Nonlinear Modelling and Control of an Autonomous Underwater Vehicle

This paper highlights nonlinear modelling and control of the Hammerhead AUV. To model the AUV, a multi-layer perceptron neural network structure has been chosen. Data from actual trials is utilised for training and validating the neural network model. A genetic algorithm (GA) based nonlinear model predictive control (MPC) algorithm is also implemented to the identified model and results are shown for two waypoint following scenarios involving step changes in reference heading. It is demonstrated that the neural network is able to model the system well and the GA-based nonlinear MPC autopilot is quite capable of dealing with nonlinear nonconventional models.