Scholarlink

<p>The growing demands of the supply chain industry and advancements in technology have highlighted the need for agile and efficient warehouse operations to support dynamic and adaptive industrial environments. However, most warehouses today are still dominated by traditional systems, which often rely on manual processes or rigid automation, making them less adaptable to rapidly changing conditions and unable to fully integrate human–machine collaboration. This study conducted a Human–Cyber–Physical System (HCPS) interface utilizing eye gaze estimation to control a robotic storage system, addressing key limitations in existing interface implementations—namely, the absence of intuitive control interfaces, reliance on multimodal sensors, and limited real-time application in industrial environments. Unlike previous HCPS interfaces that rely on intrusive wearable sensors, this study introduces a camera-based eye gaze interface requiring minimal hardware and delivering high classification accuracy. Eye gaze signals were captured using the Omron HVC-P camera module, transforming them into Cartesian coordinates via Euler angle calculations to enable precise selection of the storage rack. Experimental validation involved preprocessing facial position data to correct offsets and ensure reliable measurements, with multiple machine learning algorithms—Decision Tree, Support Vector Machine (SVM), Discriminant Analysis, Naive Bayes, and K-Nearest Neighbors (KNN)—applied to classify gaze targets. Among these, Decision Tree, Discriminant Analysis, and KNN achieved perfect classification accuracy (100%), while SVM and Naive Bayes demonstrated lower accuracies of 81.81 and 81.84%, respectively. These results demonstrate the system’s potential to enhance human–robot interaction, supporting the transition toward Industry 5.0 by fostering flexible and adaptive collaboration between humans and intelligent robotic systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Eye Gaze-Based Human–Cyber–Physical System for Robotic Storage Control: Evaluation of Machine Learning Classifiers

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Engelbert Harsandi Erik Suryadarma,
  • Pringgo Widyo Laksono,
  • Ilham Priadythama,
  • Lobes Herdiman

摘要

The growing demands of the supply chain industry and advancements in technology have highlighted the need for agile and efficient warehouse operations to support dynamic and adaptive industrial environments. However, most warehouses today are still dominated by traditional systems, which often rely on manual processes or rigid automation, making them less adaptable to rapidly changing conditions and unable to fully integrate human–machine collaboration. This study conducted a Human–Cyber–Physical System (HCPS) interface utilizing eye gaze estimation to control a robotic storage system, addressing key limitations in existing interface implementations—namely, the absence of intuitive control interfaces, reliance on multimodal sensors, and limited real-time application in industrial environments. Unlike previous HCPS interfaces that rely on intrusive wearable sensors, this study introduces a camera-based eye gaze interface requiring minimal hardware and delivering high classification accuracy. Eye gaze signals were captured using the Omron HVC-P camera module, transforming them into Cartesian coordinates via Euler angle calculations to enable precise selection of the storage rack. Experimental validation involved preprocessing facial position data to correct offsets and ensure reliable measurements, with multiple machine learning algorithms—Decision Tree, Support Vector Machine (SVM), Discriminant Analysis, Naive Bayes, and K-Nearest Neighbors (KNN)—applied to classify gaze targets. Among these, Decision Tree, Discriminant Analysis, and KNN achieved perfect classification accuracy (100%), while SVM and Naive Bayes demonstrated lower accuracies of 81.81 and 81.84%, respectively. These results demonstrate the system’s potential to enhance human–robot interaction, supporting the transition toward Industry 5.0 by fostering flexible and adaptive collaboration between humans and intelligent robotic systems.