제목: Skill-based Robot Learning

강사: 이영운 박사님 (23년 9월 부임 예정)

일시 : 2022년 2월 23일 목요일 16시

장소: 제 4공학관 D915호

Abstract: Despite the recent progress in robot learning, robotics research and benchmarks today are typically confined to simple and short-horizon tasks. However, tasks in our daily lives are much more complicated — consisting of multiple sub-tasks and requiring high dexterity skills — and the typical "learning from scratch” scheme is hardly scale to such complex long-horizon tasks. In this talk, I propose to extend the range of tasks that robots can learn by acquiring a useful skillset and efficiently harnessing these skills. As a first step, I will introduce a novel benchmark for complex long-horizon manipulation tasks, furniture assembly both in simulation and the real world. Then, I will talk about a series of skill-based reinforcement learning approaches that discover reusable skills and efficiently learn a complex long-horizon task by leveraging skills, skill priors, and a world model extracted from diverse data.

Bio: Youngwoon Lee is a postdoctoral scholar at the University of California, Berkeley working with Prof. Pieter Abbeel. His research interests are in deep reinforcement learning and imitation learning for robotics. Particularly, his research focuses on solving complex long-horizon tasks, such as furniture assembly, which requires many aspects of intelligent robots from structural reasoning to long-term planning to sophisticated control. Prior to joining UC Berkeley, he received his Ph.D. in Computer Science at the University of Southern California in 2022, advised by Prof. Joseph J. Lim, and received his B.S. and M.S. degrees in Computer Science at KAIST.