AN INTELLIGENT DRIVER WARNING SYSTEM FOR VEHICLE COLLISION-AVOIDANCE

This paper describes the basic architecture of an intelligent driver w arning system which embodies an adaptive driver model for indirect col lision avoidance, In this study, the driver modeling objective is focu sed only on longitudinal car-following, and the model inputs are chose n to be the past history of throttle angle, controlled vehicle's speed , range and range rate to the front vehicle whereas the model output i s chosen to be the current throttle angle. An artificial neural networ k called Cerebellar Model Articulation Controller (CMAC) and a convent ional linear model (CLM) are independently applied to model the real d river data taken from test track and motorway environments, The CMAC m odel is chosen because of its nonlinear modeling capability, on-line l earning convergence and minimum learning interference characteristics, whereas the linear model is chosen as a control benchmark to examine the nonlinear characteristic of the driver's behavior. The modeling ca pabilities are then evaluated based on one-step ahead prediction error performances over the training and testing sets, learning curves and correlation based model validation techniques, Modeling results sugges t that the past history of throttle angle plays a critical role in red ucing the deviation of the error correlation, which in turn suggest th at the throttle dynamics are generally slow for road driving. Also, th e time scale dependency of the model on the driver's behavior varies s ignificantly from the test track to motorway environment, In the drive r modeling experiment, the time scale was chosen such that the deviati on of the error correlation was minimized, The test track results sugg est that the chosen inputs are indeed relevant variables for modeling the driver's behavior, Unlike that of the CLM, the degree of the error deviation of the CMAC model was found to be acceptable for the test t rack scenario, implying a significant nonlinear coupling of the thrott le output with the speed, range and range rate data, Whereas for the m otorway data, the modeling performance for both models is comparable, and the time scale of the driver model is approximately three times lo nger than that used in the test track data.