Impact of Intersection Left Turn Guide Lines Configuration on Novice Drivers’ Behavior

Abstract

Novice drivers often face challenges such as misjudgment, inappropriate steering control, distraction, and insufficient speed control when making left turns at intersections, leading to safety hazards. Installing intersection guide lines offers a solution by providing clear path directions, mitigating safety concerns associated with novice drivers’ left-turn actions. This study explored the impact of intersection guide line configurations on the driving behavior of novice drivers during left turns, utilizing large, medium, and small typical intersections to create six categories of left-turn simulation scenarios in a driving simulator. Data on vehicle trajectory, steering angle, steering speed, and eye-tracking were collected and analyzed. The study revealed that guide line arrangement significantly influences novice drivers’ left-turning behavior, enhancing path guidance while reducing trajectory and steering angle fluctuations, speed variations, and driver attention dispersion. This improvement in stability is particularly notable as intersection size and the number of left-turn lanes increase. The study’s findings offer robust theoretical support and guidance for the development and widespread adoption of intersection guide lines.

1. Introduction

Intersections, frequent sites of traffic conflicts, contribute significantly to urban accidents, with over 50% occurring at these locations, causing about 20% of total fatalities [1]. Left-turn conflicts at intersections stand out as a major concern, especially given the complexities of decision-making during left turns compared to right turns. Left turns involve longer durations and greater distances but often lack specific lane markings, presenting challenges, particularly for novice drivers. In China, with 456 million drivers by 2020, those with less than one year of experience constituted 4.90%, and those with less than three years comprised 17.48%. Novice drivers, whose accident rates are 2 to 3 times higher, underscore the importance of improving left-turn safety at intersections to reduce overall accident numbers [2].

Numerous scholars have extensively investigated left-turn conflicts at intersections, employing various research methods such as on-site observations [3,4], intersection simulation experiments [5,6,7], accident report analyses [8], and computer vision technologies [9,10,11]. For instance, Yang et al. used simulation techniques to examine the impact of left-turn path variations on traffic plans, intersection safety, and efficiency [7].

To understand the influencing factors and mechanisms of left-turn conflicts, scholars have delved into modeling and analyzing drivers’ left-turn behavior and trajectories [12,13,14,15,16,17,18,19]. Some researchers utilized eye-tracking devices to analyze drivers’ behaviors [20,21,22]. To mitigate left-turn conflicts, scholars have extensively researched intersection design and signal control, encompassing improvements in intersection design, optimized signal control, lane markings, and the integration of intelligent transportation systems [23,24,25,26]. Notably, Edara et al. assessed the effectiveness of unsignalized J-turn intersection designs, revealing a 31.2% reduction in overall accident frequency and a 63.8% decrease in injury and fatal accidents [24]. The J-turn intersection design, also known as a Reduced Conflict Intersection (RCUT), alters the way left turns and through movements are handled from cross-street approaches. Instead of making direct left turns, vehicles on minor roads first make a right turn and then perform a U-turn at a designated location to continue in their desired direction. This design helps to reduce the number of conflict points and improve overall traffic safety.

Addressing traffic conflicts involving same-direction left turns, scholars advocate for the use of guide lines at intersections. Wei et al. analyzed left-turning vehicle trajectories, speed variations, and flow changes, concluding that controlling vehicle turns with left-turn guide lines effectively reduces traffic delays and conflicts [27]. Guide lines not only assist drivers in locating exit lanes promptly, improving traffic flow efficiency, but also regulate driver behavior, mitigate conflicts, and enhance left-turning vehicle safety. Proper guide line settings contribute to increased efficiency and reduced conflict risks for left-turning vehicles at intersections.

In essence, understanding the driving behavior of left-turning drivers at intersections is crucial for minimizing intersection conflicts. While existing studies often focus on trajectory models and driving simulator analyses for left turns, there is a notable gap in research on the effectiveness of guide lines, particularly in relation to novice drivers’ left-turning behavior. In this study, novice drivers, defined as those with less than one year of driving experience, are the primary focus. The research involves designing driving simulator scenarios based on intersections in Lingang of Shanghai, analyzing novice drivers’ left-turn behavior using data on trajectories, speeds, steering angles, and eye-tracking. The goal is to investigate the impact of left-turn guide line settings on novice drivers and offer experimental evidence and recommendations for the effective implementation of left-turn guide lines at intersections. This, in turn, aims to enhance the driving safety of novice drivers.

2. Material and Methods

2.1. Intersection Survey

The intersections in the Lingang area of Shanghai were selected as the research objects for this study. The DJI MINI2 unmanned aerial vehicle was employed for intersection photography to collect geometric parameters of the intersections. A total of 32 intersections in the Lingang area were investigated, covering the intersection area, geometric shape, channelization type, and other relevant aspects. Based on the area size, the intersections were categorized into three levels: small (S ≤ 4000 m²), medium (4000 m² < S ≤ 8000 m²), and large (S > 8000 m²). Subsequently, 21 intersections were selected based on intersection area, geometric shape, and channelization type (as shown in Figure 1, where the red dots represent the selected intersections). Field surveys were conducted for these 21 intersections, and 5 intersections were identified as the focus of this study (as shown in Figure 1 and detailed in

Table 1. Basic information for the selected intersections.

Intersections	Size (m2)	Type	Entrance	Lanes Classification by Function
				Left-Turn Lane	Thru Lane	Thru Lane	Thru and Right Turn	Right-Turn Lane	Total
HC Ring RD–SY RD	3000	Small	West	1				1	2
			South	1		3			4
			North	1		2		1	4
HC Ring RD–YF RD	2862	Small	South			3		1	4
			North	2		3			5
GB RD–HHW 3rd RD	6512	Medium	West	1		1		1	3
			South	1		2	1		4
			North	2		1	1		4
HC Ring RD–SG AVE	14,994	Large	East	1		3		1	5
			West	1	1	1		1	4
			South	2		3		1	6
			North	1		3		1	5
LG AVE–HC Ring RD	8084	Large	East	2		3			5
			West	1		2		1	4
			South	3				3	6

). Among these five intersections, Hucheng Ring Road–Shangyuan Road Share Area (HC Ring RD–SY RD SA) and Hucheng Ring Road–Yinfly Road (HC Ring RD–YF RD) are classified as small intersections, each with 1 and 2 left-turn lanes, respectively. Gu Brown Road–Huanhu West 3rd Road (GB RD–HHW 3rd RD) is a medium-sized intersection with 2 left-turn lanes. Hucheng Ring Road–Shengang Avenue (HC Ring RD–SG AVE) and Lingang Avenue–Hucheng Ring Road (LG AVE–HC Ring RD) are large intersections, each with 2 and 3 left-turn lanes, respectively. Specific parameter investigations were conducted for these five intersections to facilitate the later construction of driving simulator scenarios.

2.2. Driving Simulator Experiment Design

2.2.1. Experimental Equipment

The driving simulator used in this experiment incorporates high-fidelity driving scenario simulation software (SILAB 7.2) developed by the German company WIVW. This equipment is capable of presenting realistic virtual scenes to the driver, reproducing various vehicle and environmental sounds, including those of their own vehicle, to provide a more authentic driving experience, as illustrated in Figure 2. The driving simulator comprises several components, including the following: ① projectors, ② projection screen wall, ③ cockpit, ④ motion platform, ⑤ server, and ⑥ display. Three front projectors are used to project images onto a 180-degree screen wall in front of the vehicle, simulating driving conditions. Additionally, a rear projector projects images onto a screen behind the vehicle, simulating the rearview mirror’s perspective. The simulated vehicle model is a 2015 Shanghai Volkswagen Polo sedan. The cockpit is equipped with built-in speakers to simulate various sounds during the driving process, such as weather conditions and road bumps. The motion platform, with 3 degrees of freedom (up-down, forward-backward, left-right), can simulate various driving conditions. The experiment utilizes the Dikablis Pro eye-tracking system to collect data on the driver’s eye movements during the simulation.

