[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84868-en":3,"doc-seo-84868-105":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":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"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},84868,8796095461564,"Liam","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","PCBWorld: A Benchmark Environment for Engine-Grounded PCB Design Automation","PCB routing requires connecting board nets with copper traces while satisfying strict design rules, yet learning-based approaches still trail rule-based routers. PCBWorld is an open-source engine-grounded routing environment built on the KiCad EDA engine, enabling interactive agent routing using native operations and Design Rule Check (DRC) feedback. It supports RL policies and tool-using LLM agents, and provides PCBWorld-Bench with synthetic and 679 real KiCad-format boards. Evaluation uses eight engine-checked metrics and shows strong performance and zero-shot transfer.","PCBWorld: A Benchmark Environment for Engine-Grounded PCB Design Automation  \nHyungseok Song∗ LG AI Research Seoul, Republic of Korea [hyungseok.song@lgresearch.ai](hyungseok.song@lgresearch.ai)  \nJunseok Park∗ LG AI Research Seoul, Republic of Korea [frank.park227@lgresearch.ai](frank.park227@lgresearch.ai)  \nWon-Seok Choi∗  \nLG AI Research Seoul, Republic of Korea [wonseok.choi@lgresearch.ai](wonseok.choi@lgresearch.ai)  \nSeohui Bae  \nLG AI Research Seoul, Republic of Korea [seohui.bae@lgresearch.ai](seohui.bae@lgresearch.ai)  \nHan-Seul Jeong  \nLG AI Research Seoul, Republic of Korea [hanseul.jeong@lgresearch.ai](hanseul.jeong@lgresearch.ai)  \nYoungjoon Park† LG AI Research Seoul, Republic of Korea [yj.park@lgresearch.ai](yj.park@lgresearch.ai)  \nSoonyoung Lee† LG AI Research Seoul, Republic of Korea [soonyoung.lee@lgresearch.ai](soonyoung.lee@lgresearch.ai)  \narXiv :2607 .059 15v 1 [ cs .AI ] 7 Jul 2026  \nAbstract  \nPCB routing is the task of connecting the nets of a board with copper traces under strict design rules, yet learning-based methods still lag behind rule-based routers. We introduce PCBWorld, an open-source engine-grounded PCB routing environment built on the KiCad EDA engine. As a human engineer does, agents in PCBWorld interactively route a board through the engine’s native operations, using its Design Rule Check (DRC) feedback to keep the routing within the design rules. The environment supports both RL policies and tool-using LLM agents. Alongside the environment, PCBWorld-Bench provides three dataset families in KiCad’s native board format ( .kicad_pcb), covering two types of controllable synthetic instances and 679 real open-source boards. It scores any completed board with eight engine-checked evaluation metrics, regardless of the routing method. In our experiments, agents in PCBWorld consistently outperformed grid-action RL policies and open-loop LLM baselines, and an RL policy trained only on synthetic boards transferred zero-shot to real boards, approaching rule-based routers. These results position the engine-grounded, interactive approach of PCBWorld as a promising foundation for advancing the routing ability of both RL and LLM agents.  \nKeywords  \nAgentic AI, PCB Routing Benchmark, Reinforcement Learning  \n1 Introduction  \nPrinted circuit boards (PCBs) are the physical boards that mount and interconnect electronic components, forming the backbone of nearly every electronic product [11, 27, 29] . Turning a circuit design into a manufacturable board centers on routing, the task of drawing copper traces that connect the pads of each net while keeping distinct nets electrically isolated under strict design rules [61] . As modern boards grow denser, routing every net without conflicts  \n∗ Equal contribution.†Corresponding author.  \namong tightly packed traces becomes increasingly challenging. Yet industry still relies mainly on rule-based routers rooted in decadesold heuristics [15, 35, 44] . They reliably route routine portions of a board, but rarely complete diverse, complex production boardsend-to-end. Learning-based methods have been studied extensively to overcome this limit, but unlike in language, vision, and games, they are not yet competitive even with rule-based routers [5, 47] .  \nWe attribute this limitation to how the routing problem is modeled. Existing RL formulations cast routing as cell-by-cell movement on a discretized grid, whose search space grows rapidly with grid size [5, 26, 47], or delegate routing to an auto-router and restrict the agent to auxiliary decisions such as net ordering [40, 73, 74] . More recent LLM-based methods generate scripts or complete board files in a single open-loop pass [4, 66, 75]; without engine feedback, they struggle to satisfy the strict design rules. The only workflow that reliably completes production boards end-to-end remains that of a human engineer. An engineer neither leaves the whole board to an auto-router nor draws every trace entirely by hand. ","cbCaii58gY60VwvR","https://ap.wps.com/l/cbCaii58gY60VwvR","pdf",4092464,1,42,"English","en",105,"# Abstract\n# Introduction\n## Background: Why routing is hard\n## Limitations of existing RL and LLM methods\n## Human-in-the-loop engine-grounded workflow\n# PCBWorld Overview\n## Engine-grounded interactive routing loop\n## API exposure and gym support\n# PCBWorld-Bench and Evaluation\n## Dataset families and scoring metrics","[{\"question\":\"What is PCBWorld, and what problem does it address?\",\"answer\":\"PCBWorld is an open-source engine-grounded PCB routing environment that helps agents complete PCB routing under strict design rules using native KiCad engine operations and DRC feedback. It targets the gap where learning methods lag behind rule-based routers.\"},{\"question\":\"How do agents interact with PCBWorld during routing?\",\"answer\":\"Agents route step-by-step by invoking KiCad EDA engine-native routing operations. At each step they receive the updated board state, admissible actions, and DRC feedback to ensure compliance with design rules.\"},{\"question\":\"What does PCBWorld-Bench provide for evaluation?\",\"answer\":\"PCBWorld-Bench provides three dataset families in KiCad’s native .kicad_pcb format, including two controllable synthetic instance types and 679 real open-source boards. It scores completed boards using eight engine-checked evaluation metrics regardless of routing method.\"}]",1784198917,106,{"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},"pcbworld-a-benchmark-environment-for-engine-grounded-pcb-design-automation","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"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":52},"https://docshare.wps.com/document/pcbworld-a-benchmark-environment-for-engine-grounded-pcb-design-automation/84868/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What is PCBWorld, and what problem does it address?","Question",{"text":74,"@type":75},"PCBWorld is an open-source engine-grounded PCB routing environment that helps agents complete PCB routing under strict design rules using native KiCad engine operations and DRC feedback. It targets the gap where learning methods lag behind rule-based routers.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How do agents interact with PCBWorld during routing?",{"text":79,"@type":75},"Agents route step-by-step by invoking KiCad EDA engine-native routing operations. At each step they receive the updated board state, admissible actions, and DRC feedback to ensure compliance with design rules.",{"name":81,"@type":72,"acceptedAnswer":82},"What does PCBWorld-Bench provide for evaluation?",{"text":83,"@type":75},"PCBWorld-Bench provides three dataset families in KiCad’s native .kicad_pcb format, including two controllable synthetic instance types and 679 real open-source boards. It scores completed boards using eight engine-checked evaluation metrics regardless of routing method.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,127,130,134],{"id":20,"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":52,"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":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":105,"slug":137},19,"General","general"]