Cooperative Model for Idle Traffic Optimisation

Transcription

1 Cooperatve Model for Idle Traffc Optmsaton Hrvoje Barčevć, Jasmn Ćelć, Snša Vlke The growth trend and the trend of populaton mgraton nto cty centres, as well as the contnuous ncrease n the number of motor vehcles cause traffc jams and congestons n most major ctes n the world. In order to elmnate the problems caused by the crclng n search of avalable parkng space,.e. by the congeston, dfferent methods and approaches are mplemented. Frst of all, these are the approprate strateges and polces that are made n order to regulate the exstng stuaton and to demotvate the arrval of part of the populaton n the cty centre, whch does not represent a necessty. Another way to elmnate the above problem, whch s the most emphaszed n ths paper, s represented by the ntellgent parkng systems. Intellgent parkng systems are used to gude the drver to the avalable parkng space and to provde nformaton on the characterstcs of t. Normally, they are also used for parkng reservaton and payment, but n most cases they were related exclusvely to closed parkng lots and garages. Wth the technologcal progress and the development of ntellgent sensors t has become possble even to control street parkng spaces, whch s especally emphaszed n ths paper. Also, an advanced model of gudance to avalable parkng spaces usng a cooperatve approach s proposed. The smulaton model presentng the advantages of ths system compared to the classcal approach, whch are also hghlghted n ths paper, s developed. KEY WORDS ~ ~ Intellgent Transportaton Systems (ITS) ~ ~ Cooperatve approach ~ ~ Advanced Transport Informaton Systems ~ ~ Gudance and Navgaton Systems ~ ~ Routes searchng ~ ~ Parkng lots n ports Unversty of Rjeka, Faculty of Martme Studes Rjeka, Rjeka, Croata e-mal: hrvoje@pfr.hr. INTRODUCTION The development of urban centres and the ncreasng number of vehcles on the roads around the world, and ncreasng demand for parkng spaces n cty centres serously affects the qualty of traffc. Searchng for free parkng places ncreases the traffc volume whch cannot be adequately met by the exstng capacty of the transport network. Ths problem s partcularly expressed n large harbours, especally durng the tourst season (Barčevć, 200). Consequently, ths may affect traffc congeston and thus reduce the demand for travel n the desred area. Elastcty, as one of the characterstcs of ths approach for resolvng cty centre congeston, enables precse algnment to the target group of users n relaton to space and tme. The man goal of the system for nformaton and referral to the parkng spaces (Parkng Gudance and Informaton, PGI) s to reduce traffc caused by searchng for a sutable free parkng space n cty centres and large parkng area (Krpan et al., 2007). At the entrances and exts of the parkng lots or on ndvdual parkng spaces sensors or devces for countng vehcles to allow the collecton of data on the number of occuped parkng spaces are usually nstalled. Sensors used for ths purpose are usually nducton loops, nfrared, ultrasound and magnetc sensors (Zhou and L, 204), laser technology, etc. The development of nformaton and communcaton technologes and contnued efforts to reduce the dmensons of electronc components and crcuts enabled the nstallaton of a mcrocontroller, a wreless transcever and a power supply, together wth certan sensor, n the traffc nfrastructure (Jeffrey et al., 202; J et al., 204; Geng and Cassandras, 202). These ntellgent sensors provde partally or fully processed nformaton to ts users n the form of messages whch usually, amongst the other data, contan the poston of avalable parkng spaces and nformaton about possble transport routes to the same, and may provde other Trans. mart. sc. 206; 0: 5-0 TRANSACTIONS ON MARITIME SCIENCE 5

