Optimization of Hybrid Token-CDMA MAC System Using Cross-Layer Information

Transcription

1 1 Optmzaton of ybrd Token-CDMA MAC System Usng Cross-ayer Informaton I-Sheng u, Graduate Student Member, IEEE, Fambra Takawra, Member, IEEE, and ong Jun Xu, Member, IEEE Abstract Ths paper presents a ont medum access control and physcal (MAC-PY) layers soluton for optmzng the system performance n the hybrd Token-CDMA MAC system. The proposed scheme s desgned n order to provde contnuous montorng of the performance acheved by the users and adustng msson parameters usng dfferent spreadng factors. Varous performance metrcs are used to demonstrate effectveness of the cross-layer nteracton for the hybrd Token-CDMA MAC protocol. Index Terms CDMA, Cross-layer, Medum Access Control (MAC), ualty of Servce (os), token passng. T I. INTRODUCTION he past few years have seen tremendous nterest n cooperatve communcatons n the feld of wreless communcaton research: the so-called cross-layer optmzaton [1]. Varous possble cross-layer nteractons can be consdered when performng a cross-layer desgn wthn the OSI-layered model. To demonstrate, Fg. 1 dsplays the dfferent control flows needed to provde a cross layer nteracton between physcal and upper layers of two remote nodes. When two nodes communcate, the recevng one measures the physcal state, whch s generally a vector of real values. An entty named Agent Manager estmates, measures and selects the approprate values to be sent to the upper layers of the mttng node. These layers wll actuate accordngly to adapt to the actual channel condtons, performng the cross-layer nteracton. Prmtves Prmtves MAC PY Node as mtter Parameter Vectors MAC Agent Manager PY Node as recever Fg. 1. Example of cross-layer nteracton through an agent manager The amount of lterature on ths ssue s stll relatvely scarce and mostly at the physcal layers [], [3], [4], [5], [6], [7] and [8]. It has been shown n those lteratures and references theren that from the perspectve of the network, cross-layer desgn approach can beneft not only the nodes nvolved, but the whole network n many dfferent aspects. Although orgnatng from the physcal layer cooperaton, all the benefts can not be fully realzed untl proper mechansms have been ncorporated at hgher protocol layers (e.g. MAC, network) and the necessary nformaton s made avalable from the lower layers (e.g. PY). One of the most relevant areas n cross-layer optmsaton s the nteracton between physcal (PY) and medum access control (MAC) layers n wreless networks. Ths s due to the fact that the PY layer s the most tme varant entty n a wreless communcatons system and due to the proxmty of the two layers n the OSI model stack and the nherent varablty of the channel state. The MAC layer n wreless network s mplemented to enable nodes to access the avalable channel(s) whle attemptng to enforce a far and effcent usage of the channel(s). To accomplsh ths task, the MAC protocol makes use of nput or feedback nformaton that other layers of the protocol stack may forward to t drectly or ndrectly. Typcally, however, the MAC layer s mostly nterested n the nformaton t receves from the underlyng physcal (PY) layer regardng the state(s) of the channel(s) and/or the occurrence of any events that are key to ts operaton (e.g., the successful msson of a frame over the channel). Based on the feedback nformaton, the MAC protocol dynamcally adusts ts behavor n order to better allocate the channel(s) among competng nodes wthn the network. The PY layer, on the other hand, has the man ob of recevng the bts of nformaton from the MAC layer and, at the MAC's dscreton, mt the bts across the underlyng communcaton channel(s) as fast and relable as possble, accordng to approprate (de)codng and (de)modulaton schemes. The lkelhood wth whch a msson s successful wll depend on how well the sgnalng used defends aganst channel mparments and nterference from any source. In wreless ad hoc networks, n partcular, the sgnal mssons from any node can potentally nterfere wth sgnal receptons at any other node n the network. ence, the qualty of a rado lnk depends on the msson actvty n the entre system. As a result, each node's msson actvty can affect the PY-layer performance at every node n the network, whch, n turn, can affect ther MAC dynamcs. Clearly, the dynamcs of the MAC layer s tghtly connected to the dynamcs of the PY layer, and the cross-layer nteractons at each node

