Simulation-based policy evaluation for bicycle safety under cooperative perception: Linking occlusion modeling to detection performance

Urban intersections remain critical for vulnerable road user (VRU) safety, particularly under occlusion-rich conditions that limit the perception performance of connected and automated vehicles (CAVs). While cooperative perception extends the field of view of CAVs by sharing perception information among them, systematic methodologies linking occlusion modeling with VRU-focused safety assessment remain limited. This study presents an occlusion-aware framework for evaluating VRU safety under cooperative perception. Building on the open-source simulation framework FTO-Sim, geometric ray tracing models static and dynamic occlusions at an urban intersection. Traditional binary visibility metrics are extended toward a Level of Perception Potential, incorporating detection redundancy and sensor performance assumptions. In addition, VRU-specific spatio-temporal detection rates are complemented by a perception-aware VRU perception rate as well as redundancy distributions and occlusion level analysis. A case study evaluates regulatory intersection re-design measures, varying observer penetration rates and heterogeneous observer fleets. The results show that improvements in geometric visibility do not automatically translate into higher VRU detection rates, as dynamic occlusion remains a dominant constraint in dense intersection environments. Increasing homogeneous observer penetration yields marginal improvements in occlusion-critical settings, whereas heterogeneous observer perspectives substantially enhance VRU detection performance. Spatial mapping of full occlusion hotspots reveals persistent blind spots that overlap with historical accident locations, indicating locations for targeted infrastructure-based sensor deployment. The proposed framework provides a reproducible methodology for integrating occlusion modeling into VRU-focused safety assessment and supports evidence-based infrastructure and cooperative perception system design. Source code, simulation outputs and results are publicly available to support transfer to other use cases.