[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82855-en":3,"doc-seo-82855-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},82855,2336464648322,"Aria","https://ap-avatar.wpscdn.com/avatar/2200025388227c56fec?_k=1778556882303663488",8,"Research & Report","SLAM Structured and Localized Analytic Manifold Adaptation for Lifelong VPR","Lifelong Visual Place Recognition (VPR) for long-term autonomous robotics must continually adapt to new environments while avoiding catastrophic forgetting. The paper presents SLAM, a Structured and Localized Analytic Manifold adaptation framework that unifies uncertainty-aware smoothing via the Unscented transformation, manifold partitioning using a Gaussian Mixture Model (GMM), and H∞ robust bound optimization into a closed-form analytic recursion. Ablations show uncertainty smoothing plus localized mapping reaches 27.5% nominal accuracy, while the H∞ bound provides a mathematically guaranteed minimax robustness trade-off via a single regularization parameter.","SLAM: Structured and Localized Analytic Manifold Adaptation for  \nLifelong VPR  \nKenta Tsukahara, Kanji Tanaka, and Rai Hisada  \narXiv :2607 .04764v 1 [ cs .RO] 6 Jul 2026  \nAbstract—Visual Place Recognition (VPR) in lifelong deployment requires continuous adaptation to new environments without catastrophic forgetting. In this paper, we propose SLAM, a Structured and Localized Analytic Manifold adaptation framework. Our framework elegantly unifies uncertainty-aware smoothing via Unscented transformation, topological space partitioning through a Gaussian Mixture Model (GMM), and H∞ robust bound optimization into a singular, unified closed-form analytical recursion. Exhaustive ablation studies demonstrate that while the synergistic combination of uncertainty smoothing and localized mapping (U+G configuration) achieves the stateof-the-art nominal accuracy of 27.5%, the full deployment of the H∞ bound does not require an architectural split; rather, it introduces a mathematically guaranteed minimax robust bound. This formulation enables the system to seamlessly modulate the intrinsic trade-off between nominal placement precision and worst-case disturbance attenuation through a single regularization parameter.  \nIndex Terms—Class-Incremental Learning, Lifelong Visual Place Recognition, Analytic Continual Learning, H∞ Filter, Unscented Transform, Gaussian Mixture Model.  \nI. INTRODUCTION  \nVisual Place Recognition (VPR) constitutes a core capabilities for long-term autonomous mobile robotics, enabling spatial localization across varying seasons, weather conditions, and operational spans. In open-world deployments, a robotic agent is continuously exposed to non-stationary environments where new geographic regions and landmark classes appear sequentially. To maintain an up-to-date representation of the environment without relying on prohibitive storage for historical raw data, systems must leverage Class-Incremental Continual Learning (CICL) . Under standard gradient descent optimization frameworks, however, neural network weights suffer drastically from catastrophic forgetting—a phenomenon where optimizing parameters for a newly introduced task fundamentally disrupts and overwrites the decision boundaries established for historical classes.  \nTo resolve catastrophic forgetting without the burden of maintaining experience replay buffers, Analytic Continual Incremental Learning (ACIL) has emerged as an elegant alternative. ACIL maps the latent representations extracted from a frozen deep backbone into an unconstrained feature space and recursively adjusts the final classification layer. By leveraging standard ridge regression and the Woodbury matrix inversion lemma, ACIL achieves closed-form parameter updates that mathematically eliminate forgetting. This approach ensures  \nAll authors are with the Department of Mechanical Engineering, Faculty of Engineering, University of Fukui, Fukui 910-8507, Japan. (E-mail:{tsukahara, tnkknj, [hisada](hisada}@u-fukui.ac.jp)[}](hisada}@u-fukui.ac.jp)[@u-fukui.ac.jp](hisada}@u-fukui.ac.jp)).  \nexact equivalence to batch training over all accumulated data up to the current timestamp.  \nDespite its mathematical elegance, traditional ACIL algorithms face critical limitations when deployed in real-world visual lifelong loops:  \n1) Decision Boundary Over-fitting: ACIL solves a deterministic least-squares objective that inherently seeks a sharp fit to the available samples. Consequently, small observational noises or outlier visual patches can skew the analytic boundaries, degrading generalization.  \n2) Static Manifold Assumption: Standard ACIL presumes that the projected feature distribution resides within a homogeneous, isotropic manifold. When a robot traverses spatially distinct zones under varying illumination or seasonal cycles, covariate shifts introduce severe geometric distortions that a single global analytic classifier cannot resolve.  \n3) Vulnerability to Unbounded Disturbance: In the presence ","cbCaiuwCwfBwZHbW","https://ap.wps.com/l/cbCaiuwCwfBwZHbW","pdf",257812,1,6,"English","en",105,"# Introduction\n# Proposed SLAM Framework\n## Topological GACIL with GMM Partitioning\n## Unscented Perturbed ACIL Smoothing\n## H∞ Robust Bound Optimization","[{\"question\":\"Why does lifelong VPR face catastrophic forgetting in continual deployment?\",\"answer\":\"Open-world robotic agents encounter non-stationary environments where new geographic regions and landmark classes appear sequentially. Standard gradient-based learning can disrupt previously learned decision boundaries when optimizing for new tasks, causing catastrophic forgetting.\"},{\"question\":\"How does ACIL avoid forgetting compared with standard neural network training?\",\"answer\":\"Analytic Continual Incremental Learning (ACIL) freezes a deep backbone and recursively updates the final classification layer using closed-form ridge regression. Using tools like the Woodbury matrix inversion lemma, it achieves equivalence to batch training over all accumulated data up to the current time.\"},{\"question\":\"What limitations of traditional ACIL does SLAM address?\",\"answer\":\"The paper identifies over-fitting from deterministic least-squares solutions, a static manifold assumption that fails under covariate shifts across zones, and vulnerability to accumulated estimation errors under persistent noise. SLAM extends the approach with localized manifold partitioning, uncertainty smoothing, and H∞ robust optimization.\"}]",1784183459,15,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"slam-structured-and-localized-analytic-manifold-adaptation-for-lifelong-vpr","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/slam-structured-and-localized-analytic-manifold-adaptation-for-lifelong-vpr/82855/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why does lifelong VPR face catastrophic forgetting in continual deployment?","Question",{"text":75,"@type":76},"Open-world robotic agents encounter non-stationary environments where new geographic regions and landmark classes appear sequentially. Standard gradient-based learning can disrupt previously learned decision boundaries when optimizing for new tasks, causing catastrophic forgetting.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does ACIL avoid forgetting compared with standard neural network training?",{"text":80,"@type":76},"Analytic Continual Incremental Learning (ACIL) freezes a deep backbone and recursively updates the final classification layer using closed-form ridge regression. Using tools like the Woodbury matrix inversion lemma, it achieves equivalence to batch training over all accumulated data up to the current time.",{"name":82,"@type":73,"acceptedAnswer":83},"What limitations of traditional ACIL does SLAM address?",{"text":84,"@type":76},"The paper identifies over-fitting from deterministic least-squares solutions, a static manifold assumption that fails under covariate shifts across zones, and vulnerability to accumulated estimation errors under persistent noise. SLAM extends the approach with localized manifold partitioning, uncertainty smoothing, and H∞ robust optimization.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]