Channel Controlled Payload Length &Rate in MIMO Networks Using Cross-Layer

Transcription

1 Chael Cotrolled Payload Legth &Rate i MIMO Networs Usig Cross-Layer Mrs.Srimathi Mathialaga ad Dr. S. Shamugavel Abstract Today multimedia traffic is cogested due to heavy data traffic. If traffic becomes flooded, cogestio is more ad Data delivery is less. The source ca be optimized based o the chael coditio. Hece, Shao s theorems ca be ehaced to iclude chael state iformatio (CSI). We propose cross-layer desig which optimizes data rate ad payload legth based o CSI i MIMO etwors. Video summary frames are trasmitted via MIMO trasmit/receive diversity with AMC at the physical layer ad optimal payload legth algorithm ad ARQ at the data li layer ad source codig at the applicatio layer. We, assume first model the fiite-state Marov chai (FSMC) for the physical layer service ad based o that, we the characterize the Lagragia relaxatio ad Dyamic programmig to fid the optimal shortest path. The source codig as well as retrasmissio requests are based o the chael-state iformatio (CSI) ad the system cotroller is implemeted i all the four physical, MAC layer ad applicatio layers. The umerical results revealed that our proposed cross-layer desig ca efficietly achieve the distortio gai ad payload adaptatio at worst chael coditio. Idex Terms Cross-layer desig, quality-of-service (QoS), adaptive modulatio ad codig (AMC), multiple iput multiple output (MIMO), automatic repeat request (ARQ), mobile wireless etwors. About four ey words or phrases i alphabetical order, separated by commas. I. INTRODUCTION The 4G cellular stadards wor towards multi-user eviromet. The users are provided with differet data rates ad at differet terrai profiles. So, the iformatio theory bouded by Claude Shao i his wor A Mathematical Theory of Commuicatio may o loger be optimal. This divisio of codig theory ito compressio ad trasmissio is justified by the iformatio trasmissio theorems, source-chael separatio theorems that justify the use of bits as the uiversal currecy for iformatio i may cotexts. However, these theorems oly hold i the situatio where, oe trasmittig user wishes to commuicate oe receivig user.i scearios with more tha oe trasmitter(multiple-access chael), more tha oe receiver(broadcast chael) or itermediary helpers (relay chael),or more geeral etwors, compressio followed by trasmissio may o loger be optimal. Networ iformatio theory refers to these mmulti-aget commuicatio models. I multi-user sceario, user ca access the etwor uaware of chael coditios. This will mae the etwor more cogested. If cogestio becomes predomiat, QoS will be degraded. This will results i loss of data, delay ad distortio full receptio at the receiver. If the chael s badwidth is fixed, the user may suffer. So this problem ca be overcome by optimizig the physical layer with higher layers. The effective throughput is affected by a umber of parameters, icludig trasmissio rate, payload ad header size, costellatio size, trasmitted power ad received oise characteristics. Previous results revealed that careful payload legth adaptatio sigificatly improves the throughput performace at low sigal to oise ratios (SNRs), while at higher SNRs, rate adaptatio with higher payload legths provides better throughput performace. These payload sizes cover a wide rage of applicatios from various voice codecs to H.264 video coferecig applicatios alog with various data applicatios lie web browsig, FTP, etc. I [2], payload legth was cosidered as a optimizatio parameter ad tight couplig betwee payload legth ad data rate to maximize the sigle-user throughput o the AWGN ad differet fadig chaels was required. I geeral, the video summarizatio algorithm will geerate a still-image storyboard, which is composed of a collectio of saliet images extracted from the uderlyig video sequece. Some earlier wor cosidered the pacet loss factor due to usatisfactory wireless chael, where some does ot cosider. The source codig has ot bee optimized i the framewor, which might directly impact the perceptual quality of the results. I additio, the algorithm does ot guaratee a good cotet coverage aspect of the selected frames because potetial pacet loss pealty heavily biases the selectio process. Here, we propose Cross-layer desig that joitly optimizes the physical layer, MAC layer ad applicatio layer. Trasmissio of video summary at applicatio layer is based o MIMO trasmit/receive diversity with AMC at the physical layer ad payload legth ad rate adaptatio at the MAC layer ad the source codig at the applicatio layer. The rest of the paper is orgaized as follows. I sectio II, we derive the Problem Formulatio. System Overview is show i Sectio III. We have aalyzed the Related Wor i Sectio IV. Experimetal Results are show i Sectio V. Fially we cocluded i Sectio VI

2 II. PROBLEM FORMULATION I this sectio, we provide a brief explaatio about the various techiques adopted i this paper. The summary video frames are fragmeted ito pacets of smaller payload. The li adaptatio tables are used to idetify the SNR operatig regios ad correspodig data rates to optimize the video summary pacets. The optimizatio parameters are based o chael state iformatio. The physical layer is desiged adaptive modulatio codig with MIMO i order to achieve better throughput. The summary is fragmeted ito NP pacets. If the size of summary frame is less tha the pacet legth, there is o fragmetatio to be doe. The the trasmitted Pacet Error Rate () is PTras PL/L ad La LS. If the frame size is larger, the the pacet legth Np as Ls L f Np Where, La actual pacet legth. LS summary frame size. Lf fragmetatio pacet size. If the actual legth of the first pacet size is equal to the fragmetatio size, the the last pacet should be less tha the legth of the previous pacets. So, the PTras [3] is / L / N P L PTras P Where, PL is the probability that L cosecutive summary frames are lost simultaeously. P P + Μ. ( ml) β j ml m + ( a exp( )) M i g i Thus the average ca be expressed as K M ( ) (8) La P d i( ml i ) ξ ji j [ ( ) ] 2 K 2 + R Where R deotes the spectral-efficiecy of the th mode which is listed i TABLE II P deotes the probability that the SNR falls ito mode which is determied by R P P ( ) + Ρ P (4) (5) d (6) The pacet success rate is defied as the probability of receivig a pacet correctly correspodig to AMC mode, if the SNR falls i to the rage < +. The bits iside the pacets have the same bit error rate related with pacet error rate. So we ca write pacet error rate P as for pacet cotaiig La bits. L P ( BER) a () I order to guaratee the upper boud P AMC, the required BER to achieve for ay AMC mode is / La BER AMC ( ) () P AMC The pacet-error rate () for the th AMC mode for 2, 3..K ca be approximated [8] by ( ), if < < a exp(-g ), if The AMC is i mode if the SNR falls i to the rage < +. ca be expressed as below The average pacet error rate combied with MIMO[] is g P l a (2) (3) for the AMC + ( ) Ρ ( ) d PSR - (7) The effective throughout ca be defied as the umber of payload bits per secod received correctly. The payload overhead should also be tae ito accout ad for the iitial aalysis; we cosider the acowledgemets are error free. The payload overhead taes ito accout the CSMA/CA chael access time ad the header overheads as specified by the IEEE 82. protocol. Based o the data rate, the overhead is varyig.so the trasmit time of pacet icludig MAC, PHY headers ad DCF protocol overheads ca be defied as [2] P D * P to (8) Here, P to Total protocol overhead D Data rate correspodig to AMC mode. The total protocol overhead is calculated as specified by the IEEE 82. protocol. Each MAC frame cosists of MAC header or MAC protocol data uit (MPDU), variable legth frame body ad frame chec sequeces (FCS) as show i Fig.. The MAC header ad FCS cosist of 28 bytes ad the ACK is 4 bytes log. MAC Header 24byte s IP Header 2byte s RTP/UDP Header 2bytes Payload Data FCS 4bytes - 6 -

