ebook img

Continuity of the path delay operator for dynamic network loading with spillback PDF

1 MB·English
by  Ke Han
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Continuity of the path delay operator for dynamic network loading with spillback

ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009. Continuity of the path delay operator for dynamic network loading with spillback KeHana,∗,BenedettoPiccolib,TerryL.Frieszc aDepartmentofCivilandEnvironmentalEngineering,ImperialCollegeLondon,UnitedKingdom. 6 bDepartmentofMathematicalSciencesandCCIB,RutgersUniversity-Camden,USA 1 cDepartmentofIndustrialandManufacturingEngineering,PennsylvaniaStateUniversity,USA. 0 2 r a M Abstract 8 Thispaperestablishesthecontinuityofthepathdelayoperatorsfordynamicnetworkloading(DNL)problemsbased 1 on the Lighthill-Whitham-Richards model, which explicitly capture vehicle spillback. The DNL describes and pre- dicts the spatial-temporal evolution of traffic flow and congestion on a network that is consistent with established ] P routeanddeparturetimechoicesoftravelers. TheLWR-basedDNLmodelisfirstformulatedasasystemofpartial A differentialalgebraicequations(PDAEs). Wetheninvestigatethecontinuousdependenceofmergeanddivergejunc- . tionmodelswithrespecttotheirinitial/boundaryconditions,whichleadstothecontinuityofthepathdelayoperator h throughthewave-fronttrackingmethodologyandthegeneralizedtangentvectortechnique. Aspartofouranalysis t a leadinguptothemaincontinuityresult,wealsoprovideanestimationoftheminimumnetworksupplywithoutresort m toanynumericalcomputation. Inparticular,itisshownthatgridlockcanneveroccurinafinitetimehorizoninthe [ DNLmodel. 2 Keywords: pathdelayoperator,continuity,dynamicnetworkloading,LWRmodel,spillback,gridlock v 1 4 2 4 1. Introduction 0 1. Dynamictrafficassignment(DTA)isthedescriptivemodelingoftime-varyingflowsontrafficnetworksconsistent 0 with traffic flow theory and travel choice principles. DTA models describe and predict departure rates, departure 5 times and route choices of travelers over a given time horizon. It seeks to describe the dynamic evolution of traffic 1 in networks in a fashion consistent with the fundamental notions of traffic flow and travel demand; see Peeta and : v Ziliaskopoulos(2001)forsomereviewonDTAmodelsandrecentdevelopments. Dynamicuserequilibrium(DUE) i oftheopen-looptype,whichisonetypeofDTA,remainsamajormodernperspectiveontrafficmodelingthatenjoys X widescholarlysupport. Itcapturestwoaspectsofdrivingbehaviorquitewell: departuretimechoiceandroutechoice r a (Friesz et al., 1993). Within the DUE model, travel cost for the same trip purpose isidentical for all utilized route- and-departure-time choices. The relevant notion of travel cost is effective travel delay, which is a weighted sum of actualtraveltimeandarrivalpenalties. In the last two decades there have been many efforts to develop a theoretically sound formulation of dynamic networkuserequilibriumthatisalsoacanonicalformacceptabletoscholarsandpractitionersalike. Therearetwo essentialcomponentswithintheDUEmodels: (1)themathematicalexpressionofNash-likeequilibriumconditions; ∗Correspondingauthor Emailaddresses:[email protected](KeHan),[email protected](BenedettoPiccoli),[email protected](TerryL. Friesz) PreprintsubmittedtoTransportationResearchPartB March22,2016 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 and (2) a network performance model, which is, in effect, an embedded dynamic network loading (DNL) problem. The DNL model captures the relationships among link entry flow, link exit flow, link delay and path delay for any given set of path departure rates. The DNL gives rise to the notion of path delay operator, which is viewed as a mapping from the set of feasible path departure rates to the set of path travel times or, more generally, path travel costs. PropertiesofthepathdelayoperatorareoffundamentalimportancetoDUEmodels. Inparticular,continuityof the delay operators plays a key role in the existence and computation of DUE models. The existence of DUEs is typically established by converting the problem to an equivalent mathematical form and applying some version of Brouwer’sfixed-pointexistencetheorem; examplesincludeHanetal.