RSU-based Real-time Data Access in Dynamic Vehicular Networks

Transcription

1 th Internatonal IEEE Annual onference on Intellgent Transportaton Systems Madera Island, Portugal, September 9-, TB. RSU-based Real-tme Data Access n Dynamc Vehcular Networks Ka Lu kevn.lu@student.ctyu.edu.hk Vctor.S. Lee csvlee@ctyu.edu.hk Dept. of omputer Scence, ty Unversty of Hong Kong Abstract Recently, data dssemnaton n vehcular networks has receved consderable attenton. In partcular, there has been sgnfcant nterest n accessng nformaton stored at RSUs Roadsde Unts). Although data access has been extensvely studed n the network communty, unque characterstcs n vehcular networks brng us new challenges. Frst, the RSU provdes both safety-crtcal and non-safety-crtcal servces, whch should be dfferentated. Second, due to the lmted communcaton range covered by RSUs, nformaton must be provded n tme to be accessed by passng vehcles. Last but not least, n a typcal vehcular network, the traffc characterstcs such as vehcle arrval rate, dwell tme and data access pattern change from tme to tme, whch requre specal efforts n nformaton delvery to gan satsfactory performance. Ths paper frst proposes a data dssemnaton framework supported by the RSU, n whch an adaptable data dssemnaton method s presented. Next, we derve an analytcal model to form the bass of performance evaluaton. Last, we present a seres of evaluaton results to examne the system performance n dynamc traffc envronments. The results not only demonstrate the effectveness of the adaptable data dssemnaton mechansm, but also explore novel features of data dssemnaton n dynamc vehcular networks, whch gve nsghts nto new research problems. I. INTRODUTION Nowadays, the ITS Intellgent Transport Systems) are expected to offer fundamental breakthroughs n road safety, congeston reducton, drvng comfort and envronmental mprovement. In general, there are two types of communcaton: IV Inter-Vehcle ommuncaton) and RV Roadsde-to- Vehcle ommuncaton) [], []. The system contans two basc elements: OBUs On Board Unts) and RSUs Roadsde Unts). OBUs are mounted on vehcles. The connecton among OBUs forms an ad-hoc vehcular network, n whch vehcles are able to communcate wth each other. RSUs are mounted at ntersectons or along the road, actng as a server, to provde nformaton to passng vehcles. In ths study, we focus on the roadsde-to-vehcle communcaton wthn the coverage of the RSU. Note that vehcles are also able to forward or exchange nformaton outsde the servce regon of the RSU, whch are the ssues addressed n the IV model. In ndustry, numerous projects worldwde are actvely engaged n dong researches and developng nfrastructures for vehcular communcaton applcatons, such as VIS o-operatve Vehcle-Infrastructure Systems) [] and SAFESPOT Smart Vehcles on Smart Roads) [] n Europe; AHS Advanced ruse Assst Hghway System) [] and VIS Vehcle Informaton and ommuncaton System) [6] n Japan. There are also many efforts from the government n supportng and encouragng the ITS project, such as DOT Department of Transportaton) [7] n U.S. In academa, the ssue of effectve communcaton n dynamc vehcular networks has receved great attenton. From the perspectve of communcaton requrements, they are dvded nto three categores [8]: data traffc characterstcs, communcaton functonalty requrements and communcaton qualty requrements. These three categores of requrements are typcally bult upon dfferent stack layers referrng to the OSI protocol model. Although there have been extensve studes on nformaton access n vehcular networks [9], [], [], [] and [], they are lmted to the ssues of communcaton functonalty and qualty requrements, whch prmarly are the problems resded on the network layer and data lnk layer. Lttle work, however, has put effort on the problems of data traffc characterstcs on the applcaton layer. Ths study s commtted to the