Introducing our latest collaborative work, named LEGATO, which describes a universal handheld smart gripper and a motion-invariant policy, that enable efficient cross-embodiment skill transfer between human users and robots with different morphologies, enhancing scalability and versatility in robotics. Congratulations to my PhD student Mingyo Seo who has led the study. In addition, many thanks to The AI Institute, Andy Park and Yuke Zhu (co-advisor of Mingyo) for their incredible contributions and support. Many thanks to Mitch Pryor for providing access to his Spot robot. Many thanks to the Office of Naval Research for supporting this project.
Pointer to paper:
https://arxiv.org/abs/2411.03682
Project Website (includes code and various videos):
https://lnkd.in/gBeW8pQZ
Leave a Reply