Loosely-coupled lidar-inertial odometry and mapping in real time

The lidar-inertial-based simultaneous localization and mapping (SLAM) have been widely investigated in recent years. In this paper, a loosely-coupled lidar-inertial odometry and mapping method is developed for robot state estimation in real-time. The proposed lidar-inertial odometry is a loosely-coupled and nonlinear optimization-based method, fusing IMU measurements and pose of lidar odometry. Lidar odometry processing de-skewed point cloud with keyframes strategy, which significantly saves computation and allows scan matching run in real-time. Further, high-frequency robot state is obtained by imu prediction in a short time. And a sliding-window-based optimization is preformed to correct imu prediction in time. Real-world experiments and public dataset tests are performed in different scenarios to validate accuracy and effectiveness of our method.