Development of a fuzzy logic guidance system for mobile robots

The primary focus of this dissertation is the development and experimental verification of a Fuzzy Logic Guidance System for a Mobile Robot capable of autonomous control on a primitive forest trail. Because of the lack of systematic design methodologies for FLC systems, a necessary part of the work was to research and develop methods for designing and optimizing FLC systems. We developed two design methodologies during my course of study for my Ph.D. program. The first method we developed a procedure embedding human expert knowledge into a fuzzy logic controller. Computer simulations were used to determine a human response for a given set of sensor data. An optimization routine is then used to train a fuzzy logic controller to match the responses of a human controller. The second method provides an optimization algorithm that can be used to design either conventional or fuzzy logic control systems. We show how a fuzzy logic control system can be made to be equivalent to a linear conventional control system with the proper rule base, choice of membership functions and defuzzification procedure. This equivalent control system and optimization algorithm provides the foundation by which conventional control theory can be extended to develop nonlinear, fuzzy logic control systems. These methods are used to develop of a controller for a mobile robot used to skid logs. The fuzzy logic controller gives the robot the capability of operating autonomously. This fuzzy logic controller uses a sensor system that returns relatively sparse data. The sensors are organized into modules, each capable of stand-alone operation, and combined in a hierarchical control system. Performance testing of the controller in the forest environment are presented and discussed.