引用本文:陈诚,黄剑,刘磊,伍冬睿.基于数据驱动的气动柔性关节Takagi–Sugeno模糊系统建模与预测控制[J].控制理论与应用,2022,39(4):633~642.[点击复制]
CHEN Cheng,HUANG Jian,LIU Lei,Wu Dong-rui.Data-driven Takagi–Sugeno fuzzy system modeling and predictive control of a pneumatic flexible joint[J].Control Theory and Technology,2022,39(4):633~642.[点击复制]
基于数据驱动的气动柔性关节Takagi–Sugeno模糊系统建模与预测控制
Data-driven Takagi–Sugeno fuzzy system modeling and predictive control of a pneumatic flexible joint
摘要点击 2230  全文点击 524  投稿时间：2021-02-24  修订日期：2022-01-14
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2021.10156
  2022,39(4):633-642
中文关键词  柔性关节  预测控制  数据驱动  模糊系统
英文关键词  flexible joint  predictive control  data-driven  fuzzy systems
基金项目  国家自然科学基金项目(61873321, U1913207), 辽宁省自然基金资助计划项目(2020–KF–22–01)资助, 国家重点研发计划政府间/港澳台重点专项 项目(2017YFE0128300)资助.
作者单位E-mail
陈诚 华中科技大学 hust_cheng@hust.edu.cn 
黄剑* 华中科技大学 huang_jan@mail.hust.edu.cn 
刘磊 华中科技大学  
伍冬睿 华中科技大学  
中文摘要
      针对气动柔性关节动态特性复杂、难以实现高精度控制的问题, 提出一种基于Takagi–Sugeno (T–S)模糊系 统的预测控制方法. 首先, 应用MBGD–RDA算法对T–S模糊系统进行离线训练, 该算法结合了机器学习中的小批量 梯度下降算法、正则化、Droprule和AdaBound算法. 其次, 基于模糊集相似性度量方法, 对训练得到的T–S模糊系统 的模糊集进行剪枝, 对模糊规则进行合并, 简化T–S模糊系统结构. 最后, 设计了基于T–S模糊系统的单层神经网络 预测控制器. T–S模糊系统参数和预测控制器参数均能实现在线更新, 且基于李雅普诺夫理论的稳定性分析保证了 系统的稳定性. 仿真结果验证了基于数据驱动的T–S模糊系统的高精度预测性能, 且结构简化后的T–S模糊系统在 规则数减少的情况下仍能维持较高的预测精度. 实际实验中, 所提控制方法最大跟踪误差小于3?, 而传统的模糊逻 辑控制器最大跟踪误差大于5?, 这表明所提控制算法显著提升了对柔性关节的跟踪控制精度.
英文摘要
      A predictive control approach based on Takagi–Sugeno (T–S) fuzzy system is proposed to address complex dynamics of the pneumatic flexible joint and achieve high-precision control. Firstly, a data-driven training algorithm, MBGD–RDA which combines mini-batch gradient descent, regularization, DropRule, and AdaBound of machine learning, trains T–S fuzzy systems offline. Secondly, based on the similarity measure of fuzzy sets, T–S fuzzy systems’ fuzzy sets and rules are pruned and merged to simplify the structure. Finally, based on the T–S fuzzy system, a single layer neural network (SNN) predictive controller is proposed. Parameters of the T–S fuzzy system and parameters of the SNN controller can be updated online, and the stability of the system is guaranteed based on the Lyapunov theory. Simulation results validate the high-precision prediction capability of the T–S fuzzy system based on data-driven. The T–S fuzzy system with a simplified structure can maintain high prediction accuracy with fewer rules. In real-world experiments, the maximum tracking error of the proposed method is less than 3?, while the maximum tracking error of the traditional fuzzy logic controller is more than 5?, which indicates that the proposed method significantly improves the tracking control accuracy of the flexible joint.