{
  "schema_version": "paper_public_manifest_v1",
  "paper_id": "mocap_2026_10",
  "slug": "mocap",
  "title": "MoCaP：人类视频驱动零样本移动操作",
  "authors": [],
  "source": {
    "arxiv_id": "2610.20998",
    "pdf_url": "https://arxiv.org/pdf/2610.20998",
    "project_url": "",
    "github_url": "",
    "huggingface_url": "",
    "original_source": "https://arxiv.org/abs/2610.20998"
  },
  "site": {
    "post_url": "/posts/mocap",
    "canonical_url": "https://haiguangboy.com/posts/mocap",
    "cover_image": "https://static.haiguangboy.com/papers/mocap/cover.webp"
  },
  "taxonomy": {
    "domain": "embodied_ai",
    "track": "action_generation",
    "tasks": [
      "embodied_ai",
      "action_generation",
      "robotics",
      "state_prediction",
      "具身智能",
      "移动操作",
      "机器人导航",
      "人类视频",
      "mocap"
    ],
    "related_topics": [
      {
        "paper_id": "an_open_foundation_model_towards_2026_07",
        "title": "An_Open_Foundation_Model_Towards",
        "url": "https://haiguangboy.com/posts/an_open_foundation_model_towards",
        "relation": "contrast",
        "summary": "全部学习组件冻结，不做任务专门训练 contradicts 联合co-training异构人机数据是结构性次优方案",
        "strength": "strong"
      },
      {
        "paper_id": "omega_0_a_latent_predictive_world_action_model_for_concurrent_humanoid_loco_mani_2026_08",
        "title": "未来预测越准，机器人反而越迟滞",
        "url": "https://haiguangboy.com/posts/omega-0",
        "relation": "contrast",
        "summary": "底盘移动后重新观测，再做全臂规划 contradicts 问题定位:已有 VLA 要么手臂中心、要么把底盘和手臂当两个未协调的子系统",
        "strength": "strong"
      },
      {
        "paper_id": "learning_a_thousand_tasks_in_a_day_2026_08",
        "title": "1 天 1000 任务，靠的是归纳偏置",
        "url": "https://haiguangboy.com/posts/mt3-thousand-tasks",
        "relation": "same_track",
        "summary": "Zero-shot仍依赖标定、深度和现成执行模块 validates 感知依赖:仅视觉、单相机,无触觉,依赖准确分割",
        "strength": "strong"
      },
      {
        "paper_id": "dextouch_wm_learning_action_conditioned_tactile_world_models_from_human_touch_fo_2026_09",
        "title": "DexTouch-WM：人类触觉预训练",
        "url": "https://haiguangboy.com/posts/dextouch-wm",
        "relation": "same_track",
        "summary": "Zero-shot仍依赖标定、深度和现成执行模块 validates 人类数据不是无硬件、无机器人数据的直接迁移",
        "strength": "strong"
      },
      {
        "paper_id": "google_blog_gemini_robotics_er_2_20260730_2026_07",
        "title": "Introducing Gemini Robotics ER 2",
        "url": "https://haiguangboy.com/posts/google_blog_gemini_robotics_er_2_20260730",
        "relation": "same_track",
        "summary": "代码代理用执行反馈协调导航与抓取 validates 对导航从业者的直接含义:导航不是被基座吸收,是被降格成基座调用的一个工具",
        "strength": "strong"
      },
      {
        "paper_id": "mint_a_unified_model_for_world_space_camera_and_hand_motion_estimation_from_scal_2026_09",
        "title": "MINT_A_Unified_Model_for_World_Space_Camera_and_Hand_Motion_Estimation_from_Scal",
        "url": "https://haiguangboy.com/posts/mint_a_unified_model_for_world_space_camera_and_hand_motion_estimation_from_scal",
        "relation": "same_track",
        "summary": "VGGT相对位姿通过机器人深度获得尺度 validates 两阶段训练背后的判断:学会稳定的手-相机耦合几何关系,和校正相机轨迹的绝对尺度,是两个可以分开解决的子问题——前者需要大量多样但带噪声的数据,后者只需要少量高精度数据",
        "strength": "strong"
      }
    ]
  },
  "analyst_take": {
    "type": "author_opinion",
    "text": "💡 工程启发\n人类视频的用途，可以延伸到执行时直接调用的记忆与接近经验。这与《Towards Agentic Robots》讨论的执行器启动条件相呼应：选对技能，还得把机器人送到技能能执行的状态。\nMoCaP展示了一条不更新模型权重的经验复用路径。它能否扩展，仍要看换环境、换录像、统一后端后，完整操作成功率是否稳定受益。"
  },
  "ruling": {
    "importance_score": 3.0,
    "one_sentence": "MoCaP：人类视频驱动零样本移动操作"
  },
  "asset_base_url": "https://static.haiguangboy.com/papers/mocap",
  "assets": [
    {
      "type": "pdf_screenshot",
      "object_key": "papers/mocap/page_01.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/mocap/page_01.webp",
      "role": "paper_first_page",
      "size_bytes": 242744
    },
    {
      "type": "pdf_screenshot",
      "object_key": "papers/mocap/key_figure.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/mocap/key_figure.webp",
      "role": "method_figure",
      "size_bytes": 238666
    },
    {
      "type": "cover_image",
      "object_key": "papers/mocap/cover.webp",
      "content_type": "image/webp",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/mocap/cover.webp",
      "role": "post_cover",
      "size_bytes": 72092
    },
    {
      "type": "public_brief",
      "object_key": "papers/mocap/public_brief.md",
      "content_type": "text/markdown; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/mocap/public_brief.md",
      "role": "public_brief",
      "size_bytes": 4676
    },
    {
      "type": "public_brief",
      "object_key": "papers/mocap/public_brief.en.md",
      "content_type": "text/markdown; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/mocap/public_brief.en.md",
      "role": "public_brief_en",
      "size_bytes": 5867
    },
    {
      "type": "public_manifest",
      "object_key": "papers/mocap/public_manifest.en.json",
      "content_type": "application/json; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/mocap/public_manifest.en.json",
      "role": "public_manifest_en",
      "size_bytes": 6743
    },
    {
      "type": "public_manifest",
      "object_key": "papers/mocap/public_manifest.json",
      "content_type": "application/json; charset=utf-8",
      "upload_status": "uploaded",
      "bucket": "paper-assets",
      "url": "https://static.haiguangboy.com/papers/mocap/public_manifest.json",
      "role": "public_manifest",
      "size_bytes": 7217
    }
  ],
  "published_at": "2026-10-18T17:34:32+08:00",
  "created_at": "2026-10-18T17:34:32+08:00",
  "updated_at": "2026-10-18T17:34:42+08:00"
}