<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="ja">
  <id>tag:noteflowai.github.io,2026:physical-ai-radar:feed:ja</id>
  <title>Physical AI フロンティア・レーダー</title>
  <subtitle>フィジカルAIの最前線を毎日更新：8つの軸、三言語の要約、追跡可能な根拠。</subtitle>
  <updated>2026-10-12T23:40:00Z</updated>
  <link rel="self" type="application/atom+xml" href="https://noteflowai.github.io/physical-ai-radar/radar/feed.ja.xml" />
  <link rel="alternate" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/ja/" />
  <author>
    <name>Physical AI フロンティア・レーダー</name>
    <uri>https://github.com/noteflowai/physical-ai-radar</uri>
  </author>
  <generator uri="https://github.com/noteflowai/physical-ai-radar">pairadar</generator>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.13643</id>
    <title>AutoBio+: Benchmarking Dexterous Manipulation with Dynamic Experimental Phenomena in Realistic Laboratories</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.13643" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html" />
    <published>2026-10-12T23:40:00Z</published>
    <updated>2026-10-12T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.13643</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Automating biological laboratories requires robots to perform dexterous tool interactions under realistic visual conditions while handling dynamic experimental phenomena, presenting significant challenges for robotic manipulation.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.13643 · 2026-10-09&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 14 tasks&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Automating biological laboratories requires robots to perform dexterous tool interactions under realistic visual conditions while handling dynamic experimental phenomena, presenting significant challenges for robotic manipulation.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html"&gt;2026-10-13 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.13117</id>
    <title>Forward Dynamics Models</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.13117" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html" />
    <published>2026-10-12T23:40:00Z</published>
    <updated>2026-10-12T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.13117</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language-action (VLA) models and world action models (WAMs) have advanced robotic policy learning by building on pretrained vision-language models (VLMs) and video generative models (VGMs), respectively.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.13117 · 2026-10-09&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 99.4% · 13%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Vision-language-action (VLA) models and world action models (WAMs) have advanced robotic policy learning by building on pretrained vision-language models (VLMs) and video generative models (VGMs), respectively.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html"&gt;2026-10-13 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.13345</id>
    <title>Evo-RL: Iterative Robot Reinforcement Learning from Selective Experience</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.13345" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html" />
    <published>2026-10-12T23:40:00Z</published>
    <updated>2026-10-12T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.13345</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="R" label="[R]" />
    <summary type="text">Existing real-world reinforcement learning methods improve robot policies by combining human corrections with autonomous rollouts.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.13345 · 2026-10-09&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 1,000 hours&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり · コードまたは重み公開&lt;/p&gt;
&lt;blockquote&gt;Existing real-world reinforcement learning methods improve robot policies by combining human corrections with autonomous rollouts.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html"&gt;2026-10-13 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:aws-physical-ai:a07068e67300</id>
    <title>3D Gaussian Splatting in practice: Industry use cases</title>
    <link rel="alternate" href="https://aws.amazon.com/blogs/physical-ai/3d-gaussian-splatting-in-practice-industry-use-cases/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html" />
    <published>2026-10-12T23:40:00Z</published>
    <updated>2026-10-12T23:40:00Z</updated>
    <author>
      <name>AWS Physical AI Blog</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="O" label="[O]" />
    <summary type="text">Gaussian Splatting is dramatically lowering the cost and complexity of photorealistic 3D, making once impractical use cases like near real time capture, robotics, and simulation viable across media, construction, manufacturing, and real…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [O]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: AWS Physical AI Blog · 2026-10-06&lt;/p&gt;
&lt;blockquote&gt;Gaussian Splatting is dramatically lowering the cost and complexity of photorealistic 3D, making once impractical use cases like near real time capture, robotics, and simulation viable across media, construction, manufacturing, and real…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html"&gt;2026-10-13 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:fb73046bfc9e</id>
    <title>Closing the reality gap: Why real-time execution has become the new competitive advantage</title>
    <link rel="alternate" href="https://www.therobotreport.com/closing-reality-gap-why-real-time-execution-has-become-new-competitive-advantage/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html" />
    <published>2026-10-12T23:40:00Z</published>
    <updated>2026-10-12T23:40:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="M" label="[M]" />
    <summary type="text">Daifuku's Cecile Dick-Calmes said success is no longer measured solely by automation density or equipment utilization.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-12&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Daifuku&amp;#x27;s Cecile Dick-Calmes said success is no longer measured solely by automation density or equipment utilization.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html"&gt;2026-10-13 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:9c414b247031</id>
    <title>费莫一刘念邱：智驾无图的经验，也能迁移到机器人 | 物理AI 50人</title>
    <link rel="alternate" href="https://www.leiphone.com/category/transportation/Y6GLcnioywVn3u7K.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html" />
    <published>2026-10-12T23:40:00Z</published>
    <updated>2026-10-12T23:40:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">刘念邱很难被归进典型的具身创业者模板里。 他不是从机器人实验室出来的学者，也不是过去两年才进入 Physical AI 的连续创业者。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-12&lt;/p&gt;
&lt;blockquote&gt;刘念邱很难被归进典型的具身创业者模板里。 他不是从机器人实验室出来的学者，也不是过去两年才进入 Physical AI 的连续创业者。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html"&gt;2026-10-13 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:28b795de6185</id>
    <title>产业出题，科技答题：2026 IEIC国际具身智能技能大赛，用真实岗位任务考核机器人落地能力</title>
    <link rel="alternate" href="https://www.leiphone.com/category/industrynews/ZFAMCctM7KUw5gr3.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html" />
    <published>2026-10-12T23:40:00Z</published>
    <updated>2026-10-12T23:40:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="data" label="データエンジン（収集 / 精査 / スケーリング）" />
    <category term="M" label="[M]" />
    <summary type="text">实景比拼赋能科创，助力人形机器人技术规模化落地！ 编辑丨李希 为深入贯彻落实工业和信息化部、国务院国资委《 2026年度人形机器人与具身智能实景实训专项行动》部署，加快推动我国机器人与具身智能技术创新突破与规模化落地，国际具身智能技能大赛检阅赛将于2026年10月23日至25日在上海举行。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: データエンジン（収集 / 精査 / スケーリング）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-12&lt;/p&gt;
&lt;blockquote&gt;实景比拼赋能科创，助力人形机器人技术规模化落地！ 编辑丨李希 为深入贯彻落实工业和信息化部、国务院国资委《 2026年度人形机器人与具身智能实景实训专项行动》部署，加快推动我国机器人与具身智能技术创新突破与规模化落地，国际具身智能技能大赛检阅赛将于2026年10月23日至25日在上海举行。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html"&gt;2026-10-13 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:9be94d2c68ce</id>
    <title>百度世界2026将于11月12日在北京举办，主题为「下一幕，智能组织」</title>
    <link rel="alternate" href="https://www.leiphone.com/category/industrynews/ukm5yjK3oT5X29yV.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html" />
    <published>2026-10-12T23:40:00Z</published>
    <updated>2026-10-12T23:40:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="M" label="[M]" />
    <summary type="text">10月12日，百度官方宣布，百度世界2026将于11月12日在北京·国家会议中心二期举办。目前，大会官网已开启售票通道（https://baiduworld2026.events.net.cn/tickets）。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-12&lt;/p&gt;
&lt;blockquote&gt;10月12日，百度官方宣布，百度世界2026将于11月12日在北京·国家会议中心二期举办。目前，大会官网已开启售票通道（https://baiduworld2026.events.net.cn/tickets）。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-13.ja.html"&gt;2026-10-13 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12368</id>
    <title>LiteNWM: Efficient Latent World Models for Onboard Visual Navigation in the Wild</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12368" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html" />
    <published>2026-10-11T23:40:00Z</published>
    <updated>2026-10-11T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12368</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Direct visual navigation policies generate trajectories efficiently but do not explicitly evaluate their future consequences.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12368 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 17.56% · 16.2% · 43.3% · 83.3%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Direct visual navigation policies generate trajectories efficiently but do not explicitly evaluate their future consequences.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html"&gt;2026-10-12 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.11480</id>
    <title>RoboAware: Learning to Coordinate Embodied Skills from Counterfactual Outcomes</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.11480" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html" />
    <published>2026-10-11T23:40:00Z</published>
    <updated>2026-10-11T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.11480</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Embodied coding agents can combine modular robot skills with frozen end-to-end policies, yet effective composition requires anticipating which policy family will succeed in the current physical state.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.11480 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 100 tasks · 77.0% · 90.0% · 73.8%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Embodied coding agents can combine modular robot skills with frozen end-to-end policies, yet effective composition requires anticipating which policy family will succeed in the current physical state.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html"&gt;2026-10-12 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12440</id>
    <title>Generative Neural Retargeting for Human-to-Robot Dexterous Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12440" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html" />
    <published>2026-10-11T23:40:00Z</published>
    <updated>2026-10-11T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12440</name>
    </author>
    <category term="data" label="データエンジン（収集 / 精査 / スケーリング）" />
    <category term="R" label="[R]" />
    <summary type="text">Human demonstrations are a scalable data source for learning dexterous manipulation, but the embodiment gap prevents human motion from being executed directly on robots.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: データエンジン（収集 / 精査 / スケーリング）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12440 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Human demonstrations are a scalable data source for learning dexterous manipulation, but the embodiment gap prevents human motion from being executed directly on robots.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html"&gt;2026-10-12 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12407</id>
    <title>LeWAM: A JEPA World Action Model with Diffusion-Steering-Based MPC</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12407" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html" />
    <published>2026-10-11T23:40:00Z</published>
    <updated>2026-10-11T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12407</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries noisy, redundant information which can complicate downstream predictions.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12407 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;World action models (WAMs) predict actions and future observations, typically from a reconstruction-based representation that carries noisy, redundant information which can complicate downstream predictions.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html"&gt;2026-10-12 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12457</id>
    <title>SpatialHarness: Test-Time Spatial Scaffolding for Fine Robotic Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12457" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html" />
    <published>2026-10-11T23:40:00Z</published>
    <updated>2026-10-11T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12457</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="R" label="[R]" />
