Cooperative Wideband Spectrum Detection Based on Maximum Likelihood Ratio for CR Enhanced VANET

Transcription

1 Joural of Commuicatios Vol. 8, No., December 3 Cooperative Widebad Spectrum Detectio Based o Maimum Lielihood Ratio for CR Ehaced VANET Xiaorog Xu,, Jiarog Bao, Yi Luo, ad Huaia Wag College of Telecommuicatio Egieerig, Hagzhou Diazi Uiversity, Hagzhou, Zhejiag 38, Chia Electrical ad Computer Egieerig Departmet, Steves Istitute of Techology, Hoboe, NJ 73, USA {ur, baojr, luoyi}@hdu.edu.c, hwag38@steves.edu Abstract Cogitive radio (CR) ehaced vehicular ad hoc etwor (CR-VANET) is a special VANET that each vehicle could eplore local spectrum gap ad reduce data cogestio by mitigatig iterferece by dyamically accessig spectrum holes. Oe of the most importat challeges for CR-VANET is to idetify the presece of primary users (PU) over widebad spectrum at a particular time ad specific geographic locatio. I this paper, maimum lielihood ratio (MLR) detectio is proposed as a effective scheme for widebad spectrum detectio i orthogoal frequecy divisio multipleig (OFDM) based overlay CR-VANET. OR fusio rule with square-law selectio (SLS) is implemeted at roadside to mae global detectio decisio by meas of the combiatio of each vehicle s local sesig results. Hece, cooperative widebad spectrum detectio is implemeted i dyamic topology CRVANET to ehace the reliability of spectrum detectio. Specifically, robust cooperative sesig techique is ivestigated with the cosideratio of sesig chael SNR, reportig chael fadig, as well as iter-user error rate, with the purpose of eploitig diversity i multi-vehicle cooperative sesig sceario. Limitatios of cooperative widebad spectrum sesig are preseted with MLR detectio i space-time bloc code (STBC) ad time divisio multiple access (TDMA) protocols for practical cosideratio. Simulatio results are preseted to cofirm our theoretical derivatios. It is revealed that, compared with covetioal eergy detectio (ED), cooperative widebad spectrum detectio with MLR has adaptive detectio thresholds. Moreover, robust cooperative sesig by roadside SLS OR rule ad its limitatios are also performed with receiver operatio characteristics (ROC). It is prove that, MLR detectio is available i robust cooperative sesig with dyamic threshold variatios for CR-VANET practical applicatios. Ide Terms CR ehaced VANET (CR-VANET), maimum lielihood ratio (MLR), square-law selectio (SLS), cooperative widebad spectrum detectio, Iter-user error rate, receiver operatio characteristics (ROC) I. INTRODUCTION Cogitive radio (CR) ehaced vehicular commuicatios sigificatly reduce data cogestio by mitigatig iterferece ad dyamically accessig to Mauscript received July 9, 3; revised October, 3. This wor was supported i part by Natioal Natural Sciece Foudatio of Chia (Grat No. 666, 633). Prelimiary result of this paper was preseted i part at the Proceedigs of IET Iteratioal Coferece o Automatic Cotrol ad Artificial Itelligece (ACAI ). Correspodig author ur@hdu.edu.c. doi:.7/jcm Egieerig ad Techology Publishig 84 spectrum holes, which eables vehicles to opportuistically use the idle licesed spectrum i primary system, i order to improve the throughput ad combat the iterferece amog eighborhood vehicles. The dedicated short rage commuicatios (DSRC) bad has bee allocated i the USA at 5.9GHz for vehicle-toroadside (VR) ad vehicle-to-vehicle (VV) commuicatios, while IEEE 8.p worig group has developed the ehacemets of 8. stadards to support itelliget trasportatio systems (ITS), vehicular commuicatios ad iformatio applicatios []. Differet from covetioal poit-to-poit ad poit-tomultipoit commuicatio paradigm [], [3], vehicular commuicatio is characterized by highly dyamic topology, predictable movemet, ad quicly varyig etwor scale. For vehicular commuicatio applicatios, respose time ad reliability are the major characteristics requiremets [], [3]. I additio, DSRC spectrum is evisioed to become icreasig cogested, especially whe the desity of vehicles becomes higher. Furthermore, i the etreme case, such as traffic jams or emergecy icideces, the DSRC radio would be oversaturated, which results i the deterioratio of commuicatio quality-of-service (QoS) i vehicular ad hoc etwor (VANET). Cogitive radio (CR) has bee proposed to efficietly eploit the overall spectrum by allowig secodary users (SUs) to opportuistically