Lien Teams : cliquer ici.
– Titre : Generative control: zero-shot learning for feedback systems
– Résumé : Data-driven control has transformed modern automation by enabling controllers to be designed directly from system data, reducing the reliance on detailed modeling and allowing rapid deployment in complex, uncertain, or poorly understood environments. Yet, traditional data-driven approaches often remain limited to linear systems, require extensive tuning, and depend on large datasets. This talk presents Generative Control, a novel framework inspired by Generative AI that addresses these limitations by enabling zero-shot learning of feedback controllers. The approach leverages prior knowledge and in-context learning to synthesize effective control laws directly from minimal interaction with a previously unseen system. By doing so, it generates interpretable, efficient, and adaptable controllers without the need for system-specific training data. We will discuss the theoretical foundations, practical implementation, and potential applications of Generative Control, demonstrating how this paradigm opens new possibilities for fully automated, data-driven control of complex and nonlinear systems.
– Biographie : Simone Formentin is an Associate Professor at Politecnico di Milano, Italy. He received his B.Sc. and M.Sc. degrees cum laude in Automation and Control Engineering from Politecnico di Milano (2006, 2008) and his Ph.D. cum laude in Information Technology in a joint program with Johannes Kepler University of Linz, Austria (2012). He held postdoctoral positions at EPFL, Switzerland, and the University of Bergamo, Italy. He currently chairs the IEEE Technical Committee on System Identification and Adaptive Control, and represents IFAC on social media for the TC on Robust Control. He is an Associate Editor for Automatica and the European Journal of Control. His research focuses on system identification and data-driven control, with applications in automotive and financial systems.