Development of a Self-localization Technology with LiDAR Data Processing Utilizing Landmark Information

Abstract

In order to improve the self-localization accuracy of autonomous driving vehicles, we have developed the processing system of LiDAR scan data utilizing the landmark information stored in Advanced map. The data processing system consists of three contents: the optimal prediction window for landmark detection, determination of reliability information of landmark measurement, and adaptive Kalman gain according to the generated reliability information. The experimental results confirmed that high accurate self-localization can be realized with above technologies. After that, from the simulation assuming by using forward scan type LiDAR, we calculated the accuracy of the estimated position and confirmed the sufficient performance. In addition, we studied how to update of landmark information. Simulation results of them indicated that utilizing the cloud server will make possible the update of landmark information.