access the dedicated spectrum that has bee assiged to primary users (PUs). SUs are allowed to trasmit ad receive data over portios of spectrum whe PUs are iactive [4], [5]. For VV/VR commuicatios, dyamic spectrum sharig betwee SU vehicles ad the licesed PUs ca potetially improve commuicatio efficiecy as well as spectrum utilizatio. Cogitive radio ehaced vehicle commuicatios allows vehicles to opportuistically use the idle licesed spectrum i the primary systems (i.e., TV bads) to improve the throughput ad mitigate the iterferece amog the eighborhood vehicles. CR ehaced VANET (CR-VANET) is a special VANET that each vehicle could eplore local spectrum gap ad a cetral cotroller (roadside) coordiates the spectrum for cogitive vehicles to realize dyamic spectrum access (DSA), which is suitable to operate i CR-VANET sceario. To achieve the ivisibility requiremet, SU vehicles eeds to sese primary spectrum holes via cooperatio approach. This

2 Joural of Commuicatios Vol. 8, No., December 3 ivolves some sort of spectral aalysis ad sesig coordiatio [6], [7]. I geeral, whe the ature of PU sigal (TV bad) is uavailable to SU vehicle, spectral aalysis maily relies o eergy detectio (ED), with higher requiremets for sesitivity ad frequecy resolutio [5], [8]. Differet from covetioal ED local detectio, this paper implemet maimum lielihood ratio detectio (MLR) approach at SU vehicle. MLR belogs to the geeralized lielihood ratio test (GLRT) paradigm [9]. It has some uique features, such as adaptive detectio threshold that ca be chaged with the received sigal-tooise ratio (SNR). I additio, due to each SU vehicle ad roadside are distributed i the large geographic area, sesig (from PU TV bad to SU vehicles) ad reportig (from SU vehicles to roadside) chaels fadig should be cosidered simultaeously (usually those chaels fadig are time-frequecy selective (double selective)) [7]. Moreover, with fast ad agile sesig ability, SU vehicles must detect the spectrum holes opportuistically to improve the spectrum occupacy utilizatio [7]. However, oce PU returs to access the licesed bad, SU vehicles must immediately stop operatig at the licesed bad ad shift to other available spectrum holes adaptively, which is deoted as CR overlay spectrum sharig paradigm [9, ]. This fast switchig ca guaratee the miimum iterferece to PU ormal commuicatio. Hece, robust cooperative widebad sesig with MLR detectio is ivestigated to cope with the problem of SU vehicle discotiuous data trasmissio. Meawhile, this method also maitais QoS of PU commuicatio, which serves as the motivatio of this paper. The mai cotributio of this paper is described as follows. Firstly, we ivestigate MLR scheme that is implemeted by each vehicle for local spectrum detectio. OR fusio rule with square-law selectio (SLS) diversity scheme is applied at CR-VANET roadside. The, robust cooperative sesig strategy is ivestigated with the cosideratio of sesig/reportig chaels fadig ad iter-user error rates. Global detectio probability ad the correspodig ROC performaces are derived i cooperative widebad spectrum detectio sceario. Fially, we preset umerical results ad aalyze ROC performace with differet sesig/reportig SNR ad iter-user error rates to cofirm our theoretical derivatios. The remider of this paper is orgaized as follows. I Sectio II, we preset system model of CR-VANET. Vehicle MLR spectrum detectio ad roadside OR fusio process with SLS are give i Sectio III. I this sectio, robust cooperative sesig with MLR detectio is also derived, which is compared with traditioal ED detectio algorithm for differet sesig/reportig SNR ad itervehicle chael error rates. Sectio IV gives umerical results ad aalysis of MLR global detectio performace, BER performace with reportig chael SNR, as well as ROC performace of robust cooperative detectio. Fially, we provide some cocludig remars i Sectio V. II. SYETEM MODEL OF CR ENHANCED VANET This sectio describes system model of CR-VANET cooperative widebad spectrum detectio, which taes the sesig/reportig chael fadig ad iter-user error rates ito cosideratios. PUT PUT R g Roadside SUT SUD SUT SUT Roadside SUT SUT SUT (a) h, h, N Rg (b) SUD g, g, h, SUD SUD SUD PUD CR ehaced VANET g, N PUD SUD CR ehaced VANET Fig.. System model of CR ehaced VANET. (a) Cooperative widebad spectrum sesig based o local MLR detectio. (b) Roadside