Real-time decision making for autonomous mobile robot using evolution strategy and anytime sensing

This paper presents a new real-time decision making method for an autonomous mobile robot based on an integration of evolutionary action search anal anytime sensing. The anytime sensing proposed in this paper increases a sensing information, which is necessary for the action search over the search time. With this property, it could be possible for the robot to search one of the 'good' actions in less time because the computational amount increases step by step according to the increase of the sensing information by introducing the anytime sensing. As a result, we can make the search process faster in the time scale. We also analyze the proposed anytime sensing from viewpoint of Kullback-Leibler information criterion and show the usefulness through some simulation results.