Safety-Constrained Tracking Control of AUVs under Composite Disturbances via Adaptive Neural Disturbance Learning

Abstract

This paper investigates safe trajectory tracking for fully actuated autonomous underwater vehicles under composite disturbances. A deeply coupled disturbance model is first derived to characterize the effects of hydrodynamic uncertainties, environmental disturbances, and actuator disturbances. An adaptive radial basis function (RBF) neural network is then introduced for online disturbance learning. Based on the learned disturbance and a conservative residual bound, a robust control Lyapunov function-control barrier function quadratic programming (CLF-CBF-QP) controller is developed to achieve tracking control while enforcing tracking-error safety and actuator constraints. Simulation results validate the effectiveness of the proposed method.