华南理工大学学报(自然科学版) ›› 2020, Vol. 48 ›› Issue (5): 142-148.doi: 10.12141/j.issn.1000-565X.190253

• 机械工程 • 上一篇    

基于 RBF 神经网络的气动人工肌肉 PID 位置控制

刘凯1 陈伊宁1 吴阳1 王扬威2   

  1. 1. 南京航空航天大学 机电学院,江苏 南京 210016; 2. 东北林业大学 机电工程学院,黑龙江 哈尔滨 150040
  • 收稿日期:2019-05-12 修回日期:2019-12-23 出版日期:2020-05-25 发布日期:2020-05-01
  • 通信作者: 刘凯(1981-),男,博士,副教授,主要从事仿生机器人、数控技术研究。 E-mail:liukai@nuaa.edu.cn
  • 作者简介:刘凯(1981-),男,博士,副教授,主要从事仿生机器人、数控技术研究。
  • 基金资助:
    国家自然科学基金资助项目 (51405229); 江苏省自然科学基金资助项目 (BK20151470,BK20171416)

PID Position Control of Pneumatic Muscle Actuator Based on RBF Neural Network 

LIU Kai1 CHEN Yining1 WU Yang1 WANG Yangwei2   

  1. 1. College of Mechanical and Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016Jiangsu,China; 2. College of Mechanical and Electrical Engineering,Northeast Forestry University,Harbin 150040,Heilongjiang,China
  • Received:2019-05-12 Revised:2019-12-23 Online:2020-05-25 Published:2020-05-01
  • Contact: 刘凯(1981-),男,博士,副教授,主要从事仿生机器人、数控技术研究。 E-mail:liukai@nuaa.edu.cn
  • About author:刘凯(1981-),男,博士,副教授,主要从事仿生机器人、数控技术研究。
  • Supported by:
    Supported by the National Natural Science Foundation of China (51405229) and the Natural Science Founda-tion of Jiangsu Province (BK20151470,BK20171416)

摘要: 搭建了气动人工肌肉静态测试平台,在 0. 1 ~ 0. 8 MPa 气压下进行多次测量试验,对气动人工肌肉进行特性分析,根据理论模型和测试数据建立了数学模型,模型求解精度较好。考虑外负载、气源气压和系统摩擦等因素对数学模型的影响,结合RBF网络的快速学习能力设计了一种基于 RBF 网络的 PID 控制策略。在外负载50 ~200N 的条件下,搭建了气动人工肌肉动态测试平台并进行了多组位置控制试验。结果表明,传统 PID 控制只能在一定的外负载范围内实现较好的位置控制,基于RBF 网络的 PID 控制能自适应调整 PID 参数,且响应速度快,调节时间短,超调量小,能更好地补偿其数学模型误差并实现精确的位置控制。

关键词: 气动人工肌肉, RBF 神经网络, 自适应 PID, 位置控制

Abstract: The static test platform of pneumatic artificial muscle was built,and a series of measurement tests were carried out under pressure of 0. 1 ~ 0. 8 MPa to analyze the characteristics of pneumatic artificial muscle. The mathematical model,which was built based on the theoretical model and test data,shows a high accuracy of solu-tion. In consideration of the influence of external load,gas pressure and system friction on the mathematical mod-el,a PID control strategy based on RBF network was designed with the fast learning ability of RBF network. Un-der the condition of external load F = 50 ~ 200 N,the dynamic test platform was built and a number of position control tests were implemented. The results show that the traditional PID control strategy can only achieve better control accuracy within a certain range of external loads,while the proposed strategy is able to adjust the PID pa-rameters adaptively. Moreover,the proposed PID control strategy has the advantages of higher response speed,shorter adjustment time and smaller overshoot,and it can better compensate the mathematical model error and a-chieve higher control accuracy.

Key words: pneumatic artificial muscle, RBF neural network, adaptive PID, position control

中图分类号: