[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86315-en":3,"doc-seo-86315-105":29,"detail-sidebar-cat-0-en-105":90},{"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":21,"is_downloadable":21,"audit_status":21,"page_count":11,"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},86315,13056703020460,"Valentina","https://ap-avatar.wpscdn.com/avatar/be000253dac470eee5d?_k=1778207105932848923",8,"Research & Report","AutoPath Learning Transferable Goal-Conditioned Stochastic Path Prior for Safe Navigation Without Human Demonstrations","Real-time navigation in cluttered, dynamic scenes demands collision-free, dynamically feasible motion under limited sensing, yet feasible behavior is inherently multimodal due to multiple obstacle-avoiding routes. AutoPath addresses this by learning a transferable goal-conditioned stochastic path prior: a reusable distribution over goal-aligned, geometry-consistent local paths conditioned on local observations. It supports structured sampling of diverse candidates without robot-specific motion constraints. A goal-aligned canonical state removes in-plane rotational ambiguity and normalizes geometry, enabling rotation-invariant learning, while a geometry-aware polar action manifold and risk-sensitive utility shaping with multi-goal rollouts improve safety-focused planning. Experiments validate high success rates and cross-platform transfer from differential-drive to quadruped platforms without retraining.","AutoPath: Learning Transferable Goal-Conditioned Stochastic Path Prior for Safe Navigation Without Human Demonstrations  \nZiyang Zhang 1 , Boyang Zhou 1 , Zesong Yang 1 , Haocheng Peng 1 , Zeming Gai2 , Xiao Liang 1 , Yujun Shen3 , Danping Zou4 , Ruizhen Hu5 , Hujun Bao 1 , and Zhaopeng Cui 1†  \narXiv :2607 . 11739v1 [ cs .RO] 13 Jul 2026  \nAbstract—Real-time navigation in cluttered and dynamic environments requires collision-free and dynamically feasible motion under limited perception. However, feasible navigation behaviors are inherently multimodal because multiple paths may exist around obstacles. In this paper, we formulate navigation as learning a transferable goal-conditioned stochastic path prior that models a reusable distribution over goal-aligned geometryconsistent local paths conditioned on local observations. This formulation enables structured sampling of navigation candidates, allowing multiple feasible paths to be explored through sampling without relying on robot-specific motion constraints. To this end, we introduce a goal-aligned canonical state representation that removes in-plane rotational ambiguity and normalizes local geometry with respect to the goal, enabling rotation-invariant path distribution learning. We further develop a structured prior learning framework that parameterizes local paths using a geometry-aware polar action manifold and incorporates risksensitive utility shaping with multi-goal distributional rollouts for stable and safety-aware planning. Extensive experiments in dense static environments and dynamic pedestrian scenarios demonstrate that the proposed method achieves consistently high success rates with competitive efficiency while enabling crossplatform transfer of a single path prior learned on differentialdrive robots to quadruped platforms without retraining.  \nIndex Terms—Motion and Path Planning, Integrated Planning and Learning, Collision Avoidance.  \nI. INTRODUCTION  \nREAL-TIME navigation requires generating collision-free  \nand dynamically feasible trajectories under limited perception and continuously changing environments. In cluttered scenes, feasible motion is inherently multimodal: multiple paths may exist around obstacles, and safe navigation depends on structured spatial reasoning rather than a single deterministic action. Classical model-based planners offer explicit constraint handling but rely on hand-crafted heuristics or computationally intensive search. Learning-based approaches improve adaptability by predicting controls or trajectories directly from observations [1]–[5]; however, many existing methods collapse this multimodality into deterministic outputs and tightly couple spatial reasoning with platform-specific motion constraints.  \nManuscript received: March 10, 2026; Accepted May 6, 2026 . This paper was recommended for publication by Editor Aniket Bera upon evaluation of the Associate Editor and Reviewers’ comments. This work was partially supported by the NSFC (No. 62572425) and Ant Group.  \n†Corresponding author: Zhaopeng Cui.  \n1Zhejiang University. {zhangtzeyang, byzhou, zesongyang0, haochengpeng, 22521125, baohujun, [zhpcui](zhpcui}@zju.edu.cn)[}](zhpcui}@zju.edu.cn)[@zju.edu.cn](zhpcui}@zju.edu.cn)  \n2 Harbin Institute [of Technology.](of Technology. 2023311f24@stu.hit.edu.cn)[ 2023311f24@stu.hit.edu.cn](of Technology. 2023311f24@stu.hit.edu.cn)  \n[3](3 Ant Group. shenyujun0302@gmail.com)[ Ant Group.](3 Ant Group. shenyujun0302@gmail.com)[ shenyujun0302@gmail.com](3 Ant Group. shenyujun0302@gmail.com)  \n4 Shanghai [Jiao Tong University.](Jiao Tong University. dpzou@sjtu.edu.cn)[ dpzou@sjtu.edu.cn](Jiao Tong University. dpzou@sjtu.edu.cn)  \n[5](5 Shenzhen University. ruizhen.hu@gmail.com)[ Shenzhen University.](5 Shenzhen University. ruizhen.hu@gmail.com)[ ruizhen.hu@gmail.com](5 Shenzhen University. ruizhen.hu@gmail.com)[ ](5 Shenzhen University. ruizhen.hu@gmail.com)Digital Object Identifier (DOI): see top of this page.  \nFig. 1. A transferable g","cbCaiogNejUkkNAH","https://ap.wps.com/l/cbCaiogNejUkkNAH","pdf",4905894,4,1,"English","en",105,"# Introduction\n## Problem: multimodal real-time navigation\n## Limitations of classical and existing learning methods\n# Proposed Approach\n## Transferable goal-conditioned stochastic path prior\n## Goal-aligned canonical state representation\n## Structured sampling and risk-sensitive planning","[{\"question\":\"What problem does AutoPath target in real-time navigation?\",\"answer\":\"AutoPath targets collision-free, dynamically feasible navigation in cluttered and dynamic environments under limited perception, where feasible motion is inherently multimodal.\"},{\"question\":\"How does AutoPath model navigation to handle multimodality?\",\"answer\":\"It learns a transferable goal-conditioned stochastic path prior, representing a reusable distribution over geometry-consistent local paths conditioned on local observations rather than predicting a single deterministic trajectory.\"},{\"question\":\"Why is cross-platform transfer possible without retraining?\",\"answer\":\"AutoPath introduces a goal-aligned canonical state representation that removes robot-specific kinematic dependence, enabling rotation-invariant distribution learning and transfer from differential-drive robots to quadruped platforms.\"}]",1784210443,20,{"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":85,"head_meta":87,"extra_data":89,"updated_unix":27},"autopath-learning-transferable-goal-conditioned-stochastic-path-prior-for-safe-navigation-without-human-demonstrations","",{"@graph":35,"@context":84},[36,52,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":21},"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":20},"https://docshare.wps.com/document/autopath-learning-transferable-goal-conditioned-stochastic-path-prior-for-safe-navigation-without-human-demonstrations/86315/",{"url":51,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":23,"description":14,"dateModified":60,"datePublished":61,"encodingFormat":59,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":40,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-27","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":20},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What problem does AutoPath target in real-time navigation?","Question",{"text":74,"@type":75},"AutoPath targets collision-free, dynamically feasible navigation in cluttered and dynamic environments under limited perception, where feasible motion is inherently multimodal.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does AutoPath model navigation to handle multimodality?",{"text":79,"@type":75},"It learns a transferable goal-conditioned stochastic path prior, representing a reusable distribution over geometry-consistent local paths conditioned on local observations rather than predicting a single deterministic trajectory.",{"name":81,"@type":72,"acceptedAnswer":82},"Why is cross-platform transfer possible without retraining?",{"text":83,"@type":75},"AutoPath introduces a goal-aligned canonical state representation that removes robot-specific kinematic dependence, enabling rotation-invariant distribution learning and transfer from differential-drive robots to quadruped platforms.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,126,129,133],{"id":21,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"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":28,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":28,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":28,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]