Dataset · agibot-world
Over 1M robot manipulation trajectories
agibot-world/AgiBotWorld-Beta
Over 1 million trajectories from 100 robots, spanning 2976 hours and 100+ real-world scenarios.
What's inside
Contains over 1M trajectories from 100 robots with a total duration of 2,976.4 hours. Covers 100+ real scenarios across 5 domains, 200+ task types, and 87 atomic skills including tying, opening cans, and sweeping. English language specified; metadata indicates text modality and size between 100M and 1B.
What you can do with it
- Training robot manipulation policies
- Imitation learning on long-horizon and contact-rich tasks
- Evaluating planning and multi-robot coordination models
- Tasks
- Other, Robotics
- Languages
- English
- Size
- 100M–1B rows
- Downloads this month
- 185.9K
- Likes
- 92
- Access
- on request on Hugging Face
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("agibot-world/AgiBotWorld-Beta")