A Study of an Indoor AGV Logistics Mobile Robot Platform Incorporating YOLOv5-Based Package Open/Closed State Recognition
TL;DR
This study presents an indoor logistics mobile robot system that integrates an automated guided vehicle with deep learning-based object recognition and employs the YOLOv5 algorithm to classify box-sealing states in real time, and autonomously transports the identified items to the designated destinations based on the classification results.
Abstract
Logistics automation systems have recently emerged as critical technologies in smart factories for enhancing productivity and ensuring industrial safety. This study presents an indoor logistics mobile robot system that integrates an automated guided vehicle with deep learning-based object recognition. The proposed system employs the YOLOv5 algorithm to classify box-sealing states (open or closed) in real time, and autonomously transports the identified items to the designated destinations based on the classification results. An RGB sensor-based line-tracking method is adopted for autonomous driving across multiple path configurations to enable robust navigation, whereas ultrasonic sensors are used for real-time obstacle detection to mitigate collision risks and improve the operational safety and reliability. Experimental evaluations demonstrate that the object recognition model achieves approximately 99% precision and a mean average precision calculated at an intersection over union (IoU) threshold of 0.50 >98%, indicating a high detection accuracy and robustness. The results confirm that the proposed system enhances the efficiency and accuracy of logistics operations in smart factories.