Global localization algorithms for mobile robot still face many challenges when using in some indoor environments, where the layout changes frequently or there are many dynamic obstacles. Aiming at such scenarios, this paper proposed a new and convenient global localization system based on man-made landmark. The system localized based on multiple man-made landmarks attached on the unshaded ceiling, and could achieve stable pose estimation with only one camera. The whole system consisted of two main processes: map construction and global localization, according to their specific functions. During the map construction, it automatically estimated the pose of the man-made landmark in the global coordinate system, according to the camera's observation of the landmark and the pose estimation output by some laser SLAM algorithm. Given this, it then established the map of environment based on the pose of each landmark. During the global localization, it estimated the robot's pose in real time based on the camera's observation of the man-made landmarks in the map with known poses, combined with the pre-integrated information fused with the odometer and IMU. Sufficient experiments show that the localization error of the system is stable within 10 cm when running in the deployed environment, and the system can also ensure real-time pose output, which meets the application requirements of typical indoor mobile robot global localization.
» Read on@article{LWapprescom21,
author = {Zhihan Liao and Feng Wu},
doi = {10.19734/j.issn.1001-3695.2021.01.0033},
journal = {Application Research of Computers (ARC)},
number = {9},
pages = {2741-2745},
title = {Man-made landmark based convenient indoor-robot global localization system},
volume = {38},
year = {2021}
}