The collision-free motion of robot with Fuzzy neural network

Abstract

This article use intelligent method based on fuzzy neural network for real-time robot path planning. The constructed fuzzy neural network has six inputs and two outputs. The six inputs are the obstacle distance information detected by the five sonar sensors installed on the robot and the target point relative to the robot orientation information; Obstacle information obtained from the five sensors is integrated into the obstacle distance information in three directions of the left, the front and the right by the second and third layer of Fuzzy neural network. Outputs of the network are the robot's speed and angle to run the next step. The inputs of the five obstacles are measured by the sonar sensors in real time. Target points in the sampling time relative to the direction and position of the robot can be calculated by the location of the robot and the target point in the world coordinate system. Computer simulation shows that by robot's own perception of the environment, through the constant interaction with the environment by learning to repeatedly adjust the environment and its own model, robot can run without collision in unknown environment. Compared with other path planning methods, this method greatly enhances the adaptive capacity of the robot.