Research on 3D LiDAR outdoor SLAM algorithm based on LiDAR/IMU tight coupling.

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Traditional Simultaneous Localization and Mapping (SLAM) algorithms struggle with GPS signal loss and drift in outdoor environments.
• Existing SLAM methods often exhibit inaccuracies in map construction and positioning, particularly in challenging outdoor settings.

Purpose of the Study:

• To develop a robust 3D LiDAR and inertial guidance tightly coupled SLAM algorithm for enhanced outdoor localization.
• To address the limitations of traditional SLAM algorithms concerning GPS dependency and motion distortion.

Main Methods:

• Utilizes IMU forward and backward propagation for position prediction and motion distortion compensation.
• Employs Iterative Error State Kalman Filter (IESKF) for fusing LiDAR point cloud residuals and IMU data.
• Incorporates a sparse voxel near-neighbor structure (iVox) for efficient key frame selection and local map construction.

Main Results:

• The proposed algorithm demonstrates lower cumulative error and higher localization accuracy compared to FAST-LIO2 and LIO-SAM.
• Achieves improved visualization and greater robustness in real-world outdoor scenarios and on the KITTI dataset.
• Reduces computational time for point cloud alignment through an iVox-based method.

Conclusions:

• The tightly coupled 3D LiDAR-IMU SLAM algorithm offers superior performance for outdoor navigation and mapping.
• The novel fusion strategy and efficient key frame selection enhance accuracy, robustness, and computational efficiency.
• This approach provides a reliable solution for applications requiring precise outdoor localization without consistent GPS.