2 useful nformaton such as zones, total number of avalable parkng spaces n the parkng lot, cost of parkng and etc. All of ths technology does not provde any guarantee that ntellgent parkng system wll work effcently n all possble condtons (Zanella et al., 204). 2. APPLICATION OF ITS FOR LOCATING VEHICLES AND NAVIGATION One of the basc requrements for qualty performance of ntellgent parkng systems, especally n terms of on-street parkng, s the nformaton about the vehcle poston. Automatc Vehcle Locaton (Automatc Vehcle Locaton, AVL) s a range of technologes that nclude technology postonng, mappng and communcatons. AVL technology s rarely used as a standalone applcaton, but s an ntegral part of many ITS servces such as route gudes, transt passenger nformaton, fleet management of commercal vehcles, computer-asssted forwardng, congeston detecton and return of stolen vehcles. The basc components of the AVL and navgaton technologes are: locaton technology and postonng systems, maps, geographc Informaton Systems (GIS) and map-matchng, route gudance and detectng route dsplay and dstrbuton of nformaton. A specal approach for determnng the poston s n the use of moble devces (Fgure ). The man advantage of ths approach s the ablty to use the exstng nfrastructure, whch sgnfcantly reduces the costs of the establshment of such servces. Moble telephony has already allocated frequency band and a large user base that s requred to support such an dea. Fgure. Navgaton system NAVIGON on connectng moble devces. For the purpose of postonng and navgaton dfferent charts are used, usually raster or vector maps. Raster maps are bascally mages that are sutable for showng detals such as heght dfferences usng topographcal ndcatons. Vector maps are drawngs made up of ponts, lnes and polygons stored n the database n the form of x and y coordnates. Vector maps show much less graphc detals from the raster map. Raster maps have a number of dsadvantages whch are, regardless of ther benefts, almost useless n most AVL applcatons. They requre far more memory than vector maps; they need much more tme for renderng on the devce dsplay; and are fxed relatve to the orentaton because when turned text wll rotate and gong to be showed upsde. Also, regardless of the ncrease or decrease n the chart t wll show the same level of detal, and the parts of the text may become completely llegble. 3. ADVANCED MODEL REFERENCES AVAILABLE ON THE VEHICLE PARKING SPACES Smulators of traffc are an essental tool for the study and research of traffc. Mcrosmulaton software solutons are able to model the movement of ndvdual vehcles travelng a predetermned road network. Mathematcal models are essentally an ntegral part of smulators, and provde a realstc vew of the behavour of drvers n the road network. Unlke determnstc models, stochastc approach and better resoluton of mcroscopc smulaton software solutons provde a better vew of the behavour of drvers and vehcles n the transport network, especally when t comes to complex traffc problems such as the mpact of ncdents or parkng on the traffc entrely. Models of road traffc bascally consst of transported enttes, road vehcles and the road network. Each of these factors s dstngushed by specfc characterstcs that should be determned as part of the smulaton model n order to acheve realstc condtons n whch the proposed advanced system should operate. For the purpose of ths study a conceptual model of the system that refers to avalable parkng spaces by usng a cooperatve approach has been made. The proposed system s n contrast to smlar exstng solutons based prmarly on cooperaton between on-board unt (OBU) and roadsde unt (RSU). The data provded by the Advanced Traveller Informaton Systems (ATIS) s used only n cases where t s necessary (no vehcles n the communcaton range, there s no need for hstorcal data, etc.), n order to acheve effectve nformaton sharng. Informaton from the sensors, whch are part of the transport nfrastructure, are processed on the spot usng a mcrocontroller and forwarded drectly to the system n a nearby vehcle, whch then forwards the nformaton to other vehcles n the seres untl t fnds a person of nterest. So, sensors of the road unt and the vehcle have a drect ablty to process the data and sort of ntellgence to be consdered as mcrocomputer unts. A vehcle equpped wth the approprate equpment on the arrval n the area of communcaton range of RSU starts recevng the avalable nformaton and transfers them to other vehcles around t. At the same tme ths nformaton, dependng on the communcaton capabltes, forwards tself to 6 do: /toms.v05.n0.00 Hrvoje Barčevć et al.: Cooperatve Model for Idle Traffc Optmsaton

3 Fgure 2. Flowchart travel, search of avalable parkng space and parkng wth a cooperatve approach. Trans. mart. sc. 206; 0: 5-0 TRANSACTIONS ON MARITIME SCIENCE 7

