A Real-Time Embedded Localization in Indoor Environment Using LiDAR Odometry

Localization for mobile robots in a pre-structured environment with a given prior map is considered as the essential problem to perform the further autonomous navigation. For indoor environments without the access to external localization system like GPS, we present a localization method based on the Monte Carlo Localization (MCL), only utilizing a modern 2D LiDAR of high update rate and low measurement noise, to locate the mobile robot in the prior map without giving a starting point. A LiDAR pseudo-odometry is proposed to compute pose changes in movements of the robot, in which the scan point clouds are matched against a locale map to reduce the cumulative errors. In localization iterations, the LiDAR odometry provides motion data to predict the position hypotheses distribution, which is corrected by incorporating the current LiDAR observation to update and yield the localization estimates. The experiments performed on a car-like mobile robot in the real indoor environment demonstrate the accuracy and the real-time performance of the proposed localization system.