Data for learning
in the physical world.
Explore robotics training data, from recorded demonstrations to simulation. Tell us what your model needs to learn or introduce a dataset you own.
Start with the task.
A useful brief describes what is recorded and what a model should learn from it.
- Demonstrations
- What the robot does, how actions are recorded and whether outcomes are labelled.
- Sensor recordings
- Which sensors are included, their formats and how recordings are synchronised.
- Simulation
- The environments, tasks and software needed to use or reproduce the data.
The right context
makes the difference.
A recording alone does not tell the whole story. We help clarify the details that matter for evaluation.
Task and environment
What is happening, which hardware is involved and the conditions under which the data was captured.
Signals and coverage
Available modalities, timing, annotations, recorded amounts and known gaps.
Rights and evaluation
Who can license the data, permitted uses and whether a suitable sample can be shared.
Have a robotics dataset?
Start with a description of the tasks, hardware, data amounts and rights you hold. No raw files needed for the first conversation.