4 the ATIS. Based on the above cooperaton between vehcles and RSU the nformaton about the avalablty of parkng spaces s transferred. When enterng the street or arrvng n the port zone where the parkng spaces are, vehcles know n advance where to park, thus reducng the need for unnecessary crclng n search of avalable parkng space. 3.. The Calculaton of Travel Tme and Total Cost Traffc stuaton and thus travel tmes are changng durng the perod that takes tme to assgn the parkng lot. The total smulaton tme s dvded nto smaller ntervals to easly evaluate consdered travel tme. It should not be less than 5 mnutes and no more than half the tme nterval set by the orgn-destnaton (OD) matrx. Durng the smulaton process the travel tme for each edge of the network (tme between two nodes) s measured separately. Every vehcle after leavng a certan edge reports about tme spent, and then calculates the average of all travel tme spent at the edge whch then presents the resultng travel tme. Specal attenton s gven to vehcles that are retaned on an edge longer than the duraton of the evaluaton perod because t s a sgn of the presence of congeston. Ths s the report about duraton of outages. The measured travel tme s not used for the purposes of search and selecton of route n the same teraton, but ts nfluence on the next one. The development of models and smulatons was acheved by the usage of the software soluton named VISSIM. Namely, smulaton program VISSIM allows the use of not only the travel tme from the teratons that mmedately preceded the current but also all other early teratons, n two ways (PTV VISSIM 6 User Manual, 204):. Exponental smoothng - the user determnes the smoothng factor that calculates the mpact of prevous teratons so that the oldest teraton have the least mpact, 2. Method of Successve Averages (MSA) - calculates the arthmetc medum of all teratons, and on the bass that determnes the weght factor. The hgher the number of teratons that are consdered n analyss, the less the nfluence of each next teraton. Smoothed travel tme s calculated as s result from the set of values, as a weghted sum of the old teraton and current teraton. The results of smoothed value represent travel tme whch s expected n the next teraton. Travel tme for partcular edge measured n the tme nterval of the one teraton s exponentally smoothed before makng a decson about selecton of the route, n a followng way (Barcelo, 200): T a = ( - α ) T a n-, K + α TO a n,k () sgnfcance: K - ndex of evaluaton ntervals durng the perod of smulaton, n - ndex of teraton allocaton α - edge ndex TO a -measured travel tme on the edge a for perod k n teraton n T a - expected travel tme on the edge of a for perod k n teraton n a - smoothng factor. Estmated travel tme usng MSA s calculated as follows: T = ( - ) T n-, K + TO N + n meanng: N - user-determned value, K - ndex of evaluaton of the ntervals durng the smulaton n - ndex of teraton allocaton - edge ndex TO - measured travel tme on the edge for a perod k n the teraton n, T - expected travel tme on the edge and for the perod k n the teraton n, n + N - varable factor that depends on smoothng parameter N and teraton allocaton. Travel tme s not the only factor for nfluencng the choce of route. The functon of general costs s a lnear combnaton of travel tme, travel length and the fnances, such as toll payment. Unlke the tme needed to travel these factors are ndependent of the traffc stuaton and are not determned for the smulaton. Components of the cost can be determned separately for each class of vehcle that the user had generated earler. Accordng to these three factors, the overall cost C for each edge s calculated as a weghted sum of: C = α t + β s + Υ C f + C n2 N + n meanng: α, β, Υ Factors whch may, f necessary, determne the dfferent behavour of the drver n the choce of routes, s determned by the geometry of the lnk, C f sum of all costs of lnks that form an edge. Cost of the lnk s the product of ts length n klometres and the cost per klometre length specfed n ts parameters. (2) (3) 8 do: /toms.v05.n0.00 Hrvoje Barčevć et al.: Cooperatve Model for Idle Traffc Optmsaton

