变体飞行器强化学习自适应抗扰动控制方法

A novel reinforcement learning-based adaptive disturbance rejection control method is proposed to address the challenge of balancing flight safety，disturbance rejection，and optimality in the design of morphing flight vehicle control systems. Initially，a dynamic model of a morphing flight vehicle is established under uncertainty conditions. A fuzzy disturbance observer is designed based on fuzzy control theory to ensure that the estimation errors of system uncertainties and disturbances converge to the neighborhood of the origin. To tackle the issue of balancing flight safety，optimality，and disturbance rejection，the optimality problem of high-order nonlinear systems is transformed into an optimization problem for control signal design in each subsystem. The Hamilton-Jacobi-Bellman equation is solved using a reinforcement learning framework. An actor-critic strategy based on neural networks is designed to address the nonlinear characteristics of the system that are difficult to manage during equation solving. In backstepping design，the actor network generates the control signal，the critic network evaluates control performance，and the barrier Lyapunov function method is employed for stability analysis to ensure the system’s stability，optimality，and state constraints. Finally，the effectiveness of the proposed method is validated through simulation.