关键词: 折叠翼, 垂直起降飞行器, 动力学建模, 模型预测控制, 神经网络

Abstract: To address the modeling challenges and loss of control accuracy caused by pronounced aerodynamic variations and strong coupling induced by changes in rotational inertia during multi-mode flight, this paper proposes a flight control method that integrates neural-network-based aerodynamic prediction with nonlinear model predictive control (NMPC) for a foldable-wing morphing vertical take-off and landing aircraft. An all-angle-of-attack aerodynamic model incorporating rotor slipstream effects is established based on strip theory, together with a multi-rigid-body dynamic model parameterized by the wing-folding angle. These models characterize the evolution of the aerodynamic forces and flight dynamics during wing morphing. To reduce the computational burden imposed by model complexity on online optimization, a neural network is trained offline to approximate the aerodynamic forces and moments over the prescribed flight envelope and is subsequently embedded in the NMPC prediction model. A control allocation strategy that accounts for the wing-folding angle is also developed, and closed-loop trajectory tracking is achieved by solving the NMPC optimal control sequence. Simulation results demonstrate that the proposed method maintains smooth attitude transitions and accurate trajectory tracking along the prescribed reference trajectory in the presence of aerodynamic model bias and sensor noise. These results confirm its effectiveness in handling strong nonlinearities and time-varying parameters, while also indicating its potential for engineering applications in complex mission scenarios.

Key words: folding wing, vertical takeoff and landing aircraft, dynamics modeling, model predictive control, neural network