Factors Affecting Rear-End Collisions in Underground Road Junctions Using VISSIM

Abstract

Due to urban overcrowding, available land is limited and traffic congestion has increased. Underground roads are being built to mitigate traffic congestion as an alternative. Studies associated with underground roads are needed because these roads are dark and closed and have a high risk of accidents compared to surface roads. In particular, there is limited study on junctions that connect two or more underground roads. In this study, an underground road network including junctions was constructed to analyze the factors behind rear-end collisions at underground road connections. To reflect the driving behavior on underground roads, the scenario analysis was conducted by applying the speed distribution of underground roads in Korea. The results of the analysis showed that variables such as acceleration standard deviation and lateral position standard deviation are crucial for accidents on underground roads. Thus, this study can be used as a basis for traffic management and safety improvement in the operation of underground road junctions in the future.

1. Introduction

Cities provide numerous opportunities, services, and facilities to their citizens, but at the same time, their growth results in negative impacts [1], such as land-use changes and increasing population mobility. Thus, cities face problems like traffic congestion and population imbalance [2]. In dense cities, traffic problems can be exacerbated by insufficient road capacity. In particular, Asian cities such as Seoul—the capital of Republic of Korea—experience more severe problems than Western cities because of their extremely dense environment [3]. Roads in dense areas cannot meet demand because of the rapidly decreasing road capacity per person [4].

Underground spaces serve an important role in the sustainability of cities as a means of making more rational use of scarce land [5]. In particular, underground roads are the easiest way to access the saturated surface space, making them ideal for solving problems that exist in cities and their transportation systems [6]. In the 2000s, many underground highways were opened, such as the Big Dig in Boston, the A86 West Beltway Tunnel in France, and Södra Länken in Sweden, which helped regenerate cities by reducing traffic congestion on urban road networks [7].

On the other hand, underground roads are dark, closed environments, and have a higher risk of accidents than aboveground roads [8,9]. Liu et al. [10] argued that it is not reasonable to apply the design criteria for surface roads to underground roads because of their environmental differences. In addition, Qin et al. [11] showed that the visual information inside a tunnel, which is a similar environment to an underground road, affects driver perception. The lack of visibility leads to a higher risk of accidents [12].

Because underground road junctions are designed to be curved, unlike mainline roads, they have even less visibility. Thus, it is crucial to prevent incidents through traffic management. Despite these risks, there are limited studies on the determinants of accidents within underground junctions that enable access between two or more underground roads. Most of the underground roads that have been opened so far are single roads, and none of them have yet implemented junctions. However, a study considering junctions in future underground networks is necessary.

The objective of this study was to analyze the factors that influence the possibility of rear-end collisions on underground road junctions. For this purpose, a network was constructed using VISSIM version 11, a microscopic traffic simulation program. Traffic information such as the speed and acceleration of individual vehicles and surrogate safety measures (SSM) were considered among the factors that influence the occurrence of accidents at the junction of underground roads.

The paper is organized as follows. Section 2 reviews the research related to risk factors and SSM, while Section 3 describes the network and methodology used. Section 4 summarizes and presents the results, and Section 5 concludes with a final discussion of the results.

2. Literature Review

Currently, underground road junctions are not being constructed, making it challenging to analyze real-world data. Simulation-based studies are particularly useful in evaluating the safety of roads that have not yet been constructed, as they are more cost-effective than actual road construction and allow for the assessment of safety issues in advance. Numerous studies on underground roads have been conducted using simulation tools.

2.1. Evaluating the Effectiveness of Underground Roads

Driving simulators are an effective method for collecting data from subjects driving in a virtual environment to analyze behavioral patterns based on different driving conditions [13]. Several researchers have employed driving simulators to observe driver behavior in underground roads or tunnel environments similar to underground roads, focusing on how design factors influence driver behavior [10,14,15,16]. Driving simulators have also been utilized to evaluate safety under varying conditions, including weather, lighting, and human factors [17,18,19].

However, analyzing complex interactions between drivers can be challenging with driving simulators, as it typically involves one or two users at a time. In contrast, microscopic traffic simulation is well suited for analyzing intricate interactions between vehicles [13]. Therefore, traffic simulation is necessary to evaluate complex systems, such as underground road junctions.

Some studies have employed traffic simulation to assess operational effectiveness in underground roads, and many others have used it to evaluate safety in underground roads and tunnels [20,21,22,23,24,25]. Factors such as speed variance, traffic volume, travel time, and queue length have been analyzed to assess safety [20,21,22,23]. Existing research has significantly contributed to understanding underground road safety, including aspects like road design, lighting, and secondary crash risks. However, there remains a gap in research concerning the safety and operation of junctions, which are crucial components of the underground road network. Therefore, this study aims to conduct an in-depth analysis of underground road junctions and their related safety factors using traffic simulation, focusing on assessing accident risk through surrogate safety measures (SSMs) such as time-to-collision (TTC), crash potential index (CPI), and deceleration rate to avoid the crash (DRAC).

