I am currently very interested in building robust general-purpose embodied AI systems, a field that I find exceptionally exciting. Specifically, I focus on two pivotal challenges: 1) policy with self-correction: How to empower visuomotor policy learning with self-correction ability? 2) efficient policy learning: How to facilitate in-context learning for robotics, enabling the acquisition of new skills through minimal demonstrations?
Powered by Jekyll and Minimal Light theme.