Leilei Cui, Ph.D.Assistant Professor at University of New Mexico
Friday, May 1st, 2026, 3:00 - 4:00 PM
Jett Hall, room 109
Title: Seminar Title: Interplay Between Learning and Control: Robustness of Learning Algorithms and Data-Driven Controller Design
ABSTRACT: Learning and control are deeply intertwined disciplines. On one hand, control theory provides a rigorous mathematical framework for understanding, analyzing, and certifying learning algorithms. On the other hand, learning enables data-driven, model-free controller design for dynamical systems.
At the core of most learning methods lie gradient-based optimization algorithms, whose performance in practice is inevitably affected by disturbances such as data noise and computational inaccuracies. These perturbations raise a fundamental question: Can optimization algorithms maintain convergence to near-optimal solutions in the presence of disturbances? To address this question, we leverage tools from control theory—particularly input-to-state stability (ISS)—to rigorously analyze the robustness of gradient-based optimization algorithms under disturbances.
We show that, under mild assumptions, these algorithms converge to a small neighborhood of the optimal solution, provided that the perturbations remain bounded and sufficiently small. In addition, we highlight the power of learning in controller design, especially for time-delay systems, where computing optimal controllers is challenging even when accurate models are available. In such settings, we develop a direct data-driven approach that learns optimal controllers directly from data. Finally, we demonstrate the effectiveness of learning-based control through applications in robotics.
BIO: Dr. Leilei Cui is an Assistant Professor in the Department of Mechanical Engineering at the University of New Mexico. He was a Postdoctoral Associate at the Massachusetts Institute of Technology (MIT) from June 2024 to July 2025. He received his M.Sc. degree in Control Science and Engineering from Shanghai Jiao Tong University, China, in 2019, and his Ph.D. degree in Electrical Engineering from New York University in 2024.
Dr. Cui’s research lies at the intersection of control theory, optimization, and reinforcement learning, with a particular focus on applications to robotic systems. He is the recipient of the Outstanding Paper Award from the journal Control Theory and Technology, the Dante Youla Award for graduate research excellence, and the Alexander Hessel Award for the best Ph.D. dissertation in Electrical Engineering at NYU.
