Stanford researchers, led by Zipeng Fu, Qingqing Zhao, and Qi Wu, have developed HumanPlus, a humanoid robot capable of learning autonomous skills by copying humans. In other words, the team used a full-stack system for humanoids to help the robot learn motion and autonomous skills from human data.
This involved training a low-level policy in simulation via reinforcement learning using existing 40-hour human motion datasets. The policy could then be transferred to the real world and enable humanoid robots to follow human body as well as hand motions in real time using only a RGB camera, i.e. shadowing.
- Authentic Design - Features a LEGO brick model of a Tallneck, the most iconic Horizon machine in the Forbidden West with authentic Tallneck features
- Iconic Features - Perfectly capture the essence of the Tallneck with its iconic disc-shaped head, antennae protruding from the neck, tail-like...
- Detailed Display Stand - The display stand comes with a familiar landscape: a brick-built birch tree, tall grass, and a rusty stoplight with a vine...
We demonstrate the system on our customized 33-DoF 180cm humanoid, autonomously completing tasks such as wearing a shoe to stand up and walk, unloading objects from warehouse racks, folding a sweatshirt, rearranging objects, typing, and greeting another robot with 60-100% success rates using up to 40 demonstrations,” said the researchers.
[Source]
