[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84517-en":3,"doc-seo-84517-105":29,"detail-sidebar-cat-0-en-105":91},{"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":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},84517,549758146520,"Patrick","https://ap-avatar.wpscdn.com/avatar/80002397d8c0411e94?_k=1775819394049821470",8,"Research & Report","Distributionally Robust Control via Stein Variational Inference for Contact-Rich Manipulation","Reliable robotic manipulation depends on control policies that represent and adapt to uncertainty generated by contact-rich interactions. Data-driven methods can reduce uncertainty using large training sets and computation, but performance drops sharply when training samples are limited. Model-based controllers are computationally efficient yet may not capture task-relevant uncertainty well. This work formulates manipulation as distributionally robust control and introduces a deterministic Stein variational inference approach that explicitly models sensitive parameter uncertainty, improving robustness by up to 3× without sacrificing performance.","Robotics: Science and Systems 2026  \nSydney, Australia, July 13-July 17, 2026  \nDistributionally Robust Control via Stein Variational Inference for Contact-Rich Manipulation  \nHrishikesh Sathyanarayan∗ , Victor Vantilborgh†, Harish Ravichandar‡, Tom Lefebvre†, and Ian Abraham∗§  \n∗ Yale University, † Ghent University, ‡Georgia Tech, § University of Sydney  \narXiv :2605 . 19029v2 [ cs .RO] 13 Jul 2026  \nAbstract—Reliable robotic manipulation requires control policies that can accurately represent and adapt to uncertainty arising from contact-rich interactions. Modern data-driven methods mitigate uncertainty through large-scale training and computation, and degrade significantly in performance with limited number of training samples. By contrast, classical model-based controllers are computationally efficient and reliable, but their limited ability to represent task-relevant uncertainty can hinder performance in contact-rich interactions.  \nIn this work, we propose to expand the capabilities of modelbased manipulation control through more flexible uncertainty modeling that retains performance while exactly adapting to uncertainty. Our approach casts the manipulation problem asa distributionally robust control optimization and proposes a novel deterministic formulation based on Stein variational inference that preserves performance while explicitly modeling task-sensitive parameter uncertainty. As a result, the derived controllers are more aware of task sensitivities to uncertainty, yielding high reliability without compromising performance. Experimental results demonstrate up to 3× improved robustness across a range of contact-rich manipulation tasks under broad parametric uncertainty, outperforming existing modelbased control methods. Additional media and code is provided in [https://github.com/ialab-yale/stein-variational-dro.git](https://github.com/ialab-yale/stein-variational-dro.git)  \nI. INTRODUCTION  \nIn-the-wild manipulation often requires reasoning in an environment filled with uncertainty. However, the reliability and success of robotic control for manipulation is contingent on the ability to appropriately exploit and adapt to uncertainty through interacting with their environment. Modern control approaches mitigate uncertainty through vast accumulation of data and compute [9, 39, 5, 8, 6], effectively aiming to cover all possible scenarios at the expense of precision and interpretability, and their performance rapidly degrades with a reducing number of samples [22, 7] . The unifying goal is to develop zero-shot manipulation controllers that actively adapt to physical uncertainty, without the loss of performance and robustness.  \nMost prior work on uncertainty-aware manipulation control under parameter uncertainty spans data-driven methods and  \n∗ H.S and I.A are with the Department of Mechanical Engineering, Yale University, New Haven, CT, USA, email: [hrishi.sathyanarayan@yale.edu](hrishi.sathyanarayan@yale.edu)  \n†V.V and T.L are with the Department of Electromechanical, Systems and Metal Engineering, Ghent University, Belgium, email: { victor.vantilborgh, [tom.lefebvre](tom.lefebvre}@ugent.be)[}](tom.lefebvre}@ugent.be)[@ugent.be](tom.lefebvre}@ugent.be)  \n‡[H.R. is](H.R. is) with the School of Interactive Computing, Georgia Institute of Technology, Atlanta, GA, USA, email: [harish.ravichandar@cc.gatech.edu](harish.ravichandar@cc.gatech.edu)  \n§ I.A is with the Department of Electrical Engineering, University of Sydney, Australia, email: [ian.abraham@sydney.edu.au](ian.abraham@sydney.edu.au)  \nFig. 1: Within-hand dynamic positioning of a cup with unknown mass distribution and friction coefficient. Our proposed Stein Variational Distributionally Robust Optimizer (SV-DRO) applied toa within-hand positioning task via controlled sliding. The demonstration shown requires the robot to slide an object with unknown physical parameters (inertia, mass, and friction coefficient) to a goal state located at the center of the tr","cbCaik3WULtpaCk1","https://ap.wps.com/l/cbCaik3WULtpaCk1","pdf",40883235,1,15,"English","en",105,"# Introduction\n## Uncertainty in in-the-wild manipulation\n## Limits of data-driven and adaptive methods\n## Distributionally robust optimization background\n## Goal and proposed formulation","[{\"question\":\"Why do data-driven robotic manipulation methods degrade with fewer training samples?\",\"answer\":\"Their uncertainty mitigation relies on extensive data and computation to cover scenarios; when sample coverage decreases, performance drops rapidly.\"},{\"question\":\"What limitation of classical worst-case distributionally robust optimization affects contact-rich tasks?\",\"answer\":\"Worst-case uncertainty modeling can reduce task performance, producing worse overall behavior when object dynamics vary due to contact.\"},{\"question\":\"How does the proposed approach improve robustness without compromising performance?\",\"answer\":\"It formulates manipulation under parametric uncertainty as distributionally robust control and uses a deterministic Stein variational inference formulation to explicitly model task-sensitive parameter uncertainty, yielding up to 3× improved robustness across tasks.\"}]",1784196259,38,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":27},"distributionally-robust-control-via-stein-variational-inference-for-contact-rich-manipulation","",{"@graph":35,"@context":85},[36,53,68],{"@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/distributionally-robust-control-via-stein-variational-inference-for-contact-rich-manipulation/84517/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why do data-driven robotic manipulation methods degrade with fewer training samples?","Question",{"text":75,"@type":76},"Their uncertainty mitigation relies on extensive data and computation to cover scenarios; when sample coverage decreases, performance drops rapidly.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What limitation of classical worst-case distributionally robust optimization affects contact-rich tasks?",{"text":80,"@type":76},"Worst-case uncertainty modeling can reduce task performance, producing worse overall behavior when object dynamics vary due to contact.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the proposed approach improve robustness without compromising performance?",{"text":84,"@type":76},"It formulates manipulation under parametric uncertainty as distributionally robust control and uses a deterministic Stein variational inference formulation to explicitly model task-sensitive parameter uncertainty, yielding up to 3× improved robustness across tasks.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]