Dynamic traffic safety risk assessment in road tunnel entrance zone based on drivers' psychophysiological perception states: Methodology and case-study insights

Abstract

This study endeavored to accurately and comprehensively assess dynamic traffic safety risk at the road tunnel entrance zone. First, vehicle speeds and multiple types of psychophysiological indicators of drivers are collected in real-vehicle tests at different times and various measurement points in the tunnel entrance zone, and the drivers' psychophysiological perception states and change trends are analyzed. Second, a Traffic Safety Risk Value (TSRV) is quantified in terms of the difference in safe speeds and a traffic safety risk model that is established for the tunnel entrance zone. Fuzzy C-means clustering algorithm is used to divide the threshold of TSRV into three traffic safety risk levels. Subsequently, through the optimization of three machine learning models, the dynamic traffic safety risk assessment model is constructed based on the optimal Decision Tree. Through further model hyperparameter optimization and pruning, the relationship between drivers' psychophysiological perception states and traffic safety risk levels is quantified. Finally, a questionnaire survey is used to obtain drivers' subjective feelings about traffic safety risks while driving in the tunnel entrance zone. The effectiveness of the assessment model is verified by combining driver's subjective feelings and objective physiological responses. The results show that the traffic safety risk identification accuracy of the machine learning model proposed in this study reaches 95.13%, and the model can dynamically assess the real-time driving risk level. The findings have practical implications for the prevention of traffic crashes in the tunnel entrance zone and the realization of safe and stable operation of road tunnels.