Research
.『()ur,lnfofHarbinlnstituteofTech,lfJ1ogY(NewSeries),Vo1.12,No.3,2005 ;Researchontrafficcongestionmechanismandcountermeasures
;basedondynamictrafficassignment
;PEIn，一long.L4,vG一shltn
;裴玉龙,郎益顺
;(InstituteofTranspofiationResearch.HarbinInstituteofTeehnolugy,Harbin150090,China) ;Abstract:Trafficcongestioniswidelydist,‟ibutedaroundanetwork.Generally,toanalyzetraffi(congestion,
;staftctramccapacitvisadopted.Butdynamiceharacterisff.csmustbestudiedbecausecongestionisadynamic
;process.ADvnamicTrafficAssignmentmodelingfundamentalcombinedwithanurbancongestionanalysis
;metho(1isstudiedinthispaper.rhreemethodsarebasedoncongestionanalysis.andthesic(hasti(-USel‟o
pti—
;realD.‟1Amodelsareespeciallyconsidered.Correspondingly.adynamicsystemoptimalmodelissuggestedfol‟
;respondingcongestioneounternleasuresandanidealuseroptlnmlmodelforpredi(„tedcongestioncnItutel‟llle~lsnl‟e
;respectively.
;Keywords:dynamictrafficassignment;trafficcongestion;dynamicUSel‟optimal;dynami(„systemoptimal
;CLCnumber:U49lDocumentcode:AArticleID:1005—9l13(2005)034)2354)4
;Tramccongestiongreatlyinfluem‟escitydevelop—
;mentandactivityefficiencv.Itseriouslyinterfereswith
;theresidents‟workan(1lire.Toundel?standthelawof
;trafficcongestion.twocausesshouh1I)econsidered.
;Firstcauseoftrafficcongestionisirrationalplanningof
;thecityanditsroad‟snetwork.Improperfacilityusage
;withdeficientcapacityaccompaniedwithunbalanced
;flowdistributioncausessomefacilities‟tosupersatu—
;rate.AlsounsuitableTMC(Trafficnlanagementand
;contro1)strategy,notmeetingthedemandof”Static
;Management”and”DynamicControl”.1cadstoiJJ
;trafficstructureandwasteoftimeandspaceresources.
;Settingnewtrafficfacilitiesisverydiffi(uhinthecity
;becauseofaccountingforurbanareausage,compact
;layout.humanityan(1landscapes.Newfacilities‟plan—
;ningisalong—tennactionandthereforeisnotableto
;repaircurrenttrafficcrisis.ComparativelyTMCstrategy
;ismoreflexibleandadaptedtodynamictramc.Thedy ;namictrafficnetworkandTDM(trafficdemandman—
;agement)method,aswellasthecorrespondingTRG ;(trafficrouteguidance)system,havethepriorit)of ;tramflowoptimalityintheshot/tel?Il1.
;Thegeneralcongestionanalysismethodsare}rased ;onHCMornetworkcapacityconcepts..Thereforethe ;definitionof”traffic(?(ingestion”isina(?eordaneewith
;theserviceleve1.whichisgeneralb,rstati(?andrelated ;tothefacilities.Tlie(„ongcstionisl,e(?ausethestatic
;trafficdemandsexceedthefac/lities‟servicecapacity
;andflowwhenthevreachinstahiIitv.Inthismetho(1. ;facilitiesaretheanalysisitems.l‟heI”acilitvusei?sbe—
;haviorandrea(?tionareignm?e(1(vellItthe(?apa(?ifv ;Recei~(12003一()4—07
;simulationmethod.So.evelltheI)estTMCfun(tamen—
;talsareiBefficientbygeneralHCMOl?networkcapacity ;analysis.especiallyfol‟real—timeusage.
;TheDTAfdynamictrafficassignment)metho(1is ;themainmethodofdynamietram(networkanalysis. ;BecausetheATMS/ATISconceptisreal—timetrafficin—
;formationserviceandTMCimplemente(1.thedynamic ;trafficmodeliStheonlychoice.ThereforetheDTAis ;greatlyemphasized.Thenewtransportationplanning ;theoryisbeingdevelopedintheUSAImsedonthe ;DTA.
