ROBOTURK: A Crowdsourcing Platform for Robotic Skill Learning through Imitation

Ajay Mandlekar, Yuke Zhu, Animesh Garg, Jonathan Booher, Max Spero, Albert Tung, Julian Gao, John Emmons, Anchit Gupta, Emre Orbay, Silvio Savarese, Li Fei-Fei
Proceedings of The 2nd Conference on Robot Learning, PMLR 87:879-893, 2018.

Abstract

Imitation Learning has empowered recent advances in learning robotic manipulation tasks by addressing shortcomings of Reinforcement Learning such as exploration and reward specification. However, research in this area has been limited to modest-sized datasets due to the difficulty of collecting large quantities of task demonstrations through existing mechanisms. This work introduces ROBO-TURK to address this challenge. ROBOTURK is a crowdsourcing platform for high quality 6-DoF trajectory based teleoperation through the use of widely available mobile devices (e.g. iPhone). We evaluate ROBOTURK on three manipulation tasks of varying timescales (15-120s) and observe that our user interface is statistically similar to special purpose hardware such as virtual reality controllers in terms of task completion times. Furthermore, we observe that poor network conditions, such as low bandwidth and high delay links, do not substantially affect the remote users’ ability to perform task demonstrations successfully on ROBOTURK. Lastly, we demonstrate the efficacy of ROBOTURK through the collection of a pilot dataset; using ROBOTURK, we collected 137.5 hours of manipulation data from remote workers, amounting to over 2200 successful task demonstrations in 22 hours of total system usage. We show that the data obtained through ROBOTURK enables policy learning on multi-step manipulation tasks with sparse rewards and that using larger quantities of demonstrations during policy learning provides benefits in terms of both learning consistency and final performance. For additional results, videos, and to download our pilot dataset, visit roboturk.stanford.edu

Cite this Paper

BibTeX
@InProceedings{pmlr-v87-mandlekar18a, title = {ROBOTURK: A Crowdsourcing Platform for Robotic Skill Learning through Imitation}, author = {Mandlekar, Ajay and Zhu, Yuke and Garg, Animesh and Booher, Jonathan and Spero, Max and Tung, Albert and Gao, Julian and Emmons, John and Gupta, Anchit and Orbay, Emre and Savarese, Silvio and Fei-Fei, Li}, booktitle = {Proceedings of The 2nd Conference on Robot Learning}, pages = {879--893}, year = {2018}, editor = {Billard, Aude and Dragan, Anca and Peters, Jan and Morimoto, Jun}, volume = {87}, series = {Proceedings of Machine Learning Research}, month = {29--31 Oct}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v87/mandlekar18a/mandlekar18a.pdf}, url = {https://proceedings.mlr.press/v87/mandlekar18a.html}, abstract = {Imitation Learning has empowered recent advances in learning robotic manipulation tasks by addressing shortcomings of Reinforcement Learning such as exploration and reward specification. However, research in this area has been limited to modest-sized datasets due to the difficulty of collecting large quantities of task demonstrations through existing mechanisms. This work introduces ROBO-TURK to address this challenge. ROBOTURK is a crowdsourcing platform for high quality 6-DoF trajectory based teleoperation through the use of widely available mobile devices (e.g. iPhone). We evaluate ROBOTURK on three manipulation tasks of varying timescales (15-120s) and observe that our user interface is statistically similar to special purpose hardware such as virtual reality controllers in terms of task completion times. Furthermore, we observe that poor network conditions, such as low bandwidth and high delay links, do not substantially affect the remote users’ ability to perform task demonstrations successfully on ROBOTURK. Lastly, we demonstrate the efficacy of ROBOTURK through the collection of a pilot dataset; using ROBOTURK, we collected 137.5 hours of manipulation data from remote workers, amounting to over 2200 successful task demonstrations in 22 hours of total system usage. We show that the data obtained through ROBOTURK enables policy learning on multi-step manipulation tasks with sparse rewards and that using larger quantities of demonstrations during policy learning provides benefits in terms of both learning consistency and final performance. For additional results, videos, and to download our pilot dataset, visit roboturk.stanford.edu } }
Endnote
%0 Conference Paper %T ROBOTURK: A Crowdsourcing Platform for Robotic Skill Learning through Imitation %A Ajay Mandlekar %A Yuke Zhu %A Animesh Garg %A Jonathan Booher %A Max Spero %A Albert Tung %A Julian Gao %A John Emmons %A Anchit Gupta %A Emre Orbay %A Silvio Savarese %A Li Fei-Fei %B Proceedings of The 2nd Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2018 %E Aude Billard %E Anca Dragan %E Jan Peters %E Jun Morimoto %F pmlr-v87-mandlekar18a %I PMLR %P 879--893 %U https://proceedings.mlr.press/v87/mandlekar18a.html %V 87 %X Imitation Learning has empowered recent advances in learning robotic manipulation tasks by addressing shortcomings of Reinforcement Learning such as exploration and reward specification. However, research in this area has been limited to modest-sized datasets due to the difficulty of collecting large quantities of task demonstrations through existing mechanisms. This work introduces ROBO-TURK to address this challenge. ROBOTURK is a crowdsourcing platform for high quality 6-DoF trajectory based teleoperation through the use of widely available mobile devices (e.g. iPhone). We evaluate ROBOTURK on three manipulation tasks of varying timescales (15-120s) and observe that our user interface is statistically similar to special purpose hardware such as virtual reality controllers in terms of task completion times. Furthermore, we observe that poor network conditions, such as low bandwidth and high delay links, do not substantially affect the remote users’ ability to perform task demonstrations successfully on ROBOTURK. Lastly, we demonstrate the efficacy of ROBOTURK through the collection of a pilot dataset; using ROBOTURK, we collected 137.5 hours of manipulation data from remote workers, amounting to over 2200 successful task demonstrations in 22 hours of total system usage. We show that the data obtained through ROBOTURK enables policy learning on multi-step manipulation tasks with sparse rewards and that using larger quantities of demonstrations during policy learning provides benefits in terms of both learning consistency and final performance. For additional results, videos, and to download our pilot dataset, visit roboturk.stanford.edu
APA
Mandlekar, A., Zhu, Y., Garg, A., Booher, J., Spero, M., Tung, A., Gao, J., Emmons, J., Gupta, A., Orbay, E., Savarese, S. & Fei-Fei, L.. (2018). ROBOTURK: A Crowdsourcing Platform for Robotic Skill Learning through Imitation. Proceedings of The 2nd Conference on Robot Learning, in Proceedings of Machine Learning Research 87:879-893 Available from https://proceedings.mlr.press/v87/mandlekar18a.html.

Related Material