This paper presents an intelligent navigation method for navigation of a mobile vehicle in unknown environments. The proposed navigator consists of three modules: Obstacle Avoidor, Environment Evaluator and Navigation Supervisor. The Obstacle Avoidor is a fuzzy controller whose rule base is learnt through reinforcement learning. A new and powerful training method is proposed to construct the fuzzy rules automatically. The Navigation Supervisor determines the tactical requirement of avoiding obstacles or moving towards the goal location at each action step so that the vehicle can achieve its task without colliding with obstacles. The effectiveness of the learning method and the whole navigator are verified by simulation.
展开▼