(2013c);SmithandWisten(1995);Wieetal. (2002)andZhuandMarcotte(2000). Alloftheseexistencetheoriesrelyonthecontinuityofthepathdelayoperator. OnthecomputationalsideofanalyticalDUEmodels,everyestablishedalgorithmrequiresthecontinuityofthedelay operatortoensureconvergence;anincompletelistofsuchalgorithmsincludethefixed-pointalgorithm(Frieszetal., 2013),theroute-swappingalgorithm(HuangandLam,2002),thedescentmethod(HanandLo,2003),theprojection method(HanandLo,2002;Ukkusurietal.,2012),andtheproximalpointmethod(Hanetal.,2015a) Ithasbeendifficulthistoricallytoshowcontinuityofthedelayoperatorforgeneralnetworktopologiesandtraffic flow models. Over the past decade, only a few continuity results were established for some specific traffic flow models. Zhu and Marcotte (2000) use the link delay model (Friesz et al., 1993) to show the continuity of the path delay operator. Their result relies on the a priori boundedness of the path departure rates, and is later improved by a continuity result that is free of such an assumption (Han et al., 2012). In Bressan and Han (2013), continuity of thedelayoperatorisshownfornetworkswhosedynamicsaredescribedbytheLWR-Laxmodel(BressanandHan, 2011;Frieszetal.,2013),whichisavariationoftheLWRmodelthatdoesnotcapturevehiclespillback. Theirresult also relies on the a priori boundedness of path departure rates. Han et al. (2013c) consider Vickrey’s point queue model(Vickrey,1969)andshowthecontinuityofthecorrespondingpathdelayoperatorforgeneralnetworkswithout invokingtheboundednessonthepathdeparturerates. All of these aforementioned results are established for network performance models that do not capture vehicle spillback. To the best of our knowledge, there has not been any rigorous proof of the continuity result for DNL models that allow queue spillback to be explicitly captured. On the contrary, some existing studies even show that thepathtraveltimesmaydependdiscontinuouslyonthepathdeparturerates,whenphysicalqueuemodelsareused. Forexample,Szeto(2003)usesthecelltransmissionmodelandsignalcontroltoshowthatthepathtraveltimemay depend on the path departure rates in a discontinuous way. Such a finding suggests that the continuity of the delay operatorcouldverywellfailwhenspillbackispresent. Thishasbeenthemajorhurdleinshowingthecontinuityor identifyingrelevantconditionsunderwhichthecontinuityisguaranteed. Thispaperbridgesthisgapbyarticulating theseconditionsandprovidingaccompanyingproofofcontinuity. Thispaperpresents,forthefirsttime,arigorouscontinuityresultforthepathdelayoperatorbasedontheLWR network model, which explicitly captures physical queues and vehicle spillback. In showing the desired continuity, weproposeasystematicapproachforanalyzingthewell-posednessoftwospecificjunctionmodels1: amergeand a diverge model, both originally presented by Daganzo (1995). The underpinning analytical framework employs thewave-fronttrackingmethodology(Dafermos,1972;HoldenandRisebro,2002)andthetechniqueofgeneralized tangent vectors (Bressan, 1993; Bressan et al., 2000). A major portion of our proof involves the analysis of the interactionsbetweenkinematicwavesandthejunctions,whichisfrequentlyinvokedforthestudyofwell-posedness of junction models; see Garavello and Piccoli (2006) for more details. Such analysis is further complicated by the factthatvehicleturningratiosatadivergejunctionaredeterminedendogenouslybydrivers’routechoiceswithinthe DNLprocedure. Asaresult,specialtoolsaredevelopedinthispapertohandlethisuniquesituation. Asweshalllatersee,acrucialstepoftheprocessaboveistoestimateandboundfrombelowtheminimumnetwork