引用本文:杜章铭,周超,王硕.带有姿态估计反馈和数据驱动模型的微动平台角位移控制[J].控制理论与应用,2025,42(3):473~481.[点击复制]
DU Zhang-ming,ZHOU Chao,WANG Shuo.Angular control with attitudes estimation feedback and data-driven model for fine motion platform[J].Control Theory and Technology,2025,42(3):473~481.[点击复制]
带有姿态估计反馈和数据驱动模型的微动平台角位移控制
Angular control with attitudes estimation feedback and data-driven model for fine motion platform
摘要点击 40  全文点击 2  投稿时间：2023-02-25  修订日期：2024-10-27
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2023.30081
  2025,42(3):473-481
中文关键词  微动平台  角位移控制  姿态估计  直接逆控制  神经网络
英文关键词  fine motion platform  angular displacement control  attitude estimation  direct inverse control  neural network
基金项目  国家自然科学基金项目(61873268, 62033013)资助.
作者单位E-mail
杜章铭 中国科学院自动化研究所 dzm000@126.com 
周超* 中国科学院自动化研究所 chao.zhou@ia.ac.cn 
王硕 中国科学院自动化研究所  
中文摘要
      对角位移的高精度控制是微动平台在实际应用中的重要任务, 当前主流微动平台在进行角位移控制时, 对 机构的精密性较为依赖, 成本较高且不便于根据具体任务定制化. 对此, 本文提出了一种基于姿态估计反馈和数据 驱动模型的微动平台角位移控制方法, 使机械精度较差的机构也可实现高精度的角位移致动. 首先, 基于数据驱动 的方式对致动机构的正、逆运动学模型进行神经网络建模, 避开了复杂的运动学求解和自由度间解耦问题; 其次, 根据稀疏实测角位移和正运动学模型预测提供角位移的联合估计反馈. 基于此, 通过直接逆控制(DIC)和比例–积分 (PI)结合的方式实现2自由度角位移复合控制; 最后, 通过角位移跟踪控制实验对所提方法的有效性进行了验证.
英文摘要
      High-precision angular displacement actuation is an important task in applications of fine-motion platform, for which the major current products highly depends on expensive high-precision mechanism with low flexibility of customization. Therefore, this article proposes a controller based on a combination of attitude estimation feedback and datadriven models, making it possible for platforms with low mechanical precision to perform fine angular actuation. Firstly, the forward and inverse kinematics models are built with neural-networks, avoiding complex calculations for the solution of kinematic problem and the decoupling between motion axes. Secondly, a joint estimation of angular displacement is designed for feedback, based on sparse actual measurements and forward-model predictions. Then the direct-inverse-control (DIC) and the proportional-integral (PI) method are combined to perform 2-DOFs angular control. Finally, the effectiveness of proposed method is verified with angular tracking experiments.