2.2. Safety Evaluation Studies with SSMs

SSMs allow for the estimation of accidents without relying on actual accident data, making them a valuable tool for safety evaluation through simulation [24,26,27,28,29]. Upon reviewing the literature, TTC, CPI, and DRAC emerged as commonly used metrics for estimating accident risk. CPI and DRAC, in particular, can be assessed when vehicle information is collected within a connected vehicle environment that enables information exchange between vehicles.

TTC is especially useful for predicting rear-end collisions by measuring the time [30] until a collision occurs, assuming that two vehicles in the following situation maintain their current speeds. In this study, we considered TTC as an SSM to identify the crash risk at road junctions.

3. Methodology

The accident risk analysis was conducted through a microscopic traffic simulation and involved setting up an analysis network on VISSIM [31], running simulations for each scenario, generating data for analysis, and performing logistic regression analysis. Figure 1 explains the research framework and shows how the study was conducted.

3.1. Simulation Network and Scenario

Because there are currently no existing three-dimensional intersections in operational underground roads in Republic of Korea, a virtual simulation network was constructed. The simulation network was set up according to the Highway Capacity Manual [32], Explanation of the Rules for Structural and Facility Standards for Roads [33], and Design Guidelines for Underground Roads in Urban Areas [34], which are the Korean design guidelines. The network consisted of a diamond-shaped underground-underground junction with a main line of 10 km and connections of 1 km.

Underground roads and the surrounding areas, including ramp, diverge areas, and merge areas, were selected as the studied area. Figure 2 shows the simulation network used in the study and the scope of the studied area where the data were analyzed. Table 1 describes the simulation network components.

Table 2 describes the simulation scenarios set up for the analysis. The simulation was conducted by dividing the analysis into cases with and without driving speed distribution in the underground road. To collect individual vehicle information through the simulation, we set up scenarios according to the mainline traffic volume and the rate of vehicles entering the connecting road. In the case of the mainline R1 and R2 in Figure 2, the traffic volume varied from 1600 pcph (LOS B) to 4000 pcph (LOS E) depending on the level of service, and in the case of the connecting ramp, the percentage of vehicles entering the connecting ramp from the mainline varied from 15% to 40% in 5% increments. Each scenario was simulated 30 times, and 1 h of individual vehicle driving data (not including the first 600 s as a warm-up period) was used for analysis.

To construct the speed distribution, we used data from the Shinwol-Yeoui underground road in Seoul. Figure 3 shows the Shinwol-Yeoeui underground road in Seoul. The Shinwol-Yeoui underground road is a 7.53 km long road that connects Shinwol-dong, Yangcheon-gu district to Yeouido-dong, Yeongdeungpo-gu district. The road has a speed limit of 80 km/h from the Shinwol direction to Yeoui JCT and 60 km/h after Yeoui JCT. There are also two off-ramps from Yeoui JCT to Yeoui-daero and Olympic-daero, and the off-ramp to Olympic-daero is a single lane, which is similar to the simulation network in this study.

VDS (vehicle detection system) data collected for two months from 1 September 2021 to 31 October 2021 were used. Figure 4 shows the speed distribution set up for the underground road environment. Based on the average hourly section speed and traffic volume, we created speed distributions for the main line, diverging area, and road connection and applied them to VISSIM.

3.2. Data Processing

After data collection, TTC and density were estimated based on speed, location, and link information. The TTC was calculated using the speed and location information of the lead vehicle and the target vehicle, employing Equation (1).

(1)${TTC}_i\left(t\right)={\displaystyle \frac{X_{i-1}\left(t\right)-X_i\left(t\right)-L_i}{V_i\left(t\right)-V_{i-1}\left(t\right)}}\le 1.5$

where X: $V\left(t\right)$: vehicle speed;

$X\left(t\right)$: vehicle position;

$L_i$: subject vehicle’s length;

$X_{i-1}\left(t\right)-X_i\left(t\right)$: relative distance;

$V_i\left(t\right)-V_{i-1}\left(t\right)$: relative speed.

The Surrogate Safety Assessment Model (SSAM) (version 3.0), a road safety performance analysis software provided by the Federal Highway Administration (FHWA) of the United States, sets 1.5 s as the threshold for TTC. Thus, in this study, the threshold for TTC was set to 1.5 s to classify cases with a TTC ≤ 1.5 s as conflict situations and cases with TTC > 1.5 s as non-conflict situations [35].

