[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-141358-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-141358-en":131},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":74,"head_meta":76,"extra_data":78,"updated_unix":80},105,"en","whole-body-mobile-manipulation-using-offline-reinforcement-learning-on-sub-optimal-controllers","Whole-Body Mobile Manipulation using Offline Reinforcement Learning on Sub-optimal Controllers","","Mobile manipulation (MoMa) of articulated objects requires simultaneous whole-body coordination between a robot base and arms. Classical whole-body controllers can solve these tasks through hierarchical optimization, but often depend on extensive hand-tuning and can be brittle. Learning methods typically need costly whole-body teleoperation data or heavy reward engineering. The work shows that even a sub-optimal controller provides a strong structural prior for collecting task-relevant data and can be further improved with offline reinforcement learning. It introduces a two-stage pipeline that generates diverse demonstrations using a lightweight controller, then applies offline RL with action-chunked diffusion policies for robust transfer to real hardware.",{"@graph":14,"@context":73},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/whole-body-mobile-manipulation-using-offline-reinforcement-learning-on-sub-optimal-controllers/141358/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/whole-body-mobile-manipulation-using-offline-reinforcement-learning-on-sub-optimal-controllers/141358.png","ImageObject",300,407,{"name":42,"@type":43},"Kyle","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-30","2026-08-25",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",18,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"Why is whole-body coordination critical for mobile manipulation of articulated objects?","Question",{"text":63,"@type":64},"Tasks such as opening doors, drawers, and cupboards require simultaneous coordination of the robot base and arms, making whole-body synchronization necessary for success.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How does the method use sub-optimal whole-body controllers instead of teleoperation?",{"text":68,"@type":64},"A lightweight, randomized controller generates diverse demonstrations within a constrained, task-relevant region, and offline reinforcement learning improves those behaviors using reward signals.",{"name":70,"@type":61,"acceptedAnswer":71},"What is the key policy-learning mechanism introduced to handle complex coordination?",{"text":72,"@type":64},"The approach extends offline implicit Q-learning with Q-chunking so critics evaluate action chunks directly, supporting expressive action-chunked diffusion policies.","https://schema.org",{"og:url":32,"og:type":75,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":77,"canonical":32},"index,follow",{"doc_id":79,"site_id":7},141358,1787654538,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,106,111,115,120,123,127],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":25,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":102,"doc_module":4,"doc_module_name":25,"category_name":103,"show_sort_weight":104,"slug":105},6,"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":25,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":112,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":113,"slug":114},8,30,"research-report",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":25,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":25,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":25,"category_name":129,"show_sort_weight":97,"slug":130},19,"General","general",{"code":4,"msg":82,"data":132},{"doc_id":79,"user_id":133,"nickname":42,"user_avatar":134,"doc_module":4,"category_id":112,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":140,"language":141,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":12,"update_tm":80,"read_time":145,"preview_page_count":4},3985741905716,"https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d","arXiv :2604 . 12509v1 [ cs .RO] 14 Apr 2026  \nWhole-Body Mobile Manipulation using Offline Reinforcement Learning on Sub-optimal Controllers  \nSnehal Jauhri 1 , Vignesh Prasad 1 , and Georgia Chalvatzaki 1 ,2 ,3  \nFig. 1: WHOLE-MoMa policy on a real Tiago++ mobile manipulator simultaneously opening a cupboard and placing an object inside it. Videos at project  \nwebsite: [sites.google.com/view/whole-moma](sites.google.com/view/whole-moma).  \nAbstract—Mobile Manipulation (MoMa) of articulated objects, such as opening doors, drawers, and cupboards, demands simultaneous, whole-body coordination between a robot’s base and arms. Classical whole-body controllers (WBCs) can solve such problems via hierarchical optimization, but require extensive hand-tuned optimization and remain brittle. Learning-based methods, on the other hand, show strong generalization capabilities but typically rely on expensive whole-body teleoperation data or heavy reward engineering. We observe that even a sub-optimal WBC is a powerful structural prior: it can be used to collect data in a constrained, task-relevant region of the state-action space, and its behavior can still be improved upon using offline reinforcement learning. Building on this, we propose WHOLE-MoMa, a two-stage pipeline that first generates diverse demonstrations by randomizing a lightweight WBC, and then applies offline RL to identify and stitch together improved behaviors via a reward signal. To support the expressive actionchunked diffusion policies needed for complex coordination tasks, we extend offline implicit Q-learning with Q-chunking for chunklevel critic evaluation and advantage-weighted policy extraction. On three tasks of increasing difficulty using a TIAGo++ mobile manipulator in simulation, WHOLE-MoMa significantly outperforms WBC, behavior cloning, and several offline RL baselines. Policies transfer directly to the real robot without finetuning, achieving 80% success in bimanual drawer manipulation and 68% in simultaneous cupboard opening and object placement, all without any teleoperated or real-world training data.  \nIndex Terms—Mobile manipulation, whole-body control, offline reinforcement learning  \nI. INTRODUCTION  \nMobile Manipulation (MoMa) robots are central to the vision of general-purpose home assistants, given their en  \nThis research is funded by the European Union’s Horizon program under grant agreement no. 101120823, project MANiBOT and the German Research Foundation (DFG) Emmy Noether Programme (CH 2676/1-1) . Support and HPC resources provided by Erlangen National High Performance Computing Center (NHR) of Friedrich-Alexander-Universitt Erlangen-N¨urnberg (FAU), funded by federal and Bavarian authorities and the German Research Foundation (DFG)– 440719683.  \n1PEARL Lab, Computer Science Department, Technische Universitt Darmstadt, Germany 2Hessian.AI, Darmstadt, Germany 3Robotics Institute, Germany. Correspondence: snehal@robot-learning .de  \nhanced workspace and ability to perform everyday tasks in large-scale environments [1]–[3] . A key challenge of MoMa systems is the need to effectively coordinate the different embodiments of the robot, namely the manipulator arms and the movable wheeled base or legs. While several methods have been proposed to solve mobile “pick & place” style tasks [4]–[7], there remains a significant gap in robot capabilities for more challenging tasks that require simultaneous coordination of the base and arms, such as manipulating articulated objects (doors, drawers, cupboards) . Humans perform such wholebody coordination naturally, yet few existing robot learning methods specifically target this problem [8]–[10] . An often overlooked aspect is performing these tasks in a coordinated, time-efficient manner, e.g., opening a cupboard door while simultaneously placing an object inside it (Figure 1) . Our work focuses specifically on this under-explored setting of whole-body mobile manipulation for articulated object tasks requiring si","cbCaiqVrqrNAOLKa","https://ap.wps.com/l/cbCaiqVrqrNAOLKa","pdf",20411576,13,"English","# Abstract\n# I. Introduction\n## Background and challenge in whole-body mobile manipulation\n## Limitations of existing learning and controller methods\n# Contributions and approach overview","[{\"question\":\"Why is whole-body coordination critical for mobile manipulation of articulated objects?\",\"answer\":\"Tasks such as opening doors, drawers, and cupboards require simultaneous coordination of the robot base and arms, making whole-body synchronization necessary for success.\"},{\"question\":\"How does the method use sub-optimal whole-body controllers instead of teleoperation?\",\"answer\":\"A lightweight, randomized controller generates diverse demonstrations within a constrained, task-relevant region, and offline reinforcement learning improves those behaviors using reward signals.\"},{\"question\":\"What is the key policy-learning mechanism introduced to handle complex coordination?\",\"answer\":\"The approach extends offline implicit Q-learning with Q-chunking so critics evaluate action chunks directly, supporting expressive action-chunked diffusion policies.\"}]","Whole-Body Mobile Manipulation using Offline Reinforcement Learning on Sub-optimal Controllers | PDF",33]