Classification of Driving Fatigue in High-Altitude Areas

Driving fatigue is one of the main causes of traffic accidents. Thus, to prevent traffic accidents and ensure traffic safety, the properties of driving fatigue at the wheel must be determined. The Qinghai–Tibet Plateau in China is known for its high elevation, causing hypoxia, and presence of severely cold areas; all these easily lead to fatigue during driving. This, in turn, seriously affects the traffic safety on the high-altitude highway. Therefore, the factors leading to driving fatigue and the influence of high-altitude on driving fatigue affecting the driver must be further studied. In this study, we classified and quantified driving fatigue according to the driving fatigue degree. We determined three levels of driving fatigues (i.e., mild, moderate, and severe fatigues) to present their influence on drivers. Our study shows that in this high-altitude area, drivers became fatigued within a significantly shorter time.


Introduction
In 2013, 1.25 million road traffic deaths were recorded worldwide.With the growth of the population and of motorization, the amount of road traffic deaths are said to increase further [1].Driving fatigue is not only a direct cause of traffic accidents but also leads to the most vicious accidents.In the UK, Maycock [2] determined that 7% of motor vehicle accidents could be attributed to fatigue.Fletcher [3] determined that almost 40% of traffic accidents in North America were caused by driving fatigue.Recent studies show that the costs of fatigue-related accidents in the US amount to $31.1 billion [4].All these studies show that researchers have fully realized the importance of preventing driving fatigue for ensuring road safety.
Williamson et al. [5] defined driving fatigue as a state of reduced mental alertness, which impairs the performance of cognitive and psychomotor tasks, including driving.Many factors, such as the natural environment, the driver's mental status, road traffic environment, and vehicle characteristics, could influence driving fatigue.In particular, the natural environment is the most important factor in some extreme areas, including high-altitude areas.
In the Tibetan Plateau, the average elevation exceeds 4500 m.In addition, the natural environment is abominable; the climate is extreme; winters are long, cold, dry, and windy; and the extreme lowest temperature of the day is −48.1 • C. All these factors play a major role in influencing a driver.The current studies have not considered these extreme conditions in the designing and management of this area.
Drivers are unaware of driving fatigue, which decreases their ability to drive safely [6,7].Therefore, we must measure and quantify driving fatigue.Previous studies have proposed several methods for measuring driving fatigue, these include performance, perceptual, electrophysiological, psychological, and biochemical measurements [8].
Some studies have used the measurement of heart rate for measuring driving fatigue.Riemersma et al. [9] found that drivers' heart rate decreased when driving for a long time during the night; this was verified by Lal and Craig [10].Harris et al. [11] determined that heart rate variability (HRV) and the feeling of fatigue are both associated with deterioration in driving.
Furthermore, O'Hanlon [12] found changes in HRV among drivers with long test drives.Hartley and Arnold [13] found that the changes in HRV could be used for indicating driving fatigue.The importance of utilizing HRV as a physiological parameter to assess fatigue for purposes of developing fatigue counter-measurement systems was explored by Lal et al. [14].Patel et al. [15] used HRV as a measure for determining driving fatigue under laboratory conditions and determined that it showed a 90% accuracy.
This paper proposes a method to quantify and classify driving fatigue according to HRV.In addition, we propose a driving fatigue standard based on elevation.

Participants
Ten healthy volunteers (five females and five males) with an age range between 19 to 36 years (mean age of 23.9, standard deviation (SD) of 4.9) participated in this experiment.All subjects had driving licenses (mean of 2 years, SD of 3.2) and were asked to abstain from alcohol and smoking for 24 h and from caffeine-based drinks for 12 h before participating.All participants slept at least 7 h before being evaluated (mean of 7.2 h and SD of 1.2).To avoid the confounding influence of circadian rhythm [16,17] or any diurnal variation [18], all experimental sessions were conducted between 9 a.m. and 12 p.m.

Instruments
We used the CTM-3000E, which is a non-contact speedometer camera, to display real-time test data.It was deployed with the test vehicle to record real-time speed and distance.In addition, we used the PM-60A, a cardiotachometer and oximeter, to record the heartbeat rate and degree of blood oxygen saturation.Moreover, a portable GPS was used to record the elevation.

