{"id":8984,"date":"2026-06-10T01:25:28","date_gmt":"2026-06-09T17:25:28","guid":{"rendered":"https:\/\/infernews.com\/blog\/from-simulation-data-collection-to-real-world-humanoid-loco-manipulation\/"},"modified":"2026-06-10T01:29:06","modified_gmt":"2026-06-09T17:29:06","slug":"from-simulation-data-collection-to-real-world-humanoid-loco-manipulation","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/from-simulation-data-collection-to-real-world-humanoid-loco-manipulation\/","title":{"rendered":"OASIS\uff1a\u6a21\u64ec\u8a13\u7df4\u3001\u90e8\u7f72\u4eba\u5f62\u6a5f\u5668\u4eba\u6846\u67b6"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/cover-58dd3ff97642.jpg\" alt=\"Cover\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">OASIS\uff08Simulation Data Collection to Real-World Humanoid Loco-Manipulation\uff09\u7531\u4e2d\u570b\u96fb\u4fe1 TeleAI \u7b49\u6a5f\u69cb\u958b\u6e90\uff0c\u76ee\u6a19\u662f\u8b93\u4eba\u5f62\u6a5f\u5668\u4eba\u7684\u300c\u79fb\u52d5\uff0b\u64cd\u4f5c\u300d\uff08loco-manipulation\uff09\u7b56\u7565\uff0c\u9760\u6a21\u64ec\u6578\u64da\u8a13\u7df4\u5f8c\u5c31\u80fd\u76f4\u63a5\u642c\u5230 Unitree G1 \u771f\u6a5f\u57f7\u884c\uff0c\u7701\u53bb\u8cbb\u6642\u7684\u771f\u6a5f\u9059\u64cd\u4f5c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6574\u500b\u6d41\u7a0b\u6709\u4e09\u500b\u6838\u5fc3\u968e\u6bb5\u3002\u5148\u7528 3D \u751f\u6210\u6a21\u578b\u628a\u771f\u5be6\u7269\u54c1\u5f71\u50cf\u91cd\u5efa\u70ba\u53ef\u7528\u7684 3D \u8cc7\u7522\uff1b\u63a5\u8457\u900f\u904e PICO VR \u982d\u76d4\u642d\u914d Isaac Lab v2.1.0 \u9032\u884c\u6c89\u6d78\u5f0f\u9059\u64cd\u4f5c\u6536\u96c6\u793a\u7bc4\u6578\u64da\uff0c\u904e\u7a0b\u4e2d PICO \u6703\u5373\u6642\u63a5\u6536\u982d\u90e8\u76f8\u6a5f\u756b\u9762\uff0cX\u3001Y\u3001A\u3001B \u9375\u5206\u5225\u8ca0\u8cac\u9304\u5f71\u3001\u505c\u6b62\u3001\u91cd\u7f6e\u3001\u8996\u89d2\u5207\u63db\uff0c\u677f\u6a5f\u8207\u63e1\u628a\u53ef\u63a7\u5236\u593e\u722a\u958b\u5408\uff1b\u6700\u5f8c\u5728 replay \u968e\u6bb5\u52a0\u5165\u591a\u6a23\u5316\u7684\u5149\u7167\u8207\u74b0\u5883\u96a8\u6a5f\u5316\uff0c\u4e26\u8a13\u7df4\u5c64\u6b21\u5316\u8996\u89ba\u904b\u52d5\u7b56\u7565\uff08hierarchical visuomotor policy\uff09\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u8ad6\u6587\u7d50\u679c\u6307\u51fa\uff0c\u5728\u96f6\u6a23\u672c\uff08zero-shot\uff09\u90e8\u7f72\u4e0b\uff0c\u7d14\u6a21\u64ec\u6578\u64da\u8a13\u7df4\u7684\u7b56\u7565\u5728\u5927\u591a\u6578\u4efb\u52d9\u7684\u6210\u529f\u7387\uff0c\u53cd\u800c\u9ad8\u65bc\u7528\u771f\u6a5f\u9059\u64cd\u4f5c\u6578\u64da\u8a13\u7df4\u7684\u7248\u672c\uff0c\u4e3b\u8981\u539f\u56e0\u662f\u6a21\u64ec\u80fd\u6db5\u84cb\u771f\u6a5f\u96e3\u4ee5\u62cd\u651d\u7684\u5149\u7167\u8207\u74b0\u5883\u8b8a\u5316\u3002\u5718\u968a\u4e5f\u5c55\u793a\u4e86\u4e00\u9375\u91cd\u555f\u7684\u6545\u969c\u6062\u5fa9\u6a5f\u5236\uff0c\u89e3\u6c7a\u771f\u6a5f\u8dcc\u5012\u5f8c\u9700\u8981\u4eba\u5de5\u642c\u904b\u3001\u91cd\u7f6e\u5834\u666f\u7684\u75db\u9ede\u3002<\/p>\n\n\n<figure class=\"wp-block-embed-youtube wp-block-embed is-type-video is-provider-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"lyte-wrapper\" title=\"OASIS: From Simulation Data Collection to Real-World Humanoid Loco-Manipulation\" style=\"width:853px;max-width:100%;margin:5px auto;\"><div class=\"lyMe\" id=\"WYL_6A-s6sQL3c4\" itemprop=\"video\" itemscope itemtype=\"https:\/\/schema.org\/VideoObject\"><div><meta itemprop=\"thumbnailUrl\" content=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2F6A-s6sQL3c4%2Fhqdefault.jpg\" \/><meta itemprop=\"embedURL\" content=\"https:\/\/www.youtube.com\/embed\/6A-s6sQL3c4\" \/><meta itemprop=\"duration\" content=\"PT3M8S\" \/><meta itemprop=\"uploadDate\" content=\"2026-06-07T06:11:06Z\" \/><\/div><div id=\"lyte_6A-s6sQL3c4\" data-src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2F6A-s6sQL3c4%2Fhqdefault.jpg\" class=\"pL\"><div class=\"tC\"><div class=\"tT\" itemprop=\"name\">OASIS: From Simulation Data Collection to Real-World Humanoid Loco-Manipulation<\/div><\/div><div class=\"play\"><\/div><div class=\"ctrl\"><div class=\"Lctrl\"><\/div><div class=\"Rctrl\"><\/div><\/div><\/div><noscript><a href=\"https:\/\/youtu.be\/6A-s6sQL3c4\" rel=\"nofollow noopener\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/plugins\/wp-youtube-lyte\/lyteCache.php?origThumbUrl=https%3A%2F%2Fi.ytimg.com%2Fvi%2F6A-s6sQL3c4%2F0.jpg\" alt=\"OASIS: From Simulation Data Collection to Real-World Humanoid Loco-Manipulation\" width=\"853\" height=\"460\" \/><br \/>Watch this video on YouTube<\/a><\/noscript><meta itemprop=\"description\" content=\"Recent progress in robot manipulation has been largely driven by learning from large-scale demonstrations. For humanoid robot loco-manipulation tasks, however, existing data sources force an unsatisfying tradeoff between trajectory quality and scalability. Real-world teleoperation provides the highest-quality trajectories but requires dedicated physical space and time-consuming scene resets. Simulation offers an alternative way out of this dilemma: it can produce clean, embodiment-aligned data at scale without any physical hardware. In this paper, we propose OASIS, a simulation-data-driven framework for humanoid loco-manipulation. OASIS automatically reconstructs realistic object assets from real-world images using a 3D generative model. Based on these assets, trajectories are first collected through teleoperation in simulation, and then augmented under diverse domain randomizations in a post-processing stage. With the resulting simulation data, we further design a hierarchical visuomotor policy for humanoid loco-manipulation. Extensive experiments on the real humanoid robot show that, under zero-shot deployment, the policy trained on our simulation data achieves higher success rates on most tasks than that trained on real-robot teleoperation data, owing largely to the broad lighting and environmental variations covered by our simulation rendering, which real-robot data fails to capture.