從邊緣運算到物理 AI：協作機器人於智慧組裝之 具身智能技術與產線實務應用

隨著人工智慧 (AI) 技術快速發展，智慧製造正由傳統固定式自動化，逐步邁向具感知與自主決策 能力之物理 AI 系統。本文以 AI 智能風扇模組自動組裝產線為案例，探討協作型機器人結合 AI 視覺辨識、 邊緣運算與閉迴路控制之智慧組裝應用。系統採雙視覺相機架構，整合 YOLO 深度學習辨識模型與比例 控制器 (P Controller) 視覺伺服控制技術，以提升組裝精度與作業穩定性。實際量產測試結果顯示，本系 統可穩定維持高組裝成功率，並將原需分鐘等級之人工作業縮短至百秒內完成，整體效率大幅提升。研 究結果顯示，結合 AI 視覺與協作機器人之物理 AI 技術，具高度實務可行性，可有效提升產線柔性與智 慧製造效率。

With the rapid advancement of Artificial Intelligence (AI), smart manufacturing is evolving from conventional fixed-process automation toward Physical AI systems with perception and autonomous decisionmaking capabilities. This study presents an AI-enabled fan module assembly system integrating collaborative robots, AI vision recognition, edge computing, and closed-loop control for intelligent manufacturing applications. The system adopts a dual-camera architecture combined with a YOLO-based object detection model and proportional-controller visual servoing to improve assembly accuracy and operational stability. Production-line validation demonstrates consistently high assembly success rates while reducing manual assembly time from several minutes to within 100 seconds, achieving the level of significant improvement in efficiency. The results demonstrate that integrating AI vision and collaborative robots into Physical AI systems provides strong practical feasibility for enhancing production flexibility and overall manufacturing efficiency in assembly environments [1].