3 Fig.. Frame format of a data frame MPDU The effective throughput correspodig to the AMC mode is T ( L a / L a + P )* D * PSR (9) III. SYSTEM OVERVIEW The Cross-layer Desig is show i Fig.2.This cosists of a structure of physical layer, MAC layer ad applicatio layer. The MIMO diversity schemes cosists of Nt trasmit ad Nr receive ateas. As show i Fig.2 at the trasmitter side, we store the video summary frames at the buffer. Source codig is doe at applicatio layer. The video summary frames are fragmeted ito multiple pacets for trasmissio at lower layers. The fragmeted pacet legth which is othig but the payload legth at the MAC layer is adapted as per the chael state iformatio received. At the physical layer accordig to the chael coditio, MIMO combiig with AMC method is adopted. If the fragmeted pacets are received correctly, the the frames are stored at the buffer ad the video clip frames are recostructed. sigal-to-oise ratio (SNR) deoted by Ρ derived as a uified expressio [8] as follows: ( ) Ρ Μ ( ml) M i exp i i( ml βm i+ M j ) i ξ ji β j ml m + j+ml () ca be Where, (.) represets the Gamma fuctio, deotes the average SNR of the combied sigal, m deotes the fadig parameter, ξji,the multiomial expasio coefficiets b determied by ξ ji ξ with a max {, p ( i) /[( j p)! ] p a j - (M-)}, b mi{ j, (i-) (M-) }, ξj ξi, ξj / (j!), ad ξi i!. The parameters M, L ad β depeds o MIMO diversity schemes as specified i the TABLE I. Here M deotes the selectio diversity order ad L deotes the combiig diversity order β is varied oly whe STBC scheme is used. The total diversity order is determied by M x L i.e., Nt x Nr. A. MIMO Diversity Scheme The selected modulatio is trasmitted ad received via Nt trasmit ad Nr receive ateas. The differet diversity techiques [], are adopted as per the chael coditio as show i TABLE I. If perfect CSI is available at both sides of the wireless li, maximal-ratio trasmissio (MRT) also ow as beamformig ad maximal-ratio combiig (MRC) are ow as the optimal trasmit- ad receive-diversity schemes respectively. Whe the CSI is ot available at the trasmitter side, space-time bloc codig (STBC) is good choice to achieve trasmit diversity. The selectio combiig (SC) at either the trasmitter or receiver side is a good tradeoff both performace ad complexity level. TABLE I PARAMETERS FOR MIMO DIVERSITY MIMO Diversity Schemes M L β Tx-MRT/Rx- N t Tx-STBC/Rx-MRC N t N r N t Tx-SC/Rx-MRC N t N r Tx-MRT/Rx-SC N r N t Tx-STBC/Rx-SC N r N t N t Tx-SC/Rx-SC N t N r Performace Upper-boud N t N r B. Combied AMC with MIMO AMC is the powerful techique to icreases the spectral-efficiecy. The specific AMC modes are selected from the TABLE II based o the cross-layer optimizatio parameters a ad g which are the modulatio order ad the correspodig codig rate of a particular mode. Each mode cosists of a specific modulatio ad FEC code pair as i 3GPP, HILAN/2, IEEE 82.a ad IEEE 82.6 stadards. SNR rage is divided ito 7 o-overlappig itervals as < 2 < + with correspods to outage mode ad 2 7 correspod to BPSK, QPSK, QPSK, 6-QAM, 6-QAM ad 64-QAM respectively. If the SNR icreases, it will choose higher modes ad if the chael is worst, lower modes are selected. The pacet-error rate () for the th AMC mode for 2, 3...K ca be approximated by ( ), if < < a exp(-g ), if (2) The AMC is i mode, if the SNR falls i to the rage < +. ca be expressed as below: The probability desity fuctio (pdf) of the combied - 6 -

4 C. Source Codig g l a The video summary frames, which are goig to be trasmitted, are fragmeted ito multiple pacets for trasmissio at lower layers. If umber of frames of a video clip{ f,f,..f-} { of m umber of frames of its video g } summary, g,..., g m are to be trasmitted the, the lossy source codig produce the resultat cosumed bits of ith summary frame as, i,,,m-. Let Si ad Bi be the codig parameters of the lossy codig, which are used to optimize the desig. Let, Qi deote the umber of fragmeted pacets of ith summary frame ad Ni,q, deote the umber of trasmissios ad Fi,q, deote the pacet size for the qth pacet of the ith summary frame. The video summary frames, which are goig to be trasmitted, are fragmeted ito multiple pacets for trasmissio at lower layers. If umber of frames of a video clip{ f,f,..f-} { of