evaluaton of system performance n terms of nformaton access on the applcaton layer. The man contrbuton of ths paper s to present a RSU-based data dssemnaton framework to address arsng challenges n vehcular networks. In partcular, an analytcal model s derved for evaluatng real-tme data access n vehcular networks. Dynamc traffc factors are comprehensvely nvestgated n performance evaluaton. The results demonstrate that data dssemnaton n a typcal vehcular network should be adaptable to dynamc traffc envronments. The rest of ths paper s organzed as follows. Secton II descrbes the system archtecture. Secton III formulates the research problem. Secton IV derves an analytcal model and Secton V presents the performance evaluaton results and analyss. Fnally, we draw the conclusons n Secton VI. II. SYSTEM FRAMEWORK DSR Dedcated Short Range ommuncaton) s consdered to be a key wreless communcaton standard n vehcular networks. Nowadays, there are two promnent DSR versons worldwde []. One s the Euro DSR standard, whch s developed from tradtonal RFID Rado Frequency Identfcaton) technques. The other s the Amercan standard, whch s compatble wth IEEE 8.p. In ths work, we nvestgate the roadsde-to-vehcle communcaton whch follows the Amercan DSR Standard. The spectrum ranges between.8 to.9 GHz, called ITS-RS Intellgent Transportaton Systems Rado Servce), s allocated for DSR applcatons []. Accordng to the band plan, the //$6. IEEE

2 Fg.. Data Dssemnaton Model 7 MHz spectrum s dvded nto seven channels, whch are further classfed nto one H ontrol hannel) and sx SHs Servce hannels) [6]. Wth the above knowledge, we hereby present a RSUbased data dssemnaton system as shown n Fgure. Ths s a RV model where the RSU s nstalled at the road ntersecton and passng vehcles can access nformaton from the RSU. The rado coverage of the RSU s regarded as the servce regon. Only when vehcles wthn the servce regon can they access the nformaton. In addton, the RSUs are connected wth each other by wred lnks to form a backbone network. The data tems stored n the local database of a RSU are kept up-to-date by the nformaton provders from the backbone network. Typcally, the ITS system supports two categores of servces: safety-crtcal e.g. frontal collson warnng, hazard warnng, ntersecton safety coordnaton, etc.) and non-safety-crtcal e.g. moble nfotanment, advertsements, congeston advsores, etc). Accordngly, the data tems can be categorzed nto two sets: safety-crtcal and non-safetycrtcal data tems [9], []. As specfed n [6], the control channel, H, s commtted to delverng all safety-crtcal data tems. It s vable because as observed n [8], [], compared wth other nformaton, safety-crtcal data tems are generally much smaller n sze and much less n amount. In practce, provdng such a servce only takes a short perod of tme and consumes a small fracton of bandwdth. Passng vehcles are prescrbed to montor on the H upon enterng the servce regon to retreve safety-crtcal data tems [7]. The rest sx SHs are responsble for the delvery of nonsafety-crtcal data tems. The objectve of the system s not only to provde safety-crtcal servces, but also ams at makng the best utlzaton of avalable bandwdth n provdng non-safety-crtcal servces. The detaled research problems are formulated n Secton III. III. PROBLEM FORMULATION The prmary notatons presented n the paper are summarzed n Table I. The database D s dvded nto two sets: safety-crtcal data tems D s ) and non-safety-crtcal data tems D n ), where D s = {ds,ds,...,d D s s } and D n = {dn,d n,...,d D n n }. The channel set K s dvded nto one H, denoted by K c = {k c }, and sx SHs, denoted by K s = {ks,k s,...,k K s s }, where K c = and K s = 6. There are two types of data dssemnaton approaches: push-based and on-demand broadcast [8]. In push-based broadcast, the RSU broadcasts data tems perodcally whle passng vehcles retreve data tems