2 wll depend, fundamentally, on the actvty of every node n the network. Wth ths dea n mnd, ths paper descrbes a ontly optmal desgn of the medum access and physcal layer protocols for the hybrd Token-CDMA network. Usng the cross-layer nteracton, the PY layer provdes channel state nformaton to be feed backed to MAC layer and based on the nformaton, the MAC scheme accurately estmates the traffc loadng condton and modfes the msson rate by changng the spreadng factor of each msson. Therefore, the dstrbuted rate adaptaton [9], [10] through spreadng factor selecton uses both the traffc nformaton provded by the MAC algorthm and the channel estmate from the PY layer, whch consttutes the cross-layer concept. The novelty of ths work les n presentng a cross-layer nteracton between PY and MAC layers and conductng a performance analyss of the optmzed hybrd token-cdma MAC scheme. The remander of ths paper s organzed as follows. Secton II presents the comprehensve descrpton of the hybrd Token-CDMA MAC scheme. Secton III s devoted to the modelng of cross-layer framework between physcal and MAC layers for the hybrd Token-CDMA scheme. Both the smulaton model and results are presented n Secton IV and conclusons are drawn n Secton V. II. YBRID TOKEN-CDMA MAC SCEME DESCRIPTION In ths secton, a comprehensve descrpton of the hybrd MAC protocol s dscussed. Detaled dscusson on how the proposed MAC scheme deals wth ssues on token protecton, node on and lost, network ntalzaton and provson for multple rngs s presented n [11]. Ths secton presents the structure, characterstcs and propertes of the new protocol. λ p,1 Staton 1 λ p, Staton T λ p,3 Staton Token Fg.. Network model for hybrd Token-CDMA MAC protocol 3 A. Network Topology and λ p, N The hybrd MAC scheme s capable to mplement n ether ad hoc or wreless mesh networks (WMN). For WMN confguraton, statons are served as access networks utlzng non-moble relayng nodes to provde wreless backbone servces for nomadc users to access the wred Internet. In ths paper, t s assumed that the network s a dstrbuted, de-centralzed ad hoc wreless network that conssts of N statons wth ncorporaton of M CDMA codes ( M < N ). In Fg., denotes the queue model of staton and T s the N Staton 1 M token msson tme between two statons and λ p, s the packet arrval rate for staton. Each staton s assumed to have the ablty to communcate wth ts adacent statons over a sngle-hop. The token mplemented n the hybrd token-cdma scheme s used to dstrbute M CDMA codes n the network. Each staton s equpped wth two cevers; one f used for token msson and the other s used for data msson. Each staton s also assumed to be equpped wth a MUD (multple user detecton) capabltes, n order to be able to receve multple mssons smultaneously. B. Channel Access Control Once the token s generated n the network, t contnuously crculates wthn the network, followng a predetermned order. To montor the usage of the CDMA codes, the token uses the parameter NOC ( 0 NOC M ) to control the amount of traffc flow n the network. If the staton has data packets that t wshes to mt, t has to wat for the arrval of the token. When a staton s vsted by a token, t forwards the token to the successor staton wthout capturng t under two crcumstances. In the frst scenaro, the staton forwards the token f the staton tself s stll busy mttng data packets from prevous token cycle. A token cycle ( TRT ) s the tme for the token to vst all the statons n the network ( TRT = T ), T s the N = 1 tme for a token to travel from staton to ts successor staton +1. In the second scenaro, the staton forwards the token