Seeking more accurate robot navigation? This paper introduces LTA-OM, an innovative LiDAR-IMU SLAM system that enhances the accuracy and consistency of robot mapping. The research focuses on improving the simultaneous localization and mapping (SLAM) problem by effectively fusing sensor measurements from LiDAR and IMU. LTA-OM employs FAST-LIO2 and Stable Triangle Descriptor as LiDAR-IMU odometry and loop detection methods, respectively. The system achieves real-time long-term association (LTA) mapping by exploiting direct scan-to-map registration and utilizing a corrected history map to provide direct global constraints to the LIO mapping process. LTA mapping enables drift-free odometry at revisit places. Additionally, a multisession mode is designed to allow users to store results, including corrected map points, optimized odometry, and descriptor databases. The benefits of LTA-OM include high-frequency and real-time odometry, driftless odometry at revisit places, and fast loop closing convergence. LTA-OM is versatile and applicable to multiline spinning and solid-state LiDARs, as well as mobile robots and handheld platforms, it steadily outperforms other systems regarding trajectory accuracy, map consistency, and time consumption.
This paper, published in the Journal of Field Robotics, is well-suited to the journal's focus on advancing robotic technologies for field applications. The presented LiDAR-IMU SLAM system, LTA-OM, contributes significantly to the field by providing efficient, robust, and accurate localization and mapping capabilities, essential for autonomous robot navigation in complex environments.