迭代学习控制
规范(哲学)
趋同(经济学)
控制理论(社会学)
最优控制
数学
计算机科学
数学优化
应用数学
控制(管理)
人工智能
政治学
法学
经济
经济增长
出处
期刊:IEEE Transactions on Automatic Control
[Institute of Electrical and Electronics Engineers]
日期:2024-01-01
卷期号:: 1-16
标识
DOI:10.1109/tac.2024.3362857
摘要
To reduce the need for high gains (reduced control weighting) for fast convergence in Norm Optimal Iterative Learning Control (NOILC) the paper presents a simple data-driven mechanism for accelerating the convergence of low gain feedback NOILC controllers. The method uses a modification to the reference signal on each NOILC iteration using the measured tracking error from the previous iteration. The basic algorithm is equivalent to a gradient iteration combined with a NOILC iteration. The choice of design parameters is interpreted in terms of the spectrum of the error update operator and the systematic annihilation of spectral components of the error signal. The methods apply widely, including continuous and discrete-time end-point, intermediate point and signal tracking. The effects of parameter choice are revealed using examples. A robustness analysis is presented and illustrated by frequency domain robustness conditions for multi-input, multi-output discrete-time tracking, and robustness conditions for end-point problems for state space systems. Finally, the algorithm is extended to embed a number of gradient iterations within a single NOILC iteration. This makes possible the systematic manipulation of the spectrum, providing additional acceleration capabilities with the theoretical possibility of arbitrary fast convergence.
科研通智能强力驱动
Strongly Powered by AbleSci AI