Keywords: automatic train operation, on-sight train operation, LiDAR simulation, Unreal Engine, real-time simulation, virtual validation.

Recent European railway initiatives such as Shift2Rail emphasise sustainability, capacity increase, and cost efficiency through automation and digitalisation. In this context, Automatic Train Operation increasingly relies on robust environment perception for safe on-sight operation, where LiDAR sensors provide accurate geometric information for obstacle detection and infrastructure monitoring. Driven by advances in automotive and robotics research, many LiDAR simulation approaches focus either on high physical fidelity at significant computational cost or on offline dataset generation, limiting their suitability for real-time closed-loop validation. For railway-oriented Automatic Train Operation testing, however, a lightweight, engine-native, and reproducible real-time simulation is required. This paper presents a CPU-based LiDAR simulation implemented natively in Unreal Engine 5. The approach combines engine-native ray casting, JSON-configurable parameters, and UDP-based point cloud streaming with a pragmatic plausibility model. Instead of reproducing full physical sensor behaviour, selected effects such as distance attenuation, material-based reflectivity proxies, incidence-angle weighting, and measurement noise are incorporated to generate perception-relevant point cloud characteristics under real-time constraints. The simulation enables reproducible and configurable LiDAR perception for on-sight Automatic Train Operation validation and is released as part of the open-source Vir2Rail environment, which provides railway-specific assets and scenarios.

download the full-text of this paper (PDF, 14 pages, 567 Kb)

go to the previous paper
go to the next paper
return to the table of contents
return to the volume description