LIDAR-INERTIAL NAVIGATION BASED ON MAP AIDED DISTANCE CONSTRAINT AND FACTOR GRAPH OPTIMIZATION

Abstract

Abstract. The simultaneous localization and mapping (SLAM) is one of the well-developed positioning technology that provides high accuracy and reliability positioning for automatic vehicles and robotics applications. Integrating Light Detection and Ranging (LiDAR) with an Inertial Measurement Unit (IMU) has emerged as a promising technique for achieving stable navigation results in dense urban environments, outperforming vision-based or pure Inertial Navigation System (INS) solutions. However, conventional LiDAR-Inertial SLAM systems often suffer from limited perception of surrounding geometric information, resulting in unexpected and accumulating errors. In this paper, we proposed a LiDAR-Inertial SLAM scheme that utilizes a prior structural information map which is generated from opensource OpenStreetMap (OSM). In contrast to conventional solutions of OSM-aided SLAM approaches, our method extracts the vectorized models of road and building and synthetically generates dense point maps for LiDAR registration. Specifically, a structural map processing module extracts the road models and building models from OSM and generates a structure information map (SIM) with dense point clouds. Secondly, a map aided distance (MD) constraint is calculated by registering selected keyframes and the prior SIM. Finally, a factor graph optimization (FGO) algorithm is involved to integrate the relative transformation obtained from LiDAR odometry, IMU pre-integration, and the map aided distance constraints. To evaluate the proposed LiDAR-based positioning accuracy, experimental evaluation is implemented in an opensource dataset collected in the urban canyon environments. Experimental results demonstrates that with the help of the proposed MD constraint, the LiDAR-based navigation solution can achieve accurate positioning, with a root mean square deviation (RSME) of 4.7 m.