f t has fnshed mttng data packets and also occupes a code channel. In the latter scenaro, the staton has to release the code back to the token rrespectve of whether or not t has packets that t wshes to mt. Ths s the global farness algorthm [1] employed to ensure that all the statons have the same opportunty to access the network. The algorthm s mplemented to avod the staton from constantly wthholdng the code for ts own msson, thereby dsruptng the farness of the network. Once the code s released, the token ncrements ts NOC value by one. If the staton s nether mttng nor has a code channel when the token arrves, t may capture the token f the os requrement s met. If the staton fals to satsfy modfed leaky-bucket os guarantee [11], t wll then be oblged to forward the token to ts successor staton accordng to the pre-defned order. If the token has been successfully captured, the permsson for msson s now acknowledged, the staton decrements the NOC value by one, ndcatng that t has occuped a code channel. Usng that specfed code channel, the network allows the staton n the rng to send ts packets. The gated servce dscplne has been appled to awatng packets n the queue buffers and packets are then served accordng to frst-n-frst-out (FIFO) prncple. The staton forwards the token to ts successor n the pre-defned order before t begns wth ts data msson. Ths assumpton permts the token to be relayed to next staton before the actual data msson starts. The polcy of not wthholdng token whle mttng data packets has the advantage of decreasng the overhead of the token rotaton tme

3 3 and consequently leads to mprove network bandwdth utlzaton. To mt data packets, the sender staton frst looks up the code assgnment table to search for the avalable code of ts ntended recever staton. The code assgnment table s read from the token s Channel-st (C-st) when t has vsted the staton. C-st dsplays the avalable code channels for each staton. For the msson and recepton of data packets, the sender staton nserts the destnaton staton s address nto the data packet before msson. The data packets are then sent usng the receved code. The recevng staton, wth ts MUD ablty, constantly montors ts code channels to detect any ncomng data traffc destned for t. III. CROSS-AYER MODEING FRAMEWORK FOR YBRID TOKEN-CDMA MAC SYSTEM It s dscussed n secton II that the network s capable of smultaneously processng a maxmum number of M mssons,.e., the maxmum number of actve users supported by the network s M. Each actve user s data s BPSK modulated and s mtted to the recever node asynchronously. Followng the cross-layer phlosophy, assume that the MAC protocol s aware of the channel state of the lnks between nodes. Among the functons of the MAC layer, the man obectve of optmzng the packet delay s acheved where the MAC layer montors the nformaton forwarded from the PY layer and ts own layer and changes ts msson rate by selectng the approprate spreadng factors accordngly thereby reducng the mutual nterference. Recent works n cross-layer desgn of PY-MAC layers n CDMA networks demonstrated the possblty of desgnng more flexble collson recovery strateges [3] by employng sgnal processng methodologes to dscrmnate multple colldng frames on the medum. The effect leads to reducton n collson ntervals and number of remssons. In ths paper, the concept s appled to the am of enablng dynamc adaptaton of the msson parameters n order to mprove data msson performance. A. Physcal ayer propertes CDMA System It s assumed that BPSK modulaton s used n the system and all nodes mplement ampltude modulaton (AM) that s ndependent of the bt rate. Wth the mplementaton of dynamc spreadng factor adustment, the current system can be denoted as a mult-rate CDMA system that supports n (n=m) dfferent rates or subsystems. The mtted sgnal of user number k, n subsystem, s then of the form [13] r t = P a t b t cos ω t + φ (1) ak ( ) ( ) ( ) ( ) k k k c k where P s the power of each user n the subsystem and t s a pulse ampltude modulaton sgnal wth a