Comparative study on electroencephalography characteristics of professional and non-professional drivers in tunnel sections: A field investigation

Abstract

BACKGROUND:

The tunnel section is a complex traffic scenario and an accident prone area. There are differences in the performance of different driving groups in tunnel environments, which may have an impact on traffic safety.

OBJECTIVE:

To study the differences in the impact of tunnel environment on professional and non-professional drivers.

METHODS:

Based on the vehicle experimental data, the electroencephalography (EEG) power was compared for professional and non-professional drivers. The impact of illumination changes and longitudinal slope on different driving groups was analyzed.

RESULTS:

At tunnel entrance with severely reduced lighting, the adaptation time of non-professional drivers to the light environment is 1.5 times that of professional drivers. When driving on the longitudinal slope, professional drivers perform better. The greater the longitudinal slope, the more obvious the advantages of professional drivers. However, in areas with relatively good traffic conditions, professional drivers are more prone to be distracted.

CONCLUSION:

Professional and non-professional drivers have their own advantages and disadvantages in the tunnel environment. In general, the driving adaptability of professional drivers is better than that of non-professional drivers in tunnel sections. The research conclusions provide a reference for driver safety training.

Keywords

Professional drivers human factor driving safety lighting environment road alignment EEG

1 Introduction

Road tunnels are a kind of transportation facilities built under obstacles such as mountains, water, and urban buildings for vehicles to pass [1]. It is able to break through the geographical restrictions on traffic, and plays an important role in improving urban layout, optimizing road alignment and shortening access distance [2]. Road tunnels are important transportation hubs and play a crucial role in the road network [3]. Once a traffic accident occurs, it will cause regional traffic congestion at both ends of the tunnel. The “black hole effect” at the tunnel entrance (driver suddenly enters a low-light environment from a high-light environment, the field of vision is dark, and the road ahead cannot be seen for a short time) and the “white hole effect” at the tunnel exit (driver suddenly enters a high-light environment from a low-light environment, the field of vision is too bright, and the road ahead cannot be seen for a short time) are important factors affecting tunnel traffic safety [4]. Greater external illumination produces a greater illumination difference inside and outside the tunnel, and a greater threat to traffic safety [5]. At the same time, the road tunnel has a complex road alignment. Taking the water-crossing tunnel as an example, the start of the tunnel is located on one side of the water, go down to the bottom of the water, and finally go up from the bottom of the water to the other side, which leads to obvious longitudinal slope in some tunnels. Studies have shown that tunnel accident rates and accident losses are much higher than those of other roads, and that the occurrence of tunnel traffic accidents is significantly related to the light environment and the longitudinal slope [6]. Thus, an in-depth study on the physiological and psychological effects of light environment changes and longitudinal slope of tunnels on different driving groups is an important practical necessity for tunnel operation and traffic accident prevention.

In the study of tunnel traffic safety, previous research has shown that a bad driving environment (road hazards, poor illumination, traffic congestion, etc.) is significant correlation with the occurrence of traffic accidents [4]. Du et al. [7] analysed the phenomenon of transient visual impairment at highway tunnel entrances. They proposed the concept of visual shock, and used it to evaluate driver comfort passing through a tunnel entrance. Zhao et al. [8] analysed the influence of raised pavement markers (RPMs) in long tunnels on the visual perception of drivers, and found that RPMs can increase driver attention and reduce vehicle speed. Yang et al. [9] evaluated traffic safety in a tunnel based on the physiological characteristics of the driver and vehicle speed, and found that there were more road hazards at the tunnel entrance. Feng et al. [10] considered an urban underpass tunnel as the research object and analysed the psychological changes of the driver using ECG (Electrocardiograph) as the index, and found that steeper slope led to higher heart rate growth index. The physiological driver indexes used in previous research have usually been ECG and eye movement, and EEG has rarely been considered. This study analyzes the impacts of road tunnels on driver EEG, and apply brain science to tunnel traffic safety research.

EEG related research is increasingly used in the field of transportation, which is often used to study the physiological states of drivers such as tension, fatigue, relaxation, and attention. Using EEG to measure these physiological states in different driving environments can help identify risk factors that affect traffic safety. Edmund et al. [11] found that driving is a complex behavior that requires the coordinated participation of three main brain regions, especially the temporal, occipital and frontal regions. Yang et al. [12] divided driving behavior into radical stable type, conservative stable type, conservative unstable type, radical unstable type and general type by analyzing the EEG signals of drivers. Oka et al. [13] studied the changes in the EEG signals of drivers in different turning directions, and according to the results, left-turn behaviour required more visual attention than right-turn behaviour. EEG signals will be affected by the common perception of various human senses such as vision, hearing, touch, taste, etc. Compared with other physiological indicators (eye movement, ECG, skin electrical, etc.), EEG will evaluate external stimuli more comprehensively.

