VANET Multicast Routing for Congestion Control in Traffic Flow WSN

Transcription

1 VNET Mutcast Routng for Congeston Contro n Traffc Fow W Hao WU Nanjng Communcatons Insttute of Technoogy, Nanjng, Jangsu, , Chna bstract Securty data fuson s fundamenta and essenta for ntegent traffc systems. dynamc data fuson scheme s proposed to guarantee the securty and accuracy of sensed data. It s used to coect and manage mut-source traffc messages. It mproves K-neghbor nonparametrc regresson method on the bass of repeated occurrence feature of traffc fow state mode. Doube-screenng for neghbors s adopted, and dentfcaton functon based on state mode s ntroduced and traffc fow of past tme frame and traffc fows of reated turnng at upstream juncton and downstream juncton are consdered n agorthm. Ths heps to mprove capacty predcton of K-neghbor nonparametrc regresson. Fna predcton resut s gven by the weghted average method of match dstance recproca on the bass of state mode vector, thus predcton of short-tme traffc fow becomes more accurate n rea-tme. Keywords- ntegent traffc; wreess senor network; data fuson; mutcast routng; vanet I. INTRODUCTION Coecton of traffc message s a key technoogy n ntegent traffc system. Sensor network acqures accurate traffc parameters ke vehce speed, traffc fow and road occupaton rate for ntegent traffc system and these parameters are the bass for traffc management. tradtona montorng sensors ke nducton co and camera restrct the expendabty of exstng system and affect network effcency but mut-source traffc message data fuson based on W can acqure more accurate traffc message than tradtona sensor and more effectve traffc montorng and management [1], ncudng eectronc chargng, parkng management, juncton traffc gudance, energy conservaton and emsson reducton, are reazed. In the ntegent traffc scene, coected traffc message vares and orgna data sze s huge. Moreover, as vehce nodes are hghy mobe, vehce sensor network has the topoogca structure wth more changes than that of statc sensor network. though these data fuson schemes have advantages n some aspects, there are st defcences because agorthm for ntegent traffc system s too compex and energy consumpton of fuson node s too much. Mut-source traffc message data fuson based on appcaton of W n ntegent traffc system s studed n the thess and a dynamc fuson scheme s proposed. The scheme adapts to tme-space dstrbuton characterstc [6] of transportaton. Data fuson agorthm that desgn custers subject to credbty evauaton of the node can reduce energy consumpton of fuson node and mprove the reaty of sensed data. II. FUSION THEORY BSED ON EVIDENCE RESONING Evdence reasonng method proposed by Depmster Shafer [7] draws more attenton to uncertan factors and unknown factors and s more approxmate to human thnkng ogc and natura decson-makng process [8]. data fuson method s acqured on the bass of Depmster Shafer evdence reasonng theory and s apped to ntegent traffc fed. Depmster Shafer s theory s many apped to target dentfcaton and cassfcaton. In such fuson method, each sensor obtans oca decson and sends them to fuson center for fna aggregate decson. The basc concept s descrbed as foows: Suppose L s fnte proposton anguage, { w1, w2,..., w n } s possbe word set and w {=1,2,...,n} s one expanaton of L. proposton can be expressed as a subset of. Set has the foowng features: (1 fnteness. (2 Eements n the set repuse each other. Defnton 1 Suppose s dentfcaton frame, functon m: 2 [0,1] s a basc probabty assgnment, when and ony when (1 m( 0 and ( 1. If m(>0, s a Θ foca eement of the functon. Defnton 2 Functon Be: 2 [0,1] s caed beef functon. If Be(= (B, Be( here s caed beef B degree of. Defnton 3 Functon PI: 2 [0,1] s caed kehood functon. If PI(=.m(B, PI( here s caed B kehood degree of. Beef degree Be( refers to the sum of beef degrees of propostons that ceary support the proposton ' expressed by and kehood degree PI( refers to sum of kehood degrees of propostons that are expressed to potentay support proposton '. Reatons PI( Be( and PI( 1Be( are true. Combnaton rue: suppose m1 and m2 are two basc probabty assgnments on and ther foca eements are DOI / IJSSST.a IS: x onne, prnt

