New evaluation criteria in the Markov decision processes

摘要

Herein, we discuss the optimal evacuation route in the instance of fire in the framework of Markov decision processes. In general, it is impossible to recognize a priori where a fire will occur in advance. Therefore, it is crucial to consider an evacuation route that does not depend on the location of the fire. In this study, we assume that the fire location is not fixed and that it follows a uniform distribution. Accordingly, we propose new criteria to define an optimal evacuation route.