The processed data were aggregated by segment and time for analysis. The analysis area was divided into 50 m segments, with a total of 116 segments. The aggregation time interval was 1 s, and the information of 250 m worth of vehicles, including the center segment (k) and the two segments before and after it, was aggregated and used for the analysis, as illustrated in Figure 5 (i.e., the information was aggregated from 116 segments every second). Table 3 shows the final variables after data processing, and Table 4 presents a sample of the collected data.

3.3. Logistic Regression

Logistic regression analysis was performed using the processed data. Logistic regression is a methodology used to estimate the probability of an accident by considering the variables that affect the risk of it occurring. Because we categorized conflict and non-conflict situations based on the TTC in Section 3.2, logistic regression can be used to identify factors that affect the probability of an accident.

The log odds of observing an incident (Y = 1) are estimated with each combination of explanatory variables, and the goodness of fit of the variables can be increased through Pearson residual values. The parameter estimates, standard errors, odds ratios, and confidence intervals of the logistic can also be visualized [36,37].

4. Results

4.1. Simulation Result

Table 5 shows the percentage of accidents with and without speed distribution. When the average number of conflicts was determined, a total of 108,000 pieces of information were collected per hour of simulation. In the scenario without speed distribution, 36.8% of the total (39,783 on average) were conflicts, whereas in the scenario with speed distribution, 52.5% of the total (56,702 on average) were conflicts.

Figure 6 illustrates the difference in conflict rates by location between the without and with scenarios. It shows that diverge area, merge area, and ramp are the most common conflicts, with the with scenario having the highest conflict rates. This suggests that the proportion of conflicts on underground roads is higher than that on surface roads, and it is thus necessary to identify and manage variables that have a substantial effect on the safety of underground road junctions.

Table 6 presents the changes in the values according to whether speed distribution was applied as well as the R1 LOS change. When comparing the results by LOS, the average acceleration and average speed were found to gradually decrease as the traffic situation worsened. As spacing decreased, the acceleration standard deviation and density increased. When evaluating the changes in the variables depending on whether speed information was applied, we confirmed that the number of conflicts increased, speed and acceleration decreased, and density increased.

4.2. Logisitic Regression Result

To identify accident factors in the “with” scenario, logistic regression analysis was performed using Python’s sklearn.linear_model tool. Sixteen variables, including traffic dummy variables, were used in the logistic regression analysis. The variance inflation factor (VIF) test was performed to check for multicollinearity problems between the variables [38]. VIF is used to test whether the correlation between predictor variables affects the accuracy of the model and reflects the severity of multicollinearity. VIF measures the effects on model accuracy by evaluating the correlation between predictor variables (which reflects the severity of multicollinearity). Typically, a VIF value ≥ 5 means high collinearity. If VIF > 10, the correlation between variables is high, affecting the prediction accuracy of the model [39].

Here, the VIF values of all variables were < 1, indicating a low correlation. The VIFs of the average speed and average lateral position variables were 0.347 and 0.001, respectively, demonstrating a very low correlation.

Table 7 presents the estimated coefficients, Wald Chi-square, odds ratio, and p-value. The Wald Chi-square statistic and p-value were used to test the significance of variables used in the logistic regression analysis. The odds ratio is a measure that shows how many times the accident frequency increases when the unit of a variable increases by 1, assuming that the values of all variables, except for that one variable, are the same. The odds ratio was used to determine the effects of variables on accidents.

The logistic regression analysis revealed that the variables that affected the increase in accidents were the average acceleration, average density, acceleration standard deviation, and lateral standard deviation. In R1, where the diverging area exists, the possibility of an accident was found to increase from LOS B to LOS D. By contrast, the variables that affected the decrease in accidents as their value increased were the average speed, average lateral position, average distance between the vehicles, speed standard deviation, standard deviation of the distance between the vehicles, and density standard deviation. In R2, which includes the merging area, accidents were found to decrease from LOS B to LOS D, but the effect was small.

As a result of the odds ratio analysis, the variables that have a large impact per unit are the standard deviation of the lateral position and the standard deviation of acceleration. When one unit increases, the occurrence of accidents increases by 8.147 times in the standard deviation of the lateral position, and increases by 0.935 times in the standard deviation of acceleration.

The simulation results revealed that the accident rate increases when a speed distribution is applied to the underground road, compared to when it is not applied. As the traffic flow increased, the average acceleration and average speed decreased, while the standard deviations of acceleration and density increased.

When the factors affecting accidents were estimated through logistic regression analysis, variables such as acceleration, speed, lateral position, and density were found to be related to accidents. In particular, the odds ratio of the lateral position and acceleration standard deviation is higher than other variables, suggesting that special management measures are needed to manage acceleration and lateral position standard deviation.