dissemiates decisio results ad SU data trasmissio. As depicted i Fig. (a)-(b), we cosider CR ehaced VANET with overlay commuicatio paradigm, that is, multiple secodary vehicles access the licesed spectrum durig PU trasceivers are iactive. I the primary licesed TV bad, PU trasmitter (PUT) seds data to a PU destiatio (PUD) over the licesed ultra high frequecy (UHF) bad, ofte rages at 4-8MHz. The licesed PUs i UHF bads usually are DTV users [6], [7]. Meawhile, CR-VANET cosists of cogitive vehicles that are equipped with CR devices. A cogitive vehicle ca trasmit its data to its correspodig cogitive receiver vehicle opportuistically over the same UHF TV bad. Multiple CR vehicles cooperate to sese the available spectrum i order to improve the global detectio performace [6], [7], []. To be specific, the widebad cooperative sesig process ca be divided ito two phases. Firstly, all cogitive vehicles arrive withi the rage of roadside R, ad they detect the idle licesed TV bad chael via MLR spectral detectio scheme (deoted as the sesig phase). After this phase, these vehicles sed their local results to the roadside, which is idicated as the reportig g 3 Egieerig ad Techology Publishig 85

3 Joural of Commuicatios Vol. 8, No., December 3 phase. The, the cetral cotroller (roadside) implemets a special data fusio rule to mae global decisio. If the licesed chaels are available for cogitive vehicles to access, the roadside will dissemiate the decisio results i CR-VANET. Multiple cogitive vehicles will utilize these idle chaels for data trasmissio to realize overlay spectrum sharig. Otherwise, they will cotiue to detect the licesed bad ad to see available spectrum holes for opportuistic trasmissio [], []. We deote the chael betwee SU trasmitter (SUT) ad PUD as iterferig chael, while the oe betwee SUT ad its correspodig SU destiatio (SUD) as cogitive chael. We assume a total of pairs of cogitive vehicles i CR-VANET that share N subcarriers i a fied time with total badwidth B. Due to chael gais are time-variat i CR-VANET, we cosider the average iterferece ad cogitive chael gais for the -th cogitive vehicle with the -th subcarrier, which are deoted as E g, ad Eh, respectively. The phases of cooperative widebad spectrum sesig ad trasmissio are illustrated i Fig.. Time slot SU vehicles MLR detectio Roadside OR fusio with SLS diversity scheme Cooperative widebad spectrum sesig SU vehicles data trasmissio Spectrum sharig Fig.. The phases of cooperative widebad spectrum sesig ad trasmissio. III. COOPERATIVE WIDEBAND SPECTRUM DETECTION BASED ON MLR A. MLR Local Spectrum Detectio Cosider CR ehaced VANET with N subcarriers, the received sigal at the -th subcarrier for SUT ca be writte as r whs () where w (or ) deotes whether PU TV bad radio trasmits sigal s or ot, represets the additive chael oise which is comple Gaussia radom variable with zero mea ad variace. Chael fadig coefficiet h is writte as h ep( j), where the amplitude is a radom variable coforms to Rayleigh distributio, ad the phase is uiformly distributed amog the azimuth i [, ] [3]. We trasform { r, [, N ]} to { R, [, N ]} i frequecy domai by DFT, which ca be writte as R w ep( j ) S, [, N ] () where { S }, { } are the N -poit DFT of { s } ad { } respectively. Whe the time istat t, accordig to cetral limit theorem (CLT) [4], the received sigals at the -th subcarrier R, ca be regarded as comple Gaussia process R N(, ) [4], where ( w ) deotes R the variace of R, ad R s is defied as the h average SNR at the -th subcarrier. Hece, we ca R ormalize the amplitude of R as. Obviously, is the Rayleigh distributed radom variable with zero mea ad variace w. Accordig to Ref. [4], its probability desity fuctio (PDF) ad cumulative distributio fuctio (CDF) ca be writte as follows respectively fr( ) ep( ) (3) ( ) Pr{ } ( ) ep( F ) R f R u du (4) For simplicity, H (or H ) is used to represet the evet whether PU TV bad radio is actually utilizig the -th subcarrier or ot, which correspods to w (or ). Assume that the average chael gai is ow to each vehicle by chael estimatio. We ca address the MLR model as follows [4] f ( H) ep( ) f ( H ) where is the MLR threshold at the -th subcarrier. If H is true, decisio regio ca be epressed as DR { : }. The we select the upper boud of MLR threshold, that is (5) ep( ) (6) Hece, we ca easily get the decisio threshold l[ ( )]. For, is proportioal to decisio threshold, thus we get the lower boud of l( ), ad the correspodig decisio regio is DR { }. Therefore, the detectio probability, false alarm probability ad missig detectio probability at the -th subcarrier ca be writte as below respectively Pr D, f ( H) d ep( ) (7) DR Pr, ( ) ep( ) f H d (8) DR 3 Egieerig ad Techology Publishig 86

