面向战机大迎角机动过程的智能学习控制

Considering the strong nonlinearity, aerodynamic uncertainty and channel coupling characteristics of fighter dynamics at high angle of attack, an adaptive maneuver tracking control is proposed based on intelligent learning. By taking the channel coupling into a part of the total disturbance, the model is decomposed into the angle of attack subsystem, the sideslip angle subsystem and the roll angle rate subsystem. Neural networks are used to estimate aerodynamic uncertainties, and the controllers using tracking error feedback and total disturbance estimation feed-forward are designed to obtain the desired control torque. Then the aerodynamic surface deflection and thrust vector deflection are calculated based on daisy chain method. For the neural network weight update, the prediction error is constructed to reflect the estimation performance of the total disturbance, and the composite learning update law is designed combining with the tracking error. The uniformly ultimate boundedness of the closed-loop system is proved based on the Lyapunov method. Simulation and anti-disturbance parameter deviation tests are carried out for the Cobra and Herbst maneuvers, and the results show that the proposed method presents high tracking accuracy and more robust performance.