Procedures
We chose G214, a first-class highway with K145 + 000 − K795 + 000 as the experiment road, and evaluated it from 2000 to 5000 m.We chose four sections of the road with different elevations: Section 4 (K570 + 000 − K610 + 000) with an elevation of 4500-5000 m, Section 3 (K536 + 000 − K564 + 000) with an elevation of 4000-4500 m, Section 2 (K290 + 000 − K225 + 000) with an elevation of 3500-4000 m, and Section 1 (K225 + 000 − K180 + 000) with an elevation of 3000-3500 m. 80% of the all four sections are composed by straight lines and large radius of a circular curves.They have the same road alignment conditions.The longitudinal slope is not greater than 4%.The experiment was performed between 9 a.m. and 12 p.m. to ensure the same road traffic condition.The roadside landscape was monotonous with only grasslands and no landmarks.The experimental vehicle used was the Jinbei Granse multipurpose vehicle.
Before initiating the experiment, we recorded the driver's heartbeat rate and degree of blood oxygen saturation for 30 min, and helped participants get acquainted with the experimental environment and devices.We also ensured that the participants were not stressed before performing the experiment.The experiment was then started and took 1 h each time for completion.We recorded all the data automatically while observing the state of the driver and enquiring about their state of mind.

Data Collection
HRV is the measure of variations in a heartbeat and is calculated by processing the time series of R-R interval in an electrocardiogram (ECG) signal (see Figure 1).Thus, we translated the ECG signals to R-R intervals for further analysis.The procedure is as follows (see Figure 2): (1) Process the ECG data obtained in the experiment to its corresponding R-R interval series with different elevations, (2) Calculate the standard deviation of NN intervals (SDNN) and mean of R-R interval series every 30 s, N is the number of R-R interval in 30 s, that is: (3) Calculate the variable coefficient of RR interval series, that is, RRVC, every 30 s: Sustainability 2019, 11, x, 817 of 10   Before driving, a driver showed a certain amount of fatigue, which we termed as the before driving fatigue (BDF).In addition, the driving fatigue defined as the fatigue due to the driving process is termed as the DFC, and the total amount of fatigue is termed as the fatigue cumulant (FC).All these factors are derived as follows: We used the total driving fatigue degree (TDFD) and the DFD to represent the fatigue generated in the driving process: Through the aforementioned procedure, we obtained the data for the elevation of 3000-3500 m.Table 1 partially illustrates the driver fatigue parameter statistics in 3000-3500 m, from 0-35.5 min.The changing tendency of parameters was stable after 35.5 min.Figures 3-5 show that in the 3000-3500 m interval, DFC increased with time.This increase accelerated at 0-15 min, became more moderate at 15-29 min, and sharply accelerated again after 29 min.According to the drivers' subjective response during the experiment, most of them showed evident characteristics of driving fatigue, such as yawning, sore, or heavy eyes, after 32 min.Therefore, the driving fatigue could be categorized into three levels as mild, moderate, and severe by incorporating the DFD and the subjective performance of drivers.This reaction of the driver was the same as the data obtained from the inflection points on the curves.When the driver was mildly fatigued, the TDFD was 2.61 and DFD was 1.61; for moderate fatigue, the TDFD was 2.76 and DFD was 1.76; and for severe fatigue, the TDFD was 3.86 and DFD was 2.86.evident characteristics of driving fatigue, such as yawning, sore, or heavy eyes, after 32 min.Therefore, the driving fatigue could be categorized into three levels as mild, moderate, and severe by incorporating the DFD and the subjective performance of drivers.This reaction of the driver was the same as the data obtained from the inflection points on the curves.When the driver was mildly fatigued, the TDFD was 2.61 and DFD was 1.61; for moderate fatigue, the TDFD was 2.76 and DFD was 1.76; and for severe fatigue, the TDFD was 3.86 and DFD was 2.86.Based on the regression analysis, during the elevation of 3000-3500 m, we calculated the relational model between the DFC and T as follows, where T is time: Therefore, the driving fatigue could be categorized into three levels as mild, moderate, and severe by incorporating the DFD and the subjective performance of drivers.This reaction of the driver was the same as the data obtained from the inflection points on the curves.When the driver was mildly fatigued, the TDFD was 2.61 and DFD was 1.61; for moderate fatigue, the TDFD was 2.76 and DFD was 1.76; and for severe fatigue, the TDFD was 3.86 and DFD was 2.86.Based on the regression analysis, during the elevation of 3000-3500 m, we calculated the relational model between the DFC and T as follows, where T is time: incorporating the DFD and the subjective performance of drivers.This reaction of the driver was the same as the data obtained from the inflection points on the curves.When the driver was mildly fatigued, the TDFD was 2.61 and DFD was 1.61; for moderate fatigue, the TDFD was 2.76 and DFD was 1.76; and for severe fatigue, the TDFD was 3.86 and DFD was 2.86.Based on the regression analysis, during the elevation of 3000-3500 m, we calculated the relational model between the DFC and T as follows, where T is time: Based on the regression analysis, during the elevation of 3000-3500 m, we calculated the relational model between the DFC and T as follows, where T is time: and coefficient of determination R 2 = 0.699.We used Pearson's correlation coefficient to assess the correlation of driving time with the index we used.
At 3000-3500 m, a significant correlation was observed between the DFC and time (driving time) (see Table 2).Based on the regression analysis, we calculated the relational model between the TDFD and T in the 3000-3500 m elevation as follows: TDFD= 0.0001T 3 −0.0074T 2 +0.1569T+1.5914, And R 2 = 0.651.At the 3000-3500 m elevation, a significant correlation was observed between the DFD and time (driving time) (see Table 3).Based on the regression analysis, we calculated the relational model between the DFD and T at the 3000-3500 m elevation.
In the 3000-3500 m elevation, a significant correlation was observed between the TDFD and time (driving time) (see Table 4).The same procedure mentioned for Section 1 was applied to the remaining three elevated sections, that is, Sections 2-4.These sections showed the same results.

