Ronghu Chi
Title: Data-driven Iterative Learning Control
Abstract: Artificial intelligence (AI) has experienced a great resurgence with a key characteristic of “Learning”. In general case, learning refers to the action of system to adapt and change its behavior based on input/output (I/O) observations. Many control systems have this learning ability to respond to changes in its environment, including feedback control systems, adaptive control systems, or any type of artificial neural network equipped with a weight update algorithm. The primary goal of our work is centered on iterative learning control (ILC). The term “Iterative” indicates a kind of action that requires the dynamic process be repeatable, i.e., the dynamic system is deterministic and the tracking control is repeatable over a finite tracking interval. It is worth pointing out that the ILC was originally proposed for nonlinear uncertain systems directly using I/O data for the controller design without requiring the exact knowledge of the system model and thus is classified as data-driven control. However, many design and analysis methods of ILC systems still require some model information. For example, optimal ILC depends on the accuracy of the linear model of the system to guarantee the convergence. Therefore, our work mainly introduces some new design and analysis methods of data-driven ILC. First, the basic idea of data-driven ILC is introduced in this work, and then the latest advances of data-driven ILC are discussed with some significant problems of non-repetitive uncertainty, incomplete information, specified points tracking, higher-order learning algorithm, event-triggered mechanism, etc. Our work provides new insights into the learning control system design and analysis, so that many other learning methods can be incorporated in the control systems as a new branch of AI.
Biography: Ronghu Chi, Professor of the School of Automation and Electronic Engineering of Qingdao University of Science and Technology. He received the Ph.D. degree in systems engineering from Beijing Jiaotong University, Beijing China, in 2007. He was a Visiting Scholar with Nanyang Technological University, Singapore from 2011 to 2012 and a Visiting Professor with University of Alberta, Edmonton, AB, Canada from 2014 to 2015. He also serves as Deputy Secretary General of Data-driven Control, Learning and Optimization Professional Committee of China Automation Society, member of Process Control Professional Committee of China Automation Society, Director of Shandong Automation Society, Guest editor of International Journal of Automation and Computing. He was awarded the “Taishan scholarship” in 2016. He has presided over 3 general projects of NSFC, 5 provincial and ministerial level vertical projects and more than 10 horizontal projects and participated in 3 National Longitudinal topics. He has published more than 100 academic papers, including more than 70 refereed journal articles. He has successively won 2 provincial and ministerial scientific research awards and 4 municipal department level scientific research awards. More than 10 invention patents have been applied for and authorized. His current research interests include: data-driven control, iterative learning control, multi-agent systems, batch process control, etc.
Abstract: Artificial intelligence (AI) has experienced a great resurgence with a key characteristic of “Learning”. In general case, learning refers to the action of system to adapt and change its behavior based on input/output (I/O) observations. Many control systems have this learning ability to respond to changes in its environment, including feedback control systems, adaptive control systems, or any type of artificial neural network equipped with a weight update algorithm. The primary goal of our work is centered on iterative learning control (ILC). The term “Iterative” indicates a kind of action that requires the dynamic process be repeatable, i.e., the dynamic system is deterministic and the tracking control is repeatable over a finite tracking interval. It is worth pointing out that the ILC was originally proposed for nonlinear uncertain systems directly using I/O data for the controller design without requiring the exact knowledge of the system model and thus is classified as data-driven control. However, many design and analysis methods of ILC systems still require some model information. For example, optimal ILC depends on the accuracy of the linear model of the system to guarantee the convergence. Therefore, our work mainly introduces some new design and analysis methods of data-driven ILC. First, the basic idea of data-driven ILC is introduced in this work, and then the latest advances of data-driven ILC are discussed with some significant problems of non-repetitive uncertainty, incomplete information, specified points tracking, higher-order learning algorithm, event-triggered mechanism, etc. Our work provides new insights into the learning control system design and analysis, so that many other learning methods can be incorporated in the control systems as a new branch of AI.
Biography: Ronghu Chi, Professor of the School of Automation and Electronic Engineering of Qingdao University of Science and Technology. He received the Ph.D. degree in systems engineering from Beijing Jiaotong University, Beijing China, in 2007. He was a Visiting Scholar with Nanyang Technological University, Singapore from 2011 to 2012 and a Visiting Professor with University of Alberta, Edmonton, AB, Canada from 2014 to 2015. He also serves as Deputy Secretary General of Data-driven Control, Learning and Optimization Professional Committee of China Automation Society, member of Process Control Professional Committee of China Automation Society, Director of Shandong Automation Society, Guest editor of International Journal of Automation and Computing. He was awarded the “Taishan scholarship” in 2016. He has presided over 3 general projects of NSFC, 5 provincial and ministerial level vertical projects and more than 10 horizontal projects and participated in 3 National Longitudinal topics. He has published more than 100 academic papers, including more than 70 refereed journal articles. He has successively won 2 provincial and ministerial scientific research awards and 4 municipal department level scientific research awards. More than 10 invention patents have been applied for and authorized. His current research interests include: data-driven control, iterative learning control, multi-agent systems, batch process control, etc.