Modeling of drivers collision avoidance behavior based on hybrid system model: an approach with data clustering

This paper presents a development of the modeling of the human driving behavior based on the expression as hybrid dynamical system (HDS) focusing on the driver's collision avoidance behavior. The driving data are collected by using the three-dimensional driving simulator based on CAVE, which provides stereoscopic immersive vision. In our modeling, the relationship between the measured information such as the sensory information of range between cars, range rate and lateral displacement between cars and the output of driver of the steering amount are expressed by the piecewise linear (PWL) model, which is a class of HDS. Then, we solve the identification problem for the PWL model by using the combination of data clustering and support vector machine. By introducing the PWL model, it becomes possible to find not only coefficients in each submodel but also parameters in the logical (switching) conditions from the measured driving data. From the obtained results, it is found that the driver appropriately switches the 'control law' according to the sensory information. This enables us to capture not only the physical meaning of the driving skill, but also the decision-making aspect (switching conditions) in the driver's collision avoidance behavior.