rectangular ( ) pulse shape of duraton bk t s the spreadng code waveform, consstng of N perodcally repeated chps n a bnary polar format wth rectangular pulse shape of duraton T c (duraton per chp) therefore T = NTc. The modulator phase φ k T. And ( ) are modelled as ndependent random varables, unformly dstrbuted over 0,π ). Channel In moble rado envronments, the channel lnk performance s known to be dependent on the receved desred sgnal strength n whch t depends on the propagaton loss and shadowng. The sgnal strength s normally lmted by the co-channel nterference (CCI) n the system. In ths chapter, the shadowng effect (.e. slow fadng) s assumed to be a log-normal dstrbuted random varable as t s characterzed by [14]. And based on [15], the shadowng spatal correlaton s assumed to be exponental decayed that s dependent on the dstance between any two separate postons and the varaton s modelled as a Gaussan-Markov stochastc process. The channel nterference s approxmated as Gaussan, the derved equaton for the performance on a sngle path gven the ampltude from [13] s used and modfed for a mult-rate BPSK system. The bt error rate performance of user wthout antenna dversty s known to be BER ( γ ) ( γ ) = () where γ s the sgnal strength and t can be derved as 1 χ γ = K R χ N 0 R + Eb 3N where, Eb N 0 represents the sgnal-to-nose rato (SNR), K s the number of codes that are currently beng used n the system and N s the spreadng gan used by user and R s the user s bt rate. χ s a log-normal random varable representng shadowng effect of user and can be defned as Ω ( k ) 10 χ = P 10 (4) where Ω k s the receved co-channel nterference power at locaton k. Wth the spatal correlaton property of shadowng effect modelled as a Gaussan-Markov process, f the spatal dstance between k and k+1 can be represented [15] as Ω [ k ] = υ Ω [ k 1] + ( 1 υ ) V[ k ] (5) P s the nstantaneous power of user and ( ) where υ s the spatal correlaton coeffcent and the Ω k s a Gaussan receved co-channel nterference power ( ) random varable wth mean assumed to be k = and varance σ. V k s a Gaussan random varable wth mean and varance σ 1 V = ( + ) ( 1 υ ) υ σ Usng the BC error correctng code, the frame error rate can be derved as (3) (6)

4 FER = 1 1 BER 1 BER = 4 31 ( ( γ )) ( ( γ )) (7) where s the length of the (31, 16) BC blocks. B. MAC ayer Propertes Cross-ayer Interacton and Optmzaton In the ntalzaton stage, each node enters the network mts data usng a pre-defned set of msson parameters that enables the hghest bt-rate on the channel. In parallel, t lstens to the token channel for token recepton. For data msson, t s assumed that a data packet s a frame. At MAC layer, t s assumed that there are n spreadng factors havng values expressed as N1 < N < < Nn. For the msson of the frame, the spreadng factor s selected from the set { N1, N, Nn}, based on the nformaton forwarded from the PY and MAC layer. The frame error rate s determned from the msson and forwarded from the PY layer to the MAC layer (cross-layer nformaton). At the MAC layer, a rate-adaptaton algorthm s mplemented where t wll choose the sutable spreadng factor for the msson of the next frame. Rate Adaptaton Algorthm at MAC ayer The algorthm s responsble for choosng the sutable spreadng factor based on the nformaton forwarded from the PY and MAC layers tself. For each frame, the frame error rate of the frame s forwarded to the MAC layer from the PY layer. If the BER s hgh, then the algorthm wll choose a hgher spreadng factor to counter the nterference. owever, t s known that wth the ncrease n spreadng gan, the msson tme also ncreases whch leads to ncrease n packet delay. Therefore n order to effcently mantan a low BER and at the same tme acheve low packet delay, the MAC layer, n ths case, montors the load condton ( ρ ) of the node. λ ρ λ µ α, n, =, tot =, NT c 1 FER, where ρ, s the loadng condton of the node wth packet, λ, tot s the total packet arrval rate ncludng remsson packets to node, µ s the servce tme of packet, λ, n s the