Advancing Road Traffic Safety: Deep Learning for Traffic Modeling and Strategy

This dataset was created for conflict and accident modeling. Real-time traffic data were first collected through a traffic API and then integrated with the microscopic traffic simulation platform Simulation of Urban MObility to reconstruct detailed vehicle trajectories. The simulation produced trajectory files in XML format, which were subsequently converted into a CSV file (trajectories_with_angle(april_may_june).csv). Using Python-based data processing and analysis, surrogate safety indicators were then computed from the trajectory data. This process generated the final dataset ttc_lttb_final_split_by_type_yaw(april_may_june).csv, which is intended for subsequent algorithm development and modeling tasks related to traffic conflict and accident analysis. Due to privacy and safety considerations, the complete set of reconstructed pre-crash trajectories cannot be fully disclosed. Therefore, a representative example dataset (case_1.csv) is provided to illustrate the structure and format of the reconstructed trajectory data.