    <summary type="text">Frontier multimodal foundation models (e.g., GPT-6 Astra) have recently shown strong potential for direct robotic control, yet their performance on fine manipulation remains limited.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12457 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 26.7% · 66.7% · 0% · 100%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Frontier multimodal foundation models (e.g., GPT-6 Astra) have recently shown strong potential for direct robotic control, yet their performance on fine manipulation remains limited.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html"&gt;2026-10-12 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12249</id>
    <title>Real-Time Motion Planning with Dynamic Hazards: Classical vs. Learning-Based Methods</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12249" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html" />
    <published>2026-10-11T23:40:00Z</published>
    <updated>2026-10-11T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12249</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">We study real-time motion planning in dynamic hazard fields through a controlled comparison between classical planning and learning-based methods.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12249 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;We study real-time motion planning in dynamic hazard fields through a controlled comparison between classical planning and learning-based methods.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html"&gt;2026-10-12 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.11809</id>
    <title>WAND: Learning Robust Navigation under Complex Wind Disturbances and Dense Obstacles for Quadrotors</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.11809" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html" />
    <published>2026-10-11T23:40:00Z</published>
    <updated>2026-10-11T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.11809</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Robust navigation in cluttered environments remains a fundamental challenge for quadrotors, particularly when strong wind disturbances arise, which perturb vehicle dynamics, limit control authority, and substantially increase collision…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.11809 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Robust navigation in cluttered environments remains a fundamental challenge for quadrotors, particularly when strong wind disturbances arise, which perturb vehicle dynamics, limit control authority, and substantially increase collision…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html"&gt;2026-10-12 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12467</id>
    <title>CSF: Contextual Safety Filtering for Motion Generators</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12467" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html" />
    <published>2026-10-11T23:40:00Z</published>
    <updated>2026-10-11T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12467</name>
    </author>
    <category term="safety" label="安全・権限・コンプライアンス" />
    <category term="R" label="[R]" />
    <summary type="text">Text-conditioned motion generators produce trackable whole-body motion, but they have no notion of scene-dependent safety: the same action may target an object or a person.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 安全・権限・コンプライアンス&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12467 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 90% · 100%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Text-conditioned motion generators produce trackable whole-body motion, but they have no notion of scene-dependent safety: the same action may target an object or a person.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-12.ja.html"&gt;2026-10-12 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12185</id>
    <title>RESETTLE: Robotic Recovery through Disagreement-Triggered Retrieval and Efficient Corrective Control</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12185" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html" />
    <published>2026-10-10T23:40:00Z</published>
    <updated>2026-10-10T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12185</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Reliable robotic manipulation requires timely intervention to correct emerging deviations and restore progress after execution errors.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12185 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 8.70% · 6.28% · 6.83% · 74.04%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり · コードまたは重み公開 · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Reliable robotic manipulation requires timely intervention to correct emerging deviations and restore progress after execution errors.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html"&gt;2026-10-11 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.11283</id>
    <title>Being-M0.7: A Latent World-Action Model for Humanoid Robots</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.11283" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html" />
    <published>2026-10-10T23:40:00Z</published>
    <updated>2026-10-10T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.11283</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="R" label="[R]" />
    <summary type="text">Humanoid loco-manipulation requires coordinated locomotion and manipulation informed by future scene evolution and whole-body motion, yet learning these capabilities is constrained by scarce robot demonstrations.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.11283 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 10,000 hours&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Humanoid loco-manipulation requires coordinated locomotion and manipulation informed by future scene evolution and whole-body motion, yet learning these capabilities is constrained by scarce robot demonstrations.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html"&gt;2026-10-11 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12231</id>
    <title>Residual Modeling Closes the Regression and Generative Policy Gap in Robot Learning</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12231" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html" />
    <published>2026-10-10T23:40:00Z</published>
    <updated>2026-10-10T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12231</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Learning from demonstration has enabled impressive robot behaviors. A common choice for policy learning is to use diffusion or flow matching (Flow-Policies), which often outperforms direct action regression trained with mean squared error…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12231 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Learning from demonstration has enabled impressive robot behaviors. A common choice for policy learning is to use diffusion or flow matching (Flow-Policies), which often outperforms direct action regression trained with mean squared error…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html"&gt;2026-10-11 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12386</id>
    <title>ARC: A Reasoning Recipe for Robot Foundation Models</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12386" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html" />
    <published>2026-10-10T23:40:00Z</published>
    <updated>2026-10-10T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12386</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">The prevailing approach to improving robot foundation models (RFMs) relies on larger models, more robot demonstrations, and costly training at scale.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12386 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;The prevailing approach to improving robot foundation models (RFMs) relies on larger models, more robot demonstrations, and costly training at scale.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html"&gt;2026-10-11 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12046</id>
    <title>SkillWeave: Weaving Heterogeneous Demonstrations into Long-Horizon Manipulation Skills</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12046" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html" />
    <published>2026-10-10T23:40:00Z</published>
    <updated>2026-10-10T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12046</name>
    </author>
    <category term="data" label="データエンジン（収集 / 精査 / スケーリング）" />
    <category term="R" label="[R]" />
    <summary type="text">Dexterous manipulation requires both large-scale task progression and precise contact-rich interaction, making it challenging to collect demonstrations that effectively support both regimes.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: データエンジン（収集 / 精査 / スケーリング）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12046 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 27% · 65% · 87%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Dexterous manipulation requires both large-scale task progression and precise contact-rich interaction, making it challenging to collect demonstrations that effectively support both regimes.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html"&gt;2026-10-11 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12432</id>
    <title>FAITH: Feasibility-Aware Safety-Filtered RL for High-Dimensional Systems</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12432" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html" />
    <published>2026-10-10T23:40:00Z</published>
    <updated>2026-10-10T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.12432</name>
    </author>
    <category term="safety" label="安全・権限・コンプライアンス" />
    <category term="R" label="[R]" />
    <summary type="text">Safe reinforcement learning commonly places safety and task performance in the same policy objective, where they can introduce competing updates.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 安全・権限・コンプライアンス&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12432 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 99.95% · 97%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Safe reinforcement learning commonly places safety and task performance in the same policy objective, where they can introduce competing updates.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html"&gt;2026-10-11 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.11416</id>
    <title>Rewiring Semantics, Dynamics, and Control: A Simple yet Effective Action-Centric Tri-Stream Transformer</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.11416" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html" />
    <published>2026-10-10T23:40:00Z</published>
    <updated>2026-10-10T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.11416</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-Language-Action (VLA) models have emerged as a prominent framework for complex robotic manipulation, building on the strong semantic understanding of pretrained Vision-Language Models (VLMs).</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.11416 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Vision-Language-Action (VLA) models have emerged as a prominent framework for complex robotic manipulation, building on the strong semantic understanding of pretrained Vision-Language Models (VLMs).&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html"&gt;2026-10-11 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:0698575e496b</id>
    <title>Why humanoid robot demos still fail the generalization test</title>
    <link rel="alternate" href="https://www.therobotreport.com/why-humanoid-robot-demos-still-fail-the-generalization-test/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html" />
    <published>2026-10-10T23:40:00Z</published>
    <updated>2026-10-10T23:40:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">How better data and human judgment can help robots move beyond demos toward reliable real-world performance.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-09&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;How better data and human judgment can help robots move beyond demos toward reliable real-world performance.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-11.ja.html"&gt;2026-10-11 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.09254</id>
    <title>RoboRender: Robot-Oriented Video Generation for Visual Sim-to-Real Transfer</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.09254" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html" />
    <published>2026-10-10T07:15:00Z</published>
    <updated>2026-10-10T07:15:00Z</updated>
    <author>
      <name>arXiv 2610.09254</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Simulation enables large-scale, low-cost robot data generation, but policies trained in simulation often fail to transfer to the real world due to the sim-to-real visual discrepancies.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.09254 · 2026-10-07&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 71% · 7.1x · 3.6x&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Simulation enables large-scale, low-cost robot data generation, but policies trained in simulation often fail to transfer to the real world due to the sim-to-real visual discrepancies.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html"&gt;2026-10-10 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.11322</id>
