2013 年 35 巻 3 号 p. 177-187
An optimal maintenance problem is investigated for a system that deteriorates under the influence of a resettable varying environment. Since the internal true state of the system cannot be observed directly, the decision maker needs to determine the action that will minimize the discounted total expected cost over an infinite horizon by inferring the system's exact state on the basis of the available information, such as the current environment condition, the monitors' observations, and the observed system history. This procedure is formulated as a partially observable Markov decision process (POMDP) with the environment as a parameter. The structural properties of the optimal maintenance policy is studied, and a sufficient condition that ensures the optimality of the control limit policy with respect to the system's prior probability and its environment is identified.