Modeling Drivers' Behavior During Panic Braking for Brake Assist Application, Using Neural Networks and Logistic Regression and a Comparison

摘要

Researchers have shown that unskilled drivers fail to apply sufficient force on brake pedal in emergency. To solve this problem, Brake Assist System (BAS) is used to enhance the vacuum brake booster performance and results decrease in stopping distance. A major problem in BAS is to determine if a panic braking has been occurred or not. In this study, a model of drivers' behavior during a severe braking is created using both neural networks and logistic regression methods to determine the BAS threshold activation. Samples of brake pedal speed, Brake pedal displacement, and vehicle acceleration measured from panic and normal situations, will be fed for training neural networks and acquiring logistic regression equation. From both methods, the probability of a panic and normal situation will be determined. By using MATLAB software, the result from these two models is compared, and the one that is quicker in detecting panic situation and simple for implementation, will be chosen for BAS activation threshold.