Beyond Pick-and-Place: Tackling Robotic Stacking of Diverse Shapes

Alex X. Lee, Coline Manon Devin, Yuxiang Zhou, Thomas Lampe, Konstantinos Bousmalis, Jost Tobias Springenberg, Arunkumar Byravan, Abbas Abdolmaleki, Nimrod Gileadi, David Khosid, Claudio Fantacci, Jose Enrique Chen, Akhil Raju, Rae Jeong, Michael Neunert, Antoine Laurens, Stefano Saliceti, Federico Casarini, Martin Riedmiller, raia hadsell, Francesco Nori
Proceedings of the 5th Conference on Robot Learning, PMLR 164:1089-1131, 2022.

Abstract

We study the problem of robotic stacking with objects of complex geometry. We propose a challenging and diverse set of such objects that was carefully designed to require strategies beyond a simple “pick-and-place” solution. Our method is a reinforcement learning (RL) approach combined with vision-based interactive policy distillation and simulation-to-reality transfer. Our learned policies can efficiently handle multiple object combinations in the real world and exhibit a large variety of stacking skills. In a large experimental study, we investigate what choices matter for learning such general vision-based agents in simulation, and what affects optimal transfer to the real robot. We then leverage data collected by such policies and improve upon them with offline RL. A video and a blog post of our work are provided as supplementary material.

Cite this Paper

BibTeX
@InProceedings{pmlr-v164-lee22b, title = {Beyond Pick-and-Place: Tackling Robotic Stacking of Diverse Shapes}, author = {Lee, Alex X. and Devin, Coline Manon and Zhou, Yuxiang and Lampe, Thomas and Bousmalis, Konstantinos and Springenberg, Jost Tobias and Byravan, Arunkumar and Abdolmaleki, Abbas and Gileadi, Nimrod and Khosid, David and Fantacci, Claudio and Chen, Jose Enrique and Raju, Akhil and Jeong, Rae and Neunert, Michael and Laurens, Antoine and Saliceti, Stefano and Casarini, Federico and Riedmiller, Martin and hadsell, raia and Nori, Francesco}, booktitle = {Proceedings of the 5th Conference on Robot Learning}, pages = {1089--1131}, year = {2022}, editor = {Faust, Aleksandra and Hsu, David and Neumann, Gerhard}, volume = {164}, series = {Proceedings of Machine Learning Research}, month = {08--11 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v164/lee22b/lee22b.pdf}, url = {https://proceedings.mlr.press/v164/lee22b.html}, abstract = {We study the problem of robotic stacking with objects of complex geometry. We propose a challenging and diverse set of such objects that was carefully designed to require strategies beyond a simple “pick-and-place” solution. Our method is a reinforcement learning (RL) approach combined with vision-based interactive policy distillation and simulation-to-reality transfer. Our learned policies can efficiently handle multiple object combinations in the real world and exhibit a large variety of stacking skills. In a large experimental study, we investigate what choices matter for learning such general vision-based agents in simulation, and what affects optimal transfer to the real robot. We then leverage data collected by such policies and improve upon them with offline RL. A video and a blog post of our work are provided as supplementary material.} }
Endnote
%0 Conference Paper %T Beyond Pick-and-Place: Tackling Robotic Stacking of Diverse Shapes %A Alex X. Lee %A Coline Manon Devin %A Yuxiang Zhou %A Thomas Lampe %A Konstantinos Bousmalis %A Jost Tobias Springenberg %A Arunkumar Byravan %A Abbas Abdolmaleki %A Nimrod Gileadi %A David Khosid %A Claudio Fantacci %A Jose Enrique Chen %A Akhil Raju %A Rae Jeong %A Michael Neunert %A Antoine Laurens %A Stefano Saliceti %A Federico Casarini %A Martin Riedmiller %A raia hadsell %A Francesco Nori %B Proceedings of the 5th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2022 %E Aleksandra Faust %E David Hsu %E Gerhard Neumann %F pmlr-v164-lee22b %I PMLR %P 1089--1131 %U https://proceedings.mlr.press/v164/lee22b.html %V 164 %X We study the problem of robotic stacking with objects of complex geometry. We propose a challenging and diverse set of such objects that was carefully designed to require strategies beyond a simple “pick-and-place” solution. Our method is a reinforcement learning (RL) approach combined with vision-based interactive policy distillation and simulation-to-reality transfer. Our learned policies can efficiently handle multiple object combinations in the real world and exhibit a large variety of stacking skills. In a large experimental study, we investigate what choices matter for learning such general vision-based agents in simulation, and what affects optimal transfer to the real robot. We then leverage data collected by such policies and improve upon them with offline RL. A video and a blog post of our work are provided as supplementary material.
APA
Lee, A.X., Devin, C.M., Zhou, Y., Lampe, T., Bousmalis, K., Springenberg, J.T., Byravan, A., Abdolmaleki, A., Gileadi, N., Khosid, D., Fantacci, C., Chen, J.E., Raju, A., Jeong, R., Neunert, M., Laurens, A., Saliceti, S., Casarini, F., Riedmiller, M., hadsell, r. & Nori, F.. (2022). Beyond Pick-and-Place: Tackling Robotic Stacking of Diverse Shapes. Proceedings of the 5th Conference on Robot Learning, in Proceedings of Machine Learning Research 164:1089-1131 Available from https://proceedings.mlr.press/v164/lee22b.html.

Related Material