Calibration of Microscopic Car-Following Models on the Basis of the Intelligent Cruise Control Field Operation Test Data

Abstract

Simulations with vehicular traffic are an important tool forecast the impacts of Intelligent Transportation Systems on traffic flow and safety in road networks. When choosing microscopic models the focus is on describing driver behavior in car following. These models can be regarded as key component of a comprehensive simulation. For an accurate reproduction of driver behavior these models have to be calibrated. This paper presents a calibration process of two microscopic car-following models: theintelligent driver model (IDM) [10,11] and the Krauss model (SK) [5]. Simulation results of both models already were compared with data obtained from measurements on German freeways and arterials. In our paper the models are calibrated on the basis of U.S. roadway data from the Intelligent Cruise Control Filed Operational Test (ICC FOT) [3,4]. The analysis of the calibration shows that both models generally perform well in reproducing the recorded driver behabior. However, in special driving situations the models have problems to match the data, due to model limitations. Therefore an extended microscopic traffic model (SKCONT), developed by the Institute of Transportation Research of the German Aerospace Center, also is taken into cosideration. This ODE based model explicitly describes continuous driver behavior. Thus, its time discrete version better copes with the short time step size of the ICC FOT data than the two first models. The amount of data and the huge number of participating drivers in the ICC FOT also permit to obtain several parameter sets for the models, curresponding to different driving scenarios or driver groups. These results are finally presented in a comprehensive parameter statistics and show the application fields for microscopic simulation of intelligent transportation systems such as automated cruise control and driver assistance.