迭代学习控制
计算机科学
控制(管理)
人工智能
控制理论(社会学)
数学
迭代法
算法
数学优化
控制系统
理论(学习稳定性)
机器学习
控制器(灌溉)
在线学习
标识
DOI:10.1080/00207721.2025.2589463
摘要
The existing iterative learning control (ILC) algorithms for 2-D discrete systems are concerned with the strictly repetitive case, and no ILC results for 2-D nonrepetitive uncertain systems (NUS) with communication network have been reported. This article first investigates the robust and convergent properties on a networked P-type ILC law for 2-D NUS, which is subject to iteration-dependent uncertainties (IDUs), random data dropouts and quantised measurements. By using the input signals at previous iteration and the 2-D reference trajectory, the random data dropouts occurred are compensated. A logarithmic quantiser is used to describe the quantised measurements. Consequently, it is theoretically proved that the ILC tracking error and the control input in the sense of mathematical expectation robustly converge to a bounded range, the bound of which relies continuously on the bounds of the IDUs, probability of the random data dropouts, and quantisation density. If these IDUs are progressively convergent along the iteration domain, a precise tracking result can be achieved with mathematical expectation sense. Additionally, an improved ILC law with compensation technique is applicable to the 2-D NUS with coupling input. An illustrative example on the practical thermal process is given to demonstrate the validity of the presented networked P-type ILC law.
科研通智能强力驱动
Strongly Powered by AbleSci AI