Neural network feedback linearization target tracking control of underactuated autonomous underwater vehicles with a guaranteed performance

Abstract

In this paper, an adaptive neural network-based feedback-linearizing controller is proposed to force an underactuated autonomous underwater vehicle (AUV) to follow a target with desired relative distance and angles in three-dimensional space. Towards this end, a new input-output model of the AUV is developed by defining a proper set of look-ahead position variables in a cone-shaped area in the front of the vehicle. Then, a nonlinear saturated PID-type controller is designed such that target tracking errors converge to a neighborhood of the zero while the prescribed transient and steady-state performance characteristics including desirable overshoot/undershoot, convergence rate and steady-state accuracy are guaranteed. This technique effectively prevents undesirable transient state behaviour of the vehicle while reducing the risk of actuators saturation. A combination of multi-layer neural networks (NNs) and adaptive robust control (ARC) techniques is used to handle the compensation of model uncertainties including unknown parameters, time-varying environmental disturbances induced by waves and ocean currents and NN approximation errors. A Lyapunov-based stability synthesis is presented to ensure the semi-global uniform ultimate boundedness of the proposed closed-loop control system with a guaranteed prescribed performance. Finally, simulation results are given to show the effective performance of the AUV for practical applications.