A Laser SLAM Localization Method Fusing Multi-Sensor Velocity Observations

Simultaneous Localization and Mapping (SLAM) is widely used in the field of autonomous robot navigation. Laser SLAM can leverage high-precision environmental geometric information obtained from LiDAR to achieve relatively accurate pose estimation in static and structured environments. To address the decline in accuracy and robustness of the laser odometry system in dynamic environments, under sensor observation constraints, and in cases of feature loss, a scalable multi-sensor fusion system is proposed. Based on the Kalman filter framework, this system integrates external sensor observations such as IMU, GNSS, and Camera with laser odometry localization information, enhancing the system's localization accuracy, stability, and motion estimation smoothness. The fusion method allows for the flexible addition or removal of sensors with poor observations according to actual needs, thereby adapting to different application scenarios. Experimental data from campus environments demonstrate that the proposed multi-sensor fusion localization method exhibits higher localization accuracy and stability compared to a single laser odometry system.