Reward Function Using Inverse Reinforcement Learning and Fuzzy Reasoning

A reward function estimated with inverse reinforcement learning has been used to determine a method for controlling a robot. Inverse reinforcement learning requires observed sequences of actions to estimate a reward function. Few models of the sequences give the optimal motion of the robot; therefore, a suboptimal one may be given. However, the suboptimal sequences include some errors and ambiguities. In this paper, we propose a method for quantifying the ambiguity of the reward function, which is designed with inverse reinforcement learning using fuzzy reasoning.