    <title>USDCraft: Geometrically Grounded Programmatic Modeling of Articulated 3D Assets for Simulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.11322" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html" />
    <published>2026-10-10T07:15:00Z</published>
    <updated>2026-10-10T07:15:00Z</updated>
    <author>
      <name>arXiv 2610.11322</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Geometrically faithful and functional articulated 3D assets are essential for real-to-sim robot manipulation, where policies trained in simulation must transfer to physical objects.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.11322 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Geometrically faithful and functional articulated 3D assets are essential for real-to-sim robot manipulation, where policies trained in simulation must transfer to physical objects.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html"&gt;2026-10-10 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12026</id>
    <title>Humanoid World Action Model With Joint State--Action Generation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12026" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html" />
    <published>2026-10-10T07:15:00Z</published>
    <updated>2026-10-10T07:15:00Z</updated>
    <author>
      <name>arXiv 2610.12026</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Humanoid robots are a promising platform for general-purpose manipulation. Recent Vision-Language-Action (VLA) policies learn actions directly from multimodal observations, while World Action Models (WAMs) further incorporate future…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12026 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 70.6% · 43.3%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Humanoid robots are a promising platform for general-purpose manipulation. Recent Vision-Language-Action (VLA) policies learn actions directly from multimodal observations, while World Action Models (WAMs) further incorporate future…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html"&gt;2026-10-10 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12089</id>
    <title>ManiUnit: A Manipulation Skill Dataset and Benchmark for Long-Horizon Tasks</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12089" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html" />
    <published>2026-10-10T07:15:00Z</published>
    <updated>2026-10-10T07:15:00Z</updated>
    <author>
      <name>arXiv 2610.12089</name>
    </author>
    <category term="systems" label="システムとオーケストレーション（長期タスク / 複数方策）" />
    <category term="R" label="[R]" />
    <summary type="text">Long-horizon mobile manipulation requires a robot to navigate multi-room environments and execute a sequence of manipulation skills under a single natural language instruction.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: システムとオーケストレーション（長期タスク / 複数方策）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12089 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 56% · 78.7% · 49.3% · 4.0%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Long-horizon mobile manipulation requires a robot to navigate multi-room environments and execute a sequence of manipulation skills under a single natural language instruction.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html"&gt;2026-10-10 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:ieee-spectrum-robotics:96aa397c5848</id>
    <title>Video Friday: Robot Decommissioning Takes a Fun Turn</title>
    <link rel="alternate" href="https://spectrum.ieee.org/video-friday-reachy-mini-raps" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html" />
    <published>2026-10-10T07:15:00Z</published>
    <updated>2026-10-10T07:15:00Z</updated>
    <author>
      <name>IEEE Spectrum Robotics</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">It’s possible that I wasn’t the only one who thought that Figure’s recent stunt was somewhat ridiculous. Devanthro is in fact selling these plastic balls for $1000 each, and they’ll arrive in time for Christmas.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: IEEE Spectrum Robotics · 2026-10-09&lt;/p&gt;
&lt;blockquote&gt;It’s possible that I wasn’t the only one who thought that Figure’s recent stunt was somewhat ridiculous. Devanthro is in fact selling these plastic balls for $1000 each, and they’ll arrive in time for Christmas.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html"&gt;2026-10-10 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:4522d1bc988b</id>
    <title>让触觉像视觉一样可规模化，哥大李昀烛团队 FlexiTac 的开源实践 | IROS 2026</title>
    <link rel="alternate" href="https://www.leiphone.com/category/robot/Pz34omTBnW9Y0LhO.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html" />
    <published>2026-10-10T07:15:00Z</published>
    <updated>2026-10-10T07:15:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">FlexiTac压阻式传感器可兼顾柔性、低成本与一致性。机器人的视觉能力正快速增强，但 触觉仍然是具身智能系统中的薄弱环节 。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-10&lt;/p&gt;
&lt;blockquote&gt;FlexiTac压阻式传感器可兼顾柔性、低成本与一致性。机器人的视觉能力正快速增强，但 触觉仍然是具身智能系统中的薄弱环节 。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html"&gt;2026-10-10 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:2fc129443dc7</id>
    <title>这届 IROS 2026 上，具身学术与产业边界正在消融</title>
    <link rel="alternate" href="https://www.leiphone.com/category/ai/4I8YKksz9V10Ih1R.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html" />
    <published>2026-10-10T07:15:00Z</published>
    <updated>2026-10-10T07:15:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="data" label="データエンジン（収集 / 精査 / スケーリング）" />
    <category term="M" label="[M]" />
    <summary type="text">模型、数据、灵巧操作都在工程化，顶会上具身传递出的信号变了9 月 27 日到 10 月 1 日，全球机器人领域最顶级的国际会议之一 IROS 2026 将在美国匹兹堡开幕。 如果你前两年也在关注这个会，你会发现那时候大家都在 “卷概念”。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: データエンジン（収集 / 精査 / スケーリング）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-10&lt;/p&gt;
&lt;blockquote&gt;模型、数据、灵巧操作都在工程化，顶会上具身传递出的信号变了9 月 27 日到 10 月 1 日，全球机器人领域最顶级的国际会议之一 IROS 2026 将在美国匹兹堡开幕。 如果你前两年也在关注这个会，你会发现那时候大家都在 “卷概念”。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html"&gt;2026-10-10 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:6fcc505a20cb</id>
    <title>再裁员5000人！保时捷明确回归燃油车，高端车型计划涨价20%；腾讯拟发行50亿美元离岸债券，或筹钱加码AI；OpenAI全面上线GPT-6</title>
    <link rel="alternate" href="https://www.leiphone.com/category/zaobao/GGmq1RiBgXY2GSVL.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html" />
    <published>2026-10-10T07:15:00Z</published>
    <updated>2026-10-10T07:15:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="M" label="[M]" />
    <summary type="text">要闻提示 再裁员5000人！保时捷明确回归燃油车，高端车型计划涨价20% 阿里巴巴蔡崇信：开源是欧洲发展 AI 的最佳选择 腾讯拟发行50亿美元离岸债券，或筹钱加码AI 超33亿元！</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-09&lt;/p&gt;
&lt;blockquote&gt;要闻提示 再裁员5000人！保时捷明确回归燃油车，高端车型计划涨价20% 阿里巴巴蔡崇信：开源是欧洲发展 AI 的最佳选择 腾讯拟发行50亿美元离岸债券，或筹钱加码AI 超33亿元！&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-10.ja.html"&gt;2026-10-10 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.11508</id>
    <title>WARP-VLA: Wrist-Camera Adaptation for View-Robust Policy Execution in Vision-Language-Action Models</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.11508" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html" />
    <published>2026-10-09T07:28:00Z</published>
    <updated>2026-10-09T07:28:00Z</updated>
    <author>
      <name>arXiv 2610.11508</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Despite recent advances in Vision-Language-Action models (VLAs) for robotic manipulation, their performance remains sensitive to changes in camera configuration.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.11508 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 39.2% · 78.3%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり · コードまたは重み公開&lt;/p&gt;
&lt;blockquote&gt;Despite recent advances in Vision-Language-Action models (VLAs) for robotic manipulation, their performance remains sensitive to changes in camera configuration.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html"&gt;2026-10-09 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12007</id>
    <title>REACT: Rolling Denoising and Dual Decoupling for Reactive Robot Control with VLA Models</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12007" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html" />
    <published>2026-10-09T07:28:00Z</published>
    <updated>2026-10-09T07:28:00Z</updated>
    <author>
      <name>arXiv 2610.12007</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Flow-based vision-language-action (VLA) models generate action chunks for temporally coherent robot motion, but chunked control creates a fundamental closed-loop trade-off: long chunks provide smooth execution, whereas frequent replanning…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12007 · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Flow-based vision-language-action (VLA) models generate action chunks for temporally coherent robot motion, but chunked control creates a fundamental closed-loop trade-off: long chunks provide smooth execution, whereas frequent replanning…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html"&gt;2026-10-09 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.12140</id>
    <title>Leveraging Human-In-The-Loop Demonstrations in Reinforcement Learning for Digital Twin-Driven Robot Flexibility</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.12140" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html" />
    <published>2026-10-09T07:28:00Z</published>
    <updated>2026-10-09T07:28:00Z</updated>
    <author>
      <name>arXiv 2610.12140</name>
    </author>
    <category term="data" label="データエンジン（収集 / 精査 / スケーリング）" />
    <category term="R" label="[R]" />
    <summary type="text">Growing automation makes collaborative robots work in more variable environments, increasing the need for adaptation.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: データエンジン（収集 / 精査 / スケーリング）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.12140 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 100% · 17%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Growing automation makes collaborative robots work in more variable environments, increasing the need for adaptation.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html"&gt;2026-10-09 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.11952</id>
    <title>Tell Robot What Not to Do: A Negation Understanding Perspective</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.11952" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html" />
    <published>2026-10-09T07:28:00Z</published>
    <updated>2026-10-09T07:28:00Z</updated>
    <author>
      <name>arXiv 2610.11952</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Instruction following enables robots to perform diverse tasks specified in natural language, making it a fundamental capability for human-robot interaction.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.11952 · 2026-10-08&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 2.60% · 88.45% · 12.4% · 88.8%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Instruction following enables robots to perform diverse tasks specified in natural language, making it a fundamental capability for human-robot interaction.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html"&gt;2026-10-09 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:ieee-spectrum-robotics:a2b57285998e</id>
    <title>This Disembodied Hand Is All the Robot You Need</title>
    <link rel="alternate" href="https://spectrum.ieee.org/walking-robotic-hand" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html" />
    <published>2026-10-09T07:28:00Z</published>
    <updated>2026-10-09T07:28:00Z</updated>
    <author>
      <name>IEEE Spectrum Robotics</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">A hand attached to the end of an arm is a given for most humans, but what if robots didn’t have to play by the same rules as we do?</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: IEEE Spectrum Robotics · 2026-10-08&lt;/p&gt;
&lt;blockquote&gt;A hand attached to the end of an arm is a given for most humans, but what if robots didn’t have to play by the same rules as we do?&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html"&gt;2026-10-09 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:bcb1ebbc55ff</id>
    <title>Jabil discusses the pace of humanoid robot development and production</title>
    <link rel="alternate" href="https://www.therobotreport.com/jabil-discusses-pace-humanoid-robot-development-production/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html" />
    <published>2026-10-09T07:28:00Z</published>
    <updated>2026-10-09T07:28:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">Jabil, which is working with robot developers such as Apptronik, provides a design perspective on humanoid progress.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-08&lt;/p&gt;
&lt;blockquote&gt;Jabil, which is working with robot developers such as Apptronik, provides a design perspective on humanoid progress.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html"&gt;2026-10-09 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:6b2f68612d16</id>
    <title>AWS launches open-source Physical AI Toolchain for robotics</title>
    <link rel="alternate" href="https://www.therobotreport.com/aws-launches-open-source-physical-ai-toolchain-for-robotics/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html" />
