一种基于神经网络补偿动态逆误差的方法

An Adaptive Neural Network in Compensating the Modeling Error in Dynamic Inversion Control

讨论了一种基于神经网络自适应补偿动态逆误差的方法 ,并应用于超机动飞机控制器设计中。飞机的基本控制律采用非线性动态逆方法进行设计 ,对于模型不准确导致的逆误差采用神经网络进行在线补偿。仿真结果表明 ,采用神经网络补偿逆误差 ,弥补了非线性动态逆要求精确数学模型的缺点 ,而且可以简化动态逆控制律的设计 。

A neural network based method for compensating dynamic inversion error is introduced in this paper and is applied in the control of a supermaneuvering aircraft. The base control law is designed using the nonlinear dynamic inversion, and single hidden layer (SHL) neural networks are used to cancel the inversion error induced by the inaccuracy of the system model. Simulation results show that through adaptively canceling the inversion error by neural networks, the limitation of needing accurate mathematical model in the dynamic inversion method can be released. Besides this, the design of dynamic inversion control law can be greatly simplified by utilizing the online neural networks and the control effects are improved.