MirrorDuo: Reflection-Consistent Visuomotor Learning from Mirrored Demonstration Pairs

Zheyu Zhuang, Ruiyu Wang, Giovanni Luca Marchetti, Florian T. Pokorny, Danica Kragic
Proceedings of The 9th Conference on Robot Learning, PMLR 305:2391-2408, 2025.

Abstract

Image-based behaviour cloning leverages demonstrations captured from ubiquitous RGB cameras, enabling impressive visuomotor performance. However, it remains constrained by the cost of collecting sufficiently diverse demonstrations, especially for generalizing across workspace variations. We propose MirrorDuo, a mirroring-based formulation that operates on image, proprioception, and full 6-DoF end-effector action tuples, generating a mirrored counterpart for each original demonstration, effectively achieving “collect one, get one for free." It can be applied as a data augmentation strategy for existing learning pipelines, such as standard behaviour cloning or diffusion policy, or as a structural prior for reflection-equivariant policy networks. By leveraging the overlap between the original and mirrored domains, MirrorDuo achieves significantly improved performance under the same data budget when demonstrations are evenly distributed across both sides of the workspace. When demonstrations are confined to one side, MirrorDuo enables efficient skill transfer to the mirrored workspace with as few as zero or just 5 demonstrations in the target arrangement.

Cite this Paper

BibTeX
@InProceedings{pmlr-v305-zhuang25a, title = {MirrorDuo: Reflection-Consistent Visuomotor Learning from Mirrored Demonstration Pairs}, author = {Zhuang, Zheyu and Wang, Ruiyu and Marchetti, Giovanni Luca and Pokorny, Florian T. and Kragic, Danica}, booktitle = {Proceedings of The 9th Conference on Robot Learning}, pages = {2391--2408}, year = {2025}, editor = {Lim, Joseph and Song, Shuran and Park, Hae-Won}, volume = {305}, series = {Proceedings of Machine Learning Research}, month = {27--30 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v305/main/assets/zhuang25a/zhuang25a.pdf}, url = {https://proceedings.mlr.press/v305/zhuang25a.html}, abstract = {Image-based behaviour cloning leverages demonstrations captured from ubiquitous RGB cameras, enabling impressive visuomotor performance. However, it remains constrained by the cost of collecting sufficiently diverse demonstrations, especially for generalizing across workspace variations. We propose MirrorDuo, a mirroring-based formulation that operates on image, proprioception, and full 6-DoF end-effector action tuples, generating a mirrored counterpart for each original demonstration, effectively achieving “collect one, get one for free." It can be applied as a data augmentation strategy for existing learning pipelines, such as standard behaviour cloning or diffusion policy, or as a structural prior for reflection-equivariant policy networks. By leveraging the overlap between the original and mirrored domains, MirrorDuo achieves significantly improved performance under the same data budget when demonstrations are evenly distributed across both sides of the workspace. When demonstrations are confined to one side, MirrorDuo enables efficient skill transfer to the mirrored workspace with as few as zero or just 5 demonstrations in the target arrangement.} }
Endnote
%0 Conference Paper %T MirrorDuo: Reflection-Consistent Visuomotor Learning from Mirrored Demonstration Pairs %A Zheyu Zhuang %A Ruiyu Wang %A Giovanni Luca Marchetti %A Florian T. Pokorny %A Danica Kragic %B Proceedings of The 9th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Joseph Lim %E Shuran Song %E Hae-Won Park %F pmlr-v305-zhuang25a %I PMLR %P 2391--2408 %U https://proceedings.mlr.press/v305/zhuang25a.html %V 305 %X Image-based behaviour cloning leverages demonstrations captured from ubiquitous RGB cameras, enabling impressive visuomotor performance. However, it remains constrained by the cost of collecting sufficiently diverse demonstrations, especially for generalizing across workspace variations. We propose MirrorDuo, a mirroring-based formulation that operates on image, proprioception, and full 6-DoF end-effector action tuples, generating a mirrored counterpart for each original demonstration, effectively achieving “collect one, get one for free." It can be applied as a data augmentation strategy for existing learning pipelines, such as standard behaviour cloning or diffusion policy, or as a structural prior for reflection-equivariant policy networks. By leveraging the overlap between the original and mirrored domains, MirrorDuo achieves significantly improved performance under the same data budget when demonstrations are evenly distributed across both sides of the workspace. When demonstrations are confined to one side, MirrorDuo enables efficient skill transfer to the mirrored workspace with as few as zero or just 5 demonstrations in the target arrangement.
APA
Zhuang, Z., Wang, R., Marchetti, G.L., Pokorny, F.T. & Kragic, D.. (2025). MirrorDuo: Reflection-Consistent Visuomotor Learning from Mirrored Demonstration Pairs. Proceedings of The 9th Conference on Robot Learning, in Proceedings of Machine Learning Research 305:2391-2408 Available from https://proceedings.mlr.press/v305/zhuang25a.html.

Related Material