Exploration of Car-Following Model in Emergency Evacuation Situations

Abstract

This paper presents results of an exploratory study on building up the car-following model under emergency evacuation situations. A simulated scenario was used to create a driving environment in emergency evacuation conditions, and a questionnaire investigation at the Miriam ECG was used to verify the validity of the driving environment from both the subjective and objective aspects. A Backpropagation (BP) Neural Network was designed, and a gray-correlation analysis was conducted to determine which factors have more impact on the acceleration of the following car. Acceleration of the following vehicle was set as the mother factors series, and speed of the following vehicle, speed of the leading vehicle, speed difference between vehicles, headway between vehicles, and acceleration of the leading vehicle were set as the sub-factors series. Finally, acceleration of the leading vehicle, speed difference between vehicles, and speed of following vehicle were determined to be the three most important factors. These three factors were selected as input variables. The acceleration of the following car was selected as the output variable. Simulation of the BP Neural Network using data collected from the driving simulator revealed that the BP Neural Network has a high precision in the prediction of the car-following model.