Microscopic 2D-modeling of driver behavior based on trajectory data from real traffic, driving simulation and traffic simulation

Microscopic traffic flow models are an essential tool for understanding how driver behavior influences the safety and efficiency of road traffic, and how road infrastructure can be designed to support safe and efficient driving behavior. Microscopic traffic flow models describe a vehicle's trajectory in two dimensions - longitudinal and lateral - depending on surrounding traffic and road layout. To reduce modeling complexity, different model types focus on specific influencing factors. Car-following models primarily consider the effect of the leading vehicle on driver behavior, without accounting for road layout or vehicles in adjacent lanes. Speed prediction models describe speed choice based on road geometry, without incorporating surrounding traffic. Lane change models capture driver behavior before and during lane changes, taking into account vehicles in the current and adjacent lanes, but not the road layout. To calibrate and validate microscopic traffic flow models, various sources of trajectory data can be used. Real-world trajectory data yield the most realistic representation of actual driving behavior, but they do not allow for the study of behavior under controlled conditions. Driving simulator data, on the other hand, make it possible to investigate such controlled conditions. Additionally, synthetic trajectory data from traffic simulations can be used to analyze model properties under idealized or specific conditions. This dissertation investigates how trajectory data from real-world driving, driving simulation, and traffic simulation can contribute to more accurate modeling of driver behavior in two dimensions. The overarching research question is divided into five sub-questions, each addressing specific challenges associated with individual model types and data sources. For car-following models, a calibration and validation method was developed that enables the investigation of individual differences between drivers as well as their behavior in extreme traffic situations using a driving simulator. A speed prediction model was applied to evaluate a speed reduction measure in freeway off-ramps based on real-world trajectory data. For the modeling of driver behavior in freeway on-ramps, a lane change model was developed and validated using both real-world and simulated trajectory data. Furthermore, data processing methods and quality assessment procedures were developed for real-world trajectory data, enabling their use in real-time applications as well as for model calibration and validation. The models and methods developed in this dissertation can be applied in traffic simulations to predict traffic conditions, evaluate planned infrastructure or traffic control measures, or assess the impact of automated driving on traffic flow and safety. They thus make a valuable contribution towards the development of safe, efficient, and future-ready road systems.