蚁群算法及其应用研究(3)
发布时间:2021-06-06
发布时间:2021-06-06
●___Il●●●l,I—_●I——I——_ lII _l_l●__●IlIl___l _I Abstract
Abstract
Biologists’studyfoundthatnatural
pheromoneantscouldreleaseachemicalsubstanceknownascommunicationandthroughwhich
aantcolonyconductindirectcooperationinordertofindshortestpathfromnesttofood.Inspiredbytheant’s
behavior,ItalyscholarDorigoandhiscolleaguesdidsomesimulationresearchbycomputer.Forthefirsttime,theyproposed
years,antantcolonyalgorithm.Inthefollowing10colonyalgorithmwasappliedtoCombinatorialoptimization,network
SOrouting,functionoptimization,datamining,robotpathplanningand
showsthegreaton,whichadvantageofantcolonyalgorithminsolvingcomplexproblemsandabrightfutureforfurtherdevelopment.
However,therearestillsomedefectsinantcolonyalgorithmsuchascostingtoomuchtimeandeasytodropintostagnation.Thispaperfocusesontheprincipleofantcolonyoptimizationandits
antcolonyapplication.WeasdosomeresearchonimprovementuponalgorithmaswellapplicationtoTSP(TravelingSalesmanProblem)and
onMKP(Multidimensional
andKnapsackProblem).Basedaccuracyofstandarddatasetswecomparealgorithmsanalyzetheefficiencyandoriginalandproposedalgorithms.
First,anewAntColonyOptimizationalgorithmbased
diffusionmodelisproposed.It
energyconversationandonpheromoneincrementandadoptsanewpheromoneupdatemechanismbasedontransformwhichintegrates
pheromonetheimpactofglobalinformationandlocalinformationonandembodiesthepheromonedifferencefor
originalpheromonediffusionmodeland
tofaithfullyreflectadifferentpaths.Meanwhile,weimprovethepathpheromonediffusionmodelisestablishedthestrengthfieldof
pheromonediffusionwhichstrengthensthecollaborationamongants.Amutationstrategy、^,itll
result.lowercomputationalcomplexityisadoptedtooptimizeeachevolution
Second,aimingatsolvinglarge—scaleTravelingSalesmanProblemswhichconsume
alargecomputation,anewalgorithmisproposed.Itadopts
strategiestolookforansetareofmultistageoptimalsolution.Firstly,thecitiesofTSPclusteredintoseveraldistrictsbymeansofdensity-basedalgorithm.Secondly,theantalgorithm
tosolving
abasedonpheromonediffusionmodelisappliedinparallelizationthesub—problemineachdistrict.Then,alldistrictsolutionsareintegratedintosolution.
Finally,localoptimizationisconductedbymeansofPartitionoptimization.Onthis