A tightly coupled LIDAR-IMU SLAM in off-road environment

Abstract

A tightly coupled LIDAR-IMU SLAM is proposed in this paper to make precise and robust estimation on position, posture, velocity as well as bias of accelerometers and gyros in off-road environment where features are not sufficient. This method is based on optimization of residuals both produced by LIDAR point clouds and IMU integration. The first part of residuals comes from the sum of distance between current sweep point clouds and voxel centroid of relative maps which are built simultaneously. The second part of residuals comes from a pre-integration procedure which takes LIDAR and IMU calibration error into account. A series of experiments based on data collected from an intelligent vehicle platform is carried out to evaluate the SLAM system. The experimental results have proven the ability of the system for precise pose estimation. Compared with the LIDAR-only method, LIDAR-IMU SLAM shows better performance on the estimation accuracy and robustness on position and posture as well as obtaining convergent result of pitch and roll angle.