Learning pipelines bridging reinforcement learning and adaptive control
Event details
| Date | 28.10.2026 |
| Hour | 15:15 › 16:15 |
| Speaker | Dr Feiran Zhao, postdoctoral researcher at Automatic Control Lab in ETH Zurich |
| Location | |
| Category | Conferences - Seminars |
| Event Language | English |
Abstract:
The adjacent fields of reinforcement learning (RL) and adaptive control share the same objectives, yet they are separated by a wide cultural gap. In this presentation, I attempt to bridge this gap for the linear quadratic regulator (LQR) problem, which serves as a cornerstone and the benchmark for both fields. I begin by discussing different learning pipelines, including direct and indirect (model-based) approaches, as well as episodic and online (adaptive) approaches. Despite the extensive literature spanning several decades, numerous problems remain unsolved. For instance, RL methods are seldom concerned with closed-loop stability certificates or efficient implementations, while the adaptive control community has dedicated minimal effort to optimality. We address the data-driven LQR problem in an adaptive setting, which entails online recursive algorithms and closed-loop data, and we seek both algorithmic as well as closed-loop certificates. Our approach encompasses different variations of policy gradient methods going also beyond mere LQR formulations, employs a novel covariance parameterization, and leverages different kind of regularizations. Finally, all our theoretical results are validated through simulations and experiments in diverse domains, demonstrating the computational and sample efficiency of our method.
Biography:
Feiran Zhao is currently a postdoctoral researcher at Automatic Control Lab in ETH Zurich, working with Prof. Florian Dörfler. Before that, he received the Ph.D. Degree in Control Science and Engineering from the Tsinghua
University, China, in 2024. His research interests include reinforcement learning, control theory, optimization, their intersections, and applications on energy, robotics, and mobility systems.
The adjacent fields of reinforcement learning (RL) and adaptive control share the same objectives, yet they are separated by a wide cultural gap. In this presentation, I attempt to bridge this gap for the linear quadratic regulator (LQR) problem, which serves as a cornerstone and the benchmark for both fields. I begin by discussing different learning pipelines, including direct and indirect (model-based) approaches, as well as episodic and online (adaptive) approaches. Despite the extensive literature spanning several decades, numerous problems remain unsolved. For instance, RL methods are seldom concerned with closed-loop stability certificates or efficient implementations, while the adaptive control community has dedicated minimal effort to optimality. We address the data-driven LQR problem in an adaptive setting, which entails online recursive algorithms and closed-loop data, and we seek both algorithmic as well as closed-loop certificates. Our approach encompasses different variations of policy gradient methods going also beyond mere LQR formulations, employs a novel covariance parameterization, and leverages different kind of regularizations. Finally, all our theoretical results are validated through simulations and experiments in diverse domains, demonstrating the computational and sample efficiency of our method.
Biography:
Feiran Zhao is currently a postdoctoral researcher at Automatic Control Lab in ETH Zurich, working with Prof. Florian Dörfler. Before that, he received the Ph.D. Degree in Control Science and Engineering from the Tsinghua
University, China, in 2024. His research interests include reinforcement learning, control theory, optimization, their intersections, and applications on energy, robotics, and mobility systems.
Practical information
- General public
- Free