Previous studies have shown that lateral position [10,14,40,41,42] and acceleration [14,43,44,45,46,47] are variables that directly affect accidents in various environments. On surface roads, lateral position has been shown to be managed through management of edgelines, guardrail design, etc., and speed management policies have been shown to be enforced through automated speed enforcement.

5. Conclusions

Logistic regression analysis was used to examine accident risk factors in the VISSIM network. This is a proactive method for safety management and operation of underground road junctions that do not exist yet. We aimed to simulate the underground road environment by reflecting the speed and traffic volume of an actual underground road. To realize the underground road environment, we specifically configured the simulation based on VDS data of Korean underground roads.

To predict the risk of accidents in a large-depth environment, the analysis was conducted using VISSIM data, and the speed and traffic volume of the Shinwol-Yeoeui underground road were reflected in the speed distribution. Time to collision (TTC) was used as the surrogate safety measure (SSM) to determine whether there was a conflict, and each datum was aggregated by segment to collect the traffic information of the surrounding segments depending on whether there was an accident in the target segment.

Speed, acceleration, and density decreased when underground road speed information was applied and when the level of service (LOS) of the access road was worsened in the underground road environment. In addition, the simulation results show that the proportion of conflicts is higher than that of non-conflicts in scenarios that reflect the speed distribution of underground roads. Thus, it is necessary to manage accidents in underground road junctions. The analysis revealed that the acceleration standard deviation and lateral standard deviation were the main contributing factors to the accident. Variables such as acceleration and lateral position are affected when driving in a tunnel environment similar to an underground road, so management of these variables is particularly necessary when operating on an underground road.

For example, it is expected that improving lane markings in underground roads to encourage drivers to keep their lane, and maintaining uniform speeds through speed enforcement, will reduce the risk of accidents in underground roads. In the future, a more detailed analysis will be required to analyze the effects of different management strategies in underground roads.

In addition, when the LOS of road R1 with a classifier increases from B to C and from C to D, the accident risk increases considerably, so it seems that traffic operation management should be performed before LOS D. In addition, the probability of an accident decreases when driving close to the left on a junction with a right-hand curve, so installing a sufficiently wide shoulder is expected to have a positive effect on accident reduction.

Notwithstanding, this study has some limitations. Due to data acquisition difficulties, only two months of VDS data were used to adjust the simulation parameters, making it difficult to understand long-term traffic patterns and accident risk. It is expected that more detailed traffic conditions can be constructed through long-term data analysis. And the data also do not reflect the visual elements of underground roads and differ from the real environment. Therefore, future studies should further utilize driving simulator-based data to reflect lighting conditions and underground roads components. Data that reflect driver behavior are expected to provide more realistic results.

Author Contributions

Conceptualization, G.L. and C.Y.; methodology Z.P. and G.L.; validation, Z.P., G.L., C.Y. and J.-K.L.; formal analysis, Z.P. and G.L.; writing—original draft preparation, Z.P.; writing—review and editing, G.L.; supervision, G.L. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are contained within the article.

Conflicts of Interest

Author Jin-Kak Lee was employed by the company NAEIL Engineering & Consultants. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1. Research framework.

Figure 2. Simulation network (left: design of network, Right: Analysis Section of Network).

Figure 3. Shinwol–Yeoeui underground road in Seoul.

Figure 4. Speed distribution on Sinwol-Yeoui.

Figure 5. Aggregate traffic information around segments by time interval.

Figure 6. Conflict rate by scenario by location.

Table 1

Description of the simulation network.

Component	Content
Lane	Main road: 2 lanes one way
	Connection road: 1 lane
Vehicle composition	Underground road: Car only
	Ground road: All vehicle
Link Length	Main road: 10 km
	Connection road: 1 km
Designed speed	Main road: 80 km/h
	Connection road: 60 km/h
Gradient of link	Main road: 8%
	Connection road: 10%
Lane width	Main road: 3.5 m
	Connection road: 3.5 m
Capacity	Main road: 4000 pcph (veh/hour) 1
	Connection road: 1800 pcph (veh/hour) 1
Acceleration and deceleration lane	Acceleration: 205 m
	Deceleration: 180 m
Separation distance between JC	Main road: 2 km
Curve radius	Main road: 450 m
	Connection road: 250 m

¹ pcph: passenger cars per hour.

Table 2

Description of the simulation scenarios.

Speed Distribution	Volume of R1	Volume of R2	Access Rate to Ramp
Speed on surface roads(Without)	LOS B	LOS B	15% of R1 traffic	30% of R1 traffic
	LOS C	LOS C	20% of R1 traffic	35% of R1 traffic
Speed on underground(With)	LOS D	LOS D	25% of R1 traffic	40% of R1 traffic
	LOS E	LOS E

Table 3

Description of the study variables.

