How natural light influences HSR drivers' visual behavior.

Existing studies have shown that the lighting environment is essential in influencing a driver's visual behavior. Due to the pivotal role of high-speed railway (HSR) in worldwide transit, it is necessary to examine how HSR drivers' visual behavior adjust under different lighting environments. However, the methods for evaluating and categorizing lighting conditions have not been fully explored. In this study, we established a general framework for examining the impact of lighting on driver's visual behavior. The application of this framework to explore the effects of natural light on HSR drivers' visual characteristics was elaborated. Particularly, we used unsupervised machine learning methods to classify natural light conditions automatically. Specifically, Fuxing HSR simulation, illuminance meter, and Tobii Nano eye-tracker were employed to collect data. K-means clustering analysis of daily illuminance data identified 3 natural light conditions, namely low illuminance (1 -6 ), medium illuminance (6 -9 ), by and high illuminance (9 -1 ). Further, ANOVA with 3 natural light environments * 2 tunnel conditions * 4 areas of interest (AOIs) were conducted. Results manifested drivers' visual characteristics under different natural light conditions. Specifically, lower illuminance can lead to a wider average pupil diameter, while higher illuminance results in a greater number of fixations and saccades, and a shorter time to first fixation. Moreover, all the eye movement indicators are highest for the speed dial AOI. This study contributes to the field by developing a framework to examine the effects of lighting on drivers' visual behavior. The findings provide new insights into analyzing lighting environments by using machine learning methods, which servers to HSR driving safety and operational management.