Event
New Method for Fall Recovery with Constrained Headroom for Humanoid Robots
1 reports · 1 sources
Yidan Lu and other authors propose ROOF, a reinforcement learning method with contact-supported geometric escape guidance that enables humanoid robots, using only proprioception, to first make space and then get up when there is an obstacle above their heads. In simulation, the recovery success rate under a board that can only be lifted was 88.5%, compared with 18.9% without geometric guidance; under a free board with randomized size, mass, and position, it was 98.6%. The authors state that after the method was deployed to the Unitree G1, it succeeded in 95 of 150 attempts across seven types of real obstacle scenarios, including furniture and multi-object arrangements that did not appear in training.
Report timeline
- ROOF: A Learning Method for Humanoid Fall Recovery in Overhead-Constrained Environments
ROOF proposes a reinforcement learning method with contact-supported geometric escape guidance that enables proprioception-only humanoid robots to first create clearance and then complete the rise under overhead obstacles. In simulation, the recovery success rate was 88.5% under plates that can only be lifted, compared with 18.9% without geometric guidance; it was 98.6% under free plates with randomized size, mass, and placement. After deployment on a Unitree G1, the policy successfully recovered in 95 of 150 attempts across seven real-world obstacle scenes, including furniture and multi-object arrangements not seen in training.
arXiv cs.RO ·
Related topics
Interest trend
- Current
- 3
- Peak
- 4Nov 4 01:00
- Past 24 hours
- –
The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.