| 引用本文: | 牟方厉,吴 丹,董云飞.具有多层感知器力矩补偿的机器人自抗扰控制[J].控制理论与应用,2020,37(6):1397~1405.[点击复制] |
| Mou Fangli,WU Dan,DONG Yunfei.Active disturbance rejection control with multilayer perceptron compensating network for robot systems[J].Control Theory and Technology,2020,37(6):1397~1405.[点击复制] |
|
| 具有多层感知器力矩补偿的机器人自抗扰控制 |
| Active disturbance rejection control with multilayer perceptron compensating network for robot systems |
| 摘要点击 1897 全文点击 880 投稿时间：2019-05-30 修订日期：2019-10-14 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2019.90397 |
| 2020,37(6):1397-1405 |
| 中文关键词 机器人 自抗扰控制 控制器 神经网络 |
| 英文关键词 robots active disturbance rejection control (ADRC) controllers neural networks |
| 基金项目 国家自然科学基金项目(51575306)资助. |
|
| 中文摘要 |
| 本文针对机器人系统的控制特性, 提出了一种基于自抗扰控制(ADRC)的关节控制算法, 该算法可以克服
传统控制算法中存在的如系统抗干扰能力弱, 控制性能受限于建模精度, 动态性能与稳态性能难以兼顾, 控制律设
计较为复杂等问题. 针对受控系统特性给出了一套实际控制器的完整设计方法与参数整定方法, 并根据控制性能指
标设计优化函数完成了最优控制参数的优化, 在系统参数辨识的基础上利用多层感知器(MLP)设计了对建模不确
定性的补偿网络. 数值仿真和实验结果均表明该算法能够实现机器人快速稳定的轨迹跟踪, 具有良好的控制精度
与很强的抗干扰能力, 此外该算法不依赖于精确的系统模型, 降低了实际设计和应用的难度, 具有很好的工程应用
价值. |
| 英文摘要 |
| In this paper, a joint controller based on active disturbance rejection control (ADRC) method is designed for
robot manipulators. The proposed method can overcome the deficiencies that exist in traditional methods such as weak
anti-disturbance ability, control performance limited by the modeling accuracy, dynamic performance and steady-state
performance difficult to balance, and complex control law design. A complete controller design method and parameter
tuning method are given for the physical robot system. Besides, different optimization functions according to the control
objective are designed to optimize the control parameters. A deep multilayer perceptron (MLP) is designed to compensate
the modeling uncertainties based on system parameter identification. Both numerical simulation and experiment indicate the
validity of the proposed method, the results show that the ADRC controller with MLP compensating network can realize
the robot rapid and stable trajectory tracking, and have excellent control precision and strong anti-disturbance ability.
Meanwhile, as not depending on accurate system model, that reduces the difficulty of the practical design and application,
makes the proposed method having great value in engineering application. |
|
|
|
|
|