Reward AI unveils OM-1 universal robot policy

Reward AI has released OM-1, a universal robot policy it says can adapt to a new task with under 30 minutes of natural human motion data. The same policy is claimed to drive both robotic arms and humanoid robots without teleoperation or real-machine pre-collection.

• Adapting to a new task needs less than 30 minutes of human motion data
• A single policy drives both robotic arms and humanoid robots
• No teleoperation or real-machine pre-collection is required
• Training uses one unified single-stage process with RL fine-tuning