A LiDAR-Inertial-Encoder Odemetry with Global Optimization

Accurate state estimation is the key to realize robot autonomy, LiDAR-inertial odometry has become one of the most widely used methods because it is capable of accurate and fast state estimation. However, LiDAR-inertial odometry could only obtain relative pose, and global position drift is easy to occur at large scale. Driven by the objective of achieving pose estimation without long-term drift, a positioning system with global optimization integrating GNSS measurements and LiDAR-inertial-encoder odometry is proposed in this paper. LiDAR, IMU and encoder data are fused by a nonlinear optimization framework to obtain an accurate local state. The GNSS measurements are prone to electromagnetic interference, satellite rejection, and other various influencing elements. Therefore, we fuse the GNSS, IMU and encoder measurements by Error-state Kalman filter to accquire more robust and smooth global position. In the global optimization module, the local state estimation and the global positioning information are fused in a pose graph optimization to get a global pose without drifts. We evaluate the performance of our algorithm on challenging datasets. Compared with other state-of-the-art algorithms, the positioning accuracy and robustness in our algorithm has been greatly improved and the global drifts are corrected by the global optimization module.