Ji-Yuan Shi

Egocentric human videos can now directly train humanoid robots, bypassing the need for expensive robot-specific demonstrations for novel tasks.

OASIS shows that simulation can outperform real-world teleoperation in humanoid manipulation tasks by leveraging diverse environmental variations.

Robots can now learn fine-grained manipulation skills directly from human demonstrations, thanks to a new imitation learning framework that automatically figures out the right rewards and handles noisy real-world data.

Forget brittle, single-skill robots: X-Loco achieves robust, vision-based humanoid locomotion by intelligently combining multiple expert policies for fall recovery, terrain traversal, and whole-body coordination.

Humanoid robots can now reliably perform long-horizon loco-manipulation tasks in the real world thanks to a novel root-trajectory conditioned policy and persistent object estimation.