Multi-user cross-layer allocation design for LP-OFDM high-rate UWB

Transcription

1 Multi-uer cro-layer allocation deign for LP-OFDM high-rate UWB Ayman Khalil, Antoine Stephan, Matthieu Cruière, Jean-Françoi Hélard To cite thi verion: Ayman Khalil, Antoine Stephan, Matthieu Cruière, Jean-Françoi Hélard. Multi-uer cro-layer allocation deign for LP-OFDM high-rate UWB. International Sympoium for Wirele Communication Sytem, Oct 008, Reykjavik, Iceland. pp.1-5, 008, < /ISWCS >. <hal > HAL Id: hal Sumitted on Dec 008 HAL i a multi-diciplinary open acce archive for the depoit and diemination of cientific reearch document, whether they are pulihed or not. The document may come from teaching and reearch intitution in France or aroad, or from pulic or private reearch center. L archive ouverte pluridiciplinaire HAL, et detinée au dépôt et à la diffuion de document cientifique de niveau recherche, pulié ou non, émanant de étaliement d eneignement et de recherche françai ou étranger, de laoratoire pulic ou privé.

2 Multi-uer Cro-layer Allocation Deign for LP-OFDM High Data Rate UWB Sytem Ayman Khalil, Antoine Stephan, Matthieu Cruière and Jean-Françoi Hélard Intitute of Electronic and Telecommunication of Renne (IETR) INSA, 0 Avenue de Butte de Coeme, Renne, France ayman.khalil@ina-renne.fr Atract In thi paper, we invetigate a cro-layer deign for the packet cheduling and the reource allocation in UWB ytem. Thi deign conider the comination of queuing and channel tate information (CSI) which provide QoS upport for multimedia application in UWB. For the phyical layer, the ue of a linear precoded orthogonal diviion multiplexing (LP- OFDM) waveform i propoed ecaue of it ignificant performance increae compared to the WiMedia propoal. For the medium acce control layer, cheduling i performed in order to differentiate etween the different uer and to atify their quality of ervice contraint. Thi cro-layer approach optimize the ytem pectral efficiency and olve the prolem in the WiMedia olution of cohaitation of more than three uer haring the three u-and of the ame channel. Simulation reult how that the propoed cheme lead to a coniderale improvement in reource allocation and can guarantee the required quality of ervice. I. INTRODUCTION Ultra-wideand tranmiion i an emerging technology for future high-rate, hort-range wirele communication. It wide andwidth and low tranmiion power denity make it attractive to reearcher ince 00 when the Federal Communication Commiion (FCC) regulated UWB ytem y allocating the 3.1 to 10.6 GHz pectrum for unlicened ue of UWB [1]. In order to reduce interference with other exiting ytem, the FCC impoed a power pectral denity (PSD) limit of dbm/mhz. The IEEE 80.15a wirele peronal area network (WPAN) tandardization group defined a very high data rate phyical layer aed on UWB ignalling. One of the multipleacce technique conidered y the group i a multiand orthogonal frequency diviion multiplexing (MB-OFDM) upported y the Multi OFDM Alliance (MBOA) and the WiMedia forum [], [3], which merged in March 005 and are today known a the WiMedia Alliance. There have een a lot of tudie on the reource allocation in UWB ytem aed on the WiMedia olution. However, to thi date, mot reearch tudie on multiand UWB ytem have een devoted to the phyical layer iue. In [4] [6], the author propoe different u-and and power allocation trategie aed on the channel information without taking into account the uer requirement and the quality of ervice iue. On the other hand, in [7], [8], the author propoe cheduling and power allocation algorithm that provide quality of ervice for multimedia application in UWB ytem ut without having a full knowledge of the channel. The purpoe of thi paper i to propoe a cro-layer reource management aed on UWB ignalling in a multiuer context. The propoed cheme jointly exploit cheduling and u-and allocation principle to maintain an efficient ue of the pectrum in a multiple medium acce demand. In addition, we conider a phyical layer aed on a linearprecoded OFDM tranmiion cheme (LP-OFDM) alo known a pread pectrum multicarrier multiple-acce (SS- MC-MA) in the wirele context [9]. Baed on thi cheme, we hall conider the effect of the ource tatitic and cheduling in addition to the channel information provided y the phyical layer model. Therefore, we need the contriution of the MAC layer for etter QoS and higher utilization. A packet cheduler i defined y the MAC layer in order to control each uer requirement, to manage the cohaitation of more than three uer in one channel and to differentiate etween two major traffic clae: real-time or Quality of Service (QoS), and non real-time or Bet Effort (BE). The definition of thee two clae i ueful for UWB ytem, where multimedia or real-time application (video recording, A/V conferencing, etc) hould have a certain priority on data or non real-time application (file tranfer, wirele USB, etc). The remainder of thi paper i organized a follow. Section II introduce the ytem model y preenting the phyical and MAC layer condition. Section III preent the propoed cheme that reflect the new concept of allocating the reource aed on a cro-layer trategy. Section IV preent ome imulation howing the efficiency and the good performance of the new cheme and it interet for UWB ytem. Finally, ection V conclude thi paper. A. Phyical Layer II. SYSTEM MODEL 1) The WiMedia Solution: The WiMedia olution conit of comining OFDM with a multi-anding technique that divide the availale and into 14 u-and of 58 MHz, a illutrated in Fig. 1. An OFDM ignal can e tranmitted on each u-and uing a 18-point invere fat Fourrier tranform (IFFT). Out of the 18 ucarrier ued, only 100 are aigned to tranmit data.