5 Ths cost ncludes the value of the parameter surcharge C n2 addtonal cost lnk (surcharge 2) that s not loaded wth factor Υ Search and Selecton of Sutable Routes The route s a seres of edges whch descrbes the path through the network. Routes begn and end at the parkng spaces. The general cost of the route s determned by the overall cost of ts all edges (PTV VISSIM 6 User Manual, 204): C R = C a aєr (4) However, ths process can lead to useless routes jonng the group whch s generated. Useless routes are those routes that replace any exstng route of the known route, wth lnk of much greater length, and they are consdered obvous detours. The length of a seres lnk that wll be qualfed as a bypass s determned by the user. Wth dynamc assgnment drvers are choosng the route at the tme of leavng the source parkng lot. One of the basc assumptons when choosng the route n VISSIM s that not all drvers use the best route, but wll use all known pathways n a way that the larger share of the traffc wll be dstrbuted n better way. Qualty of route s evaluated by overheads, whch are contrary to the utlty, whch s part of the theory of dscrete decson. Therefore, utlty s determned as the recprocal of overheads (PTV VISSIM 6 User Manual, 204): meanng: C general costs, R route. = C j (5) Between the startng pont and the parkng lot, there s usually more than one possble path, so smulaton program VISSIM's model s used to help the drver n hs choce of route. Wthn the dynamc assgnment of VISSIM, t weghts the number of avalable routes and carres out the selecton. Route selecton s a specal case of dscrete choce because the probablty of selecton of alternatves must be calculated for a gven set of dscrete alternatves. Usng the dynamc assgnment the set of avalable routes can be dsplayed wthn orgn-destnaton set. There s no effcent algorthm n fndng the best of several routes, but there s a possblty of fndng the rght one (Van der Waerden, 202). For every OD par n dynamc teraton the best route s requred, whch solves the afore mentoned problem, because traffc stuatons change travel tme from teraton to teraton untl the certan crtera s satsfed. Ths wll be a way to fnd a dfferent best route that VISSIM wll save to a fle wth.weg extenson so that they could be avalable for future teratons. The crteron for the selecton of the best routes s general expense. Due to the fact that the weghtng coeffcent overheads depend on vehcle type, dfferent types of paths wll be selected for dfferent types of vehcles. In the frst teraton, there s no travel tme nformaton from prevous teratons and, therefore, nstead of usng the length of travel tme, the length of the route n meters s used. A set of known ways wll be used more f drvers are somehow encouraged to try new routes. For that purpose, the followng teratons of edges, on whch the vehcle dd not pass, fcttous travel tme of 0. second wll be added n order to ncrease ther attractveness durng searchng for a route. meanng: - usefulness of travel tme j, C j - overheads of travel tme j. A - functon that s most often used for choce of routes s Logt functon (Ben-Akva et al., 984): e μuj ) = e μuj meanng: usefulness of route j, ) probablty of selected route j, μ parameter of model senstvty (>0). A small value of the parameter senstvty wll result wth dstrbuton wthout larger nfluence of the utlty whle the greater value wll result n fact that almost every drver wll choose the best route. If the Logt functon, whch s translatonal rreversble, s appled wth the prevously mentoned cost functon, the model wll be assgned the same mportance wth smaller dfferences between travel tme for smaller values as wth the relatvely large (e.g. a lot of sgnfcant dfferences n the travel tmes between and 8 mnutes wll be dentfed wth almost neglgble dfference trps of 6 and 3 mnutes). In order to approxmate the actual assessment and dstrbute traffc demand to all known routes as OD pars, VISSIM uses Krchhoff dstrbuton formula (PTV VISSIM 6 User Manual, 204): (6) Trans. mart. sc. 206; 0: 5-0 TRANSACTIONS ON MARITIME SCIENCE 9