m umber of frames of its video g } summary, g,..., g m are to be trasmitted the, the lossy source codig produce the resultat cosumed bits of ith summary frame as, i,,,m-. Let Si ad Bi be the codig parameters of the lossy codig, which are used to optimize the desig. Let, Qi deote the umber of fragmeted pacets of ith summary frame ad Ni,q, deote the umber of trasmissios ad Fi,q, deote the pacet size for the qth pacet of the ith summary frame. (3) TABLE III OFDM PHY CHARACTERISTICS Parameter Value Notes tslot 9µs Slot Time tsifs 6µs SIFS Time tdifs 34µs DIFSSIFS+2*Slot CW mi 5 mi.cotetio widow size CW max 23 max.cotetio widow size tplcp_preamble 6µs PLCP preamble duratio tplcp_sig 4µs PLCP SIGNAL Field duratio tsymbol 4µs OFDM symbol iterval (iv).the OFDM PHY characteristics of IEEE 82.a are summarized i TABLE III.The Payload overhead is tae as 4 bytes as the MAC header ad FCS cosists of 28 bytes ad the ACK is of 4 bytes log. (v). We have tae the SNR of 2dB. We cosider the differet modes of AMC as the data rate of 6Mbps, 2Mbps, 8Mbps, 24Mbps, 36Mbps ad 54Mbps. The payload legth is system varied from 25 bytes to 5 bytes. (v). The MIMO parameters are M 2, L2 ad Β which i othig but Tx-SC/Rx-MRC Next we listed the operatig assumptios adopted i this paper. (i).we cosider 2 such that, PL is-2 so, o eighbourig summary frames ca be lost together durig trasmissio. (ii).we assume fadig parameter m 2 idicatig the Rayleigh fadig chael ad the average SNR db. The chael remais time Ivariat durig trasmissio of a pacet, but varies from pacet to pacet. Thus the chael ca be estimated from the received SNR, per pacet which is a radom variable with a probability desity fuctio (pdf): Ρ exp (iii).we assume fixed chael trasmissio rate, r 6 6 sym/sec ad Fixed roud trip time TRTT ms ad maximum retrasmissio umber Nmax IV. Related Wor High bit rate requiremets of a video may cogest the etwor sigificatly. It is imperative to accout for the potetial impact of each video user o the etwor statistics ad guaratee that the etwor is ot operatig beyod its capacity. Ufortuately, most etwor desigs do ot provide mechaisms for protocol layers to optimally adapt to uderlyig chael coditios ad specific applicatio requiremets. While protocol layerig is a importat abstractio that reduces etwor desig completely, it is ot well suited to wireless etwors sice the ature of the wireless medium maes it difficult to decouple the layers. Moreover, meetig the ed-to-ed performace requiremets of demadig applicatios is extremely challeges without iteractio betwee protocol layers. Video streamig over wireless etwors ca beefit substatially from a cross-layer desig. I this desig, iterdepedecies betwee layers are characterized ad exploited by adaptig to iformatio exchaged betwee layers ad buildig the appropriate amout of robustess ito each layer. For example, routig protocols ca avoid lis experiecig deep fades, or the applicatio layer ca adapt its trasmissio rate based o the uderlyig etwor throughput ad latecy. The followig subdivisios explore cross-layer framewor that icorporates adaptatio across applicatio layer with other layers to meet

5 the requiremets of QoS capabilities for video trasmissio over wireless etwors Hierarchical video codig is a smart solutio to hadle the heterogeeity of receivers i multimedia multicast trasmissio over the wired iteret such as i RLM-based schemes ad SARC. Basically i a hierarchical ecodig scheme which is aalyzed i [2], [3], [4], [5],[6], [7], [], the most relevat elemets of the video sequece are icluded i a base layer, while less relevat pieces of iformatio are put ito a secod level also deomiated as ehacemet layer. Usually base layer receives a high priority treatmet, while the other layer is delegated to a secod plae. While desigig hierarchical video codecs, some amout of overhead is itroduced due to breapoit used whe splittig the ecoded video bit stream ito the base ad ehacemet layers ad the ability of assigig differet priorities. Oe of the mai advatages of hierarchical codig is that, this techique ca be applied