Variable	Description	Variable	Description
Time frame	Simulation seconds	DensityMean	Acceleration average
Segment	Analysis points (No.)	AccelerationStd	Standard deviation of acceleration
Conflict	Conflict or not at segment m(conflict: 1, non-conflict: 0)	SpeedStd	Standard deviation of speed
AccelerationMean	Average of acceleration	PosLatStd	Standard deviation of lateral position
SpeedMean	Average speed	SpacingStd	Standard deviation of distance between vehicles
PosLatMean	Average of lateral position	DensityStd	Standard deviation of density
SpacingMean	Average of distance between vehicles

Table 4

Aggregated data sample.

Time Frame	Segment	Conflict	AccelerationMean	SpeedMean	PosLatMean	SpacingMean	DensityMean	AccelerationStd	SpeedStd	PosLatStd	SpacingStd	DensityStd
601	83	0	−0.1	78.3	0.5	30.9	48.6	0.2	1.0	7.8	0.1	12.8
601	84	0	−0.1	77.9	0.5	28.4	47.6	0.2	1.1	7.7	0.1	10.1
601	85	0	−0.1	77.0	0.5	33.0	46.4	0.1	1.0	7.5	0.1	22.5
601	86	0	0.3	79.8	0.5	35.9	46.6	0.1	0.6	4.7	0.2	26.9
601	87	0	0.1	74.1	0.5	33.6	30.5	0.0	0.7	13.4	0.2	31.0
⋮						…
4200	112	1	−1.4	57.0	0.5	61.8	34.4	0.0	3.2	21.7	0.1	40.9
4200	113	0	−1.4	57.0	0.5	61.8	34.4	0.1	3.2	21.7	0.1	40.9
4200	114	0	−1.2	58.8	0.5	64.9	35.7	0.1	3.4	22.7	0.1	43.1
4200	115	0	0.1	80.7	0.5	90.3	29.5	0.1	0.5	9.1	0.0	31.7
4200	116	0	−0.1	77.0	0.5	65.5	24.3	0.0	0.3	11.7	0.0	55.9

Table 5

Conflict rates by scenario.

	Without		With
Non-conflict	68,217	63.2%	51,298	47.5%
Conflict	39,783	36.8%	56,702	52.5%
Total	108,000	100%	108,000	100%

Table 6

Analysis results by scenario.

		Mean	LOS B	LOS C	LOS D	LOS E
Without	AccelerationMean [m/s2]	−0.1	−0.1	−0.1	−0.1	−0.1
	SpeedMean [km/h]	61.6	70.0	68.9	63.0	47.8
	PosLatMean [m]	314.9	332.6	339.5	332.7	266.8
	SpacingMean [m]	55.9	66.2	64.2	55.6	41.4
	DensityMean [veh/km]	40.3	18.4	21.3	36.6	76.0
	AccelerationStd [m/s2]	1.1	0.7	0.8	1.0	1.6
	SpeedStd [km/h]	16.6	12.6	12.8	15.9	23.2
	PosLatStd [m]	316.5	329.7	331.1	324.2	288.1
	SpacingStd [m]	54.6	65.2	60.8	51.6	43.5
	DensityStd [veh/km]	30.3	15.6	17.0	27.4	55.0
With	AccelerationMean [m/s2]	−0.1	0.0	−0.1	−0.1	−0.1
	SpeedMean [km/h]	32.2	40.7	34.0	27.1	27.0
	PosLatMean [m]	0.5	0.5	0.5	0.5	0.5
	SpacingMean [m]	29.6	40.3	31.1	23.5	23.4
	DensityMean [veh/km]	75.6	48.9	65.0	93.9	94.7
	AccelerationStd [m/s2]	1.8	1.4	1.7	2.0	2.0
	SpeedStd [km/h]	21.7	22.6	22.4	21.0	20.9
	PosLatStd [m]	0.0	0.1	0.1	0.0	0.0
	SpacingStd [m]	35.5	45.7	37.9	29.4	29.1
	DensityStd [veh/km]	35.8	26.4	32.0	42.5	42.5

Table 7

Statistic results of logistic regression models.

Variable	Coefficient	Odds Ratio	p-Value
Intercept	0.0213	1.022	<0.01
AccelerationMean	0.0742	1.077	<0.01
SpeedMean	−0.0106	0.989	<0.01
PosLatMean	−0.469	0.626	<0.01
SpacingMean	−0.0134	0.987	<0.01
DensityMean	0.0114	1.011	<0.01
AccelerationStd	0.6599	1.935	<0.01
SpeedStd	−0.0032	0.997	<0.01
PosLatStd	2.2134	9.147	<0.01
SpacingStd	−0.0075	0.993	<0.01
DensityStd	−0.0034	0.997	<0.01
R1_C	0.0321	1.033	<0.01
R1_D	0.0595	1.061	<0.01
R1_E	0.0638	1.066	<0.01
R2_C	−0.0028	0.997	<0.01
R2_D	−0.0362	0.964	<0.01
R2_E	−0.0474	0.954	<0.01

(Pseudo R-square: 0.267, log-likelihood: −131,270,000).

