Online Sensor Selection in Reinforcement Learning Environment

摘要

More sensors do not necessarily result in more appropriate state descriptions, so that a mobile robot has to select an appropriate set of sensors besides learning a state-action function in a reinforcement learning environment. We present a multi-armed bandit formulation of the problem and apply it to mobile robot navigation task. We modified the reinforcement comparison method to suit our problem and build a system where the selection of optimal set of sensors and the learning of state-action functions are done simultaneously. Our approach is evaluated on a Khepera robot simulator and the results reveal that our approach works well as an integrated learning system to identify the best set of sensors and reduce learning time.