Guide to Imitation Learning with Cyclo Intelligence

The ROBOTIS Humanoid Division is excited to release Cyclo Intelligence, an open-source Physical AI platform designed for imitation learning workflows on AI Worker.

Setting up real-robot imitation learning pipelines can be complex, requiring seamless integration from teleoperation data collection and dataset conversion to model training (supporting GR00T, PI, ACT, etc.) and real-time robot inference.

Cyclo Intelligence streamlines this entire workflow into an end-to-end (E2E) pipeline. With the latest AI Worker 2.0 update, it now natively supports ROS 2 Zenoh RMW, containerized Docker execution environments, and an intuitive Web UI.

We hope this serves as a valuable open-source tool for researchers and engineers developing Physical AI and robotic manipulation. Everything is fully open-source, so please feel free to test it out and share your feedback with us!

:clapper_board: Video: https://youtu.be/xOyVGEDR5VY
:robot: Code: GitHub - ROBOTIS-GIT/cyclo_intelligence: cyclo_intelligence · GitHub
:memo: Docs: Cyclo Intelligence | ROBOTIS Docs
:speech_balloon: Community: https://www.robotis.com/en/community.php