May 30, 2025, Webb Hall 1100

Over the past decades, the control community has increasingly focused its attention on data-driven approaches, moving away from traditional model-based methods. Early developments originated in system identification and adaptive control, while recent efforts have been strongly influenced by reinforcement learning. In this context, a popular modern trend is to compute controllers directly from data via linear matrix inequalities (LMIs), bypassing the need for explicit system identification. Building on this LMI-based paradigm, the talk will introduce the problem of designing a controller for asymptotic reference tracking and disturbance rejection from a single experiment — in short, data-driven output regulation. The approach will be presented for multivariable continuous-time linear time-invariant systems, using recent tools that connect with modern nonlinear estimation methods and reinterpret classical adaptive observers in a state-space setting. Exploiting the internal model principle, the controller design requires only prior knowledge of the frequencies of the unknown disturbance and the measured outputs. The methodology will be illustrated with realistic examples, and relevant open research questions will be discussed.

Alessandro Bosso received the Ph.D. degree in automatic control from the University of Bologna, Bologna, Italy, in 2020. Currently, he is a Tenure-Track Researcher at the Department of Electrical, Electronic, and Information Engineering (DEI), University of Bologna. He has been a visiting scholar at The Ohio State University and at the University of California, Santa Barbara. His research interests include nonlinear adaptive control, hybrid dynamical systems, and the control of mechatronic systems. He is the recipient of a Marie Skłodowska-Curie Postdoctoral Fellowship.