3 Channel 1 Channel Channel 3 Channel 4 Channel K code (MHz) Fig. 1. Channel ditriution for WiMedia olution Different data rate from 53.3 to 480 Mit/ are otained through the ue of forward error correction (FEC), frequencydomain preading (FDS) and time-domain preading (TDS). The contellation applied to the different ucarrier i either a quadrature phae-hift keying (QPSK) for the low data rate or a Dual Carrier Modulation (DCM) for the high data rate. Time-frequency code (TFC) are ued to provide frequency hopping from a u-and to another at the end of each OFDM ymol. TFC allow every uer to enefit from frequency diverity over a andwidth equal to the three u-and of one channel. In WiMedia, there i no clear olution for imultaneou tranmiion uing TFC and conflict etween uer appear when a fourth uer i added in one channel. ) LP-OFDM Sytem: The LP-OFDM cheme i applied to UWB while repecting the OFDM parameter of the WiMedia olution. The ytem evolution reduce in practice to a imple addition of a precoding lock in the tranmiion chain. Taking into account the frequency electivity and the lowtime variation of the UWB channel in an indoor environment, the preading equence are applied in the frequency domain. Thi preading component improve the ignal routne againt frequency electivity and narrowand interference, ince the ignal andwidth ecome much larger than the coherence and interference andwidth. Moreover, it increae the reource allocation flexiility a the preading code dimenion offer an additional degree of freedom [10]. We will aume that orthogonal preading equence are ued in the propoed ytem. A chematic repreentation of the LP-OFDM ignal i illutrated in Fig.. At a given time, each uer i allocated one of the firt three WiMedia u-and of 58 MHz andwidth each, in order not to increae the ytem complexity compared to the WiMedia olution. Each u-and i then divided into everal lock, each of them including a numer of ucarrier equal to the preading code length L. In a general approach, the generated ymol vector at the output of the OFDM modulator for an LP-OFDM ytem can e written a H = F MX. (1) Vector i N-dimenional, with N the numer of ued ucarrier. X = [ x1,..., x ] T k i the output of the erial-toparallel converion of the K QPSK-mapped ymol to e tranmitted. M repreent the N K precoding matrix applied to X, which precode K ymol over the N ucarrier. H Finally, F repreent the Hermitian of the N N unitary Fourier matrix that realize the multicarrier modulation. f f (MHz) 1 3 Code Time Frequency Uer 1 Uer Uer 3 Spread Symol Fig.. LP-OFDM chematic repreentation for three uer occupying the firt three u-and of the WiMedia olution Note that for implicity reaon, (1) doe not involve any guard interval contriution even if ZP ymol extenion i ued in practice a in the WiMedia olution. The generated ymol vector applied to each u-and can e retated a C1 X 1 0 O M H S = F D C X, () 0 O M C B X B where B i the numer of lock in the u-and with B L= N, C the precoding matrix containing the K precoding equence of lock, and X the K-dimenional vector containing the ymol to e tranmitted within lock. In addition, a permutation matrix, denoted D, i ued to interleave the chip reulting from the precoding proce in the frequency domain. B. MAC Layer The MAC layer model i illutrated in Fig. 3. The two main tak in thi layer are claification and cheduling. 1) QoS and ervice clae: The MAC layer i reponile for QoS upport for the multimedia application. Hence, ecaue of the variou type of multimedia or real-time application that can e ued in UWB ytem and that have trict QoS requirement (video recording, A/V conferencing, interactive gaming, etc), we define a two traffic clae model that differentiate etween two traffic type: QoS (i.e. realtime traffic) and BE (i.e. non real-time or data traffic). The main goal i to guarantee a certain amount of reource for QoS traffic and to hare the remaining amount of reource among BE traffic, in order to atify the multimedia application contraint uch a the error rate, the throughput, the delay, etc. ) Scheduler: Scheduling i a main tak in our model. Therefore, we define a cheduler that hall e ale to claify all the uer into the traffic clae (QoS or BE) and to aign a priority level for each traffic aed on the traffic requirement and contraint (i.e. type of application, delay). L

