Data-Driven Robust Attitude Control for Uncrewed Aerial Vehicles With Time-Varying Input Gain

Abstract

This article investigates the attitude control problem of nonlinear uncrewed aerial vehicles subjected to uncertain internal parameters, unknown external disturbances, multiaxis coupling, and particularly time-varying input gains. A data-driven concurrent-learning (CL)-enhanced robust compensation control method (CL-Robust) is proposed. By leveraging both real-time and historical states, a CL estimator is constructed to achieve the convergence of time-varying input gain estimation without requiring persistent excitation. Based on this data-driven framework, a robust compensation controller with antiwindup mechanism is designed for a quadcopter attitude system as a case study, which retains satisfactory performance despite time-varying control input gains during flight. Through Lyapunov stability analysis and the comparison lemma, it is proven that the CL estimation errors converge exponentially and are uniformly ultimately bounded. Furthermore, the state tracking errors of the closed-loop system can ultimately enter a compact neighborhood of the origin, whose size can be regulated via controller parameters. Numerical simulations, including ablation studies, verify the effectiveness of the proposed CL-Robust controller. Experimental results further demonstrate its superiority over conventional controllers in terms of average tracking overshoot.