Avanços na navegação autônoma de AGVS: desenvolvimento e implementação de segmentação semântica baseada em YOLOv8 para ambientes industriais

Abstract

Mobile robotics has emerged as a key area in industrial automation, especially with the use of Automated Guided Vehicles (AGVs). Inspired by the human ability to identify and distinguish floor markings in factories for safe navigation, this dissertation presents the development and implementation of an autonomous navigation system for AGVs using advanced computer vision techniques, with an emphasis on semantic segmentation through the You Only Look Once version 8 (YOLOv8n) model. The work involved creating a high-quality dataset, training the YOLOv8 model using this dataset, integrating the developed computer vision model with the PID controllers that guide the AGVs, and validating the system in real industrial environments. The implementation utilized the Robot Operating System (ROS), a middleware widely used in robotics, which efficiently managed the communication between sensors, the computer vision model, and the AGV controllers. The results demonstrated that the proposed model, generated from YOLOv8, is capable of accurately identifying and differentiating floor markings, such as pedestrian crossings and navigation areas, enabling AGVs to navigate with the same safety and efficiency as a human walking through a factory. The precision mentioned in the text refers to YOLOv8’s ability to accurately identify and differentiate factory floor markings, such as pedestrian crossings and navigation areas. This precise identification is essential to ensure that AGVs can navigate safely and efficiently, autonomously, in industrial environments. In other words, the term “precision”here refers to the accuracy with which the developed computer vision model can correctly distinguish elements of the environment, allowing AGVs to make navigation decisions comparable to those of a human who relies on the same markings. Precision is a critical factor in avoiding misinterpretation of floor markings, which could compromise the safety and effectiveness of autonomous navigation. The application of the proposed model not only improved the of logistical operations but also proved to be a practical and innovative solution to meet the demands of Industry 4.0.