio of successful coded coop- Fig. 8 Normalized throughput compariso of coded cooperatio with/without adaptive modulatio ( γ SR, γ RD, γ SD )=( γ + 5dB, γ + 5dB, γ). eratio, the first N 1 bits go through the chael fadig coefficiet of the lik from S to D (i.e. α SD ) while the secod N 2 bits are trasmitted by R over the lik from R to D with the chael fadig coefficiet α RD. The chael fadig coefficiets α SD ad α RD are idepedet idetical distributed Rayleigh variables as metioed before. So, the PER performaces of coded cooperatio are better tha direct trasmissio because the spatial diversity gai ca be obtaied by coded cooperatio. With the larger cooperatio level, more bits are trasmitted through the lik from R to D ad lower PER is achieved due to better exploit the spatial diversity gai. Moreover, whe the quality of the lik from R to D is icreased, i.e. ( γ RD, γ SD )=( γ+5db, γ), the performace of coded cooperatio is further improved. Figure 7 ad Fig. 8 show the data throughput ormalized by 1/T s of the coded cooperatio systems with/without the proposed adaptive modulatio scheme uder i differet scearios respectively. It is clear that coded cooperatio with adaptive modulatio leads to higher throughput tha

6 ZHENG et al.: ADAPTIVE MODULATION IN CODED COOPERATION UNDER RAYLEIGH FADING CHANNELS 87 Fig. 9 Normalized throughput compariso of directio trasmissio ad coded cooperatio with adaptive modulatio ( γ SR, γ RD, γ SD )=( γ + 10 db, γ, γ). Fig. 11 Probability of cooperative trasmissio with adaptive modulatio. Fig. 12 RMSE performaces with estimated CSI/SNR ( γ SR, γ RD, γ SD )=( γ, 15 db, 10 db). Fig. 10 Normalized throughput compariso of directio trasmissio ad coded cooperatio with adaptive modulatio ( γ SR, γ RD, γ SD )= γ+5db, γ + 5dB, γ). the fixed modulatio i both scearios. Next, let us compare the throughput performace of systems with adaptive modulatio i the direct trasmissio ad coded cooperatio uder two differet scearios i Fig. 9 ad Fig. 10 respectively. We ca fid that the coded cooperatio with adaptive modulatio outperforms direct trasmissio with adaptive modulatio. Compared with the sceario i which R is located close to S, the coded cooperatio with adaptive modulatio leads to more obvious gais i the low ad medium SNR regio uder the sceario where R is i the middle of S ad D. O the other had, more throughput gai is achieved for the scheme with ρ = 1/3 tha the scheme with ρ = 1/2. It is because more N 1 bits are trasmitted i the first phase with smaller cooperatio level ρ, which icreases the probability of successful cooperative trasmissios as show i Fig. 11. For example, the sceario i which R is located close to S, the throughput gai is 0.05 bits/symbol with ρ = 1/2 ad 0.13 bits/symbol with ρ = 1/3 at γ = 12 db. 5.2 With Chael ad SNR Estimatio O the assumptio of the chael ad SNR estimatio as explaied i Sect. 4, the performace of adaptive modulatio with the simplified estimatio of P 1 (γ SR ) will be studied i this part. Accordig to the coheret time of the fadig chael, here we set the widow size W as 64. I order to evaluate the accuracy of simplified estimatio, the root mea squared error (RMSE) of P 1 () isdefiedas RMS E = E[ P 1 (γ SR ) P 1 () 2 ] (9) where E[x] deotes the expectatio of the radom process x. With the estimated CSI ad SNR, the performaces is evaluated i case that oly the average SNR of the lik from S to R is varied ad the qualities of other liks are assumed to be fixed, i.e. γ RD = 15 db ad γ SR = 10 db. Firstly, Fig. 12 shows the RMSE performaces with differet cooperatio level. With the icrease of the SNR of the lik from S to R, the RMSE becomes less, which meas more reliability of

