Using Trajectory Data to Analyze Intradriver Heterogeneity in Car-Following

This paper presents a methodology for using vehicle trajectory data to study the intradriver heterogeneity of driving behavior between the acceleration process and the deceleration process. Trajectory data were collected during peak hours on Dutch Motorway A2. Criteria were proposed for the selection of subtrajectories corresponding to both the acceleration and the deceleration processes of car-following. With the application of these subtrajectories to calibrate three types of models (the Helly model, the Gipps model, and the intelligent driver model), it was found that obvious intradriver heterogeneities existed in driving behaviors between the acceleration and deceleration processes of car-following: (a) the average response time of drivers in the acceleration process was longer than that in the deceleration process according to the prediction of the Helly and intelligent driver models; (b) drivers were apt to respond more intensively to surrounding traffic in the deceleration process than they did in the acceleration process; and (c) more than 65% of drivers involved in this study drove in obviously different styles between the acceleration and deceleration processes. Moreover, a compensation for the response delay from model parameters was observed, and all three models presented low robustness in predicting driving behaviors of one car-following process with the parameters optimized from the data of other car-following processes. This work not only provides insight into intradriver heterogeneity in car-following behaviors but also suggests some important criteria for car-following modeling.