Driving Fatigue in Extra-Long Highway Tunnels and a Bayesian Networks-Based Fatigue Detection Method

Abstract

Highway tunnels are semi-enclosed traffic structures distinguished by insufficient illumination and monotonous surroundings. In these conditions, prolonged drives often culminate in driver fatigue, which has become a prominent causal factor in tunnel accidents. This study conducted real-vehicle experiments and simulated driving tests to investigate the characteristics of driving fatigue during passage through extra-long highway tunnels. Four tunnels with lengths of 5, 10, 15, and 20 km were used for the experiments. Six indicators derived from drivers’ eye movement, electrocardiogram (ECG), and electroencephalogram (EEG) signals were selected to examine the fatigue progression. The experiment results indicate that when driving in extra-long tunnels exceeds 500 s, the drivers experience mild fatigue, as evidenced by deviations of their physiological indicators from the normal ranges. After driving beyond 700 s, the mild fatigue progresses into severe fatigue, with significant deviations of the physiological indicators and dramatic increases in the fluctuations of these indicators. On exiting the tunnels, the mild fatigue recovers immediately, while the severe fatigue is not fully alleviated and continues to exert negative impacts on subsequent driving. Based on the experimental findings, a multistate Bayesian networks (B-Ns) method for detecting driving fatigue in extra-long tunnels is proposed. The method is compared with the Karolinska Sleepiness Scale (KSS) rating results and previous single-indicator detection methods, demonstrating that the proposed B-Ns method achieves precise detection outcomes and substantially enhances the accuracy of fatigue detection, particularly when drivers are experiencing severe fatigue accompanied by strong fluctuations of the physiological indicators.