When: Thursday 20th of February, 11:30am AEDT
Where: This seminar will be partially presented at the ACFR seminar area, J04 lvl 2 (Rose St Building) and partially online via Zoom. RSVP
Speaker: Prof Jongeun Choi
Title: Expert data-efficient Imitation Learning for Driver Models and Robot Manipulators
Abstract:
This seminar covers two topics in data-efficient learning. The first topic presents an inverse model predictive control (iMPC) framework that learns track-dependent MPC cost functions for expert imitation in racing sports. By integrating a scheduled cost function using Gaussian process regression and Bayesian optimization, the approach generates expert-like driving data and accurately restores trajectories—outperforming standard MPC methods.
The second topic focuses on visual robotic manipulation with spatial roto-translation equivariant, or SE(3)-equivariant methods. It examines Equivariant Descriptor Fields (EDFs), which leverage generative modeling, bi-equivariance, steerable representation, and locality for high data efficiency and robust generalization. Moreover, the Diffusion-EDFs approach combines SE(3)-equivariance with diffusion-based generative modeling, enabling effective end-to-end training with only 10 demonstrations and faster inference.
Together, these topics demonstrate how advanced data modeling techniques can significantly improve data efficiency in learning of driver models and robotic manipulators.
Bio:
Jongeun Choi received his B.S. degree in Mechanical Design and Production Engineering from Yonsei University, Seoul, Republic of Korea, in 1998, and his M.S. and Ph.D. degrees in Mechanical Engineering from the University of California, Berkeley, in 2002 and 2006, respectively. He is currently a Professor at the School of Mechanical Engineering, Yonsei University, where he has also been affiliated with the Department of Artificial Intelligence since 2020. In 2023, he served as a visiting scholar at UC Berkeley. Before joining Yonsei University, Dr. Choi was a faculty member at Michigan State University, where he worked as an Assistant Professor (2006–2012) and later as an Associate Professor (2012–2016) in the Departments of Mechanical Engineering and Electrical and Computer Engineering. Dr. Choi’s research interests include machine learning, systems and control, deep reinforcement learning, and Bayesian methods, with applications in robotics, autonomous driving, human-robot interaction, and AI in healthcare. He has been serving as a Senior Editor for the International Journal of Control, Automation, and Systems since 2023 and previously served as an Associate Editor for the IEEE Robotics and Automation Letters in 2018, the Journal of Dynamic Systems, Measurement, and Control (2014–2019), and the International Journal of Precision Engineering and Manufacturing (2017–2018). Dr. Choi has received several awards, including the Best Paper Award at the RSS 2023 Workshop on Symmetries in Robot Learning and the 12th International Conference on Ubiquitous Robots and Ambient Intelligence in 2015. His papers were finalists for the Best Student Paper Award at the 24th American Control Conference in 2005 and the Dynamic Systems and Control Conference in 2011 and 2012. He was also a recipient of the NSF CAREER Award in 2009. Dr. Choi is an ASME Fellow and a member of IEEE.