;Comparedtothestaticmode1.theDTArelieson ;itsdynamicfacets.Thestaticmodel(?0neption.while ;networkstructureisadeterminate.resultsinadefinite ;necksectionwithcongestionoccurring.TheDTAlnod—
;elanalyzescongestioninthreemethods.TheDfA ;modelhasalsoenabledcongesti(mmanagementby ;modeling.whichwillI)ediscussedlater.
;1TrafficCongestionMechanisimAnalysis
;rheDTAmodelingmetho(1analyzestt?afficconges ;tioninthreeways,traffiload,I/0(input/output)of ;theintersection.an(1intportantroutesele(?tion. ;1.1TrafficLoadandSectionCongestion
;Thetrafficlca(1consistsofalltheexistingvehi(?les ;ahmgastudiedsecti(,nCongestionhappenswhelltllP ;trafficloa(1exceedsthemaximum[)ermitte(1loa(1.AlSO. ;stochasticdisturbance.SU(?hasatraffi(?a(?(?identwhich ;resuhsinabnommlfraffi(?Ilowleadingto(„ongesti()nor
;over—saturation,canbea(„auseot‟set?tion(„ongestion.
;?
;235?
;
;JournalofHarbinInstituteofTechnology(NewSeries),Vo1.12,No.3,2005
;Trafficcongestioncausedbyaccidentsshould ;lyzedbvtrafficsimulationmodels‟.suchas
;(Input/Output)mode1.Theassignmentphase ;bythecongestionanalyzingmodelis:
;Q(t)=,(t)一x(t)Vt,
;beana.
;theI/O
;inferred
;whereQ(t)istrafficload,,(t)andX(t)areinput ;outputtrafficvolumerespectively(asshowninFig. ;一一一一一一
;Q(一一一一一.
;X(t)
;Fig.1IllustrationofI/otrafficflowmodel ;(1)
;and
;1).
;ThefuInctionreferstothetrafficflowattimete. ;qualtothedifferencebetweeninputandoutputflow. ;Thishappensonthesuppositionthatthesectionisa ;closedsystem.Thissituationmaybedescribedasa ;strictclosedsystemwithnonewtraffic.Italsomatches ;thestatictrafficstateasweanalyzethenetworkonthe ;suppositionofinputandoutputflowequalitythereby ;ignoringtheformofthetrafficassignmentmode1.How—
;everthisisincorrect.asweanalyzethenetworkbythe ;dynamictrafficmodel,thatisIVOimbalance:Q(t) ;o.evenintheformofdiscretetimemode1.Firstthis ;expressesthedifferencebetweendynamicandstatica. ;nalysis,andsecondthisshowsthecontrollablefactorof ;inputandoutputflowunderthedynamictrafficmode1. ;1.2I/ofinput/output)FlowandCongestion ;Normally.congestionwillnotoccurinthesitua. ;tionwheretheinputflowislessthanoutputflow.The ;vehicleflowentersthesectionwithvelocityV,andits ;outputflowequalstheinputflow,sothepossibilityof ;congestionislimited.Becausetheinput/outputrestric—
;tionofurbantrafficismostlycausedbyoutput,weon. ;1yconsideroutputflowhere.Thetraveltimeofthedy. ;namictrafficnetworkisnotonlyrelatedtosection
;load.butalsorelatedtoinputandoutputflow.Itis ;notnecessarytodefineroutetrafficstatebyasingle ;section.RothrockandKeeferhaveresearchedtravel ;time.flowrelationbyobservation.Theyfoundthat ;smalltrafficflowsandlongtraveltimealsoconcur ;whentakingshortobservationintervals.Ifwedefine ;congestionastimeconsummation,wemaybuildthe ;followingfunction:
;Ts(t)=F.{Q(t)}+F:{I(t),X(t)}or
;Ts(t)=F{Q(t)}+F2{I(t))}+F3:X(t)),(2)
;whereF1denotessectiontraveltime,F,andF1denote ;intersectiondelays.
