Relationship between truck driver fatigue and rear-end collision risk

Abstract

The fatigue of truck, bus, and taxi drivers has been a causal trigger for road accidents. However, the relationship between collision risk and the extent of objective fatigue has yet to be confirmed. In this study, we aimed to identify the relationship between autonomic nerve function as an objective parameter of fatigue and the extent of rear-end collision risk, which includes not only objectively risky events but also situations in which truck drivers require safety guidance from safety transport managers. Data of 33 truck driver participants (2 females, 31 males, 46.0 ± 9.1 years old, min-max: 24-65 years old) were analyzed. Drive recorder and automotive sensor data were collected over an eight-month period, and the autonomic nerve function during resting state in drivers was evaluated daily, pre-and postshift, using pulse waves and electrocardiographic waveform measurement. The rear-end collision risk Index was developed using decision tree analysis of the audiovisual drive recorder data and distance data from the front automotive sensors. The rear-end collision risk index of shift-day was positively correlated with the sympathetic nerve activity index of post-shift condition on the previous day. This suggests that fatigue-related sympathetic nerve overactivity of post-shift condition increases the rear-end collision risk in the following day. Measures, such as actively seeking rest and undertaking fatigue recovery according to the degree of sympathetic nerve activity of post-shift condition, are necessary in order to prevent truck drivers' rear-end collisions.

Increase in rear-end collision risk by acute stress-induced fatigue in on-road truck driving

“…The obtained RRI time-series were divided into windows of 120 s each, and ANF was calculated for each window. In the pre- and post-shift conditions, RRI and ANF were obtained using the fatigue stress measurement system (VM500, Fatigue Science Laboratory, Osaka, Japan), which simultaneously measures electrocardiogram and photoplethysmogram, in the resting eye-closing state for 90 s with 600 Hz sampling [ 10 ]. Under all shift conditions, any windows in which misdetection or abnormal intervals exceeded 10% of the total beats, or the calculated heart rate showed outliers, were excluded from ANF calculation due to the instability of the measurement and the low reliability of calculated ANF.…”

“…In ANF calculation, various RRI features, such as frequency-domain features were first obtained as the ANF indicators using a similar method as that described in our previous study [ 10 ]. Briefly, frequency-domain features were obtained by integrating some frequency bands in the power spectral density (PSD) based on the maximum entropy method, which is adequate for estimating PSD from short-duration RRI [ 26 ].…”

“…In addition, the LF/HF ratio, which represents the balance between sympathetic and parasympathetic nerve activity, was also calculated. LF and HF are affected by heart rate and aging [ 27 , 28 ], and as we cannot control drivers’ heart rates and ages in actual working shifts, we utilized the devised score of LF and HF using a method described in previous studies [ 10 , 29 , 30 ]. The LF deviation score LF score was defined by the following formula: where, μ LF ( age ) and σ LF ( age ) are the mean and standard deviation of age , respectively, in the LF standard distribution N LF; age that approximates a normal distribution.…”

“…For example, the Japan Trucking Association is promoting measures to check drivers’ health before they begin their shift, as a means to prevent crashes among truck drivers [ 8 , 9 ]. Moreover, our research group has also reported that evaluating drivers’ fatigue level, an indicator of physiological condition, before and after work, is effective in estimating the risk of rear-end collision by truck drivers [ 10 ].…”

“…Therefore, it is necessary to analyze data obtained during the operation of actual vehicles. Meanwhile, due to such crashes being rare occurrences, it is difficult to obtain sufficient historical data on crashes, including physiological condition data, during actual vehicle operation within the experiment period [ 10 ]. Moreover, since the degree of latent risk in the event of a crash not occurring is not represented in the crash history data, it is not suitable for evaluating the effects of changes in physiological condition on crash risk.…”