packet arrval rate to node FER, s the approxmated frame error rate of node s th packet, α, s the packet length of packet and T c s a chp duraton. In ths case, both of the BER and spreadng gan are taken nto consderaton when choosng the optmum gan settng. It s known that the packet delay s related to the loadng condton, therefore f the low loadng condton s acheved; the packet delay wll also be mnmzed. ow loadng condton s attaned by choosng the optmum spreadng factor. Before the packet s servced, the channel loadng condton s predcted from (8). Usng the set of predefned spreadng factors, each factor s mplemented to predct the total packet arrval rate and ts servce tme. Amongst all the values attaned from the predcton, the node selects the spreadng factor that acheves ε (8) the lowest load condton and mts the packet usng that settng. Ths procedure s performed before every packet msson n the system. Usng the algorthm, the nodes are provded wth the optmzaton n terms of frame error rate and packet delay. In ths way, f few nodes are actve n the network, they can explot the avalable resources by decreasng ther spreadng factors; whle as the number of nodes or traffc ntensty grows, those nodes whch suffer most from nterference can self-adust to dfferent msson parameters n order to provde a hgher level of robustness. The proposed rate adaptaton scheme s amed at provdng contnuous montorng of the performance acheved by the users and selectng better msson parameters to those who are sufferng severe sgnal degradaton due to nterference. IV. SIMUATION MODE AND RESUTS The proposed approach s valdated through extensve smulatons. Ths secton presents the acheved results n terms of performance at the physcal/mac layers. The effectveness of the cross-layer approach scheme and the orgnal hybrd MAC scheme are compared for varous performance metrcs. The metrcs used for the evaluaton are throughput and queung delay of the packets. A smulaton model usng C++ bulder software package s bult. The model s based on event drven packet level smulator for montorng and recordng results. A. Traffc model For the hybrd MAC scheme, t s assumed that the arrval traffc process s typcally descrbed by an Markov-modulated Posson processes (MMPP) [16]. Four parameters are used to represent the -state MMPP source of each traffc class. ψ ( ψ ) s defned as the mean ton rate out of the ow load Symbol TABE I SYSTEM PARAMETERS Parameter Number of nodes (N) 30 Number of codes (M) 15 G set 8,16,3,64,18 spreadng Mean packet sze ( α ), 56 bts Error correcton scheme BC (16,31) FEC correctable bts 3 bts CDMA chppng rate (1 T c ) 3.84 Mcps Modulaton BPSK Fadng channel og-normal Token walk tme ( T ) 100 us Packet buffer capacty ( ) 0000 Bad state duraton ( τ bad ) 3. ms Bad state PEP bad 1.0 Sgnal to Nose Rato (SNR) 5 db Spatal correlaton coeffcent (υ ) Varance of V r.v.( σ ) dB

5 5 (gh load) state, and λ s the mean arrval rate of the Posson process n the ow load state and λ corresponds to bursts of hgh arrval load state for node. The effectve combned Posson arrval rate for node s then gven by λ ψ + λ ψ λp, = (9) ψ + ψ For the smulaton, t s assumed that traffc arrves as packets of varyng lengths to dfferent class nodes and the packet length s geometrcally dstrbuted wth mean packet length of ξ bts. All parameters used n the smulaton are summarzed n Table I. B. Performance Analyss In Fg. 3, the code utlzaton graph comparng the performance of the cross-layer hybrd scheme wth normal hybrd scheme s shown. Ths plot provdes an ndcaton of the optmum code usage n the network. The code utlzaton s plotted aganst the offered load, whch s defned to be the traffc created for the entre network. In the fgure, the cross-layer hybrd scheme s performance s compared wth standard hybrd scheme that mplemented fxed spreadng factors (spreadng factors of 16 and 64). When the traffc load ncreases n the network, t creates hgh packet error probablty due to severe multple access nterference (MAI) as shown n Fg. 4. Ths effect s especally