Learning control of underwater robotic vehicles

A learning control approach to an underwater robotic vehicle system using neural networks is described. The objective is to have a robust control system with respect to changes in the vehicle dynamics and parameters. It has been observed that dominant vehicle dynamics vary with the vehicle velocity and that the effect of thruster dynamics becomes significant at low velocity in the vehicle control system. When it is necessary to operate the vehicle at low velocity (e.g., stationkeeping and hovering), the effect of thruster dynamics must be considered in the vehicle control system design. Results of computer simulation for pitch and altitude vehicle motion show the effectiveness of the proposed control system