Online Static Probability Map and Odometry Estimation using Automotive LiDAR for Urban Autonomous Driving

This paper presents an online static probability map and odometry estimation using automotive LiDAR for urban autonomous driving. In an urban autonomous driving environment where high-rise buildings and many vehicles exist, estimation of vehicle odometry is an important factor for perception and localization accuracy. Therefore, this study aims to increase the accuracy of vehicle odometry in vulnerable GPS situations. The algorithm consists of two LiDAR point cloud processing modules. The first module constructs static probability map (SPM) and estimates the vehicle odometry using correspondence between consecutive SPM. The second module tracks moving objects using a particle filter based geometric model-free approach (GMFA-PF). Two modules operating in parallel enable online operation by using mutual results for preprocessing. The proposed algorithm has been implemented in the robot operating system(ROS) environment and investigated via actual urban driving. Test results show that the online LiDAR pointwise estimation algorithm improves the vehicle odometry estimation and moving object tracking performance compared with the previous extended kalman filter based method.