    <published>2026-10-09T07:28:00Z</published>
    <updated>2026-10-09T07:28:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="M" label="[M]" />
    <summary type="text">The toolchain combines AWS and NVIDIA tools for synthetic data generation, model training, simulation and validation, edge deployment, and continuous improvement.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-08&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: コードまたは重み公開&lt;/p&gt;
&lt;blockquote&gt;The toolchain combines AWS and NVIDIA tools for synthetic data generation, model training, simulation and validation, edge deployment, and continuous improvement.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html"&gt;2026-10-09 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:df0a0c0416fd</id>
    <title>独家 | 清华 AIR 首届博士李健雄创立本溯智能：五源资本领投，押注动作原生具身模型</title>
    <link rel="alternate" href="https://www.leiphone.com/category/ai/WbD0Tf6P1oA2WDdq.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html" />
    <published>2026-10-09T07:28:00Z</published>
    <updated>2026-10-09T07:28:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="M" label="[M]" />
    <summary type="text">常春藤、线性、华山资本跟投，AIR副教授詹仙园任首席科学家。雷峰网·鲸犀独家获悉，清华大学智能产业研究院（AIR）首届博士生李健雄已毕业，并创立具身智能公司 本溯智能 （BASAL Intelligence，简称 BASAL）。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-08&lt;/p&gt;
&lt;blockquote&gt;常春藤、线性、华山资本跟投，AIR副教授詹仙园任首席科学家。雷峰网·鲸犀独家获悉，清华大学智能产业研究院（AIR）首届博士生李健雄已毕业，并创立具身智能公司 本溯智能 （BASAL Intelligence，简称 BASAL）。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-09.ja.html"&gt;2026-10-09 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.10479</id>
    <title>Agentic RSR: Real-to-Sim-to-Real through Scene Reconstruction and Execution-Grounded Robot Policies</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.10479" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html" />
    <published>2026-10-08T07:30:00Z</published>
    <updated>2026-10-08T07:30:00Z</updated>
    <author>
      <name>arXiv 2610.10479</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">A simulation of a real robot workspace must preserve task-relevant interactions, while policies developed in it must operate on observations available to the real robot.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.10479 · 2026-10-07&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 80%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;A simulation of a real robot workspace must preserve task-relevant interactions, while policies developed in it must operate on observations available to the real robot.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html"&gt;2026-10-08 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.10079</id>
    <title>RealtimeWAM: How Fast Can I Run My World Action Model?</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.10079" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html" />
    <published>2026-10-08T07:30:00Z</published>
    <updated>2026-10-08T07:30:00Z</updated>
    <author>
      <name>arXiv 2610.10079</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">World Action Models (WAMs) combine visual dynamics modeling with action generation, but their high inference latency limits responsive robot control.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.10079 · 2026-10-07&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 63.09 ms · 82.75% · 87.41%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;World Action Models (WAMs) combine visual dynamics modeling with action generation, but their high inference latency limits responsive robot control.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html"&gt;2026-10-08 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.10453</id>
    <title>RFPO: Rectified Flow Policy Optimization for Embodied Control</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.10453" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html" />
    <published>2026-10-08T07:30:00Z</published>
    <updated>2026-10-08T07:30:00Z</updated>
    <author>
      <name>arXiv 2610.10453</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Flow-based policies provide an expressive framework for continuous robot control, but their iterative ODE integration incurs substantial inference cost.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.10453 · 2026-10-07&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 2.4% · 98.5% · 4.39 ms · 0.08 ms&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり · コードまたは重み公開 · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Flow-based policies provide an expressive framework for continuous robot control, but their iterative ODE integration incurs substantial inference cost.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html"&gt;2026-10-08 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.08784</id>
    <title>PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.08784" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html" />
    <published>2026-10-08T07:30:00Z</published>
    <updated>2026-10-08T07:30:00Z</updated>
    <author>
      <name>arXiv 2610.08784</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="R" label="[R]" />
    <summary type="text">Pretrained robotic policies can suffer substantial performance degradation under out-of-distribution (OOD) conditions encountered during deployment, motivating post-training through real-world interaction.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.08784 · 2026-10-06&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 53.2% · 95% · 90%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Pretrained robotic policies can suffer substantial performance degradation under out-of-distribution (OOD) conditions encountered during deployment, motivating post-training through real-world interaction.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html"&gt;2026-10-08 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:4b07be24d804</id>
    <title>研究了1933篇 IROS 论文，我们看到了机器人学的六项新变化</title>
    <link rel="alternate" href="https://www.leiphone.com/category/private/wEpCJ4T8v9cJFgCn.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html" />
    <published>2026-10-08T07:30:00Z</published>
    <updated>2026-10-08T07:30:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">大模型并没有“吃掉”机器人学。9月27日，IROS 2026 将在匹兹堡开幕。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-08&lt;/p&gt;
&lt;blockquote&gt;大模型并没有“吃掉”机器人学。9月27日，IROS 2026 将在匹兹堡开幕。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html"&gt;2026-10-08 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:f05eb6e1da2e</id>
    <title>对话如祺出行COO韩锋：世界模型越强，真实数据才是真壁垒</title>
    <link rel="alternate" href="https://www.leiphone.com/category/transportation/ESCXNijR4DVx6uDl.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html" />
    <published>2026-10-08T07:30:00Z</published>
    <updated>2026-10-08T07:30:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="data" label="データエンジン（収集 / 精査 / スケーリング）" />
    <category term="M" label="[M]" />
    <summary type="text">过去一年，“物理AI”几乎成了自动驾驶、车企共同争夺的下一张船票。 世界模型、具身大模型、机器人数据……众多公司从不同切入口布局物理AI，试图成为新风口之下的主角，相比之下，如祺出行选择了一条不那么显眼的路径，不自己训练大模型，而是尝试成为模型背后的数据基础设施。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: データエンジン（収集 / 精査 / スケーリング）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-08&lt;/p&gt;
&lt;blockquote&gt;过去一年，“物理AI”几乎成了自动驾驶、车企共同争夺的下一张船票。 世界模型、具身大模型、机器人数据……众多公司从不同切入口布局物理AI，试图成为新风口之下的主角，相比之下，如祺出行选择了一条不那么显眼的路径，不自己训练大模型，而是尝试成为模型背后的数据基础设施。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html"&gt;2026-10-08 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:leiphone:bee8b4023118</id>
    <title>IROS 2026大奖揭晓：长期记忆拿下最佳论文，人形机器人打网球、端托盘双双获奖</title>
    <link rel="alternate" href="https://www.leiphone.com/category/private/VYRZFCnadDOcPDeX.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html" />
    <published>2026-10-08T07:30:00Z</published>
    <updated>2026-10-08T07:30:00Z</updated>
    <author>
      <name>雷峰网 Leiphone</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="M" label="[M]" />
    <summary type="text">在机器人领域，一个更大的模型并不能自动解决所有问题。当地时间 9 月 30 日，IROS 2026 主会在美国匹兹堡落下帷幕，本届大会的主要奖项也在最后一天正式揭晓。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: 雷峰网 Leiphone · 2026-10-08&lt;/p&gt;
&lt;blockquote&gt;在机器人领域，一个更大的模型并不能自动解决所有问题。当地时间 9 月 30 日，IROS 2026 主会在美国匹兹堡落下帷幕，本届大会的主要奖项也在最后一天正式揭晓。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html"&gt;2026-10-08 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:42327a98fe3f</id>
    <title>Harmonic Drive to share how integrated actuators can reduce engineering at RoboBusiness</title>
    <link rel="alternate" href="https://www.therobotreport.com/harmonic-drive-share-how-integrated-actuators-reduce-engineering-robobusiness/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html" />
    <published>2026-10-08T07:30:00Z</published>
    <updated>2026-10-08T07:30:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">Attendees will learn how integrated actuators can reduce engineering effort by simplifying system integration.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-07&lt;/p&gt;
&lt;blockquote&gt;Attendees will learn how integrated actuators can reduce engineering effort by simplifying system integration.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-08.ja.html"&gt;2026-10-08 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.08618</id>
    <title>Demo: Closed-Loop Sionna-Isaac Sim Co-Simulation Framework for Wireless-Aware Robot Navigation over ROS 2</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.08618" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html" />
    <published>2026-10-07T07:19:00Z</published>
    <updated>2026-10-07T07:19:00Z</updated>
    <author>
      <name>arXiv 2610.08618</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">A robot that offloads its control loop to the network carries the receiver with it, so link quality is decided by where it goes.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.08618 · 2026-10-06&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 16ms · 60 Hz · 7.4%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;A robot that offloads its control loop to the network carries the receiver with it, so link quality is decided by where it goes.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html"&gt;2026-10-07 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.07756</id>
    <title>StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.07756" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html" />
    <published>2026-10-07T07:19:00Z</published>
    <updated>2026-10-07T07:19:00Z</updated>
    <author>
      <name>arXiv 2610.07756</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language-action (VLA) models increasingly rely on diffusion- or flow-matching-based action heads to generate continuous robot actions.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.07756 · 2026-10-06&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 96.5% · 97.8% · 115.0 ms · 44.2 ms&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Vision-language-action (VLA) models increasingly rely on diffusion- or flow-matching-based action heads to generate continuous robot actions.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html"&gt;2026-10-07 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.07696</id>
    <title>ESP: Energy-Score Policy for One-Step Multimodal Action Generation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.07696" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html" />
    <published>2026-10-07T07:19:00Z</published>
    <updated>2026-10-07T07:19:00Z</updated>
    <author>
      <name>arXiv 2610.07696</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Generative action models based on diffusion and flow matching have been increasingly adopted in vision-language-action (VLA) policies for their ability to capture diverse behaviors, including multiple valid action sequences under the same…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.07696 · 2026-10-06&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Generative action models based on diffusion and flow matching have been increasingly adopted in vision-language-action (VLA) policies for their ability to capture diverse behaviors, including multiple valid action sequences under the same…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html"&gt;2026-10-07 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.07946</id>
    <title>Adapting Vision-Language-Action Models to Unknown Visual Disruptions During Execution</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.07946" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html" />
    <published>2026-10-07T07:19:00Z</published>
    <updated>2026-10-07T07:19:00Z</updated>
    <author>
      <name>arXiv 2610.07946</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Visual disruptions can arise while a robot is executing a task, leaving a vision-language-action (VLA) policy to respond without knowing the disruption type or timing.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.07946 · 2026-10-06&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 43.9% · 53.2% · 58.7% · 66.0%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Visual disruptions can arise while a robot is executing a task, leaving a vision-language-action (VLA) policy to respond without knowing the disruption type or timing.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html"&gt;2026-10-07 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.05024</id>
