A Persistent Spatial Semantic Representation for High-level Natural Language Instruction Execution

Valts Blukis, Chris Paxton, Dieter Fox, Animesh Garg, Yoav Artzi
Proceedings of the 5th Conference on Robot Learning, PMLR 164:706-717, 2022.

Abstract

Natural language provides an accessible and expressive interface to specify long-term tasks for robotic agents. However, non-experts are likely to specify such tasks with high-level instructions, which abstract over specific robot actions through several layers of abstraction. We propose that key to bridging this gap between language and robot actions over long execution horizons are persistent representations. We propose a persistent spatial semantic representation method, and show how it enables building an agent that performs hierarchical reasoning to effectively execute long-term tasks. We evaluate our approach on the ALFRED benchmark and achieve state-of-the-art results, despite completely avoiding the commonly used step-by-step instructions. https://hlsm-alfred.github.io/

Cite this Paper

BibTeX
@InProceedings{pmlr-v164-blukis22a, title = {A Persistent Spatial Semantic Representation for High-level Natural Language Instruction Execution}, author = {Blukis, Valts and Paxton, Chris and Fox, Dieter and Garg, Animesh and Artzi, Yoav}, booktitle = {Proceedings of the 5th Conference on Robot Learning}, pages = {706--717}, 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/blukis22a/blukis22a.pdf}, url = {https://proceedings.mlr.press/v164/blukis22a.html}, abstract = {Natural language provides an accessible and expressive interface to specify long-term tasks for robotic agents. However, non-experts are likely to specify such tasks with high-level instructions, which abstract over specific robot actions through several layers of abstraction. We propose that key to bridging this gap between language and robot actions over long execution horizons are persistent representations. We propose a persistent spatial semantic representation method, and show how it enables building an agent that performs hierarchical reasoning to effectively execute long-term tasks. We evaluate our approach on the ALFRED benchmark and achieve state-of-the-art results, despite completely avoiding the commonly used step-by-step instructions. https://hlsm-alfred.github.io/} }
Endnote
%0 Conference Paper %T A Persistent Spatial Semantic Representation for High-level Natural Language Instruction Execution %A Valts Blukis %A Chris Paxton %A Dieter Fox %A Animesh Garg %A Yoav Artzi %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-blukis22a %I PMLR %P 706--717 %U https://proceedings.mlr.press/v164/blukis22a.html %V 164 %X Natural language provides an accessible and expressive interface to specify long-term tasks for robotic agents. However, non-experts are likely to specify such tasks with high-level instructions, which abstract over specific robot actions through several layers of abstraction. We propose that key to bridging this gap between language and robot actions over long execution horizons are persistent representations. We propose a persistent spatial semantic representation method, and show how it enables building an agent that performs hierarchical reasoning to effectively execute long-term tasks. We evaluate our approach on the ALFRED benchmark and achieve state-of-the-art results, despite completely avoiding the commonly used step-by-step instructions. https://hlsm-alfred.github.io/
APA
Blukis, V., Paxton, C., Fox, D., Garg, A. & Artzi, Y.. (2022). A Persistent Spatial Semantic Representation for High-level Natural Language Instruction Execution. Proceedings of the 5th Conference on Robot Learning, in Proceedings of Machine Learning Research 164:706-717 Available from https://proceedings.mlr.press/v164/blukis22a.html.

Related Material