;Thethreeelementsarethecausesofcongestion. ;Sectiontravelisthedirectcauseandtheintersectionis ;theindirectcause.Wealsocananalyzethetrafficflow ;bydividingtraveltimeintotwoparts:sectionandin—
;tersectiontraveltimedrespectivelytoexpresstrafficdy. ;?
;236?
;namiccharacteristies.
;Theoccurrenceofcongestiononsectiontravelis ;relatedtotheheavytrafficnetwork.Whentheloadof ;roadtramcisuptothemaximumtolerance.congestion ;willhappen.Thecongestiondistributionoflocation ;andtimeonadayisbasicallyfixed.Congestioncaused ;byanintersectiondelayhasregularfactorsandstochas. ;ticfactors:theregulationisthesamewiththeregula—
;tionofthesection:therandomnessisbecauseofdense ;confictpoints.Ifthetrafficconflictcausessecond ;stopsoratrafficaccident,congestionmayhappen.If ;thelevelofintersectionmanagementisadvanced.this ;typeofcongestioncanbeavoidedandthedispersingof ;thecongestionwillbefaster.Thecongestioncausedby ;anintersectioncapacitydeficitwilllengthenthevehicle ;queue.andwillevencausetrafficflowtobackupmore ;andmore.Thistypeofcongestioncanonlybesolved ;bychangingtheroute.Therea1.timeinformationof ;trafficaccidentsandtramccongestionisusuallybroad—
;castthroughFMradioindomesticcities.Thesecities ;haveemphasizedtheadoptionofTRG(TrafficRoute ;Guiding)systemsandVMS(VaryingMessageSys—
;tems).
;1.3RouteSelectionandCongestion
;Trafficcongestionrevealstheactualtramcflow
;distribution.Aself-decisionandexperience--efiectar. ;chitectureareformedinthetripperwithoutthesystem ;strategyimpact:asthetrippercan‟treachthecoordi—
;nationstatewithoutthesystemprescription.Theva. ;chieverouteselectionbyself-decisionorexperience. ;effectaccordingtoadjacentnetworkload.Thismotion ;ismostusers‟decisionandkeepstothesecondWard.
;ropdiscipline.Whentoomanyusersselecttheroute. ;congestionoccurs.Sotheusers‟routeselectionshould
;bethekeycauseofformingcongestionandthemain ;factorthataffectsflowtime.spacedistribution. ;Wewouldliketoadopttheroutetimeasanele. ;mentinbuildingtrafficmodels.Theuser—optimizing
;modelhastwofoITI1S:
;1)InstantaneousUserOptimizingTime:Thetrip—
;perswhohavethesamedestinationselecttherouteac. ;cordingtotheshortesttimeaftertheirstart.andthisis ;calledInstantaneousUserOptimizingstate.Wecana1. ;soexplainthistobethattheusersselecttheirroutes ;accordingtoInstantaneousnetworkload.
;2)ActualUserOptimizingTime:Thetripperswho ;havethesamedestinationselecttherouteaccordingto ;theshortesttimebeforetheyactuallyarriveattheir ;destination.Wedefinethetimeasactualuseroptimi. ;zingtime.
;Actually,mosttrippers‟decisionoftheirroute
;selectionisaccordingtothefirsttimestandard.Wea1. ;socansupposethatthetrippers‟decisionfitsforthe
;secondtimestandard.Butasweconsidertheeffectof ;thedifferenceofthetrippers‟estimatedtime.itisran.
;
;JournalofHarbinInstituteofTechnology(NewSeries),Vo1.12,No.3,2005
;domorwecallitstochasticdistributioninmodeling.In ;consideringthis.wemayusethestochasticmodelin ;ouranalysis.