References

1. Durmisevic, S. The future of the underground space. Cities; 1999; 16, pp. 233-245. [DOI: https://dx.doi.org/10.1016/S0264-2751(99)00022-0]

2. Tsagkis, P.; Bakogiannis, E.; Nikitas, A. Analysing urban growth using machine learning and open data: An artificial neural network modelled case study of five Greek cities. Sustain. Cities Soc.; 2023; 89, 104337. [DOI: https://dx.doi.org/10.1016/j.scs.2022.104337]

3. Barter, P. Urban Transport in Asia: Problems and prospects for high-density cities1. Asia-Pac. Dev. Monit.; 2000; 2, pp. 33-66.

4. Eom, H.-J.; Cho, G.-H. Exploring thresholds of built environment characteristics for walkable communities: Empirical evidence from the Seoul Metropolitan area. Transp. Res. Part D Transp. Environ.; 2015; 40, pp. 76-86. [DOI: https://dx.doi.org/10.1016/j.trd.2015.07.005]

5. Bobylev, N. Mainstreaming sustainable development into a city’s Master plan: A case of Urban Underground Space use. Land Use Policy; 2009; 26, pp. 1128-1137. [DOI: https://dx.doi.org/10.1016/j.landusepol.2009.02.003]

6. Wang, W.; Hu, Y.; Liu, Q. A New Set of Social Benefit Evaluation Parameters for Underground Road Project. Proceedings of the 2010 International Conference on Management and Service Science; Wuhan, China, 24–26 August 2010; pp. 1-4.

7. Benardos, A.; Kazos, X.; Sotiropoulos, N. Design Analysis for the Underground Relocation of a Central Road Artery in Athens; The Society for Rock Mechanics & Engineering Geology: Singapore, 2013; pp. 557-566.

8. Feng, Z.; Yang, M.; Zhang, W.; Du, Y.; Bai, H. Effect of longitudinal slope of urban underpass tunnels on drivers’ heart rate and speed: A study based on a real vehicle experiment. Tunn. Undergr. Space Technol.; 2018; 81, pp. 525-533. [DOI: https://dx.doi.org/10.1016/j.tust.2018.08.032]

9. Yeung, J.S.; Wong, Y.D. Road traffic accidents in Singapore expressway tunnels. Tunn. Undergr. Space Technol.; 2013; 38, pp. 534-541. [DOI: https://dx.doi.org/10.1016/j.tust.2013.09.002]

10. Liu, S.; Wang, J.; Fu, T. Effects of Lane Width, Lane Position and Edge Shoulder Width on Driving Behavior in Underground Urban Expressways: A Driving Simulator Study. Int. J. Environ. Res. Public Health; 2016; 13, 1010. [DOI: https://dx.doi.org/10.3390/ijerph13101010]

11. Qin, X.; Zhang, N.; Zhang, W.; Meitner, M. How does tunnel interior color environment influence driving behavior? Quantitative analysis and assessment experiment. Tunn. Undergr. Space Technol.; 2020; 98, 103320. [DOI: https://dx.doi.org/10.1016/j.tust.2020.103320]

12. Zhang, Y. Analysis of the Relation between Highway Horizontal Curve and Traffic Safety. Proceedings of the 2009 International Conference on Measuring Technology and Mechatronics Automation-Volume 03; Zhangjiajie, China, 11–12 April 2009; IEEE Computer Society: Los Alamitos, CA, USA, 2009; pp. 479-481.

13. Sun, J.; Ma, Z.; Li, T.; Niu, D. Development and application of an integrated traffic simulation and multi-driving simulators. Simul. Model. Pract. Theory; 2015; 59, pp. 1-17. [DOI: https://dx.doi.org/10.1016/j.simpat.2015.08.003]

14. Calvi, A.; D’Amico, F. A study of the effects of road tunnel on driver behavior and road safety using driving simulator. Adv. Transp. Stud.; 2013; 30, pp. 59-76.

15. Huang, Y.; Chen, F.; Song, M.; Pan, X.; You, K. Effect evaluation of traffic guidance in urban underground road diverging and merging areas: A simulator study. Accid. Anal. Prev.; 2023; 186, 107036. [DOI: https://dx.doi.org/10.1016/j.aap.2023.107036] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36996603]

16. Xia, Z.; Lv, Y.; Pan, X.; Chen, F.; Xu, M.; Wu, G.; Feng, D. Research on Design Pattern of City Tunnel Side Wall Based on the Driver Visual Effect. Proceedings of the Advances in Human Aspects of Transportation; Los Angeles, CA, USA, 17−21 July 2017; Stanton, N.A.; Landry, S.; Di Bucchianico, G.; Vallicelli, A. Springer International Publishing: Cham, Switzerland, 2017; pp. 689-701.

