Invariance Through Latent Alignment

A robot's deployment environment often involves perceptual changes that differ from what it has experienced during training. Standard practices such as data augmentation attempt to bridge this gap by augmenting source images in an effort to extend the support of the training distribution to bet...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yoneda, Takuma, Yang, Ge, Walter, Matthew R, Stadie, Bradly
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Yoneda, Takuma
Yang, Ge
Walter, Matthew R
Stadie, Bradly
description A robot's deployment environment often involves perceptual changes that differ from what it has experienced during training. Standard practices such as data augmentation attempt to bridge this gap by augmenting source images in an effort to extend the support of the training distribution to better cover what the agent might experience at test time. In many cases, however, it is impossible to know test-time distribution-shift a priori, making these schemes infeasible. In this paper, we introduce a general approach, called Invariance Through Latent Alignment (ILA), that improves the test-time performance of a visuomotor control policy in deployment environments with unknown perceptual variations. ILA performs unsupervised adaptation at deployment-time by matching the distribution of latent features on the target domain to the agent's prior experience, without relying on paired data. Although simple, we show that this idea leads to surprising improvements on a variety of challenging adaptation scenarios, including changes in lighting conditions, the content in the scene, and camera poses. We present results on calibrated control benchmarks in simulation -- the distractor control suite -- and a physical robot under a sim-to-real setup.
doi_str_mv 10.48550/arxiv.2112.08526
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2112_08526</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2112_08526</sourcerecordid><originalsourceid>FETCH-LOGICAL-a676-8f63a11313fe007b628c8f9be41a9c296540fcd49b4cb853a43f7d64d0c3be363</originalsourceid><addsrcrecordid>eNotzjsLwjAYheEsDlL9AU4WnFuTfEmajqV4g4JL9_IlTbSgVeIF_fdep_NOh4eQCaOp0FLSOYZHd085YzylWnI1JLNNf8fQYW9dXO_D6bbbxxVeXX-Ni0O364_vGpGBx8PFjf8bkXq5qMt1Um1Xm7KoElSZSrRXgIwBA-8ozYzi2mqfGycY5pbnSgrqbStyI6zRElCAz1olWmrBOFAQkenv9qtszqE7Yng2H23z1cILZGc47A</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Invariance Through Latent Alignment</title><source>arXiv.org</source><creator>Yoneda, Takuma ; Yang, Ge ; Walter, Matthew R ; Stadie, Bradly</creator><creatorcontrib>Yoneda, Takuma ; Yang, Ge ; Walter, Matthew R ; Stadie, Bradly</creatorcontrib><description>A robot's deployment environment often involves perceptual changes that differ from what it has experienced during training. Standard practices such as data augmentation attempt to bridge this gap by augmenting source images in an effort to extend the support of the training distribution to better cover what the agent might experience at test time. In many cases, however, it is impossible to know test-time distribution-shift a priori, making these schemes infeasible. In this paper, we introduce a general approach, called Invariance Through Latent Alignment (ILA), that improves the test-time performance of a visuomotor control policy in deployment environments with unknown perceptual variations. ILA performs unsupervised adaptation at deployment-time by matching the distribution of latent features on the target domain to the agent's prior experience, without relying on paired data. Although simple, we show that this idea leads to surprising improvements on a variety of challenging adaptation scenarios, including changes in lighting conditions, the content in the scene, and camera poses. We present results on calibrated control benchmarks in simulation -- the distractor control suite -- and a physical robot under a sim-to-real setup.