An Indoor Mapping Algorithm Fusing LiDAR-IMU Tightly Coupled Fusion and Scan Context: IS-LEGO-LOAM.

Area of Science:

• Robotics and Autonomous Systems
• Computer Vision
• Sensor Fusion

Background:

• Indoor environments with sparse features (corridors, atriums) pose challenges for Simultaneous Localization and Mapping (SLAM).
• Difficulties in loop closure detection and accumulated positioning errors lead to localization drift and mapping failures.

Purpose of the Study:

• To propose an improved indoor mapping algorithm, IS-LEGO-LOAM, to address localization drift and mapping failures in feature-poor environments.
• To enhance the robustness and accuracy of LiDAR-IMU based SLAM systems in complex indoor settings.

Main Methods:

• Integration of tightly coupled LiDAR-IMU odometry with an adaptive covariance matrix to handle abnormal LiDAR echoes and sparse features.
• Introduction of Scan Context global descriptor with vector nearest neighbor search and similarity score matching to mitigate drift in large-scale scenes.

Main Results:

• The proposed IS-LEGO-LOAM algorithm demonstrates superior mapping performance compared to existing methods.
• Validation on the KITTI dataset and real-world scenarios confirms the algorithm's effectiveness in reducing localization drift and improving mapping accuracy.

Conclusions:

• IS-LEGO-LOAM effectively overcomes the limitations of traditional SLAM in sparse indoor environments.
• The algorithm offers a robust solution for accurate and reliable indoor mapping, crucial for autonomous navigation and robotics applications.