Vehicle Localization in Urban Environment Using a 2D Online Map with Building Outlines

Abstract

We discuss GPS-free localization in urban areas using a lidar and a 2D online map with building outlines. To achieve this, the boundaries of buildings extracted from the reference map are matched to 3D point cloud data provided by the lidar. The normalized mutual information between them is maximized. The matching result is used as a measurement and combined with odometry and inertial sensor measurements using an extended Kalman filter. The proposed method has been implemented and verified through an experiment with a mobile robot.