Multivariable Neurofuzzy Control of an Autonomous Underwater Vehicle

This paper discusses the application of a novel multivariable control technique to the problem of autonomous underwater vehicle (AUV) autopilot design. Based on an adaptive neural network structure a multivariable Sugeno style fuzzy inference system is tuned to produce a course-changing and roll minimizing autopilot. Simulation results, performed using a full non-linear six degree of freedom model, illustrate the effectiveness of this new approach when compared to a more traditional control approach which makes no provision for the inherent cross coupling between AUV yaw and roll channels.