AI钱塘论坛（第36期）
主题：Higher-order regulation for robot localization and machine learning
时间：2025年10月26日上午10：00-11：30
地点：第二教学楼南楼228会议室
主持人：曹九稳 教授
曹明博士
Ming Cao has since 2016 been a professor of networks and robotics with the Engineering and Technology Institute (ENTEG) at the University of Groningen, the Netherlands, where he started as an assistant professor in 2008. Since 2022 he is the director of the Jantina Tammes School of Digital Society, Technology and AI at the same university. He received the Bachelor degree in 1999 and the Master degree in 2002 from Tsinghua University, China, and the Ph.D. degree in 2007 from Yale University, USA. From 2007 to 2008, he was a Research Associate at Princeton University, USA. He worked as a research intern in 2006 at the IBM T. J. Watson Research Center, USA. He is the 2017 and inaugural recipient of the Manfred Thoma medal from the International Federation of Automatic Control (IFAC) and the 2016 recipient of the European Control Award sponsored by the European Control Association (EUCA). He is an IEEE fellow and a distinguished lecturer of the IEEE Control Systems Society. He is a Senior Editor for Systems and Control Letters, an Associate Editor for IEEE Transactions on Automatic Control, IEEE Transaction of Control of Network Systems and IEEE Robotics & Automation Magazine, and was an associate editor for IEEE Transactions on Circuits and Systems and IEEE Circuits and Systems Magazine. He is a member of the IFAC Council. His research interests include autonomous robots and multi-agent systems, complex networks and decision-making processes.
摘要：Regularization methods are commonly used in a range of engineering and computer science problems. In this talk, I propose a set of new high-order regularization techniques (HR) for two scenarios. When mobile robots move indoors in huge flat spaces or narrow corridors, their locations are difficult to calculate because most of the numerical methods encounter the so-called ill-conditioned problems, namely some location-related matrices become close to singular whose inverse are hard to compute. A typical method to solve ill-conditioned problems is regularization, and a classical regularization method is the Tikhonov regularization. I will show that the Tikhonov regularization can be seen as a low order case of our HR method, which is superior in approximating some ill-conditioned inverse problems. Moreover, we construct one a priori criterion which improves the numerical stability of the ill-conditioned problem. Since most of the regularization solutions are biased, we also discuss bias-correction techniques. Such a new HR method can provide insight into the less understood connection between regularization and learning using neural networks. More specifically, regularization can be regarded as an approximation in terms of inverse mapping. We further verify the performance of the proposed HR method by solving a classic control problem in reinforcement learning.