A METHOD TRACKING MOBILE ROBOT INDOOR ENVIRONMENT USING CEILING CAMERA USING YOLOv9

Abstract

This study presents a method for determining the position of a mobile robot within an indoor environment using a ceiling-mounted camera and the YOLOv9 deep learning network. Conventional  solutions often involve attaching a QR code tag to the robot; however, such tags are usually small, prone to noise, and affected by lighting conditions. Therefore, we propose the YOLOv9 deep learning network to track the random movement of the robot. Additionally, the robot's state while navigating through a maze is represented using optical flow methods. We also conduct transforming coordinates from the camera coordinate system to the Cartesian coordinate system to detect the current position of the mobile robot. The results indicate that the proposed solution can effectively record the entire trajectory of the robot within the maze. At a instantenous point, the optical flow method also demonstrates the robot's state during straight movement, rotation, and proximity to obstacles. These results provide a foundation for addressing advanced problems in mobile robotics, such as trajectory planning and tracking in indoor environments.