ARCH: Hierarchical Hybrid Learning for Long-Horizon Contact-Rich Robotic Assembly

Published in CoRL, 2025

Jiankai Sun, Aidan Curtis, Yang You, Yan Xu, Michael Koehle, Qianzhong Chen, Suning Huang, Leonidas Guibas, Sachin Chitta, Mac Schwager, Hui Li

arXiv / PDF / project page / BibTeX

Abstract

ARCH proposes a hierarchical modular approach for long-horizon, high-precision robotic assembly in contact-rich settings. It employs a hierarchical planning framework, including a low-level primitive library of parameterized skills and a high-level policy learned via IL. ARCH generalizes well to unseen objects and outperforms baseline methods in terms of success rate and data efficiency.

BibTeX

@inproceedings{sun2025arch,
  title     = {ARCH: Hierarchical Hybrid Learning for Long-Horizon Contact-Rich Robotic Assembly},
  author    = {Jiankai Sun and Aidan Curtis and Yang You and Yan Xu and Michael Koehle and Qianzhong Chen and Suning Huang and Leonidas Guibas and Sachin Chitta and Mac Schwager and Hui Li},
  booktitle = {Conference on Robot Learning (CoRL)},
  year      = {2025},
  eprint    = {2409.16451},
  archivePrefix = {arXiv}
}

Recommended citation: Sun, J., Curtis, A., You, Y., Xu, Y., Koehle, M., Chen, Q., Huang, S., Guibas, L., Chitta, S., Schwager, M., & Li, H. (2025). ARCH: Hierarchical Hybrid Learning for Long-Horizon Contact-Rich Robotic Assembly. CoRL 2025.