7 88 IEICE TRANS. COMMUN., VOL.E93 B, NO.1 JANUARY 2010 Fig. 13 Normalized throughput with estimated CSI/SNR ( γ SR, γ RD, γ SD )=( γ, 15 db, 10 db). Fig. 15 Normalized throughput compariso of directio trasmissio ad coded cooperatio with ideal or estimated CSI/SNR (ρ = 1/3). tio error is oly aroud 0.4 or 0.2 bits/symbol i the regio of low or medium SNR. Meawhile, the coded cooperatio systems with proposed adaptive modulatio scheme with/without simplified method still outperform the direct trasmissio whe estimated CSI ad SNR are used. 6. Coclusio Fig. 14 Normalized throughput compariso of directio trasmissio ad coded cooperatio with ideal or estimated CSI/SNR (ρ = 1/2). the proposed simplified estimatio. The, the correspodig ormalized throughout performaces are give i Fig. 13. It ca be see that the throughput differece betwee two )orp 1 (), are quite small, which demostrates the effectiveess of the simplified method. Figure 14 ad Fig. 15 compares the ormalized data throughput of three schemes with ideal or practical chael estimatio uder the sceario of ( γ SR, γ RD, γ SD )=( γ + 5dB, γ + 5dB, γ), where differet cooperatio level ρ is assumed respectively. I the case that the estimated SNR of S-to-R methods, i.e. usig P 1 (γ SR lik is kow at D, thep 1 (γ sr ) is used for throughput computatio i coded cooperatio systems. Otherwise, the simplified scheme usig P 1 () ca be applied. Firstly, i case of ideal CSI ad SNR estimatio, we ca fid that the performace degradatio due to the simplified method usig P 1 () ) is small, e.g. the throughput loss is oly 0.05 bits/symbol at γ = 12 db i case of ρ = 1/2asshowi Fig. 14 while 0.02 bits/symbol i case of ρ = 1/3 i Fig. 15. Secodly, with the estimated CSI ad SNR through pilot sequece of L = 5, the performace loss due to estima- istead of P 1 (γ SR I this paper, we have proposed the adaptive modulatio scheme that ca maximize the data throughput i systems employig coded cooperatio relayig. Also, cosiderig that the istataeous chael iformatio betwee the source ad the relay is usually ot available at the destiatio, the simplified estimatio of the chael quality betwee the source ad relay through collectig the previous trasmissio status statistically is give. Simulatio results have bee preseted to demostrate the ormalized throughput improvemet by adaptive modulatio schemes i coded cooperatio systems. Refereces [1] A. Sedoaris, E. Erkip, ad B. Aazhag, User cooperatio diversity! Parts I & II, IEEE Tras. Commu., vol.51, o.11, pp , Nov [2] J.N. Laema, D.N.C. Tse, ad G.W. Worell, Cooperative diversity i wireless etworks: Efficiet protocols ad outage behavior, IEEE Tras. If. Theory, vol.50, o.12, pp , Dec [3] T. Huter ad A. Nosratiia, Cooperatio diversity through codig, Proc. EEE It. Symp. If. Theory, pp , Lausae, Switzerlad, Jue [4] M. Jaai, A. Hedayat, T.E. Huter, ad A. Nosratiia, Coded cooperatio i wireless commuicatios: Space-time trasmissio ad iterative decodig, IEEE Tras. Sigal Process., vol.52, o.2, pp , Feb [5] Z. Li, E. Erkip, ad M. Ghosh, Adaptive modulatio for coded cooperative systems, IEEE 6th Workshop o Sigal Processig Advaces i Wireless Commuicatios, pp , Jue [6] E. Yazdia ad M.R. Pakrava, Adaptive modulatio techique for cooperative diversity i wireless fadig chaels, IEEE Persoal, Idoor ad Mobile Radio Commuicatios, pp.1 5, Sept [7] E.C. Striati, Sheg Yag, ad J.C. Belfiore, Adaptive modulatio

8 ZHENG et al.: ADAPTIVE MODULATION IN CODED COOPERATION UNDER RAYLEIGH FADING CHANNELS 89 ad codig for hybrid cooperative etworks, IEEE Iteratioal Coferece o Commuicatios, pp , Jue [8] 3GPP2-C , Ericsso, Effective SNR mappig for modellig frame error rates i multiple-state chaels, April [9] D.-J. Shi, W. Sug, ad I.-K. Kim, Simple SNR estimatio methods for QPSK modulated short bursts, IEEE GlobeCom, vol.6, pp , Nov [10] 3GPP TS , Multiplexig ad chael codig (FDD). [11] K. Brueighaus, D. Astely, T. Salzer, S. Visuri, A. Alexiou, S. Karger, ad G.-A. Seraji, Lik performace models for system level simulatios of broadbad radio access systems, IEEE Persoal, Idoor ad Mobile Radio Commuicatios, vol.4, pp , Sept Webo Wag received his B.S., M.S. ad Ph.D. degrees from Beijig Uiversity of Posts&Telecommuicatios (BUPT), Chia, i 1986, 1989 ad 1992 respectively. He is curretly a professor ad dea of school of Telecommuicatio Egieerig of BUPT. His research iterests iclude sigal processig, mobile commuicatios ad wireless etwork. Li Huag received the B.S. ad M.S. from Beijig Uiversity of Posts & Telecommuicatios, Chia, i 2002 ad 2005 respectively. Now she works o Orage Labs, Frace Telecom R&D, Beijig, Chia. Her curret research iterests lie i the field of cogitive/cooperative commuicatio schemes. Ka Zheg received the B.S., M.S. ad Ph.D. degrees from Beijig Uiversity of Posts&Telecommuicatios, Chia, i 1996, 2000 ad 2005 respectively, where he is curretly a associate professor. From April 2000 to October 2001, he was a system developmet egieer at TD-SCDMA R&D cetre of Siemes (Ltd) at Beijig, Chia. His curret research iterests lie i the field of sigal processig for digital commuicatios, with emphasis o PHY/MAC algorithms i cooperative wireless etworks. Lijie Hu received the B.S. degree from Beijig Uiversity of Posts&Telecommuicatios (BUPT), Chia, i 2007, where she is workig for M.S. degree. Her curret research iterests iclude lik adaptio algorithms i wireless cooperative systems. Lig Wag received the B.S. ad M.S. degrees from Beijig Uiversity of Posts&Telecommuicatios (BUPT), Chia, i 2005 ad 2008,respectively. Her curret research iterests iclude PHY/MAC algorithms i MIMO-OFDM systems.