Calibration of car-following models using Kalman filters

Abstract

Calibration and validation, especially of microscopic traffic flow models is a challenging task. Currently, ther is a renewed interest in results of those efforts [1, 2] as well as in developing the right tools to actually perform the calibration. The approach developed in [1] will be refined, extended and compared to an approach based on the recent development of a so called unscented Kalman filter [3]. These methods will be used to calibrate a couple of microscopic traffic flow models to two sets of freeway data, one US-American (the I-880 FSP-dataset) and an European dataset, data from the German freeway A100.