17. Domenichini, L.; La Torre, F.; Vangi, D.; Virga, A.; Branzi, V. Influence of the lighting system on the driver’s behavior in road tunnels: A driving simulator study. J. Transp. Saf. Secur.; 2017; 9, pp. 216-238. [DOI: https://dx.doi.org/10.1080/19439962.2016.1173155]

18. Zhao, L.; Jiang, H.; Hu, J. Driver’s Physiological and Mental Reaction in Tunnel. ICCTP 2011: Towards Sustainable Transportation Systems, Proceedings of the 11th International Conference of Chinese Transportation Professionals (ICCTP), Nanjing, China, 14–17 August 2011; American Society of Civil Engineers: Reston, VA, USA, 2012; pp. 1760-1766. [DOI: https://dx.doi.org/10.1061/41186(421)174]

19. Chen, F.; Peng, H.; Ma, X.; Liang, J.; Hao, W.; Pan, X. Examining the safety of trucks under crosswind at bridge-tunnel section: A driving simulator study. Tunn. Undergr. Space Technol.; 2019; 92, 103034. [DOI: https://dx.doi.org/10.1016/j.tust.2019.103034]

20. Yang, C.H.; Park, S.B.; Kim, J.G.; Moon, J.P. Development and Application of Traffic Control Framework for Underground Roads. 2022; Available online: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4291161 (accessed on 20 August 2024).

21. Xing, Y.; Lu, J.; Lu, L.; Jiang, C.; Cai, X. Traffic Safety Evaluation Model of Road Tunnels Based on VISSIM. Proceedings of the Fourth International Conference on Transportation Engineering; Chengdu, China, 19–20 October 2013; pp. 1075-1085. [DOI: https://dx.doi.org/10.1061/9780784413159.157]

22. Zhang, B.; Zhang, L.; Zhou, X.; Yu, J.; Liu, G. Traffic Accident Influence in Freeway Tunnel Simulation Analysis Based on the VISSIM; Atlantis Press: Dordrecht, The Netherlands, 2016; pp. 223-227.

23. Yi, L.; Xiang, Q.; Len, H. Impact of Lighting on Traffic Safety in Freeway Tunnel. ICCTP 2011: Towards Sustainable Transportation Systems, Proceedings of the 11th International Conference of Chinese Transportation Professionals (ICCTP), Nanjing, China, 14–17 August 2011; American Society of Civil Engineers: Reston, VA, USA, 2012; pp. 2358-2366. [DOI: https://dx.doi.org/10.1061/41186(421)235]

24. Yu, S.; Chen, Y.; Song, L.; Xuan, Z.; Li, Y. Modelling and Mitigating Secondary Crash Risk for Serial Tunnels on Freeway via Lighting-Related Microscopic Traffic Model with Inter-Lane Dependency. Int. J. Environ. Res. Public Health; 2023; 20, 3066. [DOI: https://dx.doi.org/10.3390/ijerph20043066]

25. Wang, Q.; Jiang, X.; Park, H.; Wang, M. HGV fire risk assessment method in highway tunnel based on a Bayesian network. Tunn. Undergr. Space Technol.; 2023; 140, 105247. [DOI: https://dx.doi.org/10.1016/j.tust.2023.105247]

26. Weng, J.; Meng, Q. Rear-end crash potential estimation in the work zone merging areas. J. Adv. Transp.; 2014; 48, pp. 238-249. [DOI: https://dx.doi.org/10.1002/atr.211]

27. Zhao, P.; Lee, C. Assessing rear-end collision risk of cars and heavy vehicles on freeways using a surrogate safety measure. Accid. Anal. Prev.; 2018; 113, pp. 149-158. [DOI: https://dx.doi.org/10.1016/j.aap.2018.01.033]

28. Wu, Y.; Abdel-Aty, M.; Wang, L.; Rahman, M.S. Combined connected vehicles and variable speed limit strategies to reduce rear-end crash risk under fog conditions. J. Intell. Transp. Syst.; 2020; 24, pp. 494-513. [DOI: https://dx.doi.org/10.1080/15472450.2019.1634560]

29. Xing, L.; He, J.; Abdel-Aty, M.; Wu, Y.; Yuan, J. Time-varying Analysis of Traffic Conflicts at the Upstream Approach of Toll Plaza. Accid. Anal. Prev.; 2020; 141, 105539. [DOI: https://dx.doi.org/10.1016/j.aap.2020.105539]

30. Hayward, J. Near Misses as a Measure of Safety at Urban Intersections; Pennsylvania State University, Department of Civil Engineering: University Park, PA, USA, 1971.

