GT-111/agibotworld-v3-ee

AgiBotWorld-Alpha (LeRobot v3 EE)

This is a LeRobot v3 format conversion of the AgiBotWorld-Alpha dataset, reorganized into a canonical bimanual 16D end-effector (EE) pose representation for multi-dataset robot learning.

Original Dataset

AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems

Bu, Q., Cai, J., Chen, L., et al. (AgiBot-World-Contributors). AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems. arXiv:2503.06669, 2025.

• Original Source: https://huggingface.co/datasets/agibot-world/AgiBotWorld-Alpha
• Project Website: https://agibot-world.com
• GitHub: https://github.com/OpenDriveLab/AgiBot-World

AgiBot World is a large-scale real-world robot manipulation dataset. The Alpha release contains over 1 million trajectories across 217 tasks in five deployment scenarios, featuring dual-arm manipulation with both gripper and dexterous hand embodiments.

License

This dataset is released under CC BY-NC-SA 4.0, consistent with the original AgiBotWorld-Alpha license.

Important: By downloading this dataset, you agree to the AgiBot World Community License Agreement. The original dataset is gated — please also request access at the original Hugging Face repo.

Conversion Details

What we changed

1. Canonical EE Pose Representation: The raw dataset stores per-joint proprioception (joint positions, end-effector positions, quaternions, gripper values). We merged these into a unified 16D bimanual EE pose vector:

   • [left_x, left_y, left_z, left_qw, left_qx, left_qy, left_qz, left_gripper, right_x, right_y, right_z, right_qw, right_qx, right_qy, right_qz, right_gripper]
   • Gripper values normalized to [0, 1]

2. Training-Ready Action Masking: Head position, waist position, and robot velocity dimensions are masked from policy inputs/outputs during training. Only the 16D EE pose dimensions are used for action prediction.

3. LeRobot v3 Format: Converted to the latest LeRobot v3 dataset layout with sharded video storage (MP4, AV1 codec) and Parquet-based frame data.

4. Delta Actions: During training, actions are converted to delta (relative) form in the EE pose space, with quaternion handling via geodesic interpolation.

What we preserved

• All 3 camera views: head, hand_left, hand_right (480×640 RGB video)
• Episode structure and task labels
• Frame-level timestamps

Dataset Structure

agibot_gripper/
├── data/
│   └── chunk-*/file-*.parquet   # Frame data (16D state + 16D action + task labels)
├── videos/
│   └── observation.images.head/chunk-*/file-*.mp4
│   └── observation.images.hand_left/chunk-*/file-*.mp4
│   └── observation.images.hand_right/chunk-*/file-*.mp4
├── meta/
│   ├── info.json                 # Dataset metadata
│   ├── tasks.json                # Task vocabulary
│   ├── episodes.jsonl            # Episode index
│   └── stats.json                # Normalization statistics
└── norm_stats.json               # Precomputed quantile normalization stats

Statistics

Metric	Value
Total Episodes	33,609
Total Frames	56,858,070
Robot Type	A2D (AgiBot dual-arm)
Embodiment	Gripper
State Dim	16
Action Dim	16

Usage

from lerobot.datasets import LeRobotDataset

dataset = LeRobotDataset(
    repo_id="GT-111/agibotworld-v3-ee",
    root="agibot_gripper",  # or "GT-111/agibotworld-v3-ee" for HF download
)

For training with the LWM-VLA / OpenPI framework:

from openpi.training.config import MultiDatasetPretrainDatasetSpec

MultiDatasetPretrainDatasetSpec(
    repo_id="GT-111/agibotworld-v3-ee",
    dataset_type="agibot",
    weight=0.3,
)

Citation

If you use this dataset, please cite both the original AgiBot World paper and this conversion:

@article{agibotworld2025,
  title={AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems},
  author={AgiBot-World-Contributors and Bu, Qingwen and Cai, Jisong and Chen, Li and Cui, Xiuqi and Ding, Yan and Feng, Siyuan and Gao, Shenyuan and He, Xindong and Hu, Xuan and Huang, Xu and Jiang, Shu and Jiang, Yuxin and Jing, Cheng and Li, Hongyang and Li, Jialu and Liu, Chiming and Liu, Yi and Lu, Yuxiang and others},
  journal={arXiv preprint arXiv:2503.06669},
  year={2025}
}

Version History

• v3.0 (current): LeRobot v3 conversion with canonical 16D EE pose layout