4 MAC uer 1 uer uer n.... Algo_MAC N_uer : Total numer of uer N_u : Numer of uand Uer_n_Req : Requet of uer n in term of OFDM ymol Scheduler Start QoS cla BE cla Reource Allocation Define the uer n, define Uer_n_Req PHY Channel 1 WiMedia CSI Claify the uer : QoS or BE Fig. 3. Cro-layer model After aigning thi priority level, the cheduler comine it with the channel tate information (CSI) provided y the phyical layer, in order to make the allocation deciion for each uer a explained in the next ection. No Run Algo_PHY Allocate the claification parameter (weight) q1,..., q n the different uer N_uer > N_u to Ye Run Algo_Scheduler III. PROPOSED SCHEME In a multi-uer context taking into account the QoS iue, the aim of the propoed cheme i to improve the performance of the uer having higher priority than other uer, which in turn hall improve the total ytem capacity. The total throughput in it per ymol of a LP-OFDM uer uing a zero-forcing (ZF) detection i given y [11] R = 1 + u B C c, log 1, L 1 c 1 N = = Γ 0 i= 1 L ( 1 hi, ) where Γ i the SNR gap of the quadrature amplitude modulation (QAM), B the numer of lock, C the numer of ued code in lock, L the preading code length, hi, the frequency-domain repone of ucarrier i in lock and P the power allocated to code c within lock. In UWB c, cae, thi power denity repect the following condition: C Pc, P and c, 0,, c= 1 P (3) (4) with P related to the PSD limit defined y the regulation authoritie. In order to optimize the reource allocation of the whole ytem, the um of all uer throughput mut e maximied. In other term, we have to maximie each ingle uer throughput while repecting it QoS requirement. Therefore, the prolem of haring the pectrum ecome critical, particularly when the numer of uer exceed the numer of the u-and on one channel. To olve thi prolem, we propoe a cro-layer algorithm that allocate for each uer the u-and that reflect it requirement provided y the QoS (MAC) and the appropriate throughput provided y the CSI (PHY). Algo_PHY Sand H = U = = { 1,..., S} { 1,..., K} { 1,..., N} n, Algo_Scheduler Start w h ile ( U ) Compute the uer power for each uand Pn = Hi, i H Allocate to the uer n having the highet ( q n + m ax Pn ) 1. The mot powerful uand. The highet rate (from Tale I ) atifying it requirement U = U n S = S Uer_n_Req : Requet of uer n in term of OFDM ymol BE_Req_Tot : Total numer of BE requeted OFDM ymol N_Su_Tot : Total numer of uand N_uer_QoS : Numer of QoS uer N_uer_BE : Numer of BE uer Start Guarantee the QoS uer N _ Su _ Tot - = N _ uer _ QoS S and Share the remainder uand etween BE uer proportionally uch a : Uer _ n_ n _ ym = q Uer _ n_ Re q Fig. 4. Cro-layer algorithm n

5 The algorithm i preented in Fig. 4. A illutrated in the figure, there are two main procee, each running at a level (one at MAC level and one at PHY level). At the MAC level (Algo_MAC), the uer are defined and characterized y their type and requirement that are tranlated into two effective parameter: the claification type (QoS or BE) and the numer of OFDM ymol needed (or requeted). After the claification, a weight q i attriuted for each uer (or application). More the uer i high-claified, more hi weight q i high. Then, it i checked if the numer of uer exceed the numer of availale u-and; if it i the cae, a cheduler (Algo_Scheduler) define a trategy to allocate all the uer in the availale pectrum in an efficient way that repect the claification that differentiate etween the availale uer, otherwie we move to the phyical layer (Algo_PHY) in order to allocate to the claified uer the availale u-and. To do o, a priority level i aigned to each uer in order to repect the main concept of differentiating the uer. Having two eential parameter, the claification coefficient vector (provided y the MAC layer through the weight attriution) and the channel matrix (through the CSI knowledge) of all the uer, a priority level PL, which i the comination of thee two parameter, i aigned to each uer (PL= weight + CSI). The uer that ha the higher priority level i allowed to chooe the mot powerful u-and. The procee continue a long a there are uer that are not aigned a u-and. A. Channel Model IV. SYSTEM PERFORMANCE The channel ued in thi tudy i the one adopted y the IEEE a committee for the evaluation of UWB phyical layer propoal [1]. Thi model i a modified verion of Saleh-Valenzuela model for indoor channel [13], fitting the propertie of UWB channel. Four different channel model (CM1 to CM4) are defined for the UWB ytem modelling, each with arrival rate and decay factor choen to match different uage cenario and to fit line-of-ight (LOS) and non-line-of-ight (NLOS) cae. B. Simulation Reult In thi ection, we preent the imulation reult for the propoed cro-layer allocation cheme and we compare the performance of the new cheme with the performance of WiMedia and LP-OFDM olution uing TFC. Therefore, we ue the propoed LP-OFDM data rate (ee Tale I), which are cloe to the one propoed y the WiMedia olution. The reult are performed on the firt three WiMedia u-and ( GHz) for CM1 channel model. In Fig. 5, we how through an example the trategy of cohaitating four uer in one channel. The highet priority level uer (QoS-PL1) i allocated the mot powerful u-and (u-and ), the QoS-PL i allocated u-and 1 and the BE Data rate (Mit/) TABLE I LP-OFDM SYSTEM DATA RATES Modulation Coding Rate Load (C) Coded it Per ymol 51. QPSK 1/ QPSK 1/ QPSK 1/ QPSK 1/ QPSK 1/ QPSK 1/ QPSK / QPSK 3/ uer (BE-PL3) have to hare u-and 3 in a proportion depending on their priority level (in thi example they have the ame priority level). In thi cae, the BE uer are forced to ue half of the allocated rate and they will have a delay that i two time greater than the QoS uer delay. Thi degradation in performance of BE uer i acceptale, ecaue of the tolerance of non-real time application to the delay and the jitter. In Fig. 6, the cae of three uer tranmitting imultaneouly in the firt channel i hown. The three uer have different data rate in order to how the advantage of QoS uer on BE uer in term of error rate. The QoS uer i allocated the highet rate from Tale I (460 Mit/) and the BE uer are allocated the econd data rate (409 Mit/). A illutrated in the figure, the QoS uer outperform the BE uer although it i tranmitting at a higher data rate. In Fig. 7, we compare the performance of the three uer cheme having a ame data rate of 307 Mit/ with the performance of a ingle uer adopted y WiMedia (at 30 Mit/) and LP-OFDM (at 307 Mit/) olution. Note that for the ingle uer olution, TFC aed i exploited for the comparion ecaue it offer etter performance. For real-time uer (QoS-PL1) compared to the LP-OFDM olution, and a.5 db gain compared to the WiMedia olution. For other uer, the performance i cloe to that of the WiMedia and LP-OFDM olution. In Fig. 8, we preent the performance of the four uer cheme tranmitting imultaneouly in the firt channel at the ame data rate of 307 Mit/; two QoS uer with two different priority level and two BE uer with the ame priority level. In the ame channel, the et u-and are guaranteed for the two QoS uer repectively and the lat u-and i hared etween the two BE uer (a illutrated in Fig. 5). Hence, note that the rate of the BE uer i divided y 4 = 10, the cro-layer cheme offer a db gain for the a factor of two, and actually equal to 307 / = Mit/. A illutrated in the figure, the QoS contraint i repected and the QoS uer outperform the BE uer.

