Learning value functions with relational state representations for guiding task-and-motion planning

Beomjoon Kim, Luke Shimanuki
Proceedings of the Conference on Robot Learning, PMLR 100:955-968, 2020.

Abstract

We propose a novel relational state representation and an action-value function learning algorithm that learns from planning experience for geometric task-and-motion planning (GTAMP) problems, in which the goal is to move several objects to regions in the presence of movable obstacles. The representation encodes information about which objects occlude the manipulation of other objects and is encoded using a small set of predicates. It supports efficient learning, using graph neural networks, of an action-value function that can be used to guide a GTAMP solver. Importantly, it enables learning from planning experience on simple problems and generalizing to more complex problems and even across substantially different geometric environments. We demonstrate the method in two challenging GTAMP domains.

Cite this Paper

BibTeX
@InProceedings{pmlr-v100-kim20a, title = {Learning value functions with relational state representations for guiding task-and-motion planning}, author = {Kim, Beomjoon and Shimanuki, Luke}, booktitle = {Proceedings of the Conference on Robot Learning}, pages = {955--968}, year = {2020}, editor = {Kaelbling, Leslie Pack and Kragic, Danica and Sugiura, Komei}, volume = {100}, series = {Proceedings of Machine Learning Research}, month = {30 Oct--01 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v100/kim20a/kim20a.pdf}, url = {https://proceedings.mlr.press/v100/kim20a.html}, abstract = {We propose a novel relational state representation and an action-value function learning algorithm that learns from planning experience for geometric task-and-motion planning (GTAMP) problems, in which the goal is to move several objects to regions in the presence of movable obstacles. The representation encodes information about which objects occlude the manipulation of other objects and is encoded using a small set of predicates. It supports efficient learning, using graph neural networks, of an action-value function that can be used to guide a GTAMP solver. Importantly, it enables learning from planning experience on simple problems and generalizing to more complex problems and even across substantially different geometric environments. We demonstrate the method in two challenging GTAMP domains.} }
Endnote
%0 Conference Paper %T Learning value functions with relational state representations for guiding task-and-motion planning %A Beomjoon Kim %A Luke Shimanuki %B Proceedings of the Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2020 %E Leslie Pack Kaelbling %E Danica Kragic %E Komei Sugiura %F pmlr-v100-kim20a %I PMLR %P 955--968 %U https://proceedings.mlr.press/v100/kim20a.html %V 100 %X We propose a novel relational state representation and an action-value function learning algorithm that learns from planning experience for geometric task-and-motion planning (GTAMP) problems, in which the goal is to move several objects to regions in the presence of movable obstacles. The representation encodes information about which objects occlude the manipulation of other objects and is encoded using a small set of predicates. It supports efficient learning, using graph neural networks, of an action-value function that can be used to guide a GTAMP solver. Importantly, it enables learning from planning experience on simple problems and generalizing to more complex problems and even across substantially different geometric environments. We demonstrate the method in two challenging GTAMP domains.
APA
Kim, B. & Shimanuki, L.. (2020). Learning value functions with relational state representations for guiding task-and-motion planning. Proceedings of the Conference on Robot Learning, in Proceedings of Machine Learning Research 100:955-968 Available from https://proceedings.mlr.press/v100/kim20a.html.

Related Material