{"id":9169,"date":"2026-06-16T06:14:39","date_gmt":"2026-06-15T22:14:39","guid":{"rendered":"https:\/\/infernews.com\/blog\/translating-actions-into-motion-and-contact-images-for-embodied-world-models\/"},"modified":"2026-06-16T06:15:05","modified_gmt":"2026-06-15T22:15:05","slug":"translating-actions-into-motion-and-contact-images-for-embodied-world-models","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/translating-actions-into-motion-and-contact-images-for-embodied-world-models\/","title":{"rendered":"iMaC\uff1a\u628a\u6a5f\u68b0\u81c2\u52d5\u4f5c\u8b8a\u6210\u53ef\u9810\u6e2c\u5f71\u50cf"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/pipeline-d17aa82e2c08.jpg\" alt=\"overview\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6642\u4e0d\u5c11 action-conditioned video models \u6703\u628a\u672a\u4f86\u52d5\u4f5c\u58d3\u6210 compact vectors\uff0c\u518d\u7d93 learned conditioning modules \u4ea4\u7d66\u6a21\u578b\u8655\u7406\uff1b\u4f5c\u8005\u8a8d\u70ba\u9019\u7a2e\u505a\u6cd5\u8981\u6a21\u578b\u81ea\u884c\u731c\u6e2c\u7d30\u5fae\u7a7a\u9593\u5f8c\u679c\uff0c\u9047\u5230 real manipulation \u6642\uff0c\u5e7e\u5398\u7c73\u5dee\u8ddd\u5df2\u8db3\u4ee5\u6539\u8b8a\u63a5\u89f8\u3001\u7269\u4ef6\u79fb\u52d5\u8207\u4efb\u52d9\u6210\u6557\u3002iMaC \u5c6c\u65bc<strong>\u4e16\u754c\u6a21\u578b\u8207\u5f71\u7247\u751f\u6210\u6a21\u578b<\/strong>\uff0c\u6838\u5fc3\u662f\u628a future joint actions \u8f49\u6210 image-like controls\uff0c\u6e1b\u5c11\u300c\u52d5\u4f5c\u6709\u8f38\u5165\uff0c\u4f46\u7a7a\u9593\u95dc\u4fc2\u8868\u9054\u4e0d\u8db3\u300d\u7684\u554f\u984c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9019\u500b\u9805\u76ee\u7684\u65b9\u6cd5\u76f8\u7576\u5177\u9ad4\uff1a\u5148\u5229\u7528 robot URDF \u8207 forward kinematics\uff0c\u6e32\u67d3 future robot-observation control videos\uff0c\u4e5f\u5c31\u662f motion images\uff1b\u4e4b\u5f8c\u518d\u52a0\u5165 depth \u4f5c\u70ba\u8f14\u52a9\u8a0a\u865f\uff0c\u914d\u5408 3D pointclouds \u5efa\u7acb two-stream geometry controls\uff0c\u4e5f\u5c31\u662f contact images\u3002\u820a\u7bc4\u5f0f\u4e3b\u8981\u9760\u62bd\u8c61\u5411\u91cf\u689d\u4ef6\u5316\uff0ciMaC \u5247\u628a\u300c\u672a\u4f86\u6a5f\u68b0\u81c2\u6703\u51fa\u73fe\u5728\u54ea\u88e1\u3001\u5982\u4f55\u63a5\u8fd1\u5834\u666f\u300d\u76f4\u63a5\u8b8a\u6210\u53ef\u898b\u63a7\u5236\uff0c\u9019\u662f\u5b83\u6700\u6e05\u6670\u7684\u6280\u8853\u5206\u91ce\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">GitHub \u5132\u5b58\u5eab\u63d0\u4f9b training\u3001preprocessing \u8207 inference code\uff0c\u8986\u84cb RND-mix stage-one\u3001stage-two\uff0c\u4ee5\u53ca WorldArena \u4e09\u689d\u6d41\u7a0b\u3002\u60f3\u8a66\u9019\u500b\u9805\u76ee\u7684\u4eba\uff0c\u6703\u5148\u7531\u8cc7\u6599\u524d\u8655\u7406\u3001depth \u8207 3D condition \u5efa\u7acb\u958b\u59cb\uff0c\u518d\u8dd1 validation inference \u770b\u751f\u6210\u5f71\u7247\u662f\u5426\u8ddf\u52d5\u4f5c\u4e00\u81f4\uff1b\u82e5\u672c\u8eab\u505a robotic policy evaluation\uff0c\u9084\u53ef\u4ee5\u63a5\u5230 WorldArena \u6216 online RND evaluation \u5834\u666f\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u628a actions \u8f49\u6210 motion images \u8207 contact