31. PTV. Vissim 2020 User Manual; PTV AG: Karlsruhe, Germany, 2020.

32. Ministry of Land, Transport and Martime Affairs. Korean Highway Capacity Manual; Ministry of Land, Transport and Maritime Affairs: Sejong, Republic of Korea, 2013.

33. Ministry of Land, Infrastructure and Transport. Explanation of the Rules for Structural and Facility Standards for Roads; Ministry of Land, Infrastructure and Transport: Sejong, Republic of Korea, 2020.

34. Ministry of Land, Infrastructure and Transport. Design Guidelines for Underground Roads in Urban Areas; Ministry of Land, Infrastructure and Transport: Sejong, Republic of Korea, 2016.

35. Pu, L.; Joshi, R. Siemens Energy and Automation. Business Unit Intelligent Transportation Systems. Surrogate Safety Assessment Model (SSAM)—Software User Manual; Federal Highway Administration: McLean, VA, USA, 2008.

36. Kuhnert, P.M.; Do, K.-A.; McClure, R. Combining non-parametric models with logistic regression: An application to motor vehicle injury data. Comput. Stat. Data Anal.; 2000; 34, pp. 371-386. [DOI: https://dx.doi.org/10.1016/S0167-9473(99)00099-7]

37. Ma, L.; Yan, X. Examining the nonparametric effect of drivers’ age in rear-end accidents through an additive logistic regression model. Accid. Anal. Prev.; 2014; 67, pp. 129-136. [DOI: https://dx.doi.org/10.1016/j.aap.2014.02.021] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24642249]

38. Greene, W.H. LIMDEP Econometric Modeling Guide, Version 8.0; Econometric Software: Plainview, NY, USA, 2002.

39. Uddin, M.A.; Debnath, M.; Roy, S.; Adiba, S.; Talukder, M.M.A. Identifying the Smoking and Smokeless Tobacco-Related Predictors on Frequencies of Heavy Vehicle Traffic Accidents in Bangladesh: Linear and Binary Logistic Regression-Based Approach. Adv. Civ. Eng.; 2023; 2023, e7116057. [DOI: https://dx.doi.org/10.1155/2023/7116057]

40. van Driel, C.J.G.; Davidse, R.J.; van Maarseveen, M.F.A.M. The effects of an edgeline on speed and lateral position: A meta-analysis. Accid. Anal. Prev.; 2004; 36, pp. 671-682. [DOI: https://dx.doi.org/10.1016/S0001-4575(03)00090-3] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15094422]

41. Bella, F. Driver perception of roadside configurations on two-lane rural roads: Effects on speed and lateral placement. Accid. Anal. Prev.; 2013; 50, pp. 251-262. [DOI: https://dx.doi.org/10.1016/j.aap.2012.04.015]

42. Charly, A.; Mathew, T.V. Estimation of traffic conflicts using precise lateral position and width of vehicles for safety assessment. Accid. Anal. Prev.; 2019; 132, 105264. [DOI: https://dx.doi.org/10.1016/j.aap.2019.105264]

43. Guo, M.; Zhao, X.; Yao, Y.; Yan, P.; Su, Y.; Bi, C.; Wu, D. A study of freeway crash risk prediction and interpretation based on risky driving behavior and traffic flow data. Accid. Anal. Prev.; 2021; 160, 106328. [DOI: https://dx.doi.org/10.1016/j.aap.2021.106328]

44. Hamzeie, R.; Savolainen, P.T.; Gates, T.J. Driver speed selection and crash risk: Insights from the naturalistic driving study. J. Saf. Res.; 2017; 63, pp. 187-194. [DOI: https://dx.doi.org/10.1016/j.jsr.2017.10.007]

45. Chen, Q.; Huang, H.; Li, Y.; Lee, J.; Long, K.; Gu, R.; Zhai, X. Modeling accident risks in different lane-changing behavioral patterns. Anal. Methods Accid. Res.; 2021; 30, 100159. [DOI: https://dx.doi.org/10.1016/j.amar.2021.100159]

46. Waizman, G.; Shoval, S.; Benenson, I. Traffic accident risk assessment with dynamic microsimulation model using range-range rate graphs. Accid. Anal. Prev.; 2018; 119, pp. 248-262. [DOI: https://dx.doi.org/10.1016/j.aap.2018.07.027]

47. Liu, P.; Zhang, X.; Wang, W.; Xu, C. Driver Response to Automated Speed Enforcement on Rural Highways in China. Transp. Res. Rec.; 2011; 2265, pp. 109-117. [DOI: https://dx.doi.org/10.3141/2265-12]