evdent under heavy traffc load, and t s known that low spreadng factor always generated hgh packet error rate, therefore t reaches maxmum code utlzaton before the hgh spreadng factor and cross-layer schemes due to hgh remsson rate n whch caused by hgh packet error rate. Usng the rate adaptaton algorthm and cross-layer dalogue, cross-layer hybrd scheme s able to mantan optmum code utlzaton under any load condtons. r spreadng factor scheme suffered the worst performance when load ncreases. The hgh spreadng factor and cross-layer schemes exhbt relatvely consstent error probablty due to ther hgh spreadng gan and flexble gan settngs. The algorthm fnds the equlbrum between the packet error rate and spreadng factor, t wll choose lower spreadng factor f t s needed to mnmze delay performance, under the condton that packet error rate s stll wthn tolerable range. Ths s llustrated n delay performance shown n Fg. 6. Fg. 4. Packet error probablty of standard hybrd protocol wth dfferent spreadng gan settngs and hybrd MAC protocol wth cross-layer optmzaton wth 15 CDMA codes assgned n the system under dfferent load condtons Throughput performance of all the schemes s shown n Fg.5. It s clearly shown from the fgures that the cross-layer scheme acheved better throughput performance than other two schemes. It s llustrated n the fgure that as the load ncreases, the performance for low spreadng factor hybrd scheme quckly deterorates as ts packet error probablty ncreases drastcally. The packets are therefore sufferng from hgh remsson rate, whch consequently leads to a decrease n throughput as shown n Fg. 5. Fg. 3 Code utlzaton of standard hybrd protocol wth dfferent spreadng gan settngs and hybrd MAC protocol wth cross-layer optmzaton wth 15 CDMA codes assgned n the system under dfferent load condtons Fg. 4 dsplays the packet error probablty for all schemes. Ths metrc reflects the overall system performance of the schemes. It s clearly shown n the fgures that the packet error probablty ncreases wth an ncrease n load and the low Fg. 5. Throughput of standard hybrd protocol wth dfferent spreadng gan settngs and hybrd MAC protocol wth cross-layer optmzaton wth 15 CDMA codes assgned n the system under dfferent load condtons

6 6 The packet delay s defned as the tme perod from the tme when a packet arrves at the packet buffer of a node to the tme t s successfully mtted to the ntended recevng staton. Fg. 6 dsplays the mean packet delay for all schemes. From the fgure, t s clearly ndcated that the cross-layer hybrd scheme outperformed the other two schemes under any load condtons. For the ncrease n load condton, low spreadng factor scheme suffered the worst performance due to ts hgh packet error rate as dscussed prevously. Fg. 6. Mean packet delay of standard hybrd protocol wth dfferent spreadng gan settngs and hybrd MAC protocol wth cross-layer optmzaton wth 15 CDMA codes assgned n the system under dfferent load condtons V. CONCUSION Cross-layer technques n whch dfferent layers of wreless communcaton systems nterchange control nformaton n order to optmze the use of the scarce bandwdth are proven to be a relatvely unexplored research area where tremendous potental benefts can be acheved. It s dscussed n ths paper that the nteracton between PY and MAC layers has been the frst explored ssue n ths area. In ths paper, a PY-MAC cross-layer optmzaton for hybrd token-cdma based wreless networks has been presented. To account for the effects of both cross-layer nteractons and the nterference among all nodes, a novel model was ntroduced wth whch topology and PY/MAC-layer aspects are naturally ncorporated nto the nodes. The model s used to buld a brdge between the physcal and MAC layers and to balance the effcency and farness of resource allocaton. In partcular, the necessary and suffcent condton for fndng an optmum system performance for the hybrd scheme s nvestgated when rate-adaptaton algorthm s used. The cross-layer nteracton proves to mprove the spectrum effcency, keepng the