• [2025-08] CLONE is accepted by CoRL!

• [2025-07] We release CLONE's early checkpoint and deploy document.

• [2025-06] We release the paper and demos.

CLONE employs an MoE-based policy with closed-loop error correction for holistic humanoid teleoperation, enabling capabilities previously unattainable with existing systems—such as whole-body coordination and long-horizon task execution. Using only minimal input from a commercial MR headset, CLONE significantly improves tracking precision over existing open-loop approaches, opening new possibilities for practical humanoid deployment in unstructured environments.

• Release the paper with demos.
• Release early checkpoints(used for real-world deployment).
• Release instructions for real-world deployment.
• Release training and inference code with Roboverse.
• Release enhanced checkpoints.
• Full codebase release and further updates.

If you find our work helpful, please cite:

@misc{li2025clone,
  title={CLONE: Closed-Loop Whole-Body Humanoid Teleoperation for Long-Horizon Tasks}, 
  author={Yixuan Li and Yutang Lin and Jieming Cui and Tengyu Liu and Wei Liang and Yixin Zhu and Siyuan Huang},
  journal={arXiv preprint arXiv:2506.08931}, 
  year={2025}
}

You may also find CLONE's implementation in RoboVerse soon.

This project is released under CC BY-NC 4.0 license.

• RoboVerse: Together with the RoboVerse team, we release the training and inference code.
  • Thanks to Charlie Cheng for providing the Docker version of this project.
• PHC: We use PHC for reference motion retargeting.
• OmniH2O: We use OmniH2O as our codebase.
• FAST_LIO: We use the FAST_LIO as the Lidar Odometry.
• VisionProTeleop: We use VisionProTeleop to teleoperate with Apple VisionPro.