2.2.2. Experimental Participants

The participants in this experiment were novice drivers recruited on the campus of Shanghai Maritime University due to experimental and economic constraints. The study included 25 novice drivers, comprising 20 males and 5 females. The participants’ ages ranged from 20 to 32 years with an average age of 24 years. This study primarily selected novice drivers for testing based on the time they obtained their driver’s license and their actual driving experience. Initially, drivers who had obtained their driver’s license within the past 12 months were chosen for testing. If the drivers had obtained their driver’s license within the past 3 years and had less than 1 year of driving experience, they were also selected as test participants.

2.2.3. Experimental Scenario Design

The experimental scenarios were designed using the simulation scenario editor of SILAB 7.2.

Table 2. Designed simulation scenarios.

Intersection		Small	Medium		Large
Number of left-turn lanes		Single	Single	Dual	Single	Dual	Triple
Major scenarios number		1	2	3	4	5	6
Guide line setting	N: No guide line	1-A: S-Single-N	2-A: M-Single-N	3-A: M-Dual-N	4-A: L-Single-N	5-A: L-Dual-N	6-A: L-Triple-N
	O: One guide line	1-B: S-Single-O	2-B: M-Single-O	3-B: M-Dual-O	4-B: L-Single-O	5-B: L-Dual-O	6-B: L-Triple-O
	F: Full guide lines	1-C: S-Single-F	2-C: M-Single-F	3-C: M-Dual-F	4-C: L-Single-F	5-C: L-Dual-F	6-C: L-Triple-F

outlines the specific scenarios for the simulation experiment, encompassing 6 major categories and 18 subcategories. The scenarios were horizontally differentiated by intersection size (small, medium, or large) and the number of left-turn lanes (single, dual, or triple). Guide line settings further distinguished the scenarios: N signified intersections without guide lines, O designated the innermost left-turn lane with a guide line, and F represented guide lines for all left-turn lanes. Specific configurations are depicted in Figure 3. Additionally, some other non-player character (NPC) vehicles were added to the simulation scenarios to create a more realistic driving environment. However, it should be noted that in the intersection scenario with two or three left-turn lanes, the test vehicle was set to drive in the inner left-turn lane. As a result, the interaction between the test vehicle and any left-turning vehicle to its left was not considered.

2.2.4. Experimental Procedure

To optimize the experiment’s efficiency, participants were grouped, and simulation processes for each scenario were streamlined. Novice drivers sequentially completed simulations for the six scenario categories, each comprising three sub-scenarios with distinct guide line settings. The estimated time per participant was 40–50 min, with an expected total duration of approximately 1 h, accounting for preparation time.

Before the experiment, participants were asked to sign an informed consent form and complete a questionnaire that included their name, gender, age, and driving experience. The consent form detailed the experiment tasks and precautions. Researchers then checked equipment connections, calibrated instruments, and explained the simulator’s operation to participants. Finally, each participant underwent a 5–10 min session in the driving simulator to quickly familiarize themselves with its operation and environment, helping to minimize subjective differences in the perception of the simulator’s impact on the results.

The experiment began with drivers engaging in simulations for the six scenario categories, illustrated in Figure 4. Each category featured three intersections: Intersection 1 without guide lines (N), Intersection 2 with one guide line (O), and Intersection 3 with full guide lines (F). The green dot marks the starting point, the red dot indicates the endpoint (after passing through Intersection 3), and black arrows depict the vehicle’s direction. The pink line represents the vehicle’s path. Intersections were spaced 300 m apart, and each intersection was controlled by traffic signals. Since the six scenarios were not continuous, after completing each one, the driver was required to immediately evaluate the completed scenario, fill out a questionnaire, and prepare for the next simulation. This process was repeated until the driver completed simulations for all scenarios.

2.3. Data Extraction

In the driving simulator, real-time vehicle motion and driver gaze point data were continuously collected and recorded. Given the diverse and extensive data, relevant indicators, such as steering angle, speed, lateral acceleration, driver gaze points, and fixation time, were extracted for each driver across various scenarios to analyze novice driving behavior. Steering angle and speed were utilized to assess driving behavior characteristics in different simulation scenarios. Lateral acceleration, distance traveled, and operational time provided insights into vehicle trajectory characteristics.

