Technology Of Motion Planning For Autonomous Underwater Vehicle Based On Fuzzy Theory And Reinforcement Learning

First, in my paper , to the question about underwater robot local path planning , the real-time path planning system based on fuzzy control is designed .This system solves the path planning problem in the complicated environment .The simulation results verified the way presented in this paper is effective.Second, Consider that the fuzzy planning system is depending on the experience and knowledge of operator. It is insufficient that the fuzzy rule may exist, and confirm that the perfect fuzzy regular work load is very large. So, consider improving the method of planning. Reinforcement learning is a simple kind of study mechanism , and it only need one strengthen signal to judge the quality of movements, does not need many people' s intervention very much . However, practise and realize the local path planning system simply with reinforcement learning, the time for study is certainly long , and study bad possibility . The pluses and minuses practised on the basis of the fuzzy theory and Reinforcement learning, so conbine the two together, make the elementary fuzzy rule first, then the robot utilizes reinforcement learning to be practised and study independently on the elementary and fuzzy and regular basis, and adjust the corresponding fuzzy rule. Inthis way, can reduce a large amount of artificial work . Meanwhile, the environment that the robot is in is not unalterable, many unknown factors can't foresee for people, and it can be studied according to the concrete environment while utilizing this way underwater robot. To revise the improper function, it has strengthened the adaptive capacity of the robot. Because has already confirmed the elementary fuzzy rule in advance, the time for study of robot will be shortened greatly . The simulation results verified the way presented in this paper is effective.