Investigating heterogeneous car-following behaviors of different vehicle types, traffic densities and road types

Abstract For realistic traffic flow simulation and analysis, it is essential to discuss car-following behaviors in detail, while most studies only concentrate on the heterogeneity according to vehicle types based on the unstable observable variables. To overcome such shortcoming, this paper introduces car-following properties to represent car-following behaviors in different driving conditions. Based on a naturalistic trajectory dataset named The Next Generation Simulation Program (NGSIM), vehicles are divided into 7 classes based on driving conditions. The car-following model parameters of each vehicle, which have practical meanings, are considered as the car-following properties, and are calibrated by genetic algorithm (GA). And then, the heterogeneity of property is discussed based on Kolmogorov-Smirnov test (KS-test) and the rangeability to understand car-following behaviors. Results show the heterogeneity of car-following behaviors according to vehicle type and road type, while traffic density has tiny effect on car-following behaviors. And car-following properties are also heterogeneous, as the comfortable deceleration is stable, while other properties change differently with driving conditions. Besides, there is a similar pattern of change on vehicle properties and vehicle trajectories. Furthermore, potential applications of the research are provided, for instance, model more realistic traffic simulations and give support to adaptive cruise control systems (ACC) for vehicles in different conditions.