Research on Local Path Planning for the Mobile Robot Based on QL-anfis Algorithm

Abstract

A local path planning method for the mobile robot is proposed based on QL-anfis algorithm to resolve the common problems in traditional method, such as the dimension disaster, the slow learning speed and deadlock avoidance of QL algorithm. Firstly, state-action pairs <; s, a> are learned by QL algorithm, and the obstacle avoidance table of the robot and the counter of the robot reaching the target are designed to solve the obstacle avoidance and deadlock problem in complex environment. Then anfis is used to learn <; s, a> to get the best <; s, a> instead of the iteration update of Q matrix, which can speed up the speed of path planning and avoid the excessive dimension of Q matrix. Simulation results show that efficiency of obtaining the optimal <; s, a> and planning speed of the optimal or sub-optimal path for the robot can be speeded up by using the learning of anfis system instead of Q matrix replacement.