Estimation of steering model depending on gazing distance and its evaluation by using driving simulator

Abstract

To analyze the driving behavior is important for designing a comfortable automatic driving. So far, conventional PID models was widely used to steer the vehicle. It has been assumed that the driver looks at the constant point in conventional PID model. However, all drivers have experiences that the gazing distance changes according to the travel routes, driving situations and driver's characteristics. The PID gain of the steering model is assumed to be modified depending on driver's gazing distance. The facts are familiar, but the relationship between driver's steering gain and the gazing distance has not been clarified yet. If this relation becomes clear, it can be helpful for the comfortable driving assistance and/or automatic driving systems to provide for each driver. The aim of this study is to design a steering model estimated by Particle Swarm Optimization (PSO) depending on gazing distance . The feature of the model is on-line measurement of gazing distance incorporating the gazing distance explicitly into a model. The isolated driving and the following driving are evaluated using a Honda driving simulator. The effectiveness of the model is discussed individually.