4 Joural of Commuicatios Vol. 8, No., December 3 Prm, PrD, ep( ) (9) It is apparet that a larger umber of MLR threshold results i a more accurate decisio, as epressed i Eq. (5). However, as see from Eq. (7) to Eq. (9), a large umber of leads to large umber of ad small value of Pr D,. Hece, it is essetial to mae a tradeoff betwee detectio probability ad sesig accuracy. The upper boud of detectio threshold with the relatioship of missig probability threshold Pr ca be writte as ( )l( Pr ) () Substitute i Eq. () ito Eq. (6), it is easily to obtai MLR threshold upper boud, which ca be epressed as ( Pr ) () For is proportioal to as Eq. (6) idicates, it is obvious that ( )l( ) ( )l[ u ( )] () where the upper boud of MLR threshold is show i Eq. (), that is, ( Pr ) u. It is apparet that MLR threshold is related with the missig probability threshold Pr ad SU vehicle average received SNR at the -th subcarrier. I geeral, { S } i Eq. () is uow because PU TV radio sigal is usually ucertai, which iduces that { } is hardly to obtai. Hece, we utilize maimum lielihood (ML) algorithm to estimate [3, 4], that is ˆ arg ma f (, H ) (3) Substitute Eq. (3) ito Eq. (5), it yields to the fial decisio criterio for MLR, which is show as f (4) f H ( H) ep( ) ( ) where [, u ]. From Eq. (6) to Eq. (9), it is ow that oce MLR threshold is selected, the decisio threshold at the -th subcarrier ca also be determied. If the average received SNR is give, the the detectio probability Pr D,, false alarm probability Pr, as well as missig detectio probability Pr miss, ca be obtaied simultaeously. B. Roadside OR Fusio Process with SLS Each SU vehicle seds local detectio result to roadside that performs global decisio judgmet by OR fusio rule with square-law selectio (SLS). I SLS diversity scheme, suppose that each vehicle s sesig/reportig chaels are idepedet ad idetical distributed (i.i.d), chael with the maimum output decisio statistic from the square-law devices (squaread-itegrate operatio per chael) should be selected at roadside [8]. Specifically, assume { y } deotes the received sigal eergy from square-law device at each vehicle, the the statistic epressio ysls ma{ y, y, y } is to be selected at roadside. SLS diversity scheme is also deoted as OR rule, which meas roadside ifers the presece of PU sigal whe there eists the maimum decisio statistic from cogitive vehicles. Accordig to Ref. [, 5], OR rule is very coservative for CR vehicles to access the licesed PU bad, which guaratees the miimizatio of iterferece to PU. For may cases of practical iterest, OR rule obtais better performace tha other fusio rules. I this paper, cogitive OFDM-based overlay sceario is implemeted i CR ehaced VANET, which is proposed as a promisig approach for ehacig spectral efficiecy [9, 4, 6]. For cogitive OFDM-based CR- VANET, a few subcarriers are selected to report the istataeous SU vehicle local decisio results. To avoid the iterferece geerated by other SU vehicles, each subcarrier is eclusively assiged to oe SU vehicle for opportuistic utilizatio [6, 7]. Due to subcarriers are orthogoal i OFDM, differet SU vehicles are oly allowed to trasmit their local MLR detectio results via orthogoal subcarriers. Hece, total subcarriers N is equivalet to the umbers of SU vehicles i reportig phase [, 5]. Suppose the local decisio of the -th SU vehicle is give by H : y d,,,, H : y (5) where y is the received sigal eergy at the -th SU vehicle, ad idicates the decisio threshold with specific detectio scheme (i.e., ED or MLR). Each SU vehicle seds local decisio result d to roadside for fusio, ad roadside performs the global decisio with specific fusio rule. Assume OR fusio rule with SLS at roadside, ad each SU vehicle performs local MLR detectio, the the global false alarm probability Q ca be evaluated usig the CDF of hypothesis testig H [8,9], yieldig SLS y give by biary Q F ( H ) Pr (6),SLS y,sls, 3 Egieerig ad Techology Publishig 87

