| pretty_name: RoboCasa GR1 EEF33 LeRobot v3 | |
| task_categories: | |
| - robotics | |
| # RoboCasa GR1 EEF33 (LeRobot v3) | |
| OpenWAM-ready conversion of the 24k-trajectory `gr1_unified.*` subset from | |
| [`nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim`](https://huggingface.co/datasets/nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim). | |
| The original LeRobot v2.0 payload is preserved and enriched with: | |
| - `eef_action` | |
| - `observation.eef_state` | |
| Both use the robot-base-frame layout: | |
| ```text | |
| [L xyz3, L rot6d6, L hand6, R xyz3, R rot6d6, R hand6, waist3] | |
| ``` | |
| The conversion uses the exact RoboCasa GR1 MuJoCo model for forward kinematics, | |
| then re-indexes the dataset to the LeRobot v3 metadata/path contract. | |
| `normalization_stats.npy` contains one global `eef` statistics block pooled | |
| from state and action rows; rot6d dimensions are identity-pinned. | |
| Conversion implementation: OpenWAM PR #44, commit `e36dd50`. | |