Enhanced Localization of Driverless Trains using Multi-Scale Grid LiDAR Fiducial Marker Features*

This paper presents a comprehensive investigation into the driverless train localization technique employing fiducial markers, enhanced by LiDAR sensor technology. The primary contribution lies in the development of a novel multi-scale grid fiducial marker LiDAR template encoding methodology, aimed at improving detection robustness across diverse environmental conditions across multiple ranges. By encoding templates at multiple scales, the proposed approach accommodates variations in marker appearance, sparseness, and orientation, thereby enhancing detection accuracy and reliability by LiDAR. Furthermore, the study delves into the impact of fiducial marker LiDAR template selection based on the distance from the LiDAR sensor, elucidating its influence on detection performance. In this study, a novel approach for train localization using LiDAR technology and ArUco fiducial markers is proposed. The method addresses the challenges inherent in comparing structured image-based fiducial markers with unstructured LiDAR data. By generating ArUco templates from LiDAR data and employing multi-scale grid score decoding, successful marker identification and localization up to 50 meters are achieved. The findings demonstrate the feasibility and effectiveness of utilizing LiDAR for fiducial marker-based train localization, paving the way for enhanced railway navigation systems.