    <title>LightVLN: Efficient Aerial Vision-and-Language Navigation with Compact Memory and History-Guided Local Aggregation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.05024" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html" />
    <published>2026-10-07T07:19:00Z</published>
    <updated>2026-10-07T07:19:00Z</updated>
    <author>
      <name>arXiv 2610.05024</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Aerial vision-and-language navigation (VLN) enables unmanned aerial vehicles to execute long-horizon natural-language instructions from visual observations in complex three-dimensional environments.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.05024 · 2026-10-04&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 50.93% · 36.14% · 25.83% · 16 GB&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Aerial vision-and-language navigation (VLN) enables unmanned aerial vehicles to execute long-horizon natural-language instructions from visual observations in complex three-dimensional environments.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html"&gt;2026-10-07 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.08444</id>
    <title>ActTune: Action-Aware Precision and GPU Operating-Point Adaptation for Energy-Efficient Vision-Language-Action Inference</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.08444" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html" />
    <published>2026-10-07T07:19:00Z</published>
    <updated>2026-10-07T07:19:00Z</updated>
    <author>
      <name>arXiv 2610.08444</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language-action (VLA) policies repeatedly invoke inference to control robots, making graphics processing unit (GPU) energy a recurring cost of task execution.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.08444 · 2026-10-06&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Vision-language-action (VLA) policies repeatedly invoke inference to control robots, making graphics processing unit (GPU) energy a recurring cost of task execution.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html"&gt;2026-10-07 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:monoist:41b504f26025</id>
    <title>ルネサスが低耐圧GaNデバイス市場に参入、業界トップクラスの性能指数を実現</title>
    <link rel="alternate" href="https://monoist.itmedia.co.jp/mn/articles/2610/07/news043.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html" />
    <published>2026-10-07T07:19:00Z</published>
    <updated>2026-10-07T07:19:00Z</updated>
    <author>
      <name>MONOist (ITmedia)</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">ルネサス エレクトロニクスは、同社初となる100V耐圧のEモードGaNパワートランジスタ製品群を発売した。AIデータセンターやヒューマノイドロボットなど向けに、超高速スイッチングと優れた放熱性能を実現する。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: MONOist (ITmedia) · 2026-10-07&lt;/p&gt;
&lt;blockquote&gt;ルネサス エレクトロニクスは、同社初となる100V耐圧のEモードGaNパワートランジスタ製品群を発売した。AIデータセンターやヒューマノイドロボットなど向けに、超高速スイッチングと優れた放熱性能を実現する。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html"&gt;2026-10-07 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:04735ae4ac05</id>
    <title>Boston Dynamics appoints former Amazon exec Rohit Prasad as new CEO</title>
    <link rel="alternate" href="https://www.therobotreport.com/boston-dynamics-appoints-former-amazon-executive-rohit-prasad-new-ceo/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html" />
    <published>2026-10-07T07:19:00Z</published>
    <updated>2026-10-07T07:19:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">Before Boston Dynamics, Prasad served as senior vice president and head scientist for Alexa and artificial general intelligence at Amazon.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-06&lt;/p&gt;
&lt;blockquote&gt;Before Boston Dynamics, Prasad served as senior vice president and head scientist for Alexa and artificial general intelligence at Amazon.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-07.ja.html"&gt;2026-10-07 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.02832</id>
    <title>FastOPD: On-Policy Distillation for Lightweight VLA Deployment</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.02832" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html" />
    <published>2026-10-05T23:40:00Z</published>
    <updated>2026-10-05T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.02832</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-Language-Action (VLA) foundation models have scaled rapidly to enhance manipulation performance and generalizability, but this scaling incurs high computational costs that render real-world deployment increasingly challenging.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.02832 · 2026-10-02&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 84% · 78.1%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Vision-Language-Action (VLA) foundation models have scaled rapidly to enhance manipulation performance and generalizability, but this scaling incurs high computational costs that render real-world deployment increasingly challenging.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html"&gt;2026-10-06 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.03620</id>
    <title>UniIntervene++: An Adaptive Intervention Agent for Efficient Real-World Reinforcement Learning</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.03620" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html" />
    <published>2026-10-05T23:40:00Z</published>
    <updated>2026-10-05T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.03620</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="R" label="[R]" />
    <summary type="text">Online reinforcement learning (RL) enables robot policies to improve through physical interaction, but the assistance they require changes as their competence evolves.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.03620 · 2026-10-02&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 89.67% · 0.77% · 94.6%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり · コードまたは重み公開&lt;/p&gt;
&lt;blockquote&gt;Online reinforcement learning (RL) enables robot policies to improve through physical interaction, but the assistance they require changes as their competence evolves.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html"&gt;2026-10-06 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.03476</id>
    <title>MobiAgent: Dual-Loop Recursive Policy Self-Improvement for Long-Horizon Mobile Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.03476" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html" />
    <published>2026-10-05T23:40:00Z</published>
    <updated>2026-10-05T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.03476</name>
    </author>
    <category term="systems" label="システムとオーケストレーション（長期タスク / 複数方策）" />
    <category term="R" label="[R]" />
    <summary type="text">Long-horizon mobile manipulation presents significant challenges due to compounding execution errors and capacity interference between locomotion and arm control.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: システムとオーケストレーション（長期タスク / 複数方策）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.03476 · 2026-10-02&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 7.50% · 27.50% · 32.5% · 57.5%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Long-horizon mobile manipulation presents significant challenges due to compounding execution errors and capacity interference between locomotion and arm control.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html"&gt;2026-10-06 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.03607</id>
    <title>World Action Learning via Interaction-Centric Spectral Latent Guidance</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.03607" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html" />
    <published>2026-10-05T23:40:00Z</published>
    <updated>2026-10-05T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.03607</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Learning general-purpose robot policies requires large-scale real-world interaction data, yet collecting robot demonstrations remains expensive and difficult to scale.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.03607 · 2026-10-02&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 99.20% · 93.80% · 57.7%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Learning general-purpose robot policies requires large-scale real-world interaction data, yet collecting robot demonstrations remains expensive and difficult to scale.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html"&gt;2026-10-06 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.02784</id>
    <title>SimpleTouch: Can Vision-Language-Action Models Master Contact-Rich Manipulation Without Tactile Policy Pretraining?</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.02784" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html" />
    <published>2026-10-05T23:40:00Z</published>
    <updated>2026-10-05T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.02784</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Tactile sensing provides essential contact information for robotic manipulation, yet incorporating it into pretrained vision-language-action (VLA) models remains challenging.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.02784 · 2026-10-02&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 50 demonstrations · 77.5% · 45.2% · 66.7%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Tactile sensing provides essential contact information for robotic manipulation, yet incorporating it into pretrained vision-language-action (VLA) models remains challenging.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html"&gt;2026-10-06 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.03283</id>
    <title>HexVIO: Towards All-Day Stereo-Inertial Tracking Through Commodity DSPs</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.03283" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html" />
    <published>2026-10-05T23:40:00Z</published>
    <updated>2026-10-05T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.03283</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">The ability of a device to localize itself within its surroundings is a fundamental prerequisite for spatial computing.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.03283 · 2026-10-02&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 67% · 86% · 30 fps · 0.83 W&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;The ability of a device to localize itself within its surroundings is a fundamental prerequisite for spatial computing.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html"&gt;2026-10-06 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:077068b67d2b</id>
    <title>FCC robot restrictions could accelerate shift to local AI</title>
    <link rel="alternate" href="https://www.therobotreport.com/fcc-robot-restrictions-could-accelerate-shift-to-local-ai/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html" />
    <published>2026-10-05T23:40:00Z</published>
    <updated>2026-10-05T23:40:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="safety" label="安全・権限・コンプライアンス" />
    <category term="M" label="[M]" />
    <summary type="text">Companies deploying physical AI today need to decide which AI workloads belong on the robot, what can remain in the cloud, and whether the underlying technology stack can meet rising expectations for security, safety, and supply-chain…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 安全・権限・コンプライアンス&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-05&lt;/p&gt;
&lt;blockquote&gt;Companies deploying physical AI today need to decide which AI workloads belong on the robot, what can remain in the cloud, and whether the underlying technology stack can meet rising expectations for security, safety, and supply-chain…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html"&gt;2026-10-06 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:monoist:7aaa16ae267f</id>
    <title>川崎重工とNVIDIAが目指す「次世代デジタルシップヤード構想」とは何か</title>
    <link rel="alternate" href="https://monoist.itmedia.co.jp/mn/articles/2610/06/news009.html" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html" />
    <published>2026-10-05T23:40:00Z</published>
    <updated>2026-10-05T23:40:00Z</updated>
    <author>
      <name>MONOist (ITmedia)</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="M" label="[M]" />
    <summary type="text">川崎重工は、NVIDIAのフィジカルAIとデジタルツイン技術を活用し、「次世代デジタルシップヤード」の実現を目指すプロジェクトを開始した。海外で進むデジタルシップヤード構想と比べて、川崎重工とNVIDIAの取り組みにはどのような違いがあるのだろうのか。</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: MONOist (ITmedia) · 2026-10-05&lt;/p&gt;
&lt;blockquote&gt;川崎重工は、NVIDIAのフィジカルAIとデジタルツイン技術を活用し、「次世代デジタルシップヤード」の実現を目指すプロジェクトを開始した。海外で進むデジタルシップヤード構想と比べて、川崎重工とNVIDIAの取り組みにはどのような違いがあるのだろうのか。&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-06.ja.html"&gt;2026-10-06 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.01351</id>
    <title>Is Success All You Need? Investigating the Impact of Input Perturbations on VLA Behaviour in Tabletop Manipulation Tasks</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.01351" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html" />
    <published>2026-10-04T23:40:00Z</published>
    <updated>2026-10-04T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.01351</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-Language-Action (VLA) models have achieved high task success rates on robot manipulation task benchmarks.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.01351 · 2026-10-01&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Vision-Language-Action (VLA) models have achieved high task success rates on robot manipulation task benchmarks.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html"&gt;2026-10-05 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.01083</id>
