GA optimized fuzzy control of an autonomous mobile robot
In this thesis, a unique two-stage fuzzy controller for an autonomous mobile robot is proposed. An unknown environment presents many challenges for a robot operating with only a limited sensor range. Environments contain objects of various size which can be situated in any location. In some cases, narrow passageways must be navigated by a robot in order for it to obtain its goal. It is with these challenges in mind that the two-stage controller was designed. We introduce a freespace parameter, which is a measure of the restrictiveness of the immediate environment. The incorporation of this parameter allows a robot to move at a speed suitable for navigation in its immediate surroundings. Parameters affecting the robot's performance are identified and optimized by use of a genetic algorithm (GA) and a special training environment. Membership functions, sensor weights, and behaviour-switching threshold values are all optimized through the GA process.