The performance and management of control surfaces on a near-surface autonomous underwater vehicle (AUV) were examined in terms of hydrodynamics, modelling, and control. Experiments were conducted using a one-quarter scale physical model of the International Submarine Engineering (ISE) Mark II DOLPHIN AUV. The experiments involved extensive tests in wind tunnels and a tow tank and included both force measurement and flow visualization studies. The experiment results were used to mathematically describe the performance of AU V control surfaces for use in simulation. Additionally, based on the new control surface hydrodynamics information, enhancements were made to the vehicle controller. For the various vehicle and controller configurations, the overall vehicle performance was evaluated through simulation using representative manoeuvres and operating conditions.

In straight and level flight, the performance of the forward control surfaces (bowplanes) was well described through existing semi-empirical predictions. Trailing vortices shed by the forward control surfaces significantly affected the performance of the aft control surfaces (sternplanes). This interaction was a strong function of vehicle orientation and was accurately predicted by a simple potential flow model. Significant changes were noted to bowplane performance when the vehicle was oriented with trim or yaw. The influence of waves on the planes, including the dependence on vehicle speed and depth, agreed well with analytical predictions.

In simulation, the effect on vehicle performance of changes in the control surface modelling, control surface configuration, and the controller design were studied. Although significantly different mathematical models of control surface performance were developed, they were found to have similar effects on the overall vehicle performance. Changes to the control surface configuration had a very significant effect on performance. In particular, increasing the bowplane span and adding dihedral to the bowplanes or sternplanes were both found to improve manoeuvring and minimize the effect of waves on the vehicle. Only minor differences in performance were noted between different PD controller implementations. The PD controllers were sensitive to modeling errors and exhibited unstable behavior in one instance. A linear quadratic gaussian controller with loop transfer recovery (LQG/LTR) was more robust and was further improved with the incorporation of sliding mode control.