Professional drivers are important participants in road traffic and also a key research object for traffic safety in road tunnels. In Hong Kong, professional drivers account for a large proportion of traffic accidents in road tunnels with 70% of all traffic accidents involving commercial vehicles [14]. Thus, classifying the drivers, studying the impact of road tunnels on different driver groups (professional and non-professional drivers), and analysing safety problems for different driver groups in road tunnels can provide a theoretical basis for improving tunnel driving adaptability.

The previous study has shown that there are significant differences in driving performance among different driving groups in certain scenarios [15]. At present, research on the human factors of tunnel drivers is not in-depth, and there is a lack of research on different driving groups. The little research that is available are mostly focused on eye movement and ECG. Therefore, this study is based on EEG characteristics and focuses on professional and non-professional drivers to analyze the driving performance of different driving groups in tunnel scenarios. It aims to provide a reference basis for the safe operation of tunnels. As stated below, these are actually research questions.

How does tunnel lighting and road alignment affect driving technique or traffic safety among professional and non-professional drivers?

What are the differences of EEG signals between professional and non-professional drivers in different sections of road tunnels?

2 Test design

2.1 Test tunnel

The test tunnel is the Jiaozhou Bay Tunnel, located between the Qingdao and Huangdao Districts. It is an important channel connecting the two areas. The total length of the line is 7.8 km. The maximum speed in the tunnel is 80 km/h. The tunnel is in normal operation and other vehicles can pass freely during the test. To prevent traffic flow interference as much as possible for the test drivers, the test was conducted during a common period (9 am to 11 am and 2 pm to 4 pm) of the working day, with drivers driving in free-flow conditions. The test period is from July to August 2021.

2.2 Test equipment

The test vehicle is a 2019 Chevrolet Cavalier, which is a five-seat family passenger car, and the vehicle’s appearance, power, steering, braking and other performance are in good condition. The illumination measurement equipment is a TES-1339R illuminance meter, with a sampling rate of 5 samples/s, a resolution of 0.01 Lux, and measurement accuracy of 3%. During data collection, a light sensor is fixed to the right side of the front windshield, and the data collector is placed in the vehicle and operated by the experimental assistant (the staff member who operate the test equipment). The EEG measurement equipment is an EEG instrument with a sampling rate of 258 samples/s. The International 10–20 system electrode locations method is used, with a total of 16 measurement channels (AF3, F3, F7, AF4, F4, F8, P3, P7, P4, P8, FC5, T7, FC6, T8, O1 and O2). The location of each measurement channel is shown in Fig. 1. Measurement channels AF3, F3, F7, AF4, F4, and F8 are located in the frontal lobe (physical movement and thinking function centre). Measurement channels P3, P7, P4, and P8 are located in the parietal lobe (physical sensory function centre). Measurement channels FC5, T7, FC6, and T8 are located in the temporal lobe (auditory function centre), and measurement channels O1 and O2 are located in the occipital lobe (visual function centre). Other auxiliary equipment includes a slope meter, a driving recorder, and two video recorders. The slope meter is used to measure the target tunnel, and the measurement results are consistent with the longitudinal slope value in the Jiaozhou Bay Tunnel Engineering Manual [16]. The driving recorder was fixed on the front windshield of the test vehicle to record the driving speed and the traffic in front of the vehicle in real time. The video recorders were fixed on the sides of the vehicle to record real-time traffic on both sides. When processing the test data, the video is referred to, and when other vehicles’ overtaking, lane changing, and other behaviour interfere with the test driver, the corresponding data segment will be removed.

Fig. 1

Laboratory equipment.

2.3 Test time and test driver

The “black/white hole effect” is most severe in sunny. To ensure the universality of the results of this study, the experiment was conducted in sunny summer. Before the test, the external illuminance was measured to ensure that it was greater than 40000 lux [17]. To prevent traffic flow interference as much as possible for the test drivers, the test was conducted during a common period of the working day. Sufficient experimental sample size is a prerequisite for reliable conclusions. The sample size calculation method is as follows: $N = \frac{Z^{2} σ^{2}}{E^{2}}$ (1) where N denotes the sample size, Z represents the standard normal distribution statistic, σ signifies the standard deviation, and E denotes the maximum allowable error. Generally, if the significance level is 10%, then Z = 1.25, and the value range of σ is 0.25∼0.5 [18]. Considering the limitation of the number of test drivers, σ was 0.38, E was 10%. The minimum required sample size was 23 [4].

In this study, effective test data of 52 drivers are obtained, which satisfies the requirement of the number of samples, including 24 professional drivers and 28 non-professional drivers. Professional drivers come from a taxi company and have formal training, in a driving occupation. Non-professional drivers are offered online recruitment, who have vehicle driving qualifications but have not received driving vocational training. The professional drivers were 24–50 years old (36.9±4.6 years), and the male to female ratio is 3:1. The non-professional drivers were 20–53 years old (33.1±8.6 years), and the male to female ratio is 2.5:1. The age distribution and gender ratio of the test drivers were consistent with the Chinese driving population. The test drivers were in good health, with no history of mental disease. To ensure the accuracy of the test, for three days before the vehicle experiment, all drivers had adequate sleep (self-reported number of hours per night, with each person per night sleeping for 6–8.5 hours, with an average of 7.6 hours) and did not consume alcohol or drugs. Drivers gave their informed consent before participating in the study. After the experiment, each driver was compensated 200 Chinese Yuan Renminbi (RMB) for their participation in the research. This study was conducted in accordance with the ethical principles of the current Declaration of Helsinki.