supply,whichisdefinedintermsoflocalvehicledensitiesinthesamewayasinLebacqueandKhoshyaran(1999). Infact,ifthesupplysomewheretendstozero(thatis,whentrafficapproachesthejamdensity),thewell-posednessof thedivergejunctionmayfail,aswedemonstrateinSection4.2.1. Thishasalsobeenconfirmedbytheearlierstudy ofSzeto(2003), whereawaveofjamdensityistriggeredbyasignalredlightandcausesspillbackattheupstream junction,leadingtoajumpinthepathtraveltimes.Remarkably,inthispaperweareabletoshowthat(1)ifthesupply 1Well-posednessofamodelreferstothepropertythatthebehaviorofthesolutionhardlychangeswhenthereisaslightchangeintheini- tial/boundaryconditions. 2 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 isboundedawayfromzero(thatis,trafficisboundedawayfromthejamdensity),thenthedivergejunctionmodelis wellposed;and(2)thedesiredboundednessofthesupplyisanaturalconsequenceofthedynamicnetworkloading procedurethatinvolvesonlythemergeanddivergejunctionmodelswestudyhere. Thisisahighlynon-trivialresult becauseitnotonlyplaysaroleinthecontinuityproof,butalsoimpliesthatgridlockcanneveroccurinthenetwork loadingprocedureinanyfinitetimehorizon. ThefinalcontinuityresultispresentedinSection5.4,followinganumberofpreliminaryresultssetoutinprevious sections.Althoughourcontinuityresultisestablishedonlyfornetworksconsistingofsimplemergeanddivergenodes, itcanbeextendedtonetworkswithmorecomplextopologiesusingtheprocedureofdecomposinganyjunctioninto several simple merge and diverge nodes (Daganzo, 1995). Moreover, the analytical framework employed by this papercanbeinvokedtotreatotherandmoregeneraljunctiontopologiesand/ormerginganddivergingrules,andthe techniquesemployedtoanalyzewaveinteractionswillremainvalid. Themaincontributionsmadeinthispaperinclude: • formulation of the LWR-based dynamic network loading (DNL) model with spillback as a system of partial differentialalgebraicequations(PDAEs); • acontinuityresultforthepathdelayoperatorbasedontheaforementionedDNLmodel; • a novel method for estimating the network supply, which shows that gridlock can never occur within a finite timehorizon. Therestofthispaperisorganizedasfollows. Section2recapssomeessentialknowledgeandnotionsregarding theLWRnetworkmodelandtheDNLprocedure. Section3articulatesthemathematicalcontentsoftheDNLmodel byformulatingitasaPDAEsystem. Section4introducesthemergeanddivergemodelsandestablishestheirwell- posedness. Section 5 provides a final proof of continuity and some discussions. Finally, we offer some concluding remarksinSection6. 2. LWR-baseddynamicnetworkloading 2.1. Delayoperatoranddynamicnetworkloading Throughout this paper, the time interval of analysis is a single commuting period expressed as [0, T] for some T > 0. WeletPbethesetofallpathsemployedbytravelers. Foreachpath p ∈ Pwedefinethepathdeparturerate whichisafunctionofdeparturetimet∈[0, T]: hp(·): [0, T] → R+ whereR+denotesthesetofnon-negativerealnumbers. Eachpathdepartureratehp(t)isinterpretedasatime-varying pathflowmeasuredattheentranceofthefirstarcoftherelevantpath,andtheunitforh (t)isvehiclesperunittime. p We next define h(·) = {h (·) : p ∈ P} to be a vector of departure rates. Therefore, h = h(·) can be viewed as a p vector-valuedfunctionoft,thedeparturetime2. Thesinglemostcrucialingredienttisthepathdelayoperator,whichmapsagivenvectorofdeparturerateshtoa vectorofpathtraveltimes. Morespecifically,welet D (t, h) ∀t∈[t , t ], ∀p∈P p 0 f bethepathtraveltimeofadriverdepartingattimetandfollowingpath p,giventhedeparturerateshassociatedwith allthepathsinthenetwork. WethendefinethepathdelayoperatorD(·)bylettingD(h) = {D (·, h) : p ∈ P},which p isavectorconsistingoftime-dependentpathtraveltimesD (t, h). p 2Fornotationconvenienceandwithoutcausinganyconfusion,wewillsometimesusehinsteadofh(·)todenotepathflowvectors. 