Microscopic Traffic Simulation of Automated Driving: Modeling and Evaluation of Traffic Performance

To use microscopic traffic simulation to evaluate the impact of ADSs on traffic performance, driving models need to be able to simulate driving decisions and behavioral patterns of ADSs. Driving models have been proposed specifically for ADSs, however, it remains to be validated whether these driving models when used in combination with traditional human driving models adequately simulate mixed traffic that includes human drivers and ADSs. Ideally, a clear interpretation of the behavioral assumptions for each type of vehicle should be possible, as these determine the simulation results. However, it is challenging to compare behavioral assumptions when using different driving models to describe different vehicle types. Empirical research has validated that some driving models, such as the intelligent driver car-following model (IDM), are well-suited for describing both human or automated driving when calibrated with the proper data.

The aim of this thesis is two fold: to further develop microscopic traffic simulation for the study of mixed traffic, and to evaluate the effects of mixed traffic on motorway traffic performance. To enhance the modeling of mixed traffic, a model for perception is proposed which allows the explicit inclusion of perception errors in driving decisions. Its use, in combination with driving models capable of describing both human and automated driving, enables to make distinctions between human drivers and ADSs both in perception capabilities and in driving behavior. This modeling approach focuses on describing essential differences to simulate mixed traffic and removes risks involved in using different driving models.

Simulation experiments are conducted using state-of-the-art tools to evaluate the modeling of perception errors on traffic flow dynamics and to evaluate the effects of mixed traffic on motorway traffic performance.