All drivers were trained to ensure understanding of the test process and driving route. After the test, the staff will manually check the experimental data. When the raw EEG data shows a large area of bad segments caused by poor contact of the measuring electrodes or scalp shaking, the data will be discarded and the test will be repeated. A series of processing such as bad segment elimination, EEG segmentation, filtering, EEG frequency segmentation, baseline calibration, and artifact removal are performed on the collected EEG data, and then the EEG power spectrum and EEG power are further extracted. After that, the Laida criterion method (Equation 1) was used to eliminate outliers in the EEG data. Relevant studies have shown that EEG slow waves (θ wave and α wave) can be used to detect driver distraction states [19]. In this study, the ratio of EEG slow wave power during the test was compared with the resting state. The test data of the driver’s distracted driving were screened and deleted. $| X_{i} - \bar{X} | > 3 σ$ (2) where X_i is the data of each test sample; X is the arithmetic mean of each sample; σ is the standard deviation of the experimental sample.

3 Results and analysis

3.1 Analysis of driving environment in tunnel

The illumination data from the test were processed, the illumination and longitudinal slope distribution of the road tunnel is presented in Fig. 2.

It is observed in Fig. 2 that at the tunnel entrance, the illumination decreases sharply from above 50,000 Lux to below 100 Lux with a longitudinal slope of –4%. At the tunnel exit, the illumination increases sharply from below 100 Lux to above 50,000 Lux with a longitudinal slope of 3.5%. In the middle of the tunnel, the longitudinal slope changed several times and the illumination remains at 80 Lux. Therefore, in studying the influence of illumination changes on drivers in road tunnels, we focus only on the entrance and exit sections. When studying the influence of longitudinal slope changes on drivers, we focus only on the middle section. The longitudinal slope and length distribution of each section of the Jiaozhou Bay Tunnel from Huangdao District to Qingdao District are shown in Table 1.

Fig. 2

Illumination and longitudinal slope distribution of test tunnel.

Table 1

Longitudinal slope distribution of test tunnel

Longitudinal slope (%)	–4.00	–2.47	–3.90	–2.88	–0.30	2.23	3.45	0.30	2.15	3.50
Length (m)	797	386	750	765	875	1510	815	895	750	314

Table 1 shows that one round trip from Huangdao District to Qingdao District passes through 10×2 longitudinal slope sections ranging from –4.00% to 4.00%.

3.2 Analysis of impact of illumination changes on professional and non-professional drivers

An EEG measures spontaneous rhythmic neural electrical activity; its frequency ranges from 0.5–32 cycles per second. According to the frequency classification, an EEG can be divided into four bands: δ wave (0.5–4 Hz), θ wave (4–8 Hz), α wave (8–13 Hz), and β wave (13–32 Hz) [20].

A brain electrical activity map (BEAM) is a plan view showing a human brain in different colours according to the power of the EEG signals in different bands. A BEAM clearly reflects the activity in different brain regions. The length of the tunnel entrance section and the exit section is 150 m respectively. The driver will pass through the area in a very short time, and the activity of the EEG will change dramatically in a short time, which is unfavorable to traffic safety. The darker the red of the BEAMs, the more active the EEG, the more nervous the driver, the greater the driving load, and the poor the traffic safety [9]. Blue of the BEAMs means the driver is calmer, with a low driving load and good traffic safety. The BEAMs of two types of drivers at the tunnel entrance and exit are shown in Figs. 3 and 4.

Fig. 3

Comparison of BEAM between professional and non-professional drivers at tunnel entrance.

Fig. 4

Comparison of BEAM between professional and non-professional drivers at tunnel exit.

It is observed in Fig. 3 that the brain activity of drivers changes with a sudden decrease in illumination at the tunnel entrance. As a driver approaches the tunnel entrance, brain activity increases, and the driver is more nervous. After the tunnel entrance, the brain activity gradually decreases, and the driver is calmer. The professional driver brain is highly active until [–25 m, 75 m]. The non-professional driver brain is highly active until [–25 m, 125 m]. The length of the road section in which non-professional drivers exhibit high brain activity is approximately 1.5 times longer than for professional drivers, indicating that professional drivers adapt faster to sudden decreases in illumination than non-professional drivers. BEAMs have a deeper red for non-professional drivers than for professional drivers in the same area, indicating significantly greater brain activity, and a stronger ability of professional drivers to resist the “black hole effect”.

