Hybrid Steering Model depending on Driver's Gazing Point to detect inattentive driving using Machine Learning

The modeling of driving behaviors is important to analyze and design comfortable functions for the driving systems. The estimation method of the non-linear steering model using the eye tracking information was studied using the heuristic search algorithm. However, the model was limited to gaze and the inattentive driving was not modeled. It is considered that the steering model is consists of a hybrid steering model that switches the controllers according to the eye tracking information is classified in the effective/peripheral viewing field. In this paper, an estimation method of the hybrid system focusing on the effective visual field during driving is proposed. The hybrid model is constructed by steering model depending on the gazing distance and simple on-off controller. This estimation algorithm consists of 2-steps; clustering which classifies data by k-means method and the estimation of the parameters by Particle Swarm Optimization. The experiment with the long driving course consisting of five-curves and straight lines is demonstrated by HONDA driving simulator.