Estimating Traffic Demand using Uplink Information from Infrared Vehicle Detectors

Transcription

1 Estimatig Traffic Dema usig Uplik Iformatio from Ifrare Vehicle Detectors Yoshito Mashiyama (1), Masao Kuwahara (2), Toshio Yoshii (2), Satoshi Niikura (3) a Toshihiko Oa (4) (1) Matsushita Commuicatio Iustrial Co., Lt Tsuashima-higashi, Kohoku-ku, Yokohama 223, Japa Tel: Fax: masiyama@is.mci.mei.co.jp (2) Uiversity of Tokyo Roppogi Miato-ku Tokyo, Japa Tel: Fax: (3) Kaagawa Prefectural Police Heaquarters 2-4 Kaiga-ori, Naka-ku, Yokohama 231, Japa Tel: ext.4726 Fax: ext.4730 (4) Matsushita Commuicatio Iustrial Co., Lt Tsuashima-higashi, Kohoku-ku, Yokohama 223, Japa Tel: Fax: a@is.mci.mei.co.jp SUMMARY Curretly, various traffic coitios are measure by usig ifferet etectors such as loopcoil a so o. I Japa, ifrare vehicle etectors (IRVD) were itruce a istalle i urba areas sice This vehicle etector has two fuctios, oe is coutig vehicle umbers, a the other is trasmittig traffic iformatio to each river. I this paper, i orer to estimate a OD matrix, we gathere a store uplik iformatio from ifrare vehicle etectors i Yokohama, Japa. INTRODUCTION Sice we have become a avace iformatio-oriete society i recet years, traffic sigal cotrol systems, roa traffic iformatio provisio systems, two way commuicatio betwee roa, vehicle a car avigatio systems usig satellite commuicatio are spreaig. It is ecessary, therefore, to correctly aalyze a estimate traffic pheomeos a coitios. Aitioally traffic pheomeos are yamic a traffic coitios are chageable. Uer these circumstaces ifrare vehicle etectors, oe of the meths of etectio have bee istalle alog oriary roas. The fuctio of this etector is to provie two-way commuicatio betwee vehicle a river i aitio to the covetioal meth. The movemet of each vehicle which carries a car avigatio system, ca be caught by usig two-way commuicatio, so as various iformatio such travel time, routig a vehicle types ca be gathere. I this paper, we propose a estimatio meth, to ivestigate the characteristics of the etector, aalyze iformatio o the etector, a use these to put together cocers about reucig traffic jams a O-D estimatig.

2 O-D ESTIMATING Fist of all, it is assume as X as estimates O-D ema from lik o to lik, there is relatio betwee each lik a each ifrare vehicle etector, which is show i figure 1. Therefore, X is traffic ema which is the origi poit of etector o a the estiatio poit of etector. For simplicity we eglect subscriptios of time. Ifrare Vehicle Detector Lik Figure 1 Locatio of Ifrare Vehicle Detectors o Network We ca oly kow the O-D patters of uplik ata from vehicles with a i-vehicle uit (ODPU) at each time zoe. We CANNOT recogize all O-D patters of every vehicle (ODPE). However traffic volume at each lik is measure by usig aother fuctio of IRVD. Accorigly, we will be able to estimate the ODPE usig the istributio of ODPU. I this sectio we propose a estimatig meth which is base o the follow assumptios. Assumptio 1 : There is o measuremet error i traffic iformatio which is traffic coutig a two way commuicatio betwee ifrare vehicle etector a i-vehicle uit Assumptio 2 : The ODPU represets the ODPE. The traffic volume at estiatio lik which coul be observe by usig the coutig fuctio of IRVD, is etermie as X. If the p is give the rate of traffic ema X ca be writte as follows. X = X p If the p is ukow, the ODPU represets the ODPE i assumptio 2, estimate by p = : the observe vehicles o estiatio lik of the ODPU i a (1) (2)