3 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 2.2. TheLighthill-Whitham-Richardsmodelonnetworks WerecapthenetworkextensionoftheLWRmodel(LighthillandWhitham,1955;Richards,1956),whichcaptures theformation,propagation,anddissipationofspatialqueuesandvehiclespillback. Discussionprovidedbelowrelies ongeneralassumptionsonthefundamentaldiagramandthejunctionmodel,andinvolvesnoadhoctreatmentofflow propagation,flowconservation,linkdelay,orothercriticalmodelfeatures. We consider a road link expressed as a spatial interval [a, b] ⊂ R. The partial differential equation (PDE) representationoftheLWRmodelisthefollowingscalarconservationlaw ∂ρ(t, x)+∂ f(cid:0)ρ(t, x)(cid:1) = 0 (t, x)∈[0, T]×[a, b] (2.1) t x withappropriateinitialandboundaryconditions,whichwillbediscussedindetaillater. Here,ρ(t, x)denotesvehicle density at a given point in the space-time domain. The fundamental diagram f(·) : [0, ρjam] → [0, C] expresses vehicle flow at (t, x) as a function of ρ(t, x), where ρjam denotes the jam density, andC denotes the flow capacity. Throughoutthispaper,weimposethefollowingmildassumptionon f(·): (F).Thefundamentaldiagram f(·)iscontinuous,concaveandvanishesatρ=0andρ=ρjam. AnessentialcomponentofthenetworkextensionoftheLWRmodelisthespecificationofboundaryconditions ataroadjunction. Derivationoftheboundaryconditionsshouldnotonlyobeyphysicalrealism,suchasthatenforced by entropy conditions (Garavello and Piccoli, 2006; Holden and Risebro, 1995), but also reflect certain behavioral and operational considerations, such as vehicle turning preferences (Daganzo, 1995), driving priorities (Coclite et al., 2005), and signal controls (Han et al., 2014). Articulation of a junction model is facilitated by the notion of Riemann Problem, which is an initial value problem at the junction with constant initial condition on each incident link. ThereexistanumberofjunctionmodelsthatyielddifferentsolutionsofthesameRiemannProblem. Inoneline of research, an entropy condition is defined based on a minimization/maximization problem (Holden and Risebro, 1995). In another line of research, the boundary conditions are defined using link demand and supply (Lebacque andKhoshyaran,1999),whichrepresentthelink’ssendingandreceivingcapacities. Modelsfollowingthisapproach includeDaganzo(1995);JinandZhang(2003)andJin(2010). ThesolutionofaRiemannProblemisgivenbythe RiemannSolver(RS),tobedefinedbelow. 2.2.1. TheRiemannSolver WeconsiderageneralroadjunctionJwithmincomingroadsandnoutgoingroads,asshowninFigure1. m m+n . . . . . . J . . . . . . 2 m+2 1 m+1 Figure1:Aroadjunctionwithmincominglinksandnoutgoinglinks. We denote by I1,...,Im the incoming links and by Im+1,...,Im+n the outgoing links. In addition, for every i ∈ {1,...,m+n},thedynamiconI isgovernedbytheLWRmodel i ∂ρ(t, x)+∂ f(cid:0)ρ(t, x)(cid:1) = 0 (t, x)∈[0, T]×[a, b] (2.2) t i x i i i i wherelinkI isexpressedasthespatialinterval[a, b],andwealwaysusethesubscript‘i’toindicatedependenceon i i i thelinkI. Theinitialconditionforthisconservationlawis i ρ(0, x) = ρˆ (x) x∈[a, b] (2.3) i i i i 4 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 Noticethattheabove(m+n)initialvalueproblemsarecoupledtogetherviatheboundaryconditionstobespecified atthejunction. SuchasystemofcouplingequationsiscommonlyanalyzedusingtheRiemannProblem. Definition2.1. (RiemannProblem)TheRiemannProblematthejunctionJisdefinedtobeaninitialvalueproblem onanetworkconsistingofthesinglejunction J withmincominglinksandnoutgoinglinks,allextendingtoinfinity, suchthattheinitialdensitiesareconstantsoneachlink:  ρi(0, x) ≡ ρˆi x∈(−∞, bi], i∈{1, ..., m} ρ (0, x) ≡ ρˆ x∈[a , +∞), j∈{m+1, ..., m+n} j j j whereρˆ ∈[0, ρjam]areconstants,k=1,...,m+n. k k A Riemann Solver for the junction J is a mapping that, given any (m+n)-tuple of Riemann initial conditions (cid:0)ρˆ1,...,ρˆm+n(cid:1), provides a unique (m + n)-tuple of boundary conditions (cid:0)ρ1,...,ρm+n(cid:1) such that one can solve the initial-boundary value problem for each link, and the resulting solutions constitute a weak entropy solution of the