Reducingtransitfleetemissionsthroughvehicleretrofits,replacements,andusagechangesovermultipletimeperiodsTimonH.Stasko,H.OliverGao*SchoolofCivilandEnvironmentalEngineering,CornellUniversity,Ithaca,NY14853,USAarticleinfoKeywords:TransitvehicleschedulingVehiclereplacementEmissionsDieselretrofitGreenfleetmanagementSustainabilityabstractBustransitisoftenpromotedasagreenformoftransportation,butsurprisinglylittleresearchhasbeendoneonhowtoruntransitsystemsinagreenmanner.Bothvehicletaskassignmentandpurchasemodelsaregenerallyconstructedtominimizefinancialcosts.Integratingvehicletaskassignmentwithpurchasedecisionsismadechallengingbythedifferenttimescalesinvolved.Anintegerprogrammingapproachisusedtocombinevehi-clepurchase,retrofitandaggregatedtaskassignmentdecisions.Theformulationisdesignedtooperateinsequencewithtraditionalvehicletaskassignmentmodels,toaddemissionsandlongtermfinancialcostelementstotheobjective,whilemaintainingcom-putationaltractabilityandfeasibleinputdatarequirements.Inacasestudy,atransitagencysavesmoneyinthelongtermbyusingstimulusmoneytobuyCNGinfrastructureinsteadofpurchasingonlynewbuses.Carbonpricesupto$400/(tonCO2equivalent)donotchangevehiclepurchasedecisions,buthighercarbonpricescancausemoredieselhybridpurchases,atahighmarginalcost.AlthoughthemotivationandnumericalcasestudyarefromtheUStransitindustry,themodelisformulatedtobewidelyapplicabletogreenfleetmanagementinmultiplecontexts.2010ElsevierLtd.Allrightsreserved.1.IntroductionBustransitisoftenpromotedasagreenformoftransportation,butsurprisinglylittleresearchhasbeendoneonhowtoruntransitsystemsinagreenmanner.Researchontransitvehicleassignment(theprocessofassigningvehiclestoroutes)focusesonminimizingcapitaland/oroperatingcosts,orfactorsrelatedtocost,suchasthenumberofvehiclesrequired.Evenwithoutincludingrelatedproblemssuchasdriverscheduling,vehicleassignmentproblemscanbeverylargeanddifficulttosolve(Klieweretal.,2006;BanihashemiandHaghani,2000;Haghanietal.,2003).LiandHead(2009)standsoutfromothervehicleassignmentresearchbyincludingemissionsconstraintsandpenaltiesinitsformulations.Unliketypicaltransitvehi-cleassignmentmodels,theyexplicitlymodelvehiclepurchases,limitedbyacapitalbudget.Transitvehicleassignmentmodels,includingthatofLiandHead(2009),aregenerallyconstructedtooptimizethevehiclemovementsneededtocoverallroutesonatimetable,whichisusuallyonedaylong.Intransit,thetimetablestendtorepeatregularly(perhapswithvariantsdependingondayoftheweek).Itisacommonpracticetoreworktransitvehi-cleassignmentsseveraltimesperyear,asadjustmentstoroutesareintroduced(BanihashemiandHaghani,2000).LiandHead(2009)donotforecasthowoperationalcostswillchangeduetofuturerouteadjustments,orotherfactorssuchasshiftingfuelprices.Theimpactsofcurrentvehiclepurchaseandusagedecisionsonfuturecapitalexpendituresarenotconsidered.Thefinegrainednatureofthetime–spacenetwork,whichisessentialtoshorttermvehicleassignment1361-9209/$-seefrontmatter2010ElsevierLtd.Allrightsreserved.doi:10.1016/j.trd.2010.03.004*Correspondingauthor.Tel.:+16072548334;fax:+16072559004.E-mailaddresses:ths9@cornell.edu(T.H.Stasko),hg55@cornell.edu(H.OliverGao).TransportationResearchPartD15(2010)254–262ContentslistsavailableatScienceDirectTransportationResearchPartDjournalhomepage:ficult.Asaresult,vehiclepurchasedecisionsaremadewithaveryshortplan-ninghorizoncomparedtothelifeofthevehicle.Simmsetal.(1984)presentedalongtermmodelforplanningthetimingofbuspurchases,aswellasthedegreetowhichbusesareusedinagivenperiod.Thismodeldoesnotdistinguishbe-tweendifferentbustechnologies,orbetweendifferentdutycycles,anditdoesnotincludeemissionsobjectivesorconstraints.Thispaperproposesanintegerprogrammingmethodofmakingvehiclepurchasedecisionsoveralongtimehorizon,whilesimultaneouslymakingvehicletaskassignmentsatanaggregatedlevel.Theintegerprogramminimizesoperationalcosts,pluspenaltiesforemissions,givencapitalbudgetconstraints.Retrofitsareincorporatedasanalternatemethodofreducingemissions.Retrofitsmayonlybecompatiblewithsomevehicles,andtheymayonlybecompatiblewithsometasks(dependingontheexhausttemperaturesgenerated).GaoandStasko(2009)presentedamodelfordevelopingretrofitandreplacementstrategies,butitonlyconsideredretrofitsandreplacementsmadeatonepointintime,anddidnotincludetheconnectedproblemofvehicletaskassignment.Section2presentstheintegerprogrammingmethodology,whileSection3describesacasestudyforatransitoperator,andSection4outlinesconclusionsandpotentialextensions.2.Modelformulation2.1.ExpandingontraditionalvehicleschedulingAlthoughthemotivationandcasestudyarefromthetransitindustry,themodelisformulatedtobewidelyapplicabletomultipletypesoffleets.Thegoalistobeabletomodelvehiclepurchase,retrofit,andaggregatedtaskassignmentdecisionsovermultipleperiods,whileconsideringbothfinancialcostsandemissions.Perhapsthemoststraightforwardwaytolengthentheplanninghorizonofatraditionalvehicleassignmentmodelwouldbetoincludemultipleconnectednetworks.Eachnetworkcouldrepresentatypicaldayinalonger(perhapsquar-terly)period.Th