\"><\/div><\/div><div class=\"lL\" style=\"max-width:100%;width:853px;margin:5px auto;\"><\/div><figcaption><\/figcaption><\/figure>\n\n\n<p class=\"wp-block-paragraph\"><strong>\u5e7e\u500b\u91cd\u9ede\u6458\u8981\uff1a<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u9805\u76ee\u5b9a\u4f4d<\/strong>\uff1a\u6a21\u64ec\u6578\u64da\u9a45\u52d5\u7684\u4eba\u5f62\u6a5f\u5668\u4eba loco-manipulation \u6846\u67b6\uff0c\u4e26\u975e\u901a\u7528 LLM \u6216\u591a\u6a21\u614b\u6a21\u578b<\/li>\n\n\n\n<li><strong>\u5275\u65b0\u8655<\/strong>\uff1a\u4ee5 3D \u751f\u6210\u6a21\u578b\u91cd\u5efa\u8cc7\u7522\uff0c\u52a0\u4e0a\u5f8c\u8655\u7406\u968e\u6bb5\u7684 domain randomization\uff0c\u89e3\u6c7a\u6a21\u64ec\u5230\u771f\u6a5f\u7684\u5dee\u8ddd<\/li>\n\n\n\n<li><strong>\u76f8\u95dc\u6a21\u578b\uff0f\u7d44\u4ef6<\/strong>\uff1aUnitree G1 \u4eba\u5f62\u6a5f\u5668\u4eba\u3001Isaac Lab \u6a21\u64ec\u5668\u3001PICO VR\u3001PICO SDK\u3001GMR\u3001XRoboToolkit<\/li>\n\n\n\n<li><strong>\u786c\u9ad4\u8981\u6c42<\/strong>\uff1aPC \u5fc5\u9808\u5b89\u88dd Ubuntu 22.04\uff0cPICO \u9808\u8207 PC \u8655\u65bc\u540c\u4e00\u7db2\u6bb5<\/li>\n\n\n\n<li><strong>\u9069\u7528\u5834\u666f<\/strong>\uff1a\u7814\u7a76 humanoid \u79fb\u52d5\u64cd\u4f5c\u3001\u60f3\u7528\u6a21\u64ec\u53d6\u4ee3\u90e8\u5206\u771f\u6a5f\u9059\u64cd\u4f5c\u6210\u672c\u3001Unitree G1 \u958b\u767c\u8005<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u9069\u5408\u6a5f\u5668\u4eba\u7814\u7a76\u5718\u968a\u3001\u4eba\u5f62\u6a5f\u5668\u4eba\u958b\u767c\u8005\uff0c\u4ee5\u53ca\u60f3\u5617\u8a66 sim-to-real \u8def\u7dda\u7684\u4eba\u3002\u4e0d\u904e\u6574\u5957\u74b0\u5883\u5c0d VR \u8a2d\u5099\u8207 Linux \u7248\u672c\u8981\u6c42\u56b4\u683c\uff0c\u7d14 Windows \u6216 Mac \u7528\u6236\u9580\u6abb\u504f\u9ad8\uff0c\u5efa\u8b70\u5148\u8a55\u4f30\u624b\u4e0a\u786c\u9ad4\u662f\u5426\u5230\u4f4d\u518d\u6295\u5165\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/TeleHuman\/OASIS\" rel=\"noopener noreferrer\">https:\/\/github.com\/TeleHuman\/OASIS<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Paper\uff1a<\/strong> <a href=\"https:\/\/arxiv.org\/pdf\/2606.08548\" rel=\"noopener noreferrer\">https:\/\/arxiv.org\/pdf\/2606.08548<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>OASIS \u662f\u4e00\u5957\u91dd\u5c0d Unitree G1 \u4eba\u5f62\u6a5f\u5668\u4eba\u7684\u6a21\u64ec\u6578\u64da\u6536\u96c6\u8207\u8996\u89ba\u904b\u52d5\u7b56\u7565\u6846\u67b6\uff0c\u4e3b\u6253\u96f6\u6a23\u672c\u771f\u6a5f\u9077\u79fb\u3002<\/p>\n","protected":false},"author":8,"featured_media":8983,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,184,197],"tags":[],"class_list":["post-8984","post","type-post","status-publish","format-standard","hentry","category-133","category-robotic","category-framework"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8984","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=8984"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8984\/revisions"}],"predecessor-version":[{"id":8986,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8984\/revisions\/8986"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/8983"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=8984"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=8984"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=8984"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}