基于RBF网络的导弹滑模动态逆控制律设计

Design of sliding model and dynamic inverse control law for a missile based on RBF neural-networks

讨论了一种基于径向基函数(RBF,Radial Basic Function)神经网络的导弹滑模动态逆控制律.导弹的基本控制律采用动态逆方法设计,对慢回路设计神经网络滑模控制器以补偿整个控制系统的不确定性.即用RBF神经网络逼近导弹慢模态不确定性的数学模型,并将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;滑模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 radial basic function（RBF） neural networks based sliding model control and dynamic inverse control approach to a missile was presented.The basic control law was designed by dynamic inversion,and neural networks based sliding model and dynamic inverse controller was designed for the slow loop to compensate the uncertainty of the whole control system.The RBF neural networks were used to approximate the uncertainty of slow states model of missile and the approximation errors of the neural networks were introduced to the design of adaptive adjust law to improve the quality of the systems.Sliding model controller was used to attenuate the uncertainty of model and the approximation errors of the neural networks.The controller could guarantee stability of overall system and attenuate effect of uncertainty of model and approximation errors of neural networks to a prescribed level.Finally,simulation results show the effectiveness of the control method.