Examination of driving behavior simulation using risk potential parameter estimation method

Abstract

In the risk potential driver model used to generate path control targets for autonomous vehicles, traffic elements such as obstacles and traffic rules are represented by risk potential, and safe routes are determined by synthesizing them. Therefore, it is possible to calculate the control target path adapted to the individual by quantifying the feel of risk that the human driver perceives and using it as a parameter of the driver model. This paper describes a method for estimating risk potential parameters using human driving maneuver in the overtaking situation and confirm of effectivity. A travel path measured by the simple driving simulator was functionally approximated to create a smooth trajectory as a sigmoid curve, and the steering operation to obtain this trajectory was calculated. The risk potential parameter was calculated by the nonlinear least squares method using the relationship between the deviation of the risk received from the left and right and the steering angle. Estimated parameters are defined as random variables. Also, the representative values for the parameters are average values and apply to the driver model. As a result, it was confirmed that the driving behavior of each driver could be reproduced with high accuracy.