Hybridization ofverylargeneighborhoodsearchforready-mixedconcretedelivery problems

Companies intheconcreteindustryarefacingthefollowingschedulingproblemonadailybasis:concrete produced atseveralplantshastobedeliveredatcustomersʹconstructionsitesusingaheterogeneous fleet ofvehiclesinatimely,butcost-effectivemanner.Thedistributionofready-mixedconcrete(RMC)is a highlycomplexprobleminlogisticsandcombinatorialoptimization. This paperproposestwohybridsolutionproceduresfordealingwiththisproblem.Theyarebasedona combination ofanexactalgorithmandavariableneighborhoodsearch(VNS)approach.TheVNSisused at firsttogeneratefeasiblesolutionsandistryingtofurtherimprovethem.Theexactmethodisbasedon a mixedintegerlinearprogramming(MILP)formulation,whichissolved(afteranappropriatedvariable fixing phase)byusingageneral-purposeMILPsolver.Anapproachbasedonverylargeneighborhood search (VLNS)determineswhichvariablesaresupposedtobefixed.Inasense,theapproachesfollows a localbranchingscheme.ThehybridmetaheuristicsarecomparedwiththepureVNSapproachandthe conclusion isthatthenewmetaheuristicsoutperformtheVNSifappliedsolely.