Real-time IMU Bias Calibration for LiDAR-Inertial SLAM

LiDAR-Inertial SLAM requires accurate initial values of the IMU bias, while the bias of gyroscope and accelerometer can change over time. This paper proposes a real-time and robust method for calibrating IMU bias in LiDAR-Inertial odometry, without requiring the sensor to be static. Our approach separates LiDAR odometry and pre-integrated IMU measurements into two parts for pose estimation, and then tightly integrates the results to solve for gyroscope and accelerometer bias using maximum a posteriori probability (MAP) estimation. Experimental results in various scenarios validate the ability of our method to achieve fast and robust IMU bias calibration in both static and dynamic states.