of nterest wthout submttng any query. In on-demand broadcast, the RSU broadcasts data tems only n response to explct queres receved from vehcles. The uploadng of queres competes for the upload bandwdth allocated by the H. We defne the uploadng capacty,, as the maxmum number of queres whch can be uploaded n a broadcast tck. In practce, ether TDM Tme Dvson Multplexng) or FDM Frequency Dvson Multplexng) can be appled to support query uploadng n H. In correspondence wth the two data dssemnaton approaches, we further classfy the non-safety-crtcal data tems n D n nto two sets: push-based data tems, D np D np = {dnp,d np,...,d np }) and on-demand data tems, D no D no = {dno,d no,...,d D no no }), where D np D n and D no D n. Each data tem dn s ether n the pushbased or the on-demand set, namely, D np Dno = D n and D np Dno = /. Moreover, the servce channels, K s, are also dvded nto two sets: push channel K sp ) and on-demand channel K so ), where K sp = {ksp,k sp,...,k K sp sp } K sp K s ) and K so = {kso,k so,...,k K so so } K s ). Each servce channel s ether n the push or the on-demand channel set, namely, K sp Kso = K s and K sp Kso = /. As dscussed n Secton II, compared wth the safetycrtcal servce, the dssemnaton of non-safety-crtcal data tems s typcally much more tme and bandwdth consumng, whch brngs more challenges. Therefore, n the followng of ths work, we concentrate on the nvestgaton of system performance n terms of dssemnatng data tems n D n. Generally, we clam that t s crtcal for the data dssemnaton servce to be adaptable to the dynamc traffc patterns. For nstance, durng rush hours, the vehcle arrval rate s hgh, whch results n a hgh data access rate. Meanwhle, the heavy traffc condton also mples longer dwell tme of vehcles because of the congeston), gven a looser deadlne for data dssemnaton. Intutvely, n ths case, the pushbased servce should take more responsblty. Ths s because frstly, the push-based servce releves the burden of query uploadng, whch would otherwse defeat the performance of on-demand servce due to the overwhelmng uploadng of queres. Secondly, although the data tems are broadcast wth fxed perod n the push-based servce, the relatve looser servce deadlne makes ths drawback not so senstve. In contrast to the heavy traffc scenaro, n lght traffc envronments, the data access rate s very low. Meanwhle, the dwell tme of vehcles wll be much shorter due to the congeston free condton), whch results n a tghter deadlne for data dssemnaton. Apparently, n such a scenaro, the effect of on-demand servce shall be more sgnfcant because

3 explct queres wll remnd the server to make a tmely response. Moreover, query uploadng wll not be a bottleneck of the on-demand servce when the data access rate s low. Wth the above observatons, to acheve satsfactory performance n dynamc vehcular networks, t s essental to strke a balance between push-based and on-demand broadcast, whch bols down to the problem of dynamc data and channel allocaton between push-based and on-demand servces. Gven an access attempt for data tem dn, no matter whether dn s classfed nto the push-based set dn D np ) or the on-demand set dn D no), as long as t s faled to be accessed before the vehcle leaves the servce regon, t s an nstance of mssng deadlne. The total number of nstances of mssng deadlne dvded by the total number of data access attempts s called deadlne mss rato. To sum up, the research problem of ths work s to nvestgate the system performance n terms of the deadlne mss rato by dynamcally allocatng the data tems nto D np and D no, and splttng the channels nto K sp and K so, n order to provde the two types of servces. IV. ANALYTIAL MODEL In ths secton, we present an analytcal model to analyze the system performance n terms of expected deadlne mss rato. Snce the accessed data tem dn s ether n the pushbased set dn D np) or n the on-demand