Robust Human Tracking Using a 3D LiDAR and Point Cloud Projection for Human-Following Robots

Area of Science:

• Robotics
• Computer Vision
• Human-Robot Interaction

Background:

• Human tracking is crucial for mobile robots operating alongside people.
• Existing methods using 2D laser range finders (LRFs) struggle with variations in human height.
• 3D LiDAR offers richer environmental data but requires adapted processing for human tracking.

Purpose of the Study:

• To develop a human-tracking method using 3D LiDAR that is resilient to differences in human height.
• To adapt 3D point cloud data for established 2D human-tracking algorithms.
• To enhance the reliability of human following for mobile robots.

Main Methods:

• Utilized a 3D LiDAR to capture environmental point cloud data.
• Projected the 3D point cloud onto a single horizontal plane, similar to 2D LRF data.
• Investigated and identified the optimal projection range (top 30% of point clouds) for stable human tracking.

Main Results:

• The proposed method significantly reduced outlier points in path-following experiments, decreasing from 3.63% to 1.75%.
• Tracking stability was enhanced by projecting a specific portion of the 3D point cloud data.
• Demonstrated improved robustness against variations in human height compared to traditional methods.

Conclusions:

• The developed 3D LiDAR projection method effectively addresses height variability challenges in human tracking.
• This technique enhances the robustness and reliability of mobile robots for human following tasks.
• The findings contribute to more dependable human-robot collaboration in dynamic environments.