5 Joural of Commuicatios Vol. 8, No., December 3 Similarly, the global missig probability with SLS ca be obtaied as Q Pr (7) miss,sls m, where Pr, ad Pr m, are give i Eq. (8) ad Eq. (9) respectively. C. Robust Cooperative Sesig Based o MLR The local detectio results will also eperiece fadig ad shadowig durig the reportig process, which deteriorates the trasmissio reliability ad affects the global detectio result ievitably [, 8]. A cross-layer approach is proposed to reduce the ureliable reportig by aomaly spectrum usage detectio i multihop CR etwors [7]. Accordig to the Ref. [, 8], MLR sesig results are reported from SU vehicles to roadside through the dedicated reportig chaels i cooperative widebad sesig phase. Hece, the global false alarm probability Q ad missig detectio probability Q miss epressed i Eq. (6) ad Eq. (7) are rewritte as follows. Q [( Pr )( Pr ) Pr Pr ] (8), e,, e, Q [Pr ( Pr ) ( Pr ) Pr ] (9) miss m, e, m, e, where Pr e, deotes the reportig error probability durig the reportig trasmissio of local MLR decisio [,8]. Suppose each SU vehicle has idetical local sesig result ad the local MLR decisio eperieces i.i.d fadig report chaels, amely, Pre, Pre, for,,,. As a result, the global false alarm probability has lower boud, which ca be writte as Q Q ( Pr ) Pr. e e For practical cosideratio, all the local MLR detectio results from SU vehicles to roadside are assumed i accordace with TDMA protocol, hece, roadside gathers the orthogoal decisios without mutual iterferece. The receptio performace of multiple SU vehicles i TDMA is the same as that of oe SU. Oe SU biary hypothesis testig problem is just the same as biary phase shift eyig (BPS) modulatio durig reportig trasmissio. Cosider the correspodig symbol error rate (SER) of BPS modulatio over Rayleigh fadig chael [, 3] show as Pre,TDMA () where deotes the average received SNR. O the other had, SU vehicles ca cooperate with each other to form a virtual atea array, thus cooperative diversity ca be achieved i this way []- [], [8]-[9]. Space-time (ST) coded cooperative sesig is implemeted to realize trasmit diversity [9], []. However, the virtual atea array formed by vehicle cooperatio is differet from a real trasmit atea array formed by multiple ateas at a trasmitter. This is because iter-user chael fadig should be cosidered i a virtual atea array [8]. I distributed STBC, cooperative SU vehicles echage local MLR detectio iformatio, which is performed via sedig request-to-sed (RTS) ad clear-to-sed (CTS) frames betwee SU vehicles ad the roadside. I two vehicles cooperatio sceario, if both SU vehicles correctly decode the sigals trasmitted from the other side, the STBC ca be implemeted. Otherwise, SU vehicles will report their ow decisios to the roadside by TDMA protocol. Hece, MLR detectio decisios are reported to the roadside by either direct trasmissio usig TDMA protocol or trasmit diversity via STBC, based o the quality of iter-vehicle chaels. Ref. [], [8] give the SER epressio of BPS modulatio usig STBC over Rayleigh flat fadig reportig chael, which is deoted as Pre,STBC 4 () where is the umber of cooperated SU vehicles, ad. Let represet the possible error rate occurred i iter-vehicle chael, the ( ) deotes the probability of the cooperated vehicles both correctly decode the received sigals that come from each other. Hece, the reportig error rate is epressed as Pr Pr ( )Pr () e e,stbc e,tdma The proposed vehicle cooperatio is combied TDMA protocol with STBC trasmit diversity. For good itervehicle chael, approaches to zero, ad Pr Pr, e e,stbc which idicates the two cooperative vehicles could always correctly decode the received sigals. The trasmit diversity gai could be obtaied i reportig chael. Apply Eq. () ito Eq. (8) ad Eq. (9), we ca obtai the global false alarm probability ad missig detectio probability of robust cooperative widebad spectrum sesig with MLR local detectio. I multiple SU vehicles sceario, collaborative clusters are formed uder the cotrol of roadside, which meas some closely located SU vehicles cooperate with each other to form a cluster [6], [8]-[9]. STBC could be implemeted i those clusters. For differet clusters, TDMA could also be implemeted i the process of reportig. That is to say, each vehicle cluster will be assiged a time slot to report the decisios, ad roadside will receive the sigal from oe cluster oly i a give time slot without causig iterferece to other clusters [], [8]-[9]. 3 Egieerig ad Techology Publishig 88

