Landmark-based Automated Guided Vehicle Localization Algorithm for Warehouse Application

Automated guided vehicle (AGV) is a solution for warehouse goods transportation, but robot localization is crucial for this application and existing methods are expensive. Therefore, in this paper, a low-cost landmark based AGV algorithm localization algorithm with single camera is proposed for warehouse application. The proposed algorithm includes the computer vision algorithm to recognize the landmark and estimate the distance between the landmark and AGV with single camera. Previous localization algorithm based on triangulation is using three landmarks for localization, the proposed localization algorithm uses only two landmarks which is based on concept of intersection of two circles. The landmarks in the scene were detected with Canny edge detection method and transformed back to straight square from skewed image with perspective transform to provide consistent landmark recognition result. The landmark then was recognized with Tesseract open source character recognition library and custom trained database. The performance of the proposed algorithm was evaluated using images captured by a single camera setup on a trolley and maneuvered through the library and laboratory at Universiti Sains Malaysia with landmarks. The recognition accuracy for landmark is 93.26% overall. The average error of the localization algorithm was 237.29mm and standard deviation 184.27mm. As a conclusion, landmark based AGV localization algorithm for warehouse application was successfully developed.