HannesImitation is an imitation learning-based approach that trains a single grasping policy across diverse objects and environments. The learned policy is deployed on the Hannes prosthetic hand, enabling control of wrist orientation and hand closure. The main contributions are

• HannesImitationPolicy: The novel application of Diffusion Policy to control a prosthetic hand equipped with and anthropomorphic wrist.
• HannesImitationDataset: A novel dataset for imitation learning with the Hannes prosthetic hand for grasping on unstructured scenarios.

The HannesImitation project builds on top of Diffusion Policy. To set up the environment and dependencies, please follow the installation instructions provided in the Diffusion Policy repository: https://github.com/real-stanford/diffusion_policy.

📄 Preprint available on arXiv: https://arxiv.org/abs/2508.00491.

The HannesImitationDataset is available at: https://github.com/hsp-iit/HannesImitation

If you found this useful, you could cite the article as follows:

@inproceedings{...,
  title={HannesImitation: Grasping with the Hannes Prosthetic Hand via Imitation Learning},
  author={...},
  booktitle={...},
  pages={...},
  year={...},
  organization={...}
}

If you have any questions, feel free to contact Carlo Alessi.