Determination of driver's behavior for car-following model using GPS data

Abstract

Car following is the major driving task reflecting individual vehicle movements and interactions among vehicles in a single lane, resulting in overall characteristics of traffic flow. Thus, determination of the driver's behavior is fundamental to accurate explanation of traffic flow. In this research, the driver's behavior was determined under different traffic conditions in Thailand, especially expressway and surface street using GPS technology. The vehicle trajectories were collected by RTK-GPS device at 0.1 second interval with accuracy of less than +- 1 cm. The speed and acceleration profile were then constructed. Drivers behaviors described in the GM car following model parameters, namely reaction time and sensitivity parameters, were determined using Graphical method and Linear regression method, respectively. The results indicate that the reaction time of a driver varies by traffic and road conditions. In congested traffic condition, the reaction time for acceleration is longer than that for deceleration. However, the reaction time for acceleration is shorter than that for deceleration in uncongested traffic condition. The result of the 5th GM model calibration shows that the sensitivity parameter of drivers is inconsistent. Compared among subject drivers, some drivers are more aggressive in a more critical (close spacing, high speed) situation but less aggressive in a non-critical situation. The research brings about the better understanding of driver's behaviors