set dn D no), the expected deadlne mss rato s analyzed n two cases. ase : The data tem s n the push-based set Denote the requred data tem as dnp ). dnp would be faled to be accessed by a vehcle f t was not broadcast durng the dwell tme of ths vehcle. Denote the broadcast perod of dnp as ld np ) and denote the dwell tme of the vehcle as t. Obvously, f t ldnp ), d np can certanly be accessed by ths vehcle. Otherwse t < ldnp )), the chance for dnp to be accessed s calculated by t ldnp )dt, whch t equals ldnp). So, the probablty that a partcular vehcle wth the dwell tme t fals to access dnp s calculated by: { t ldnp) P d np t)= t ldnp) t < ldnp) ) For all passng vehcles whch attempt to access dnp, the probablty that dnp cannot be accessed s the deadlne mss rato of dnp, denoted by R dnp. Denote the probablty densty functon of the dwell tme t as f t), accordng to equaton : R d np t)= P dnp t) f t)dt = ldnp) f t)dt + ld np) = ld np) t ld np) ) f t)dt t ld np) ) f t)dt ase : The data tem s n the on-demand set Denote the requred data tem as dno D no ) wth the access probablty of pdno ). Denote the nter arrval tme of vehcles as τ, and then, the vehcle arrval rate s ) TABLE I ASUMMARY OF NOTATIONS Notatons Descrptons Notes D the set of all data tems D s the set of safety-crtcal data tems D n the set of non-safety-crtcal D = D s Dn data tems and D s Dn = / D np the set of push-based D no non-safety-crtcal data tems the set of on-demand non-safety-crtcal data tems D np Dno = D n and D np Dno = / K the set of channels n the system K = 7 K c the set of control channels K c = K s the set of servce channels K s = 6 K sp the set of push channels K so the set of on-demand channels K sp Kso = K s and K sp Kso = / uploadng capacty maxmum number of queres can be uploaded n a tme unt ) pdn) access probablty of the th dn D n non-safety-crtcal data tem t dwell tme of vehcles exponentally dstrbuted wth parameter μ τ nter arrval tme of vehcles exponentally dstrbuted wth parameter λ τ. Accordngly, the arrval rate of query for d no s calculated by τ pd no). Denote the probablty that a query can be uploaded successfully as P success and consder that only one broadcast tck s avalable for vehcles to submt ther queres, then, as long as the total arrval rate of queres s not τ pd no) ), all queres can be uploaded successfully P success = ). Otherwse hgher than the uploadng capacty D no D no τ pd no) > ), P success = P success s calculated by: P success = τ pd no) τ τ pdno) τ <. In concluson, pdno) pdno) Note that a vehcle would stand a chance to access ts requred dno only on condton that ts query for d no s uploaded successfully. Wth ths precondton, denote O d no as the probablty that, n a broadcast tck, dno s not selected to broadcast n any of the on-demand channel. Suppose the selecton of data tems n each broadcast tck s ndependent, then, n t broadcast tcks, the probablty that d no s not selected to broadcast s equal to O d no ) t. So, the probablty that a partcular vehcle wth the dwell tme t fals to access dno s calculated by: ) ) t P d no t,τ)= O d no Psuccess + P success ) ) We have the followng equaton to calculate O d no. Due to space lmtatons, the proof can be found n Appendx I from the supplementary materal, whch s avalable at [9].

4 O d no = ) Dno s= ) Dno q= Kso a s m) m= Kso b q r ) r= D no > ) ) ) where a s,as,...as, s Dno ) s the s th combnaton { of choosng data tems from the set of {p dno ) ) ), p d no,..., p d D no no } {p dno ) } } D no ) and K so a s m s the probablty of the s th combnaton. m= b q,bq,...bq s the q th combnaton of choosng data tems from the set of {p dno ) ) ), p d no,..., p d D no no } and r= b q r s the probablty of the q th combnaton. Lastly, for all passng vehcles whch attempt to access dno, the probablty that d no cannot be accessed s the deadlne mss rato of dno, denoted by R d no t,τ). Denote the