Evaluating the impacts of in-vehicle warnings at roundabouts with drivers’ performance and eye movement data

Abstract

Over half of all traffic accidents that result in fatalities and injuries are intersection related. Roundabout, as a special type of unsignalized intersection, is a challenging scenario for human drivers. With the development of intelligent vehicle technology, more vehicles are equipped with sensors that can monitor both traffic environments and drivers’ status and generate real-time warnings to assist drivers in responding to hazardous situations. This study aims to investigate drivers’ reactions to an in-vehicle advanced warning system through a driving simulation study. A real-world roundabout was built and calibrated in the simulator and both driving performance and eye movement data were collected from the experiments. The results indicated that advanced warnings can effectively influence vehicle speed, steering wheel control, and drivers’ attention on different areas of interests (AOIs). It was found that proper warning time was critical to improve drivers’ safety and comfort. Gender differences were also identified from both types of data. Finally, to better facilitate the design of the personalized warning systems, machine learning models were developed to predict drivers’ perceived risk and minimum TTC. The prediction model for minimum TTC achieved 0.111  of mean square error (MSE) and the risk classifier had 83.5% overall accuracy.