无人机路径规划优化方法

0Abstract—Pathplanningofunmannedcombataerialvehicle(UCAV)aimstoseekanoptimalflightrouteconsideringthreatsandconstraintsalongthewaytowardstheterminaltarget.Thispaperproposedanovelprobabilitydensitymodeltotransformtheinitialpathplanningtaskintoanumericalproblem,whichshowshigheraccuracyincomparisonwiththetraditionalcircletreatmodel.Thewell-knownArtificialBeeColonyalgorithm(ABC)isusedtosettlethiscorrespondingoptimizationproblemandcomparisonsaremadebetweentheproposedalgorithmandotherintelligencealgorithmsregardingconvergencerateandefficiencyinvariousseriesofcombatfields.ExperimentalresultsverifiedwithstatisticalsignificancethesuperiorityofABCfortheUCAVpathplanningproblem.I.INTRODUCTIONINrecentyears,unmannedcombataerialvehicle(UCAV)hasbeenofhighinteresttomanymilitaryorganizationsthroughouttheworld,consideringitsabilitytoworkindangerousorcomplicatedenvironments[1]–[3].PathplanningsystemisacrucialcomponentofautonomouscontrolmoduleinUCAV,whichprovidesanoptimalpathfromthestartingpointtothedesireddestination.Duringtheflight,artificialthreatsandsomenaturalconstraintsaresupposedtoavoid.ForaUCAVpathplanningtask,theoptimalsolutioncorrespondstoonethatminimizesthetraveleddistance,averagealtitude,consumptionoffuel,exposuretoradarorartillery,etc[4].Asguardinganddefendingweaponsdevelop,thecomplexityofmodelingtheseartificialthreatssignificantlygrows.Tocopewiththeincreasingcomplexity,researchershavegraduallyshiftedtheirinterestawayfromdeterministicalgorithms.Toavoidaninefficientenumeratingprocess,intelligencealgorithmshavebeendevelopedandinvestigatedinrecentyears,suchasgeneticalgorithm(GA)[5],differentialevolutionalgorithm(DE)[6],particleswarmoptimizationalgorithm(PSO)[7]aswellastherecentlyproposedartificialbeecolonyalgorithm(ABC)[8].PSOisManuscriptreceivedMarch1,2013.Thisworkwassponsoredinpartby5thand6thNationalCollegeStudents'InnovativeandEntrepreneurialTrainingProgramsandsupportedbytheDepartmentofAdvancedEngineeringinBeihangUniversity.BaiLiiswiththeSchoolofAdvancedEngineering,BeihangUniversity,Beijing,100191,China(E-mail:[e-mail]).LigangGongiswiththeSchoolofAutomationScienceandElectricalEngineering,BeihangUniversity,Beijing,100191,China(E-mail:[e-mail]).ChunhuiZhaoiswithStateKeyLaboratoryofIndustrialControlTechnology,DepartmentofControlScienceandEngineering,ZhejiangUniversity,Hangzhou,310027,China(Tel:[phone];E-mail:[e-mail]).inspiredbythesocialbehaviorofbirdflocking,whereaswarmofparticlesmoveinthesearchspaceforappropriatesolutionsandeveryparticleownsapositionvectoraswellasavelocityvector.Eachparticlerecordsitsownbestpositionsofarandaglobalcurrentbestpositionisreadilyavailableforadjustmentsinvectorsofparticles.ABCwasfirstlyproposedbyDervisKarabogain2005,whichimitatedtheforagingbehaviorofbeeswarms.Inthisalgorithm,bothlocalexploitationandglobalexplorationareconductedineachiteration.Itworkswellinexplorationandhasarousedgreatconcernsincethen.Beforetheadoptionofthesealgorithms,anappropriatemathematicalmodelneedstobebuilt,whichtransfersthepathplanningtaskintoanumericaloptimizationproblem.Itisnotedthat,asinpreviouswork,thethreatzonearoundathreatpointhasalwaysbeendescribedbythetraditionalcirclemodel,whereresearchersseldomconsideredthedifferencesinsideoroutsidethecircle[9]–[11].Forthespecificproblemconcernedinthispaper,variousalgorithmssuchasGA[12],ImmuneGA(I-GA)[13],PSO[14],Quantum-behavedPSO(Q-PSO)[15],Master-slaveparallelvector-evaluatedGA(MPV-GA)[16],andChaoticABC(C-ABC)[9]havebeendeveloped.Viewingpreviouswork,wefindthatPSOdefectstoavoidgettingtrappedinlocaloptimums.Ontheotherhand,thetruepotentialityofABCwasseldomcarefullyconsideredforsuchreal-timeproblems,andfewrigorousexperimentalcomparisonsweremadeamongtheperformancesoftheseintelligencealgorithms.Inthispaper,anovelprobabilitydensitymodelisintroducedtodifferentiatedifferentlocationsbytheirdistancestowardsthethreatcenter.AclearinsightintothecapabilityofABCisalsoprovidedforthisUCAVpathplanningproblem.Simulationresultsverify,withstatisticalsignificance,pathsobtainedbyABCaresuperiortothosebyPSO.Itisnotedthatthisworkpreliminarilyfocusesonpathplanningintwodimensions.Therestofthispaperisorganizedasfollows.InSectionII,principleoftheproposedenvironmentalmodelisintroduced.TheprocedureofABCiselaboratedinSectionIII.Severalwell-designedcomparablesimulationresultsarepresentedandcarefullydiscussedinSectionIV.Andthenafinalconclusionisdrawnintheend.II.MODELOFUCAVPATHPLANNINGForthisspecificUCAVpathplanningproblem,researchershavebeenpursuingmodelsthatwelldescribethetrueenvironment,whichaccountforartificialthreatsandnaturalconstraints.Inthiswork,thefamous2Dmodelin[9]isbasicallyused,andtheprobabilitydensitycirclemodelUnmannedCombatAerialVehiclesPathPlanningUsingaNovelProbabilityDensityModelBasedonArtificialBeeColonyAlgorithmBaiLi,LigangGong,andChunhuiZhaoreplacestheoriginalcirclemodeltocharacterizethreatzones.AsinFig.1,apaththatlinksupstartingpointandterminaldestinationisneededtoobtain.Atfirst,segmentSTwhichconnectingthestartingandterminalpointsisdrawn.T