Neural Learning Control of Fighter Aircraft at High-Angle-of-Attack Maneuvers With Online Recorded Data and Disturbance Observer

Considering the strong nonlinearities, aerodynamic uncertainties, and external disturbances exhibited by fighter aircrafts during high angle-of-attack (AOA) maneuvers, this article proposes a neural learning control strategy based on online recorded data and a disturbance observer (DOB). The channel coupling is incorporated into the total disturbance, and the model is decomposed into the AOA, sideslip, and roll-rate subsystems. The neural network approximates system uncertainties, while an online recorded data scheme constructs the prediction error combined with the tracking error to form the adaptive law. On this basis, the DOB is designed to estimate composite disturbances in real time and incorporated into the online recorded data model as a feedforward compensation to obtain the desired control moments. An allocation optimization model is further established to compute aerodynamic control surface deflections and thrust vectoring angles online. Lyapunov analysis proves the uniform ultimate boundedness of the closed loop system. Finally, simulations of the Herbst maneuver under disturbance torques verify that the proposed strategy achieves high learning performance and accurate tracking of maneuver commands.