"Pipe from the Past": Vector control of subway train clearance using 3D LiDAR data

The author shares their experience participating in the "Leaders of Digital Transformation" hackathon, which challenged teams to develop a space monitoring algorithm for autonomous subway trains. The primary goal was to detect obstacles up to 150 meters away in complete tunnel darkness using a Hesai 128 3D LiDAR. Although the solution did not make the top 10 due to data formatting issues, the author introduced an architectural approach called "Pipe from the Past." Instead of traditional frame-by-frame analysis, the project utilizes a digital twin concept of the infrastructure based on spatial memory. The developer has open-sourced the core system under the MIT license and is inviting the community to collaborate on refining the solution for industrial use. The article highlights the importance of robust algorithms for rail-based ADAS systems and offers an alternative perspective on processing LiDAR data.