Research on multimodal data enhanced SLAM algorithm for global mapping of underground coal mines

Area of Science:

• Robotics
• Geospatial Engineering
• Autonomous Systems

Background:

• Inaccurate global mapping in coal mines hinders autonomous driving systems.
• Existing SLAM methods struggle with the unique challenges of underground environments like coal mines.

Purpose of the Study:

• To develop an enhanced SLAM global mapping method for coal mines using multimodal data.
• To improve pose estimation and map establishment for autonomous driving in underground roadways.

Main Methods:

• A frontend fusing LiDAR, inertial measurement unit (IMU), and wheel odometry data using an iterated error-state Kalman filter.
• A backend employing spherical targets for global position constraints and loop closure detection within a multi-factor pose graph.
• Integration of GNSS/UWB signals for point-based map corrections.

Main Results:

• The proposed method effectively mitigates LiDAR degeneration through sensor fusion.
• Accurate pose estimation and map establishment were achieved in simulated and real-world coal mine environments.
• Demonstrated practicability and performance in large-scale underground mapping.

Conclusions:

• The enhanced multimodal SLAM method significantly improves global mapping accuracy in coal mines.
• The approach provides a reliable foundation for autonomous driving systems in subterranean environments.
• The method shows strong potential for practical applications in underground mining operations.