简介:Thispaperpresentsanewhybridgeneticalgorithmforthevertexcoverproblemsinwhichscan-repairandlocalimprovementtechniquesareusedforlocaloptimization.Withthehybridapproach,geneticalgorithmsareusedtoperformglobalexplorationinapopulation,whileneighborhoodsearchmethodsareusedtoperformlocalexploitationaroundthechromosomes.Theexperimentalresultsindicatethathybridgeneticalgorithmscanobtainsolutionsofexcellentqualitytotheprobleminstanceswithdifferentsizes.Thepuregeneticalgorithmsareoutperformedbytheneighborhoodsearchheuristicsprocedurescombinedwithgeneticalgorithms.
简介:Adiscretedifferentialevolutionalgorithmcombinedwiththebranchandboundmethodisdevelopedtosolvetheintegerlinearbilevelprogrammingproblems,inwhichbothupperlevelandlowerlevelvariablesareforcedtobeinteger.Anintegercodingforupperlevelvariablesisadopted,andthenadiscretedifferentialevolutionalgorithmwithanimprovedfeasibility-basedcomparisonisdevelopedtodirectlyexploretheintegersolutionattheupperlevel.Foragivenupperlevelintegervariable,thelowerlevelintegerprogrammingproblemissolvedbytheexistingbranchandboundalgorithmtoobtaintheoptimalintegersolutionatthelowerlevel.Inthesameframeworkofthealgorithm,twootherconstrainthandlingmethods,i.e.thepenaltyfunctionmethodandthefeasibility-basedcomparisonmethodarealsotested.Theexperimentalresultsdemonstratethatthediscretedifferentialevolutionalgorithmwithdifferentconstrainthandlingmethodsiseffectiveinfindingtheglobaloptimalintegersolutions,buttheimprovedconstrainthandlingmethodperformsbetterthantwocomparedconstrainthandlingmethods.
简介:InordertoovercometheshortcomingoftheclassicalHungarianalgorithmthatitcanonlysolvetheproblemswherethetotalcostisthesumofthatofeachjob,animprovedHungarianalgorithmisproposedandusedtosolvetheassignmentproblemofserial-parallelsystems.Firstofall,byreplacingparalleljobswithvirtualjobs,theproposedalgorithmconvertstheserial-parallelsystemintoapureserialsystem,wheretheclassicalHungarianalgorithmcanbeusedtogenerateatemporalassignmentplanviaoptimization.Afterwards,theassignmentplanisvalidatedbycheckingwhetherthevirtualjobscanberealizedbyrealjobsthroughlocalsearching.Iftheassignmentplanisnotvalid,theconvertedsystemwillbeadaptedbyadjustingtheparametersofvirtualjobs,andthenbeoptimizedagain.Throughiterativesearching,thevalidoptimalassignmentplancaneventuallybeobtained.Toevaluatetheproposedalgorithm,thevalidoptimalassignmentplanisappliedtolaborallocationofamanufacturingsystemwhichisatypicalserial-parallelsystem.
简介:Nystrommethodisanewmethodforsolvingelectromagneticscatteringproblems.Thispapergivesthedetaileddescriptiononhigh-orderNystrommethodusedfortheelectricfieldintegralequationofelectromagneticscatteringproblems.ThenumericalsolutionsoftwoexamplesarecorrectcomparedwithMethodOfMoment(MOM).
简介:Amemeticalgorithm(MA)foramulti-moderesourceconstrainedprojectschedulingproblem(MRCPSP)isproposed.WeuseanewfitnessfunctionandtwoveryeffectivelocalsearchproceduresintheproposedMA.Thefitnessfunctionmakesuseofamechanismcalled'strategicoscillation'tomakethesearchprocesshaveahigherprobabilitytovisitsolutionsarounda'feasibleboundary'.Oneofthelocalsearchproceduresaimsatimprovingthelowerboundofprojectmakespantobelessthanaknownupperbound,andanotheraimsatimprovingasolutionofanMRCPSPinstanceacceptinginfeasiblesolutionsbasedonthenewfitnessfunctioninthesearchprocess.AdetailedcomputationalexperimentissetupusinginstancesfromtheprobleminstancelibraryPSPLIB.ComputationalresultsshowthattheproposedMAisverycompetitivewiththestate-of-the-artalgorithms.TheMAobtainsimprovedsolutionsforoneinstanceofsetJ30.
简介:Aself-adaptivelargeneighborhoodsearchmethodforschedulingnjobsonmnon-identicalparallelmachineswithmultipletimewindowsispresented.Theproblems'anotherfeatureliesinoversubscription,namelynotalljobscanbescheduledwithinspecifiedschedulinghorizonsduetothelimitedmachinecapacity.Theobjectiveisthustomaximizetheoverallprofitsofprocessedjobswhilerespectingmachineconstraints.Afirst-infirst-outheuristicisappliedtofindaninitialsolution,andthenalargeneighborhoodsearchprocedureisemployedtorelaxandreoptimizecumbersomesolutions.Amachinelearningmechanismisalsointroducedtoconvergeonthemostefficientneighborhoodsfortheproblem.Extensivecomputationalresultsarepresentedbasedondatafromanapplicationinvolvingthedailyobservationschedulingofafleetofearthobservingsatellites.Themethodrapidlysolvesmostprobleminstancestooptimalornearoptimalandshowsarobustperformanceinsensitiveanalysis.
简介:Amodifiedbottleneck-based(MB)heuristicforlarge-scalejob-shopschedulingproblemswithawell-definedbottleneckissuggested,whichissimplerbutmoretailoredthantheshiftingbottleneck(SB)procedure.Inthisalgorithm,thebottleneckisfirstscheduledoptimallywhilethenon-bottleneckmachinesaresubordinatedaroundthesolutionsofthebottleneckschedulebysomeeffectivedispatchingrules.ComputationalresultsindicatethattheMBheuristiccanachieveabettertradeoffbetweensolutionqualityandcomputationaltimecomparedtoSBprocedureformedium-sizeproblems.Furthermore,itcanobtainagoodsolutioninashorttimeforlarge-scalejob-shopschedulingproblems.
简介:Themixedl1/H2optimizationproblemforMIMO(multipleinput-multipleoutput)discrete-timesystemsisconsidered.Thisproblemisformulatedasminimizingthel1-normofaclosed-looptransfermatrixwhilemaintainingtheH2-normofanotherclosed-looptransfermatrixatprescribedlevel.ThecontinuitypropertyoftheoptimalvalueinrespecttochangesintheH2-normconstraintisstudied.Theexistenceoftheoptimalsolutionsofmixedll/H2problemisproved.Becausethesolutionofthemixedl1/H2problemisbasedonthescaled-Qmethod,itavoidsthezerointerpolationdifficulties.Theconvergentupperandlowerboundscanbeobtainedbysolvingasequenceoffinitedimensionalnonlinearprogrammingforwhichmanyefficientnumericaloptimizationalgorithmsexist.