Similarly, in Fig. 4, the professional driver EEG signals is highly active until [–50 m, 0 m], compared with [–100 m, 0 m] for the non-professional driver. At the tunnel exit, professional drivers adapt significantly faster to sudden increases in illumination than non-professional drivers, with a stronger ability to resist the “white hole effect”.

Comparing Figs. 3 and 4, driver brain activity at the tunnel entrance is higher than at the exit, and the duration of a highly active brain state is longer than at the exit of the tunnel, indicating that the influence of the “black hole effect” on the driver is greater than that of the “white hole effect”. This trend is more obvious for professional drivers.

In terms of brain regions, when passing through the entrance and exit of the road tunnel, the EEG activity of each brain region changes to some extent for both professional and non-professional drivers, indicating that navigating the tunnel entrance and exit is a complex behaviour integrating information perception, judgment analysis, and instruction execution. The most active brain region is the occipital lobe (visual function centre), which suggests a strong influence on vision at the entrance and exit of the tunnel.

To quantitatively compare the effects of illumination changes in the road tunnel on professional and non-professional drivers, the EEG power increase and illumination change coefficient are introduced, calculated as $G_{i} = \frac{n_{i, 2} - n_{i, 1}}{n_{i, 1}} \times 100 {%$ (3) where G_i is the EEG power increase for a driver at a certain time, (i =δ wave, θ wave, α wave, β wave). The β wave power increase (G_β) represents the psychological load of the driver [9]. With the increase of G_β, the more nervous the driver, the worse the driving safety. n_i, 1 is the EEG power of a driver in a resting state. n_i, 2 is the EEG power of a driver at a certain time. $Δ I = I_{2 -} I_{1}$ (4) $Q = ln | Δ I |$ (5) where ΔI is the illumination variable quantity (lux); I₂ is the illuminance of the current cycle (lux); I₁ is the illuminance of the previous cycle (lux); Q is the illumination change coefficient. A larger illumination change coefficient indicates a more intense illuminance change; a smaller illumination change coefficient indicates a slower illuminance change.

The variation in illumination and the β wave power increase for professional and non-professional drivers at tunnel entrance and exit, respectively, are shown in Figs. 5 and 6.

Fig. 5

Comparison of G_β between professional and non-professional drivers at tunnel entrance.

It is observed in Fig. 5 that the G_β for professional drivers increases rapidly approaching the tunnel entrance, and reaches the maximum at [0 m, 10 m]. After the tunnel entrance, the G_β for professional drivers gradually decreases. A similar change in G_β was observed for non-professional drivers. The range of G_β change was significantly smaller for professional drivers than for non-professional drivers. The maximum G_β for non-professional drivers (289.0%) was 1.38 times greater than for professional drivers (210.0%). The average G_β for non-professional drivers (158.5%) was 1.47 times greater than for professional drivers (107.5%). This indicates that professional drivers were more calm than non-professional drivers with a sudden decrease in illumination at the tunnel entrance. The illumination changes the most at [0 m, 10 m], which coincides with the peak G_β for both driver types. The main illumination change area (ΔI > 1000 lux) is located at [–30 m, 70 m], and is generally consistent with the area of high brain activity for professional drivers in Fig. 3.

It is observed in Fig. 6 that passing through the tunnel exit, the G_β for professional drivers first increases and then decreases, and reaches the maximum at [–10 m, 0 m]. A similar change in G_β was observed for non-professional drivers. The G_β change range for professional drivers was significantly smaller than for non-professional drivers. The maximum G_β for non-professional drivers (249.3%) was 1.67 times greater than for professional drivers (149.7%). The average G_β for non-professional drivers (128.5%) was 1.43 times greater than for professional drivers (89.8%). This indicates that professional drivers experienced less stress than non-professional drivers with a sudden increase in illumination at the tunnel exit. The illumination changes the most at [0 m, 10 m]; the G_β is high in this area for both driver types. The main illumination change area (ΔI > 1000 lux) is located at [–70 m, 30 m], the same as at the entrance.

Fig. 6

Comparison of G_β between professional and non-professional drivers at tunnel exit.

In Figs. 5 and 6, the change in G_β with a change in illuminance shows the same trend for professional and non-professional drivers. With a greater illumination change closer to the tunnel entrance, G_β is higher. Farther from the tunnel entrance, the illumination change is smaller and G_β is lower. For professional drivers, the maximum G_β at the tunnel entrance (210.0%) is 1.40 times greater than at the exit (149.7%). For non-professional drivers, the maximum G_β at the entrance (289.0%) is 1.16 times greater than at the exit (249.3%). This indicates that the influence of the “black hole effect” at the tunnel entrance is greater than that of the “white hole effect” at the exit. The difference is more pronounced for professional drivers. Professional drivers are less affected by illumination changes than non-professional drivers, which may present advantages regarding traffic safety related risks.

Considering the illumination change coefficient as the independent variable and the G_β for professional and non-professional drivers as the dependent variable, the two variables are fitted. The scatter and fitting curves of the illumination change coefficient and G_β are shown in Figs. 7 and 8.

