[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85504-en":3,"doc-seo-85504-105":28,"detail-sidebar-cat-0-en-105":89},{"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":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":11,"language":21,"language_code":22,"site_id":23,"html_lang":22,"table_of_contents":24,"faqs":25,"seo_title":13,"seo_description":14,"update_tm":26,"read_time":27},85504,34359740700684,"Finn","https://ap-avatar.wpscdn.com/avatar/1f400023980c374ae676?_k=1777273430885731487",8,"Research & Report","PUMA Perception driven Unified Foothold Prior for Mobility Augmented Quadruped Parkour","Quadruped parkour for agile locomotion is limited by robots’ difficulty in perceiving terrain and selecting adaptive footholds during obstacle traversal. Existing hierarchical approaches depend on precomputed footholds, reducing real-time adaptability and constraining reinforcement learning exploration. PUMA introduces an end-to-end single-stage learning framework that unifies visual perception with foothold priors. Egocentric polar foothold priors guide velocity tracking, while Probability Annealing Selection transitions from ground-truth to predicted priors for stable convergence.","PUMA: Perception-driven Unified Foothold Prior for Mobility  \nAugmented Quadruped Parkour  \narXiv :2601 . 15995v2 [ cs .RO] 13 Jul 2026  \nLiang Wang 1 ,2 , Kanzhong Yao2 , Yang Liu2 , Weikai Qin2 , Jun Wu 1 , Zhe Sun∗2 and Qiuguo Zhu∗1  \nFig. 1: PUMA enables quadruped robots to fuse proprioception with visual perception to estimate adaptive footholds for traversing complex discrete terrains. Top Row: The robot twists its posture to forcefully kick off the inclined wall, propelling itself across a wide gap. Middle Row: The robot sequentially traverses uneven stepping stones. Bottom Row: The robot leverages an inclined wall to surmount a high platform.  \nAbstract—Parkour tasks for quadrupeds have emerged as a promising benchmark for agile locomotion. While human athletes can effectively perceive environmental characteristics to select appropriate footholds for obstacle traversal, endowing legged robots with similar perceptual reasoning remains a significant challenge. Existing methods often rely on hierarchical controllers that follow pre-computed footholds, thereby constraining the robot’s real-time adaptability and the exploratory potential of reinforcement learning. To overcome these challenges, we present PUMA, an end-to-end learning framework that integrates visual perception and foothold priors into a single-stage training process. This approach leverages terrain features to estimate  \nManuscript received: January 22, 2026; Revised: May 7, 2026; Accepted: June 25, 2026 .  \nThis paper was recommended for publication by Editor Olivier Stasse upon evaluation of the Associate Editor and Reviewers comments. This work was supported by the “Leading Goose” R&D Program of Zhejiang (Grant No. 2023C01177), the National Key R&D Program of China (Grant No. 2022YFB4701502), and the 2035 Key Technological Innovation Program of Ningbo City (Grant No. 2024Z300) .  \n1The authors are with Institute of Cyber-Systems and Control, Zhejiang University, 310027, China [3210102182@zju.edu.cn](3210102182@zju.edu.cn).  \n2Institute of Artificial Intelligence (TeleAI), China Telecom.  \n∗ Corresponding authors: Qiuguo Zhu ([qgzhu@zju.edu.cn](qgzhu@zju.edu.cn)) and Zhe Sun ( [sunzhe@nwpu.edu.cn](sunzhe@nwpu.edu.cn)) .  \nProject website: [https://puma-parkour.github.io/](https://puma-parkour.github.io/) .  \nDigital Object Identifier (DOI): see top of this page.  \negocentric polar foothold priors, composed of relative distance and heading, guiding the robot in active posture adaptation for parkour tasks. Extensive experiments conducted in simulation and real-world environments across various discrete complex terrains demonstrate PUMA’s exceptional agility and robustness in challenging scenarios.  \nIndex Terms: foothold prior, visual perception, reinforcement learning.  \nI. INTRODUCTION  \nIn recent years, quadruped robots have demonstrated remarkable athletic performance: a low-cost quadruped robot can leap over long gaps, climb over high obstacles, and traverse complex discrete terrain composed of stepping stones and sloped platforms [1]–[6] . To navigate complex terrains in parkour tasks, many studies leverage exteroceptive sensors (e.g., cameras, LiDAR) to acquire terrain information. The sensory data is typically fused with proprioception to train dynamic locomotion policies via learning-based methods. Approaches include both decoupled hierarchical frameworks separating control and perception modules [7], [8], and end-to-end visualaided policies [9], [10], establishing a strong foundation for agile legged locomotion.  \nHowever, even with the integration of exteroceptive sensing, robots still struggle to fully comprehend or exploit terrain features. Human parkour athletes can leverage environmental features to extend locomotive potential beyond inherent physical limits, exemplified by kicking off a wall to gain extra height and reach otherwise inaccessible elevations. This ability to strategically leverage terrain features for accomplishing locomotion ta","cbCaic5LyUQdIP14","https://ap.wps.com/l/cbCaic5LyUQdIP14","pdf",8600458,1,"English","en",105,"# Introduction\n## Motivation and challenges\n## Prior work and limitations\n## PUMA framework and approach","[{\"question\":\"What problem does PUMA address in quadruped parkour locomotion?\",\"answer\":\"PUMA targets the challenge of enabling legged robots to perceive terrain characteristics and select adaptive footholds in dynamic parkour tasks, where onboard perception only covers part of the environment.\"},{\"question\":\"How does PUMA differ from hierarchical foothold-tracking methods?\",\"answer\":\"Unlike methods that enforce direct tracking of planned foothold targets, PUMA uses egocentric polar footholds as motion priors for velocity tracking and feeds estimated polar footholds directly into the actor network.\"},{\"question\":\"How does PUMA stabilize training in its unified single-stage framework?\",\"answer\":\"PUMA uses Probability Annealing Selection (PAS) to gradually transition from ground-truth footholds to predicted footholds during training, improving stable convergence.\"}]",1784204057,20,{"code":4,"msg":29,"data":30},"ok",{"site_id":23,"language":22,"slug":31,"title":13,"keywords":32,"description":14,"schema_data":33,"social_meta":84,"head_meta":86,"extra_data":88,"updated_unix":26},"puma-perception-driven-unified-foothold-prior-for-mobility-augmented-quadruped-parkour","",{"@graph":34,"@context":83},[35,52,66],{"@type":36,"itemListElement":37},"BreadcrumbList",[38,42,46,49],{"item":39,"name":40,"@type":41,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":43,"name":44,"@type":41,"position":45},"https://docshare.wps.com/document/","Document",2,{"item":47,"name":12,"@type":41,"position":48},"https://docshare.wps.com/document/research-report/",3,{"item":50,"name":13,"@type":41,"position":51},"https://docshare.wps.com/document/puma-perception-driven-unified-foothold-prior-for-mobility-augmented-quadruped-parkour/85504/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":22,"description":14,"dateModified":60,"datePublished":60,"encodingFormat":59,"isAccessibleForFree":61,"interactionStatistic":62},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":39,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":63,"interactionType":64,"userInteractionCount":4},"InteractionCounter",{"@type":65},"ViewAction",{"@type":67,"mainEntity":68},"FAQPage",[69,75,79],{"name":70,"@type":71,"acceptedAnswer":72},"What problem does PUMA address in quadruped parkour locomotion?","Question",{"text":73,"@type":74},"PUMA targets the challenge of enabling legged robots to perceive terrain characteristics and select adaptive footholds in dynamic parkour tasks, where onboard perception only covers part of the environment.","Answer",{"name":76,"@type":71,"acceptedAnswer":77},"How does PUMA differ from hierarchical foothold-tracking methods?",{"text":78,"@type":74},"Unlike methods that enforce direct tracking of planned foothold targets, PUMA uses egocentric polar footholds as motion priors for velocity tracking and feeds estimated polar footholds directly into the actor network.",{"name":80,"@type":71,"acceptedAnswer":81},"How does PUMA stabilize training in its unified single-stage framework?",{"text":82,"@type":74},"PUMA uses Probability Annealing Selection (PAS) to gradually transition from ground-truth footholds to predicted footholds during training, improving stable convergence.","https://schema.org",{"og:url":50,"og:type":85,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":87,"canonical":50},"index,follow",{"doc_id":7,"site_id":23},{"code":4,"msg":5,"data":90},[91,95,99,103,108,113,118,121,125,128,132],{"id":20,"doc_module":4,"doc_module_name":44,"category_name":92,"show_sort_weight":93,"slug":94},"Story & Novel",90,"story-novel",{"id":45,"doc_module":4,"doc_module_name":44,"category_name":96,"show_sort_weight":97,"slug":98},"Literature",80,"literature",{"id":51,"doc_module":4,"doc_module_name":44,"category_name":100,"show_sort_weight":101,"slug":102},"Exam",70,"exam",{"id":104,"doc_module":4,"doc_module_name":44,"category_name":105,"show_sort_weight":106,"slug":107},5,"Comic",60,"comic",{"id":109,"doc_module":4,"doc_module_name":44,"category_name":110,"show_sort_weight":111,"slug":112},6,"Technology",50,"technology",{"id":114,"doc_module":4,"doc_module_name":44,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":44,"category_name":12,"show_sort_weight":119,"slug":120},30,"research-report",{"id":122,"doc_module":4,"doc_module_name":44,"category_name":123,"show_sort_weight":27,"slug":124},9,"Religion & Spirituality","religion-spirituality",{"id":27,"doc_module":4,"doc_module_name":44,"category_name":126,"show_sort_weight":27,"slug":127},"World Cup","world-cup",{"id":129,"doc_module":4,"doc_module_name":44,"category_name":130,"show_sort_weight":129,"slug":131},10,"Lifestyle","lifestyle",{"id":133,"doc_module":4,"doc_module_name":44,"category_name":134,"show_sort_weight":104,"slug":135},19,"General","general"]