;Asthenetworklcadisnorma1.thedifferencebe—
;tweenthetwotypesoftimeshouldbesmal1.Otherwise ;thedifferencewillbecomedistinctasthenetworklcad ;becomesheavyorabnorma1.Itisbecausethisdistinct ;differencecausesthenetworkcongestion.Weusethe ;assignmentmodeltoillustratethetramcnetworklcad: ;rainf(Q,,,x)dw,|,x,e.EJFTd,?Jo
;T:y(Q,,,)denotesstochastictraveltimeestimation ;forODpairwhentakingrouter.Generallytheestima—
;tionvaluehasanexpectation,andifwecandecidethe ;routeselectionrate,thentheODpairusingrouterat ;timetissettledby:
;(t)=f(t)?P(t),(4)
;whereZ(t)isanODpairthatselectsrouterattimet. ;p~r(t)istheselectionprobabilityofrouterbythe ;same0Dattimet.
;po(t)asthevalueisconfirmed.theotheranaly—
;siselementscanbecalculatedinthesamewaywiththe ;determinateuseroptimizingassignmentmode1. ;2ApplicationofDynamicTrafficAssignmentin ;TrafficCongestionCountermeasure
;Besidesusingtheapplicationinanalyzingtraffic ;congestionasthemainmethodofanalyzingdynamic ;tr蛳cnetwork.thedynamictramcassignmentroodelis ;alsoresearchedintheestablishmentofcongestion ;countermeasures.
;Thecountermeasureofanurbantrafficnetworkis ;dividedintotwotypes:theresponsemodelandthe ;forecastmode1.Thesetwomodelsareusuallycombined ;inurbantrafficnetworks.Nextwewilldefinethesetwo ;modelsanddiscussthedynamicassignmentmode1. ;2.1ResponseCongestionCountermeasurewith ;CongestionChargeandDSOAssignment
;Theresponsecongestioncountermeasurelooksat ;theareathathashighfrequentcongestionorhasheavv ;congestion.Thistypeofcountermeasureiswidelva—
;dopted.Themainpracticaltechniqueofcountermeas—
;ureiscongestioncharge？.Thistechniqueisbasedon
;trafficmanagement,dependingonthedegreeofcon—
;gestion,expectingtooptimizethetrafficsysteminthe ;directlyimpactedarea.whosemaintheorybasisisthe ;dynamicsystemoptimumassignmentmode1.
;Sincecostisusedasafactortoadjustthedistri—
;butionoftramcflow.theobjectfunctionmustbetrans—
;formedintoacostform,namelytotransformtraveltime ;ordelayintocost.Aftertakingthesystemoptimumas—
;signment,wecanacquirethetramccostintheopti—
;mizedsystem.Referringtothecostinthecongestion ;system,wecandeterminethetaxrate.TheobjectiSto ;reducefrequentfacilityusagethroughtaxandtoadjust ;thesystemstate.Thechargerateshouhlbechanged ;accordingtotime,soitisnecessarytousethedynamic ;assignmentmode1.
;Themathematiccostformulaforoptimizingthe ;objectfunctionconsideringtaxiSlistedbelow: ;Cs(t)=s+[Ts(t)+(t)],(5)
;Hereo/isafixedcost,melythecostofsectionsunder ;thesituationoffreeflow;isthetime(?()sttransform ;rate,whichisindependentoftime;T(t)isthediffer—
;enceofinstantaneoususeroptimumassignmentand ;systemoptimumassignment,sodynamicsystemopti—
;mumassignmentiscarrie(1outtwice.Thefirstistode—
;termines(t)throughthesystemassignment,these(.一
;ondistodeterminethesmallesttotalcostthroughthe ;dynamicsystemoptimumtrafficassignment,all(]use ;thesetwotodeterminetheaveragetaxrate. ;2.2ForecastingCongestionCountermeasurewith ;TrafficGuidingandForecastingDUoAs?
