基于GEV分布参数的路网模式分类(IJEM-V2-N3-4)

I.J.EngineeringandManufacturing,2012,3,21-29PublishedOnlineJune2012inMECS()DOI:10.5815/ijem.2012.03.04Availableonlineat’anRoad,Shanghai,201804,P.R.ofChinaAbstractNetworkpatternmentionedinthispaperisreferredtothegeographicallayoutandstructureofanetwork,whichisrelatedtotheconnection,direction,andcombinationfeaturesofroadsinaroadnetwork.Aquantitativemethodisproposedinthispapertoclassifypatternsofnetworks,throughwhichanetworkcouldbeidentifiedquantitativelytobeoneofthethreestandardpatterns,i.e.,gridnetwork,circle+radialnetworkandtree-patternednetwork.MetricdistancesofshortestpathsaretakenasthemainfeaturesofthenetworksandaredescribedbyparametersthroughGeneralizedExtremeValue(GEV)fitting.Thecriteriaforpatternclassificationwereestablishedaccordingtotheclusteranalysisoftheparameterscalculatedfromasetoftrialnetworks.Sixrealnetworkswerecalculatedusingthemethodandtheirpatternsareidentifiedaccordingtotheproposedcriteria.Itturnedoutthatthemethodcouldcapturethefeaturesofthenetworkpatternswell.Thismethodmaysetathresholdofthemoregeneralanddeepstudiesofnetworkpatternclassification,whichmayofferhelptoroadnetworkplanningandassessment.IndexTerms:Roadnetworkpattern;networktopology;classification;GEVdistribution©2012PublishedbyMECSPublisher.Selectionand/orpeerreviewunderresponsibilityoftheResearchAssociationofModernEducationandComputerScience.1.IntroductionOnemaytellthenameofacityimmediatelyonlybytakingalookatitsroadnetworkbecausethenetworkofacityhasitsownpattern,whichcouldalsobenamedasmorphology.Asisconcernedhere,networkpatternisreferredtothegeographicallayoutandstructureofanetwork,whichisrelatedtotheconnection,direction,andcombinationofroadsinaroadnetwork.Roadnetworksofcitiesgettheirownpatternsthroughyearsofevolution.Lookingintothehistoryofurbanevolution,itisobviousoftheinteractionbetweenurbanlandexpandingandroadnetworkgrowing,whichisalsomentionedasachickenandeggproblem,i.e.,thegrowthoftrafficnetworkdependsonurbanlandusepattern,andnevertheless,hasobviousinfluencesonurbanlanduseinreturn.Thecircularinfluences*Correspondingauthor.E-mailaddress:[e-mail],[e-mail]22RoadNetworkPatternClassificationUsingGEVDistributionParameterscanbeshownmoreclearlyinFig.1.Fig.1Interactionsbetweenroadnetworkpatternandurbanlanduse.Differentpatternsofroadnetworksmayhavedifferentinfluencesonbothurbanlanduseevolutionandnetworkperformances.Intheearlyyearsofthe20thcentury,plannersandarchitectsputforwardamethodbywhichstrictlyhierarchicalroadsareorganizedintoatreepatternednetworkbecausetheythoughtanetworkinthatpatterncouldbemoreefficientandsafeforthevehicles‎[1].Accordingtothistheory,ITEhaspublishedRecommendedPracticeforSubdivisionStreets‎[2],inordertoimprovethetrafficsafetyofcommunitystreetsthroughhierarchicalandunconnectednetworklayout.Recently,asaccessibilityofmulti-modeltrafficbecomemoreandmoreconcerned,trafficplannersstarttoconsiderincreasingthecapacityofawholenetworkbystrengthenitsdensityandconnectivity‎[2],andhighlyconnectedpatternbecomesafavorableoneinnetworkplanning‎[3].Inspiteofthedevelopmentthathasbeengainedinthenetworkpatternstudy,theinsightintonetworkpatternsisstillnotclearandonlyafewresearchershaveworkedonit.SunandLovegrovehavestudiedfourpatternsofnetworksandcomparedtheirlevelsoftrafficsafety‎[4].Sunmeasuredurbanrailtransitnetworkpatternusingfractalapproachanddescribedtheinteractionsandmorphologicalsimilaritiesbetweenurbanlandusemorphologyandurbantransportnetworkpattern‎[5].Oneoftheveryimportantproblemsduringthestudyofnetworkpatternisthatthereisnotyetastandardforpatternclassificationwhichisagreedbymostresearchers.Furthermore,thereisnotevenaneffectivemethodtodistinguishbetweentypesofpatterns.TakeSun’sresearch‎[4]forinstance,thefourtypesofnetworkwerejustputoutbytheauthor,thusmaybedifficultforotherresearchesorcitiestofollow.Tosolvethiskindofproblems,thispaperproposesaquantitativemethodtoidentifydifferenttypesofnetworkpatterns.Thismethodcouldcapturethemainfeaturesofanetwork’spatternanditprovedouttobeabletodistinguishdifferenttypesofnetworkpatterns.Thefollowingsectionwilltalkaboutmodelsusedtodescribenetworks.Inthethirdsection,thenetworkpatternclassificationmethodandrelatedcriteriawillbeproposedusingseveraltrialnetworksamples.Intheforthsection,actualroadnetworkswillbecalculatedinordertovalidatethemethod.Conclusionswillbestatedinthefinalsection.2.NetworkDescriptionUsingGevModelInthispaper,networkisabstractedasaplanargraph:G={V,E},whereVisacollectionofnodesconnectedbylinks(edges)E.Nodesrepresentintersectionswhilelinksrepresentroadsegmentsbetweenintersections.Adjacentnodesaredefinedasthenodesthatconnectwitheachotherwithoutpassingthroughothernodes.Theadjacentrelationscanberepresentedbythetopologicaladjacencymatrix{aij}:1,ifthereisanedgejoiningnodeitonodej;,otherwise;ija(1)andcanals