3D LiDAR Point Cloud Registration Based on IMU Preintegration in Coal Mine Roadways

Area of Science:

• Robotics
• Sensor Fusion
• Computer Vision

Background:

• 3D LiDAR point cloud registration is crucial for robot perception and SLAM.
• Coal mine environments present challenges like sparseness, motion distortion, and weak features.
• Traditional methods struggle with accuracy, z-axis drift, and map ghosting in these conditions.

Purpose of the Study:

• To develop an improved point cloud registration method for robots operating in challenging environments.
• To address limitations of traditional methods in coal mine roadways.
• To enhance the accuracy and robustness of LiDAR-based odometry and SLAM.

Main Methods:

• Proposed a method integrating Inertial Measurement Unit (IMU) preintegration with LiDAR data.
• Utilized IMU linear interpolation to correct motion distortion and RANSAC for ground segmentation.
• Extracted feature corner and plane points to construct a LiDAR point cloud registration error function.
• Employed Gaussian Newton optimization for refining LiDAR odometry frame constraints.

Main Results:

• The proposed method demonstrated higher registration accuracy and success rates compared to traditional approaches.
• Achieved improved computational efficiency.
• The resulting LiDAR odometry provided more authentic robot trajectories with higher accuracy and reduced absolute position/pose errors.

Conclusions:

• The IMU-assisted point cloud registration method effectively overcomes challenges in feature-degraded environments.
• This approach significantly enhances the performance of LiDAR odometry and SLAM systems for mobile robots.
• The method offers a more reliable solution for real-time environment perception in complex settings.