6 ) = U k U k j meanng: usefulness of route j, ) probablty of choosng the route j, k parameter of model. In ths form of the utlty the dstrbuton rato s determned, and t does not represent the absolute dfference utlty. Ths means that n the above mentoned case there wll be lttle dfference n the allocaton of traffc for the paths whose travel tmes are 6 and 3 mnutes, but t wll take a lot more traffc to the path whose travel tme s 8 mnutes compared to the path whose travel tme lasts mnutes. Krchhoff dstrbuton formula can be expressed wth Logt functon f the utlty functon s dsplayed n the logarthmc form: (7) approach, wth the emphass on the synergstc effect of the optmal route selecton and the correspondng avalable parkng space. The proposed advanced system model whch refers to the avalable parkng spaces should also encourage the creaton of new strateges and polces for parkng, whch would reduce the costs for cty authortes, entrepreneurs and busness companes, developers and users, and mprove the qualty of parkng servces and acheve substantal benefts for the whole communty. It should speed up the process of development of future systems that use smlar technology and provde approprate nformaton to traffc engneers, desgners and planners for fndng a parkng lot and become a sold support for the development of ntellgent ctes. Ths s of partcular nterest n tourst destnatons (passenger harbours) n Croata, where there s an evdent peak traffc demand n the tme of tourst season. REFERENCES Barčevć, H., (200), Tehnologje kopnenog prometa, Rjeka: Pomorsk fakultet u Rjec and Glosa d.d. Barcelo, J., (200), Fundamentals of Traffc Smulatons, vol. 45., New York: Sprnger. ) = = = U k k log Uj e where C j s overall cost of the route j. For the devaton dstrbuton n overlappng route VISSIM offers expanson of route model wth factor dentty. Based on the current stuaton n the smulaton model and pre-set crtera the parkng space for vehcles enterng the modelled transport network s dynamcally determned. In the case where one of the vehcle wthout route gudance (RG) occupes a parkng place desgnated for the vehcle wth the RG, RG vehcle wll select new avalable parkng space and calculate the approprate route to the same usng the same algorthms as n the tme of entry nto the transportaton network smulaton model. 4. CONCLUSION U k j k log Uj e -k log Cj e Intellgent parkng system ncludes varous subsystems, and thus n ts mplementaton nevtably nclude knowledge from the doman of ITS. Cooperatve approach offers quck access to nformaton for users, where nformaton s gven as needed wthout necessarly drect contact wth advanced nformaton systems. Vehcles and sensors n the nfrastructure, whch now actually represent a sort of mcrocomputers, are able to analyse and process nformaton and forward them to each other for the purpose of better management. Ths resulted n the proposton of the advanced model of the system usng cooperatve -k log Cj e (8) Ben-Akva, M. E., Bergman, M. J., Daly, A. J. and Ramaswamy, R., (984), Modellng Inter Urban Route Choce Behavor, Nnth Internatonal Symposum on Transportaton and Traffc Theory, VNU Scence Press, pp Geng, Y. and Cassandras, G. C., (202), A New 'Smart Parkng' System Infrastructure and Implementaton, Proceedngs of EWGT202-5 th Meetng of the EURO Workng Group on Transportaton, Pars, France, September 202, pp , Jeffrey, J., Patl, R. G., Narahar, S. K. K., Ddag, Y., Bapat, J. and Das, D., (202), Smart Parkng System Usng Wreless Sensor Networks", 6 th Internatonal Conference on Sensor Technologes and Applcatons SENSORCOMM 202, August 9-24, Rome, Italy pp J, Z., Ganchev, I., O Droma, M., Zhao, L. and Zhang, X., (204), A Cloud-Based Car Parkng Mddleware for IoT-Based Smart Ctes: Desgn and Implementaton, Sensors, 4, pp , Krpan, Lj., Barčevć, H. and Frka, D., (2007), Drectonal Parkng Garage System- Plot Project of Applcaton Wthn the Cty of Rjeka ZIRP - Transportn lanc razvtak prometa, Zagreb, Croata, Aprl 5. PTV VISSIM 6 User Manual, (204), Karlsruhe, Germany: Planung Transport Verkehr AG. Van der Waerden, P., (202), Pamela, a Parkng Analyss Model for Predctng Effects n Local Areas, PhD Thess, Endhoven, Netherland: Technsche Unverstet Endhoven, Facultet Bouwkunda. Zanella, A., Bu, N., Castellan, A., Vangelsta, L. and Zorz, M., (204), Internet of Thngs for Smart Ctyes, IEEE Internet of Thngs Journal, (), pp , Zhou, F. and L, Q., (204), Parkng Gudance System Based on ZgBee and Geomagnetc Sensor Technology, 3 th Internatonal Symposum on Dstrbuted Computng and Applcatons to Busness, Engneerng and Scence, Xan Nng, Chna, November 24-27, pp , 0 do: /toms.v05.n0.00 Hrvoje Barčevć et al.: Cooperatve Model for Idle Traffc Optmsaton