| 引用本文: | 李国军,王亢,卢甜甜,董齐芬,陈东杰.带有误差修正项的自适应学习控制[J].控制理论与应用,2021,38(2):275~286.[点击复制] |
| LI Guo-jun,WANG Kang,LU Tian-tian,DONG Qi-fen,CHEN Dong-jie.Adaptive learning control with rectifying error actions[J].Control Theory and Technology,2021,38(2):275~286.[点击复制] |
|
| 带有误差修正项的自适应学习控制 |
| Adaptive learning control with rectifying error actions |
| 摘要点击 1690 全文点击 767 投稿时间：2020-04-22 修订日期：2020-08-16 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2020.00208 |
| 2021,38(2):275-286 |
| 中文关键词 学习控制 自适应控制 初始修正 收敛 |
| 英文关键词 learning control adaptive control initial rectifying convergence |
| 基金项目 浙江省自然科学基金项目(LQ18G010001), 北京东方计量测试研究所刘尚合院士专家工作站静电研究基金项目(BOIMTLSHJD20182001)资助. |
|
| 中文摘要 |
| 本文讨论了二阶系统在任意初态偏差下的自适应控制问题, 借助学习控制及其初始修正的思想, 提出了两
种带有修正初态偏差功能的自适应控制策略: 一阶吸引子控制器和零阶吸引子控制器. 两种控制器都是将整个控
制过程分成若干个等长时间的子过程, 在每个子过程中控制算法都会进行误差校正和参数学习. 其中, 一阶吸引子
控制器在每个子过程中同时修正所有状态偏差; 而零阶吸引子控制器在每个子过程中先修正高阶状态偏差, 再修
正低阶状态偏差. 并且两种控制器在控制过程中, 都利用反正切函数对控制量进行连续化处理, 解决了控制过程中
的颤振问题. 最后, 通过计算机仿真验证了算法的有效性. |
| 英文摘要 |
| This paper discusses the adaptive control for second-order systems with any initial state error. By means of
learning control and initial shift rectifying, two adaptive control strategies are presented in the paper: first-order attractor
controller and zero-order attractor controller. Both of them divide the whole control process into many equal time subprocesses.
The parameter learning is carried out in each subprocess to speed up the convergence speed. Furthermore, the
each controller has a term that doesn’t stop rectifying the initial shift of each sub-process until the system has completed
tracking. Among them, the first-order attractor controller rectifies all the initial shifts of each sub-process at the same time;
but the zero-order attractor controller corrects the initial velocity shift of each sub-process at first, and then corrects the
displacement shift of the moment at which the controller has finished rectifying the initial velocity shift. In the control
process, a continuous controller is designed through using the arctangent function that can avoid the chattering problem.
Finally, two examples are given and the tracking performance of the proposed approach is demonstrated through computer
simulations. |
|
|
|
|
|