probablty densty functon of vehcle nter arrval tme, τ, as f τ), accordng to equaton and : R d no t,τ)= P d no t,τ) f τ) f t)dτdt = ) t N O d no f τ)dτ + N O d no where pdno) ) t τ N + τ N ) ) f τ)dτ f t)dt N = 6) To sum up, derved from equaton and 6, for all data tems dn D n ), the expected deadlne mss rato s: E [Rt,r)] = D n = = D no pd n ) R d n t,τ) pdnp) R d np t)+ D no pdno) R d no t,τ) pdnp) P dnp t) f t)dt+ pdno ) P dno t,τ) f τ) f t)dτdt V. PERFORMANE EVALUATION A. Parameter Settngs Based upon the analytcal model, a seres of results are presented n ths secton to gve a clear exposton of the effects of dynamc traffc patterns to the system performance. The results are generated by the MATLAB wth specfc parameter settngs. Partcularly, we examne a typcal traffc model n whch the arrval of vehcles follows the Posson process wth the parameter λ. So, the nter arrval tme, τ, follows the Exponental dstrbuton and f τ) =λ e λ τ τ ). Also, suppose the dwell tme t of vehcles follows the Exponental dstrbuton wth the parameter μ μ > ) 7) Expected deadlne mss rato Fg.. 6 Number of push data tems Expected Deadlne Mss Rato n Default Settng where f t)=μe μt t ). Wthout loss of generalty, the Flat broadcast polcy s assumed n the push-based servce where ldnp )=. Last, assume the data access pattern K sp follows the commonly used Zpf dstrbuton [] wth the parameter θ: pdn)= /)θ d Dn n D n ). Note that for a / j) θ j= data tem n such a dstrbuton, the smaller ts ID s, the hgher ts access probablty s. Based on equaton 7, the expected deadlne mss rato s calculated as follows. The calculaton can be found n Appendx II of the supplementary materal [9]. E [Rt,τ)] = where pdnp ) + μ K sp D no pdno) + N λ e 7 e 8 )) Ksp + μ Dnp )) λ N) μ μ lno d no pdno) N =, p dnp ) ) ) = p d n, p dno j = p and Dnp, j Dno d np+ j n 8) To produce a stressful settng for system performance evaluaton, the default parameters are set as follows: D n = 8, θ =.8, t = 8, λ = 6 and =. As specfed n Secton III, the 6 servce channels and 8 data tems are classfed nto the push-based and on-demand servces: K sp + K so = 6 and D np + D no = 8. The expected deadlne mss rato s calculated by changng the proporton of the number of push channels to the number of on-demand channels, as well as the proporton of the number of push data tems to the number of on-demand data tems. For nstance, when the number of push channels changes from to 6, to mantan a total of 6 servce channels, the number of on-demand channel changes from 6 to correspondngly. The evaluaton result n the default settng s shown n Fgure. The x axs represents the number of data tems allocated nto the push-based set. The more data tems are allocated nto the push-based set, the fewer data tems reman n the on-demand set, and vce versa. The y axs represents the number of push channels. Lkewse, the more channels are )

5 TABLE II SUMMARY FOR THE EFFET OF DWELL TIME t Mn-expected hannel Data tem allocaton mss rato allocaton 6. K sp = D np = {dn,d n,...,dn }, D np = = D no =,dn 6,...,dn 8 }, D no = K sp = D np = {dn,d n,...,dn }, D np = = D no = {dn,dn,...,dn 8 }, D no =. K sp = = D np = {dn,d n,...,d6 n }, = 6 D no = {dn 7,d8 n,...,d8 n }, D no = TABLE III SUMMARY FOR THE EFFET OF DATA AESS PATTERNS Expected deadlne mss rato Number of push data tems 8 θ Mn-expected hannel Data tem allocaton mss rato allocaton.. K sp = 6 D np = {dn,d n,...,d n 8 }, D np = 8 = D no = /, D no =.8.6 K sp = D np = {dn,d n,...,d n }, D np = = D no =,dn,...,dn 8 }, D no =..76 K sp = D np = {dn,d n,...,d n }, D np = = D no = {dn,dn,...,dn 8 }, D no = 69 classfed nto the push set, the fewer channels reman n the on-demand set, and vce versa. Accordng to the result, there are two extreme cases where the expected deadlne mss rato equals. One case s that all the data tems are allocated nto the push set, whlst none of the channels are classfed