Asking Easy Questions: A User-Friendly Approach to Active Reward Learning

Erdem B\iy\ik, Malayandi Palan, Nicholas C. Landolfi, Dylan P. Losey, Dorsa Sadigh
Proceedings of the Conference on Robot Learning, PMLR 100:1177-1190, 2020.

Abstract

Robots can learn the right reward function by querying a human expert. Existing approaches attempt to choose questions where the robot is most uncertain about the human’s response; however, they do not consider how easy it will be for the human to answer! In this paper we explore an information gain formulation for optimally selecting questions that naturally account for the human’s ability to answer. Our approach identifies questions that optimize the trade-off between robot and human uncertainty, and determines when these questions become redundant or costly. Simulations and a user study show our method not only produces easy questions, but also ultimately results in faster reward learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v100-b-iy-ik20a, title = {Asking Easy Questions: A User-Friendly Approach to Active Reward Learning}, author = {B\iy\ik, Erdem and Palan, Malayandi and Landolfi, Nicholas C. and Losey, Dylan P. and Sadigh, Dorsa}, booktitle = {Proceedings of the Conference on Robot Learning}, pages = {1177--1190}, 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/b-iy-ik20a/b-iy-ik20a.pdf}, url = {https://proceedings.mlr.press/v100/b-iy-ik20a.html}, abstract = {Robots can learn the right reward function by querying a human expert. Existing approaches attempt to choose questions where the robot is most uncertain about the human’s response; however, they do not consider how easy it will be for the human to answer! In this paper we explore an information gain formulation for optimally selecting questions that naturally account for the human’s ability to answer. Our approach identifies questions that optimize the trade-off between robot and human uncertainty, and determines when these questions become redundant or costly. Simulations and a user study show our method not only produces easy questions, but also ultimately results in faster reward learning.} }
Endnote
%0 Conference Paper %T Asking Easy Questions: A User-Friendly Approach to Active Reward Learning %A Erdem B\iy\ik %A Malayandi Palan %A Nicholas C. Landolfi %A Dylan P. Losey %A Dorsa Sadigh %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-b-iy-ik20a %I PMLR %P 1177--1190 %U https://proceedings.mlr.press/v100/b-iy-ik20a.html %V 100 %X Robots can learn the right reward function by querying a human expert. Existing approaches attempt to choose questions where the robot is most uncertain about the human’s response; however, they do not consider how easy it will be for the human to answer! In this paper we explore an information gain formulation for optimally selecting questions that naturally account for the human’s ability to answer. Our approach identifies questions that optimize the trade-off between robot and human uncertainty, and determines when these questions become redundant or costly. Simulations and a user study show our method not only produces easy questions, but also ultimately results in faster reward learning.
APA
B\iy\ik, E., Palan, M., Landolfi, N.C., Losey, D.P. & Sadigh, D.. (2020). Asking Easy Questions: A User-Friendly Approach to Active Reward Learning. Proceedings of the Conference on Robot Learning, in Proceedings of Machine Learning Research 100:1177-1190 Available from https://proceedings.mlr.press/v100/b-iy-ik20a.html.

Related Material