Monday, November 16, 2026 | 11:00 a.m.
Department of Electronics, Information and Bioengineering - Politecnico di Milano
Schiavoni Seminar Room (Building 20A)
Speaker: Sergio Lucia (TU Dortmund University)
Abstract
Model predictive control has become a powerful framework for decision-making in dynamic systems, but its computational demands still limit its widespread deployment, especially for nonlinear systems subject to uncertainty. Learning-based methods promise to close this gap, most prominently by training neural networks that approximate the optimal control law. However, such approximations typically require large amounts of data, and the safety, robustness, and stability properties of the original MPC formulation are not automatically inherited by the resulting closed-loop system.
In this talk, Prof. Lucia will first briefly review how probabilistic and deterministic verification techniques can certify relevant closed-loop properties despite approximation errors and additional uncertainty. The main part of the talk will then focus on a different route to fast and reliable control: instead of learning the solution of the optimization problem, we learn the iterative steps of a constrained optimization solver. He will present a framework for learned iterative solvers that can be trained in a self-supervised manner without pre-computed optimal solutions. He will also discuss recent extensions that make the approach scalable to long prediction horizons and applicable to mixed-integer formulations. In this way, learned iterative solvers combine the speed of learning-based methods with the accuracy and reliability of optimization-based control.
The proposed techniques will be illustrated through simulation case studies involving both linear and nonlinear systems.
Short Bio
Sergio Lucia received the M.Sc. degree in Electrical Engineering from the University of Zaragoza, Spain, in 2010 and the Dr.-Ing. degree in Optimization and Automatic Control from the TU Dortmund University, Germany, in 2014. He then joined the Otto-von-Guericke University Magdeburg and visited the Massachusetts Institute of Technology as a Postdoctoral Fellow. From 2017 to 2020, he served as an Assistant Professor at TU Berlin before returning to TU Dortmund University as an Associate Professor in 2020. Since 2023, he has been a Full Professor and Head of the Chair of Process Automation Systems at TU Dortmund University.His research focuses on decision making under uncertainty and the integration of machine learning with control theory. Additionally, he serves as an Associate Editor for the Journal of Process Control, and as Chair of the IFAC TC on Optimal Control.