Fig. 7

Fitting curve between illumination change coefficient and G_β at tunnel entrance.

Fig. 8

Fitting curve between illumination change coefficient and G_β at tunnel exit.

The relationship model between the illumination change coefficient and G_β at the tunnel entrance for professional and non-professional drivers, respectively, is shown in Equations (6) and (7). The determination coefficient R² is high (0.82 and 0.87, respectively), indicating that the reliability of the model is good. $G_{β 1} = 64.69 - 1.54 Q + 1.49 Q^{2}$ (6) $G_{β 2} = 94.16 + 0.90 Q + 1.79 Q^{2}$ (7) where G_β1 is the β wave power increase for professional drivers, and G_β2 is the β wave power increase for non-professional drivers.

The relationship model between the illumination change coefficient and G_β at the tunnel exit for professional and non-professional drivers is shown in Equations (8) and (9), respectively. The determination coefficient R² is high (0.92 and 0.94, respectively), indicating that the reliability of the model is good. $G_{β 1} = 60.78 + 8.83 Q - 1.91 Q^{2} + 0.20 Q^{3}$ (8) $G_{β 2} = 68.36 + 5.39 Q + 4.02 Q^{2} - 1.29 Q^{3} + 0.11 Q^{4}$ (9)

3.3 Analysis of impact of longitudinal slope on professional and non-professional drivers

In the middle section of the test tunnel, the illumination is constant, and the driver is only affected by the longitudinal slope. The test data in the middle section of the tunnel were processed, the G_β for professional and non-professional drivers in different longitudinal slope sections is shown in Fig. 9 and Table 2.

Fig. 9

G_β of professional and non-professional drivers at different longitudinal slope sections.

Table 2

G_β for professional and non-professional drivers, and significance

downhill	longitudinal	driver group	${\bar{G}}_{β}$	ratio of ${\bar{G}}_{β}$	median	maximum	minimum	standard	P
/uphill	slope (%)							deviation	value
—	[–4,4]	non-professional drivers	26.5	1.49	24	85	0	17.8	0.000*
		professional drivers	17.8		16	60	–6	12.3
downhill	[–4,0]	non-professional drivers	28.4	1.41	25.5	85	0	18.4	0.002*
		professional drivers	20.2		18	60	–5	13.5
uphill	[0,4]	non-professional drivers	24.5	1.59	22	83	1	17.0	0.000*
		professional drivers	15.4		14	46	–6	10.5
downhill	[–4,–3.5]	non-professional drivers	44.9	1.49	41.5	85	18	19.3	0.000*
		professional drivers	30.2		27	60	7	14.8
	[–3.5,–3]	non-professional drivers	36.9	1.42	39.5	70	14	15.4	0.000*
		professional drivers	26.0		22	55	9	11.4
	[–3,–2.5]	non-professional drivers	27.5	1.29	29.5	54	6	12.6	0.008*
		professional drivers	21.3		21.5	41	5	9.8
	[–2.5,–2]	non-professional drivers	23.5	1.28	24	50	5	11.2	0.050*
		professional drivers	18.3		16	33	2	9.7
	[–0.5,0]	non-professional drivers	9.2	1.51	6.5	25	0	7.4	0.376
		professional drivers	6.1		7	16	–5	4.7
uphill	[0,0.5]	non-professional drivers	9.1	1.71	9	27	1	7.1	0.214
		professional drivers	5.3		4	14	–6	5.1
	[2,2.5, 2,2.5]	non-professional drivers	17.2	1.50	15	41	4	9.4	0.063
		professional drivers	11.4		10.5	26	2	7.0
	[2.5,3, 2.5,3]	non-professional drivers	23.8	1.61	23	49	8	11.0	0.004*
		professional drivers	14.8		14.5	28	5	6.1
	[3,3.5, 3,3.5]	non-professional drivers	33.3	1.65	30	59	5	14.9	0.000*
		professional drivers	20.2		19	38	5	8.9
	[3.5,4, 3.5,4]	non-professional drivers	39.3	1.54	35.5	83	8	19.2	0.000*
		professional drivers	25.6		23	46	5	10.8

It is observed in Fig. 9 that in downhill and uphill sections, G_β increases for professional and non-professional drivers with an increase in the longitudinal slope. In each longitudinal slope section, the G_β for professional drivers is lower than for non-professional drivers, which indicates that non-professional drivers are more nervous than professional drivers at the same longitudinal slope. When the longitudinal slope was [–0.5%, 0] or [0, 0.5%], the G_β for some professional drivers was less than 0, indicating lower brain activity than in a resting state, and a high probability of distracted driving.

