Robots that can adapt like animals.
Identifieur interne : 000051 ( PubMed/Corpus ); précédent : 000050; suivant : 000052Robots that can adapt like animals.
Auteurs : Antoine Cully ; Jeff Clune ; Danesh Tarapore ; Jean-Baptiste MouretSource :
- Nature [ 1476-4687 ] ; 2015.
English descriptors
- KwdEn :
- MESH :
- injuries : Extremities.
- instrumentation : Robotics.
- methods : Biomimetics, Robotics.
- physiopathology : Extremities.
- Adaptation, Physiological, Algorithms, Animals, Artificial Intelligence, Behavior, Animal, Dogs, Motor Skills, Time Factors.
Abstract
Robots have transformed many industries, most notably manufacturing, and have the power to deliver tremendous benefits to society, such as in search and rescue, disaster response, health care and transportation. They are also invaluable tools for scientific exploration in environments inaccessible to humans, from distant planets to deep oceans. A major obstacle to their widespread adoption in more complex environments outside factories is their fragility. Whereas animals can quickly adapt to injuries, current robots cannot 'think outside the box' to find a compensatory behaviour when they are damaged: they are limited to their pre-specified self-sensing abilities, can diagnose only anticipated failure modes, and require a pre-programmed contingency plan for every type of potential damage, an impracticality for complex robots. A promising approach to reducing robot fragility involves having robots learn appropriate behaviours in response to damage, but current techniques are slow even with small, constrained search spaces. Here we introduce an intelligent trial-and-error algorithm that allows robots to adapt to damage in less than two minutes in large search spaces without requiring self-diagnosis or pre-specified contingency plans. Before the robot is deployed, it uses a novel technique to create a detailed map of the space of high-performing behaviours. This map represents the robot's prior knowledge about what behaviours it can perform and their value. When the robot is damaged, it uses this prior knowledge to guide a trial-and-error learning algorithm that conducts intelligent experiments to rapidly discover a behaviour that compensates for the damage. Experiments reveal successful adaptations for a legged robot injured in five different ways, including damaged, broken, and missing legs, and for a robotic arm with joints broken in 14 different ways. This new algorithm will enable more robust, effective, autonomous robots, and may shed light on the principles that animals use to adapt to injury.
DOI: 10.1038/nature14422
PubMed: 26017452
Links to Exploration step
pubmed:26017452Le document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en">Robots that can adapt like animals.</title>
<author><name sortKey="Cully, Antoine" sort="Cully, Antoine" uniqKey="Cully A" first="Antoine" last="Cully">Antoine Cully</name>
<affiliation><nlm:affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France.</nlm:affiliation>
</affiliation>
</author>
<author><name sortKey="Clune, Jeff" sort="Clune, Jeff" uniqKey="Clune J" first="Jeff" last="Clune">Jeff Clune</name>
<affiliation><nlm:affiliation>Department of Computer Science, University of Wyoming, Laramie, Wyoming 82071, USA.</nlm:affiliation>
</affiliation>
</author>
<author><name sortKey="Tarapore, Danesh" sort="Tarapore, Danesh" uniqKey="Tarapore D" first="Danesh" last="Tarapore">Danesh Tarapore</name>
<affiliation><nlm:affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France.</nlm:affiliation>
</affiliation>
</author>
<author><name sortKey="Mouret, Jean Baptiste" sort="Mouret, Jean Baptiste" uniqKey="Mouret J" first="Jean-Baptiste" last="Mouret">Jean-Baptiste Mouret</name>
<affiliation><nlm:affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [3] Inria, Team Larsen, Villers-lès-Nancy, F-54600, France [4] CNRS, Loria, UMR 7503, Vandœuvre-lès-Nancy, F-54500, France [5] Université de Lorraine, Loria, UMR 7503, Vandœuvre-lès-Nancy, F-54500, France.</nlm:affiliation>
</affiliation>
</author>
</titleStmt>
<publicationStmt><idno type="wicri:source">PubMed</idno>
<date when="2015">2015</date>
<idno type="doi">10.1038/nature14422</idno>
<idno type="RBID">pubmed:26017452</idno>
<idno type="pmid">26017452</idno>
