[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-141361-105":59,"doc-detail-141361-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","learn2assemble-with-structured-representations-and-search-for-robotic-architectural-construction","Learn2Assemble with Structured Representations and Search for Robotic Architectural Construction","","Autonomous robotic assembly depends on coordinated high-level action planning and reliable manipulation execution. Learning to assemble complex 3D structures requires connecting target designs with available building blocks while producing valid assembly sequences that satisfy structural stability and feasibility constraints. To tackle the combinatorial explosion, the approach introduces a multi-head attention graph representation trained with reinforcement learning to encode spatial relations and propose meaningful assembly actions. Structured representations are combined with model-free RL and Monte-Carlo planning, supported by robot-in-the-loop geometric execution. A hierarchical control framework sequences modules to build arbitrary 3D designs and ensures feasibility.",{"@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/learn2assemble-with-structured-representations-and-search-for-robotic-architectural-construction/141361/",{"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/learn2assemble-with-structured-representations-and-search-for-robotic-architectural-construction/141361.png","ImageObject",300,407,{"name":92,"@type":93},"Fahsai","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-03","2026-08-25",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",17,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does the Learn2Assemble approach address?","Question",{"text":113,"@type":114},"It addresses autonomous assembly of complex 3D structures using robotic manipulators without an a priori task plan or fixed goal poses for sequential pick-and-place actions.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"How does the method represent and decide assembly actions?",{"text":118,"@type":114},"It uses a graph-based representation with a multi-head attention GNN, trained via reinforcement learning to encode spatial relations and explore feasible actions.",{"name":120,"@type":111,"acceptedAnswer":121},"How are feasibility and stability ensured during execution?",{"text":122,"@type":114},"A hierarchical framework combines learned long-horizon sequencing with Monte-Carlo planning, and uses robot-in-the-loop geometric motion generation to produce reachable, feasible actions that preserve assembly stability.","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},141361,1787654899,{"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":130,"read_time":145,"preview_page_count":4},549768702563,"https://ap-avatar.wpscdn.com/avatar/8000c4aa63b76e948b?x-image-process=image/resize,m_fixed,w_180,h_180&k=1786536092046926083","Learn2Assemble with Structured Representationsand Search for Robotic Architectural Construction  \nNiklas Funk, Georgia Chalvatzaki, Boris Belousov, and Jan Peters  \nDepartment of Computer Science, Technical University of Darmstadt, Germany fniklas,georgia,[boris](borisg@robot-learning.de)[g](borisg@robot-learning.de)[@robot-learning.de](borisg@robot-learning.de), [mail@jan-peters.net](mail@jan-peters.net)  \nAbstract: Autonomous robotic assembly requires a well-orchestrated sequence of high-level actions and smooth manipulation executions. Learning to assemble complex 3D structures remains a challenging problem that requires drawing connections between target designs and building blocks, and creating valid assembly sequences considering structural stability and feasibility. To address the combinatorial complexity of the assembly tasks, we propose a multi-head attention graph representation that can be trained with reinforcement learning (RL) to encode the spatial relations and provide meaningful assembly actions.  \nCombining structured representations with model-free RL and Monte-Carlo planning allows agents to operate with various target shapes and building block types. We design a hierarchical control framework that learns to sequence the building blocks to construct arbitrary 3D designs and ensures their feasibility, as we plan the geometric execution with the robot-in-the-loop. We demonstrate the flexibility of the proposed structured representation and our algorithmic solution in a series of simulated 3D assembly tasks with robotic evaluation, which showcases our method's ability to learn to construct stable structures with a large number of building blocks. Code and videos are available at: [https://sites.google.com/view/learn2assemble](https://sites.google.com/view/learn2assemble)  \nKeywords: Structured representations, Autonomous assembly, Manipulation  \n1 Introduction  \nConstruction and manufacturing are becoming increasingly automated in the last decades. However, there is an essential need for sustainable autonomous architectural assembly [1], where a gamechanger would come with intelligent robot assembly abilities that optimally decide over plans, actions, execution, and efficiency [2] . In this work, our main focus is the combinatorial optimization problem of autonomously assembling complex structures with robotic manipulators without an a priori defined task plan and goal poses for the sequential picking-placing actions. For constructing abstract target designs, we must consider the combinatorics of the growing action space w.r.t. the number of available modules and the size of the structure. Therefore, an effective representation of the assembly problem is essential. Moreover, the geometric execution of the picking and placing actions by the robot imposes constraints to the assembly sequence, as those actions are subject to the kinematic feasibility in the robot's workspace. Eventually, the problem of assembling structures lies in the area of long-horizon manipulation tasks, where most methods in the literature consider a known task plan, and focus on fine manipulability and structural stability, or learn action sequences from demonstrations [3, 4, 5, 6, 7, 8] .  \nIn this paper, we propose a novel algorithmic solution for robotic assembly that combines highlevel decision-making on the construction sequence with the geometric execution by the robot that should ensure feasibility and stability. We propose a graph-based representation that captures the relations between target shapes and available building blocks. Notably, we design a multi-head attention-based graph neural network (GNN) architecture with a purposefully induced inductive bias for encoding the structural representation of the assembly task. The GNN is trained through reinforcement learning (RL) to explore feasible actions, resulting in an expressive and flexible representation. When combined as prior to Monte Carlo Tree Search (MCTS), it extrapolates to o","cbCaifOaV6AquLZG","https://ap.wps.com/l/cbCaifOaV6AquLZG","pdf",1711383,11,"English","# Introduction\n## Problem setting and motivation\n## Proposed approach and contributions\n## Benchmarking and evaluation setup","[{\"question\":\"What problem does the Learn2Assemble approach address?\",\"answer\":\"It addresses autonomous assembly of complex 3D structures using robotic manipulators without an a priori task plan or fixed goal poses for sequential pick-and-place actions.\"},{\"question\":\"How does the method represent and decide assembly actions?\",\"answer\":\"It uses a graph-based representation with a multi-head attention GNN, trained via reinforcement learning to encode spatial relations and explore feasible actions.\"},{\"question\":\"How are feasibility and stability ensured during execution?\",\"answer\":\"A hierarchical framework combines learned long-horizon sequencing with Monte-Carlo planning, and uses robot-in-the-loop geometric motion generation to produce reachable, feasible actions that preserve assembly stability.\"}]","Learn2Assemble with Structured Representations and Search for Robotic Architectural Construction | PDF",28]