About the speaker
Filippo Fabiani is an Assistant Professor (RTD-B) in the DYSCO (DYnamical Systems, Control and Optimization) Research Unit at the IMT School for Advanced Studies Lucca (IT). He received the B.Sc. degree in Bio-Engineering, the M.Sc. degree in Automatic Control Engineering, and the Ph.D. degree (cum laude) in Automatic Control, all from the University of Pisa (IT), in 2012, 2015, and 2019 respectively. In 2017-2018, he visited the Delft Center for Systems and Control at TU Delft (NL), where in 2018-2019 he spent one year as post-doctoral Research Fellow. Successively, in 2019-2022 he was Post-Doctoral Research Assistant in the Control Group at the Department of Engineering Science, University of Oxford (UK). His research to date has been at the intersection of machine learning, game theory and control engineering, and concentrates on developing both data-driven theoretical tools and algorithmic methods for modern control design. These include optimal control of multi-agent and uncertain systems such as power and traffic network, and emerging control applications in smart grids and smart cities.
Abstract
As traditional energy systems undergo a revolution, the renewed interest in data driven learning techniques is reshaping conventional paradigms for their identification and control, where the convergence of accessible, large datasets and advanced machine learning offers transformative opportunities. On the other hand, accompanying these learning-based identification and control methodologies with rigorous performance certificates is an essential requirement to ensure safe operations in the energy systems of the future. Motivated by their flexibility and almost inexpensive evaluation, this two-part seminar presents recently developed neural network-based approaches to address two fundamental problems in the systems-and-control community: system identification and control. In the first part we consider the problem of designing a neural network surrogate of an unknown dynamical system from a finite number of data points such that the model obtained is also suitable for optimal control design. To this end we propose a specific, yet easy-to-train, neural network architecture that, under a careful choice of its weights, produces a hybrid system model with structural properties that are highly favourable when used as part of a finite horizon optimal control problem. Specifically, we show that optimal solutions with strong optimality guarantees can be computed via nonlinear programming. In the second part, instead, we consider the design of stabilizing neural network controllers for both deterministic and uncertain systems. Specifically, we focus on the problem of certifying the approximation quality of a neural network with rectified linear units in replicating the action of traditional optimization-based controllers, such as MPC. We develop an offline methodology requiring the construction and solution of mixed-integer linear programs so that, in case the resulting optimal solution meets certain problem-dependent value, the stability of the closed-loop system with neural network controller is guaranteed.