{"id":10334,"date":"2026-03-09T12:22:28","date_gmt":"2026-03-09T12:22:28","guid":{"rendered":"https:\/\/cyber-valley.de\/news\/from-perception-to-constrained-decisions\/"},"modified":"2026-04-20T08:42:35","modified_gmt":"2026-04-20T08:42:35","slug":"from-perception-to-constrained-decisions","status":"publish","type":"news","link":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/","title":{"rendered":"From perception to constrained decisions"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\"><br\/><strong>Georg Martius\u2019 project, \u201cEmbedding combinatorial reasoning into learning control policies\u201d addressed a practical gap in AI&#8217;s basic machinery: connecting data-driven neural networks with decision-making so they can adhere to clear structures, rules, and constraints. The project developed foundations for integrating combinatorial optimization into learning systems. It has potential applications wherever planning and resource allocation are central, such as in robotics and logistics. The project was funded by the Cyber\u00a0Valley Research Fund and was carried out between 2021 and 2024.    <\/strong><\/p>\n\n<p class=\"wp-block-paragraph\">Martius\u2019 project addressed a key limitation of today\u2019s machine-learning systems. While neural networks can learn powerful representations from data (for example, from images or sensor signals), many real-world decisions must also follow strict rules, such as budgets, capacities, safety limits, or \u201cmust\/ must not\u201d constraints. These rule-based decisions are often handled by specialized optimization software, but such tools are traditionally difficult to integrate into learning-based systems.  <\/p>\n\n<p class=\"wp-block-paragraph\">Martius\u2019 research created methods that make it easier to combine neural networks with combinatorial optimization components. This enables systems to learn from data while also drawing on proven optimization techniques to make decisions that respect constraints. Rather than replacing classical solvers, the project developed ways to use them as trainable building blocks inside larger learning pipelines.  <\/p>\n\n<p class=\"wp-block-paragraph\">The project also explored \u201cneuro-algorithmic\u201d approaches to decision-making and control, investigating how structured algorithms can improve generalization to new problems compared with purely data-driven approaches.<\/p>\n\n<p class=\"wp-block-paragraph\">Overall, the project provides a foundation for building AI systems that connect perception with constraint-aware decision modules. This is an important step toward robust solutions in areas such as robotics, planning, logistics, and resource allocation. <\/p>\n\n<h4 class=\"wp-block-heading\">This project produced the following peer-reviewed publications: <\/h4>\n\n<p class=\"wp-block-paragraph\">Anselm Paulus, Michal Rol\u00ednek, V\u00edt Musil, Brandon Amos, and Georg Martius. \u201cCombOptNet: Fit the Right NP-Hard Problem by Learning Integer Programming Constraints\u201d. In: Proceedings of the 38th International Conference on Machine Learning. Vol.    139. Proceedings of Machine Learning Research. PMLR, July 2021, pp. 8443\u20138453. url: <a href=\"https:\/\/proceedings.mlr.press\/v139\/paulus21a.html\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/proceedings.mlr.press\/v139\/paulus21a.html<\/a>.   <\/p>\n\n<p class=\"wp-block-paragraph\">Marin Vlastelica, Michal Rolinek, and Georg Martius. \u201cNeuro-algorithmic Policies Enable Fast Combinatorial Generalization\u201d. In: Proceedings of the 2021 International Conference on Machine Learning (ICML). July 2021. url: <a href=\"https:\/\/proceedings.mlr.press\/v139\/vlastelica21a.html\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/proceedings.mlr.press\/v139\/vlastelica21a.html<\/a>.     <\/p>\n\n<p class=\"wp-block-paragraph\">Marco Bagatella, Mirek Ol\u0161\u00e1k, Michal Rol\u00ednek, and Georg Martius. \u201cPlanning from Pixels in Environments with Combinatorially Hard Search Spaces\u201d. In: Advances in Information Processing Systems (NeurIPS 2021).   2021. url: <a href=\"https:\/\/openreview.net\/forum?id=XgGUUaKgips\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/openreview.net\/forum?id=XgGUUaKgips<\/a>. <\/p>\n\n<p class=\"wp-block-paragraph\">Subham Sahoo, Anselm Paulus, Marin Vlastelica, V\u00edt Musil, Volodymyr Kuleshov, and Georg Martius. \u201cBackpropagation through Combinatorial Algorithms: Identity with Projection Works\u201d. In: Proceedings of the Eleventh International Conference on Learning Representations. May 2023. url: <a href=\"https:\/\/openreview.net\/forum?id=JZMR727O29\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/openreview.net\/forum?id=JZMR727O29<\/a>.     <\/p>\n\n<p class=\"wp-block-paragraph\">A. Paulus, G. Martius, and V. Musil. \u201cLPGD: A General Framework for Backpropagation through Embedded Optimization Layers\u201d. In: Proceedings of the 41st International Conference on Machine Learning (ICML). Vol.    235. Proceedings of Machine Learning Research. PMLR, July 2024, pp. 39989\u201340014. url: <a href=\"https:\/\/proceedings.mlr.press\/v235\/paulus24a.html\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/proceedings.mlr.press\/v235\/paulus24a.html<\/a>.   <\/p>\n\n<p class=\"wp-block-paragraph\"><a><strong>Georg Martius is a Professor in the Department of Computer Science at the University of T\u00fcbingen.<\/strong><\/a><\/p>\n\n<h4 class=\"wp-block-heading\">\u00a0<\/h4>\n\n<h4 class=\"wp-block-heading\">About the Cyber\u00a0Valley Research Fund<\/h4>\n\n<p class=\"wp-block-paragraph\">The Cyber\u00a0Valley Research Fund was established to support Cyber\u00a0Valley research groups undertake basic research in the fields of artificial intelligence and robotics. The fund totaled five million euros, including contributions from six of Cyber\u00a0Valley\u2019s founding corporate partners: Amazon, BMW, Bosch, IAV, Mercedes-Benz, Porsche, and ZF. It supported 20 research projects, the first of which began in 2020, and the final of which will conclude in 2026.  <\/p>\n","protected":false},"featured_media":10335,"parent":0,"template":"","meta":{"_acf_changed":false},"cv-tag":[289,296,298,317],"location":[],"network-role":[413],"news-category":[],"user-group":[245,247,410,412],"class_list":["post-10334","news","type-news","status-publish","has-post-thumbnail","hentry","cv-tag-max-planck-institute-for-intelligent-systems","cv-tag-machine-learning","cv-tag-robotics","cv-tag-research-fund","network-role-partner","user-group-researcher","user-group-phd-student"],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.8 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>From perception to constrained decisions - Cyber Valley<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/\" \/>\n<meta property=\"og:locale\" content=\"de_DE\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"From perception to constrained decisions - Cyber Valley\" \/>\n<meta property=\"og:description\" content=\"Georg Martius\u2019 project, \u201cEmbedding combinatorial reasoning into learning control policies\u201d addressed a practical gap in AI&#8217;s basic machinery: connecting data-driven neural networks with decision-making so they can adhere to clear structures, rules, and constraints. The project developed foundations for integrating combinatorial optimization into learning systems. It has potential applications wherever planning and resource allocation are [&hellip;]\" \/>\n<meta property=\"og:url\" content=\"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/\" \/>\n<meta property=\"og:site_name\" content=\"Cyber Valley\" \/>\n<meta property=\"article:modified_time\" content=\"2026-04-20T08:42:35+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/cyber-valley.de\/wp-content\/uploads\/2026\/03\/thumb_ticker_lg_From_perception_to_constrained_decisions.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1000\" \/>\n\t<meta property=\"og:image:height\" content=\"523\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Gesch\u00e4tzte Lesezeit\" \/>\n\t<meta name=\"twitter:data1\" content=\"3\u00a0Minuten\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/news\\\/from-perception-to-constrained-decisions\\\/\",\"url\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/news\\\/from-perception-to-constrained-decisions\\\/\",\"name\":\"From perception to constrained decisions - Cyber Valley\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/news\\\/from-perception-to-constrained-decisions\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/news\\\/from-perception-to-constrained-decisions\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/cyber-valley.de\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/thumb_ticker_lg_From_perception_to_constrained_decisions.png\",\"datePublished\":\"2026-03-09T12:22:28+00:00\",\"dateModified\":\"2026-04-20T08:42:35+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/news\\\/from-perception-to-constrained-decisions\\\/#breadcrumb\"},\"inLanguage\":\"de\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/cyber-valley.de\\\/de\\\/news\\\/from-perception-to-constrained-decisions\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"de\",\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/news\\\/from-perception-to-constrained-decisions\\\/#primaryimage\",\"url\":\"https:\\\/\\\/cyber-valley.de\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/thumb_ticker_lg_From_perception_to_constrained_decisions.png\",\"contentUrl\":\"https:\\\/\\\/cyber-valley.de\\\/wp-content\\\/uploads\\\/2026\\\/03\\\/thumb_ticker_lg_From_perception_to_constrained_decisions.png\",\"width\":1000,\"height\":523},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/news\\\/from-perception-to-constrained-decisions\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"From perception to constrained decisions\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/#website\",\"url\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/\",\"name\":\"Cyber Valley\",\"description\":\"From lab to market\",\"publisher\":{\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"de\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/#organization\",\"name\":\"Cyber Valley GmbH\",\"url\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"de\",\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/cyber-valley.de\\\/wp-content\\\/uploads\\\/2026\\\/05\\\/CV_Logo-Lockup_RGB_20210412.svg\",\"contentUrl\":\"https:\\\/\\\/cyber-valley.de\\\/wp-content\\\/uploads\\\/2026\\\/05\\\/CV_Logo-Lockup_RGB_20210412.svg\",\"width\":1112,\"height\":317,\"caption\":\"Cyber Valley