to all ecodig schemes, such as H.26, H.263, MPEG-,MPEG-2,MPEG-4,H.264 amog others. The authors cosidered H.264 video ecodig stadard, also ow as MPEG-4 AVC which is highly efficiet by offerig perceptually equivalet video quality at about /3 to ½ of the bit rates offered by the MPEG-2 format [3], [6], [7].I a ut shell,h.264 cosists of differet layers. First, the Video Cotet layer (VCL) cotais the specificatio of the core video compressio egies that achieve basic fuctios such as motio compesatio, trasform codig ad etropy codig. This layer is trasport-uaware ad its highest data structure is the video slice, a collectio of the coded Macroblocs(MBs) i sca order. Secod, the Networ Abstractio layer(nal) is resposible for the ecapsulatio of the coded slices ito trasport etities of the uderlyig protocols. Each slice header acts as a resychroizatio marer, which allows the slices to be idepedetly decodable, ad to be trasported out of order ad still be decoded correctly at the decoder. A set of error resiliece techiques such as data-partitioig, which is a effective applicatio-level framig techique, which divides the compressed data ito separate uits of differet importace had bee proposed. Data partitioig creates more tha oe bit strig (partitio) per slice ad allocates all symbols of a slice ito a idividual partitio with a close relatioship [6], [7].The applicatio layer passes its traffic iformatio (the priority of the stream) with their QoS requiremets to the MAC layer, which maps these partitios to differet traffic categories to improve the perceived video quality. I [6], they have preseted a prelimiary evaluatio of ARSM, which proved effective i adaptig the chael rate taig ito accout the varyig chael coditios. ARSM (Auto Rate Selectio for Multicast) is a adaptive mechaism i which the AP selects the PHY data rate to be used for the multicast service. The PHY data rate to be used is determied by taig ito accout the chael coditios perceived by each ad every MT (Mobile Termial) belogig to a give multicast group. IV.EXIMENTAL RESULTS I this sectio, we preset experimetal results for cross layer desig with MIMO Chaels. Experimets are desiged usig H.264/AVCJM.2 for the video clip called Glassgow which is a typical test clip. For ease of compariso we summarize the first 3 frames ito 3 frames. For QP adaptatio we use differet values of QP ca be chose from, 2, 3, 4, 5 whe delay budget is set equal to the delay time accordig to differet etwor coditios. For without QP adaptatio we use the fixed values lie or 2 or 3 or 4 etc. Dis tortio QP adapt wo mimo o Qp adapt wo mimo QP adapt w mimo o Qp adapt w mimo Delay(Secod) Fig.3. Distortio vs Delay compariso for system model with ad without MIMO Fig.3 plots the system based o MIMO trasmit/receive diversity. The system almost wors similar for QP adaptatio i both with ad without MIMO. The performace which is show i rouded odes wor better compared with all other systems as the distortio gai upto 6% is achieved. I Fig.4, we have show the effective throughput for data rate of 6 Mbps. Iitially icrease i payload legth icreases the effective throughput but it reaches a optimum value from 4 bytes payload as further icrease would icrease the pacet error rate. Fig.4.Payload Legth Adaptatio for date rate of 6Mbps usig MIMO Chaels effective throughput (Mbps) data rate - 54 Mbps payload legth (byts) Fig.5.Payload Legth Adaptatio for date rate of 6Mbps usig MIMO Chaels

6 I Fig.5, the effective throughput is 34.5Mbps for 54Mbps date. I Fig.6 we compare the data rate for of 6Mbps, 9Mbps ad 2Mbps.Selectig the lowest data rate is too coservative while higher data rate ca sigificatly reduces the effective throughput. effective throughput (Mbps) data rate-6 Mbps data rate-9 Mbps data rate-2 Mbps payload legth (byts) Fig.6. Payload Legth Adaptatio for differet date rates usig MIMO Chaels The expected distortio E[D]of the video clip which is show i Fig.3 ca be calculated []by ~ E[ D] E[ D( f, f )] ml i+ i [ j, ib ib ] ρia } i b j li b {( ρ ) d f g~ ( S ) () ib a The expected distortio