Faculty Recruiting Support CICS

Learning-Augmented Linear Quadratic Control

28 Oct
Friday, 10/28/2022 1:00pm to 2:00pm
Lederle Graduate Research Center, Room A215; Zoom
Data Science Deep Dive

Abstract: Recently, augmenting learning-based predictions, advice, and black-box AI tools to classic online decision-making algorithms attracts growing interests. Integrating AI techniques into control systems, on the one hand, provides a new approach to handle uncertainties caused by renewable resources and human behaviors, but on the other hand, creates practical issues such as reliability, privacy, and scalability, etc. to the AI-integrated control algorithms. 

In this talk, I will present novel problems raised in designing learning-augmented model predictive control algorithms. First, I will introduce a problem in linear quadratic control, where untrusted AI predictions of system perturbations are available. We show that it is possible to design an online algorithm with performance guarantees even with large prediction error. Second, I will present a nonlinear setting wherein pre-trained black-box RL algorithms are available. Next, I will show how these problems relate to practical applications in smart grids and demonstrate that the learning-augmented methods help improve previous classical results. Finally, I will discuss interesting potential research directions down the road.

Bio: Tongxin Li is a tenure-track assistant professor in the School of Data Science (SDS), The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen). Prior to joining SDS, he received his Ph.D. degree from the California Institute of Technology (Caltech) in Computing and Mathematical Sciences in 2022. He graduated from CUHK in a dual-degree program and obtained a B.Eng. in information engineering, a B.Sc. in mathematics and a M.Phil. in information engineering. His research interests focus on interdisciplinary topics in smart grids, machine learning, control, and optimization, with applications to power systems and sustainability. In particular, he is interested in developing trustworthy artificial intelligence and machine learning techniques that improve the sustainability, robustness, scalability, privacy, and resilience of smart grids. He has been invited to give talks at various international conferences and meetings such as the INFORMS Annual Meeting and INFORMS Applied Probability Conference and has published articles on various peer-reviewed journals and conferences. During his Ph.D., he interned as an applied scientist at AWS security in the summers of 2020 and 2021. He has participated in various projects on power systems in collaboration with NREL, Pasadena Water and Power, and Caltech Facilities. He is a recipient of the 2021 Impact Grants from the Resnick Sustainability Institute.

Join the Seminar

The Data Science Deep Dive is free and open to the public. If you are interested in giving a talk, please email Mohammad Hajiesmaili or Adam Lechowicz. Note that in addition to being a public lecture series, the Data Science Deep Dive is also a seminar (CompSci 692K, Algorithms with Predictions Seminar) that can be taken for credit.