Dynamically Dimensioned Search Parameter Calibration for Microscopic Traffic Simulation Models and their Effect on Measures of Safety Performance

摘要

The Pareto Archived Dynamically Dimensioned Search (PA-DDS) algorithm is introduced in thispaper as a method to calibrate microscopic traffic simulation platforms. This algorithm wasoriginally developed to calibrate hydrological rainfall/runoff microscopic simulation platforms. Inthis study, the algorithm is applied to the VISSIM traffic simulation platform to calibrated freewaydriving behaviour. Data from the Federal Highway Administration (FHWA) Next GenerationSimulation (NG-SIM) is used. The following three objectives are used in the calibration: ⅰ) rootmean-squared-percentage-error (RMSPE) of speed, ⅱ) RMSPE volume, and ⅲ) RMSPE CrashPotential Index per vehicle (a surrogate safety performance metric). Four other experimentswere also undertaken, and are: 1) single-criteria using RMSPE speed, 2) single-criteria usingRMSPE volume, 3) single criteria using RMSPE CPI/vehicle, and 4) weighted summation(RMSPE speed + RMSPE volume + RMSPE CPI/vehicle). The case study demonstrates thatthe PA-DDS algorithm provides acceptable errors for all three objectives compared to the othermethods.