Identification and verification of a longitudinal human driving model for collision warning and avoidance systems

The main contribution of this paper is the identification of a human driving model based on field measured car-following data, and the verification of the model's performance in a microscopic traffic simulator. We first examined the SAVME database and obtained a well-defined set of data under closing-in and decelerating lead-vehicle scenarios, i.e. the transient manoeuvre starts with a large range and negative range rate toward the equilibrium point with proper range and zero range rate. Subsequently, the ICC FOT database is used to extract model parameters under normal highway driving conditions. These well-defined data sets are then used to test the flexibility of several existing driving models, i.e. the model parameters are tuned to fit these data. The Gipps model was found to be able to fit the highest number of manoeuvres, and the identified parameters are used to represent human-controlled vehicles, which are deterministic but have different attributes (aggressiveness, target speed, etc.). The Gipps model and the parameter sets are then implemented in a microscopic traffic simulator. Macroscopic and microscopic behaviours of these simulated human-controlled vehicles are presented.