PILCO Based Attitude Control for Morphing Aircraft

Morphing aircraft have garnered significant attention due to their adaptability across a broad spectrum of operational scenarios. These aircraft exhibit superior aerodynamic performance compared to conventional fixedwing designs. However, achieving stable attitude control is challenging owing to the system's inherent complex nonlinear dynamics. We explore the implementation of a sophisticated and efficient reinforcement learning approach for attitude control in morphing aircraft. Initially, the nonlinear three-degree-of-freedom dynamics of the 'XunYing' telescopic wing morphing aircraft are modeled using the Newton-Euler method. Subsequently, an attitude control system for the morphing aircraft is designed, utilizing the Probabilistic Inference for Learning Control (PILCO) method. The controller parameters are optimized using a Gaussian process (GP) and the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm, accommodating multiple operational points. Finally, real-time simulations of the morphing aircraft are conducted across various scenarios. The experimental outcomes demonstrate that the proposed control scheme effectively satisfies the attitude control demands of morphing aircraft.