    <title>WBAG: A Whole-Body and Attached-Geometry Safety Framework for Vision-Language-Action Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.01083" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html" />
    <published>2026-10-04T23:40:00Z</published>
    <updated>2026-10-04T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.01083</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language-action (VLA) policies have demonstrated impressive capabilities in generalizable robotic manipulation, but their deployment in the real world remains challenging due to potential collisions involving different parts of the…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.01083 · 2026-10-01&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Vision-language-action (VLA) policies have demonstrated impressive capabilities in generalizable robotic manipulation, but their deployment in the real world remains challenging due to potential collisions involving different parts of the…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html"&gt;2026-10-05 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.00981</id>
    <title>NarrativeFlow: Flow-Based Vision-Language-Action Model Using Robot Velocity Fields</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.00981" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html" />
    <published>2026-10-04T23:40:00Z</published>
    <updated>2026-10-04T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.00981</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">We focus on language-conditioned flow-based manipulation, where robot flows (robot velocity fields) serve as embodiment-agnostic, motion-centric representations for leveraging data collected from multiple robot platforms.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.00981 · 2026-10-01&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;We focus on language-conditioned flow-based manipulation, where robot flows (robot velocity fields) serve as embodiment-agnostic, motion-centric representations for leveraging data collected from multiple robot platforms.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html"&gt;2026-10-05 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.00982</id>
    <title>Divide-and-Remember: Recursive Action-Relevant Memory for Long-Horizon VLA Policies</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.00982" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html" />
    <published>2026-10-04T23:40:00Z</published>
    <updated>2026-10-04T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.00982</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language-action (VLA) models struggle on history-dependent manipulation tasks, where the current observation alone does not determine the action, and the policy needs a memory of the history.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.00982 · 2026-10-01&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · コードまたは重み公開&lt;/p&gt;
&lt;blockquote&gt;Vision-language-action (VLA) models struggle on history-dependent manipulation tasks, where the current observation alone does not determine the action, and the policy needs a memory of the history.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html"&gt;2026-10-05 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.01260</id>
    <title>PROMO: Preference-conditioned Multi-Objective Reinforcement Learning for Quadrupedal Robots</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.01260" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html" />
    <published>2026-10-04T23:40:00Z</published>
    <updated>2026-10-04T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.01260</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="R" label="[R]" />
    <summary type="text">Quadrupedal locomotion requires balancing conflicting objectives such as command tracking, stability, and energy efficiency, yet conventional reinforcement learning (RL) hardcodes these priorities into a fixed scalar reward at training…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.01260 · 2026-10-01&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 30.4% · 38.7% · 59.0%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · コードまたは重み公開&lt;/p&gt;
&lt;blockquote&gt;Quadrupedal locomotion requires balancing conflicting objectives such as command tracking, stability, and energy efficiency, yet conventional reinforcement learning (RL) hardcodes these priorities into a fixed scalar reward at training…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html"&gt;2026-10-05 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.00899</id>
    <title>TOAST: Stochastic Robot Action Tokenization for Autoregressive Vision-Language-Action Models</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.00899" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html" />
    <published>2026-10-04T23:40:00Z</published>
    <updated>2026-10-04T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.00899</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Autoregressive Vision-Language-Action models often represent continuous robot actions as discrete token sequences, enabling action prediction with standard next-token objectives.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.00899 · 2026-10-01&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Autoregressive Vision-Language-Action models often represent continuous robot actions as discrete token sequences, enabling action prediction with standard next-token objectives.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html"&gt;2026-10-05 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.02054</id>
    <title>UniWAM: Unified World-Action Model</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.02054" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html" />
    <published>2026-10-04T23:40:00Z</published>
    <updated>2026-10-04T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.02054</name>
    </author>
    <category term="data" label="データエンジン（収集 / 精査 / スケーリング）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language-action models benefit from the understanding and reasoning capabilities of pretrained vision-language models, but action-only supervision provides limited grounding in world dynamics.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: データエンジン（収集 / 精査 / スケーリング）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.02054 · 2026-10-01&lt;/p&gt;
&lt;blockquote&gt;Vision-language-action models benefit from the understanding and reasoning capabilities of pretrained vision-language models, but action-only supervision provides limited grounding in world dynamics.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html"&gt;2026-10-05 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.02196</id>
    <title>InterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.02196" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html" />
    <published>2026-10-04T23:40:00Z</published>
    <updated>2026-10-04T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.02196</name>
    </author>
    <category term="systems" label="システムとオーケストレーション（長期タスク / 複数方策）" />
    <category term="R" label="[R]" />
    <summary type="text">We study test-time evolution for humanoid loco-manipulation: solving tasks that a controller was never trained for by repurposing its existing skills, improving from its own attempts, and retaining what it learns, without retraining.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: システムとオーケストレーション（長期タスク / 複数方策）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.02196 · 2026-10-01&lt;/p&gt;
&lt;blockquote&gt;We study test-time evolution for humanoid loco-manipulation: solving tasks that a controller was never trained for by repurposing its existing skills, improving from its own attempts, and retaining what it learns, without retraining.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-05.ja.html"&gt;2026-10-05 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.01612</id>
    <title>ReCo: Response-Consistent Locomotion with Policy-Aware MPC for Legged Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.01612" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html" />
    <published>2026-10-03T23:40:00Z</published>
    <updated>2026-10-03T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.01612</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Continuous legged manipulation requires accurate end-effector tracking while the base keeps walking. Combining reinforcement learning (RL) with model predictive control (MPC) suits this task: the learned policy provides robust locomotion,…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.01612 · 2026-10-01&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 28.7% · 27.4%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Continuous legged manipulation requires accurate end-effector tracking while the base keeps walking. Combining reinforcement learning (RL) with model predictive control (MPC) suits this task: the learned policy provides robust locomotion,…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html"&gt;2026-10-04 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.02161</id>
    <title>DuoMind: Enabling Distributed Multi-Robot Coordination with Semantic Communication</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.02161" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html" />
    <published>2026-10-03T23:40:00Z</published>
    <updated>2026-10-03T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.02161</name>
    </author>
    <category term="systems" label="システムとオーケストレーション（長期タスク / 複数方策）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language models (VLMs) and vision-language-action models (VLAs) have recently driven rapid progress in general-purpose robots, yet most progress has focused on single-robot settings.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: システムとオーケストレーション（長期タスク / 複数方策）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.02161 · 2026-10-01&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Vision-language models (VLMs) and vision-language-action models (VLAs) have recently driven rapid progress in general-purpose robots, yet most progress has focused on single-robot settings.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html"&gt;2026-10-04 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.01856</id>
    <title>ChunkVLA-AM: Parallel Action Chunking for Vision-Language-Action Robot Control in Additive Manufacturing</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.01856" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html" />
    <published>2026-10-03T23:40:00Z</published>
    <updated>2026-10-03T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.01856</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language-action (VLA) models unify visual perception, language understanding, and action generation, offering new opportunities for automation in additive manufacturing (AM).</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.01856 · 2026-10-01&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 92.9%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Vision-language-action (VLA) models unify visual perception, language understanding, and action generation, offering new opportunities for automation in additive manufacturing (AM).&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html"&gt;2026-10-04 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.39403</id>
    <title>IronMind: Scaling Humanoid Dexterous Manipulation via Camera-Space Ego-Centric Pretraining</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.39403" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html" />
    <published>2026-10-03T23:40:00Z</published>
    <updated>2026-10-03T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.39403</name>
    </author>
    <category term="data" label="データエンジン（収集 / 精査 / スケーリング）" />
    <category term="R" label="[R]" />
    <summary type="text">Egocentric human video offers a scalable data source for dexterous manipulation, yet using it to train humanoid robots presents two challenges: (1) an embodiment gap, as human hands differ structurally from robot end-effectors and…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: データエンジン（収集 / 精査 / スケーリング）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.39403 · 2026-09-30&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 10,000 hours · 55.0% · 11.7% · 5,000 hours&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり&lt;/p&gt;
&lt;blockquote&gt;Egocentric human video offers a scalable data source for dexterous manipulation, yet using it to train humanoid robots presents two challenges: (1) an embodiment gap, as human hands differ structurally from robot end-effectors and…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html"&gt;2026-10-04 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.38905</id>
    <title>EmbodiRSI: Recursive Self-Improvement for Data-Efficient Robot Adaptation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.38905" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html" />
    <published>2026-10-03T23:40:00Z</published>