2 1,...,p and B 1,..., Bq respectvey. Be 1, Be2 and Be =Be Be are beef functon m(m(b<1 1 j 1 2 Bj nduced by m 1, m2 and m =m1 m2 so m =m1 m2 s defned as: m ( =0 m (=K m 1( m 2 (B j, Bj 1 K [ m 1( m(b j]. Bj The rue above can be generazed to mutpe m functons and Be functons and apped to ntegrate opnons of mutpe experts. In Depmster Shafer s theory, a pece of evdence can determne a basc probabty assgnment and further determne a beef functon. Therefore, Depmster Shafer combnaton rue s caed evdence combnaton rue. III. DT FUSION SCHEME. Basc Idea Coecton and management of traffc message are based on data meshng system. Traffc network s meshed subject to specfc prncpe and method [9, 10]. Custer head node n the custer area acqures sampng data of other members and fuses them nto non-redundant data set. Then the data package woud be transmtted to aggregate pont/base staton by makng tme stamp and geoogca poston on GPS. Base staton dstngushed fused data from dfferent custer areas accordng to tmestamp and geoogca poston and upoad then to appcaton ayer. ppcaton ayer uses credbty evauaton and evdence functon reabty dstrbuton to cacuate fuson resut and makes the fna decson. Fow of fuson scheme s as shown n Fgure 1. The scheme mprove the accuracy of mut-source traffc message data and proong the survva tme of network. Fgure 1. Data fuson scheme B. Data Fuson Process Custer head node n the custer area perodcay sends nqury message MSGREQ to other member nodes to synchronze tme. In a sampng term t, member node (vehce or facty on road sde sends response message REP to custer head and formazed descrpton of the message mode s as foows: m MSG _ REP D data Pos sensd Token D : preambe of m that refects data type sensed by node and nfers actua source data. It s dfferent from preambe of other data packages. :the dentty of sensor node data E( d sens, K, BS, where, K, BSs the symmetrc key between and BS to protect the prvacy of sensed data. d sens s the actua data sensed by, ncudng degree of support for the data vaue. Pos : traffc mesh coordnate ( x, y, to descrbe geoogca poston of message source and poston the target [11]. sensd : dentfer of current data package, dfferng from dfferent messages sent by the same sensor node. In sensd F( sensd 'modm expresson, functon F s monotone ncreasng functon and sensd ' s the dentfer of ast data package. Token : message verfcaton doman, Token SIG( D data Pos sensd, K _ TP, where, SIG s the dgta sgnature of node, K _ TP s prvate key for verfcaton of and they are both ssued by verfcaton center at network ayer. In tme nterva t, the sensed data woud be forwarded to custer head node from the coecton poston or va ntermedate node. In the ntegent traffc envronment, sensor node can capture varous data smutaneousy (ke ar fow, temperature and humdty. In order to dentfy data wth dfferent attrbutes, when source node forms preambe, specfc separator s nserted between the data wth dfferent attrbutes to dfferentate attrbutes and node types of data. The preambe regon vaue D n data package repaces actua data and executes data fuson. Custer head node woud sort out data packages wth the same attrbute nto same group and fuse them nto a new message []. In addton, the sensed ranges of members n the same custer area woud overap due to the dstance. Vaue of Pos area woud be expressed as sensed range n traffc network wth crce center of ( x, y and radus of parameter R. In ths range, other modes can acqure the same data. Therefore, ntroduced parameters D and Pos can avod data beng fused n dfferent groups and formng redundant data set. Suppose a group of aggregate data sets s composed of peces of dfferent messages, custer head fuses data and head DOI / IJSSST.a IS: x onne, prnt

