Determination of Risk Perception of Drivers Using Fuzzy-Clustering Analysis for Road Safety

Abstract

This study aims at investigating drivers' risk perception ability. To achieve this objective, the risk sensitivity and risk judgment thresholds of drivers of different ages were calculated. Five scenarios of intersections with risks were established for the driving simulator experiment. The driving behavior data of fourteen younger drivers and fourteen elderly drivers during the risky event and the subjective evaluation of risk by an expert driver were collected. The expert driver's conflict degree and subjective feeling were combined to classify the risk level of driving scene; then, fuzzy signal detection was used to calculate the driver's risk sensitivity (d') and judgment threshold ( \beta ). The \beta with the greatest difference was selected for cluster analysis and drivers were divided into four types according to the threshold. Finally, a driver classification discriminant model was constructed based on Fisher discriminant analysis. The results show that d' and \beta of younger drivers are both better than those of elderly drivers, younger drivers can detect and respond risks in time, while elderly drivers need be closer to risks and have intuitive feelings to make judgments. The results of the cluster analysis showed that younger drivers account for a large proportion of the sensitive type, indicating that younger drivers can find risks more sensitively in risk scenarios than elderly drivers, while elderly drivers easily ignore risks due to physical and psychological weakness. The correlation analysis showed that age, saccade amplitude and heart rate are the main factors that affect the risk judgment threshold.