Vehicle shape influence on virtual forward simulation of car crashes

Abstract Progressive deployment of active safety systems requires precise tools to assess collision mitigation performance. Virtual Forward Simulation (VFS) plays a pivotal role in the prospective assessment of Advanced Driver Assistance Systems (ADAS). Reliable VFS routines depend on validated models capable of resolving collision mechanics, particularly when ADAS interventions fail to avoid a collision and the focus shifts to evaluating impact mitigation. However, software tools often omit exact vehicle geometries in accident reconstruction practice. Consequently, when graphical data is compiled for detailed accident records, specifically in the Pre-Crash Matrix (PCM) format, vehicle shapes are typically simplified. Although encoding realistic vehicle shapes demands substantial technical effort during database compilation, this study demonstrates that such fidelity is essential for ensuring high-fidelity outcomes. Data fidelity requirements are identified by investigating the influence of vehicle shape on the reliability of 2D car-to-car collision simulations. A validated tool based on a Reduced Order Dynamic Model (RODM) was used to simulate road collisions using progressively simplified vehicle geometries. Simulations were statistically analyzed and verified through a real-world case study. Findings indicate that while simpler momentum-based models remain insensitive to geometric variations, advanced models like RODM require detailed shapes to accurately resolve impact dynamics. Neglecting this leads to a significant underestimation of key parameters, specifically $$\Delta V$$ (by more than 5 km/h) and occupant injury risk (by up to 60% in real-world scenarios). By defining these data fidelity requirements, this work provides a methodological framework to bridge the gap between current database standards and high-fidelity virtual safety research.