Skip to content

Dataset · AlanaSky

Real robot trajectories for reinforcement learning

AlanaSky/XA1

Video recordings and real robot action trajectories for offline-to-online RL research.

What's inside

Prepared as part of the Robo-ValueRL project, this dataset contains 10k to 100k video records. It features heterogeneous trajectories from real robots designed for studying value estimation methods and policy adaptation.

What you can do with it

  • Offline pre-training of robot control policies with subsequent online adaptation
  • Training and validation of value estimation algorithms in robotics
Size
10K–100K rows
Downloads this month
58.1K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("AlanaSky/XA1")