Neural Network Predictive Control of Explorer Class Autonomous Underwater Vehicle

This paper investigates neural network predictive control (NNPC) of Explorer Class Autonomous Underwater Vehicle (AUV) in path following missions. A non-linear dynamic model for the Explorer class AUV at the Australian Maritime College, University of Tasmania is developed based on analytical approach and it is approximated by a neural network via a training process before implemented in a predictive control approach. The fundamental control objectives are to maintain the AUV at desired forward velocity known as surge velocity as well as the position and heading angle in the horizontal plane, which then will be integrated into the cascade control to guide the AUV following a desired path. The effectiveness of the developed NNPC is validated by comparison with conventional Proportional Integral Derivative (PID) controller through numerical simulations.