Collect, format, and manage robot training data — from teleoperation episode recording to HDF5, RLDS, and LeRobot dataset pipelines, quality control, and long-term storage.
Three pages that cover the full arc from first episode to production-ready dataset.
Understand the three dominant formats for robot training data, when to use each, and how to convert between them.
Collection GuideEnd-to-end guide from hardware setup through episode recording, validation, and dataset versioning.
Strategy GuideScale from hundreds to millions of demonstrations — infrastructure, team structure, and quality gates.
Every data topic has the full set of guide types.
LeRobot Data (HuggingFace)
RLDS
ALOHA Data (HDF5 / Zarr)
DROID Dataset
Open X-Embodiment
Data Pipeline & Quality
OpenArm, ALOHA, Franka, UR5e, and more.
CategoryBimanual, glove-based, VR, and leader-follower setups.
CategoryAllegro, Inspire, LEAP, and tactile grippers.
CategoryBooster, Unitree H1/G1, and Figure 02 deployment.
HubReturn to the main guides index.