Journal of Korean Society of Transportation ISSN:1229-1366(Print) 2234-4217(Online) 대한교통학회지

Recently demand for traffic simulation for development and validation of the safety performance of new system increases along with the development of autonomous vehicles. However, the existing traffic simulations with the current car-following models have limitations as they cannot simulate risky situations including vehicle collisions. In this paper, we propose a car-following model to demonstrate the driver’s behavior and simulate traffic accidents in line with the demand. Among the various types of accidents, we focus especially on the rear-end collision caused by the driver’s distraction. To model the driver’s distraction, we firstly analyze the diver’s glance behavior using the 100-Car Naturalistic Driving Study data provided by the Virginia Tech Transportation Institute. Various elements affecting the driver’s glance behavior are divided into road environmental and vehicle movement variables according to their temporal range and analyzed differently. Based on the result, a hierarchical decision tree analysis is conducted to derive the driving scenarios by glance behavior. Eleven scenarios are derived, and the driver glance models are developed for each of them. We propose the distracted-OFFA by combining the driver glance model with the base car-following model of Oversaturated Freeway Flow Algorithm (OFFA). The results show that the proposed model can regenerate risk situations including rear-end collisions, and this study has contribution in that it shows the potential to simulate risk situations based on driver’s behavior and to predict accident occurrences. Accordingly, meaningful achievements are expected with further researches.