images\uff0c\u7a7a\u9593\u689d\u4ef6\u66f4\u660e\u78ba<\/li>\n\n\n\n<li>\u7528 depth encoding \u548c 3D pointclouds \u5f37\u5316 robot-scene \u5e7e\u4f55\u7406\u89e3<\/li>\n\n\n\n<li>\u52a0\u5165 training-time rollout strategy\uff0c\u76ee\u6a19\u662f\u652f\u63f4\u66f4\u9577\u6642\u5e8f\u751f\u6210\u4e26\u6e1b\u5c11 exposure bias<\/li>\n\n\n\n<li>\u5132\u5b58\u5eab\u540c\u6642\u6db5\u84cb\u8a13\u7df4\u3001\u524d\u8655\u7406\u3001\u63a8\u8ad6\uff0c\u4e0d\u53ea\u662f\u8ad6\u6587\u5c55\u793a\u6a21\u578b<\/li>\n\n\n\n<li>\u76f8\u95dc\u7d44\u4ef6\u5305\u62ec Wan transformer variants\u3001Diffusion inference pipelines\u3001RobotWin 2.0\u3001WorldArena<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u6027\u80fd\u65b9\u9762\uff0c\u8ad6\u6587\u6307\u51fa\u5b83\u5728\u516b\u500b\u9577\u6642\u5e8f\u771f\u5be6\u6a5f\u68b0\u4eba\u64cd\u4f5c\u4efb\u52d9\u4e2d\uff0cworld-model success estimates \u8207\u771f\u5be6 policy performance \u5448\u5f37\u6b63\u76f8\u95dc\u3002\u9019\u500b\u7d50\u679c\u7684\u50f9\u503c\u4e0d\u5728\u65bc\u53d6\u4ee3\u771f\u6a5f\u6e2c\u8a66\uff0c\u800c\u662f\u5728\u6b63\u5f0f\u843d\u6a5f\u524d\uff0c\u5148\u7528\u751f\u6210\u5f0f world model \u7be9\u9078 policy checkpoints\uff1b\u5c0d\u7814\u7a76 embodied evaluation\u3001robotics \u8207\u4e16\u754c\u6a21\u578b\u7684\u4eba\u4f86\u8aaa\uff0ciMaC \u5c6c\u65bc\u76f8\u7576\u503c\u5f97\u8ddf\u9032\u7684\u4e00\u500b\u65b9\u5411\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/imac-wm\/iMac\" rel=\"noopener noreferrer\">https:\/\/github.com\/imac-wm\/iMac<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Paper\uff1a<\/strong> <a href=\"https:\/\/arxiv.org\/pdf\/2606.09813\" rel=\"noopener noreferrer\">https:\/\/arxiv.org\/pdf\/2606.09813<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>iMaC \u4e0d\u662f\u53ea\u9935\u52d5\u4f5c\u5411\u91cf\u5165\u6a21\u578b\uff0c\u800c\u662f\u628a\u672a\u4f86\u52d5\u4f5c\u8f49\u6210\u5f71\u50cf\u63a7\u5236\u3002\u9019\u7a2e world model \u66f4\u8cbc\u8fd1\u6a5f\u68b0\u81c2\u8207\u5834\u666f\u7684\u7a7a\u9593\u4e92\u52d5\u3002<\/p>\n","protected":false},"author":8,"featured_media":9168,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,33,31,30,132,162,166,159,167,150,76,127,146,141,184,186,196],"tags":[],"class_list":["post-9169","post","type-post","status-publish","format-standard","hentry","category-133","category-stablediffusion","category-video","category-image","category-3d","category-ai-productions","category-mac","category-vibe-coding","category-win","category-150","category-76","category-127","category-146","category-141","category-robotic","category-186","category-196"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9169","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=9169"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9169\/revisions"}],"predecessor-version":[{"id":9171,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9169\/revisions\/9171"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9168"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9169"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9169"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9169"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}