A memeticalgorithmforthemulti-compartmentvehicleroutingproblemwith stochastic demands

The multi-compartmentvehicleroutingproblem(MC-VRP)consistsofdesigningtransportationroutes to satisfythedemandsofasetofcustomersforseveralproductsthat,becauseofincompatibilitycon- straints, mustbeloadedinindependentvehiclecompartments.Despiteitswidepracticalapplicability the MC-VRPhasnotreceivedmuchattentionintheliterature,andthefewexistingmethodsassume perfect knowledgeofthecustomerdemands,regardlessoftheirstochasticnature.Thispaperextends the MC-VRPbyintroducinguncertaintyonwhatitisknownastheMC-VRPwithstochasticdemands (MC-VRPSD). TheMC-VRPSDismodeledasastochasticprogramwithrecourseandsolvedbymeans of amemeticalgorithm.Theproposedmemeticalgorithmcouplesgeneticoperatorsandlocalsearch procedures proventobeeffectiveondeterministicroutingproblemswithanovelindividualevaluation and reparationstrategythataccountsforthestochasticnatureoftheproblem.Thealgorithmwastested on instancesofupto484customers,anditsresultswerecomparedtothoseobtainedbyasavings-based heuristic andamemeticalgorithm(MA/SCS)fortheMC-VRPthatusesasparecapacitystrategytohandle demand fluctuations.InadditiontoeffectivelysolvetheMC-VRPSD,theproposedMA/SCSalsoimproved 14 bestknownsolutionsina40-problemtestbedfortheMC-VRP.