GmbH\"},\"image\":{\"@id\":\"https:\\\/\\\/cyber-valley.de\\\/de\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/www.linkedin.com\\\/company\\\/cybervalley\\\/\",\"https:\\\/\\\/www.instagram.com\\\/cybervalley_bw\\\/\",\"https:\\\/\\\/www.youtube.com\\\/channel\\\/UCPi8donVDkcOeM4C0CfP5fg\",\"https:\\\/\\\/bsky.app\\\/profile\\\/cyber-valley.bsky.social\"]}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"From perception to constrained decisions - Cyber Valley","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/","og_locale":"de_DE","og_type":"article","og_title":"From perception to constrained decisions - Cyber Valley","og_description":"Georg Martius\u2019 project, \u201cEmbedding combinatorial reasoning into learning control policies\u201d addressed a practical gap in AI&#8217;s basic machinery: connecting data-driven neural networks with decision-making so they can adhere to clear structures, rules, and constraints. The project developed foundations for integrating combinatorial optimization into learning systems. It has potential applications wherever planning and resource allocation are [&hellip;]","og_url":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/","og_site_name":"Cyber Valley","article_modified_time":"2026-04-20T08:42:35+00:00","og_image":[{"width":1000,"height":523,"url":"https:\/\/cyber-valley.de\/wp-content\/uploads\/2026\/03\/thumb_ticker_lg_From_perception_to_constrained_decisions.png","type":"image\/png"}],"twitter_card":"summary_large_image","twitter_misc":{"Gesch\u00e4tzte Lesezeit":"3\u00a0Minuten"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/","url":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/","name":"From perception to constrained decisions - Cyber Valley","isPartOf":{"@id":"https:\/\/cyber-valley.de\/de\/#website"},"primaryImageOfPage":{"@id":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/#primaryimage"},"image":{"@id":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/#primaryimage"},"thumbnailUrl":"https:\/\/cyber-valley.de\/wp-content\/uploads\/2026\/03\/thumb_ticker_lg_From_perception_to_constrained_decisions.png","datePublished":"2026-03-09T12:22:28+00:00","dateModified":"2026-04-20T08:42:35+00:00","breadcrumb":{"@id":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/#breadcrumb"},"inLanguage":"de","potentialAction":[{"@type":"ReadAction","target":["https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/"]}]},{"@type":"ImageObject","inLanguage":"de","@id":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/#primaryimage","url":"https:\/\/cyber-valley.de\/wp-content\/uploads\/2026\/03\/thumb_ticker_lg_From_perception_to_constrained_decisions.png","contentUrl":"https:\/\/cyber-valley.de\/wp-content\/uploads\/2026\/03\/thumb_ticker_lg_From_perception_to_constrained_decisions.png","width":1000,"height":523},{"@type":"BreadcrumbList","@id":"https:\/\/cyber-valley.de\/de\/news\/from-perception-to-constrained-decisions\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/cyber-valley.de\/de\/"},{"@type":"ListItem","position":2,"name":"From perception to constrained decisions"}]},{"@type":"WebSite","@id":"https:\/\/cyber-valley.de\/de\/#website","url":"https:\/\/cyber-valley.de\/de\/","name":"Cyber Valley","description":"From lab to market","publisher":{"@id":"https:\/\/cyber-valley.de\/de\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/cyber-valley.de\/de\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"de"},{"@type":"Organization","@id":"https:\/\/cyber-valley.de\/de\/#organization","name":"Cyber Valley GmbH","url":"https:\/\/cyber-valley.de\/de\/","logo":{"@type":"ImageObject","inLanguage":"de","@id":"https:\/\/cyber-valley.de\/de\/#\/schema\/logo\/image\/","url":"https:\/\/cyber-valley.de\/wp-content\/uploads\/2026\/05\/CV_Logo-Lockup_RGB_20210412.svg","contentUrl":"https:\/\/cyber-valley.de\/wp-content\/uploads\/2026\/05\/CV_Logo-Lockup_RGB_20210412.svg","width":1112,"height":317,"caption":"Cyber Valley GmbH"},"image":{"@id":"https:\/\/cyber-valley.de\/de\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.linkedin.com\/company\/cybervalley\/","https:\/\/www.instagram.com\/cybervalley_bw\/","https:\/\/www.youtube.com\/channel\/UCPi8donVDkcOeM4C0CfP5fg","https:\/\/bsky.app\/profile\/cyber-valley.bsky.social"]}]}},"_links":{"self":[{"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/news\/10334","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/news"}],"about":[{"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/types\/news"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/media\/10335"}],"wp:attachment":[{"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/media?parent=10334"}],"wp:term":[{"taxonomy":"cv-tag","embeddable":true,"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/cv-tag?post=10334"},{"taxonomy":"location","embeddable":true,"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/location?post=10334"},{"taxonomy":"network-role","embeddable":true,"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/network-role?post=10334"},{"taxonomy":"news-category","embeddable":true,"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/news-category?post=10334"},{"taxonomy":"user-group","embeddable":true,"href":"https:\/\/cyber-valley.de\/de\/wp-json\/wp\/v2\/user-group?post=10334"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}