RiemannProblematthejunction3. AprecisedefinitionoftheRiemannSolverisgivenasfollows. Definition2.2. (RiemannSolver)ARiemannSolverforthejunctionJwithmincominglinksandnoutgoinglinksis amapping (cid:89)m+n(cid:104) (cid:105) (cid:89)m+n(cid:104) (cid:105) RS : 0, ρjam −→ 0, ρjam k k k=1 k=1 (cid:0) (cid:1) (cid:0) (cid:1) ρˆ1, ..., ρˆm+n (cid:55)→ ρ1, ..., ρm+n which relates Riemann initial data ρˆ = (cid:0)ρˆ1,...,ρˆm+n(cid:1) to boundary conditions ρ = (cid:0)ρ1,...,ρm+n(cid:1), such that the followinghold. (i) ThesolutionoftheRiemannProblemrestrictedoneachlinkI isgivenbythesolutionoftheinitial-boundary k valueproblemwithinitialconditionρˆ andboundaryconditionρ ,k=1,...,m+n. k k (ii) TheRankine-Hugoniotcondition(flowconservation)holds: (cid:88)m (cid:88)m+n f(cid:0)ρ(cid:1) = f (cid:0)ρ (cid:1) (2.4) i i j j i=1 j=m+1 (iii) Theconsistencyconditionholds: RS(cid:2)RS[ρˆ](cid:3) = RS[ρˆ] (2.5) ThreeconditionsmustbesatisfiedbytheRiemannSolver(RS).Item(i)aboverequiresthattheboundarycondition oneachlinkmustbeproperlygivensothattheinitialvalueproblemsnotonlyhavewell-definedsolutions,butthese solutionsmustalsobecompatibleandformasensiblesolutionatthejunction.(2.4)simplystipulatestheconservation offlowacrossthejunction. (2.5)isadesirablepropertyandissometimesreferredtoastheinvarianceproperty(Jin, 2010). Remark2.3. ForthesameRiemannProblem,thereexistmanyRiemannSolversthatsatisfyconditions(i)-(iii)above. Despite their varying forms, most existing Riemann Solvers rely on a flow maximization problem at the relevant junctionsubjecttoconstraintsrelatedtoturningratio,right-of-way,orsignalcontrols;see(Cocliteetal.,2005;Han etal.,2014;HoldenandRisebro,1995;JinandZhang,2003)andJin(2010). 3WereferthereadertoHoldenandRisebro(1995)foradefinitionofweakentropysolutionatajunction 5 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 2.2.2. Thelinkdemandandsupply ForeachlinkI,weletρc bethecriticaldensityatwhichtheflowismaximized. ThedemandD(t)forincoming i i i linksI andthesupplyS (t)foroutgoinglinksI aredefinedintermsofthedensityneartheexitandentranceofthe i j j link,respectively(LebacqueandKhoshyaran,1999):  Di(t) = Di(cid:16)ρi(t, bi−)(cid:17) = Cfi(cid:0)ρ(t, b−)(cid:1) iiff ρρi((tt,, bbi−−)) <≥ ρρcic (2.6) i i i i i i  Sj(t) = Sj(cid:16)ρj(t, aj+)(cid:17) = Cf j(cid:0)ρ (t, a +)(cid:1) iiff ρρj((tt,, aaj++)) ≤> ρρccj (2.7) j j j j j j Inprose,thedemandrepresentsthemaximumflowatwhichcarscanbedischargedfromtheincominglink;andthe supply represents the maximum flow at which cars can enter the outgoing link. Notice that the demand and supply are both expressed as functions of density, and they are always greater than or equal to the fundamental diagram f(·)or f (·); seeFigure2foranillustration. Inoursubsequentpresentation, withoutcausingconfusionwewilluse i j notationsD(t)andD(ρ)interchangeablywheretheformerindicatesthedemandasatime-varyingquantity,andthe i i latteremphasizesdemandasafunctionofdensity. Thesameappliestothesupply. f(!) D(!) S(!) ! 0 !c !jam Figure2:Demandandsupplyasfunctionsofdensity. 3. DynamicnetworkloadingproblemformulatedasaPDAEsystem TheaimofthissectionistoformulatetheLWR-baseddynamicnetworkloading(DNL)problemasasystemof partialdifferentialalgebraicequations(PDAEs). TheproposedPDAEsystemusesvehicledensityandqueuesasthe primary unknown variables, and computes link dynamics, flow propagation, and path delay for any given vector of departurerates. ThePDAEsystemcapturesvehiclespillbackexplicitly,andaccommodatesawiderangeofjunction typesandRiemannSolvers. WeconsideranetworkG(A, V)expressedasadirectedgraphwithAbeingthesetoflinksandVbeingtheset of nodes. Let P be the set of paths employed by travelers, and W be the set of origin-destination pairs. Each path p∈Pisexpressedasanorderedsetoflinksittraverses: p = {I , I , ...,I } 1 2 m(p) wherem(p)isthenumberoflinksinthispath. Thereareseveralcrucialcomponentsofacompletenetworkloading model,eachofwhichiselaboratedinasubsectionbelow. Throughouttherestofthispaper,foreachnodev∈V,we denotebyIvthesetofincominglinksandOvthesetofoutgoinglinks. 