Caltech Home > PMA Home > Master Calendar > Computing + Mathematical Sciences Lecture
open search form
Wednesday, January 15, 2020
4:00 PM - 5:00 PM
Annenberg 213

Computing + Mathematical Sciences Lecture

Data-Driven Control for Societal-Scale Cyber-Physical Systems
Yuanyuan Shi, Fifth-year PhD student, Department of Electrical and Computer Engineering, University of Washington,
Speaker's Bio:
Yuanyuan Shi is a fifth-year PhD student in the Department of Electrical and Computer Engineering at the University of Washington, advised by Prof. Baosen Zhang. She is also pursuing a master’s degree in Statistics at UW. Her research interests lie in cyber-physical and energy systems, from the perspective of machine learning, optimization, and control. In the past, she has interned at DeepMind, JD.com, and Doosan GridTech. She has been named as one of the Rising Stars in EECS by MIT in 2018.

Decisions on how to best operate large-scale cyber-physical systems (CPS) such as power system, large commercial buildings and supply chain system are becoming increasingly challenging because of the growing system complexity, rich interaction between agents, and various model and environmental uncertainties. The key to solving these challenges lies in a better understanding of the physical system and through the interrogation of increasingly available operations data.

In the first part of this talk, I will present a new data-driven control framework with Input Convex Neural Network (ICNN) for system identification, which obtains both good predictive accuracy and tractable computational complexity. It shows preferable performance in building HVAC control, leveraging the prior knowledge that the underlying physics are in fact convex. Besides the single system complexity, the control of societal-scale CPS is further complicated by the interactions of multiple strategic agents. In the second part, I will introduce our recent work on analyzing the dynamics of learning agents under limited information feedback in Cournot games, which is an essential model for electricity market. We prove the general convergence of no-regret learning algorithms to the unique Nash Equilibrium, and compare the convergence rate under different information feedback.

For more information, please contact Zelda Wong by phone at 626.395.2464 or by email at zeldaw@caltech.edu.