<idno type="wicri:Area/PubMed/Corpus">000051</idno>
<idno type="wicri:explorRef" wicri:stream="PubMed" wicri:step="Corpus" wicri:corpus="PubMed">000051</idno>
</publicationStmt>
<sourceDesc><biblStruct><analytic><title xml:lang="en">Robots that can adapt like animals.</title>
<author><name sortKey="Cully, Antoine" sort="Cully, Antoine" uniqKey="Cully A" first="Antoine" last="Cully">Antoine Cully</name>
<affiliation><nlm:affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France.</nlm:affiliation>
</affiliation>
</author>
<author><name sortKey="Clune, Jeff" sort="Clune, Jeff" uniqKey="Clune J" first="Jeff" last="Clune">Jeff Clune</name>
<affiliation><nlm:affiliation>Department of Computer Science, University of Wyoming, Laramie, Wyoming 82071, USA.</nlm:affiliation>
</affiliation>
</author>
<author><name sortKey="Tarapore, Danesh" sort="Tarapore, Danesh" uniqKey="Tarapore D" first="Danesh" last="Tarapore">Danesh Tarapore</name>
<affiliation><nlm:affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France.</nlm:affiliation>
</affiliation>
</author>
<author><name sortKey="Mouret, Jean Baptiste" sort="Mouret, Jean Baptiste" uniqKey="Mouret J" first="Jean-Baptiste" last="Mouret">Jean-Baptiste Mouret</name>
<affiliation><nlm:affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [3] Inria, Team Larsen, Villers-lès-Nancy, F-54600, France [4] CNRS, Loria, UMR 7503, Vandœuvre-lès-Nancy, F-54500, France [5] Université de Lorraine, Loria, UMR 7503, Vandœuvre-lès-Nancy, F-54500, France.</nlm:affiliation>
</affiliation>
</author>
</analytic>
<series><title level="j">Nature</title>
<idno type="eISSN">1476-4687</idno>
<imprint><date when="2015" type="published">2015</date>
</imprint>
</series>
</biblStruct>
</sourceDesc>
</fileDesc>
<profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>Adaptation, Physiological</term>
<term>Algorithms</term>
<term>Animals</term>
<term>Artificial Intelligence</term>
<term>Behavior, Animal</term>
<term>Biomimetics (methods)</term>
<term>Dogs</term>
<term>Extremities (injuries)</term>
<term>Extremities (physiopathology)</term>
<term>Motor Skills</term>
<term>Robotics (instrumentation)</term>
<term>Robotics (methods)</term>
<term>Time Factors</term>
</keywords>
<keywords scheme="MESH" qualifier="injuries" xml:lang="en"><term>Extremities</term>
</keywords>
<keywords scheme="MESH" qualifier="instrumentation" xml:lang="en"><term>Robotics</term>
</keywords>
<keywords scheme="MESH" qualifier="methods" xml:lang="en"><term>Biomimetics</term>
<term>Robotics</term>
</keywords>
<keywords scheme="MESH" qualifier="physiopathology" xml:lang="en"><term>Extremities</term>
</keywords>
<keywords scheme="MESH" xml:lang="en"><term>Adaptation, Physiological</term>
<term>Algorithms</term>
<term>Animals</term>
<term>Artificial Intelligence</term>
<term>Behavior, Animal</term>
<term>Dogs</term>
<term>Motor Skills</term>
<term>Time Factors</term>
</keywords>
</textClass>
</profileDesc>
</teiHeader>
<front><div type="abstract" xml:lang="en">Robots have transformed many industries, most notably manufacturing, and have the power to deliver tremendous benefits to society, such as in search and rescue, disaster response, health care and transportation. They are also invaluable tools for scientific exploration in environments inaccessible to humans, from distant planets to deep oceans. A major obstacle to their widespread adoption in more complex environments outside factories is their fragility. Whereas animals can quickly adapt to injuries, current robots cannot 'think outside the box' to find a compensatory behaviour when they are damaged: they are limited to their pre-specified self-sensing abilities, can diagnose only anticipated failure modes, and require a pre-programmed contingency plan for every type of potential damage, an impracticality for complex robots. A promising approach to reducing robot fragility involves having robots learn appropriate behaviours in response to damage, but current techniques are slow even with small, constrained search spaces. Here we introduce an intelligent trial-and-error algorithm that allows robots to adapt to damage in less than two minutes in large search spaces without requiring self-diagnosis or pre-specified contingency plans. Before the robot is deployed, it uses a novel