Table 3

Significance analysis for professional and non-professional drivers in downhill and uphill sections

slope	[–4,0]	[0,4]	[–4,–3.5]	[3.5,4]	[–3.5,–3]	[3,3.5]	[–3,–2.5]	[2.5,3]	[–2.5,–2]	[2,2.5]	[–0.5,0]	[0,0.5]
P value
drivers
non-professional drivers	0.069		0.289		0.386		0.262		0.028*		0.957
professional drivers	0.003*		0.153		0.040*		0.011*		0.036*		0.558

From Table 2, affected by the longitudinal slope in the tunnel, the G_β for non-professional drivers (26.5%) was 1.49 times greater than for professional drivers (17.8%). The influence of longitudinal slope on professional drivers was less than on non-professional drivers. The G_β for professional drivers in the downhill section (20.2%) was 1.31 times greater than that in the uphill section (15.4%). The G_β for non-professional drivers in the downhill section (28.4%) was 1.16 times greater than in the uphill section (24.5%). This shows that the influence of the uphill section on professional and non-professional drivers was less than that of the downhill section. Professional drivers were more sensitive to the difference in uphill and downhill sections. When the longitudinal slope was [–2.5%, –2% ], [–3%, –2.5% ], [–3.5%, 3% ], and [–4%, –3.5% ], the G_β for non-professional drivers was 1.28, 1.29, 1.42, and 1.49 times greater, respectively, than that for professional drivers. With an increase in downhill longitudinal slope, the difference in G_β between professional and non-professional drivers gradually increased. It can be considered that professional drivers are somewhat more adaptable to a large longitudinal slope than non-professional drivers. With a greater longitudinal slope, the advantages of professional drivers are more obvious. In each longitudinal slope section, the longitudinal slope with the largest ratio of G_β was [–0.5%, 0] and [0, 0.5%], which may be related to the low G_β of some professional drivers.

When the longitudinal slope was [–4%, 4% ], [–4%, 0], and [0, 4%], the P value of the G_β for professional and non-professional drivers was less than 0.05, indicating significant differences in the G_β for professional and non-professional drivers in the downhill section, the uphill section, and the entire tunnel. Except for three small longitudinal slope sections ([–0.5%, 0], [0, 0.5%], and [2%, 2.5%]), there were significant differences in G_β for professional and non-professional drivers. There was no significant difference in EEG characteristics between professional and non-professional drivers in the small longitudinal slope sections, while there was a significant difference in the large longitudinal slope sections. This indicates that professional drivers have an advantage in maintaining traffic safety when driving in large longitudinal slope sections.

To further study the difference in the influence of longitudinal slope on professional and non-professional drivers, the differences of the G_β in the downhill and uphill sections was analysed for both driver types. The results are shown in Table 3.

In Table 3, the P values for professional and non-professional drivers in the downhill and uphill sections are 0.069 and 0.003, respectively. There is a significant difference between the G_β for professional drivers in the downhill section and the uphill section, whereas there is no significant difference in the G_β for non-professional drivers.

With longitudinal slopes of 2–2.5%, 2.5–3%, and 3–3.5%, there is a significant difference in the G_β for professional drivers in the uphill and downhill sections. When the longitudinal slope is 2–2.5%, there is a significant difference in the G_β for non-professional drivers in the downhill and uphill sections. In each slope section, the differences of G_β for non-professional drivers in downhill and uphill sections is significantly less than for professional drivers, indicating that professional drivers are more sensitive to the difference between downhill and uphill sections. Intuitively, the road training and additional driving experience professional drivers have would de-sensitize them to these changes (they adapt more quickly than non-professional drivers). The reason for this phenomenon may be related to the difference in accident risk between uphill and downhill. Relevant research shows that the parking distance of the downhill section > the horizontal section > the uphill section, and the accident rate in the downhill area is higher than that in the uphill area. Professional drivers have more experience, so they will be more alert to the downhill area.

Considering longitudinal slope as the independent variable and the G_β for professional and non-professional drivers as the dependent variable, the two variables are fitted. The scatter and fitting curves for longitudinal slope and G_β are shown in Fig. 10.

Fig. 10

Fitting curve between longitudinal slope and G_β in road tunnels.

The relationship model between longitudinal slope and G_β for professional and non-professional drivers is shown in Equations (10 and 11), respectively. The determination coefficient R² is high (0.98 and 0.99, respectively), indicating that the reliability of the model is good. $G_{β 1} = 6.12 - 0.93 S + 1.58 S^{2}$ (10) $G_{β 2} = 8.57 - 0.76 S + 2.40 S^{2}$ (11) where S is the longitudinal slope (%).

4 Discussion

This study indicates that G_β of professional drivers exhibit significant differences in uphill and downhill sections, whereas non-professional drivers exhibit no significant difference. There are significant differences in driving psychology and behavior between the two types of driver groups, which shows that it is necessary to study different driver groups. With sudden changes in illumination in road tunnels, the degree and duration of stress are significantly lower for professional drivers than for non-professional drivers, possibly related to driving experience. Further research will be conducted in future studies. Medic et al. [22] reported that driving ability is positively correlated with driving experience, which is consistent with the speculation of this study.

