Driver/vehicle Response Diagnostic System for Vehicle Following Based on Gaussian Mixture Model

It is well known that not all drivers drive the same and the same driver has different driving characteristics with different vehicles. Identifying these characteristics that are unique to each driver/vehicle response opens the way for more personalized and accurate driver assistance systems. In this paper we consider the problem of identifying the driver/vehicle characteristics by processing real data offline. We propose the use of a Gaussian Mixture Model (GMM) together with additional logic and appropriate thresholds. We concentrate our efforts on identifying the driver/vehicle response model in the vehicle following case. Model training using data retrieved through experiments along with comparing data sets for different drivers indicates that the system is capable of identifying the driver/vehicle response characteristics and detecting deviations from normal driving behavior. The system has been demonstrated to distinguish between drivers after it learned their characteristics.