Gradient Descent-Based Adaptive Neural Network Backstepping Control for High-Speed Aircraft

The uncertainty of parameters, such as moment of inertia, product of inertia, aerodynamic torque coefficient, and external disturbance, will cause potential risks and problems to the aircraft dynamics model. In this paper, the challenge of attitude control in high-speed aircraft with model uncertainty is tackled, introducing an adaptive neural network-driven backstepping control strategy. Firstly, the dynamics model of the aircraft is analyzed in detail, and the uncertainty factors are modeled and described. Then, an adaptive neural network backstepping controller is designed, which uses radial basis neural network and gradient descent algorithm to estimate and compensate the uncertainties in the model in real time, so as to achieve accurate attitude control of the aircraft. Finally, the proposed method is validated and evaluated by numerical simulation experiments. The results show that the backstepping controller of the adaptive neural network can effectively suppress the model uncertainty in the attitude loop of the aircraft.