On the use of neural techniques for modeling and control the diving motion of the underwater vehicle Phantom 500

Neural techniques for modeling and control of the diving motion of an underwater vehicle are presented in this article. The system under consideration is a remotely operated vehicle the Phantom 500. The direct training method is used to elaborate the direct and inverse neural models with the consideration of system non linearity and unmeasured disturbances presence. Three different inverse model based structures are applied for control: An open loop structure, the direct inverse control, and two closed loop structures, the internal model control and additive feedforward control. The performances are compared to a conventional PI cascade control.