Learning to Tidy Up from Household Demonstrations
Transactions on Robot Learning, 4(2), pp. 112--131, 2025
Abstract
Tidying a home requires knowing where every object belongs, yet each household has its own conventions. We present a method that learns a household's placement preferences from a handful of demonstrations and generalizes them to new objects. Across ten real homes, the robot put away 87% of objects where their owners expected them.
BibTeX
@article{brown2025tidy,
author = "Brown$^{*}$, Bob and C{\^o}t{\'e}$^{*}$, Carol and Adams, Alice",
title = "Learning to Tidy Up from Household Demonstrations",
journal = "Transactions on Robot Learning",
volume = "4",
number = "2",
pages = "112--131",
month = "June",
issn = "2999-0001",
year = "2025",
doi = "10.5555/example.trl.2025.0412",
abstract = "Tidying a home requires knowing where every object belongs, yet each household has its own conventions. We present a method that learns a household's placement preferences from a handful of demonstrations and generalizes them to new objects. Across ten real homes, the robot put away 87\% of objects where their owners expected them.",
project = "homebot"
}