6 100% 90% 80% 70% 60% 50% 40% 30% 0% 10% 0% 10-1 Uer 1 (QoS-PL1) Uer (QoS-PL) Uer 3 (BE-PL3) Uer 4 (BE-PL3) Su-and 1 Su-and Su-and 3 Fig. 5. Su-and allocation for four uer QoS-PL1-307 Mit/ QoS-PL- 307 Mit/ BE-PL Mit/ BE-PL Mit/ E/N0 (db) Fig. 8. Performance in a four uer configuration QoS-PL1-460Mit/ BE-PL- 409Mit/ BE-PL3-409 Mit/ ACKNOWLEDGMENT The reearch leading to thee reult ha received funding from the European Community' Seventh Framework Programme FP7/ under grant agreement n alo referred a OMEGA E/N0 (db) Fig. 6. Performance of three uer tranmitting imultaneouly with different data rate QoS-PL1-307 Mit/ BE-PL- 307 Mit/ BE-PL3-307 Mit/ ingle uer LP-OFDM with TFC- 307Mit/ ingle uer WiMedia with TFC- 30Mit/ E/N0 (db) Fig. 7. Performance of the cro-layer cheme compared to LP-OFDM and WiMedia olution uing TFC V. CONCLUSION In thi paper, we propoed a new cro-layer deign for UWB ytem. Thi deign comine the phyical layer information (CSI) and the MAC layer parameter (priority level and cheduling deciion). Thi comination olve the prolem of the cohaitation of more than three uer in one channel defined in the WiMedia olution and it manage the uer requirement and contraint, o that QoS uer have advantage on BE uer in term of error rate and quantity of reource. We howed that the propoed cheme improve the ytem performance without increaing it complexity, and it outperform the WiMedia and LP-OFDM olution for multimedia uer. REFERENCES [1] Firt report and order, reviion of part 15 of the commiion rule regarding ultra-wideand tranmiion ytem, FCC, ET Docket , Fe. 14, 00. [] A. Batra et al., Multi- OFDM phyical layer propoal for IEEE tak group 3a, IEEE document P /0493r1, Texa Intrument et al., Sept [3] WiMedia Alliance, Inc., Multiand OFDM phyical layer pecification, Releae 1.1, July 005. [4] Z. Chen, D. Wang and G. Ding, An OFDM-UWB Scheme with Adaptive Carrier Selection and Power Allocation, in Proc. IEEE Intern. Conference on Wirele Communication, Networking and Moile Computing (WiCOM 06), pp.1 4, China, Sept [5] W.P Siriwongpairat, Z. Han and K. J Ray Liu, Power controlled channel allocation for multi-uer multiand UWB ytem, IEEE Tran. Wirele Communication, vol. 6, no., pp , Fe [6] A. Stephan, J-Y. Baudai and J-F. Helard, Efficient Allocation Algorithm for Multicarrier Spread Spectrum cheme in UWB application, in Proc. IEEE Intern. Conference on Ultra-Wideand (ICUWB 07), pp , Singapore, Sept 007. [7] Y. Chu and A. Ganz, Adaptive MAC Protocol for QoS upport in UWB-aed Wirele Network, in Proc. IEEE Intern. Conference on Communication (ICC 06) pp , Itanul, June 006. [8] J. Cai, K.H Liu and X. Shen, Power Allocation and Scheduling for Ultra-Wideand Wirele Network, IEEE Tran. Vehicular Technology, vol. 57, no., pp , Mar 008. [9] S. Kaier and K. Fazel, A flexile pread-pectrum multi-carrier multiple-acce ytem for multi-media application, in Proc. IEEE Inter. Sympoium on Peronal, Indoor and Moile Radio communication (PIMRC 97), pp , Finland, Sept [10] A. Stephan, E. Gueguen, M. Cruiere, J-Y. Baudai and J-F. Helard, Optimization of Linear Precoded OFDM for High-Data- Rate UWB ytem, EURASIP Journal on Wirele Communication and Networking, vol. 008, Article ID 31757, 008. [11] M. Cruiere, J-Y. Baudai and J-F. Helard, Loading algorithm for adaptive SS-MC-MA ytem over wireline channel: comparion with DMT, European Tranaction on Telecommunication, vol. 17, no. 6, pp , June 006. [1] J. Foeter, Channel Modeling u-committee report (final), IEEE P /490rl-SG3a,003. [13] A. Saleh and R. Valenzuela, A tatitical model for indoor multipath propagation, IEEE J. on elected Area in Communication, vol. 5, no., pp , Fe