A feedforward neural network for identification and adaptive control of autonomous underwater vehicles

This paper describes a method for accurate identification of dynamical systems using backpropagation neural network. A network structure is proposed to realize the identification network, with which the motion of the controlled object can be simulated. This network is introduced into a neural-network-based control system called “self-organizing neural-net-controller system” (SONCS), which has been developed as an adaptive control system for autonomous underwater vehicles (AUVs). On the advantage of the network´s simulating capability, the controller in the SONCS can be quickly adapted through the process called “imaginary training”. The efficiency of the proposed identification network is examined through the application of heading control of an AUV