Bayesian active learning for collaborative task specification using equivalence regions

Nils Wilde, Dana Kulic, Stephen L J Smith

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Specifying complex task behaviors while ensuring good robot performance may be difficult for untrained users. We study a framework for users to specify rules for acceptable behavior in a shared environment such as industrial facilities. As non-expert users might have little intuition about how their specification impacts the robot's performance, we design a learning system that interacts with users to find an optimal solution. Using active preference learning, we iteratively show alternative paths that the robot could take on an interface. From the user feedback on ranking the alternatives, we learn about the weights that users place on each part of their specification. We extend the user model from our previous work to a discrete Bayesian learning model and introduce a greedy algorithm for proposing alternative that operates on the notion of equivalence regions of user weights. We prove that using this algorithm, the revision active learning process converges on the user-optimal path. Using simulations performed in realistic industrial environments, we demonstrate the convergence and robustness of our approach.

Original languageEnglish
Article number8633961
Pages (from-to)1691-1698
Number of pages8
JournalIEEE Robotics and Automation Letters
Volume4
Issue number2
DOIs
Publication statusPublished - 1 Apr 2019

Keywords

  • Cognitive human-robot interaction
  • motion and path planning

Cite this

@article{8cef16a2fada4519851fcde04d7c537c,
title = "Bayesian active learning for collaborative task specification using equivalence regions",
abstract = "Specifying complex task behaviors while ensuring good robot performance may be difficult for untrained users. We study a framework for users to specify rules for acceptable behavior in a shared environment such as industrial facilities. As non-expert users might have little intuition about how their specification impacts the robot's performance, we design a learning system that interacts with users to find an optimal solution. Using active preference learning, we iteratively show alternative paths that the robot could take on an interface. From the user feedback on ranking the alternatives, we learn about the weights that users place on each part of their specification. We extend the user model from our previous work to a discrete Bayesian learning model and introduce a greedy algorithm for proposing alternative that operates on the notion of equivalence regions of user weights. We prove that using this algorithm, the revision active learning process converges on the user-optimal path. Using simulations performed in realistic industrial environments, we demonstrate the convergence and robustness of our approach.",
keywords = "Cognitive human-robot interaction, motion and path planning",
author = "Nils Wilde and Dana Kulic and Smith, {Stephen L J}",
year = "2019",
month = "4",
day = "1",
doi = "10.1109/LRA.2019.2897342",
language = "English",
volume = "4",
pages = "1691--1698",
journal = "IEEE Robotics and Automation Letters",
issn = "2377-3766",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
number = "2",

}

Bayesian active learning for collaborative task specification using equivalence regions. / Wilde, Nils; Kulic, Dana; Smith, Stephen L J.

In: IEEE Robotics and Automation Letters, Vol. 4, No. 2, 8633961, 01.04.2019, p. 1691-1698.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Bayesian active learning for collaborative task specification using equivalence regions

AU - Wilde, Nils

AU - Kulic, Dana

AU - Smith, Stephen L J

PY - 2019/4/1

Y1 - 2019/4/1

N2 - Specifying complex task behaviors while ensuring good robot performance may be difficult for untrained users. We study a framework for users to specify rules for acceptable behavior in a shared environment such as industrial facilities. As non-expert users might have little intuition about how their specification impacts the robot's performance, we design a learning system that interacts with users to find an optimal solution. Using active preference learning, we iteratively show alternative paths that the robot could take on an interface. From the user feedback on ranking the alternatives, we learn about the weights that users place on each part of their specification. We extend the user model from our previous work to a discrete Bayesian learning model and introduce a greedy algorithm for proposing alternative that operates on the notion of equivalence regions of user weights. We prove that using this algorithm, the revision active learning process converges on the user-optimal path. Using simulations performed in realistic industrial environments, we demonstrate the convergence and robustness of our approach.

AB - Specifying complex task behaviors while ensuring good robot performance may be difficult for untrained users. We study a framework for users to specify rules for acceptable behavior in a shared environment such as industrial facilities. As non-expert users might have little intuition about how their specification impacts the robot's performance, we design a learning system that interacts with users to find an optimal solution. Using active preference learning, we iteratively show alternative paths that the robot could take on an interface. From the user feedback on ranking the alternatives, we learn about the weights that users place on each part of their specification. We extend the user model from our previous work to a discrete Bayesian learning model and introduce a greedy algorithm for proposing alternative that operates on the notion of equivalence regions of user weights. We prove that using this algorithm, the revision active learning process converges on the user-optimal path. Using simulations performed in realistic industrial environments, we demonstrate the convergence and robustness of our approach.

KW - Cognitive human-robot interaction

KW - motion and path planning

UR - http://www.scopus.com/inward/record.url?scp=85063311433&partnerID=8YFLogxK

U2 - 10.1109/LRA.2019.2897342

DO - 10.1109/LRA.2019.2897342

M3 - Article

VL - 4

SP - 1691

EP - 1698

JO - IEEE Robotics and Automation Letters

JF - IEEE Robotics and Automation Letters

SN - 2377-3766

IS - 2

M1 - 8633961

ER -