Robot AI
Robot learning datasets
Explore robot learning datasets with published size, modalities, formats, embodiments, licences and commercial-use permissions.
Records last verified Oct 13, 2026 · Send a correction
26 sourced datasets and releases. Licence and commercial-use flags describe the checked release; collections and conversions can have different terms. Episodes, trajectories and hours are separate measurements. Size notes preserve bounds, subsets and source disagreements.
| Published size | Embodiments | Modalities | Format | |||||
|---|---|---|---|---|---|---|---|---|
| RoboCasa365 | CC BY 4.0 | Commercial use allowed | See size notes | Franka Panda with Omron mobile base (simulated) | rgb, proprioception, language | LeRobot | Sim | Feb 18, 2026 |
| Galaxea Open-World Dataset | CC BY-NC-SA 4.0 | Non-commercial use | 2.87 TB | Galaxea R1-Lite | rgb, proprioception, imu, language | RLDS, LeRobot, Parquet, MP4 | Real | Sep 9, 2025 |
| AgiBot World (Beta) | CC BY-NC-SA 4.0 | Non-commercial use | 2,976.4 h · 1,003,672 trajectories · 48.1 TB | AgiBot G1 | rgb, depth, proprioception | HDF5, MP4, PNG, JSON | Real | Mar 1, 2025 |
| AgiBot World Alpha | CC BY-NC-SA 4.0 | Non-commercial use | 595.31 h · 9.66 TB | AgiBot G1 | rgb, depth, proprioception, language | HDF5, MP4, PNG, JSON | Real | Dec 30, 2024 |
| 1X World Model Compression Challenge Dataset | Apache-2.0 | Commercial use allowed | See size notes | 1X humanoid (current v2.0 model unspecified) | rgb, proprioception | Binary arrays, JSON | Real | Not dated |
| AgiBot World 2026 | CC BY-NC-SA 4.0 | Non-commercial use | 13.6 TB | AGIBOT G2 | rgb, depth, proprioception, language | LeRobot, Parquet, MP4 | Real | Not dated |
| Arena-GR1-Manipulation-Task | CC-BY-4.0 | Commercial use allowed | 50 trajectories | Fourier GR1 (Isaac Lab simulation) | rgb, proprioception, language | HDF5, LeRobot | Sim | Not dated |
| ARIO (All Robots In One) | Not established | Commercial use: not established | See size notes | AgileX Cobot Magic | Not specified | Not specified | Mixed | Not dated |
| BEHAVIOR-1K 2025 Challenge Demos | MIT | Commercial use allowed | 10,000 episodes · 1.95 TB | Galaxea R1 Pro (2025 Challenge simulation) | rgb, proprioception | LeRobot, Parquet, MP4 | Sim | Not dated |
| BEHAVIOR-1K 2026 Challenge Demos | MIT | Commercial use allowed | 20,000 episodes · 3.26 TB | Galaxea R1 Pro (OmniGibson challenge simulation) | rgb, depth, proprioception, language | LeRobot, Parquet, MP4 | Sim | Not dated |
| BridgeData V2 | CC BY 4.0 | Commercial use allowed | 60,096 trajectories | WidowX 250 | rgb, depth, language | JPEG | Real | Not dated |
| DROID | Not established | Commercial use: not established | 350 h · 76,000 trajectories · 1.7 TB | Franka Panda | rgb, proprioception, language | RLDS, TFDS, MP4 | Real | Not dated |
| DROID 1.0.1 (LeRobot conversion) | Apache-2.0 | Commercial use: not established | 95,600 episodes | Franka Panda | rgb, proprioception, language | LeRobot, Parquet, MP4 | Real | Not dated |
| EgoDex | CC-by-NC-ND | Non-commercial use | 829 h | Not specified | rgb, human-pose, language | HDF5, MP4 | Real | Not dated |
| Fourier ActionNet Dataset | CC BY-NC-SA 4.0 | Non-commercial use | 2.74 TB | GR2, GR1-T1 / GR1-T2 | rgb, depth, proprioception, language | HDF5, MP4, MKV, JSON | Real | Not dated |
| G1_WBT_Inspire_Put_Clothes_Into_Basket | Apache-2.0 | Commercial use allowed | 300 episodes | Unitree G1 with Inspire hands | rgb, proprioception | LeRobot, Parquet, MP4 | Real | Not dated |
| HIW-500 (LeRobot) | CC BY 4.0 | Commercial use allowed | 23,743 episodes · 2.15 TB | G1 | rgb, proprioception, imu, language | LeRobot, Parquet, MP4 | Real | Not dated |
| Humanoid Everyday | Not established | Commercial use: not established | 10,300 trajectories | G1 (Dex3-1 hands), H1 (INSPIRE hands) | rgb, depth, lidar, tactile, imu, proprioception, language | Not specified | Real | Not dated |
| LeRobot ALOHA Static Coffee | MIT | Commercial use allowed | 50 episodes | ALOHA (bimanual) | rgb, proprioception | LeRobot, Parquet, MP4 | Real | Not dated |
| LIBERO Demonstrations | CC BY 4.0 | Commercial use allowed | See size notes | Franka Panda (simulated) | rgb, proprioception, language | HDF5 | Sim | Not dated |
| NVIDIA GR00T Teleop Sim (GR1) | CC BY-NC 4.0 | Non-commercial use | See size notes | GR1 (simulated Fourier hands) | rgb, proprioception, language | HDF5, LeRobot | Sim | Not dated |
| Open X-Embodiment | Not established | Commercial use: not established | See size notes | Not specified | Not specified | TFDS | Real | Not dated |
| RH20T | Not established | Commercial use: not established | 40 TB | Not specified | rgb, depth, infrared, proprioception, force-torque, audio, tactile | TAR.GZ | Real | Not dated |
| RoboMIND | Apache-2.0 | Commercial use allowed | 107,000 trajectories · 12.3 TB | Franka Emika Panda, UR5e, AgileX Cobot Magic V2.0, Tien Kung humanoid (model not stated) | rgb, proprioception, language | HDF5 | Mixed | Not dated |
| RoboMIND 2.0 | Not established | Commercial use: not established | See size notes | Franka (dual-arm), UR5e (dual-arm), AgileX, ARX, Tian Yi (collection generation unspecified), Tien Kung (collection generation unspecified) | rgb, depth, tactile, force-torque, proprioception, language | HDF5 | Mixed | Not dated |
| RoboTwin 2.0 Dataset | MIT | Commercial use allowed | 1.53 TB | Aloha-AgileX (simulated default bimanual embodiment), Franka (simulated dual-arm; model unspecified), UR5 (simulated dual-arm), Piper (simulated dual-arm), ARX-X5 (simulated dual-arm) | rgb, proprioception, language | HDF5, LeRobot | Sim | Not dated |