{"id":6797,"date":"2025-05-27T05:48:00","date_gmt":"2025-05-26T20:48:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=6797"},"modified":"2025-05-25T08:53:11","modified_gmt":"2025-05-24T23:53:11","slug":"dreamgen-unlocking-generalization-in-robot-learning-through-neural-trajectories","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=6797","title":{"rendered":"DreamGen: Unlocking Generalization in Robot Learning through Neural Trajectories"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>DreamGen: Unlocking Generalization in Robot Learning through Neural Trajectories\u00a0<\/strong>[120.3]<br>DreamGen\u306f\u3001\u30cb\u30e5\u30fc\u30e9\u30eb\u30c8\u30e9\u30b8\u30a7\u30af\u30c8\u30ea\u3092\u901a\u3058\u3066\u884c\u52d5\u3084\u74b0\u5883\u3092\u4e00\u822c\u5316\u3059\u308b\u30ed\u30dc\u30c3\u30c8\u30dd\u30ea\u30b7\u30fc\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u305f\u3081\u306e\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u3060\u3002 \u79c1\u305f\u3061\u306e\u7814\u7a76\u306f\u3001\u624b\u4f5c\u696d\u306b\u3088\u308b\u30c7\u30fc\u30bf\u53ce\u96c6\u3092\u8d85\u3048\u3066\u3001\u30ed\u30dc\u30c3\u30c8\u5b66\u7fd2\u3092\u30b9\u30b1\u30fc\u30eb\u3059\u308b\u305f\u3081\u306e\u3001\u6709\u671b\u306a\u65b0\u305f\u306a\u8ef8\u3092\u78ba\u7acb\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2505.12705v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2505.12705v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 19 May 2025 04:55:39 GMT)<\/li>\n\n\n\n<li>\u300cThis pipeline is designed to be general-purpose across different robots, environments, and tasks. (1) We fine-tune video world models on a target robot to capture the dynamics and kinematics of the specific embodiment; (2) we prompt the model with pairs of initial frames and language instructions to generate large volumes of robot videos, capturing both familiar behaviors from fine-tuning and novel ones in unseen settings; (3) we then extract pseudo-actions using either a latent action model [13] or an inverse dynamics model (IDM)[14]; (4) finally, we use the resulting video-action sequence pairs, dubbed neural trajectories, for training downstream visuomotor policies.\u300d\u3068\u52d5\u753b\u751f\u6210\u30e2\u30c7\u30eb\u3092\u6d3b\u7528\u3057\u305f\u30c7\u30fc\u30bf\u5408\u6210\u624b\u6cd5\u306e\u63d0\u6848\u3002\u30a4\u30e1\u30fc\u30b8\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u306e\u3088\u3046\u3067\u9762\u767d\u3044\u3002<\/li>\n\n\n\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30b5\u30a4\u30c8\u306f<a href=\"https:\/\/research.nvidia.com\/labs\/gear\/dreamgen\/\">DreamGen<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[92,390,449,554],"class_list":["post-6797","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-data-augmentation","tag-synthetic-data","tag-world-model","tag-554"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6797","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=6797"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6797\/revisions"}],"predecessor-version":[{"id":6798,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6797\/revisions\/6798"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6797"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6797"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6797"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}