Learning Diverse Natural Behaviors for Enhancing the Agility of Quadrupedal Robots

Learning Diverse Natural Behaviors for Enhancing the Agility of Quadrupedal Robots | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Learning Diverse Natural Behaviors for Enhancing the Agility of Quadrupedal Robots Chunlin Chen, Huiqiao Fu, Haoyu Dong, Wentao Xu, Zhehao Zhou, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6354356/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Achieving animal-like agility is a longstanding goal in quadrupedal robotics. While recent studies have successfully demonstrated imitation of specific behaviors, enabling robots to replicate a broader range of natural behaviors in real-world environments remains an open challenge. Here we propose an integrated controller comprising a Basic Behavior Controller (BBC) and a Task-Specific Controller (TSC) which can effectively learn diverse natural quadrupedal behaviors in an enhanced simulator and efficiently transfer them to the real world. Specifically, the BBC is trained using a novel semi-supervised generative adversarial imitation learning algorithm to extract diverse behavioral styles from raw motion capture data of real dogs, enabling smooth behavior transitions by adjusting discrete and continuous latent variable inputs. The TSC, trained via privileged learning with depth images as input, coordinates the BBC to efficiently perform various tasks. Additionally, we employ evolutionary adversarial simulator identification to optimize the simulator, aligning it closely with reality. After training, the robot exhibits diverse natural behaviors, successfully completing the quadrupedal agility challenge at an average speed of 1.1 m/s and achieving a peak speed of 3.2 m/s during hurdling. This work represents a substantial step toward animal-like agility in quadrupedal robots, opening avenues for their deployment in increasingly complex real-world environments. Physical sciences/Mathematics and computing/Computer science Physical sciences/Mathematics and computing/Information technology Imitation learning Locomotion Natural Behaviors Quadrupedal robots Full Text Additional Declarations There is NO Competing Interest. Supplementary Files Supplementary.pdf Supplementary Materials for Learning Diverse Natural Behaviors for Enhancing the Agility of Quadrupedal Robots SupplementaryVideo1quadrupedalagilitychallenge.mp4 Supplementary Video 1 Quadrupedal Agility Challenge SupplementaryVideo2boxjumping.mp4 Supplementary Video 2 Box Jumping SupplementaryVideo3hurdling.mp4 Supplementary Video 3 Hurdling SupplementaryVideo4comparisonwithbaselines.mp4 Supplementary Video 4 Comparison With Baselines SupplementaryVideo5naturalbehaviorsinthewild.mp4 Supplementary Video 5 Natural Behaviors in the Wild Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6354356","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":453164038,"identity":"a556da17-b9e9-4bd0-8cf1-d24b2e2e4053","order_by":0,"name":"Chunlin Chen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAvklEQVRIiWNgGAWjYDACdgY2IGnDwNgApHiI0sIM1pJGupbDEA5RWvibmZ89+LjjfB7zjATGB2/bGOTNCWmROMxmbjjzzO1ixhkJzIZz2xgMdzYQ0GLAzMMmzdt2O7FxRgKIwZBgcIAYLX/bzoG0sP8mXgtj2wGwLcxEaQH6xUyyty05sbHnYbPknHMShhsIaeFvb34m8bPNLnFje/LBD2/KbOQJ2gIHhg3gyJQgVj0QyJOgdhSMglEwCkYYAACjWjob1ERRnwAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-3929-4707","institution":"Nanjing University","correspondingAuthor":true,"prefix":"","firstName":"Chunlin","middleName":"","lastName":"Chen","suffix":""},{"id":453164039,"identity":"bd1d37ae-68b5-4f8c-9f4f-f075ba872732","order_by":1,"name":"Huiqiao Fu","email":"","orcid":"https://orcid.org/0000-0001-9403-2449","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Huiqiao","middleName":"","lastName":"Fu","suffix":""},{"id":453164040,"identity":"d055928e-97e4-4603-9b09-41e877f0a43f","order_by":2,"name":"Haoyu Dong","email":"","orcid":"https://orcid.org/0009-0006-2197-2441","institution":"Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Haoyu","middleName":"","lastName":"Dong","suffix":""},{"id":453164041,"identity":"04695f65-a9a9-4150-84e8-701e30c6d7bb","order_by":3,"name":"Wentao Xu","email":"","orcid":"https://orcid.org/0009-0000-4739-8595","institution":"Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Wentao","middleName":"","lastName":"Xu","suffix":""},{"id":453164042,"identity":"33c37b80-b9d4-43db-baaa-dbb3b8fcba4e","order_by":4,"name":"Zhehao Zhou","email":"","orcid":"https://orcid.org/0009-0008-2788-5589","institution":"Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Zhehao","middleName":"","lastName":"Zhou","suffix":""},{"id":453164043,"identity":"1e827f1f-13c1-4205-9777-b3b2b4d836fe","order_by":5,"name":"Guizhou Deng","email":"","orcid":"https://orcid.org/0000-0001-6589-4402","institution":"Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Guizhou","middleName":"","lastName":"Deng","suffix":""},{"id":453164044,"identity":"83ffea9a-7918-44ee-a739-7e1f7d0aec82","order_by":6,"name":"Kaiqiang Tang","email":"","orcid":"https://orcid.org/0000-0002-7456-0962","institution":"Nanjing University","correspondingAuthor":false,"prefix":"","firstName":"Kaiqiang","middleName":"","lastName":"Tang","suffix":""},{"id":453164045,"identity":"3519ad13-4219-491f-85c4-77aa104cc74d","order_by":7,"name":"Daoyi Dong","email":"","orcid":"https://orcid.org/0000-0002-7425-3559","institution":"University of Technology Sydney","correspondingAuthor":false,"prefix":"","firstName":"Daoyi","middleName":"","lastName":"Dong","suffix":""}],"badges":[],"createdAt":"2025-04-01 15:01:56","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6354356/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6354356/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82762267,"identity":"fe9d93dd-ec68-4ba2-81b5-535802a9cc7f","added_by":"auto","created_at":"2025-05-15 03:29:28","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3213933,"visible":true,"origin":"","legend":"Article File","description":"","filename":"Manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6354356/v1_covered_36c196e1-756f-4a74-ad86-933b467a0a8f.pdf"},{"id":82761768,"identity":"e6aa5cfd-cf23-4ac7-a212-c072bb68b658","added_by":"auto","created_at":"2025-05-15 03:21:23","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":7609268,"visible":true,"origin":"","legend":"Supplementary Materials for Learning Diverse Natural Behaviors for Enhancing the Agility of Quadrupedal Robots","description":"","filename":"Supplementary.