Current Institution: University of Illinois at Urbana-Champaign
Bio: Yingying Li is currently a postdoc researcher at the Coordinated Science Laboratory (CSL) and the Department of Industrial and Enterprise Systems Engineering (ISE) at the University of Illinois at Urbana-Champaign. She received her Ph.D. at Harvard University in 2021 and her B.S. degree at the University of Science and Technology of China (USTC) in 2015, both in Applied Mathematics. She was also a research intern at MIT-IBM Watson AI Lab in the summer of 2020. Her research interests lie in the intersection of control, machine/reinforcement learning, and optimization, with applications in smart grids, smart cities, and robotics. Her work was selected as the Editor’s Choice by Automatica. She was also selected as a Future Digileader by the KTH Royal Institute of Technology in 2019.
Abstract: Safe Adaptive Learning for Linear Quadratic Regulators with Constraints
This paper considers single-trajectory adaptive/online learning for linear quadratic regulator (LQR) with an unknown system and constraints on the states and actions. The major challenges are two-fold: 1) how to ensure safety without restarting the system, and 2) how to mitigate the inherent tension among exploration, exploitation, and safety. To tackle these challenges, we propose a single-trajectory learning-based control algorithm that guarantees safety with high probability. Safety is achieved by robust certainty equivalence and a SafeTransit algorithm. Further, we provide a sublinear regret bound compared with the optimal safe linear policy. When developing the regret bound, we also establish a novel estimation error bound for nonlinear policies, which can be interesting on its own. Lastly, we test our algorithm in numerical experiments.