具有重构功能的基于RBF神经网络直接自适应飞控系统

反馈线性化方法鲁棒性研究一直是近些年来的研究热点。针对基于反馈线性化方法的鲁棒控制，提出了一种将反馈线性化和RBF神经网络直接自适应控制相结合的综合控制方法，利用李亚普诺夫稳定性定理推导了神经网络权值的自适应规律，保证了闭环系统的稳定性。该方法应用于某型号飞机舵面故障状态仿真，结果表明RBF神经网络自适应控制方法补偿作用显著，相当于系统具有一定的重构功能。

Reconfigurable Direct Adaptive Flight Control System Based on RBF Neural Network

As a common nonlinear control method, feedback linearization requires accurate plant mathematical model, which is usually difficult to be obtained in reality, so that the robustness of feedback linearization is not guaranteed. For overcoming this shortcoming, we present a new method. This method integrates feedback linearization and direct adaptive control based on RBF(Radial Basis Function) neural network. An adaptive weight adjustment rule is derived using Lyapunov theory to guarantee system closed-loop stability. This method was applied to the simulation of one type of fighter with aileron failure. Results show that the compensation of this method to the control system is significant; it is equivalent to adding certain reconfigurable capability to the system.