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6354356/v1/19751f38ac68afe338b7d4cd.pdf"},{"id":82761775,"identity":"9ed89aa8-432d-4e38-b715-72b015812ee6","added_by":"auto","created_at":"2025-05-15 03:21:24","extension":"mp4","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":21178480,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Video 1 Quadrupedal Agility Challenge\u003c/p\u003e","description":"","filename":"SupplementaryVideo1quadrupedalagilitychallenge.mp4","url":"https://assets-eu.researchsquare.com/files/rs-6354356/v1/fdf093390e6aafc6f2971fd8.mp4"},{"id":82761772,"identity":"d8f7c104-1560-48b0-8472-ea9a8f4a233c","added_by":"auto","created_at":"2025-05-15 03:21:23","extension":"mp4","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":5266532,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Video 2 Box Jumping\u003c/p\u003e","description":"","filename":"SupplementaryVideo2boxjumping.mp4","url":"https://assets-eu.researchsquare.com/files/rs-6354356/v1/29816925b37867be618e1fff.mp4"},{"id":82761773,"identity":"479ad900-906a-45fc-b9ba-3be3c923bb51","added_by":"auto","created_at":"2025-05-15 03:21:23","extension":"mp4","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":12822091,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Video 3 Hurdling\u003c/p\u003e","description":"","filename":"SupplementaryVideo3hurdling.mp4","url":"https://assets-eu.researchsquare.com/files/rs-6354356/v1/8dc91cb79d7a26c0988b5a70.mp4"},{"id":82761771,"identity":"3c224fb0-f851-408a-a8b0-13f2b52e9925","added_by":"auto","created_at":"2025-05-15 03:21:23","extension":"mp4","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":19817599,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Video 4 Comparison With Baselines\u003c/p\u003e","description":"","filename":"SupplementaryVideo4comparisonwithbaselines.mp4","url":"https://assets-eu.researchsquare.com/files/rs-6354356/v1/4a0468323d81dfe655efed36.mp4"},{"id":82761774,"identity":"a0818be5-f32b-493d-b0e0-c7a0287e4432","added_by":"auto","created_at":"2025-05-15 03:21:24","extension":"mp4","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":11309125,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Video 5 Natural Behaviors in the Wild\u003c/p\u003e","description":"","filename":"SupplementaryVideo5naturalbehaviorsinthewild.mp4","url":"https://assets-eu.researchsquare.com/files/rs-6354356/v1/5a6691879ce857bd200bd470.mp4"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Learning Diverse Natural Behaviors for Enhancing the Agility of Quadrupedal Robots","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Imitation learning, Locomotion, Natural Behaviors, Quadrupedal robots","lastPublishedDoi":"10.21203/rs.3.rs-6354356/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6354356/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Achieving animal-like agility is a longstanding goal in quadrupedal robotics. While recent studies have successfully demonstrated imitation of specific behaviors, enabling robots to replicate a broader range of natural behaviors in real-world environments remains an open challenge. Here we propose an integrated controller comprising a Basic Behavior Controller (BBC) and a Task-Specific Controller (TSC) which can effectively learn diverse natural quadrupedal behaviors in an enhanced simulator and efficiently transfer them to the real world. Specifically, the BBC is trained using a novel semi-supervised generative adversarial imitation learning algorithm to extract diverse behavioral styles from raw motion capture data of real dogs, enabling smooth behavior transitions by adjusting discrete and continuous latent variable inputs. The TSC, trained via privileged learning with depth images as input, coordinates the BBC to efficiently perform various tasks. Additionally, we employ evolutionary adversarial simulator identification to optimize the simulator, aligning it closely with reality. After training, the robot exhibits diverse natural behaviors, successfully completing the quadrupedal agility challenge at an average speed of 1.1 m/s and achieving a peak speed of 3.2 m/s during hurdling. This work represents a substantial step toward animal-like agility in quadrupedal robots, opening avenues for their deployment in increasingly complex real-world environments.","manuscriptTitle":"Learning Diverse Natural Behaviors for Enhancing the Agility of Quadrupedal Robots","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-15 03:21:18","doi":"10.21203/rs.3.rs-6354356/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"15911633-5b1d-4461-a691-900a3c4ad7a5","owner":[],"postedDate":"May 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":48190151,"name":"Physical sciences/Mathematics and computing/Computer science"},{"id":48190152,"name":"Physical sciences/Mathematics and computing/Information technology"}],"tags":[],"updatedAt":"2025-05-15T03:21:18+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-15 03:21:18","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6354356","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6354356","identity":"rs-6354356","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}