Autonomous Underwater Vehicles Attitude control using Neuro-Adaptive Generalized Dynamic Inversion

Abstract

A novel two loops control architecture is presented for motion control of Autonomous Underwater Vehicle (AUVs). The outer (slow) positional loop incorporates classical Proportional-Derivative control to generate reference pitch and yaw tilting commands based on the positional errors. The reference attitude commands are fed to the inner (fast) attitude control loop, which utilizes Neuro-Adaptive Generalized Dynamic Inversion (NAGDI) control to generate the elevator and rudder deflections. The baseline Generalized Dynamic Inversion (GDI) control composed of particular and auxiliary parts. The particular part is designed by dynamically scaled generalized inversion of the attitude error dynamics, and the auxiliary part is constructed via a Lyapunov control function to provide global asymptotic stability to the angular body rate dynamics. A discontinuous control based on the concept of Sliding Mode Control (SMC) theory is augmented with baseline GDI. The modulation gain of discontinuous term is made adaptive to avoid chattering. Furthermore, online estimation of the unknown nonlinear attitude dynamics is achieved through radial basis function neural networks to obtain the resultant NAGDI control law. The proposed control law guarantees semi-global practically stable attitude tracking. Computer simulations are conducted on a six degrees of freedom simulator of the Monterey Bay Aquarium Research Institute AUV under nominal and perturbed marine environments.