6 Detectio probability BER Joural of Commuicatios Vol. 8, No., December 3 IV. NUMERICAL RESULTS AND ANALYSIS I this sectio, we evaluate cooperative widebad spectrum sesig performace based o the proposed MLR local detectio i CR-VANET. BER performace with reportig chael SNR ad the correspodig ROC performace with differet sesig/reportig SNR ad iter-vehicle error rates are preseted respectively. Numerical results are implemeted with SU cadidate vehicles. Refer to Ref. [6,8,,4,8], parameter settig are give as below: The umber of SU vehicles: / / 4. The time-badwidth product for ED is m 5, ad ED thresholds are set to be equivalet to MLR thresholds. Amplitude of received sigal for the -th subcarrier: / / 3. Accordig to Eq. (4), MLR decisio 3 8 thresholds: / e / e. 4 9 Cooperative widebad spectrum sesig combied with TDMA protocol ad STBC trasmit diversity, ad iter-vehicle error rate:.3/.3. Sesig chael SNR / db, reportig chael SNR 5 / 3dB, reportig chael error 3 4 probability Pr /. e Roadside OR rule with SLS diversity scheme is implemeted. detectio threshold variatios for SU vehicle local decisio i CR-VANET. Fig. 4 shows reportig chael BER performace with differet iter-vehicle chael qualities i two/four cooperative vehicles sceario. The case of from to represets STBC ad TDMA respectively. However, it is apparet that the smaller value of iter-vehicle chael error rate leads to better BER performace at the same umber of cooperative vehicle sceario. From Eq. (), for a very poor iter-vehicle chael whe.3, it is still obtai 3dB codig gai, which is better tha TDMA ( ) []. O the other had, with the same itervehicle chael error rate, the icreasig of collaborative vehicle umbers could improve BER performace, especially for smaller values. For the case of.3, two ad four cooperative vehicles just have the same BER performace i reportig chaels. Hece, reportig chael BER performace could be improved with the decreasig of iter-vehicle chael error rate ad the icreasig of cooperative vehicle umbers ED threshold=e 8 / SNR (db) MLR threshold= MLR threshold=e 3 /4 MLR threshold=e 8 /9 ED threshold= ED threshold=e 3 /4 Fig. 3. Local detectio performace compariso betwee MLR ad ED. Fig. 3 idicates SU local detectio performace compariso betwee MLR ad ED with the same decisio threshold. As show i Eq. (4), the threshold i MLR is the epoetial fuctio of the squared amplitude of received sigals. I the case of smaller threshold, ED has better detectio performace. However, whe the decisio threshold icreases sharply, ED would degrade detectio performace obviously. Due to the decisio threshold of MLR is associated with the received SNR of SU vehicles (sesig chael SNR), hece, the rapid fluctuatio of threshold iflueces the detectio performace slightly for MLR local detectio. Therefore, MLR is fit for adaptive spectrum sesig with dyamic -5 Two cooperative vehicles,stbc Two cooperative vehicles,=. Two cooperative vehicles,=.3-6 Two cooperative vehicles,tdma Four cooperative vehicles,stbc Four cooperative vehicles,=. Four cooperative vehicles,=.3 Four cooperative vehicles,tdma Reportig chael SNR (db) Fig. 4. Reportig chael BER performace with differet iter-vehicle chael qualities. Fig. 5 illustrates ROC of cooperative sesig with differet iter-vehicle chael error rate ad reportig chael SNR based o local MLR detectio. Give sesig chael SNR db ad reportig chael SNR 5 / 3dB. It is obvious that, the icreasig of reportig chael SNR ad the reductio of iter-vehicle chael error rate lead to better ROC performace for two SU vehicles cooperatio situatio. I additio, reportig chael SNR 3dB with TDMA performs eve better tha reportig chael SNR 5dB with STBC, which meas improvig chael SNR is more efficiet tha the reductio of iter-chael error rate for the global ehacemet of ROC performace. For reportig chael SNR 3dB, curve with.3 approimately approaches to the curve with STBC, which reflects that iter-vehicle chael error rate cotributes little to the improvemet of cooperative widebad spectrum 3 Egieerig ad Techology Publishig 89