3 time zoe : the observe vehicles that lae goe through estiatio lik As the is the umber (OD traffic ema) of vehicles i the, it follows the variace of (, p ). Therefore the variace of p 1 p (1 ) Var[ p ] = Var Var[ ] = 2 = (3) Estimate OD traffic volume X is calculate as X = X p by settig p that is estimate by eq (3) ito p i eq (1). If the X a p are iepeet i the variace of X a the traffic volume X is followe by Poisso istributio, E [ X ] = Var[ X ] = X Var X is writte as are efie, a [ ] Var[ X ] = Var[ X p ] 2 = E[ p ] Var[ X ] + E[ X 2 ] Var[ p ] = p = X 2 p Var [ X ] + E[ X ] p + X 2 p (1 p (1 p ) EXPERIMENT SYSTEMS The experimet system cofiguratios show as Figure 2. ) (4) THE CHARACTERISTICS OF AN INFRARED VEHICLE DETECTOR A ifrare vehicle etector has two fuctios. Oe woul measure traffic volume similar to tay's covetioal etectors. The seco woul collect vehicle ietificatio umbers a trasmit etaile traffic iformatio, which iclues vehicle ietificatio umbers, previous-passig etector umbers, origi-passig etector umbers, car types, travel time a more, to iiviual rivers, by commuicatig irectly to the i-vehicle uits their cars are equippe with. CENTRAL SYSTEMS Cetral Systems are compose of the traffic iformatio host computer, traffic iformatio frot-e processors a iformatio moitors, all combie with the LAN (Local Area Network). The traffic iformatio host computer is coecte to existig traffic cotrol systems so traffic cogestio iformatio a other ata ca be trasferre. The fuctio of the traffic iformatio frot-e processor is to receive such uplik iformatio as vehicle

4 ietificatio umbers from ifrare vehicle etectors, the calculate travel time matchig these ietificatio umbers. Aitioally, it shoul collect owlik iformatio eite by the traffic iformatio host computer a se that ata to each ifrare vehicle etector. Kaagawa Prefectural Police Heaquarters Traffic Cotrol Ceter Traffic Iformatio Host Computer Traffic Iformatio Frot-e Processor Iformatio Moitor Multiprexer/Trasmitter Egie Start Geetate iitial ID Uplik (Iitial ID) Get ew ID & store time Dowlik (Give ID&Det.No.) Iitial Passig Det. Ifrare Vehicle Detecor Uplik (Prev.DetNo, Travel time,give ID) Passig Det. Calculatig Travel Time Dowlik (Det.No.) Egie Stop Delete ID Ifrare Vehicle Detecor Figure 2 Experimet System Cofiguratio a Commuicatio Sequece ANALYZING USING DATA Coitios of usig ata are as follows. Date; from 22 March 1998, to 3r April 1998, 14 ays. Place; All ifrare vehicle etectors (about 1,100) istalle i Kaagawa Pref., Japa Data; Uplik iformatio UPLINK INFORMATION Uplik iformatio cosists of vehicle ietificatio umbers, preset passig etector umbers, previous-passig etector umbers, previous-passig prefecture umbers, origietector umbers, car types, travel time, a others. PRIMITIVE ANALYZING DATA For the uplik iformatio from all ifrare vehicle etectors, more tha 15,000 to 17,500 cases per ay are coute, a have bee icreasig every moth. Not oly oes this ata cotai vali iformatio, but it also ivolves a lot of errors a icomplete ata. Therefore erroeous ata such as abormal preset-passig etectios, abormal previous-passig etectios, abormal origi etectios a others, are iscare from a variety of view poits. We got vali ata, which amoute to 40% of the raw ata, fialy expa ata up to 70% usig a meth which ca repair icomplete ata.

5 RESULT OF UPLINK INFORMATION RESULT OF BETWEEN UPLINK INFORMATION AND TRAFFIC VOLUME The uplik iformatio a traffic volume are compare i a poit. As a result from figure 3, they are t relate. The rate of the vehicles with i-vehicle uits is 0.2% i this area Error Vehicle Truck Other Uplik (x 500) Traffic Volume Figure Result of Uplik Iformatio RESULT OF CAR TYPES The left-ha figure 3 expresses the percetage of car types. 80% of the total are oriary vehicles. Most i-vehicle uits are use for persoal use. Figure 4 Sample of Tracig of Uplik Iformatio RESULT OF TRACING DATA This trip shows a tracig vehicle which the most uplik iformatio i 22 March I this sample expresses the istace betwee origi a estiatio which is 23.2km a the travel time is 29 miutes. Furthermore, all vehicles trips are raw i a similar maer.

6 CONCLUSIONS I this paper, we aalyze the uplik iformatio of ifrare vehicle etectors, a some of the uplik iformatio was complemete. As well as the uplik iformatio is useful because of estimatig OD, traffic emas a travel time easily. As a result, we propose the followig cocers regarig the future. - It will be ecessary to check o abormal iformatio, regarig tracig ata of each vehicle. - The esity of IRVD is scattere a the umber of uplik iformatio is very small i the preset coitio of our coutry. - It will be require to pick up tracig errors geometrically usig igital map ata. After this, we ca estimate a OD matrix a accurate preictio of traffic arrival flows. REFERENCES 1) Kuwahara. M, Mugikura T., Niikura S. a Oa T., Estimatig OD matrix usig Beaco Iformatio, ) Oa.T, Takeuchi.K, Yoshio Y.a Niikura S., Evaluatio of Measure Travel Time Utilizig Two-way Commuicatio i UTMS, proceeig of the 3r Worl Cogress o ITS, Orlao, 1996