简介:Recentextensivemeasurementsofreal-lifetrafficdemonstratethattheprobabilitydensityfunctionofthetrafficinnon-Gaussian.Ifatrafficmodeldoesnotcapturethischaracteristics,anyanalyticalorsimulationresultswillnotbeaccurate.Inthiswork,westudytheimpactofnon-Gaussiantrafficonnetworkperformance,andpresentanapproachthatcanaccuratelymodelthemarginaldistributionofreal-lifetraffic.Boththelong-andshort-rangeautocorrelationsarealsoaccounted.Weshowthattheremovalofnon-Gaussiancomponentsoftheprocessdoesnotchangeitscorrelationstructure,andwevalidateourpromisingprocedurebysimulations.
简介:Sinceitisdifficulttofitmeasuredparametersusingtheconventionaltrafficmodel,anewtrafficdensityandaveragespeedmodelisintroducedinthispaper.Todeterminetrafficmodelstructuresaccurately,amodelidentificationmethodforuncertainnonlinearsystemisdeveloped.Tosimplifyuncertainnonlinearproblem,thispaperpresentsanewrobustcriteriontoidentifythemulti-sectiontrafficmodelstructureoffreewayefficiently.Inthenewmodelidentificationcriterion,numericallyefficientU-Dfactorizationisusedtoavoidcomputingthedeterminantvaluesoftwocomplexmatrices.ByestimatingthevaluesofU-Dfactorofdatamatrix,boththeupperandlowerboundsofsystemuncertaintiesaredescribed.Thusamodelstructureidentificationalgorithmisproposed.Comparisonsbetweenidentificationoutputsandsimulationoutputsoftrafficstatesshowthatthetrafficstatescanbeaccuratelypredictedbymeansofthenewtrafficmodelsandthestructureidentificationcriterion.
简介:Car-followingtheoryisanimportantresearchdirectioninthefieldofintelligenttransportationsystems(ITS),itdescribestheone-by-onefollowingprocessofvehiclesonthesamelaneintrafficflow,andoneofitsimportantissuesiscongestioncontrol.Toexplorethestrategyforcontrollingtrafficcongestion,thispaperintroducesandanalyzessomeclassiccar-followingmodels,andgivesasystematicreviewoftheirdevelopments.Moreover,inordertointroducetheapproachtoanalyzethestability,takingthefullvelocitydifference(FVD)modelforexample,thelocalandasymptoticstabilityanalysisisdiscussed,whilethecorrespondingnonlinearanalysisisalsoconducted.Then,someperspectivesofthecar-followingmodelaregiveninthefinal.
简介:Inordertocontrolthelarge-scaleurbantrafficnetworkthroughhierarchicalordecentralizedmethods,itisnecessarytoexploitanetworkpartitionmethod,whichshouldbebotheffectiveinextractingsubnetworksandfasttocompute.Inthispaper,anewapproachtocalculatethecorrelationdegree,whichdeterminesthedesireforinterconnectionbetweentwoadjacentintersections,isfirstproposed.Itisusedasaweightofalinkinanurbantrafficnetwork,whichconsidersboththephysicalcharacteristicsandthedynamictrafficinformationofthelink.Then,afastnetworkdivisionapproachbyoptimizingthemodularity,whichisacriteriontodistinguishthequalityofthepartitionresults,isappliedtoidentifythesubnetworksforlarge-scaleurbantrafficnetworks.Finally,anapplicationtoaspecifiedurbantrafficnetworkisinvestigatedusingtheproposedalgorithm.Theresultsshowthatitisaneffectiveandefficientmethodforpartitioningurbantrafficnetworksautomaticallyinrealworld.
简介:Withthemulti-tierpricingschemeprovidedbymostofthecloudserviceproviders(CSPs),theclouduserstypicallyselectahighenoughtransmissionserviceleveltoensurethequalityofservice(QoS),duetotheseverepenaltyofmissingthetransmissiondeadline.Thisleadstotheso-calledover-provisioningproblem,whichincreasesthetransmissioncostoftheclouduser.Giventhefactthatcloudusersmaynotbeawareoftheirtrafficdemandbeforeaccessingthenetwork,theover-provisioningproblembecomesmoreserious.Inthispaper,weinvestigatehowtoreducethetransmissioncostfromtheperspectiveofcloudusers,especiallywhentheyarenotawareoftheirtrafficdemandbeforethetransmissiondeadline.Thekeyideaistosplitalong-termtransmissionrequestintoseveralshortones.Byselectingthemostsuitabletransmissionservicelevelforeachshort-termrequest,acost-efiqcientinter-datacentertransmissionservicelevelselectionframeworkisobtained.Wefurtherformulatethetransmissionservicelevelselectionproblemasalinearprogrammingproblemandresolveitinanon-linestylewithLyapunovoptimization.Weevaluatetheproposedapproachwithrealtrafficdata.Theexperimentalresultsshowthatourmethodcanreducethetransmissioncostbyupto65.04%.