变结构神经网络自适应鲁棒控制  

Variable Structure Neural Network Adaptive Robust Control

作　　者：陈杰[1,2] 李志平[1,2] 张国柱[1,2] 

机构地区：[1]北京理工大学自动化学院,北京100081 [2]北京理工大学复杂系统智能控制与决策教育部重点实验室,北京100081

出　　处：《自动化学报》2010年第1期174-178,共5页Acta Automatica Sinica

基　　金：北京市重点实验室项目(SYS100070417);国家杰出青年科学基金(60925011)资助~~

摘　　要：针对一类不确定非线性系统,提出一种变结构神经网络自适应鲁棒控制(Variable structure neural network adaptive robust control,VSNNARC)方法.其中变结构神经网络用于在线辨识系统未知非线性函数,该网络利用节点激活与催眠技术进行动态调节,减小网络规模与计算量;自适应鲁棒控制用于网络权值学习与系统建模误差及外部扰动补偿.采用Lyapunov稳定性分析法,给出网络权值自适应律的形式以及鲁棒控制项的设计方法.该方法不仅能保证系统的稳定性,也能保证系统具有很好的瞬态性能.将该方法应用到转台伺服系统的位置跟踪控制中,实际运行结果表明,该方法使系统具有很强的鲁棒性及良好的跟踪效果.A variable structure neural network adaptive robust control （VSNNARC） is proposed for a class of uncertain nonlinear SISO systems. A neural network which is adjusted by using nodes activation and passivation techniques to minish the size of the neural network and computation load is used as an opproximator for the system unknown nonlinear functions. And the adaptive robust control is used for the weight learning and compensation to the modeling error and extern disturbances. The adaptation law of neural networks weights and the design method of robust controller are given out based on the Lyapunov stability analysis. Furthermore, the proposed controller can guarantee not only global stability but also transient performance. Finally, the controller is applied to a position tracking system of a turntable. The experimental results show that the system can perform good tracking and has strong robustness.

关 键 词：自适应鲁棒 神经网络 双轴转台 函数逼近器 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]