Determining the Positions and Orientations of Multiple Robots in Dynamic Environments by Using Real-Time Visual Imagery

RoboCup, which emerged by combining the fields of artificial intelligence and robotics, has been one of the most interesting areas in the world in recent years. Real-time and accurate position detection is one of the most important issues for RoboCup, which operates in a dynamic environment. This project explores the utilization of deep learning and image processing approaches for robot localization. It is aimed to determine the position and angle information of the robots by using Convolutional Neural Networks (CNN) and object detection algorithms. In the first stage of the study, the position information of each robot and ball, represented by different colors on the playing field will be obtained by using image processing algorithms. In addition, the robot orientation will be determined by capturing images of the robots from different angles and positions, and then training the CNN using deep learning algorithms in the MATLAB environment. This study aims to find the location and angle of the robots during the game from a camera on the football field as quickly and accurately as possible.