Metamodel-based Calibration of Microscopic Traffic Simulation: A Vissim Case Study

Model calibration ensures the reliability of traffic simulation models for designing and evaluating traffic management schemes. This paper introduces a metamodel-based method for calibrating a microscopic simulation model developed with Vissim. Initially, a Vissim simulation was devised for an expressway merging zone. A parameter sensitivity analysis was conducted using a backpropagation neural network (BPNN). Several key parameters of the Vissim model were identified based on sensitivity indices. A metamodel called the Kriging model was created to show how Vissim parameter inputs affect simulation outputs. Based on the metamodel, the differential evolution (DE) algorithm was applied to find the best parameters for the VISSIM mode. The average travel time simulation of the calibrated Vissim model deviated from field survey data by less than 3%. The statistical analyses affirm the efficacy of the Kriging-DE calibration method. The metamodel-based method offers a promising tool for model calibration in traffic management.