;signment
;Themoreemcientmanagementmethodtosolxe ;citytrafficcongestionisadoptingaDriverRouteGuM—
;ingSystem(DRG).DRGisthroughobservingand ;forecastingthetimeandthelocationofcongestion【)【?一
;currence,offeringtheoptimumrouteandtriptimeto ;thetripper,whoseobjectistodispersecongestionand ;toimprovetheefficiencyofthetraf6s,rstem..DRG ;isanadvancedtramcmanagementmethodan(1the ;mainpartofATMS/ATIS.Trafficguiding.especially ;theguidingofspecialvehiclesisadoptedwidelyallo—
;vertheworld.FMforecastingisitsprima~status.an(1 ;moreadvancedstatusesarebasedonGPS.vehi(.1ecar—
;riedsystemsandelectronicinformation.whichiswidP—
;lyusedinciviltrafficsystemsabroad.TakingthPad—
;vantageoftrafficguiding,manyvehi(„lemanufa(?turers
;andtransportationorganizationsstrengthentheirre—
;searchprograms.suchastheJapaneseToyotaandUS ;GMcorporations.
;Theforecastingcongestioncountermeasureisfo—
;cusedonreal—timeinformationcollectionanddisposa1. ;forecastingthetramcflowdistributionintervals.and ;determiningthelocationandtimethatcongestionwill ;occur.Moreover,throughoptimumassignmentofthe ;usersactualtraveltime,suggestingtheusersadopted ;routeandacquiringtheminimumtraveltime.this ;countermeasureisconsideredasanloreadvan(?e(1con—
;gestioncountermeasuremode1.an(1t?andispersethe ;congestion.Becausetraveltimeisatorecastedindex.
;theforecastingDUOassignmentmodelisadopted. ;Thetraveltimeoffi)recastingtheDU0assignment ;isf1)recastedorusestheactua1trave1time.whichis ;differentfromtheDSOassignment.Inthefore(„dsting
;DUOassignmentmode1.supposingthattheUSCl‟Sae—
;tualtraveltimeonaselectedrouteismat(hedtothe ;?
;237?
;
;JournalofHarbinInstituteTechnology(NewSeries),Vo1.12,No.3,2005
;forecastedtime.notcurrenttrafficdistribution,moreo—
;vertheuserscanacquireeachminimumtraveltime ;throughacorrespondingtriproute.Theforecastedmin—
;imumtraveltimemayberevealedwiththerecursivee—
;quationbelow:
;(t)=(t)+叼[t+(t)],(6)
;where(t)denotesactualtraveltimefromstartpoint ;tosectioniattimet:l【l‟(t)denotesactualtravel
;timefromstartpointtosection，一lattimet;77lt+
;“(t)]isactualtraveltimeonsectiona.
;Theobjectofrouteguidingistomakealltrippers ;achievetheactualDUOstatus.Theprincipleofroute ;selectionalsocanbedefinedasbelow:ToallODpairs ;thathavethesamedestination,onanytime,thetravel ;timeequalstotheshortestpredictedtimefornodesa—
;longtheselectedroute.Wecallthistypeofflowdistri—
;butionasadynamicuseroptimizingstatebasedonroad ;section.Thedefinitionisthemainprincipleofdynamic ;trafficguiding.
;3ExperimentofSmallRoadNetwork
;Asmallroadnetwork
;,
;isdesignedwithtwelveroad
;sectionsandninenodes(asshowninFig.2).The ;dynamictrafficassignmentoftwoOD‟sonseven
;periodsoftimeislisted(asshowninTab.1).Inthe ;assignment,itisassumedthattheinitialtrafficloadof ;eachroadsectioniszero.all(1therearefun(?tionsbe—
;tweenthetraveltimeandtheinputandoutputofthe ;trafficvolume.
;Fig.2Smallexperimentnetwork
;Tab.1DynamicODtableofeachperiodoftime ;Theassignmentresultissh0wninTat).2.It【anbe
;seenfromthetablethat<lynamictraffi<?assignnientisrea—
;sonableanditisadvantageousforobtainingakey. ;Tab.2RouteassignmentresultsofinstantaneousDUO ;4Conclusion
;[3:
;Asweallknow,thecausesoftrafficcongestionr.] ;arenotlimitedtothefactorswehavediscussed.0ther ;causesaredeficientparkingspacesandpedestrian ;walkways.aswellasv