| 摘要: | 本研究提出了一種以數位孿生 (Digital Twin)驅動，結合YOLO視覺辨識模型的IoT設備智能換電池機器人系統，名為BRIDRobot。該系統旨在解決目前在更換IOT感測器電池的方法中，存在的人工高依賴度、效率低和安全性不足等問題，因感測器大多裝於橋梁、風力塔與高樓等高架結構中。本研究利用機械手臂結合數位孿生技術，再通過YOLO視覺辨識模型以辨識標籤的方式定位電池，並達到精準換電之功能。該系統採用了高效的YOLOv8n模型，能在各種光照條件下準確識標籤位置，並且與深度攝影機結合，獲取標籤的三維空間座標。數位孿生部分則提供了機械手臂動作的虛擬環境模擬與閉環校正機制。系統在 Moveit 中建立與真實硬體一致的數位孿生模型，先行模擬換電姿態並輸出末端執行器的預期座標，來確保換電的精準性和安全性。
為評估系統效能，本研究分別進行視覺辨識模組測試、機械臂抓取測試與整體系統驗證。實驗結果顯示，視覺辨識模組在不同光照條件下的檢測準確率均超過90%，證明其具備環境適應性。整體驗證中，本研究之系統能夠在遠程操作下高效完成智能換電流程，智能換電成功率達 91.6%。
本研究的創新之處在於以數位孿生為核心，串聯 ROS 2 通訊架構與 YOLO 視覺辨識，實現高效且精準的自動換電，不僅提升作業精度，也降低對專業人力的依賴與成本。未來，該架構可延伸至其他結構健康監測場域，具備相當的應用價值與發展潛力。
;This study proposes a Digital Twin (DT)-driven intelligent battery replacement robotic system for IoT devices, named BRIDRobot. The system aims to address the limitations of current battery replacement methods for IoT sensors, which are characterized by high dependence on manual labor, low efficiency, and safety concerns, as most sensors are installed on elevated structures such as bridges, wind turbine towers, and high-rise buildings. In this research, a robotic arm integrated with Digital Twin technology is combined with a YOLO-based vision recognition model to localize batteries through label detection, thereby enabling precise battery replacement. The system adopts the efficient YOLOv8n model, which achieves robust detection accuracy under varying lighting conditions, and is further integrated with a depth camera to acquire three-dimensional spatial coordinates of the labels. The Digital Twin module provides a virtual simulation environment and a closed-loop calibration mechanism for the robotic arm’s operations. Within MoveIt, a Digital Twin model consistent with the physical hardware is established to simulate battery replacement postures in advance and output the expected end-effector coordinates, ensuring accuracy and safety during the replacement process.
To evaluate system performance, this study conducted experiments including tests of the vision recognition module, robotic arm grasping, and overall system validation. Experimental results show that the vision recognition module achieved a detection accuracy exceeding 90% under different illumination conditions, demonstrating strong environmental adaptability. In the overall system validation, the proposed system successfully completed the intelligent battery replacement process under remote operation, achieving a success rate of 91.6%.
The novelty of this research lies in the Digital Twin-centered architecture that integrates the ROS 2 communication framework with YOLO-based vision recognition, thereby realizing efficient and precise autonomous battery replacement. This not only enhances operational accuracy but also reduces reliance on specialized manpower and associated costs. In the future, the proposed framework can be extended to other structural health monitoring scenarios, demonstrating significant application value and development potential. |