Sudden changes in illumination and the complex longitudinal slope are two important factors threatening traffic safety in road tunnels. Comparing Fig. 8 with Fig. 12, it is found that the impact of illumination change on the driver at the tunnel entrance is intense and short, whereas the impact on the driver at the middle of the tunnel is relatively mild and continuous. Illumination adjustment facilities such as shading sheds can decrease sudden illumination changes at tunnel entrances and exits. Since non-professional drivers have a greater G_β at tunnel entrances and exits, the potential driving risk is greater than that of professional drivers, non-professional drivers should be strongly considered when determining illumination adjustment facilities. Zhao et al. [8] reported that visual guiding facilities can enhance driver perception of the tunnel contour and reduce the impact of large longitudinal slopes. In addition, appropriate visual guiding facilities can improve the environment inside the tunnel and reduce driving fatigue [23, 24]. As non-professional drivers are more nervous in large longitudinal slope sections, and professional drivers have a more obvious tendency for fatigued driving in road tunnels, non-professional drivers should be strongly considered when determining visual guiding facilities in large longitudinal slope sections, and professional drivers should be strongly considered in areas with high incidence of driving distracted.

Due to the limitation of resources, this study has limitations. In the future research, we need to focus on the following issues. (1) The sleep quantity of the driver may have an impact on the driver, and more in-depth research is needed in subsequent studies. (2) The psychophysiological differences between professional drivers and non-professional drivers may be related to factors such as driving experience and driving habits, in addition to whether or not they have undergone relevant training. However, driving experience is difficult to quantify. Driving experience may be determined by a combination of factors such as driving talent, driving age, total driving hours, etc., which will be the focus of our research. (2) The psychophysiological activities of drivers are very complex and closely related to tunnel traffic safety [25]. Due to the limitation of experimental conditions and the length of the paper, only the EEG characteristics of drivers were studied in this paper. In future studies, it is necessary to comprehensively analyze the EEG, eye movement, ECG and other psychophysiological parameters, and more comprehensively evaluate the impact of the curve and longitudinal slope combination on tunnel driving safety. (3) The traffic environment in tunnels is complicated, drivers may be affected by multiple factors including the light environment, road alignment, weather, temperature, traffic flow, and vehicle speed. Thus, further research on the influence of multiple factors is necessary.

5 Conclusions

The influence and duration of sudden changes in illumination on EEG signals are smaller for professional drivers than for non-professional drivers. Professional drivers are less affected by and more adaptable to the “black hole effect” and the “white hole effect”. The “black hole effect” at the tunnel entrance has a greater impact on the driver than the “white hole effect” at the exit. This phenomenon is more obvious for professional drivers. Professional drivers have certain advantages in perceiving illumination changes.

In large longitudinal slope sections, there are significant differences in the G_β for professional and non-professional drivers. In downhill and uphill sections, with the same longitudinal slope, the G_β for professional drivers is generally significantly different; for non-professional drivers, the G_β is generally not significantly different. At the same longitudinal slope, the G_β for non-professional drivers is greater than for professional drivers; non-professional drivers are more nervous than professional drivers. Professional drivers have a stronger adaptability to large longitudinal slopes than non-professional drivers. The larger the longitudinal slope, the greater the difference in driving ability between professional and non-professional drivers, and the more prominent the advantages of professional drivers.

Generally, the adaptability of professional drivers in road tunnels is better than that of non-professional drivers. To improve the driving adaptability of non-professional drivers in tunnel sections, (1) tunnel driving simulation training should be increased; (2) tunnel driving simulation training should focus on sudden changes in illumination and large longitudinal slopes.

Ethical approval

Not applicable.

Informed consent

Drivers gave their informed consent before participating in the study.

Conflicts of interest

None to report.

Footnotes

Acknowledgments

None to report.

Funding

This research was funded by the National Natural Science Foundation of China (No. 52072291), and China Scholarship Council (No. 202306950072).

References

Jiao

, Du

, Wang

. Research on drivers’ visual characteristics in different curvatures and turningconditions of the extra-long urban underwater tunnels. Tunn Undergr Sp Tech. 2020;99:103360. doi: 10.1016/j.tust.2020.103360.

Chen

, Chen

, Xing

. Comparison and analysis of crash frequency and rate in cross-river tunnels usingrandom-parameter models. Journal of Transportation Safety & Security. 2020;14(2):1–25.

Pervez

, Huang

, Lee

. Crash analysis of expressway long tunnels using a seven-zone analytic approach. Journalof Transportation Safety & Security. 2021;13;108–22. doi: 10.1080/19439962.2019.1605642.

Yang

, Du

, Zhang

. Research on drivers’ eye movement characteristics in the combined section of longitudinalslope and curve in the extra-long underwater tunnel. Tunnelling and Underground Space Technology. 2023;136:105078. doi: 10.1016/j.tust.2023.105078.

