简介:Learningiswidelyusedinintelligentplanningtoshortentheplanningprocessorimprovetheplanquality.Thispaperaimsatintroducinglearningandfatigueintotheclassicalhierarchicaltasknetwork(HTN)planningprocesssoastocreatebetterhighqualityplansquickly.TheprocessofHTNplanningismappedduringadepth-firstsearchprocessinaproblem-solvingagent,andthemodelsoflearninginHTNplanningisconductedsimilartothelearningdepth-firstsearch(LDFS).Basedonthemodels,alearningmethodintegratingHTNplanningandLDFSispresented,andafatiguemechanismisintroducedtobalanceexplorationandexploitationinlearning.Finally,experimentsintwoclassicaldomainsarecarriedoutinordertovalidatetheeffectivenessoftheproposedlearningandfatigueinspiredmethod.