研究会 (2025 年 11 月 15 日)

Abstract:
1. In the analysis and synthesis of dynamical systems, model-based approaches
   that rely on mathematical models have long been the standard. In contrast,
   data-driven approaches, which utilize measurement data obtained from the
   system, have recently gained attention as they enable analysis and synthesis
   directly from data, even when accurate modeling is challenging. In this talk,
   I present our recent work on data-driven analysis and synthesis, with a focus
   on positive systems. In particular, I discuss data informativity—a property
   of a dataset that indicates whether it contains sufficient information to
   solve a specific analysis or synthesis problem—and show how a linear
   programming (LP) framework can be employed in this context. Finally, I
   illustrate its application to the positive stabilization of networked
   systems.

   Bio:
   Takumi Iwata received the Ph.D. degree in engineering from Nagoya University,
   Nagoya, Japan, in 2024. Since 2024, he has been an Assistant Professor
   (Special Appointment) at the School of Informatics and Data Science,
   Hiroshima University, Higashi-Hiroshima, Japan. His research interests
   include data-driven analysis and control of dynamical systems.

2. Contraction analysis has emerged as a useful tool in nonlinear control,
   providing attractor-independent guarantees on asymptotic behaviors through
   differential conditions that are well suited for computational verification
   and controller synthesis. However, many systems cannot be made contractive,
   which has motivated the search for generalized notions capturing weaker
   convergence properties. One such extension is k-contraction, which can
   guarantee interesting convergence properties without requiring an
   exponential decrease of the distance between any two system trajectories.
     In this talk, I will introduce the concept of k-contraction and discuss
   the types of guarantees it can provide on a system's asymptotic behavior.
   I will then present a set of generalized Lyapunov inequalities that extend
   classical contraction conditions to the k-contraction setting and enable
   the transition from analysis to control design. Finally, I will show how
   quasi-LMI and BMI formulations can be used to synthesize linear feedback
   laws ensuring k-contractivity of the closed-loop dynamics, and show their
   potential application in the context of multi-stability.

   Bio:
   Samuele Zoboli received his B.Sc. degree in Electronics Engineering from
   the University of Modena and Reggio Emilia, Italy, in 2016, and his M.Sc.
   degree in Automation Engineering from the University of Bologna, Italy, in
   2019. He obtained his Ph.D. in Automatic Control from the University of
   Lyon 1, France, in 2023, where his research focused on deriving robustness
   and stability guarantees for systems controlled by deep neural network
   feedback laws using nonlinear control theory. From 2023 to 2025, he was a
   postdoctoral researcher in the MAC team at LAAS-CNRS, France, working on
   partial stability for nonlinear systems and event-triggering strategies in
   deep neural network controllers. Since October 2025, he has been a permanent
   researcher at CNRS, based at LAAS-CNRS. His research interests include
   nonlinear systems control, multi-agent synchronization, and control-oriented
   artificial intelligence.