7 Joural of Commuicatios Vol. 8, No., December 3 detectio performace at higher reportig chael SNR regio. boud is., if Pr e is 4, the boud turs to be.. However, MLR has o Q boud. It idicates that, if sesig chael SNR is high eough while reportig error is very low, Q of cooperative sesig based o MLR could approach to lower value tha that of cooperative detectio based o ED. Therefore, MLR achieves better ROC performace tha ED i this situatio. I additio, cooperative widebad spectrum detectio based o MLR has adaptive detectio threshold that is related with average sesig chael SNR. This feature eables it to be suitable i dyamic topology CR- VANET. Fig. 5. ROC of cooperative sesig with differet iter-vehicle chael error rate ad reportig chael SNR. Missig detectio probability, Q miss ED,sesig SNR db,reportig error -3 ED,sesig SNR db,reportig error -3 ED,sesig SNR db,reportig error -4 ED,sesig SNR db,reportig error -4 MLR,sesig SNR db,reportig error -3 MLR,sesig SNR db,reportig error -3 MLR,sesig SNR db,reportig error -4 MLR,sesig SNR db,reportig error lse alarm probability, Q Fig. 6. Compariso of cooperative widebad detectio ROC performace betwee MLR ad ED with differet sesig chael SNR ad reportig chael error rates. Compariso of cooperative widebad detectio ROC performace betwee MLR ad ED with differet sesig chael SNR ad reportig chael error rates is show i Fig. 6. It is maily give by umerical results implemeted i Eq. (8) ad Eq. (9) respectively. For two cooperative SU vehicles, both implemet MLR or ED for local detectio. The sesig chael is eperieced Rayleigh flat fadig with chael SNR db ad db respectively, ad the reportig chael 3 error probability Pr e is assumed to be ad 4 respectively. From this figure, we fid that ED performs better tha MLR i smaller decisio threshold regio, which is cosistet with Fig. 3. However, Q with ED has its limitatio, which teds to a lower boud that is related with the umber of cooperated vehicles ad 3 reportig error rate Pr e. i.e., whe Pr e is, the lower V. CONCLUSIONS I this paper, we maily propose adaptive threshold local MLR detectio as a effective scheme for cooperative widebad spectrum detectio i CR ehaced VANET. OR fusio process with SLS is implemeted at the roadside for local decisio gatherig. With the cosideratio of reportig chael fadig ad iter-vehicle chael error rates, robust cooperative widebad spectrum detectio is tae ito cosideratio i practical dyamic topology CR-VANET scearios. Simulatio results idicate that, compared with traditioal ED, widebad spectrum detectio with MLR has adaptive detectio threshold i accordace with sesig chael SNR. Reportig chael BER performace as well as ROC performace of two SU vehicles cooperatio are all preseted to cofirm our derivatio. Due to dyamic threshold variatios, robust cooperative widebad spectrum sesig strategy based o local MLR detectio is feasible to CR-VANET practical applicatios. ACNOWLEDGMENT The authors would lie to greatly appreciate aoymous reviewers for their valuable commets ad costructive suggestios i helpig to improve the quality of this paper. REFERENCES [] D. Jiag ad L. Delgrossi, IEEE 8.p: Towards a iteratioal stadard for wireless access i vehicular eviromets, i Proc. IEEE 67th Vehicular Techology Coferece, May 8, pp [] H. Hartestei ad. P. Laberteau, A tutorial survey o vehicular Ad Hoc etwors, IEEE Commuicatios Magazie, vol. 46, o. 6, pp. 64-7, Jue 8. [3] E. Schoch, F. argl, M. Weber, ad T. Leimuller, Commuicatio patters i VANETs, IEEE Commuicatios Magazie, vol. 46, o., pp. 9-5, November 8. [4] J. Mitola ad G. Q. Maguire, Cogitive radio: Maig software radios more persoal, IEEE Persoal Commuicatios, vol. 6, o. 4, pp. 3-8, August 999. [5] I. F. Ayildiz, W. Y. Lee, M. C. Vura, ad S. Mohaty, Net geeratio/dyamic spectrum access/cogitive radio wireless etwors: A survey, Computer Networs, vol. 5, o. 3, pp. 7-59, September 6. 3 Egieerig ad Techology Publishig 8

8 Joural of Commuicatios Vol. 8, No., December 3 [6] X. Y. Wag ad P. H. Ho, A ovel sesig coordiatio framewor for CR-VANETs, IEEE Trasactios o Vehicular Techology, vol. 59, o. 4, pp , April. [7] H. S. Li ad D.. Iric, Collaborative spectrum sesig i cogitive radio vehicular Ad Hoc etwors: Belief propagatio o highway, i Proc. IEEE 7st Vehicular Techology Coferece, May, pp. -5. [8] F. F. Digham, M. S. Alouii, ad M.. Simo, O the eergy detectio of uow sigals over fadig chaels, IEEE Trasactios o Commuicatios, vol. 55, o., pp. -4, Jauary 7. [9] R. Zhag, T. J. Lim, Y. C. Liag, ad Y. Zeg, Multi-atea based spectrum sesig for cogitive radios: A GLRT approach, IEEE Trasactios