日本フィジカルAI新聞

世界のフィジカルAIを、日本語で。

エッジロボティクス/機能安全arXiv:2604.20193

エッジロボティクス向けLLM誘導型安全エージェント:ISO準拠の認識・計算・制御アーキテクチャ

LLM-Guided Safety Agent for Edge Robotics with an ISO-Compliant Perception-Compute-Control Architecture

シェア:XThreadsFacebookLINEはてブBluesky

自然言語の安全規制を実行可能な述語に変換し、冗長なエッジランタイム上で展開することで、人間とロボットの相互作用における機能安全を実現するLLM誘導型安全エージェントを提案した。

詳しい要約

1. どんなもの?

- LLM-guided safety agent for edge robotics - ISO-compliant perception-compute-control architecture - Translates natural-language safety regulations into executable predicates - Deployed on redundant heterogeneous edge runtime - Prototyped on dual-RK3588 platform - Evaluated in human-robot interaction scenarios - Aims for ISO 13849 Category 3 and PL d

2. 先行研究と比べてどこがすごい?

- Addresses gap between probabilistic AI perception and deterministic industrial standards - Enables practical edge implementation of safety-critical embodied AI - Uses cost-effective hardware (dual-RK3588) to achieve ISO 13849 Category 3 and PL d - Prior work likely lacks LLM-guided translation of safety regulations into executable predicates - Prior work may not integrate redundant heterogeneous edge runtime for fault tolerance

3. 技術・手法の肝は?

- LLM-guided translation of natural-language safety regulations into executable predicates - Redundant heterogeneous edge runtime for deployment - Symmetric dual-modular redundancy design with parallel independent execution - Low-latency perception, computation, and control - Fault-tolerant closed-loop execution under edge constraints - Prototyped on dual-RK3588 platform

4. どうやって有効だと検証した?

- Prototyped on dual-RK3588 platform - Evaluated in representative human-robot interaction scenarios - Results demonstrate practical edge implementation path toward ISO 13849 Category 3 and PL d - Validation details (metrics, baselines) not specified in abstract

5. 議論はある?

- Challenges: probabilistic AI perception vs. deterministic industrial standards - Edge constraints: low-latency, fault tolerance, cost-effective hardware - Practical deployment of safety-critical embodied AI - Limitations and broader implications not discussed in abstract

6. 次に読むべき論文は?

- ISO 13849 standard (functional safety) - LLM-guided safety agents - Edge robotics architectures - Dual-modular redundancy designs - Human-robot interaction safety - Embodied AI safety

※ AIが要旨から生成した要約です。正確性は原文をご確認ください。

著者: Xu Huang, Ruofan Zhang, Lu Cheng, Yuefeng Song, Xu Huang, Huayu Zhang, Sheng Yin, Anyang Liang, Chen Qian, Yin Zhou, Xiaoyun Yuan, Yuan Cheng

分類: cs.RO

原文アブストラクト

Ensuring functional safety in human-robot interaction is challenging because AI perception is inherently probabilistic, whereas industrial standards require deterministic behavior. We present an LLM-guided safety agent for edge robotics, built on an ISO-compliant low-latency perception-compute-control architecture. Our method translates natural-language safety regulations into executable predicates and deploys them through a redundant heterogeneous edge runtime. For fault-tolerant closed-loop execution under edge constraints, we adopt a symmetric dual-modular redundancy design with parallel independent execution for low-latency perception, computation, and control. We prototype the system on a dual-RK3588 platform and evaluate it in representative human-robot interaction scenarios. The results demonstrate a practical edge implementation path toward ISO 13849 Category 3 and PL d using cost-effective hardware, supporting practical deployment of safety-critical embodied AI.

  1. 1Boston Dynamics、車輪付きAtlasの研究試作機を紹介研究・技術
  2. 2newmoとKDDIが資本業務提携 2028年に堺市で自動運転タクシー商用化目指す提携・買収
  3. 3GXOとApptronik、Apolloの物流現場試験を拡大 最大1,000台の展開を構想量産・展開
  4. 4NVIDIA、CoRLでGR00T 2とIsaac Lab 3.0を発表新製品・発表
  5. 5UC BerkeleyとNVIDIAの研究者、AIエージェントでロボット制御コードを仮想環境で改善研究・技術

PR・アフィリエイト広告

MINISFORUM(ミニPC)

MINISFORUM(ミニPC)の公式広告バナーMINISFORUM公式ストアを見る ↗
PR本紙発行元 EmplifAI