北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2013, Vol. 36 ›› Issue (3): 16-19.doi: 10.13190/jbupt.201303.16.005

• 论文 • 上一篇    下一篇

无模型动态摩擦的自回归小波神经补偿控制

褚明, 陈钢, 贾庆轩, 孙汉旭   

  1. 北京邮电大学 自动化学院, 北京 100876
  • 收稿日期:2012-08-27 出版日期:2013-06-30 发布日期:2013-06-30
  • 作者简介:褚 明(1983—), 男, 讲师, 博士, E-mail:chuming_bupt@bupt.edu.cn.
  • 基金资助:

    高等学校博士学科点专项科研基金项目(20110005120004); 总装科技创新工程项目(ZYX12010001); 国家自然科学基金项目(61175080); 国家重点基础研究发展计划项目(2013CB733000)

Compensation Control for Model-Free Dynamic Friction Using Self-Recurrent Wavelet Neural Networks

CHU Ming, CHEN Gang, JIA Qing-xuan, SUN Han-xu   

  1. Automation School,Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2012-08-27 Online:2013-06-30 Published:2013-06-30

摘要:

针对低速伺服系统的摩擦补偿问题,提出一种基于自回归小波神经网络的智能控制算法,无需预知系统的动力模型参数,仅通过闭环位置反馈,网络即能利用极少的神经元和迭代次数实现对非线性摩擦的高精度补偿. Lyapunov稳定性分析结果证明了跟踪误差和网络权值的有界收敛性. 某型机器人关节的伺服实验结果表明,引入自回归小波神经补偿算法后的伺服定位精度得以大幅度提高.

关键词: 无模型, 摩擦补偿, 自回归小波神经网络, 智能控制

Abstract:

An intelligence control algorithm for friction compensation of low-speed servo system is proposed based on self-recurrent wavelet neural networks. There’s of no necessary to predict the system dynamic model parameters,and the high-precision compensation of nonlinear friction is realized by using few neurons and iterations through only position feedback. Lyapunov stability analysis shows the bounded convergence of tracking error and network weights. Also the servo experiments from a robot joint show that the servo positioning accuracy can be greatly improved by introducing the proposed compensation algorithm.

Key words: model-free, friction compensation, self-recurrent wavelet neural networks, intelligence control

中图分类号: