A neural network controller for diving of a variable mass autonomous underwater vehicle

In general, traditional controllers used for underwater vehicles are complex, non-adaptive and somewhat slow. On the other hand, it is difficult to accurately determine the hydrodynamic coefficients and the dynamics of underwater vehicles. They are highly nonlinear; therefore, intelligent methods are suitable choice for their control. In this paper, an intelligent neural network method for diving of a variable mass underwater vehicle is presented. The control scheme is capable of learning and adapting to changes in the vehicle dynamics and parameters. The control scheme consists of a gain tuning neural network and a variable gain PID controller. This neural network is trained so that the error between the plant output and reference signal is minimized. The results of this control scheme are compared with a constant gain PID controller. It is shown that the presented control scheme is better and more robust against disturbance than the conventional controller.