Two-Stage Fuzzy Logic-Based Controller for Mobile Robot Navigation

Abstract

A two-stage fuzzy inference system for a sonar sensor-based mobile robot is presented. The motion of the robot has been simulated as well as experimentally tested. Target seeking, obstacle avoidance, and emergency behaviors have been defined in a hierarchical fashion. The first stage of the fuzzy control system outputs an angular velocity based on the robot's five sonar sensor readings. The angular velocity and a measure of average object proximity, or freespace, is then fed into the second stage of the fuzzy controller and linear velocity is output. The freespace value is a measure of the openness of the immediate environment. Incorporating this variable in the fuzzy controller enables the robot to adapt to the environment and move at a more suitable speed. In our tests, the robot was able to react both quickly and correctly to the perceived sensor data as it navigated its way through a variety of environments. Testing has identified several system parameters that, when modified, can significantly affect the performance of the controller. A solid foundation is laid for possible optimization of these parameters via a learning algorithm