An efficient end-to-end EKF-SLAM architecture based on LiDAR, GNSS, and IMU data sensor fusion for autonomous ground vehicles

The autonomous ground vehicle’s successful navigation with a high level of performance is dependent on accurate state estimation, which may help in providing excellent decision-making, planning, and control tasks. Outside factors like air bias and multipath effects have an impact on the GPS data, obtaining accurate pose estimation remains challenging. To obtain a highly precise pose estimation, the authors propose using an end-to-end simultaneous localization and mapping architecture based on scan matching and an extended Kalman filter to perform a successful prediction using lidar, GNSS and IMU data sensor fusion. In terms of pose estimation efficiency, the obtained results from EKF are compared to the UKF filter in different sequences on the Kitti dataset to more thoroughly evaluate our designed approach. The EKF filter has a good improvement, with approximately 0.29, 0.31, 0.24, 0.34, and 0.27 of the mean RMSE error of the trajectories 09_30_drive_0018, 09_30_drive_0020, 09_30_drive_0027, 10_03_drive_0033, and 10_03_drive_0034, respectively. The proposed approach was evaluated not only on KITTI dataset but also using our data recording platform called KONA Platform.