The eye-tracker captured various data related to the driver’s visual behavior, including gaze direction, gaze point, gaze duration, head position, head direction, and eye movements such as blinks, gazes, and saccades. The eye-tracker data were collected at a frequency of 60 Hz, and then gaze point positions and fixation time for each AOI were extracted from these data to compare and analyze the impact of guide lines on novice driver gaze points. Before recording began, two calibration cards were used to ensure the accuracy of the eye movement data captured by the eye-tracker. One calibration card was placed in the lower left corner of the vehicle’s front window, and the other was positioned in the lower center of the front window. This calibration process is essential for precise tracking of the driver’s eye movements. This study focused exclusively on the left half of the windshield for analysis, excluding the right half, the inside and outside of the car, and the left and right sides. To further analyze the impact of left-turn guide line settings on novice driver behavior, the driver’s focus area was segmented into five sub-regions: Area 1 (upper-left), Area 2 (upper-right), Areas 3 and 4 (guide line constraints), and Area 5 (lower-left), as depicted in Figure 5. These five areas were designated as areas of interest (AOI) during the driver’s left-turn process. Simultaneously, Area 3 was identified as the effective fixation area for guide lines, while Areas 1, 2, 4, and 5 were considered areas with minimal or no influence from the guide lines.

2.4. Data Analysis Methods

The study aimed to assess the influence of left-turn guide line settings on novice drivers’ behavior by analyzing vehicle trajectory characteristics in various scenarios. For ease of analysis, it is necessary to standardize the displacement data in the left-turn vehicle trajectories of all participants across all scenarios. Based on this standardized data, the X-axis is set perpendicular to the direction of the intersection’s entry lane (lateral direction), and the Y-axis parallel to the direction of the intersection’s entry lane (longitudinal direction). Based on the X-Y coordinate system, basic descriptive statistics methods were used to conduct statistical analysis on the left-turn vehicle trajectory data. Specifically, an error band plot of the participants’ left-turn vehicle trajectory data was then generated to statistically analyze the fluctuations in their left-turn trajectories. There have been numerous studies on left-turn driving behavior and vehicle characteristics [28,29,30,31,32,33]. Some focus on vehicle control systems to explore dynamic responses during turns, while others examine driver behavior to analyze trajectory characteristics. Regardless of the approach, steering angle and speed are consistently key evaluation indicators. Drawing from relevant research [30,31,32], this study adopted commonly used evaluation metrics such as steering angle, difference in speed, and position change of vehicle trajectory to analyze the impact of guide line settings on the left-turn driving behavior of novice drivers.

Additionally, the Wilcoxon rank-sum test was used to analyze the standard deviation of vehicle trajectories in different designed scenarios for all participants. This was done to determine the effectiveness of left-turn guide lines and evaluate whether guide line settings would lead to significant differences in convergence of vehicle trajectories for novice drivers. Eye-tracker data were utilized to assess the effect of guide lines on novice drivers’ gaze areas and fixation times, evaluating the efficacy of guide lines in assisting novice drivers during left turns.

3. Results

3.1. Impact Analysis Based on Trajectory Characteristics

In the preceding sections, simulation scenarios were categorized into six major groups based on intersection size and left-turn lane numbers. Each category was then subdivided into three specific scenarios based on guide line settings, creating a total of 18 simulation scenarios. To facilitate analysis, the subsequent sections will explore the influence of guide lines on novice drivers’ driving behavior, categorized by different intersection sizes (small, medium, and large).

3.1.1. Small Intersection

Scenario 1 consists of three sub-scenarios based on guide line (1-A: no guide line, 1-B: one guide line, and 1-C: full guide lines); Figure 6 compares the trajectory characteristics of vehicles in each sub-scenario. In Figure 6, the X-axis represents the standardized lateral distance of the vehicle during a left turn, and the Y-axis represents the standardized longitudinal distance of the vehicle during a left turn. Initially, as vehicles enter the small intersection with a single left-turn lane, the fluctuation of vehicle trajectories along Y direction is relatively small. However, as vehicles approach the exit, the fluctuation range significantly increases. Notably, in scenarios with no guide line, one guide line, and full guide lines, the fluctuation ranges are similar, with maximum fluctuation ranges of 0.137, 0.132, and 0.130, respectively. This suggests that in small intersections with a single left-turn lane, guide lines have a relatively minor impact on the left-turn driving behavior of novice drivers.

