DreamDojo captures action labels for a Fourier GR-1 action space
NVIDIA's DreamDojo work disclosed an in-lab robot-action data collection setup.
Why it matters
NVIDIA's DreamDojo research used MANUS gloves and a Vive tracker to collect in-lab action labels retargeted to a Fourier GR-1 action space.
Evidence notes
- Operators used MANUS gloves and a Vive tracker to capture hand and body motion labels.
- The recorded labels were retargeted to a Fourier GR-1 action space for world-model research.
Company context
NVIDIA is a compute and AI infrastructure company providing the simulation, training, edge-compute and model platforms used across modern robotics. Its stack includes tools for robot learning, digital twins, perception and on-device inference.