일시: 2025. 9. 3. (수요일), 오후 4시

장소: 공학원, 452B호

Speaker: 위영선 박사과정 (University of Michigan, Ann Arbor)

Title: Multi-modal Representation Learning for Contact-rich and Dexterous Robotic Manipulation

Abstract:

Dexterity refers to a robot’s ability to effectively perform tasks with its hands and tools. In this talk, I will share how my research explores ways for robots to achieve dexterity by leveraging force and touch sensing to understand, plan, and control physical interactions with the world. The talk begins with how we can perceive contact through multi-modal 3D object representations, then moves to how we can plan and control contact using a modular approach grounded in the natural boundaries of contact. Finally, I will show how robots can learn from human touch through a tactile glove developed at Meta.

Bio:

Youngsun Wi is a Ph.D. candidate in the Robotics Department at the University of Michigan, advised by Prof. Nima Fazeli. She earned her bachelor’s degree in Mechanical Engineering from KAIST and her master’s degree in Mechanical Engineering from Michigan. During her Ph.D., she has published in leading international robotics conferences, including CoRL, ICRA, and RSS. She has also completed internships at three U.S. industry research labs (Amazon Robotics AI, NVIDIA Robotics, and Meta FAIR) and collaborated with DeepMind. At Michigan, she served as an officer in Women in Robotics & Engineering (WiRE+), supporting efforts to foster a diverse and inclusive robotics community.