Point cloud map creation based on laser and IMU information fusion

YiBing Zhao,ZhenQiang Ma,WeiQi Wang,Bin Li,ShuYong Xing

DOI: https://doi.org/10.1109/cvci54083.2021.9661174

2021-10-29

Abstract:It is very necessary for driverless vehicles to understand the surrounding environment information in real time. However, the data observed by a single sensor always has the uncertain noise, which leads to map drift and undesirable mapping effect. Multi-sensor fusion technology can effectively make up for the above shortcomings and improve the mapping effect. The paper proposes an algorithm based on the fusion of inertial navigation sensor and laser odometry, in the data preprocessing stage, the IMU data information is used to perform linear interpolation calculations on the position of each frame of laser point cloud. The calculated position data information is used to remove motion distortion for each frame of laser point cloud and improve the matching accuracy. In the process of mapping, the error state Kalman filter algorithm was employed to fuse the position of the laser odometry and the IMU, and the improved algorithm reduces the average error of the global trajectory by 3.60005 m. By intermittently initializing the IMU odometry, the cumulative error of the IMU is reduced, the IMU odometry was re-initialized by applying the current position information output of the laser odometry as the initial position information. The paper conducts effective experiments on the KITTI urban road data- set. The results prove that the map creation method based on the fusion of laser odometry and IMU can effectively reduce map drift, improve map resolution, and its output of the driving trajectory information is more accurate.