OM-1
OM-1 is a general-purpose robot manipulation policy from Reward AI that learns directly from human demonstrations, with no teleoperation or on-robot training data needed. Using a wearable Omnibody Hand mocap device to capture images, tactile signals, inter-finger proximity, and hand pose trajectories, OM-1 lets robots learn contact-rich manipulation tasks from under 30 minutes of demonstration data and perform them at human speed across different robot bodies.
- Type
- Structured data
- Weights
- Closed
More from Reward AI
3
About Reward AI
1
Reward AI builds general-purpose robot intelligence software, including its OM-1 model trained from human demonstration data to run on any robot body.