A Neural Network Online Controller for Autonomous Underwater Vehicle

Abstract

Designing a control law for Autonomous Underwater Vehicle (AUV) has been a considerable challenge from classical and modern control view point. Neural Network based control is seen as an emerging technology for intelligent control of complex system. Here we consider an approach to model the controller for the AUV using Recurrent Neural Networks (RNN). RNN had been selected to model the system as it has very good capability to incorporate the dynamics of the system. The AUV dynamic equations had been modeled to obtain the data required for training the neural network. A controller has been developed which can learn change in the dynamics on the fly. Results have been shown for online learning controller. Back Propagation Algorithm had been used in upgrading the controller weights during online learning control techniques.