Navigation system for a semi-autonomous shuttle car in room and pillar coal mines based on 2D LiDAR scanners

Autonomous navigation in underground environments remains a hot topic for the autonomous vehicle engineering community. There is a wealth of information available in the literature for implementing hardware and software solutions for robust navigation in an underground environment. Well-established algorithms are widely used for real-time data processing for self-awareness, situational awareness, and autonomous navigation. However, the geometry of the underground environment is rarely considered for the development of these approaches, in favor of more generalized approaches. In the mining sector, where production vehicles are utilized inside highly structured underground environments, the integration of underground geometry into the development of the navigation system could be beneficial. The inherent simplicity of such environments can significantly mitigate the complexity of the navigation system, as well as the composition and number of the multisensory modalities required for automating mining equipment. This paper presents a GPS-denied navigation system developed for underground coal mines that use the room and pillar mining method. Leveraging this simple and repetitive pattern, the navigation system uses only a few 2D LiDAR scanners that provide the information needed for autonomous navigation in underground room and pillar systems. Under the proposed approach, the calculation of an absolute global position is unnecessary, and navigation can be performed using only the relative position of the vehicle with respect to the immediate environment. The performance of this system is evaluated based on a lab-scale shuttle car prototype that is set to navigate a mock mine setup for a number of different scenarios simulating common missions that a shuttle car may undertake in a typical room and pillar section. The results show a minimum success ratio of 84%.