Real-time dynamic fuzzy Q-learning and control of mobile robots

In this paper, a dynamic fuzzy Q-learning (DFQL) method capable of navigating a mobile robot efficiently is presented. The fuzzy rules for navigation can be generated and tuned automatically based on Q-learning. Continuous-valued states and actions are handled using fuzzy reasoning. Prior knowledge can be embedded into the fuzzy rules for rapid and safe learning. The eligibility trace method is employed in our algorithm, leading to faster learning and alleviating the experimentation-sensitive problem where an arbitrarily bad training policy might result in a non-optimal policy. Experimental results demonstrate that the robot is able to learn the appropriate navigation policy with a few trials.