Results
As shown in the chart, with increasing elevation, the driving fatigue accumulated faster.A higher elevation implies lesser oxygen in the air, thus leading to hypoxia, which in turn causes driving fatigue.This shows that elevation greatly influenced driving fatigue.With higher altitude, the influence was more significant (see Table 5, Figure 6).The use of this method effectively quantifies and classifies driving fatigue.
To analyze the DFD and TDFD, we determined that in high altitude areas (above 3000 m), driving fatigue was more critical than in a normal area.Furthermore, the two indexes show a significant change in areas above 4000 m, implying that the driving fatigue in this area was generated and accumulated at an extremely fast rate (see Figures 7 and 8).Researchers should also consider altitude as an important factor in the design and management of a highway.The same procedure mentioned for section 1 was applied to the remaining three elevated sections, that is, sections 2-4.These sections showed the same results.

Results
As shown in the chart, with increasing elevation, the driving fatigue accumulated faster.A higher elevation implies lesser oxygen in the air, thus leading to hypoxia, which in turn causes driving fatigue.This shows that elevation greatly influenced driving fatigue.With higher altitude, the influence was more significant (see Table 5, Figure 6).The use of this method effectively quantifies and classifies driving fatigue.
To analyze the DFD and TDFD, we determined that in high altitude areas (above 3000 m), driving fatigue was more critical than in a normal area.Furthermore, the two indexes show a significant change in areas above 4000 m, implying that the driving fatigue in this area was generated and accumulated at an extremely fast rate(see Figure 7 ,Figure 8).Researchers should also consider altitude as an important factor in the design and management of a highway.The same procedure mentioned for section 1 was applied to the remaining three elevated sections, that is, sections 2-4.These sections showed the same results.

Results
As shown in the chart, with increasing elevation, the driving fatigue accumulated faster.A higher elevation implies lesser oxygen in the air, thus leading to hypoxia, which in turn causes driving fatigue.This shows that elevation greatly influenced driving fatigue.With higher altitude, the influence was more significant (see Table 5, Figure 6).The use of this method effectively quantifies and classifies driving fatigue.
To analyze the DFD and TDFD, we determined that in high altitude areas (above 3000 m), driving fatigue was more critical than in a normal area.Furthermore, the two indexes show a significant change in areas above 4000 m, implying that the driving fatigue in this area was generated and accumulated at an extremely fast rate(see Figure 7 ,Figure 8).Researchers should also consider altitude as an important factor in the design and management of a highway.

Figure 2 .Figure 1 . 10 Figure 1 .
Figure 2. The procedure for calculating the driving fatigue degree (DFD).1) Process the ECG data obtained in the experiment to its corresponding R-R interval series with different elevations, 2) Calculate the standard deviation of NN intervals (SDNN) and mean of R-R interval series every 30 s, N is the number of R-R interval in 30 s, that is: N

Figure 6 .
Figure 6.Driver fatigue time with altitude variation.

Figure 6 .
Figure 6.Driver fatigue time with altitude variation.

Figure 6 .
Figure 6.Driver fatigue time with altitude variation.

Table 2 .
Correlations between DFC and T.

Table 3 .
Correlations between DFD and T.

Table 4 .
Correlations between TDFD and T.

Table 5 .
Initial time of driver fatigue of different levels.

Table 5 .
Initial time of driver fatigue of different levels.

Table 5 .
Initial time of driver fatigue of different levels.