Events

CILVR Seminar: Learning and control of system dynamics from sensory input

Speaker: Oumayma Bounou

Location: 60 Fifth Avenue, Room 7th floor open space
Videoconference link: https://nyu.zoom.us/s/94752404862

Date: Wednesday, April 2, 2025

Identifying an effective model of a dynamical system from sensory data—and using it for future state prediction and control—is highly valuable in fluid dynamics, robotics, and engineering. However, it remains a challenging task. Borrowing from the Koopman formalism, we present a method to learn a representation of the underlying state of a system directly from raw measurements. In this learned representation space, we constrain the dynamics model to be linear, enabling the resulting representation and model to be used in a linear control framework. In this talk, I will present our approach for learning such dynamics models and their validation through simulation experiments of classical dynamical systems.