Zhang

, Wang

, Shen

, Liu

, Feng

. Drivers’ car-following behaviours in low-illumination conditions. Ergonomics. 2021;64(2):199–211. doi: 10.1080/00140139.2020.1821097.

Pan

, Zhang

, Wang

. Lane-changing risk analysis in undersea tunnels based on fuzzy inference. IEEE Access.. 1951;8:2–20. doi: 10.1109/ACCESS.2020.2968584.

, Zheng

, Ran

, Zhao

. Drivers’ visual comfort at highway tunnel portals: A quantitative analysisbased on visual oscillation. Transport Res D-Tr E. 2014;31(8):37–47. doi: 10.1016/j.trd.2014.05.012.

Zhao

, Ju

, Li

Safety of raised pavement markers in freeway tunnels based on driving behavior. Accident AnalPrev. 2020;145:105708.. doi: 10.1016/J.AAP.2020.105708.

Yang

, Du

, Jiao

, Pan

. Analysis of EEG characteristics of drivers and driving safety in undersea tunnel. Int J Environ Res Public Health. 2021;18:9810. doi: 10.3390/IJERPH18189810.

10.

Feng

, Yang

, Kumfer

. Effect of longitudinal slope of urban underpass tunnels on drivers’ heart rate andspeed: A study based on a real vehicle experiment. Tunn Undergr Sp Tech. 2018;81:525–33. doi: 10.1016/j.tust.2018.08.032.

11.

Edmund

, Arnau

, Reiser

Evaluating mental load during realistic driving simulations by means of round theear electrodes. Front Neurosci-Switz. 2019;13:940. doi: 10.3389/fnins.2019.00940.

12.

Yang

, Guan

, Ma

, Li

. Comparison among driving state prediction models for car-following condition based onEEG and driving features. Accident Anal Prev. 2019;133:105296. doi: 10.1016/j.aa2019.105296.

13.

Oka

, Yoshino

, Yamamoto

. Greater activity in the frontal cortex on left curves: A vector-based fNIRS studyof left and right curve driving. PloS One. 2015;10(5):e0127594. 10.1371/journal.pone.0127594.

14.

Chen

, Sze

, Lu

. Safety of professional drivers in an ageing society-A driving simulator study. Transport ResF-Traf. 2019;67(C):101–12. 10.1016/j.trf.2019.10.006.

15.

Yeung

, Wong

. Effects of driver age and experience in abrupt onset hazards. Accid Anal Prev. 2015;78:110–7. doi: 10.1016/j.aa2015.02.024.

16.

Song

, Zhou

The overall design of Qingdao Jiaozhou Bay subsea tunnel. Adv Mater Res. 2012;1494:2971-6. doi:10.4028/www.scientific.net/AMR.368-373.2971

17.

Wang

, Du

, Jiao

, Zheng

, Ni

. Drivers’ visual load at different time periods in entrance and exit zones ofextra-long tunnel. Traffic Inj Prev.. 2020;21:539–44. doi: 10.1080/15389588.2020.1821196.

18.

Lyu

, Wang

, Wu

, Peng

, Thomas

. Using naturalistic driving data to identify driving style based onlongitudinal driving operation conditions. Journal of Intelligent and Connected Vehicles. 2022;5(1):17–35. 10.1108/JICV-07-2021-0008.

19.

Marzieh

, Saeid

, Mahdi

, et al. , Effect of tonal noise and task difficulty on electroencephalography (EEG)and cognitive performance. International Journal of Occupational Safety and Ergonomics. 2021;28(3):31–31. doi: 10.1080/10803548.2021.1901432.

20.

Lee

, Chalmers

, Burton

, Zilberg

. Electrophysiological brain-cardiac coupling in train drivers duringmonotonous driving. Int J Environ Res Public Health. 2021;18:3741. doi: 10.3390/ijerph18073741.

21.

Mousa

, Murugappan

, Badlishah

Wavelet packet transform based driver distraction level classification using EEG. Math Probl Eng. 2013. Article ID 297587. doi:10.1155/2013/297587

22.

Medic

, Mikov

, Glavaski

. The effects of aging and driving experience on reaction times of professionaldrivers. Work.. 2020;66(2):405–19. doi: 10.3233/WOR-203181.

23.

Lantieri

, Costa

, Vignali

, Acerra

, Marchetti

, Simone

. Flashing in-curb LEDs and beacons atunsignalized crosswalks and driver’s visual attention to pedestrians during nighttime. Ergonomics. 2021;64(3):330–41. doi: 10.1080/00140139.2020.1834149.

24.

Zheng

, Du

, Yan

. The impact of rhythm-based visual reference system in long highway tunnels. Safety Sci. 2017;95:75–82. doi: 10.1016/j.ssci.2017.02.006.

25.

Yang

, Du

, Huang

. Analysis of visual load on curve and longitudinal slope combination of extralong underwater tunnel based on visual sensitive region. Transportation Research Record. 2023;0(0). doi: 10.1177/03611981231188369.