EdgeVolution: Democratizing Multi-Objective Neural Architecture Search and End-to-End Deployment on Microcontrollers

EdgeVolution: Democratizing Multi-Objective Neural Architecture Search and End-to-End Deployment on Microcontrollers | 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 EdgeVolution: Democratizing Multi-Objective Neural Architecture Search and End-to-End Deployment on Microcontrollers René Groh, Stefan Dendorfer, Mateo Avila Pava, Fabio Egle, Andreas Kist This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7032386/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 Edge AI holds great potential for extending the use of artificial neural networks to resource-constrained edge devices, such as microcontrollers. Despite this potential, optimizing and deploying neural networks on these platforms remains challenging due to a lack of tools for hardware-specific adaptation, leading to reproducibility issues and suboptimal performance. To address these challenges, we present EdgeVolution, an end-to-end hardware-in-the-loop platform that facilitates multi-objective optimization, neural architecture selection, and direct deployment onto target hardware. We demonstrate the versatility of EdgeVolution through four application use cases, showcasing its wide-ranging applicability. By offering a generic and adaptable pipeline, EdgeVolution enables the creation and deployment of neural network models tailored to specific datasets, classification tasks, and hardware constraints, thereby improving accessibility, performance, and reproducibility for AI applications on edge devices. Physical sciences/Mathematics and computing/Computer science Physical sciences/Engineering/Electrical and electronic engineering Neural Architecture Search Hardware-In-The-Loop Optimization Microcontroller Edge AI Full Text Additional Declarations There is NO Competing Interest. Supplementary Files EdgeVolutionsupplementaryinformation.pdf Appendix 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-7032386","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":489118602,"identity":"6202ad60-ae7a-4c80-9225-fada85aa9ad3","order_by":0,"name":"René Groh","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwElEQVRIiWNgGAWjYDADA/bGB0CKmRQtPIcNSNUikUykFvP23oMPGH7dkzOXfMz4mYfBWo6gFpkz55INGPuKjS1nJzNL8zCkGxPUIiGRYybB2JOQuOF2/gGglsOJDQS1yL8x/wHWcvMw82+glnrCWiR4zBgYfgC13GBmA9mSQNhhPDnGEokNCcYGZ5LZLOcYpBsStoX9jOGHD38S5AyOH2a+8abCWp6gLWCQ2AZjGRCnAQj+EK1yFIyCUTAKRiIAAP4jNbtwWY0UAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-3405-1311","institution":"Friedrich-Alexander-Universität Erlangen-Nürnberg","correspondingAuthor":true,"prefix":"","firstName":"René","middleName":"","lastName":"Groh","suffix":""},{"id":489118603,"identity":"4f081d09-c66b-4395-a929-e42e391774c1","order_by":1,"name":"Stefan Dendorfer","email":"","orcid":"","institution":"Friedrich-Alexander-Universität Erlangen-Nürnberg","correspondingAuthor":false,"prefix":"","firstName":"Stefan","middleName":"","lastName":"Dendorfer","suffix":""},{"id":489118604,"identity":"c27f2323-0f4a-4da4-9c8a-0c87a6f84710","order_by":2,"name":"Mateo Avila Pava","email":"","orcid":"","institution":"Friedrich-Alexander-Universität Erlangen-Nürnberg","correspondingAuthor":false,"prefix":"","firstName":"Mateo","middleName":"Avila","lastName":"Pava","suffix":""},{"id":489118605,"identity":"d2a91986-3abb-4250-a4e5-bfe0c2d51b66","order_by":3,"name":"Fabio Egle","email":"","orcid":"https://orcid.org/0000-0003-2591-3255","institution":"Friedrich-Alexander-Universität Erlangen-Nürnberg","correspondingAuthor":false,"prefix":"","firstName":"Fabio","middleName":"","lastName":"Egle","suffix":""},{"id":489118606,"identity":"9b82a78d-1451-4370-b8cd-c2a0daa2912d","order_by":4,"name":"Andreas Kist","email":"","orcid":"https://orcid.org/0000-0003-3643-7776","institution":"Friedrich-Alexander-Universität Erlangen-Nürnberg","correspondingAuthor":false,"prefix":"","firstName":"Andreas","middleName":"","lastName":"Kist","suffix":""}],"badges":[],"createdAt":"2025-07-02 20:10:32","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7032386/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7032386/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":94826610,"identity":"71326f95-8f73-499b-87c9-33e1ebe97fdd","added_by":"auto","created_at":"2025-10-31 06:53:35","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":24093350,"visible":true,"origin":"","legend":"Article File","description":"","filename":"EdgeVolutionmanuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7032386/v1_covered_af7d5bc0-35eb-4aee-a6be-92105c03a24b.pdf"},{"id":94819625,"identity":"98ca8381-79ff-4336-88c9-27afffc43816","added_by":"auto","created_at":"2025-10-31 05:46:32","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2374985,"visible":true,"origin":"","legend":"Appendix","description":"","filename":"EdgeVolutionsupplementaryinformation.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7032386/v1/5bd3c29291eef4d0c7f15309.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"EdgeVolution: Democratizing Multi-Objective Neural Architecture Search and End-to-End Deployment on Microcontrollers","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"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":"Neural Architecture Search, Hardware-In-The-Loop Optimization, Microcontroller, Edge AI","lastPublishedDoi":"10.21203/rs.3.rs-7032386/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7032386/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Edge AI holds great potential for extending the use of artificial neural networks to resource-constrained edge devices, such as microcontrollers. Despite this potential, optimizing and deploying neural networks on these platforms remains challenging due to a lack of tools for hardware-specific adaptation, leading to reproducibility issues and suboptimal performance. To address these challenges, we present EdgeVolution, an end-to-end hardware-in-the-loop platform that facilitates multi-objective optimization, neural architecture selection, and direct deployment onto target hardware. We demonstrate the versatility of EdgeVolution through four application use cases, showcasing its wide-ranging applicability. By offering a generic and adaptable pipeline, EdgeVolution enables the creation and deployment of neural network models tailored to specific datasets, classification tasks, and hardware constraints, thereby improving accessibility, performance, and reproducibility for AI applications on edge devices.","manuscriptTitle":"EdgeVolution: Democratizing Multi-Objective Neural Architecture Search and End-to-End Deployment on Microcontrollers","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-31 05:46:27","doi":"10.21203/rs.3.rs-7032386/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"communications-engineering","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"commseng","sideBox":"Learn more about [Communications Engineering](http://link.springer.com/journal/44172)","snPcode":"44172","submissionUrl":"https://mts-commseng.nature.com/cgi-bin/main.plex","title":"Communications Engineering","twitterHandle":"@commseng","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"68590df8-09e7-487a-a901-af3f2c7ef411","owner":[],"postedDate":"October 31st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":51926164,"name":"Physical sciences/Mathematics and computing/Computer science"},{"id":51926165,"name":"Physical sciences/Engineering/Electrical and electronic engineering"}],"tags":[],"updatedAt":"2025-12-04T22:40:38+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-31 05:46:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7032386","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7032386","identity":"rs-7032386","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}