AbstractThis thesis is concerned with the modelling of non-linear systems, identification of control strategies and designing fuzzy controllers for uncertain systems using fuzzy logic techniques assisted by other conventional methods. The application of the proposed approaches are tested and evaluated on an underwater vehicle, where limited knowledge of the vehicle's dynamic characteristics is available.
In particular the fuzzy and neuro-fuzzy techniques for modelling non-linear systems are reviewed. The combination of function approximation using neuro-fuzzy techniques and new approach to knowledge acquisition using a fuzzy supervised scheduling system is presented.
The identification and modelling of control strategies is also studied. Using fuzzy clustering techniques a systematic approach to identify and qualify the data in H-dimensional space that represents these strategies is presented and evaluated.
The design of robust and tuneable controllers for non-linear systems is also proposed and developed using a combination of the Taguchi design of experiments method and the fuzzy combined scheduling system approach.
The main contributions in this thesis are; the development of a hybrid fuzzy and neuro-fuzzy approach to model non-linear systems with the application to model the yaw dynamics of an underwater vehicle; an algorithmic methodology for identification and modelling of a complex system's control actions with application to construct control strategies for "avoid objects" task; an innovative contribution to the problem of determining the optimal parameters of fuzzy and fuzzy-like PD controllers in terms of robustness and tuning characteristics and the successful implementation of the fuzzy-like PD controller for course- changing and course-keeping in an underwater vehicle.
|Date of Award