3 Pos doman vaue of messages and forms new message maggr and sends t to base staton. Formazed descrpton of message mode s as foows: m MSG _ REP data Pos aggr BS head aggr aggr tmestamp Token ' IdLst, where head : dentty of custer head head dataaggr m. data Pos ( wmposx.., wmposy.. aggr w 1 w 1 ( w s the D doman data ength of message m, w w 1 tme stamp : tmestamp, fused data are cassfed by precse tmestamp and Pos aggr. Token ' : verfcaton sgn, Token ' SIG( head data aggr Posaggr tmestamp IdLst, K head _ TP, K head _ TP s the prvate key for verfcaton of head. IdLst : s the sequence of message sources that orgna data passed n aggregaton. IdLst E( head K head, BS, K head, BS s the symmetrc key between head and BS to protect prvacy of fused data. fter recevng the message m aggr, base staton starts verfyng t. It the message passes verfcaton, t woud accept data and upoad then to appcaton ayer of fuson center or maggr woud be abandoned and credbty of reated node woud be updated. C. Cacuaton and Evauaton of Fuson Resut ppcaton ayer assocates data and cacuate evdence functon on the bass of receved fuson data and fnay make decson. Frsty, st a the essenta attrbute vaues of sensng objects n dentfcaton frame to obtan decson cassfcaton set U [13] specfed by the dentfcaton frame, the set of attrbute subsets n. For exampe, check whether there s car on road, the dentfcaton frame s {0, 1}, where 0 s no car and 1 s car. Set U of the dentfcaton frame s {{0},{1}, {0 or 1}, }, where s zero eement. Evdence functon s the foundaton of evdence theory. Defne an evdence functon m for each pece of evdence to judge the event and the evdence functon woud map set U between [0,1], m:2 [0,1]. If s non-nu evdence functon vaue, m( s the support degree of defned functon m for. It can be nferred from defnton 1 that m( 0, m( 1. Gven mutpe sensors and 2 mutpe data source, the data sources make judgments n the same dentfcaton frame respectvey accordng to the evdence functon m they defned and then the data woud be effectvey combned accordng to evdence combnaton rue [14]. Probabty dstrbuton method based on credbty s ntroduced n the thess. Intaze the credbty of data source node r as 5. When the test resut of source node s consstent wth appcaton ayer, the credbty woud add 1 or subtract 1. then change defnton weght coeffcent w of functon for r as foows: 0, r g或 0 r g w, r g且 0 (1 max( r g Nsuccess Ntota Where, max( r s the maxmum r n the hstorca records of the area, g s constant coeffcent, suppose g 5. s the test accuracy of source node, whch s the proporton of the number Nsuccess that test resut of s consstent wth judgment of appcaton ayer n the prevous number N tota. When there are n data sources, beef functon Be ( of source node for evdence can be obtaned va equaton (1. ' Be( m( wm(,,1n Be ( 1Be (, Suppose I s the maxmum subset of event n U. Beef functons of n data sources are combned va equaton (2 and beef functon s updated subject to equaton (3. n ' ' ' Be(B m( 1 j... m( n j k m( j 1...p B 1 n ' 1 k [ m ( j] k 1 且 1...p 1 Gven B1,, Bm are mutuay excusve, t can be obtaned: PI( I 1 Be(B j B1B 2...Bm I When PI( I s ess than the crtca vaue 0, execute decson H0 or execute H 1. From perspectve of updated evdence functon, when data source s weght coeffcent w 1, support of for s decreasng actuay but the support for I ( I s ncreasng. fter turns of sampng, f the credbty r of s hgher, the generated weght (2 (3 DOI / IJSSST.a IS: x onne, prnt

