Learning Arbitrary-Goal Fabric Folding with One Hour of Real Robot Experience

Robert Lee, Daniel Ward, Vibhavari Dasagi, Akansel Cosgun, Juxi Leitner, Peter Corke
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:2317-2327, 2021.

Abstract

Manipulating deformable objects, such as fabric, is a long standing problem in robotics, with state estimation and control posing a significant challenge for traditional methods. In this paper, we show that it is possible to learn fabric folding skills in only an hour of self-supervised real robot experience, without human supervision or simulation. Our approach relies on fully convolutional networks and the manipulation of visual inputs to exploit learned features, allowing us to create an expressive goal-conditioned pick and place policy that can be trained efficiently with real world robot data only. Folding skills are learned with only a sparse reward function and thus do not require reward function engineering, merely an image of the goal configuration. We demonstrate our method on a set of towel-folding tasks, and show that our approach is able to discover sequential folding strategies, purely from trial-and-error. We achieve state-of-the-art results without the need for demonstrations or simulation, used in prior approaches.

Cite this Paper


BibTeX
@InProceedings{pmlr-v155-lee21a, title = {Learning Arbitrary-Goal Fabric Folding with One Hour of Real Robot Experience}, author = {Lee, Robert and Ward, Daniel and Dasagi, Vibhavari and Cosgun, Akansel and Leitner, Juxi and Corke, Peter}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {2317--2327}, 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/lee21a/lee21a.pdf}, url = {https://proceedings.mlr.press/v155/lee21a.html}, abstract = {Manipulating deformable objects, such as fabric, is a long standing problem in robotics, with state estimation and control posing a significant challenge for traditional methods. In this paper, we show that it is possible to learn fabric folding skills in only an hour of self-supervised real robot experience, without human supervision or simulation. Our approach relies on fully convolutional networks and the manipulation of visual inputs to exploit learned features, allowing us to create an expressive goal-conditioned pick and place policy that can be trained efficiently with real world robot data only. Folding skills are learned with only a sparse reward function and thus do not require reward function engineering, merely an image of the goal configuration. We demonstrate our method on a set of towel-folding tasks, and show that our approach is able to discover sequential folding strategies, purely from trial-and-error. We achieve state-of-the-art results without the need for demonstrations or simulation, used in prior approaches.} }
Endnote
%0 Conference Paper %T Learning Arbitrary-Goal Fabric Folding with One Hour of Real Robot Experience %A Robert Lee %A Daniel Ward %A Vibhavari Dasagi %A Akansel Cosgun %A Juxi Leitner %A Peter Corke %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-lee21a %I PMLR %P 2317--2327 %U https://proceedings.mlr.press/v155/lee21a.html %V 155 %X Manipulating deformable objects, such as fabric, is a long standing problem in robotics, with state estimation and control posing a significant challenge for traditional methods. In this paper, we show that it is possible to learn fabric folding skills in only an hour of self-supervised real robot experience, without human supervision or simulation. Our approach relies on fully convolutional networks and the manipulation of visual inputs to exploit learned features, allowing us to create an expressive goal-conditioned pick and place policy that can be trained efficiently with real world robot data only. Folding skills are learned with only a sparse reward function and thus do not require reward function engineering, merely an image of the goal configuration. We demonstrate our method on a set of towel-folding tasks, and show that our approach is able to discover sequential folding strategies, purely from trial-and-error. We achieve state-of-the-art results without the need for demonstrations or simulation, used in prior approaches.
APA
Lee, R., Ward, D., Dasagi, V., Cosgun, A., Leitner, J. & Corke, P.. (2021). Learning Arbitrary-Goal Fabric Folding with One Hour of Real Robot Experience. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:2317-2327 Available from https://proceedings.mlr.press/v155/lee21a.html.

Related Material