A Tightly Coupled LiDAR-Inertial SLAM for Perceptually Degraded Scenes

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Simultaneous Localization and Mapping (SLAM) and six degrees of freedom (6DOF) state estimation are crucial for robot navigation.
• Perceptually degraded environments like tunnels pose significant challenges for existing SLAM algorithms.
• Robust state estimation and high-performance SLAM are essential for autonomous systems.

Purpose of the Study:

• To develop a robust SLAM algorithm for perceptually degraded scenes using LiDAR-IMU fusion.
• To enhance the accuracy and real-time performance of six degrees of freedom (6DOF) state estimation.
• To improve the global consistency of maps constructed by robots in challenging environments.

Main Methods:

• A tightly coupled LiDAR-IMU fusion SLAM algorithm comprising front-end iterative Kalman filtering and back-end pose graph optimization.
• Front-end: Iterative Kalman filter for LiDAR-Inertial Odometry (LIO) to enhance attitude accuracy and system robustness.
• Back-end: Keyframe selection, loop detection, and ground constraints for real-time performance and improved 6DOF state estimation accuracy.

Main Results:

• The proposed algorithm demonstrates superior accuracy, real-time performance, and robustness compared to existing LiDAR-SLAM methods in degraded scenes.
• Effectively reduces cumulative system error in challenging environments.
• Ensures global consistency of constructed maps.

Conclusions:

• The tightly coupled LiDAR-IMU fusion SLAM algorithm offers a robust solution for robot navigation in perceptually degraded environments.
• The method significantly improves 6DOF state estimation and mapping performance.
• The algorithm provides a reliable foundation for autonomous systems operating in complex, unstructured settings.