观察过程状态数为可数的POMDP的算法

Algorithm of Discounted Model of POMDP with Denumerable States in Observation Process

给出了观察过程状态数为可数时，部分可观的马尔可夫决策规划（ＰＯＭＤＰ）折扣模型的逼近法和算法．算法保证了能求到ε－最优的平稳策略和ε－最优的期望费用函数．

The iteration method for the POMDP with denumerable states in observation process was studied. It developed an operator nH and a policy iteration to compute ε-optimal policies.