Give robots experience.
Search robot datasets by task, inspect the episodes that are actually previewable, and go straight to the original source.
- 42
- 0
- 4.3M
- 39K
Static frame · no local clipOpen X-Embodiment
- Task
- Cross-embodiment robot learning
- Robot
- 22 robot embodiments
- Data
- RGB Video, Depth, Joint Positions, Actions
- Episodes reported
- 1.0M
- Hours reported
- 5.6K h
- Source
- Real-world
- License
- Open Source
Newest in the catalog
Metadata is taken from each dataset's own source. Nothing here is a usage ranking.
Real-world IndexedOpen X-Embodiment
Source: Open X-Embodiment Collaboration
- 1.0M
- 5.6K
- 3
Real-world IndexedBridgeData V2
Source: UC Berkeley RAIL
- 60K
- 333.9
- 4
Real-world IndexedLeRobot Community SO-100 Datasets
Source: Hugging Face LeRobot
- 23K
- 481
- 2
Real-world IndexedDROID
Source: Stanford IRIS & REAL Labs
- 76K
- 350
- 3
Real-world IndexedRT-1 Robot Action Dataset
Source: Google DeepMind Robotics
- 130K
- 17.4K
- 1
Real-world IndexedWaymo Open Dataset (Perception)
Source: Waymo
- 1.1K
- 6.4
- 5
Simulation IndexedLIBERO
Source: UT Austin RobIn Lab
- 6.5K
- 36.1
- 2
Real-world IndexedMobile ALOHA
Source: Stanford IRIS & REAL Labs
- 1.1K
- 5.8
- 4
Observation
Camera, depth or audio streams
State
Joint positions, pose, force
Action
Commanded deltas over time
Result
Success label or reward
Not every dataset provides all four. Each dataset page lists only the data its source actually publishes.
Browse by task
Counts reflect datasets currently in the catalog.
Browse by robot type
Hours as reported by each dataset's source.
Contribute
RobotData is early. Three things help most right now.
Share a dataset
Link a dataset you know about — source URL, task, robot and a short description. No re-upload required.
Request data
Describe the task, robot, data you need and what you intend to use it for.
Share experiment results
Which dataset you used, your setup, the result, and the limitations you hit.