3.1.2. Medium Intersection

Based on the number of guide lines, Scenario 2 and Scenario 3 can be denoted as 2-A, 3-A (no guide line), 2-B, 3-B (one guide line), and 2-C, 3-C (full guide lines). The trajectory characteristics of vehicles in each sub-scenario are compared, as illustrated in Figure 7, where the X-axis represents the standardized lateral distance of the vehicle during a left turn, and the Y-axis represents the standardized longitudinal distance of the vehicle during a left turn.

In Scenario 2, as vehicles enter the intersection, the lateral fluctuation range of vehicle trajectories is relatively small. However, as they proceed towards the exit lane, the range significantly increases, becoming notably larger in Scenario 2-A compared to 2-B and 2-C, with maximum ranges of approximately 0.115, 0.081, and 0.087, respectively. Compared to Scenario 1, lateral fluctuation range is significantly reduced in Scenarios 2-B and 2-C, indicating a noticeable impact of guide line settings on novice drivers’ left-turn behavior at medium intersections.

In Scenario 3, as vehicles enter the intersection, the lateral fluctuation range significantly increases. Scenario 3-A has the largest range, followed by 3-B and the smallest in 3-C, with maximum ranges of 0.171, 0.163, and 0.124, respectively. Guide line settings have a certain impact on novice drivers’ left-turn behavior, becoming more pronounced with an increase in the number of guide lines in Scenario 3. Compared to Scenario 2, the lateral fluctuation range at the exit lane in Scenario 3 is significantly larger, particularly in 3-A and 3-B, emphasizing the superior guiding effect of full guide lines in scenarios with dual left-turn lanes.

3.1.3. Large Intersection

In the analysis of large intersections, three scenarios were considered: a single left-turn lane (Scenario 4), dual left-turn lanes (Scenario 5), and triple left-turn lanes (Scenario 6). Each scenario was further divided into three sub-scenarios based on guide line settings: no guide line (A), one guide line (B), and full guide lines (C). The trajectory features of vehicles in each sub-scenario were compared, as shown in Figure 8, where the X-axis represents the standardized lateral distance of the vehicle during a left turn, and the Y-axis represents the standardized longitudinal distance of the vehicle during a left turn.

In Scenario 4 (single left-turn lanes, Figure 8a), the lateral fluctuation range of vehicle trajectories gradually increases towards the exit, being smallest in Scenario 4-C. Maximum lateral fluctuation ranges in the three sub-scenarios are 0.082, 0.080, and 0.056, respectively. Compared to Scenario 2, lateral fluctuation ranges notably reduce in Scenario 4, especially with full guide lines, signifying a significant impact on novice drivers’ left-turn behavior.

In Scenario 5 (dual left-turn lanes, Figure 8b), as vehicles enter the intersection, lateral fluctuation range significantly increases. In the absence of guide lines, it is the largest, followed by one guide line. It is the smallest in the full guide line scenario. Maximum lateral fluctuation ranges are 0.152, 0.135, and 0.075, respectively. This indicates a significant impact of guide line settings on novice drivers in Scenario 5, becoming more pronounced with more guide lines. Compared to Scenario 3, lateral fluctuation ranges slightly reduce in Scenario 5, especially with full guide lines, suggesting a more pronounced impact on novice drivers’ left-turn behavior.

In Scenario 6 (triple left-turn lanes, Figure 8c), as vehicles enter the intersection, lateral fluctuation range increases significantly. In the absence of guide lines, it increases more noticeably than in scenarios with one guide line and full guide lines, especially at the intersection exit. Maximum lateral fluctuation ranges are 0.174, 0.098, and 0.083, respectively. This indicates a significant impact of guide line settings on novice drivers in Scenario 6, increasing with more guide lines, though with a relatively small magnitude.