as a push channel x=8, y=). In contrast, the other extreme case s that all the data tems are allocated nto the ondemand set, whlst none of the channels are classfed as an on-demand channel x=, y=6). Apparently, they are the two worst stuatons for data and channel allocaton, whch result n a % deadlne mss rato. Furthermore, the result also reflects that nether a pure push-based servce x=8, y=6) nor a pure on-demand servce x=, y=) can gve the mnmum expected deadlne mss rato. Only an approprate classfcaton of data tems and channels can acheve the best performance. B. Examnaton of Dynamc Traffc Patterns In the followng evaluaton, we examne two of the representatve factors, vehcle dwell tme and data access patterns, to capture the dynamc traffc characterstcs. Effect of dwell tme omparng the results n Fgure, both push-based and on-demand servces have better performance when the dwell tme s gettng longer. Ths s because longer dwell tme of vehcles gves a looser deadlne for data dssemnaton. Moreover, we also fnd that n a long dwell tme envronment t = ), the system performance s better when the push-based servce domnates. Ths suggests that when the deadlne s not so tght, compared to the on-demand servce, the push-based servce has more sgnfcant contrbuton to the system performance. The detal summary can be found n Table II. It clearly shows that the expected deadlne mss rato s gettng lower wth an ncreasng dwell tme. Meanwhle, the statstcs also suggest that, to have better performance, Fg.. Expected deadlne mss rato Expected deadlne mss rato Expected deadlne mss rato a) t = Number of push data tems b) t = Expected Deadlne Mss Rato under Dfferent Dwell Tme Number of push data tems a) θ = Number of push data tems b) θ =. Fg.. Expected Deadlne Mss Rato under Dfferent Data Access Patterns

6 more data tems should be classfed nto the push-based set n a longer dwell tme envronment. Effect of data access pattern In the Zpf dstrbuton, the data access pattern s gettng skewed wth an ncreasng value of θ []. omparng the results n Fgure, the system performance s mproved sgnfcantly n a skewed data access pattern. Ths s consstent wth the fndngs n prevous work []. Both push-based and on-demand servces can beneft from a skewed data access pattern, because there s a hgher potental to satsfy more data access attempts by broadcastng one data tem. Table III summarzes the effect of data access patterns. Accordng to the statstcs, n a less skewed data access pattern θ =. ), a pure push-based servce x = 8, y = 6) gves the mnmum expected deadlne mss rato. However, when the data access pattern s gettng skewed, the mportance of the on-demand servce becomes sgnfcant. When θ equals., more than 8% of data tems are allocated nto the on-demand servce. Ths s because n a hghly skewed data access pattern, the overwhelmng majorty of data access attempts concentrate on qute a few data tems hot data tems). The servce to these hot data tems s crtcal to the system performance. Therefore, only less than % of data tems are classfed nto the push-based set, whch promses a short broadcast perod for each push-based data tem and hence, results n satsfactory servce to those hot data tems. On the other hand, although most of the data tems are classfed nto the on-demand servce, the total query arrval rate remans low due to the very low access probablty of these data tems. Ths ensures the performance of on-demand servce. VI. ONLUSIONS AND FUTURE WORKS Wth efforts from dfferent partes ncludng automoble manufacturers, governments, and academa n dong researches and developng nfrastructures n vehcular networks, numerous potental applcatons lke road safety and value-added servces can be envsoned n the near future. In ths work, we frst presented a RSU-based data dssemnaton framework and nvestgated that unque characterstcs n vehcular networks brought new challenges n data