    <updated>2026-10-03T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.38905</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Adapting robot manipulation policies to new tasks and environments remains highly data-intensive, while the data needed for further improvement depends on the policy's current capabilities and failure modes.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.38905 · 2026-09-30&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 50.4% · 83.5% · 83.1% · 75.0%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Adapting robot manipulation policies to new tasks and environments remains highly data-intensive, while the data needed for further improvement depends on the policy&amp;#x27;s current capabilities and failure modes.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html"&gt;2026-10-04 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.39611</id>
    <title>Towards Agile Vision-Based Multi-UAV Flight: Revisiting State Estimation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.39611" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html" />
    <published>2026-10-03T23:40:00Z</published>
    <updated>2026-10-03T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.39611</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Agile multi-UAV flight requires accurate and low-latency onboard estimation of the kinematic states of neighboring UAVs for collision avoidance, motion coordination, etc.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.39611 · 2026-09-30&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 40% · 57% · 300 ms&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Agile multi-UAV flight requires accurate and low-latency onboard estimation of the kinematic states of neighboring UAVs for collision avoidance, motion coordination, etc.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html"&gt;2026-10-04 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.39600</id>
    <title>GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.39600" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html" />
    <published>2026-10-03T23:40:00Z</published>
    <updated>2026-10-03T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.39600</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="R" label="[R]" />
    <summary type="text">Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.39600 · 2026-09-30&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 72.42% · 71.35% · 61.75% · 53.32%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html"&gt;2026-10-04 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.38873</id>
    <title>DODGER: Safety-Guided Reinforcement Learning for Robot Navigation Among Dynamic Obstacles</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.38873" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html" />
    <published>2026-10-03T23:40:00Z</published>
    <updated>2026-10-03T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.38873</name>
    </author>
    <category term="safety" label="安全・権限・コンプライアンス" />
    <category term="R" label="[R]" />
    <summary type="text">Robots operating in human-centered environments must safely navigate among multiple dynamic obstacles to avoid collisions with people and surrounding infrastructure.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 安全・権限・コンプライアンス&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.38873 · 2026-09-30&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Robots operating in human-centered environments must safely navigate among multiple dynamic obstacles to avoid collisions with people and surrounding infrastructure.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-04.ja.html"&gt;2026-10-04 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.00575</id>
    <title>Token-World: World Modeling in Vision-Language Model Token Space for Robot Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.00575" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html" />
    <published>2026-10-02T23:40:00Z</published>
    <updated>2026-10-02T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.00575</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">A common approach to world-model simulation for vision-language-action (VLA) systems is to predict future RGB observations and then re-encode them into policy inputs, introducing an indirect interface between simulation and downstream…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.00575 · 2026-09-30&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;A common approach to world-model simulation for vision-language-action (VLA) systems is to predict future RGB observations and then re-encode them into policy inputs, introducing an indirect interface between simulation and downstream…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html"&gt;2026-10-03 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.00360</id>
    <title>DexPolicy: Scheduled Exploration for Trajectory-Guided Dexterous Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.00360" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html" />
    <published>2026-10-02T23:40:00Z</published>
    <updated>2026-10-02T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.00360</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="R" label="[R]" />
    <summary type="text">Reinforcement learning (RL) for dexterous manipulation must discover finger-object contacts and then control the object precisely; the action noise that serves the first goal can interfere with the second.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.00360 · 2026-09-30&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 49.4% · 68.1% · 14.1% · 45.4%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · コードまたは重み公開&lt;/p&gt;
&lt;blockquote&gt;Reinforcement learning (RL) for dexterous manipulation must discover finger-object contacts and then control the object precisely; the action noise that serves the first goal can interfere with the second.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html"&gt;2026-10-03 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.00864</id>
    <title>Kinematic MeanFlow: One-Step Action Generation Policy for Robotic Foundation Models</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.00864" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html" />
    <published>2026-10-02T23:40:00Z</published>
    <updated>2026-10-02T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.00864</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">In this paper, we study how to achieve one-step action generation in Robotic Foundation Models (RFMs), aiming to overcome the high inference latency of multi-step flow matching.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.00864 · 2026-10-01&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 67.5% · 74.4% · 30.3% · 54.9%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · コードまたは重み公開 · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;In this paper, we study how to achieve one-step action generation in Robotic Foundation Models (RFMs), aiming to overcome the high inference latency of multi-step flow matching.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html"&gt;2026-10-03 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2610.00355</id>
    <title>IndoorBEV: A Lightweight Real-Time LiDAR BEV Perception System for Indoor Mobile Robots</title>
    <link rel="alternate" href="https://arxiv.org/abs/2610.00355" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html" />
    <published>2026-10-02T23:40:00Z</published>
    <updated>2026-10-02T23:40:00Z</updated>
    <author>
      <name>arXiv 2610.00355</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Efficient indoor LiDAR perception is challenging because mobile robots must understand cluttered three-dimensional environments under strict latency and memory constraints.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2610.00355 · 2026-09-30&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 0.6M parameters · 169.6 ms · 200 ms&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり · コードまたは重み公開 · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Efficient indoor LiDAR perception is challenging because mobile robots must understand cluttered three-dimensional environments under strict latency and memory constraints.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html"&gt;2026-10-03 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:aws-physical-ai:b9f6fd764f44</id>
    <title>Isaac Lab on AWS: From Simulation to Registered Policy</title>
    <link rel="alternate" href="https://aws.amazon.com/blogs/physical-ai/isaac-lab-on-aws-from-simulation-to-registered-policy/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html" />
    <published>2026-10-02T23:40:00Z</published>
    <updated>2026-10-02T23:40:00Z</updated>
    <author>
      <name>AWS Physical AI Blog</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="O" label="[O]" />
    <summary type="text">Introduction: The missing layer in Physical AI on AWS Physical AI systems, those that perceive, reason, and act in the physical world, are moving fast from research into production deployments.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [O]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: AWS Physical AI Blog · 2026-10-02&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Introduction: The missing layer in Physical AI on AWS Physical AI systems, those that perceive, reason, and act in the physical world, are moving fast from research into production deployments.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html"&gt;2026-10-03 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:ieee-spectrum-robotics:00bb782fa200</id>
    <title>Video Friday: Albatross Falls, Spins, Self-Rights, and Sails Away</title>
    <link rel="alternate" href="https://spectrum.ieee.org/video-friday-bioinspired-robotics" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html" />
    <published>2026-10-02T23:40:00Z</published>
    <updated>2026-10-02T23:40:00Z</updated>
    <author>
      <name>IEEE Spectrum Robotics</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">We’ve developed ALBATROSS, a hybrid aerial–marine robot that can be released from the air, autorotate down to the water without a parachute, passively self-right after landing, and then reuse the same rigid wings as sails for autonomous…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: IEEE Spectrum Robotics · 2026-10-02&lt;/p&gt;
&lt;blockquote&gt;We’ve developed ALBATROSS, a hybrid aerial–marine robot that can be released from the air, autorotate down to the water without a parachute, passively self-right after landing, and then reuse the same rigid wings as sails for autonomous…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html"&gt;2026-10-03 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:013398c4f77b</id>
    <title>Runway introduces Praxis-1 world action model for robotics</title>
    <link rel="alternate" href="https://www.therobotreport.com/runway-introduces-praxis-1-world-action-model-robotics/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html" />
    <published>2026-10-02T23:40:00Z</published>
    <updated>2026-10-02T23:40:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="M" label="[M]" />
    <summary type="text">When Runway does release Praxis-1 publicly, it plans to ship it with open weights rather than as a closed model.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-02&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: コードまたは重み公開&lt;/p&gt;
&lt;blockquote&gt;When Runway does release Praxis-1 publicly, it plans to ship it with open weights rather than as a closed model.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html"&gt;2026-10-03 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:71461f9591ea</id>
    <title>Inside Omron’s next-generation LD mobile robots</title>
    <link rel="alternate" href="https://www.therobotreport.com/inside-omrons-next-generation-ld-mobile-robots/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html" />
    <published>2026-10-02T23:40:00Z</published>
    <updated>2026-10-02T23:40:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="safety" label="安全・権限・コンプライアンス" />
    <category term="M" label="[M]" />
    <summary type="text">In mobile robotics, software, safety, and fleet management are becoming just as important as the robot itself, according to an OMROM project manager.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 安全・権限・コンプライアンス&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-02&lt;/p&gt;
&lt;blockquote&gt;In mobile robotics, software, safety, and fleet management are becoming just as important as the robot itself, according to an OMROM project manager.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-03.ja.html"&gt;2026-10-03 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.39198</id>
    <title>DSDyn-VLA: A Dual-Stream Dynamic Manipulation Framework with Motion Perception, Future Awareness, and Realtime Correction</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.39198" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html" />
    <published>2026-10-01T23:40:00Z</published>
    <updated>2026-10-01T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.39198</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">While Vision-Language-Action (VLA) models excel in static tasks, they struggle in dynamic environments where objects are in motion (e.g., conveyor belt manipulation).</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.39198 · 2026-09-30&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり · コードまたは重み公開 · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;While Vision-Language-Action (VLA) models excel in static tasks, they struggle in dynamic environments where objects are in motion (e.g., conveyor belt manipulation).&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html"&gt;2026-10-02 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.39763</id>
