Gliding Vehicle Controller Based on Neural Network Compensating Dynamic Inversion Error

Abstract

Aiming at the attitude control problem of the gliding vehicle with uncertainty and strong coupling, a dynamic inversion control method based on RBF neural network compensated inversion error is discussed. Firstly, based on the double time scale separation hypothesis, the controlled aircraft model is divided into a fast and slow state subsystem. Then the dynamic inversion control laws are designed for each of the two subsystems. The weight update rule of the RBF neural network is derived online based on the Lyapunov stability principle to compensate for the inversion error caused by the modeling error and uncertainty. The simulation verifies the effectiveness of the control law. The robustness of the control system after neural network compensation is significantly improved.