Implementación del algoritmo YOLOv8 para detección de peatones en la plataforma robótica Turtlebot3

Abstract

In industrial robotics one of the key aspects is the detection of objects, especially for mobile robots, many logistics areas of the industry have a high flow of operators, so it is essential that these robots can detect them to guarantee their safety, for example. On the other hand, pedestrian detection continues to be a challenge due to the number of variables to be analyzed, the speed of inference, environmental occlusion, among others. Therefore, the following work presents the implementation of the YOLOv8 algorithm on the Turtlebot3 robotic platform. . for the detection of pedestrians and the STOP traffic sign in a controlled environment, which can be applied in mobile cargo robots, as well as multipurpose robots at a logistics level. An autonomous system was obtained that detects pedestrians in real time in a controlled environment, as well as a “stop” traffic sign; the ROS Nav library was also implemented for the autonomous navigation of the robot capable of setting destinations, tracing routes and avoiding . obstacles. The YOLOv8 algorithm was selected for detection given its inference speed and accuracy, a detailed explanation of the implementation and development process is offered through the platform and the use of the ROS environment. The system achieves real-time detection with 15 FPS video at 480p and an inference speed of 10ms with the Yolov8n model.