TriFinger: An Open-Source Robot for Learning Dexterity

Manuel Wuthrich, Felix Widmaier, Felix Grimminger, Shruti Joshi, Vaibhav Agrawal, Bilal Hammoud, Majid Khadiv, Miroslav Bogdanovic, Vincent Berenz, Julian Viereck, Maximilien Naveau, Ludovic Righetti, Bernhard Schölkopf, Stefan Bauer
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:1871-1882, 2021.

Abstract

Dexterous object manipulation is still an open problem in robotics, despite the rapid progress in machine learning during the past decade. We argue that a key issue which has hindered progress is the high cost of experimentation on real systems, in terms of both time and money. We address this problem by proposing a novel open-source robotic platform, consisting of hardware and software, to drastically reduce the cost of experimentation. The hardware is inexpensive yet highly dynamic, robust, and capable of complex contact interaction with external objects. The software allows for 1-kilohertz real-time control and performs safety checks to prevent the hardware from breaking. These properties enable the platform to run without human supervision. In addition, we provide easy-to-use C++ and Python interfaces. We illustrate the potential of the proposed platform by performing an object-manipulation task using an optimal-control algorithm and training a learning-based method directly on the real system.

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-wuthrich21a, title = {TriFinger: An Open-Source Robot for Learning Dexterity}, author = {Wuthrich, Manuel and Widmaier, Felix and Grimminger, Felix and Joshi, Shruti and Agrawal, Vaibhav and Hammoud, Bilal and Khadiv, Majid and Bogdanovic, Miroslav and Berenz, Vincent and Viereck, Julian and Naveau, Maximilien and Righetti, Ludovic and Sch\"{o}lkopf, Bernhard and Bauer, Stefan}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {1871--1882}, year = {2021}, editor = {Kober, Jens and Ramos, Fabio and Tomlin, Claire}, volume = {155}, series = {Proceedings of Machine Learning Research}, month = {16--18 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v155/wuthrich21a/wuthrich21a.pdf}, url = {https://proceedings.mlr.press/v155/wuthrich21a.html}, abstract = {Dexterous object manipulation is still an open problem in robotics, despite the rapid progress in machine learning during the past decade. We argue that a key issue which has hindered progress is the high cost of experimentation on real systems, in terms of both time and money. We address this problem by proposing a novel open-source robotic platform, consisting of hardware and software, to drastically reduce the cost of experimentation. The hardware is inexpensive yet highly dynamic, robust, and capable of complex contact interaction with external objects. The software allows for 1-kilohertz real-time control and performs safety checks to prevent the hardware from breaking. These properties enable the platform to run without human supervision. In addition, we provide easy-to-use C++ and Python interfaces. We illustrate the potential of the proposed platform by performing an object-manipulation task using an optimal-control algorithm and training a learning-based method directly on the real system.} }
Endnote
%0 Conference Paper %T TriFinger: An Open-Source Robot for Learning Dexterity %A Manuel Wuthrich %A Felix Widmaier %A Felix Grimminger %A Shruti Joshi %A Vaibhav Agrawal %A Bilal Hammoud %A Majid Khadiv %A Miroslav Bogdanovic %A Vincent Berenz %A Julian Viereck %A Maximilien Naveau %A Ludovic Righetti %A Bernhard Schölkopf %A Stefan Bauer %B Proceedings of the 2020 Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2021 %E Jens Kober %E Fabio Ramos %E Claire Tomlin %F pmlr-v155-wuthrich21a %I PMLR %P 1871--1882 %U https://proceedings.mlr.press/v155/wuthrich21a.html %V 155 %X Dexterous object manipulation is still an open problem in robotics, despite the rapid progress in machine learning during the past decade. We argue that a key issue which has hindered progress is the high cost of experimentation on real systems, in terms of both time and money. We address this problem by proposing a novel open-source robotic platform, consisting of hardware and software, to drastically reduce the cost of experimentation. The hardware is inexpensive yet highly dynamic, robust, and capable of complex contact interaction with external objects. The software allows for 1-kilohertz real-time control and performs safety checks to prevent the hardware from breaking. These properties enable the platform to run without human supervision. In addition, we provide easy-to-use C++ and Python interfaces. We illustrate the potential of the proposed platform by performing an object-manipulation task using an optimal-control algorithm and training a learning-based method directly on the real system.
APA
Wuthrich, M., Widmaier, F., Grimminger, F., Joshi, S., Agrawal, V., Hammoud, B., Khadiv, M., Bogdanovic, M., Berenz, V., Viereck, J., Naveau, M., Righetti, L., Schölkopf, B. & Bauer, S.. (2021). TriFinger: An Open-Source Robot for Learning Dexterity. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:1871-1882 Available from https://proceedings.mlr.press/v155/wuthrich21a.html.

Related Material