Fuzzy mobile robot navigation and sensor integration

Abstract

A two-layer fuzzy inference system which is capable of integrating the distance readings from different sensors and mapping the integrated results to the motion of the robot is proposed. The navigation system fuzzifies and defuzzifies the readings of the sensors and the left and right clearances of the robot were found as the result of the first-layered fuzzy system. These two signals, together with the direction of the goal relative to the robot, were used as the input of the second-stage inference system, and the linear velocity and the turning rate of the robot were returned as the final output. With the collision avoiding, obstacle following and goal trading behaviors employed inside the second-stage fuzzy inference system, robust mobile robot navigation in unknown environments can be realized. Furthermore, a multiple state property of the second-stage inference system with each state accompanied by a unique fuzzy associative memory was introduced to avoid dead end situations where the mobile robot becomes trapped between an obstacle and the goal.