Calibration of a microscopic traffic simulator

Abstract

A systematic calibration study was performed on a microscopic traffic simulator- MITSIM. An optimization based framework was developed for calibration. Car- Following model parameters were identified for calibration and experimental design methodology was used to determine the set of sensitive parameters. Calibration was performed by minimizing the deviation between the simulated and observed values of speed. Two different objective function forms were formulated for quantifying the deviation between the simulated and observed values. The search space and the optimum parameter values for the two objective function forms were compared. The effect of stochasticity in calibrating the parameter values was also studied. Stochasticity was found to have a significant impact on the optimal parameter values. It was found that though calibration is an intricate process, the performance of the simulator can be substantially improved by an appropriate calibration study.