Visually-Grounded Library of Behaviors for Manipulating Diverse Objects across Diverse Configurations and Views

Jingyun Yang, Hsiao-Yu Tung, Yunchu Zhang, Gaurav Pathak, Ashwini Pokle, Christopher G Atkeson, Katerina Fragkiadaki
Proceedings of the 5th Conference on Robot Learning, PMLR 164:695-705, 2022.

Abstract

We propose a visually-grounded library of behaviors approach for learning to manipulate diverse objects across varying initial and goal configurations and camera placements. Our key innovation is to disentangle the standard image-to-action mapping into two separate modules that use different types of perceptual input: (1) a behavior selector which conditions on intrinsic and semantically-rich object appearance features to select the behaviors that can successfully perform the desired tasks on the object in hand, and (2) a library of behaviors each of which conditions on extrinsic and abstract object properties, such as object location and pose, to predict actions to execute over time. The selector uses a semantically-rich 3D object feature representation extracted from images in a differential end-to-end manner. This representation is trained to be view-invariant and affordance-aware using self-supervision, by predicting varying views and successful object manipulations. We test our framework on pushing and grasping diverse objects in simulation as well as transporting rigid, granular, and liquid food ingredients in a real robot setup. Our model outperforms image-to-action mappings that do not factorize static and dynamic object properties. We further ablate the contribution of the selector’s input and show the benefits of the proposed view-predictive, affordance-aware 3D visual object representations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v164-yang22c, title = {Visually-Grounded Library of Behaviors for Manipulating Diverse Objects across Diverse Configurations and Views}, author = {Yang, Jingyun and Tung, Hsiao-Yu and Zhang, Yunchu and Pathak, Gaurav and Pokle, Ashwini and Atkeson, Christopher G and Fragkiadaki, Katerina}, booktitle = {Proceedings of the 5th Conference on Robot Learning}, pages = {695--705}, 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/yang22c/yang22c.pdf}, url = {https://proceedings.mlr.press/v164/yang22c.html}, abstract = {We propose a visually-grounded library of behaviors approach for learning to manipulate diverse objects across varying initial and goal configurations and camera placements. Our key innovation is to disentangle the standard image-to-action mapping into two separate modules that use different types of perceptual input: (1) a behavior selector which conditions on intrinsic and semantically-rich object appearance features to select the behaviors that can successfully perform the desired tasks on the object in hand, and (2) a library of behaviors each of which conditions on extrinsic and abstract object properties, such as object location and pose, to predict actions to execute over time. The selector uses a semantically-rich 3D object feature representation extracted from images in a differential end-to-end manner. This representation is trained to be view-invariant and affordance-aware using self-supervision, by predicting varying views and successful object manipulations. We test our framework on pushing and grasping diverse objects in simulation as well as transporting rigid, granular, and liquid food ingredients in a real robot setup. Our model outperforms image-to-action mappings that do not factorize static and dynamic object properties. We further ablate the contribution of the selector’s input and show the benefits of the proposed view-predictive, affordance-aware 3D visual object representations.} }
Endnote
%0 Conference Paper %T Visually-Grounded Library of Behaviors for Manipulating Diverse Objects across Diverse Configurations and Views %A Jingyun Yang %A Hsiao-Yu Tung %A Yunchu Zhang %A Gaurav Pathak %A Ashwini Pokle %A Christopher G Atkeson %A Katerina Fragkiadaki %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-yang22c %I PMLR %P 695--705 %U https://proceedings.mlr.press/v164/yang22c.html %V 164 %X We propose a visually-grounded library of behaviors approach for learning to manipulate diverse objects across varying initial and goal configurations and camera placements. Our key innovation is to disentangle the standard image-to-action mapping into two separate modules that use different types of perceptual input: (1) a behavior selector which conditions on intrinsic and semantically-rich object appearance features to select the behaviors that can successfully perform the desired tasks on the object in hand, and (2) a library of behaviors each of which conditions on extrinsic and abstract object properties, such as object location and pose, to predict actions to execute over time. The selector uses a semantically-rich 3D object feature representation extracted from images in a differential end-to-end manner. This representation is trained to be view-invariant and affordance-aware using self-supervision, by predicting varying views and successful object manipulations. We test our framework on pushing and grasping diverse objects in simulation as well as transporting rigid, granular, and liquid food ingredients in a real robot setup. Our model outperforms image-to-action mappings that do not factorize static and dynamic object properties. We further ablate the contribution of the selector’s input and show the benefits of the proposed view-predictive, affordance-aware 3D visual object representations.
APA
Yang, J., Tung, H., Zhang, Y., Pathak, G., Pokle, A., Atkeson, C.G. & Fragkiadaki, K.. (2022). Visually-Grounded Library of Behaviors for Manipulating Diverse Objects across Diverse Configurations and Views. Proceedings of the 5th Conference on Robot Learning, in Proceedings of Machine Learning Research 164:695-705 Available from https://proceedings.mlr.press/v164/yang22c.html.

Related Material