The primary objective for this Exploratory Advanced Research project is to characterize driver behavior under naturalistic driving experiences with respect to critical parameters related to freeway driving. To accomplish such an objective, the research team will develop agents that will learn individual driver's temporal actions for any given traffic state retrieved from the naturalistic database. These characteristics, or driving rules, of the agents will be coded in the VisSim environment to test and study the collective effects of these learned behaviors with multiple drivers involved and under different situations. Analysis of trained agents' characteristics will provide the industry with methods for developing more accurate and sensitive traffic simulation models, which could lead to future research for developing new generations of traffic simulation models to accurately model impacts of driver behavior in traffic during incidents and complex traffic situations.
The primary objective for this Exploratory Advanced Research project is to characterize driver behavior under naturalistic driving experiences with respect to critical parameters related to freeway driving.
This research provides a foundation for agent-based modeling of driver behavior based on naturalistic data through an integrated framework for safety and operation analysis. Lateral vehicle action was simulated in a microscopic traffic behavior modeling environment, bringing new insights to the modeling of driver maneuvering behavior during safety-critical events. Agents developed and evaluated in the VisSim (a German acronym that refers to a simulation model for traffic in cities) simulation platform revealed a close resemblance to real driver data. The project team improved car-following models through development of a hybrid model for greater accuracy and flexibility and through the addition of the new "passing and hook-following" thresholds. They used the model to simulate vehicle actions in safety-critical events, developed agent-based simulation components integrated with the VisSim simulation package through its driver model, and developed and implemented a robust activation mechanism for agent-based simulation based on discriminant analysis. The investigators also identified key future research issues: adaptability of agents in real time and human factors issues related to warning individual drivers about changes in their driving behavior that might lead to safety-critical events.