</description><identifier>DOI: 10.48550/arxiv.2112.08526</identifier><language>eng</language><subject>Computer Science - Learning ; Computer Science - Robotics</subject><creationdate>2021-12</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2112.08526$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2112.08526$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Yoneda, Takuma</creatorcontrib><creatorcontrib>Yang, Ge</creatorcontrib><creatorcontrib>Walter, Matthew R</creatorcontrib><creatorcontrib>Stadie, Bradly</creatorcontrib><title>Invariance Through Latent Alignment</title><description>A robot's deployment environment often involves perceptual changes that differ from what it has experienced during training. Standard practices such as data augmentation attempt to bridge this gap by augmenting source images in an effort to extend the support of the training distribution to better cover what the agent might experience at test time. In many cases, however, it is impossible to know test-time distribution-shift a priori, making these schemes infeasible. In this paper, we introduce a general approach, called Invariance Through Latent Alignment (ILA), that improves the test-time performance of a visuomotor control policy in deployment environments with unknown perceptual variations. ILA performs unsupervised adaptation at deployment-time by matching the distribution of latent features on the target domain to the agent's prior experience, without relying on paired data. Although simple, we show that this idea leads to surprising improvements on a variety of challenging adaptation scenarios, including changes in lighting conditions, the content in the scene, and camera poses. We present results on calibrated control benchmarks in simulation -- the distractor control suite -- and a physical robot under a sim-to-real setup.</description><subject>Computer Science - Learning</subject><subject>Computer Science - Robotics</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotzjsLwjAYheEsDlL9AU4WnFuTfEmajqV4g4JL9_IlTbSgVeIF_fdep_NOh4eQCaOp0FLSOYZHd085YzylWnI1JLNNf8fQYW9dXO_D6bbbxxVeXX-Ni0O364_vGpGBx8PFjf8bkXq5qMt1Um1Xm7KoElSZSrRXgIwBA-8ozYzi2mqfGycY5pbnSgrqbStyI6zRElCAz1olWmrBOFAQkenv9qtszqE7Yng2H23z1cILZGc47A</recordid><startdate>20211215</startdate><enddate>20211215</enddate><creator>Yoneda, Takuma</creator><creator>Yang, Ge</creator><creator>Walter, Matthew R</creator><creator>Stadie, Bradly</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20211215</creationdate><title>Invariance Through Latent Alignment</title><author>Yoneda, Takuma ; Yang, Ge ; Walter, Matthew R ; Stadie, Bradly</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a676-8f63a11313fe007b628c8f9be41a9c296540fcd49b4cb853a43f7d64d0c3be363</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Computer Science - Learning</topic><topic>Computer Science - Robotics</topic><toplevel>online_resources</toplevel><creatorcontrib>Yoneda, Takuma</creatorcontrib><creatorcontrib>Yang, Ge</creatorcontrib><creatorcontrib>Walter, Matthew R</creatorcontrib><creatorcontrib>Stadie, Bradly</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Yoneda, Takuma</au><au>Yang, Ge</au><au>Walter, Matthew R</au><au>Stadie, Bradly</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Invariance Through Latent Alignment</atitle><date>2021-12-15</date><risdate>2021</risdate><abstract>A robot's deployment environment often involves perceptual changes that differ from what it has experienced during training. Standard practices such as data augmentation attempt to bridge this gap by augmenting source images in an effort to extend the support of the training distribution to better cover what the agent might experience at test time. In many cases, however, it is impossible to know test-time distribution-shift a priori, making these schemes infeasible. In this paper, we introduce a general approach, called Invariance Through Latent Alignment (ILA), that improves the test-time performance of a visuomotor control policy in deployment environments with unknown perceptual variations. ILA performs unsupervised adaptation at deployment-time by matching the distribution of latent features on the target domain to the agent's prior experience, without relying on paired data. Although simple, we show that this idea leads to surprising improvements on a variety of challenging adaptation scenarios, including changes in lighting conditions, the content in the scene, and camera poses. We present results on calibrated control benchmarks in simulation -- the distractor control suite -- and a physical robot under a sim-to-real setup.</abstract><doi>10.48550/arxiv.2112.08526</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2112.08526
ispartof
issn
language eng
recordid cdi_arxiv_primary_2112_08526
source arXiv.org
subjects Computer Science - Learning
Computer Science - Robotics
title Invariance Through Latent Alignment
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-06T02%3A54%3A16IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Invariance%20Through%20Latent%20Alignment&rft.au=Yoneda,%20Takuma&rft.date=2021-12-15&rft_id=info:doi/10.48550/arxiv.2112.08526&rft_dat=%3Carxiv_GOX%3E2112_08526%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true