FUZZY LOGIC AND REINFORCEMENT LEARNING BASED APPROACHES FOR MOBILE ROBOT NAVIGATION IN UNKNOWN ENVIRONMENT

摘要

Fuzzy logic controller promises an efficient solution for the mobile robot navigation. However, it is difficult to maintain the correctness, consistency and completeness of the fuzzy rule-base constructed and tuned by a human operator. Reinforcement learning method is a type of machine learning. This approach is often used in the field of robotics. It aims of learning the fuzzy rules automatically and to generate a control law for a mobile robot in unknown environment when we assume that the only obtained information is a scalar signal which is a reward or punishment. The process of learning consists to improve the choice of actions to maximize rewards. It is an intelligent navigation method for an autonomous mobile robot. In this paper, the Q-learning algorithm of reinforcement learning and fuzzy controllers are used for the mobile robot navigation. In order to improve the mobile robot performances, an optimization of fuzzy controllers will be discussed for the robot navigation; based on prior knowledge introduced by a fuzzy inference system so that the initial behavior is acceptable. Simulation results show the obtained behaviors using the three approaches and the effectiveness of the optimization method presenting significant improvements of the robot behaviors and the speed of learning. The results are compared and discussed.