LiDAR-Inertial Odometry with Uncertainty IEKF and Ground Constraint

Over the past few decades, LiDAR-inertial odometry algorithms have advanced significantly, showing satisfactory performance in Unmanned Ground Vehicles (UGVs), Unmanned Aerial Vehicle (UAV s), and various mobile robots across diverse environments. For mobile robots, the continuous acquisition of LiDAR points results in motion distortion due to the movement of the sensor. Most existing algorithms correct motion distortion primarily through Inertial Measurement Unit (IMU) integration or pre-integration. However, these approaches often neglect the uncertainties associated with acquisition time, range measurements, and pose transformations of each point. To address this, we introduce a novel LiDAR-inertial odometry algorithm with uncertainty Iterated Extended Kalman Filter (U - IEKF). By integrating state covariance with acquisition time and point range, our method employs weighted adjustments to point-plane residuals, enhancing accuracy. Additionally, our algorithm incorporates ground constraints through ground segmentation, specifically improving state estimation for ground mobile robots. Validation on public datasets demonstrates our method's superiority, with average error reductions of 47.6% and 11.6% on the UrbanNav and UrbanLoco datasets compared to our baseline.