of the video clip should be with i tolerable time delay. So time delay should be bouded with i some delay budget Tmax as Mi E[D], s.t T T max () So, eq () ca be writte as Mi E[D], s.t T T max ad Max (G i, G i- G i+-l ), i [,m-] (2) Where, T max is the give delay budget for deliverig the whole video clip. T is the delay i trasmittig the whole summary frame which ca be expressed as m Q N i i, q Fi, q( Si, Bi) T [ + R A, C i q i, q, r i, q, i, q, T RTT ] (3) Where, TRTT is the maximum allowed RTT to get the acowledgemet pacet via the feedbac chael before a retrasmissio trial. V CONCLUSIONS I this paper, we have cosidered wireless video streamig applicatio. The goal is to deliver a video sequece with miimum delay ad distortio costraits usig miimum required trasmissio eergy to the receiver through multipath fadig chael. Our formulatio cosiders the tradeoffs i the selectio of source codig parameters lie source codig, payload adaptatio, ad physical layer adaptatio (AMC with MIMO). Oe of the mai lessos i the paper is that payload legth adaptatio at MAC yields higher throughput based o the CSI iformatio obtaied. The MIMO employed at the physical layer wors well compared with other systems. So, the etwors employed with MIMO perform better i order to yield higher throughput. Our system ca be readily exteded to multi-user sceario lie CDMA system. I additio to optimizatio parameters from the three layers, other layers iformatio ca also be icluded i order to give better results. REFERENCES [] A. L. Tole do, X. Wag ad Be Lu, A Cross- Layer TCP Modellig Framewor For MIMO Wireless Systems, IEEE Tras. Wireless Commu.,vol.5,o.4, pp , Apr. 26. [2] M. V. D.Schaar, S. Krishamachari, S. Choi ad X. Xu, Adaptive Cross-Layer Protectio Strategies forrobust Scalable Video Trasmissio Over 82.WLANs, IEEE J.Select.Areas Commu., vol.2, o., pp , Dec.23. [3] [3]D.Wu,S. Ci ad H. Wag, Cross-Layer Optimizatio for Video Summary Trasmissio over Wireless Networs, IEEE J.Select.Areas Commu., vol.25, o.4, pp.84-85, May.27. [4] W. Kumwilaisa, Y. T. Hou, Q.Zhag, W. Zhu, C. C. Jay Kuo ad Ya-Qi Zhag, A Cross-Layer Quality-of-Service Mappig for Video Delivery i Wireless Networs, IEEE J.Select.Areas Commu., vol.2,o., pp , Dec.23. [5] H. P. Shiag ad M. V. D. Schaar, Multi-User Video Streamig Over Multi-Hop Wireless Networs:A Distributed, Cross-Layer Approach Based o Priority Queuig, IEEE J. Select.Areas Commu.,vol.25, o.4, pp , May.27. [6] J.Villalo,P.Cueca,L.O.Barbosa,Y.Seo adt. Turletti, Cross-Layer Architecture For Adaptive Video Multicast Streamig Over Multirate Wireless LANs, IEEE J.Select.Areas Commu., vol.25,o.4, pp.699-7, May.27. [7] P.Pahalawatta,R.Berry, T. Pappas ada.katsaggelos, Cotet-Aware Resource Allocatio ad Pacet Schedulig for Video Trasmissio over Wireless Networs, IEEE J. Select.Areas Commu., vol.25, o.4, pp , May.27. [8] J.Tag ad X.Zhag, Cross-Layer Modelig for quality of Service Guaratees over Wireless Lis, IEEE Tras. Wireless Commu., vol.6, o.2, pp , Dec. 27. [9] Q. Liu, S. Zhou, ad G. B. Giaais, Cross-layer combiig of adaptive modulatio ad codig with trucated ARQ over wireless lis, IEEE Tras. Wireless Commu.,vol. 3, o. 5, pp , Sept. 24. [] Srimathimathialaga ad Dr.S.Shamugavel, A Cross-Layer Desig Approach usig MIMO System for Wireless Networs, i Proc. IEEE It. Cof. Idustrial Iformatio, ICIIS28. [] D.Jurca ad P.Frossard, Video Pacet Selectio ad Schedulig for MultipatStreamig, IEEETras.Multimedia.vol.9,o.3,pp ,Ap r. 27. [2] S.Choudhury ad J.D.Gibso, Payload Legth ad Rate adaptatio for MultimediaCommuicatios i Wireless LANs, IEEE

7 Iput Output Applicatio Layer (Source codig) Applicatio Layer (Source codig) S i N max MAC Layer (Payload legth ad data rate adaptatio) ARQ MAC Layer (Payload legth ad data rate adaptatio) A i, C i, System Cotroller AMC Trasmit diversity MIMO chael Receive diversity AMC CSI feedbac Fig.2. Cross-layer Desig Mod TABLE II AMC MODES Mode Mode Mode2 Mode3 Mode4 Mode5 Mode6 Modulatio BPSK QPSK QPSK 6-QAM 6-QAM 64-QAM Codig Rate /2 /2 3/4 /2 3/4 3/4 (C ) Data Rate D (Mbps) R (bytes/symbol) a g