Roadway Traffic Collision Analysis and Mitigation Using Data Analytics and Microscopic Driver Behavioral Modeling Open Access
Downloadable ContentDownload PDF
With the billions of vehicle miles traveled and millions of vehicular collisions occurring annually, collision mitigation and prevention are of top priority. To fully grasp the impact of collision occurrence on roadways, both the macroscopic and microscopic scales need to be addressed. On the macroscopic side, the traffic parameters and dynamics that lead to and result from collision formation need to be understood from the link level to the network level. On the microscopic side, the behavioral actions of individual drivers and the interactions amongst drivers should be considered. Despite the great strides that have been made in data analytics and traffic modeling, there is still room for improvement and a gap still exists between these two scales. With this motivation, the main objective of this research is to develop, calibrate, and validate a microscopic traffic simulation model capable of accurately capturing car-following, lane-changing, and collision movements with and without the incorporation of connected and autonomous vehicles on both microscopic and macroscopic scales. To accomplish such an objective, this dissertation first calculates the collision involvement rates and produces macroscopic fundamental diagrams for a variety of real-world study locations. This allows for insight to be gained into the effects of link and network level aggregation of target metrics. Next, a microscopic traffic simulation model is implemented that incorporates the psychology and heterogeneity of human drivers with acceleration, vehicle dynamics’ limitations, lane-changing with a lateral trajectory component, collision formation, post-collision movement, and a basic connected and autonomous technology module. A driving simulator experiment is used to collect data from young drivers (ages 15 – 23) during car-following and near-collision scenarios with various audio and audio-visual collision warnings. Using this data, the implemented traffic simulation model is calibrated. The calibrated model is then validated against real-world collision rate data. Finally, a brief case study is used to achieve a proof of concept for the implementation of connected and autonomous vehicles with the inclusive microscopic traffic simulation model.Results show that the implemented model can accurately reproduce collision formations and collision rates for the selected testing scenarios. This leads to the conclusion that the presented model has high potential regarding uses as a simulation tool. Differences between high school and college student drivers’ response times and collision occurrence are found for the tested collision warnings. These differences suggest that auditory with visual informative alphanumeric warnings are best suited for college students and auditory with non-informative symbolic visual warnings are best suited for high school students. The simulation results using calibrated parameters from the driving simulator experiment suggest that collision involvement rates for both rear-end and angled collisions reduce to zero with the use of an auditory with non-informative alphanumeric visual warning.