6 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 3.1. Within-linkdynamics ForeachI ∈A,thedensitydynamicisgovernedbythescalarconservationlaw i ∂ρ(t, x)+∂ (cid:2)ρ(t, x)·v(cid:0)ρ(t, x)(cid:1)(cid:3) = 0 (t, x)∈[0, T]×[a, b] (3.8) t i x i i i i i subject to initial condition and boundary conditions to be determined in Section 3.2. The fundamental diagram f(ρ(cid:1) = ρ ·v(cid:0)ρ(cid:1) satisfies condition (F) stated at the beginning of Section 2.2. In order to explicitly incorporate i i i i i drivers’routechoices, forevery p ∈ Psuchthat I ∈ pweintroducethefunctionµp(t, x), (t, x) ∈ [0, T]×[a, b], i i i i which represents, in every unit of flow f(ρ(t, x)), the fraction associated with path p. We call these variables path i i disaggregation variables (PDV). For each car moving along the link I, its surrounding traffic can be distinguished i by path (e.g. 20% following path p , 30% following path p , 50% following path p ). As this car moves, such a 1 2 3 compositionwillnotchangesinceitssurroundingtrafficallmoveatthesamespeedunderthefirst-in-first-out(FIFO) principle (i.e. no overtaking is allowed). In mathematical terms, this means that the path disaggregation variables, µp(·, ·),areconstantsalongthetrajectoriesofcars(t, x(t))inthespace-timediagram,where x(·)isthetrajectoryofa i movingcaronlinkI. Thatis, i dµp(cid:0)t, x(t)(cid:1) = 0 ∀p suchthat I ∈ p, dt i i which,accordingtothechainrule,becomes ∂µp(cid:0)t, x(t)(cid:1)+∂ µp(t, x)· d x(t) = 0, t i x i dt whichfurtherleadstoanothersetofpartialdifferentialequationsonlinkI: i ∂µp(t, x)+v(cid:0)ρ(t, x)(cid:1)·∂ µp(t, x) = 0 ∀p suchthat I ∈ p (3.9) t i i i x i i Here,ρ(t, x)isthesolutionof(3.8). Thefollowingobviousidentityholds i (cid:88) µp(t, x) = 1 whenever ρ(t, x) > 0 (3.10) i i p(cid:51)Ii where p (cid:51) I means“path pcontains(ortraverses)linkI”,andthesummationappearingin(3.10)iswithrespectto i i allsuch p. Byconvention,ifρ(t, x)=0,thenµp(t, x)=0forall p(cid:51) I. i i i 3.2. Boundaryconditionsatanordinarynode Forreasonthatwillbecomeclearlater,weintroducetheconceptofanordinarynode. Anordinarynodeisneither theoriginnorthedestinationofanytrip. WeusethenotationVotorepresentthesetofordinarynodesinthenetwork. As mentioned earlier, the partial differential equations on links incident to J are all coupled together through a given junction model, i.e., a Riemann Solver. A common prerequisite for applying the Riemann Solver is the determinationoftheflowdistribution(turningratio)matrix(Cocliteetal.,2005), whichreliesonknowledgeofthe PDVsµp(t, b)forallI ∈IJ. Wedefinethetime-dependentflowdistributionmatrixassociatedwithJ: i i i AJ(t) = (cid:8)αJ(t)(cid:9)∈[0, 1]|IJ|+|OJ| (3.11) ij where by convention, we use subscript i to indicate incoming links, and j to indicate outgoing links. Each element αJ(t)representstheturningratiosofcarsdischargedfromI thatenterdownstreamlinkI .Then,forallpthattraverses ij i j J,thefollowingholds. (cid:88) αJ(t)= µp(t, b) ∀I ∈IJ, I ∈OJ (3.12) ij i i i j p(cid:51)Ii,Ij (cid:80) ItcanbeeasilyverifiedthatαJ(t)∈[0, 1]and αJ(t)≡1accordingto(3.10). ij j ij WearenowreadytoexpresstheboundaryconditionsfortheordinaryjunctionJ ∈Vo. Let |I(cid:89)J|+|OJ| |I(cid:89)J|+|OJ| RSAJ : [0, ρjam]→ [0, ρjam] k k k=1 k=1 7 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 be a given Riemann Solver. Notice that the dependence of the Riemann Solver on AJ has been indicated with a superscript. TheboundaryconditionsforPDEs(3.8)read ρ (t, b ) = RSAJ(cid:104)(cid:0)ρ(t, b−)(cid:1) , (cid:0)ρ (t, a +)(cid:1) (cid:105) ∀I ∈IJ (3.13) k k k i i Ii∈IJ j j Ij∈OJ k ρ(t, a) = RSAJ(cid:104)(cid:0)ρ(t, b−)(cid:1) , (cid:0)ρ (t, a +)(cid:1) (cid:105) ∀I ∈OJ (3.14) l l l i i Ii∈IJ j j Ij∈OJ l whereRSAJ[·]denotesthek-thcomponentofthemapping,k=1, ..., |IJ|+|OJ|. k