基于RBF神经网络的导弹鲁棒动态逆控制  

RBF Neural-Network-Based Robust Dynamic Inverse Control for a Missile

作　　者：杨志峰[1] 雷虎民[1] 李庆良[1] 李炯[1] 

机构地区：[1]空军工程大学导弹学院,陕西713800

出　　处：《宇航学报》2010年第10期2295-2301,共7页Journal of Astronautics

基　　金：航空科学基金(20090196005)

摘　　要：讨论了一种基于神经网络的导弹鲁棒动态逆控制方法。导弹的基本控制律采用动态逆方设计,针对存在动态逆误差的慢回路设计神经网络鲁棒逆控制器。用RBF神经网络逼近导弹慢模态数学模型,并把逼近误差引入到网络权值的调节律以改善系统的动态性能;鲁棒控制器用于减弱模型不确定性及神经网络的逼近误差对跟踪精度的影响。所设计的控制器不仅保证了闭环系统的稳定性,而且使模型不确定性及神经网络的逼近误差对跟踪精度的影响减小到给定的性能指标。最后通过仿真分析,验证了该方法的有效性。A neural network based robust dynamic inverse control approach to a missile is presented in this paper. A basic control law is designed by dynamic inversion,and a neural network-based robust dynamic inverse controller is designed for the slow loop including dynamic inverse errors. The model of slow models missile is approximated by RBF neural networks,and adjust laws used to improve the quality of the system are designed by approximation errors of the neural networks. Then a robust controller is introduced to attenuate uncertainties of model and the effect of approximation errors of the neural networks on tracking accuracy. The controller can guarantee stability of the overall system and attenuate the effect of uncertainties of model and approximation errors of the neural networks on tracking accuracy to a prescribed level. Finally, simulation results show the effectiveness of the control method.

关 键 词：导弹 动态逆控制 神经网络 鲁棒控制 

分 类 号：V249.122.2[航空宇航科学与技术—飞行器设计]