Detection of high-speed railway driver fatigue across shifts scheduling

Irregular shift scheduling is closely related to high-speed rail (HSR) driver fatigue, which affects driving performance and HSR operation safety. This paper attempts to identify the difference of driver fatigue between different shifts through the research conducted on an HSR simulator. A total of 28 time-domain, frequency-domain and nonlinear heart rate variability (HRV) features are extracted. Mann-Whitney U test is used to analyze whether HRV characteristics differ in different shifts. Five classifiers (CNN, LSTM, RNN, RF and SVM) are used for fatigue ternary classification. The results show that most of the HRV features are statistically different across different shifts, and the drivers may be more tired during morning shifts driving. RF and CNN perform best in the ternary classification of driver fatigue. The results of this study are helpful to understand the time-varying difference of driver fatigue and provide reference for optimizing driver scheduling.