Derivative-free Learning of Interpretable Control Policies (Prof. Ali Mesbah, Department of Chemical and Biomolecular Engineering, University of California Berkeley)
Systems & Control Seminar
Abstract
The design of learning-based and robust/stochastic control policies generally relies on the selection of several continuous, discrete, and/or categorical design choices and tuning parameters that often influence the closed-loop control performance and constraint satisfaction in non-trivial and non-convex ways. In this talk, we will discuss how the control policy search problem can be cast as a derivative-free optimization problem, wherein by selecting optimization-based control law parameterizations the size of the design space can be significantly reduced, yielding easier to implement and more interpretable control policies. We will then focus on Bayesian optimization (BO) and its extensions that can effectively handle black-box performance functions and constraints, derived from noisy closed-loop observations, whose derivatives may not be accessible due to the implicit nature of optimization-based control laws. Finally, we will discuss example applications of BO approaches to control policy learning in the context of biomanufacturing systems for deep space missions, as well as embedded control of point-of-care biomedical systems.
Biographical information
Ali Mesbah is Associate Professor of Chemical and Biomolecular Engineering at the University of California at Berkeley. Before joining UC Berkeley, Dr. Mesbah was a senior postdoctoral associate at MIT. He holds a Ph.D. degree in Systems and Control and a Master’s degree in Chemical Engineering, both from Delft University of Technology. Dr. Mesbah is a senior member of the IEEE and AIChE. He serves on the Editorial Boards of the IEEE Transactions on Control Systems Technology, IEEE Control Systems Letters, and IEEE Transactions on Radiation and Plasma Medical Sciences. Dr. Mesbah is recipient of the Alexander von Humboldt Research Fellowship in 2023, the Best Application Paper Award of the IFAC World Congress in 2020, the AIChE's 35 Under 35 Award in 2017, the IEEE Control Systems Outstanding Paper Award in 2017, and the AIChE CAST W. David Smith, Jr. Publication Award in 2015. His research interests lie at the intersection of optimal control, machine learning, and applied mathematics, with applications to learning-based analysis, optimization, and predictive control of materials processing and manufacturing systems.
Termin
27. Feb. 202416:00 - 17:00
