Cascade ADRC with neural network-based ESO for hypersonic vehicle

In this paper, active disturbance rejection control (ADRC) based on a neural network has been investigated for the attitude control of the hypersonic vehicle (HV) with uncertain disturbances, which are regarded as a strongly time-varying, nonlinear, and coupled system. The structure of nonlinear state error feedback (NLSEF) with an Extended State Observer (NLSEF+ESO) utilized in ADRC is considered to have good disturbance resistance ability in engineering applications with less dependence on the mathematical model of the system. However, the strong coupling of the HV makes it complicated to separately design ADRC for each channel. In addition, the bandwidth and parameters of the ESO can seriously affect the performance of the ADRC, while jitter occurs when they are not well matched. A cascade active-rejection control scheme is designed by introducing the Radial Basis Function (RBF) Neural Network to substitute the ESO in ADRC, which mitigates the shortcoming of ADRC in addressing the control problems of the MIMO system with coupling disturbances. The NNESO can adapt well to disturbance characteristics through online training and fitting and can effectively reduce the jitter of the control. The stability of the NNESO is proved by Lyapunov stability theory, and the numerical simulations are presented to demonstrate the effectiveness of our theoretical results. In summary, the proposed NNESO-based cascade ADRC is an effective method for solving the problem of HV control with better disturbance resistance.