Learning Robotic Manipulation of Granular Media

Connor Schenck, Jonathan Tompson, Sergey Levine, Dieter Fox
; Proceedings of the 1st Annual Conference on Robot Learning, PMLR 78:239-248, 2017.

Abstract

In this paper, we examine the problem of robotic manipulation of granular media. We evaluate multiple predictive models used to infer the dynamics of scooping and dumping actions. These models are evaluated on a task that involves manipulating the media in order to deform it into a desired shape. Our best performing model is based on a highly-tailored convolutional network architecture with domain-specific optimizations, which we show accurately models the physical interaction of the robotic scoop with the underlying media. We empirically demonstrate that explicitly predicting physical mechanics results in a policy that out-performs both a hand-crafted dynamics baseline, and a “value-network”, which must otherwise implicitly predict the same mechanics in order to produce accurate value estimates.

Cite this Paper


BibTeX
@InProceedings{pmlr-v78-schenck17a, title = {Learning Robotic Manipulation of Granular Media}, author = {Connor Schenck and Jonathan Tompson and Sergey Levine and Dieter Fox}, booktitle = {Proceedings of the 1st Annual Conference on Robot Learning}, pages = {239--248}, year = {2017}, editor = {Sergey Levine and Vincent Vanhoucke and Ken Goldberg}, volume = {78}, series = {Proceedings of Machine Learning Research}, month = {13--15 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v78/schenck17a/schenck17a.pdf}, url = {http://proceedings.mlr.press/v78/schenck17a.html}, abstract = {In this paper, we examine the problem of robotic manipulation of granular media. We evaluate multiple predictive models used to infer the dynamics of scooping and dumping actions. These models are evaluated on a task that involves manipulating the media in order to deform it into a desired shape. Our best performing model is based on a highly-tailored convolutional network architecture with domain-specific optimizations, which we show accurately models the physical interaction of the robotic scoop with the underlying media. We empirically demonstrate that explicitly predicting physical mechanics results in a policy that out-performs both a hand-crafted dynamics baseline, and a “value-network”, which must otherwise implicitly predict the same mechanics in order to produce accurate value estimates.} }
Endnote
%0 Conference Paper %T Learning Robotic Manipulation of Granular Media %A Connor Schenck %A Jonathan Tompson %A Sergey Levine %A Dieter Fox %B Proceedings of the 1st Annual Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2017 %E Sergey Levine %E Vincent Vanhoucke %E Ken Goldberg %F pmlr-v78-schenck17a %I PMLR %J Proceedings of Machine Learning Research %P 239--248 %U http://proceedings.mlr.press %V 78 %W PMLR %X In this paper, we examine the problem of robotic manipulation of granular media. We evaluate multiple predictive models used to infer the dynamics of scooping and dumping actions. These models are evaluated on a task that involves manipulating the media in order to deform it into a desired shape. Our best performing model is based on a highly-tailored convolutional network architecture with domain-specific optimizations, which we show accurately models the physical interaction of the robotic scoop with the underlying media. We empirically demonstrate that explicitly predicting physical mechanics results in a policy that out-performs both a hand-crafted dynamics baseline, and a “value-network”, which must otherwise implicitly predict the same mechanics in order to produce accurate value estimates.
APA
Schenck, C., Tompson, J., Levine, S. & Fox, D.. (2017). Learning Robotic Manipulation of Granular Media. Proceedings of the 1st Annual Conference on Robot Learning, in PMLR 78:239-248

Related Material