4 coeffcent w s more approxmate to 1. On the contrary, t woud be more approxmate to 0. Once evdence functon ' m(s cose to 0, the evdence s hghy key to be negected and the support degree of the assurng event decnes. Reabty dstrbuton of evdence functon s reazed on the bass of credbty evauaton, whch weakens the mpact of ow reabty of dshonest evdence on the fuson resut. IV. PERFORMNCE SIMULTION the experment data come from rea-tme montorng for the traffc fow on Tbet Road Secton of Shangha Cty. Consderng traffc fow s perodc, montorng tme was from June 6, 2015 to June 13, 2016, ncudng 5 busness days and three days of festva. s there were few cars at nght, actua use vaue of traffc data s sma, the montorng tme frame was from 7:00am to 20:00pm and tmer nterva s 2mn. 30 orgna data sampes of traffc fow were obtaned. The data of the frst 7 days were used to estabsh hstorca sampe database and the data of the ast day served as test data. Damnghu Road Secton Dagram s as shown n Fgure 1. Traffc fow predcton method n the thess ncudes 5 parameters, ncudng, n, m, j, b and k. s the dmenson of state vector. State vector dmenson s drecty reated to predcton accuracy and agorthm effcent. n s the number of dots after passng the frst state mode match screenng. k s the number of dots after passng the second state mode match screen. n and k drecty affect predcton accuracy and agorthm effcency and excessvey bg or sma n and k woud reduce predcton accuracy. ccordng to the experment data above, vaues of parameters, n and k are obtaned. It can be seen from Fgure 2, when parameter ncreases from 2 to 6, predcton error decnes sgnfcanty; when ncreases agan, the predcton accuracy remans on the same eve because ncrement of requres more cacuaton. Its best match number s around 4. The fgure shows, even when the match number ncreases, the predcton effect woud not be mproved dramatcay and t woud may brng forth reverse effect so suppose =4 n the actua appcaton. Fgure 2. Impact of on predcton accuracy Observe Fgure 2. On the premse of confrmed state vector and predcton agorthm, when n ncreases from 40 to 50, predcton error drops sgnfcanty; when t ncrease from 50 to 65, the predcton error rses graduay but the rsng speed s sow, whch means the best match number n s 50. fter the frst neghbor number s confrmed, experment of the second neghbor number was conducted, k s the number of dots that passed the second state mode match screenng. s shown n Fgure 4, of k s excessvey bg, the predcton functon s excessve smooth and predcton accuracy drops. However, k cannot be excessvey sma, whch woud ncrease the rato of chance factor and affect predcton accuracy. Suppose k s 9 here. Fgure 3. Impact of n on predcton accuracy Fgure 4. Impact of k on predcton accuracy m s the number of reated turnngs at upstream juncton and j s the number of reated turnngs at downstream juncton. It s shown n Fgure 1 that there are 3 reated turnngs at upstream juncton and 3 reated turnngs at downstream juncton of Damnghu Secton, whch means m j 3. In the practca appcaton, traffc fow of current road secton at next moment s not ony reated to the traffc fow at ths moment but aso reated to reated turnngs at upstream and downstream junctons of current road secton. However, when weght vares, stuaton of Damnghu Secton sha be consdered overa, b s 0.5, a1 a2 am 0.2 and c1 c2 cm 0.3here. We compared tradtona K-neghbor agorthm and mproved K-neghbor agorthm based on state mode, substtuted parameters nto formuas to predct traffc fow at DOI / IJSSST.a IS: x onne, prnt