Comparing lateral fluctuation ranges of vehicle trajectories in scenarios with different left-turn lane numbers in large intersections yields the following observations: in Scenario 4, lateral fluctuation ranges remain relatively stable across the three guide line settings, with a slight guiding advantage in the full guide lines. In Scenario 5, the absence of guide lines leads to a significant increase in the lateral fluctuation range compared to scenarios with one guide line, and scenarios with one guide line exhibit a notably larger range than those with full guide lines. This emphasizes the significant impact of guide line settings on novice drivers in Scenario 5, particularly with full guide lines. In Scenario 6, guide line settings significantly impact novice drivers, and the impact increases with the number of guide lines, albeit with a relatively small magnitude.

3.1.4. Wilcoxon Rank-Sum Test for Trajectories

To further analyze the impact of guide lines on novice drivers’ left-turn behavior, a Wilcoxon rank-sum test was employed to compare the vehicle trajectory data in different scenarios, assessing the significance of similarities in vehicle trajectories across different scenarios. For ease of analysis, comparisons were made based on the number of left-turn lanes: single left-turn lane, dual left-turn lanes, and triple left-turn lanes.

The Wilcoxon rank-sum test results for trajectory data in scenarios with a single left-turn lane are shown in

Table 3. Wilcoxon rank-sum test results for trajectories in single left-turn lane scenarios.

Scenario	S-Single-N	S- Single-O	M-Single-N	M-Single-O	L-Single-N	L-Single-O
S-Single-O	p = 0.6776
S-Single-F	p = 0.3258	p = 0.5453
M-Single-O			p = 0.0284
M-Single-F			p = 0.1509	p = 0.1124
L-Single-O					p = 0.9397
L-Single-F					p = 0.0082	p = 0.0065

. The test results indicate that there is no significant difference in vehicle trajectory among different guide line settings for novice drivers in small intersections (all p-values > 0.05). This suggests that guide lines have minimal impact on novice drivers’ behavior in small intersections, consistent with the analysis of trajectory fluctuation.

In medium intersections with a single left-turn lane, there is a significant difference in vehicle trajectory between scenarios with one guide line and scenarios without guide lines at the 5% confidence level (p-value = 0.028). This suggests that in medium intersections with a single left-turn lane, the presence of one guide line has a certain impact on vehicle trajectories. However, there is no significant difference in vehicle trajectory between scenarios with full guide lines and scenarios without guide lines (p-value = 0.15). This result is slightly different from the analysis of trajectory fluctuation ranges, but the difference is within an acceptable range. The maximum difference between the two methods is 0.028, indicating overall consistency in conclusions.

In large intersections with a single left-turn lane, there is no significant difference in vehicle trajectory between scenarios with no guide line and scenarios with one guide line (p-value = 0.940). However, there is a significant difference in vehicle trajectory between scenarios with no guide line or one guide line and scenarios with full guide lines (p-values are 0.0082 and 0.0065, respectively). This suggests that in large intersections, full guidelines have a noticeable impact on novice drivers’ left-turn behavior. This result aligns with the previous analysis of trajectory fluctuation ranges.

The Wilcoxon rank-sum test results for trajectory data in scenarios with dual left-turn lanes are presented in

Table 4. Wilcoxon rank-sum test results for trajectories in the dual left-turn lanes scenarios.

Scenario	M-Dual-N	M-Dual-O	L-Dual-O	L-Dual-F
M-Dual-O	p = 0.8501
M-Dual-F	p = 0.0082	p = 0.0041
L-Dual-O			p = 0.0005
L-Dual-F			p = 0.0005	p = 0.2899

. According to the test results, in medium intersections with dual left-turn lanes, there is no significant difference in vehicle trajectory between scenarios with no guide line and scenarios with one guide line at the 5% confidence level (p-value = 0.8501). This indicates that in medium intersections with two left-turn lanes, the influence of having one guide line on the vehicle trajectories of novice drivers is negligible. However, there is a significant difference in vehicle trajectory between scenarios with no guide line and scenarios with full guide lines at the 5% confidence level (p-value = 0.0082). This suggests that full guide lines have a noticeable impact on novice drivers’ driving behavior. Additionally, there is a significant difference in vehicle trajectory between scenarios with one guide line and scenarios with full guide lines at the 5% confidence level (p-value = 0.0041). This implies that having full guide lines has a more significant impact on novice drivers’ driving behavior compared to having one guide line. In large intersections with dual left-turn lanes, there is a significant difference in vehicle trajectory between scenarios with no guide line and scenarios with one guide line, as well as between scenarios with no guide line and scenarios with full guide lines (both p-values are 0.0005). This indicates that in large intersections with two left-turn lanes, guide line settings significantly impact novice drivers’ behavior. However, there is no significant difference in vehicle trajectory between scenarios with one guide line and scenarios with full guide lines at the 5% confidence level (p-value = 0.2899). This suggests that in these two scenarios, the impact of the number of guide lines can be ignored. Nevertheless, this result is inconsistent with the analysis of trajectory fluctuation ranges, and the possible reason for this inconsistency is that, in the case of two left-turn lanes, the presence of preceding vehicles and vehicles in adjacent lanes has a direct impact on the test vehicle, making the effects of having one guide line and having full guide lines similar.

