部分可观察马尔可夫决策规划──首达目标模型

PARTIALLY OBSERVABLE MARKOV DECISION PROGRAMMING: FIRST PASSAGE PROBLEM

部分可观察马尔可夫决策规划──首达目标模型刘迪芬（湖南师范大学数学系，长沙４１００８１）刘建庸，刘克（中国科学院应用数学研究所，北京１０００８０）ＰＡＲＴＩＡＬＬＹＯＢＳＥＲＶＡＢＬＥＭＡＲＫＯＶＤＥＣＩＳＩＯＮＰＲＯＧＲＡＭＭＩＮＧ：ＦＩＲＳＴＰＡ...

In this paper, we discuss the first passage problem of Markov decision programming with incomplete state information. It was` shown that the problem can be transformed to Markov decision programming with complete state information. We discuss the existence conditions for existence of an optimal stationary policy. Finally, we discuss the algorithm.