Rodrigues Network for Learning Robot Actions

Published in ICLR, 2026

Jialiang Zhang*, Haoran Geng*, Yang You*, Congyue Deng, Pieter Abbeel, Jitendra Malik, Leonidas Guibas

arXiv / PDF / BibTeX

Abstract

This work introduces the Neural Rodrigues Operator, a learnable generalization of the classical Rodrigues’ rotation formula, to embed kinematic inductive bias directly into neural networks. Built upon this operator, the proposed Rodrigues Network (RodriNet) effectively models articulated actions and significantly improves performance across tasks like forward kinematics prediction, imitation learning for robot manipulation, and 3D hand pose estimation from images.

BibTeX

@inproceedings{zhang2026rodrigues,
  title     = {Rodrigues Network for Learning Robot Actions},
  author    = {Jialiang Zhang and Haoran Geng and Yang You and Congyue Deng and Pieter Abbeel and Jitendra Malik and Leonidas Guibas},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year      = {2026},
  eprint    = {2506.02618},
  archivePrefix = {arXiv}
}

Recommended citation: Zhang, J., Geng, H., You, Y., Deng, C., Abbeel, P., Malik, J., & Guibas, L. (2025). Rodrigues Network for Learning Robot Actions. arXiv preprint arXiv:2506.02618.