Efficient Doppler LiDAR Odometry Using Scan Slicing and Vehicle Kinematics

In contrast to traditional 3-D light detection and ranging (LiDAR), the newly introduced 4-D LiDAR not only provides conventional point cloud data but also offers Doppler velocity measurements of the point cloud. This capability allows the sensor to provide additional effective constraints in scenes with degraded geometric features. However, the vast quantity of point cloud data generated by this type of LiDAR poses a challenge to the efficiency of LiDAR odometry. To fully leverage the characteristics of 4-D LiDAR and achieve efficient pose estimation, we propose an efficient Doppler LiDAR odometry using scan slicing and vehicle kinematics based on iterative error state Kalman filter (IESKF). By slicing the point cloud according to the time sequence, the number of LiDAR points processed at each iteration is reduced. Based on this, we introduce the kinematic model of vehicles as a constraint to obtain the linear velocity and angular velocity of the LiDAR by decomposing the Doppler velocity. The linear velocity and angular velocity can be used to predict the vehicle state and compensate for the motion of the point cloud, which can improve the robustness of the system. Meanwhile, we utilize the Doppler velocity weight to mitigate adverse effects on pose estimation from dynamic points in the IESKF. In our experiments, we demonstrate that the efficiency of the proposed algorithm is approximately five times faster than the state-of-the-art (SOTA) methods while maintaining comparable accuracy. The implementation code of this article will be open-source at https://github.com/NEU-REAL/4DLO.