Learning to Map Natural Language Instructions to Physical Quadcopter Control using Simulated Flight

Valts Blukis, Yannick Terme, Eyvind Niklasson, Ross A. Knepper, Yoav Artzi
; Proceedings of the Conference on Robot Learning, PMLR 100:1415-1438, 2020.

Abstract

We propose a joint simulation and real-world learning framework for mapping navigation instructions and raw first-person observations to continuous control. Our model estimates the need for environment exploration, predicts the likelihood of visiting environment positions during execution, and controls the agent to both explore and visit high-likelihood positions. We introduce Supervised Reinforcement Asynchronous Learning (SuReAL). Learning uses both simulation and real environments without requiring autonomous flight in the physical environment during training, and combines supervised learning for predicting positions to visit and reinforcement learning for continuous control. We evaluate our approach on a natural language instruction-following task with a physical quadcopter, and demonstrate effective execution and exploration behavior.

Cite this Paper


BibTeX
@InProceedings{pmlr-v100-blukis20a, title = {Learning to Map Natural Language Instructions to Physical Quadcopter Control using Simulated Flight}, author = {Blukis, Valts and Terme, Yannick and Niklasson, Eyvind and Knepper, Ross A. and Artzi, Yoav}, booktitle = {Proceedings of the Conference on Robot Learning}, pages = {1415--1438}, year = {2020}, editor = {Leslie Pack Kaelbling and Danica Kragic and Komei Sugiura}, volume = {100}, series = {Proceedings of Machine Learning Research}, address = {}, month = {30 Oct--01 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v100/blukis20a/blukis20a.pdf}, url = {http://proceedings.mlr.press/v100/blukis20a.html}, abstract = {We propose a joint simulation and real-world learning framework for mapping navigation instructions and raw first-person observations to continuous control. Our model estimates the need for environment exploration, predicts the likelihood of visiting environment positions during execution, and controls the agent to both explore and visit high-likelihood positions. We introduce Supervised Reinforcement Asynchronous Learning (SuReAL). Learning uses both simulation and real environments without requiring autonomous flight in the physical environment during training, and combines supervised learning for predicting positions to visit and reinforcement learning for continuous control. We evaluate our approach on a natural language instruction-following task with a physical quadcopter, and demonstrate effective execution and exploration behavior.} }
Endnote
%0 Conference Paper %T Learning to Map Natural Language Instructions to Physical Quadcopter Control using Simulated Flight %A Valts Blukis %A Yannick Terme %A Eyvind Niklasson %A Ross A. Knepper %A Yoav Artzi %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-blukis20a %I PMLR %J Proceedings of Machine Learning Research %P 1415--1438 %U http://proceedings.mlr.press %V 100 %W PMLR %X We propose a joint simulation and real-world learning framework for mapping navigation instructions and raw first-person observations to continuous control. Our model estimates the need for environment exploration, predicts the likelihood of visiting environment positions during execution, and controls the agent to both explore and visit high-likelihood positions. We introduce Supervised Reinforcement Asynchronous Learning (SuReAL). Learning uses both simulation and real environments without requiring autonomous flight in the physical environment during training, and combines supervised learning for predicting positions to visit and reinforcement learning for continuous control. We evaluate our approach on a natural language instruction-following task with a physical quadcopter, and demonstrate effective execution and exploration behavior.
APA
Blukis, V., Terme, Y., Niklasson, E., Knepper, R.A. & Artzi, Y.. (2020). Learning to Map Natural Language Instructions to Physical Quadcopter Control using Simulated Flight. Proceedings of the Conference on Robot Learning, in PMLR 100:1415-1438

Related Material