FastUMI: A Scalable and Hardware-Independent Universal Manipulation Interface with Dataset

Zhaxizhuom Zhaxizhuoma, Kehui Liu, Chuyue Guan, Zhongjie Jia, Ziniu Wu, Xin Liu, Tianyu Wang, Shuai Liang, Pengan CHEN, Pingrui Zhang, Haoming Song, Delin Qu, Dong Wang, Zhigang Wang, Nieqing Cao, Yan Ding, Bin Zhao, Xuelong Li
Proceedings of The 9th Conference on Robot Learning, PMLR 305:3069-3093, 2025.

Abstract

Real-world manipulation datasets for robotic arms remain scarce due to the high costs, rigid hardware dependencies, and complex setup procedures associated with existing data collection methods. We introduce, a redesigned Universal Manipulation Interface (UMI) that addresses these challenges, enabling low-cost, scalable, and rapid deployment across heterogeneous platforms. FastUMI achieves this through: (i) hardware decoupling via extensive mechanical reengineering, which removes dependence on specialized robotic components while preserving a consistent visual perspective; (ii) replacement of complex visual–inertial odometry with a commercial off-the-shelf tracker, simplifying the software stack without compromising pose estimation accuracy; and (iii) the provision of an integrated ecosystem that streamlines data acquisition, automates quality control, and ensures compatibility with both standard and enhanced imitation-learning pipelines. To facilitate further research, we release an open-access dataset comprising over 15,000 real-world demonstrations spanning 24 tasks constituting one of the most extensive UMI-like resources to date. Empirical evaluations show that FastUMI supports rapid deployment, reduces operational overhead, and delivers robust performance across diverse manipulation scenarios, advancing scalable data-driven robotic learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v305-zhaxizhuoma25a, title = {FastUMI: A Scalable and Hardware-Independent Universal Manipulation Interface with Dataset}, author = {Zhaxizhuoma, Zhaxizhuom and Liu, Kehui and Guan, Chuyue and Jia, Zhongjie and Wu, Ziniu and Liu, Xin and Wang, Tianyu and Liang, Shuai and CHEN, Pengan and Zhang, Pingrui and Song, Haoming and Qu, Delin and Wang, Dong and Wang, Zhigang and Cao, Nieqing and Ding, Yan and Zhao, Bin and Li, Xuelong}, booktitle = {Proceedings of The 9th Conference on Robot Learning}, pages = {3069--3093}, 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/zhaxizhuoma25a/zhaxizhuoma25a.pdf}, url = {https://proceedings.mlr.press/v305/zhaxizhuoma25a.html}, abstract = {Real-world manipulation datasets for robotic arms remain scarce due to the high costs, rigid hardware dependencies, and complex setup procedures associated with existing data collection methods. We introduce, a redesigned Universal Manipulation Interface (UMI) that addresses these challenges, enabling low-cost, scalable, and rapid deployment across heterogeneous platforms. FastUMI achieves this through: (i) hardware decoupling via extensive mechanical reengineering, which removes dependence on specialized robotic components while preserving a consistent visual perspective; (ii) replacement of complex visual–inertial odometry with a commercial off-the-shelf tracker, simplifying the software stack without compromising pose estimation accuracy; and (iii) the provision of an integrated ecosystem that streamlines data acquisition, automates quality control, and ensures compatibility with both standard and enhanced imitation-learning pipelines. To facilitate further research, we release an open-access dataset comprising over 15,000 real-world demonstrations spanning 24 tasks constituting one of the most extensive UMI-like resources to date. Empirical evaluations show that FastUMI supports rapid deployment, reduces operational overhead, and delivers robust performance across diverse manipulation scenarios, advancing scalable data-driven robotic learning.} }
Endnote
%0 Conference Paper %T FastUMI: A Scalable and Hardware-Independent Universal Manipulation Interface with Dataset %A Zhaxizhuom Zhaxizhuoma %A Kehui Liu %A Chuyue Guan %A Zhongjie Jia %A Ziniu Wu %A Xin Liu %A Tianyu Wang %A Shuai Liang %A Pengan CHEN %A Pingrui Zhang %A Haoming Song %A Delin Qu %A Dong Wang %A Zhigang Wang %A Nieqing Cao %A Yan Ding %A Bin Zhao %A Xuelong Li %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-zhaxizhuoma25a %I PMLR %P 3069--3093 %U https://proceedings.mlr.press/v305/zhaxizhuoma25a.html %V 305 %X Real-world manipulation datasets for robotic arms remain scarce due to the high costs, rigid hardware dependencies, and complex setup procedures associated with existing data collection methods. We introduce, a redesigned Universal Manipulation Interface (UMI) that addresses these challenges, enabling low-cost, scalable, and rapid deployment across heterogeneous platforms. FastUMI achieves this through: (i) hardware decoupling via extensive mechanical reengineering, which removes dependence on specialized robotic components while preserving a consistent visual perspective; (ii) replacement of complex visual–inertial odometry with a commercial off-the-shelf tracker, simplifying the software stack without compromising pose estimation accuracy; and (iii) the provision of an integrated ecosystem that streamlines data acquisition, automates quality control, and ensures compatibility with both standard and enhanced imitation-learning pipelines. To facilitate further research, we release an open-access dataset comprising over 15,000 real-world demonstrations spanning 24 tasks constituting one of the most extensive UMI-like resources to date. Empirical evaluations show that FastUMI supports rapid deployment, reduces operational overhead, and delivers robust performance across diverse manipulation scenarios, advancing scalable data-driven robotic learning.
APA
Zhaxizhuoma, Z., Liu, K., Guan, C., Jia, Z., Wu, Z., Liu, X., Wang, T., Liang, S., CHEN, P., Zhang, P., Song, H., Qu, D., Wang, D., Wang, Z., Cao, N., Ding, Y., Zhao, B. & Li, X.. (2025). FastUMI: A Scalable and Hardware-Independent Universal Manipulation Interface with Dataset. Proceedings of The 9th Conference on Robot Learning, in Proceedings of Machine Learning Research 305:3069-3093 Available from https://proceedings.mlr.press/v305/zhaxizhuoma25a.html.

Related Material