Remark3.1. Intuitively,(3.13)-(3.14)meanthat,giventhecurrenttrafficstates(cid:0)ρ(t, b−)(cid:1) and(cid:0)ρ (t, a +)(cid:1) i i Ii∈IJ j j Ij∈OJ adjacenttothejunctionJ,theRiemannSolverRSAJ specifies,foreachincidentlinkI orI,thecorrespondingbound- k l aryconditionsρ (t, b )orρ(t, a). Inprose,ateachtimeinstancetheRiemannSolverinspectsthetrafficconditions k k l l near the junction, and proposes the discharging (receiving) flows of its incoming (outgoing) links. Such a process isbasedontheflowdistributionmatrix AJ andoftenreflectstrafficcontrolmeasuresatjunctions. Furthermore,the Riemann Solver operates with knowledge of every link incident to the junction, thus the boundary condition of any relevantlinkisdeterminedjointlybyallthelinksconnectedtothesamejunction. Therefore, theLWRequationson all the links are coupled together through this mechanism. For this reason the LWR-based DNL models are highly challenging,boththeoreticallyandcomputationally. TheupstreamboundaryconditionsassociatedwithPDEs(3.9)are: f(cid:0)ρ(t, b)(cid:1)·µp(t, b) µpj(t, aj) = i if (cid:0)ρi(t, ai)(cid:1) i ∀p suchthat {Ii, Ij}⊂ p, ∀Ij ∈OJ (3.15) j j j where the numerator f(cid:0)ρ(t, b)(cid:1)·µp(t, b) expresses the exit flow on link I associated with path p, which, by flow i i i i i i conservation,isequaltotheenteringflowoflinkI associatedwiththesamepath p;thedenominatorrepresentsthe j totalenteringflowoflinkI . j Remark3.2. Unlikethedensity-basedPDE,thePDV-basedPDEdoesnothaveanydownstreamboundarycondition due to the fact that the traveling speeds of the PDVs are the same as the car speeds (they can be interpreted as Lagrangianlabelsthattravelwiththecars);thusinformationregardingthePDVsdoesnotpropagatebackwardsor spillsovertoupstreamlinks. 3.3. Flowdistributionatoriginordestinationnodes Weconsideranodev ∈ Vthatiseithertheoriginorthedestinationofsomepath p. Oneimmediateobservation isthattheflowconservationconstraint(2.4)nolongerholdsatsuchanodesincevehicleseitherare‘generated’(ifv isanorigin)or‘vanish’(ifvisadestination). Asimpleandeffectivewaytocircumventthisissueistointroducea virtuallink. Avirtuallinkisanimaginaryroadwithcertainlengthandfundamentaldiagram,andservesasabuffer betweenanordinarynodeandanorigin/destination; seeFigure3foranillustration. Byintroducingvirtuallinksto theoriginalnetwork,weobtainanaugmentednetworkG(A˜, V˜)inwhichallroadjunctionsareordinary,andhence fallwithinthescopeoftheprevioussection. s t J J Figure3:Illustrationofthevirtuallinks. Left:avirtuallinkconnectinganorigin(s)toanordinaryjunctionJ. Right:avirtuallinkconnectinga destination(t)toanordinaryjunctionJ. LetusdenotebySthesetoforiginsintheaugmentednetworkG(A˜, V˜). Foranys∈S,wedenotebyPs ⊂Pthe setofpathsthatoriginatefroms,andbyI thevirtuallinkincidenttothisorigin. Foreach p∈Pswedenotebyh (t) s p 8 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 thedeparturerate(pathflow)along p. Itisexpectedthatabuffer(point)queuemayformat sincasethereceiving capacityofthedownstreamI isinsufficienttoaccommodateallthedeparturerates(cid:80) h (t). Forthisbufferqueue, s p∈Ps p denotedq (t),weemployaVickrey-typedynamic(Vickrey,1969);thatis, s  ddtqs(t) = p(cid:88)∈Pshp(t)−Smsi(nt)(cid:110)p(cid:88)∈Pshp(t), Ss(t)(cid:111) iiff qqss((tt)) >= 00 (3.16) whereS (t)denotesthesupplyofthevirtuallink I . Theonlydifferencebetween(3.16)andVickrey’smodelisthe s s time-varyingdownstreamreceivingcapacityprovidedbythevirtuallink. 