2025 年 16 巻 論文ID: PP4059
A proactive approach to evaluating roundabout safety under non-lane-based, highly mixed traffic conditions has been presented in this study using vehicular trajectory data from UAV videos and advanced statistical modeling. Extreme Value Theory (EVT) was applied to determine appropriate threshold values for surrogate safety measures (SSMs), particularly Time-to-Collision (TTC) and MaxDeltaV, ensuring suitability for heterogeneous traffic environments. The study utilized a bivariate Peak-Over-Threshold (POT) model to assess joint exceedances of conflict indicators. Mean residual life (MRL) and threshold stability (TS) plots were utilized to select threshold values. Conflict mapping using the Surrogate Safety Assessment Model (SSAM) revealed that lane-change and rear-end conflicts dominated (41.9% and 32.8%, respectively), whereas crossing conflicts (25.3%) were concentrated at roundabout exits and entrances. The study provides insights for localized safety interventions, emphasizing the need for threshold customization for mixed traffic conditions. Findings offer a data-driven foundation for enhancing roundabout safety in low- and middle-income countries.
