[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-327452-105":59,"doc-detail-327452-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","cofl-s-spatially-queryable-sector-flow-fields-for-local-language-conditioned-navigation","CoFL-S - Spatially Queryable Sector Flow Fields for Local Language-Conditioned Navigation","","Vision-Language Navigation increasingly emphasizes instruction reasoning, memory, global map construction, and instruction decomposition, while low-level action representation remains less studied. CoFL-S introduces a low-level vision-language-action framework that predicts a language-conditioned flow field over the robot’s local visible sector and generates continuous trajectories by rolling out the field. Training converts VLN-CE whole-episode supervision into frame-level local supervision with aligned sub-instructions and dense flow-field targets. Evaluation adds a continuous-time Habitat benchmark and uses a shared velocity-command controller for decomposition-independent closed-loop comparison across planner frequencies, yielding consistent gains over baselines and improved zero-shot real-world execution.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/cofl-s-spatially-queryable-sector-flow-fields-for-local-language-conditioned-navigation/327452/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/cofl-s-spatially-queryable-sector-flow-fields-for-local-language-conditioned-navigation/327452.png","ImageObject",300,407,{"name":92,"@type":93},"Ivy","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-07","2026-09-21",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",13,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does CoFL-S address in vision-language navigation?","Question",{"text":113,"@type":114},"CoFL-S targets the underexplored low-level action representation, aiming to determine how local language intent is converted into executable robot motion rather than only high-level instruction reasoning.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How does CoFL-S generate trajectories from language-conditioned predictions?",{"text":118,"@type":114},"CoFL-S predicts a language-conditioned sector flow field over the robot’s local visible area and produces continuous trajectories by rolling out the predicted field.",{"name":120,"@type":111,"acceptedAnswer":121},"How is the training supervision constructed for CoFL-S?",{"text":122,"@type":114},"Each VLN-CE episode supervision is converted into frame-level local supervision, using aligned sub-instructions and matched action, trajectory, and dense flow-field targets.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},327452,1790534749,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":145,"read_time":146,"preview_page_count":4},549758252649,"https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819","CoFL-S: Spatially Queryable Sector Flow Fields for Local Language-Conditioned Navigation  \nHaokun Liu, Zhaoqi Ma, Yicheng Chen, Wentao Zhang, Masaki Kitagawa, Zicen Xiong,  \nJinjie Li, Moju Zhao  \nDragon Lab, Department of Mechanical Engineering, The University of Tokyo  \narXiv :2607 .02222v 1 [ cs .RO] 2 Jul 2026  \n(a) CoFL-S framework (c) Habitat-based continuous-time benchmark  \nEgo-Centric Observation CoFL-S Sector Flow Field  \nRGB Depth  \nLow-Level Instruction  \n“Walk through the first door on the right.”  \nVisionlanguage encoder  \nCoFL-S decoder  \nTrajectory Rollout  \n\"(!!) !!\"\\#!! \"(!!)∆&  \n!!\"\\# = !! + $(!!)∆(  \n(b) Continuous-control supervision  \nVLN-CE episodes  \nRGB-D + instruction at each frame  \nInstructions  \nProcedural annotation  \nGoal grounding  \nFlow fields  \nReplay  \n+  \nFine-Grained R2R Landmark RxR  \nR2R val-unseen RxR val-unseen  \nRGB-D Observation Local sub-instruction  \nRGB-D + Local sub-instruction  \nContinuous control rollout  \nContinuous-time Habitat  \n60 Hz state / 30 Hz observation  \nFine-Grained R2R  \n+  \nLandmark RxR  \nVelocity Controller  \nBenchmark baselines and CoFL-S under the same protocol  \nAction token Action chunk CoFL-S (Flow field)  \nt t+1 t+2 t+3 … t t + H $\"!% '(  \nFine-grained Annotation  \nExpand  \n…  \n…  \n(d) Zero-shot real-world closed-loop execution  \n\n| \u003Cbr>| | \u003Cbr>\u003Cbr>go forward | | \u003Cbr>\u003Cbr>\u003Cbr>keep going straight pass the parking place | | \u003Cbr>\u003Cbr>turn left to enter the square | | \u003Cbr>|\n| --- | --- | --- | --- | --- | --- | --- | --- | --- |\n\nFigure 1: Overview of CoFL-S. CoFL-S learns an ego-centric sector flow field from RGB-D observations and local language instructions, and extracts continuous control through field rollout. The framework is trained with dense sector flow field supervision and evaluated on a Habitat-based benchmark built from R2R/RxR episodes and zero-shot real-world closed-loop execution.  \nAbstract: Vision-Language Navigation has increasingly emphasized high-level instruction reasoning, memory, global map construction, and instruction decomposition, while the low-level action representation remains comparatively underexplored. We propose CoFL-S, a low-level vision-language-action framework that predicts a language-conditioned flow field over the robot’s local visible sector and generates continuous trajectories by rolling out the predicted field. To train this low-level representation, we convert each VLN-CE episode, originally a whole-episode instruction paired with an action sequence, into frame-level local supervision with aligned sub-instructions and matched action, trajectory, and dense flow-field targets. For evaluation, we introduce a continuous-time Habitat benchmark that isolates low-level action interfaces from instruction decomposition and executes all methods through a shared velocity-command controller, enabling decomposition-independent closed-loop comparison across different planner frequencies rather than fixed discrete forward-and-turn transitions in VLNCE. Under matched encoders and training settings, CoFL-S consistently outperforms action-token and action-chunk baselines across planner frequencies in the continuous-time Habitat benchmark, and zero-shot real-world closed-loop deployment further shows its advantage over both baselines beyond simulation.  \nKeywords: Action Representation, Vision-Language Navigation, Flow Fields  \n1 Introduction  \nVision-Language Navigation (VLN) is an embodied AI problem that couples high-level semantic understanding, visual perception, and motion grounding: an agent must interpret natural-language instructions, perceive its surroundings, and convert the intended semantic progress into executable motion. Since benchmark settings such as R2R [1], RxR [2], and VLN-CE [3] established the task, recent progress has substantially strengthened the perception-and-reasoning side of this pipeline through semantic grounding, memory construction, map reconstruction, and long-horizon instruction decomposition [4, 5, 6,","cbCail2XpGvFXJym","https://ap.wps.com/l/cbCail2XpGvFXJym","pdf",11001540,27,"English","# Introduction\n## Problem motivation: low-level action interfaces in VLN\n## Existing executable interfaces and their limitations\n## Spatial guidance as a complementary view\n## Proposed approach: CoFL-S","[{\"question\":\"What problem does CoFL-S address in vision-language navigation?\",\"answer\":\"CoFL-S targets the underexplored low-level action representation, aiming to determine how local language intent is converted into executable robot motion rather than only high-level instruction reasoning.\"},{\"question\":\"How does CoFL-S generate trajectories from language-conditioned predictions?\",\"answer\":\"CoFL-S predicts a language-conditioned sector flow field over the robot’s local visible area and produces continuous trajectories by rolling out the predicted field.\"},{\"question\":\"How is the training supervision constructed for CoFL-S?\",\"answer\":\"Each VLN-CE episode supervision is converted into frame-level local supervision, using aligned sub-instructions and matched action, trajectory, and dense flow-field targets.\"}]","CoFL-S - Spatially Queryable Sector Flow Fields for Local Language-Conditioned Navigation | PDF",1789977232,68]