o Commuicatios, vol. 58, o., pp , Jauary. []. B. Letaief ad W. Zhag, Cooperative commuicatios for cogitive radio etwors, i Proc. IEEE, vol. 97, May 9, o. 5, pp [] Y. L. Zou, J. Zhu, B. Y. Zheg, ad Y. D. Yao, A adaptive cooperatio diversity scheme with best relay selectio i cogitive radio etwors, IEEE Trasactios o Sigal Processig, vol. 58, o.. pp , October. [] Y. L. Zou, Y. D. Yao, ad B. Y. Zheg, Cooperative relay techiques for cogitive radio systems: Spectrum sesig ad secodary user trasmissios, IEEE Commuicatios Magazie, vol. 5, o. 4, pp. 98-3, April. [3] J. G. Proais, Digital Commuicatios, 4 th ed. Beijig: Publishig House of Electroics Idustry, 4, pp , 66-8, [4] T. Luo, T. Jiag, W. Xiag, ad H. H. Che, A subcarrier allocatio scheme for cogitive radio systems based o multicarrier modulatio, IEEE Trasactios o Wireless Commuicatios, vol. 7, o. 9, pp , September 8. [5] A. Ghasemi ad E. S. Sousa, Opportuistic spectrum access i fadig chaels through collaborative sesig, Joural of Commuicatios, vol., o., pp. 7-8, February 7. [6] E. Yaacoub ad Z. Dawy, A survey o upli resource allocatio i OFDMA wireless etwors, IEEE Commuicatios Surveys & Tutorials, vol. 4, o., pp , Secod Quarter. [7] L. Qia, X. Li, ad S. Wei, Aomaly spectrum usage detectio i multihop cogitive radio etwors: A cross-layer approach, Joural of Commuicatios, vol. 8, o. 4, pp , April 3. [8] W. Zhag ad. B. Letaief. Cooperative spectrum sesig with trasmit ad relay diversity i cogitive radio etwors, IEEE Trasactios o Wireless Commuicatios, vol. 7, o., pp , December 8. [9] W. Y. Lee ad I. F. Ayildiz, Optimal spectrum sesig framewor for cogitive radio etwors, IEEE Trasactios o Wireless Commuicatios, vol. 7, o., pp , October 8. [] I. F. Ayildiz, B. F. Lo, ad R. Balarisha, Cooperative spectrum sesig i cogitive radio etwors: A survey, Physical Commuicatios, vol. 4, pp. 4-6,. Xiaorog Xu is with the College of Telecommuicatio Egieerig, Hagzhou Diazi Uiversity (HDU), Hagzhou, Chia. He was a research scholar with the Electrical ad Computer Egieerig Departmet, Steves Istitute of Techology (SIT), Hoboe, NJ, USA, durig 3-4. He received the B. Eg. degree i Commuicatio Egieerig ad M. Eg. degree i Commuicatio ad Iformatio System from HDU, Hagzhou, Chia, i 4 ad 7, respectively. He received Ph.D. degree major i Sigal ad Iformatio Processig from Najig Uiversity of Posts ad Telecommuicatios (NJUPT), Najig, Chia, i. Previously, from to 3, he was worig as a postdoctoral researcher i the Istitute of Iformatio ad Commuicatio Egieerig, Zhejiag Uiversity (ZJU), Hagzhou, Chia. Curretly, he is worig as a college teacher i HDU. Dr. Xu s research iterests emphasize o Cogitive Radio Networs, Cooperative Commuicatios, CR ehaced VANET, CR based WSN, as well as Compressed Sesig theory. Jiarog Bao is with the College of Telecommuicatio Egieerig, Hagzhou Diazi Uiversity (HDU), Hagzhou, Chia, as associate professor. He received the B. Eg. degree i Polymer Material Egieerig ad M. Eg. degree i Commuicatio ad Iformatio System from Zhejiag Uiversity of Techology, Hagzhou, Chia, i ad 3, respectively. He received Ph.D. degree major i Iformatio ad Commuicatio Egieerig from Tsighua Uiversity, Beijig, Chia, i 9. Previously, from to 3, he was worig as a postdoctoral researcher i the Istitute of Iformatio ad Commuicatio Egieerig, Zhejiag Uiversity (ZJU), Hagzhou, Chia. Curretly, he is worig as a college teacher i HDU. Dr. Bao s research iterests emphasize o Cogitive Radio Networs, LDPC codes i aerospace commuicatios. Yi Luo is with the College of Telecommuicatio Egieerig, Hagzhou Diazi Uiversity (HDU), Hagzhou, Chia, as associate professor. He received the B. Eg. degree i Commuicatio Egieerig ad M. Eg. degree i Commuicatio ad Iformatio System from HDU, Hagzhou, Chia, i ad 4, respectively. Curretly, he is worig as a college teacher i HDU. Prof. Luo s research iterests emphasize o Cogitive Radio Networs, Embedded System Desig. Huaia Wag is with the Electrical ad Computer Egieerig Departmet, Steves Istitute of Techology (SIT), Hoboe, NJ, USA. He received the B. Eg. degree i Iformatio Egieerig from Southeast Uiversity (SEU), Najig, Chia, i. He is curretly worig toward his Master ad Ph. D. degrees at the Electrical ad Computer Egieerig Departmet of SIT. His research iterests emphasize o Cogitive Radio Networs, Optimizatio i Commuicatios. 3 Egieerig ad Techology Publishig 8