technique to create a detailed map of the space of high-performing behaviours. This map represents the robot's prior knowledge about what behaviours it can perform and their value. When the robot is damaged, it uses this prior knowledge to guide a trial-and-error learning algorithm that conducts intelligent experiments to rapidly discover a behaviour that compensates for the damage. Experiments reveal successful adaptations for a legged robot injured in five different ways, including damaged, broken, and missing legs, and for a robotic arm with joints broken in 14 different ways. This new algorithm will enable more robust, effective, autonomous robots, and may shed light on the principles that animals use to adapt to injury.</div>
</front>
</TEI>
<pubmed><MedlineCitation Owner="NLM" Status="MEDLINE"><PMID Version="1">26017452</PMID>
<DateCreated><Year>2015</Year>
<Month>05</Month>
<Day>28</Day>
</DateCreated>
<DateCompleted><Year>2015</Year>
<Month>06</Month>
<Day>29</Day>
</DateCompleted>
<Article PubModel="Print"><Journal><ISSN IssnType="Electronic">1476-4687</ISSN>
<JournalIssue CitedMedium="Internet"><Volume>521</Volume>
<Issue>7553</Issue>
<PubDate><Year>2015</Year>
<Month>May</Month>
<Day>28</Day>
</PubDate>
</JournalIssue>
<Title>Nature</Title>
<ISOAbbreviation>Nature</ISOAbbreviation>
</Journal>
<ArticleTitle>Robots that can adapt like animals.</ArticleTitle>
<Pagination><MedlinePgn>503-7</MedlinePgn>
</Pagination>
<ELocationID EIdType="doi" ValidYN="Y">10.1038/nature14422</ELocationID>
<Abstract><AbstractText>Robots have transformed many industries, most notably manufacturing, and have the power to deliver tremendous benefits to society, such as in search and rescue, disaster response, health care and transportation. They are also invaluable tools for scientific exploration in environments inaccessible to humans, from distant planets to deep oceans. A major obstacle to their widespread adoption in more complex environments outside factories is their fragility. Whereas animals can quickly adapt to injuries, current robots cannot 'think outside the box' to find a compensatory behaviour when they are damaged: they are limited to their pre-specified self-sensing abilities, can diagnose only anticipated failure modes, and require a pre-programmed contingency plan for every type of potential damage, an impracticality for complex robots. A promising approach to reducing robot fragility involves having robots learn appropriate behaviours in response to damage, but current techniques are slow even with small, constrained search spaces. Here we introduce an intelligent trial-and-error algorithm that allows robots to adapt to damage in less than two minutes in large search spaces without requiring self-diagnosis or pre-specified contingency plans. Before the robot is deployed, it uses a novel technique to create a detailed map of the space of high-performing behaviours. This map represents the robot's prior knowledge about what behaviours it can perform and their value. When the robot is damaged, it uses this prior knowledge to guide a trial-and-error learning algorithm that conducts intelligent experiments to rapidly discover a behaviour that compensates for the damage. Experiments reveal successful adaptations for a legged robot injured in five different ways, including damaged, broken, and missing legs, and for a robotic arm with joints broken in 14 different ways. This new algorithm will enable more robust, effective, autonomous robots, and may shed light on the principles that animals use to adapt to injury.</AbstractText>
</Abstract>
<AuthorList CompleteYN="Y"><Author ValidYN="Y"><LastName>Cully</LastName>
<ForeName>Antoine</ForeName>
<Initials>A</Initials>
<AffiliationInfo><Affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France.</Affiliation>
</AffiliationInfo>
</Author>
<Author ValidYN="Y"><LastName>Clune</LastName>
<ForeName>Jeff</ForeName>
<Initials>J</Initials>
<AffiliationInfo><Affiliation>Department of Computer Science, University of Wyoming, Laramie, Wyoming 82071, USA.</Affiliation>
</AffiliationInfo>
</Author>
<Author ValidYN="Y"><LastName>Tarapore</LastName>
<ForeName>Danesh</ForeName>
<Initials>D</Initials>
<AffiliationInfo><Affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France.</Affiliation>
