Assessing safety using traffic simulation is becoming increasingly feasible with advancements in methodological frameworks and tools, emphasizing the critical importance of accuracy and reliability. This study aims to bridge the gap between simulation models and real-world safety observations, contributing to the advancement of more robust safety assessment methodologies. It presents a comprehensive comparative analysis of traffic safety metrics derived from both simulated and real-world data, employing clustering technique to identify safety patterns. Using Aimsun Next, simulation data were analyzed in the Surrogate Safety Assessment Model (SSAM) to extract traffic conflicts, which were then converted into crash risk levels. Real-world crash data from the Hellenic Statistical Authority (ELSTAT) encompassed various crash types involving at least one slightly injured individual between 2017 and 2019. Specifically, observational data encompassed speed limits, road lengths, injuries, vehicles involved, and crash counts, while simulation metrics included flow, capacity, and crash risk. The analysis of simulation and observational data revealed two distinct clusters: roads with low and high crash risks, clearly distinguished with minimal overlap. Comparison of clustering results demonstrated approximately 87.7% accuracy in predicting road crash risk classifications through traffic simulation, confirming its reliability for safety assessment. The study also highlights the importance of thorough calibration; roads inaccurately predicted lacked sufficient traffic data, underscoring the need for robust calibration to enhance safety assessment. This study validates a framework ready for future research applications in scenarios where direct observation is impractical, enhancing road safety and guiding interventions within evolving traffic conditions and technologies.
| ID | pj262 |
| DOI | |
| Tags |
