Generating 3D fundamental map by large-scale SLAM and graph-based optimization focused on road center line

Abstract

The paper presents a method to generate a large-scale 3D fundamental map from a running vehicle. To create an easy-to-use approach for frequent updates, we propose a system to utilize simultaneous localization and mapping (SLAM), which is robot mapping technology. In traditional methods, special machines or many manual operations cause higher mapping costs. The existing mobile mapping method (MMS) requires manual anchoring point measurement for ensuring accuracy. To solve this problem, we propose a 3D map optimization method by using road information from the standard map issued by the Geospatial Information Authority of Japan. From the SLAM result, the road center line of 3D shape map is estimated by assuming the car is running on road. Pose graph optimization between the estimated road center line and that of the standard map corrects cumulative distortion of the SLAM result. The experimental results of on-vehicle 3D LIDAR observation show that the proposed system could correct the cumulative distortion of the SLAM results and automatically generate a large-scale 3D map assuring reference accuracy.