Adaptive Neural Nonlinear Dynamic Inversion Control of Aircraft in Icing Conditions

Aircraft icing poses a significant challenge when flying in adverse weather conditions. It not only degrades performance and handling quality but also introduces control issues that can lead to potentially catastrophic crashes. The accumulation of ice on wings and tails alters the static and dynamic derivatives of the aircraft, introducing uncertainties into the system. In light of this situation, the development of an autopilot system capable of mitigating these uncertainties becomes of paramount importance. In this paper, we present the dynamic response of aircraft under icing conditions. Subsequently, we propose a novel approach, the Adaptive Neural Nonlinear Dynamic Inversion Adaptive(AN-NDI) Flight Control System, designed to effectively reject uncertainties associated with icing. To model these uncertainties, we employ Radial Basis Functions (RBF) that capture the complex and non-linear nature of the problem. The center of the RBF is continuously updated using streaming flight data, ensuring accurate representation of the icing effects. The desired control inputs are then extracted using the Lyapunov function, ensuring robust stability and performance. Our primary focus is on maintaining altitude and forward velocity at trim conditions during the cruise phase, a critical aspect of flight safety. Through simulation results, we demonstrate the efficacy of the designed control system in handling icing situations adeptly. The estimated states of the aircraft validate the effectiveness of the proposed approach, showcasing its capability to counteract the adverse effects of icing. This research contributes to the advancement of autopilot systems by addressing the challenges posed by aircraft icing. By leveraging neural networks, adaptive control techniques, and continuously updating the RBF center with streaming data, our approach offers a promising solution for enhancing flight safety and performance in icing conditions.