基于改进CMAC的高阶柔性直线系统跟踪控制  

Tracking control based on improved CMAC for high-order flexible linear system

作　　者：于士贤 付兴建[1] 

机构地区：[1]北京信息科技大学自动化学院,北京100192

出　　处：《北京信息科技大学学报（自然科学版）》2018年第1期18-22,共5页Journal of Beijing Information Science and Technology University

基　　金：国家自然科学基金资助项目(61573230)

摘　　要：针对高阶柔性直线系统的跟踪控制问题,设计了一种将改进的小脑模型关节控制器(Cerebellar Model Articulation Controller,CMAC)神经网络与传统控制器并联的控制方案。基于参数自适应算法和梯度下降法提出了一种高斯基函数CMAC神经网络的参数自适应更新律;利用符号距离的概念减少了神经网络的输入维数;并利用自适应遗传算法对神经网络的学习率进行了优化。仿真实验表明改进的高斯基函数CMAC具有比传统CMAC更好的学习能力,控制方案实现了高阶柔性直线系统的无差跟踪控制。In this paper,a tracking control method combining improved CMAC（ Cerebellar Model Articulation Controller） with traditional controller for high-order flexible linear system is given. Based on parameter adaptive algorithm and gradient descent method,a learning algorithm of CMAC with Gauss basis function neural network is presented. The signed distance is adopted to reduce the dimension of the neural network input; and the learning rate of neural network is optimized by using adaptive genetic algorithm. Finally,the simulation results show that the improved CMAC with Gauss basis function has better learning ability than the traditional CMAC,and the designed controller realizes the steady zeroerror tracking control for the high-order flexible linear system.

关 键 词：小脑模型关节控制器 高斯基函数 自适应遗传算法 高阶柔性直线系统 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]