Examinando por Materia "SLAM"

Mostrando 1 - 2 de 2
  • No hay miniatura disponible
    Ítem
    Feasibility of using Monocular Visual SLAM with ROS for robot navigation in warehouses
    (Universidad EAFIT, 2023) Parada Cuadros, Oswaldo José; Castaño Cano, Davinson; Castaño Cano, Davinson
    In recent years, the use of robots in industrial facilities has become widespread. Particularly, the logistics and inventory management industry has benefited from technological advances, however, the high costs of these technologies prevent many companies from adopting this degree of automation. In this research, we propose the feasibility of using monocular Visual SLAM as an alternative to reduce costs in the localization hardware used for robot navigation in warehouses. For that, we study open source implementations, especially those based on ORB-SLAM2, ROS and Ubuntu. Then, we propose a testing environment compose by a Turtle Bot 2 robot and a location that ensembles some of the characteristics of warehouses. Our analysis shows that it is possible to use monocular Visual SLAM in warehouses, however, we mention some elements in the software and hardware that should be fixed or improved in current implementations in order to robustly and safely operate robots in an indoor storage environment.
  • No hay miniatura disponible
    Ítem
    Slam y Groundtruth usando marcadores
    (Universidad EAFIT, 2024) Escobar Acevedo, Julián; Castaño Cano, Davinson
    The present thesis project focuses on the comparison of a visual SLAM method based on ArUco markers, using these markers for both the generation of the ground truth and the execution of SLAM. The research is carried out in the environment of EAFIT University. A marker is placed on top of the robot and a panoramic camera covering the entire area where the robot is located is used for the ground truth method. For the SLAM method, a camera located on the robot is used. During the experiments, markers are distributed around the perimeter of the study site. The camera integrated into the robot allows the detection of the perimeter markers. Initially, a map of the area is created using the ROS working environment, which is compared with real measurements for validation. Subsequently, the location of the robot is determined using both methods mentioned. The obtained data are compared to analyze the variance, as well as the mean squared error (MSE), root mean squared error (RMSE), and mean absolute error (MAE). Finally, conclusions are drawn about the effectiveness of implementing this visual SLAM method either as ground truth or on the robot itself. This study contributes to the advancement in understanding and applying simultaneous localization and mapping techniques in robotic environments using ArUco markers.