Neural Networks Based Attitude Decoupling Control for AUV with X-Shaped Fins

Attitude control in three-dimensional space for AUV (autonomous underwater vehicle) with x-shaped fins is complicated but advantageous. Yaw, pitch and roll angles of the vehicle are all associated with deflection angle of each fin while navigating underwater. In this paper, a spatial motion mathematic model of the vehicle is built by using theorem of momentum and angular momentum, and the hydrodynamic forces acting on x-shaped fins and three-blade propeller are investigated to clarify complex principle of the vehicle motion. In addition, the nonlinear dynamics equation which indicates the coupling relationship between attitude angles of vehicle and rotation angles of x-shaped fins is derived by detailed deduction. Moreover, a decoupling controller based on artificial neural networks is developed to address the coupling issue exposed in attitude control. The neural networks based controller periodically calculates and outputs deflection angles of fins according to the attitude angles measured with magnetic compass, thus the vehicles orientation can be maintained. By on-line training, twenty four weights in this controller converged according to index function.