5 next moment and we used Matab n smuaton and obtaned better predcton resut. Therefore, predcton performance of K-neghbor agorthm based on state mode that s proposed n the thess s better. V. CONCLUSIONS Tradtona K-neghbor nonparametrc method s mproved on the bass of repeated occurrence feature of traffc fow state mode n the thess. Doube-neghbor nonparametrc regresson method s adopted, dentfcaton functon based on state mode s ntroduced n neghbor nonparametrc regresson method, traffc fow of past tme frame and traffc fows of reated turnng at upstream juncton and downstream juncton are consdered n agorthm so the predcton capacty of K-neghbor nonparametrc regresson method s mproved and fna predcton resut s gven by the weghted average method of match dstance recproca on the bass of state mode vector. Fnay, based on the anayss of measured traffc fow predcton resut, predcton of short-tme traffc fow obtaned by the mproved doube K-neghbor nonparametrc regresson method s more accurate and rea-tme. t s an effectve method to predct short-tme traffc fow and ts predcton resut s bass for the traffc gudance and contro servce of traffc management department and the method s vta for the traffc contro and gudance. CKNOWLEDGMENTS The hgh eve taents scentfc research project. REFERENCES [1] Chm T W, Yu S M, Hu L C K, et a. VSPN: VNET-Based Secure and Prvacy-Preservng Navgaton[J]. IEEE Transactons on Computers, 2014, 63(2: [2] khtar N, Ergen S C, Ozkasap O. Vehce Mobty and Communcaton Channe Modes for Reastc and Effcent Hghway VNET Smuaton[J]. IEEE Transactons on Vehcuar Technoogy, 2015, 64(1: [3] ndreas S, Pasca R, dehed G, et a. Comparatve nayss of Varous Routng Protocos n VNET[C]// Internatona Conference on dvanced Computng & Communcaton Technooges. IEEE, 2015: [4] ndreas S, Pasca R, dehed G, et a. Comparatve nayss of Varous Routng Protocos n VNET[C]// Internatona Conference on dvanced Computng & Communcaton Technooges. IEEE, 2015: [5] Mejr M N, Ben-Othman J, Hamd M. Survey on VNET securty chaenges and possbe cryptographc soutons[j]. Vehcuar Communcatons, 2014, 1(2: [6] Wang X, Qan H. Constructng a VNET based on custer chans[j]. Internatona Journa of Communcaton Systems, 2014, 27(11: [7] Btam S, Meouk, Zeaday S. VNET-coud: a generc coud computng mode for vehcuar d Hoc networks[j]. IEEE Wreess Communcatons, 2015, 22(1: [8] Wang M, Shan H, Lu R, et a. Rea-Tme Path Pannng Based on Hybrd-VNET-Enhanced Transportaton System[J]. IEEE Transactons on Vehcuar Technoogy, 2015, 64(5: [9] Dooan R, Muntean G M. VNET-Enabed Eco-Frendy Road Characterstcs-ware Routng for Vehcuar Traffc[C]// IEEE Vehcuar Technoogy Conference. IEEE, 2015:1-5. [10] Sngh S, grawa S. VNET routng protocos: Issues and chaenges[c]// Engneerng and Computatona Scences. IEEE, 2014:1-5. [11] Hussan R O H. Cooperaton-ware VNET Couds: Provdng Secure Coud Servces to Vehcuar d Hoc Networks[J]. Journa of Informaton Processng Systems, 2014, 10(1: [] Wang M, Zhang R, Shen X. Mobty-ware Coordnated EV Chargng n VNET-Enhanced Smart Grd[M]// Mobe Eectrc Vehces [13] Yan T, Zhang W, Wang G. Grd-Based On-Road Locazaton System n VNET wth Lnear Error Propagaton[J]. IEEE Transactons on Wreess Communcatons, 2014, 13(2: [14] Ln C C, Deng D J. Optma Two-Lane Pacement for Hybrd VNET-Sensor Networks[J]. IEEE Transactons on Industra Eectroncs, 2015:1-1. [15] Jnyu Hu and Zhwe Gao. Dstncton mmune genes of hepattsnduced heptatoceuar carcnoma[j]. Bonformatcs, 20, 28(24: [16] Jang, D., Yng, X., Han, Y., & Lv, Z. (2016. Coaboratve muthop routng n cogntve wreess networks. Wreess persona communcatons, 86(2, [17] Lv, Z., Tek,., Da Sva, F., Empereur-Mot, C., Chavent, M., & Baaden, M. (2013. Game on, scence-how vdeo game technoogy may hep boogsts tacke vsuazaton chaenges. PoS one, 8(3, e [18] Jang, D., Xu, Z., & Lv, Z. (2015. mutcast devery approach wth mnmum energy consumpton for wreess mut-hop networks. Teecommuncaton systems, 1-. [19] Lv, Z., Chrvea, J., & Gagardo, P. (2016. Bgdata Orented Mutmeda Mobe Heath ppcatons. Journa of medca systems, 40(5, DOI / IJSSST.a IS: x onne, prnt