{"id":11253,"date":"2024-10-14T13:23:34","date_gmt":"2024-10-14T11:23:34","guid":{"rendered":"https:\/\/www.isir.upmc.fr\/?post_type=isir_news&#038;p=11253"},"modified":"2025-02-07T10:37:48","modified_gmt":"2025-02-07T09:37:48","slug":"ingredients-for-motion-planning-powered-reinforcement-learning-hdr-defence-of-nicolas-perrin-gilbert","status":"publish","type":"isir_news","link":"https:\/\/www.isir.upmc.fr\/news\/ingredients-for-motion-planning-powered-reinforcement-learning-hdr-defence-of-nicolas-perrin-gilbert\/?lang=en","title":{"rendered":"\u00ab\u00a0Ingredients for Motion Planning-powered Reinforcement Learning\u00a0\u00bb, HDR defence of Nicolas Perrin-Gilbert"},"content":{"rendered":"\n<div class=\"wp-block-media-text is-stacked-on-mobile\" style=\"grid-template-columns:26% auto\"><figure class=\"wp-block-media-text__media\"><img loading=\"lazy\" decoding=\"async\" width=\"391\" height=\"391\" src=\"https:\/\/www.isir.upmc.fr\/wp-content\/uploads\/2022\/06\/photo_perso_visage.jpg\" alt=\"\" class=\"wp-image-8257 size-full\" srcset=\"https:\/\/www.isir.upmc.fr\/wp-content\/uploads\/2022\/06\/photo_perso_visage.jpg 391w, https:\/\/www.isir.upmc.fr\/wp-content\/uploads\/2022\/06\/photo_perso_visage-300x300.jpg 300w, https:\/\/www.isir.upmc.fr\/wp-content\/uploads\/2022\/06\/photo_perso_visage-150x150.jpg 150w\" sizes=\"auto, (max-width: 391px) 100vw, 391px\" \/><\/figure><div class=\"wp-block-media-text__content\">\n<p class=\"wp-block-paragraph\"><strong>Nicolas Perrin-Gilbert<\/strong>, CNRS research fellow, will be defending his habilitation to supervise research (HDR) on Thursday 17 October at 4.30pm at ISIR, on the Pierre and Marie Curie campus of Sorbonne University.<\/p>\n<\/div><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Title of work:&nbsp;<strong>\u201cIngredients for Motion Planning-powered Reinforcement Learning\u201d<\/strong>.&nbsp;<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The members of the jury are as follows:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">&#8211; Aleksandra Faust, Google Deepmind (Reporter),<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">&#8211; Matthieu Geist, University of Lorraine &amp; Cohere (Examiner),<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">&#8211; Nicolas Mansard, LAAS-CNRS ( Reporter),<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">&#8211; Jochen J. Steil, Technical University of Brunswick ( Reporter),<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">&#8211; Nicolas Thome, Sorbonne University (Examiner).<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Summary<\/strong>:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">My HDR focuses on four main contributions, quite distinct, but all having a link with a common objective, that of improving exploration in reinforcement learning through the use of movement planning techniques based on random sampling.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The first contribution concerns geometric transformations between continuous movements and discrete sequences of contacts, which enables us to approach the problems of generating locomotion movements from a new angle. The second contribution proposes a technique for generalising or imitating trajectories based on the application of successive diffeomorphic transformations. The third deals with the management of successive sequences of goals, in particular the notions of intermediate goals and ways of considering them in order to achieve an overall goal more efficiently. Finally, the fourth contribution presents an \u2018off-policy\u2019 reinforcement learning algorithm, aimed at improving the training of control policies when a significant part of the data comes from exploration trajectories.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Contact<\/strong>: <a href=\"https:\/\/www.isir.upmc.fr\/personnel\/perrin\/?lang=en\" target=\"_blank\" rel=\"noreferrer noopener\">Nicolas Perrin-Gilbert<\/a>, CNRS Research Fellow<\/p>\n\n\n\n<hr class=\"wp-block-separator has-text-color has-white-color has-alpha-channel-opacity has-white-background-color has-background\"\/>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Published on 14\/10\/2024.<\/em><\/p>\n","protected":false},"featured_media":10015,"template":"","isir_news_categories":[260],"class_list":["post-11253","isir_news","type-isir_news","status-publish","has-post-thumbnail","hentry","isir_news_categories-soutenance"],"_links":{"self":[{"href":"https:\/\/www.isir.upmc.fr\/wp-json\/wp\/v2\/isir_news\/11253","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.isir.upmc.fr\/wp-json\/wp\/v2\/isir_news"}],"about":[{"href":"https:\/\/www.isir.upmc.fr\/wp-json\/wp\/v2\/types\/isir_news"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.isir.upmc.fr\/wp-json\/wp\/v2\/media\/10015"}],"wp:attachment":[{"href":"https:\/\/www.isir.upmc.fr\/wp-json\/wp\/v2\/media?parent=11253"}],"wp:term":[{"taxonomy":"isir_news_categories","embeddable":true,"href":"https:\/\/www.isir.upmc.fr\/wp-json\/wp\/v2\/isir_news_categories?post=11253"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}