4 It remains to determine the dynamics for the path disaggregation variables (PDV). More precisely, we need to determine µp(t, a ) for the virtual link I where p ∈ Ps, and x = a is the upstream boundary of I . This will be s s s s s achievedusingtheVickrey-typedynamic(3.16)andtheFIFOprinciple. Specifically,wedefinethequeueexittime function λ (t) where t denotes the time at which drivers depart and join the point queue, if any; λ (t) expresses the s s timeatwhichthesamegroupofdriversexitthepointqueue. Clearly,FIFOdictatesthat (cid:90) t (cid:88)h (τ)dτ = (cid:90) λs(t) f (cid:0)ρ (τ, a )(cid:1)dτ (3.17) p s s s 0 p∈Ps 0 wherethetwointegrandsontheleftandrighthandsidesoftheequationareflowenteringthequeueandflowleaving thequeue,respectively. Wemaydeterminethepathdisaggregationvariablesas: µsp(cid:0)λs(t), as(cid:1) = (cid:80) hp(ht)(t) ∀p∈Ps (3.18) q∈Ps q Noticethat,if(cid:80) h (t)=0,thentheflowleavingthepointqueueattimeλ (t)isalsozero;thusthereisnoneedto q∈Ps q s determinethepathdisaggregationvariables. Therefore,theidentity(3.18)iswelldefinedandmeaningful. 3.4. Calculationofpathtraveltimes Withallprecedingdiscussions,wemayfinallyexpressthepathtraveltimes,whicharetheoutputsofacomplete DNLmodel.Thepathtraveltimeconsistsoflinktraveltimespluspossiblequeuingtimeattheorigin.Mathematically, thelinkexittimefunctionλ(t)foranyI isdefined,inawaysimilarto(3.17),as i i (cid:90) t f(cid:0)ρ(τ, a)(cid:1)dτ = (cid:90) λi(t) f(cid:0)ρ(τ, b)(cid:1)dτ (3.19) i i i i i i 0 0 Forapathexpressedas p={I , I , ..., I },thetimetotraverseitiscalculatedas 1 2 m(p) λ ◦λ ◦λ ...◦λ (t) (3.20) s 1 2 m(p) . where f ◦g(t) = g(f(t))meansthecompositionoftwofunctions. Thisisduetotheassumptionthatcarsleavingthe previouslink(orqueue)immediatelyenterthenextlinkwithoutanydelay. 3.5. ThePDAEsystem WearenowreadytopresentagenericPDAEsystemforthedynamicnetworkloadingprocedure. Letusbeginby summarizingsomekeynotations. 4Therighthandsideoftheordinarydifferentialequation(3.16)isdiscontinuous.Ananalyticaltreatmentofthisirregularequationisprovided byHanetal.(2013a,b)usingthevariationalformulation. 9 ARTICLELINK:http://www.sciencedirect.com/science/article/pii/S0191261515002039 PLEASECITETHISARTICLEAS Han,K.,Piccoli,B.,Friesz,T.L.,2015. Continuityofthepathdelayoperatorfordynamicnetworkloadingwith spillback. TransportationResearchPartB,DOI:10.1016/j.trb.2015.09.009 G(A, V) theoriginalnetworkwithlinksetAandnodesetV; VL thesetofvirtuallinks; G(A˜, V˜) theaugmentednetworkincludingvirtuallinks; S thesetoforiginsinG(A˜, V˜); Ps thesetofpathsoriginatingfroms∈S; Vo thesetofordinaryjunctionsinG(A˜, V˜); IJ thesetofincominglinksofajunctionJ ∈Vo; OJ thesetofoutgoinglinksofajunctionJ ∈Vo; AJ(t) theflowdistributionmatrixassociatedwithjunctionJ; RSAJ theRiemannSolverforjunctionJ,whichdependsonAJ. WealsolistsomekeyvariablesofthePDAEsystembelow. h (t) thepathdepartureratealong p∈P; p ρ(t, x) thevehicledensityonlinkI ∈A˜; i i µp(t, x) theproportionofflowonlinkI associatedwithpath p(pathdisaggregationvariable); i i q (t) thepointqueueattheorigins∈S; s λ (t) thepointqueueexittimefunctionatorigins∈S. s Givenanyvectorofpathdepartureratesh = (cid:0)h (·) : p ∈ P),theproposedPDAEsystemforcalculatingpathtravel p timesissummarizedasfollows.  dqdst(t) = p(cid:88)∈Pshp(t)−Smsi(nt)(cid:110)p(cid:88)∈Pshp(t), Ss(t)(cid:111) qqss((tt))>=00 ∀s∈S (3.21) (cid:90) t (cid:88)h (τ)dτ = (cid:90) λs(t) f (cid:0)ρ (τ, a )(cid:1)dτ ∀s∈S (3.22) p s s s 0 p∈Ps 0 (cid:90) t f(cid:0)ρ(τ, a)(cid:1)dτ = (cid:90) λi(t) f(cid:0)ρ(τ, b)(cid:1)dτ ∀I ∈A˜ (3.23) i i i i i i i 0 0 ∂ρ(t, x)+∂ (cid:2)ρ(t, x)·v(cid:0)ρ(t, x)(cid:1)(cid:3)=0 (t, x)∈[0, T]×[a, b] (3.24) t i x i i i i i ∂µp(cid:0)t, x(cid:1)+v(cid:0)ρ(t, x)(cid:1)·∂ µp(cid:0)t, x(cid:1)=0 (t, x)∈[0, T]×[a, b] (3.25) t i i i x i i i µsp(cid:0)λs(t), as(cid:1) = (cid:80) hp(ht)(t) ∀s∈S, p∈Ps (3.26) q∈Ps q f(cid:0)ρ(t, b)(cid:1)·µp(t, b) µpj(t, aj) = i if (cid:0)ρi(t, ai)(cid:1) i ∀p⊃{Ii, Ij} (3.27) j j j (cid:88) AJ(t)=(cid:8)αJ(t)(cid:9), αJ(t)= µp(t, b) ∀I ∈IJ, I ∈OJ (3.28) ij ij i i i j p(cid:51)Ii,Ij ρ (t, b ) = RSAJ(cid:104)(cid:0)ρ(t, b−)(cid:1) , (cid:0)ρ (t, a +)(cid:1) (cid:105) ∀I ∈IJ (3.29) k k k i i Ii∈IJ j j Ij∈OJ k ρ(t, a) = RSAJ(cid:104)(cid:0)ρ(t, b−)(cid:1) , (cid:0)ρ (t, a +)(cid:1) (cid:105) ∀I ∈OJ (3.30) l l l i i Ii∈IJ j j Ij∈OJ l D (t, h) = λ ◦λ ◦λ ...◦λ (t) ∀p∈P, ∀t∈[0, T] (3.31) p s 1 2 m(p) 10

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.