简介:Adaptivecritic(AC)basedcontrollersaretypicallydiscreteand/oryieldauniformlyultimatelyboundedstabilityresultbecauseofthepresenceofdisturbancesandunknownapproximationerrors.Acontinuous-timeACcontrollerisdevelopedthatyieldsasymptotictrackingofaclassofuncertainnonlinearsystemswithboundeddisturbances.TheproposedAC-basedcontrollerconsistsoftwoneuralnetworks(NNs)-anactionNN,alsocalledtheactor,whichapproximatestheplantdynamicsandgeneratesappropriatecon...
简介:Theadaptivecriticheuristichasbeenapopularalgorithminreinforcementlearning(RL)andapproximatedynamicprogramming(ADP)alike.ItisoneofthefirstRLandADPalgorithms.RLandADPalgorithmsareparticularlyusefulforsolvingMarkovdecisionprocesses(MDPs)thatsufferfromthecursesofdimensionalityandmodeling.Manyreal-worldproblems,however,tendtobesemi-Markovdecisionprocesses(SMDPs)inwhichthetimespentineachtransitionoftheunderlyingMarkovchainsisitselfarandomvariab...
简介:Approximatedynamicprogramming(ADP)formulationimplementedwithanadaptivecritic(AC)-basedneuralnetwork(NN)structurehasevolvedasapowerfultechniqueforsolvingtheHamilton-Jacobi-Bellman(HJB)equations.AsinterestinADPandtheACsolutionsareescalatingwithtime,thereisadireneedtoconsiderpossibleenablingfactorsfortheirimplementations.AtypicalACstructureconsistsoftwointeractingNNs,whichiscomputationallyexpensive.Inthispaper,anewarchitecture,calledthe'cost-functio...
简介:摘要;为更加全面评价兰州城市交通拥堵情况,本文构建交通趋势的指标,采用CRITIC_TOPSIS方法,定量分析兰州交通拥堵演变趋势,并将此方法应用于兰州四条主要的交通干道中进行趋势评价,并对缓解交通拥堵问题提出相应建议。