A Separation Principle for data-driven predictive control - the role of models and beyond

Abstract
In this talk I shall  discuss recent approaches to data-driven predictive control, providing a bridge between behavioral and stochastic approachers. As a byproduct I will formulate  a separation principle which elucidate the role of models, and their uncertainty in the control design problem, shedding light  on the model-free vs model-based dilemma. 

Our framework is constructive in the sense that, not only  commonly used  regularization follows from first principles, but it also provides data-driven closed form expressions for regularization terms, thereby eliminating the need for costly and often unrealistic tuning of hyper-parameters.  

Extensions to the time-varying/adaptive setup and non-linear predictive control shall be discussed. 

 

Short Bio
Alessandro Chiuso received his Master degree (Laurea) in  1996, from the University of Padova and the PhD (Dottorato) in 2000 from the  University of Bologna. He has been long term visitor with several international institutions, among which  Washington University St. Louis, KTH Stockholm, UCLA, ETH Zurich.  He joined the University of Padova  as an Assistant Professor in 2001, Associate Professor in 2006 and then Full Professor since 2017. He currently serves as Editor (System Identification and Filtering) for Automatica. He has served as an Associate Editor for several prestigious journals (among which Automatica, IEEE Transactions on Automatic Control, IEEE Transactions on Control Systems Technology, European Journal of Control) and  has been active in Conference organization (among with General Chair of SYSID 2021, IPC co-chair of SYSID 2024).  He is a Fellow of IEEE (Class 2022). His research interests are mainly at the intersection of Machine Learning, Estimation, Identification and Control.