</AffiliationInfo>
</Author>
<Author ValidYN="Y"><LastName>Mouret</LastName>
<ForeName>Jean-Baptiste</ForeName>
<Initials>JB</Initials>
<AffiliationInfo><Affiliation>1] Sorbonne Universités, Université Pierre et Marie Curie (UPMC), Paris 06, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [2] CNRS, UMR 7222, Institut des Systèmes Intelligents et de Robotique (ISIR), F-75005, Paris, France [3] Inria, Team Larsen, Villers-lès-Nancy, F-54600, France [4] CNRS, Loria, UMR 7503, Vandœuvre-lès-Nancy, F-54500, France [5] Université de Lorraine, Loria, UMR 7503, Vandœuvre-lès-Nancy, F-54500, France.</Affiliation>
</AffiliationInfo>
</Author>
</AuthorList>
<Language>eng</Language>
<PublicationTypeList><PublicationType UI="D016428">Journal Article</PublicationType>
<PublicationType UI="D013485">Research Support, Non-U.S. Gov't</PublicationType>
</PublicationTypeList>
</Article>
<MedlineJournalInfo><Country>England</Country>
<MedlineTA>Nature</MedlineTA>
<NlmUniqueID>0410462</NlmUniqueID>
<ISSNLinking>0028-0836</ISSNLinking>
</MedlineJournalInfo>
<CitationSubset>IM</CitationSubset>
<CommentsCorrectionsList><CommentsCorrections RefType="CommentIn"><RefSource>Nature. 2015 May 28;521(7553):426-7</RefSource>
<PMID Version="1">26017437</PMID>
</CommentsCorrections>
</CommentsCorrectionsList>
<MeshHeadingList><MeshHeading><DescriptorName MajorTopicYN="Y" UI="D000222">Adaptation, Physiological</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D000465">Algorithms</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D000818">Animals</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="Y" UI="D001185">Artificial Intelligence</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D001522">Behavior, Animal</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D032701">Biomimetics</DescriptorName>
<QualifierName MajorTopicYN="Y" UI="Q000379">methods</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D004285">Dogs</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D005121">Extremities</DescriptorName>
<QualifierName MajorTopicYN="Y" UI="Q000293">injuries</QualifierName>
<QualifierName MajorTopicYN="N" UI="Q000503">physiopathology</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D009048">Motor Skills</DescriptorName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D012371">Robotics</DescriptorName>
<QualifierName MajorTopicYN="Y" UI="Q000295">instrumentation</QualifierName>
<QualifierName MajorTopicYN="Y" UI="Q000379">methods</QualifierName>
</MeshHeading>
<MeshHeading><DescriptorName MajorTopicYN="N" UI="D013997">Time Factors</DescriptorName>
</MeshHeading>
</MeshHeadingList>
</MedlineCitation>
<PubmedData><History><PubMedPubDate PubStatus="received"><Year>2014</Year>
<Month>12</Month>
<Day>24</Day>
</PubMedPubDate>
<PubMedPubDate PubStatus="accepted"><Year>2015</Year>
<Month>3</Month>
<Day>17</Day>
</PubMedPubDate>
<PubMedPubDate PubStatus="entrez"><Year>2015</Year>
<Month>5</Month>
<Day>29</Day>
<Hour>6</Hour>
<Minute>0</Minute>
</PubMedPubDate>
<PubMedPubDate PubStatus="pubmed"><Year>2015</Year>
<Month>5</Month>
<Day>29</Day>
<Hour>6</Hour>
<Minute>0</Minute>
</PubMedPubDate>
<PubMedPubDate PubStatus="medline"><Year>2015</Year>
<Month>6</Month>
<Day>30</Day>
<Hour>6</Hour>
<Minute>0</Minute>
</PubMedPubDate>
</History>
<PublicationStatus>ppublish</PublicationStatus>
<ArticleIdList><ArticleId IdType="pii">nature14422</ArticleId>
<ArticleId IdType="doi">10.1038/nature14422</ArticleId>
<ArticleId IdType="pubmed">26017452</ArticleId>
</ArticleIdList>
</PubmedData>
</pubmed>
</record>
Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Wicri/Lorraine/explor/InforLorV4/Data/PubMed/Corpus
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000051 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/PubMed/Corpus/biblio.hfd -nk 000051 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien |wiki= Wicri/Lorraine |area= InforLorV4 |flux= PubMed |étape= Corpus |type= RBID |clé= pubmed:26017452 |texte= Robots that can adapt like animals. }}
Pour générer des pages wiki
HfdIndexSelect -h $EXPLOR_AREA/Data/PubMed/Corpus/RBID.i -Sk "pubmed:26017452" \ | HfdSelect -Kh $EXPLOR_AREA/Data/PubMed/Corpus/biblio.hfd \ | NlmPubMed2Wicri -a InforLorV4
![]() | This area was generated with Dilib version V0.6.33. | ![]() |