Digital Map Generation and Localization for Vehicles in Urban Intersections using LiDAR and GNSS Data

Abstract

Automated vehicles require robust and precise ego-localization, especially when operating in challenging environments such as urban intersections. One possibility to improve localization accuracy is the utilization of high-definition digital maps combined with on-board sensors capable of recognizing the mapped features. In this paper, a framework for the creation of digital maps based on LiDAR data and for localization using these maps is presented. Graph SLAM is used to calculate optimal maps given LiDAR, IMU, odometry and GNSS data and a particle filter-based localization approach is proposed for localization. The performance of the framework is experimentally demonstrated by conducting mapping and localization tests in an outdoor test environment.