    <title>DiffWAM: A Fast and Efficient Navigation World Action Model</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.39763" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html" />
    <published>2026-10-01T23:40:00Z</published>
    <updated>2026-10-01T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.39763</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Pretrained video foundation models encode rich semantic and spatiotemporal priors for embodied navigation, yet converting these priors into UAV motion typically requires expensive future-video synthesis and geometric reconstruction.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.39763 · 2026-09-30&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 74.40%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Pretrained video foundation models encode rich semantic and spatiotemporal priors for embodied navigation, yet converting these priors into UAV motion typically requires expensive future-video synthesis and geometric reconstruction.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html"&gt;2026-10-02 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.37771</id>
    <title>Faster and Better? Benchmark Bugs and Design Limitations Distort the Evaluation of Vision-Language-Action Acceleration</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.37771" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html" />
    <published>2026-10-01T23:40:00Z</published>
    <updated>2026-10-01T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.37771</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Simulated manipulation benchmarks are the standard tool for evaluating vision-language-action (VLA) policies and the acceleration methods that reduce their inference latency for on-robot deployment.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.37771 · 2026-09-29&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · コードまたは重み公開 · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Simulated manipulation benchmarks are the standard tool for evaluating vision-language-action (VLA) policies and the acceleration methods that reduce their inference latency for on-robot deployment.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html"&gt;2026-10-02 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.39822</id>
    <title>Toward Real-Time VLAs: Stage-Aware Two-Step Flow Denoising and System-Level Evaluation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.39822" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html" />
    <published>2026-10-01T23:40:00Z</published>
    <updated>2026-10-01T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.39822</name>
    </author>
    <category term="foundation" label="基盤モデル（VLA / WAM）" />
    <category term="R" label="[R]" />
    <summary type="text">Vision-language-action (VLA) models face a timing gap between low-rate inference and high-rate robot execution.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 基盤モデル（VLA / WAM）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.39822 · 2026-09-30&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 61.557 ms · 21.956 ms&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Vision-language-action (VLA) models face a timing gap between low-rate inference and high-rate robot execution.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html"&gt;2026-10-02 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:ieee-spectrum-robotics:73878f37956b</id>
    <title>Atlas Robot’s New Hand May Outperform Humanlike Designs</title>
    <link rel="alternate" href="https://spectrum.ieee.org/robust-robot-hand" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html" />
    <published>2026-10-01T23:40:00Z</published>
    <updated>2026-10-01T23:40:00Z</updated>
    <author>
      <name>IEEE Spectrum Robotics</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">Today, Boston Dynamics is announcing a redesigned hand for its Atlas robot . The previous generation could do some crazy superhuman things with its three fingers, but it was never intended for mass production in the tens or hundreds of…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: IEEE Spectrum Robotics · 2026-10-01&lt;/p&gt;
&lt;blockquote&gt;Today, Boston Dynamics is announcing a redesigned hand for its Atlas robot . The previous generation could do some crazy superhuman things with its three fingers, but it was never intended for mass production in the tens or hundreds of…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html"&gt;2026-10-02 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:147d35dd5178</id>
    <title>Boston Dynamics drops pinkie on new humanoid hand</title>
    <link rel="alternate" href="https://www.therobotreport.com/boston-dynamics-drops-pinkie-on-new-humanoid-hand/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html" />
    <published>2026-10-01T23:40:00Z</published>
    <updated>2026-10-01T23:40:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="hardware" label="ハードウェアとサプライチェーン（機体 / コスト）" />
    <category term="M" label="[M]" />
    <summary type="text">Boston Dynamics engineers taped their pinkies and ring fingers together for a day to learn what they weren’t able to do.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: ハードウェアとサプライチェーン（機体 / コスト）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-01&lt;/p&gt;
&lt;blockquote&gt;Boston Dynamics engineers taped their pinkies and ring fingers together for a day to learn what they weren’t able to do.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html"&gt;2026-10-02 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:99805b279c40</id>
    <title>Your Robot’s Safety Functions Already Work. What If the Input Lies?</title>
    <link rel="alternate" href="https://www.therobotreport.com/your-robots-safety-functions-already-work-what-if-the-input-lies/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html" />
    <published>2026-10-01T23:40:00Z</published>
    <updated>2026-10-01T23:40:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="safety" label="安全・権限・コンプライアンス" />
    <category term="M" label="[M]" />
    <summary type="text">A robot can follow its safety rules and still act on manipulated information. See what this means for robot safety and how teams can strengthen their testing and evidence.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 安全・権限・コンプライアンス&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-01&lt;/p&gt;
&lt;blockquote&gt;A robot can follow its safety rules and still act on manipulated information. See what this means for robot safety and how teams can strengthen their testing and evidence.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html"&gt;2026-10-02 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:the-robot-report:7d547966b72a</id>
    <title>ANYbotics launches Shift to streamline robot fleet operations, scale inspections</title>
    <link rel="alternate" href="https://www.therobotreport.com/anybotics-launches-shift-streamline-robot-fleet-operations-scale-autonomous-inspections/" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html" />
    <published>2026-10-01T23:40:00Z</published>
    <updated>2026-10-01T23:40:00Z</updated>
    <author>
      <name>The Robot Report</name>
    </author>
    <category term="systems" label="システムとオーケストレーション（長期タスク / 複数方策）" />
    <category term="M" label="[M]" />
    <summary type="text">ANYbotics designed Shift to help maintenance teams run industrial inspections, detect failing equipment, and turn data into work orders.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: システムとオーケストレーション（長期タスク / 複数方策）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [M]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: The Robot Report · 2026-10-01&lt;/p&gt;
&lt;blockquote&gt;ANYbotics designed Shift to help maintenance teams run industrial inspections, detect failing equipment, and turn data into work orders.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-02.ja.html"&gt;2026-10-02 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.38059</id>
    <title>WorldLine: Action-Driven Visual Simulation for Robotic Manipulation</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.38059" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-01.ja.html" />
    <published>2026-09-30T23:40:00Z</published>
    <updated>2026-09-30T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.38059</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Real-world robot learning is constrained by the cost of collecting experience and evaluating candidate behaviors.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.38059 · 2026-09-29&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 10,000 hours · 2,000 hours · 74%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Real-world robot learning is constrained by the cost of collecting experience and evaluating candidate behaviors.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-01.ja.html"&gt;2026-10-01 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.37560</id>
    <title>RoboFin3D: A Sim-to-Real Platform for Robotic Surface Finishing</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.37560" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-01.ja.html" />
    <published>2026-09-30T23:40:00Z</published>
    <updated>2026-09-30T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.37560</name>
    </author>
    <category term="simeval" label="シミュレーションと評価（sim2real / 閉ループ）" />
    <category term="R" label="[R]" />
    <summary type="text">Grinding and sanding are fundamental processes in industrial robotic surface finishing. However, physical trials are expensive and consume workpieces, making reproducible experiments difficult.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: シミュレーションと評価（sim2real / 閉ループ）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.37560 · 2026-09-29&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 77.15% · 84.47% · 97.41%&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 定量結果あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;Grinding and sanding are fundamental processes in industrial robotic surface finishing. However, physical trials are expensive and consume workpieces, making reproducible experiments difficult.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-01.ja.html"&gt;2026-10-01 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.36588</id>
    <title>Cooperative Multi-Agent Vision-Language-Action Models via Reinforced Fine Tuning</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.36588" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-01.ja.html" />
    <published>2026-09-30T23:40:00Z</published>
    <updated>2026-09-30T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.36588</name>
    </author>
    <category term="training" label="学習と自己改善（RL / 事後学習）" />
    <category term="R" label="[R]" />
    <summary type="text">We study reinforcement learning (RL) methods for cooperative multi-agent Vision-Language-Action (VLA) models. This problem is challenging because VLAs are pretrained on large-scale single-agent data and therefore lack the fine-grained…</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: 学習と自己改善（RL / 事後学習）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.36588 · 2026-09-29&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 11 tasks&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり&lt;/p&gt;
&lt;blockquote&gt;We study reinforcement learning (RL) methods for cooperative multi-agent Vision-Language-Action (VLA) models. This problem is challenging because VLAs are pretrained on large-scale single-agent data and therefore lack the fine-grained…&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-01.ja.html"&gt;2026-10-01 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
  <entry>
    <id>tag:noteflowai.github.io,2026:physical-ai-radar:arxiv:2609.37772</id>
    <title>Urgent Actions Go First: Urgency-Aware Denoising for Real-Time VLA Control</title>
    <link rel="alternate" href="https://arxiv.org/abs/2609.37772" />
    <link rel="related" type="text/html" href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-01.ja.html" />
    <published>2026-09-30T23:40:00Z</published>
    <updated>2026-09-30T23:40:00Z</updated>
    <author>
      <name>arXiv 2609.37772</name>
    </author>
    <category term="edge" label="エッジとリアルタイム（オンデバイス推論 / 制御周期）" />
    <category term="R" label="[R]" />
    <summary type="text">Diffusion and flow-matching Vision-Language-Action (VLA) policies generate action chunks through iterative denoising, incurring substantial inference latency that severely limits real-time robotic control.</summary>
    <content type="html">&lt;p&gt;&lt;b&gt;軸&lt;/b&gt;: エッジとリアルタイム（オンデバイス推論 / 制御周期）&lt;br&gt;&lt;b&gt;根拠&lt;/b&gt;: [R]&lt;br&gt;&lt;b&gt;出典&lt;/b&gt;: arXiv 2609.37772 · 2026-09-29&lt;br&gt;&lt;b&gt;主要数値&lt;/b&gt;: 1.89x&lt;br&gt;&lt;b&gt;シグナル&lt;/b&gt;: 閉ループ成功率あり · 実機実験あり · 遅延・制御周期の数値あり&lt;/p&gt;
&lt;blockquote&gt;Diffusion and flow-matching Vision-Language-Action (VLA) policies generate action chunks through iterative denoising, incurring substantial inference latency that severely limits real-time robotic control.&lt;/blockquote&gt;
&lt;p&gt;&lt;a href="https://noteflowai.github.io/physical-ai-radar/radar/daily/2026-10-01.ja.html"&gt;2026-10-01 のレーダー全文&lt;/a&gt;&lt;/p&gt;</content>
  </entry>
</feed>