packet delay at the mnmum possble value for dfferent code and traffc load settngs. REFERENCES [1] S. Shakkotta, TS. Rappaport, and PC. Karlsson, Croass-layer desgn for wreless networks, IEEE Commun. Mag., vol. 41, no. 10, pp , Oct, 003. [] C, Sacch, GB. Scorza, F. Granell, and F. Natale, A ont MAC-PY approach for Medum Access Control n VBR MC-CDMA broadband ndoor connectons, IEEE Wreless Personal Communcatons, vol. 36, no.1, pp , 006. [3] G. Dmc, N. Sdropoulos, and R. Zhang, MAC-PY cross-layer desgn, n IEEE Sgnal Processng Mag, Sep., pp , 004. [4]. u, S. Zhou, and G. Gannaks, Cross-layer combng of adaptve modulaton and codng wth truncated AR over wreless lnks, IEEE Transacton on Wreless Comms, vol. 3, pp , Sep., 004. [5]. Wang and M. Aburgheff, Cross-layer sgnallng for next generaton wreless systems, IEEE WCNC 03, vol., no.1, pp , 003. [6] E. Yeh, and A. Cohen, A fundamental cross-layer approach to uplnk resource allocaton, IEEE MICOM, vol. 1, no. 6, pp , 003. [7] M. Schaar, S. Krshnamachar, S. Cho and X. Xu, Adaptve cross-layer protecton strateges for robust scalable vdeo msson over WANs, IEEE Journal on Selected Areas n Comms., vol. 1, no. 10,, pp , Dec 003. [8] G. Camero, J. Ruela, and M. Rcardo, Cross-layer desgn n 4G wreless termnals, IEEE Wreless Commun, vol. 11, No., pp. 7 14, Aprl, 004. [9] F. Adach, M. Sawahash, and K. Okawa, Tree-structured generaton of orthogonal spreadng codes wth dfferent lengths for forward lnk of DS-CDMA moble rado, Electroncs etters, vol.33, no.1, pp. 7 8, [10] F. Adach, M. Sawahash, and. Suda, Wdeband DS-CDMA for next generaton moble communcatons systems, IEEE Comms Mag., vol.36, no.1, pp , [11] Y.S. u, F. Takawra, and. Xu, ybrd Token CDMA MAC Protocol for Wreless Networks, submtted to IEEE Transacton on Moble Computng, 006. Avalable: u07%0ieee%0journal%0ybrd%0token-cdma%0mac% 0.pdf [1] J. S. C. Chen, I. Cdon, and Y. Ofek, A local farness algorthm for ggabt AN s/man s wth spatal reuse, IEEE Journal on Selected Areas n Communcatons, vol. 11, pp , October, [13] T. Ottosson, and A. Svensson, Mult rate schemes n DS/CDMA systems, IEEE VTC, vol. 1, pp , July, [14] G. Stuber, Prncples of Moble Communcaton, nd Edton, Boston, Kluwer Academc Publshes, 001. [15] M. Gudmundson, Correlaton model for shadow fadng n moble rado systems, IEEE Electronc etters, Vol. 7, pp , [16]. effes and D. M. ucanton, A Markov Modulated characterzaton of Packetzed Voce and Data Traffc and Related Statstcal Multplexer Performance, IEEE Journal on Selected Areas n Communcaton, Vol. 4, No. 6, pp , Sept Prncpal Author: Mr. Y-Sheng u attaned hs BScEng degree n 001 and MScEng degree wth Cum aude n 003 at the School of Electrcal, Electronc and Computer Engneerng at the Unversty of Natal, oward College Campus, Durban. e s currently dong hs doctoral degree, researchng medum access control (MAC) protocols for wreless networks n the Rado Access Technology Research Centre at the Unversty of KwaZulu-Natal, oward College Campus. Co-author: Professor Fambra Takawra s the senor Professor n the faculty of Electronc Engneerng, and head of the School of Electrcal and Electronc Engneerng at the Unversty of KwaZulu-Natal. e s also the drector of the Centre for Rado Access Technology (CRAT) at the same unversty. e holds a BScEng wth frst class honours n Electrcal Engneerng from Manchester Unversty and a PhD degree from Unversty of Cambrdge, Unted Kngdom. s research nterests are n the general areas of adaptve sgnal processng, dgtal and wreless communcatons and data networks. Co-author: Professor ong-jun Xu s the assocate Professor n the School of Electrcal, Electronc and Computer Engneerng at the Unversty of KwaZulu-Natal, oward College Campus. e receved the BSc degree n 1984 from the Unversty of Guln Technology and the MSc degree from the Insttute of Telecontrol and Telemeasure n Sh Jan Zhuang, 1989, and the PhD degree from the Beng Unversty of Aernautcs and Astronautcs n Beng, s research nterests are n the area of dgtal and wreless communcatons and dgtal systems.