dssemnaton. Next, an analytcal model was derved to examne the system performance n terms of provdng data servces wth tme-constrants. Fnally, the evaluaton results and analyss demonstrated that data dssemnaton n vehcular networks should be adaptable to the dynamc traffc patterns. Partcularly, dynamc channel and data allocaton s a crtcal, yet effectve mechansm n provdng a hybrd schedulng between push-based and on-demand servces. AKNOWLEDGMENT The work descrbed n ths paper was partally supported by a grant from the Research Grants ouncl of the Hong Kong Specal Admnstratve Regon, hna [Project No. tyu 7]. REFERENES []. HANG, R. HENG, H. SHIH, and Y. HEN, Maxmum freedom last schedulng algorthm for downlnks of DSR networks, IEEE Transactons on Intellgent Transportaton Systems, vol. 8, no., pp., 7. [] P. Yanln, Z. Abchar, and J. hang, Roadsde-aded routng RAR) n vehcular networks, n Proceedngs of the IEEE Internatonal onference on ommuncatons I 6), Istanbul, Turkey, 6, pp [] VIS, ooperatve Vehcle Infrastructure Systems, webste, access date: th, Mar,. [] SAFESPOT, ooperatve vehcles and road nfrastructure for road safety, webste, access date: th, Mar,. [] AHSRA: advanced cruse-assstant hghway system research assocaton, webste, access date: th, Mar,. [6] ITS Homepage, Road Bureau, MLIT, webste, access date: th, Mar,. [7] U.S. Department of Transportaton: ntellgent transportaton systems, webste, access date: th, Mar,. [8] M. Mors Mahmod, I. Khall, E. Uhlemann, and N. Nygren, Wreless strateges for future and emergng ITS applcatons, n Proceedngs of the World ongress on Intellgent Transport Systems ITS), New York ty, USA, 8, pp.. [9] O. Maeshma, S. a, T. Honda, and H. Urayama, A roadsde-tovehcle communcaton system for vehcle safety usng dual frequency channels, n Proceedngs of the IEEE Intellgent Transportaton Systems onference ITS 7), Seattle, USA, 7, pp. 9. [] T. Mak, K. Laberteaux, and R. Sengupta, A mult-channel VANET provdng concurrent safety and commercal servces, n Proceedngs of the AM Internatonal Workshop on Vehcular Ad Hoc Networks VANET ), ologne, Germany,, p. 9. [] G. Korkmaz, E. Ekc, and F. Ozguner, Internet access protocol provdng QoS n vehcular networks wth nfrastructure support, n Proceedngs of the IEEE Intellgent Transportaton Systems onference ITS 6), Toronto, anada, 6, pp. 7. [] M. Jhang and W. Lao, On cooperatve and opportunstc channel access for vehcle to roadsde VR) communcatons, n Proceedngs of the IEEE Global Telecommuncatons onference GLOBEOM 8), New Orleans, USA, 8, pp.. [] K. Kuramoto, K. Fujmura, and T. Hasegawa, The mult-class zone ITS communcaton scheme for real-tme communcatons n ntersectons, n Proceedngs of the IEEE Intellgent Transportaton Systems onference ITS 7), Seattle, USA, 7, pp.. [] K. Blstrup, A survey regardng wreless communcaton standards ntended for a hgh-speed vehcle envronment, Techncal Report IDE7, Tech. Rep. [] F -: Amendment of the commsson s rules regardng dedcated short-range communcaton servces n the.8-.9 GHz band, Federal ommuncatons ommsson, f Report and Order, Dec.. [6] F -, Federal ommuncatons ommsson, f Report and Order, Feb.. [7] Q. Xu, T. Mak, J. Ko, and R. Sengupta, Vehcle-to-vehcle safety messagng n DSR, n Proceedngs of the AM Internatonal Workshop on Vehcular Ad Hoc Networks VANET ), Phladelpha, USA,, pp [8] D. Aksoy and M. Leung, Pull vs push: a quanttatve comparson for data broadcast, n Proceedngs of the IEEE Global Telecommuncatons onference GLOBEOM ), Dallas, USA,, pp [9] Report on RSU-based real-tme data access n dynamc vehcular networks, kalu/report.pdf, webste, access date: 6th, Jul.,. [] G. Zpf, Human Behavor and the Prncple of Least Effort: An Introducton to Human Ecology. ambrdge, Mass.: Addson-Wesley Press, 99. [] V. Lee, X. Wu, and J. NG, Schedulng real-tme requests n ondemand data broadcast envronments, Real-Tme Systems, vol., no., pp. 8 99, 6. 6