Speaker: Yuzhe Qin University of California
Time: 2023-04-19 16:00-2023-04-19 17:00
Venue: C19-2 or 腾讯会议：https://meeting.tencent.com/dm/4PL7WZy1BQ5V
Dexterous manipulation using multi-fingered hands poses significant challenges in robotics. The intricate contact behavior is difficult to model, limiting the efficacy of model-based approaches. Additionally, the high degree of freedom (DoF) actuation and discontinuous interaction patterns increase the sample complexity for training model-free policies. However, the human hand's similar morphology presents a unique opportunity to learn from human demonstrations. How and where can we collect human demonstration data? What learning paradigm can we use to acquire robot manipulation skills from human data? In what aspects can human demonstrations benefit dexterous manipulation? This talk will introduce several of our works that are advancing the frontier of this research area.
Yuzhe Qin is a third-year Ph.D. student at the University of California, San Diego, co-advised by Prof. Hao Su and Prof. Xiaolong Wang, and has previous experience interning at the NVIDIA Seattle Robotics Lab and Google Robotics. He obtains a bachelor's degree from the Department of Mechanical Engineering at Shanghai Jiao Tong University. His research centers on the development of robots capable of learning manipulation through interactions.