Localization of Humans in Warehouse based on Rack Detection
Karel Košnar, Gaël Ecorchard, and Libor Přeučil
Czech Technical University in Prague, Czech Institute of Informatics, Robotics and Cybernetics, Czechia
This paper proposes a method for human localization in the automated warehouses. As the environment is highly dynamic due to changes in the racks content as well as the position of the rack, the proposed method uses the racks as landmarks for the localization. The fixed structure of the shelves and stands of the rack is used to detect the rack itself regardless of the content. The structure is coded as the graph and the detection is performed by finding the isomorphism between the rack model and the detected structure of the environment. The relative position is obtained by solving a PnP problem. The global position of the rack is gathered from the warehouse system by reading the marker identifying the rack. Experiments show that the localization has a mean error of 15 cm and is much more robust than the localization based on the marker itself.