Automated driving is regarded as a potentially disruptive technology, and is expected to bring fundamental changes to our transportation systems, ranging from increased road capacities over reduced accident rates to impacts on travel behavior and car ownership. Even though partially automated driving is already possible with current factory-fresh vehicles, anticipating the long-term impacts of vehicle automation is vital for both traffic scientists and road authorities, especially when considering that there will most likely be a long gradual transition from manual to automated driving, and thus a long period of time in which both human-driven and automated vehicles share our roads. Since field experiments are scarcely feasible in this particular context, simulations have emerged as an important instrument to perform such kind of investigations. The present thesis adopts this methodology, and develops a conceptual framework to facilitate the ex-ante evaluation of automated driving and its impacts on traffic flow. The proposed framework is formulated in a rather generic way, and is intended to enable traffic scientists to systematically augment existing models for driving behavior with a variety of behavioral traits and technology-appropriate assumptions to distinguish between human-driven and automated vehicles, but also between different degrees of vehicle automation. The first part of this thesis focuses on the various factors governing the behavior of human drivers, and incorporates some of these factors into a microscopic car-following model. Thereby, a particular focus lies on aspects related to distracted driving, and a novel approach is presented to incorporate the dynamics associated therewith into traffic simulations. Furthermore, we propose a meta-model to differentiate between different degrees of vehicle automation, but also to take into account technological constraints and limitations associated with automated driving. Subsequently, the microscopic traffic simulator TraffSim is introduced, which integrates all models developed in the scope of this thesis into a single simulation framework. In the final part of this thesis, we illustrate this framework by evaluating the impacts of automated driving on the stability of traffic flow, traffic safety, and efficiency under consideration of mixed traffic flows, that is, assuming different penetration rates of automated vehicles. For our simulations we consider a single-lane scenario comprising a platoon of vehicles following an exogenous leader as well as a large-scale scenario with multi-lane traffic and road intersections. Our findings provide evidence that automated driving may indeed result in safer and more efficient traffic operations, however, they also reveal that vehicle automation might initially even have a slight negative effect on traffic flow stability, especially at low penetration rates. With regard to traffic safety, our results indicate that even at very high penetration rates of automated vehicles and a high level of automation there still remains a residual risk of road accidents caused by distracted drivers, suggesting that a full elimination of human error seems to be a fundamental prerequisite to reach the ultimate goal of “vision zero” on our roads, i.e. to effectively avoid fatalities and serious injuries in road traffic. Overall, the findings presented in this thesis provide reasonable grounds to believe that humans might actually not be too bad at driving at all, and that considering the “human factor” is therefore indispensable when studying the potential impacts of automated driving on traffic safety and efficiency.
|Publication status||Published - 2019|