The feasibility of using a genetic algorithm to optimize a fuzzy fixed rule based autopilot is considered in this paper. Simulation results are presented to show the applicability of the approach. It is concluded such a procedure gives more credence to the resulting design than can be achieved by totally heuristic methods.

As a result of the general recognition given to the robustness qualities of fuzzy control algorithms, studies have been undertaken in the marine field to develop autopilots based on this approach. Indeed, a fuzzy autopilot was recently installed in a small leisure vessel and very successful fullscale sea trials were undertaken. In the past, the main problem in designing a fuzzy autopilot, or any other fuzzy controller, has been the reliance on purely heuristic methods to obtain an optimum solution. To compensate for such shortcomings in the design cycle analytical approaches are now being proposed.

This paper is concerned with an investigation into the use of artificial neural networks in the design of fuzzy autopilots for controlling the yaw dynamics of a modern Royal Navy warship model. Results are presented to show the viability of such an approach and that effective designs can be produced.