P-ACT-based traffic conflict risk analysis in highway work zones in mixed traffic conditions

This repository contains the Python source code developed for the trajectory-based traffic conflict and crash-risk analyses reported in the associated study on highway work zones under mixed-traffic conditions. The workflow is organized into four scripts. ACT_and_RiskType.py computes Anticipated Collision Time (ACT) from vehicle trajectories and classifies interaction types using vehicle geometry and collision-course information. EVT_ACT_Threshold.py implements the Extreme Value Theory (EVT) threshold-selection procedure by extracting conflict minima, fitting Generalized Pareto Distributions to threshold exceedances, examining mean residual life and parameter-stability diagnostics, and selecting the minimum-AIC threshold within a reviewed stable interval. Probabilistic_ACT (P-ACT).py extends deterministic ACT to an uncertainty-aware probabilistic formulation using an assumed joint Gaussian measurement-error model and site-specific EVT-derived ACT thresholds. The probabilistic workflow also includes sensitivity analysis of the assumed measurement uncertainties by varying the standard deviations associated with position, vehicle size, speed, acceleration, and yaw rate individually, as well as varying all uncertainty parameters simultaneously, from 50% to 150% of their baseline values. Sensitivity outputs support site-specific and pooled comparisons of changes in mean P-ACT, together with video-level baseline comparisons and multiple-testing-adjusted significance assessment. Analysis_Plot_Mapping.py performs the downstream analysis of ACT and P-ACT, including spatial aggregation and prioritization, ACT–P-ACT comparison, bin-width sensitivity assessment, conflict-type and vehicle-type summaries, kinematic profiling, and generation of publication-ready figures and supporting output tables. The repository contains code only; the underlying trajectory and derived research datasets are not included. Users wishing to reproduce or adapt the analyses should provide their own trajectory data in a structure compatible with the input variables expected by the scripts and adjust the input/output paths and configuration parameters as required.