Title of article :
NumericalanalysisofcontinuoustimeMarkovdecisionprocesses
over finitehorizons
Author/Authors :
Peter Buchholz، نويسنده , , IngoSchulz، نويسنده ,
Issue Information :
ماهنامه با شماره پیاپی سال 2011
Abstract :
ContinuoustimeMarkovdecisionprocesses(CTMDPs)withafinitestateandactionspacehavebeen
consideredforalongtime.Itisknownthatunderfairlygeneralconditionstherewardgainedovera
finite horizoncanbemaximizedbyaso-calledpiecewiseconstantpolicywhichchangesonlyfinitely
often inafiniteinterval.Althoughthisresultisavailableformorethan30years,numericalanalysis
approachestocomputetheoptimalpolicyandrewardarerestrictedtodiscretizationmethodswhich
are knowntoconvergetothetruesolutionifthediscretizationstepgoestozero.Inthispaper,we
present anewmethodthatisbasedonuniformizationoftheCTMDPandallowsonetocomputean
e optimalpolicyuptoapredefinedprecisioninanumericallystablewayusingadaptivetimesteps.
Keywords :
Finite horizon , Error bounds , Uniformization , Continuous time Markov decision processes
Journal title :
Computers and Operations Research
Journal title :
Computers and Operations Research
