Driving Feature Extraction from High and Low Skilled Drivers in Curve Sections Based on Machine Learning

Abstract

We have developed a new driving assistance system that can help low-skilled drivers improve their driving skills. We did this in three steps. First, we developed a statistical method to extract distinctions between high- and low-skilled drivers on the basis of AdaBoost, which selects a small number of critical operation features between high- and low-skilled drivers. Second, we built a driving skill evaluation model on the basis of the extracted features. Finally, we performed a series of experiments using a driving simulator, in which advice based on extracted features was supplied to low-skilled drivers and was expected to improve their driving skill. We also proposed an index for evaluating driving skill change, and results show the advice effectively improved the drivers' driving skills.