Scaled Model Development and Testing for Automatic Train Operation: Creating a Digital Twin

Abstract This paper explores the fundamental aspects of on-sight Automatic Train Operation (ATO) in environments without external protection. Development and testing face significant challenges due to the lack of available test vehicles and tracks, caused by high costs and logistical or operational constraints. A scaled model approach is proposed to address these challenges. The primary objective is to demonstrate how scaled models can overcome these barriers, focusing on object detection, localization and digital twins. The research employs a scaled model to address the key challenges associated with on-sight ATO in unprotected environments. The implementation of object detection was achieved through the utilisation of LiDAR sensors, enabling real-time obstacle identification similar to full-scale systems. Localization was achieved using ultrasonic sensors as a scaled GPS substitute, and a camera-based approach. A digital twin of the laboratory was created in Unreal Engine, using an in-house virtual railways environment to simulate vehicle sensors and compare with real-world data in further work. The scaled model demonstrated real-time object detection using LiDAR and precise localization with ultrasonic and camera-based methods. The digital twin accurately simulated the environment, enabling comparisons between simulated and real-world sensor data, providing insights that are difficult to obtain from real-world testing alone. The study confirms that scaled model development, offers a practical solution for advancing on-sight ATO. Moreover, the results suggest that further exploration of these capabilities within scaled models is essential for optimizing the product development process of ATO functions, potentially leading to more efficient and cost-effective innovations in the field.