A Global Initial Pose Estimation Method for Mobile Robots Based on Magnetic Matching and 2D LiDAR Features

With the increasing demand for indoor global localization in the Global Navigation Satellite System (GNSS) denial environments, accurate initial pose estimation has become crucial, as it forms the foundation for subsequent trajectory tracking. To address the challenges of infrastructure dependence, high sensor costs, and large data volumes in current indoor localization research, we propose a global initial pose estimation method based on magnetic matching and 2D LiDAR line features. After constructing the magnetic map and grid occupancy map of the environment, the method first uses magnetic information for coarse localization. Then, line features from LiDAR point clouds and the scene are extracted for angular estimation. Finally, the pose is refined using corner points. This approach enables accurate initial pose estimation of a mobile robot at any location within a global map. The algorithm was tested in an indoor corridor environment, with an average position estimation accuracy of 0.164 m and an average angular accuracy of 1.7°.