The Wilcoxon rank-sum test results for trajectory data in scenarios with triple left-turn lanes are presented in

Table 5. Wilcoxon rank-sum test results for trajectories at the triple left-turn lanes scenarios.

Scenarios	L-Triple-N	L-Triple-O
L-Triple-O	p = 0.0065
L-Triple-F	p = 0.0413	p = 0.1510

. According to the test results, in large intersections with triple left-turn lanes, there is a significant difference in vehicle trajectory characteristics between scenarios with no guide lines and scenarios with one guide line at the 5% confidence level (p-value = 0.0065). This suggests that in large intersections with three left-turn lanes, having one guide line has an impact on the vehicle trajectories of novice drivers. There is also a significant difference in vehicle trajectory characteristics between scenarios with no guide lines and scenarios with full guide lines at the 5% confidence level (p-value = 0.0413). This indicates that full guide lines have an impact on the driving behavior of novice drivers. However, there is no significant difference in vehicle trajectory characteristics between scenarios with one guide line and scenarios with full guide lines at the 5% confidence level (p-value = 0.1510). This suggests that in this scenario, increasing the number of guide lines from one to two has a minimal impact on the driving behavior of novice drivers.

3.2. Impact Analysis Based on Steering Angle and Speed Difference

Vehicle steering angle and speed difference are two important indicators reflecting vehicle manipulation characteristics and driver behavior [28,29,30,31]. In this section, the influence of left-turn guide lines on novice drivers is analyzed based on these two indicators across different scenarios in small, medium, and large intersections.

3.2.1. Small Intersections

Steering angle. Figure 9 illustrates steering angle variation and range during left-turn maneuvers by novice drivers in different guide line scenarios at small intersections. The steering angles during left turns are generally similar across the three scenarios. In the straight section, the steering angle variation is smaller; in the circular section, the steering angle increases. The maximum rate of change in steering angle is 123.49%, 94.35%, and 74.59% for Scenarios 1-A, 1-B, and 1-C, respectively; Scenarios 1-A and 1-B occur while entering the circular section, while Scenario 1-C occurs while leaving, indicating a relatively stable left-turn process for novice drivers with full guide lines in small intersections. The range of steering angles during the left turn is comparable across the scenarios, with maximum ranges of 0.41 rad, 0.34 rad, and 0.30 rad, respectively. This suggests similar stability in the left-turn process across the scenarios, with relatively better stability observed with full guide lines.

Speed difference. Figure 10 illustrates speed difference distribution during left turns by novice drivers in different guide line scenarios at small intersections. Notably, the speed differences are relatively large when vehicles are entering or about to leave the straight section of the intersection, while the fluctuation in speed difference is smaller in the middle section. Both one guide line and full guide line scenarios show reduced fluctuations. The mean speed differences in the three scenarios are 1.523 km/h, 0.357 km/h, and 0.437 km/h, respectively, highlighting the impact of guide line settings on novice drivers’ velocity stability during left turns.

3.2.2. Medium Intersections

Steering angle. Figure 11 depicts steering angle variations during left turns by novice drivers in different guide line scenarios at medium intersections. In Scenario 2, steering angles are generally similar across different guide line scenarios: in the straight section, the steering angle variation is smaller; in the circular section, the steering angle increases. However, in Scenarios 2-B and 2-C, the fluctuation range of steering angles is relatively small, with maximum ranges of 0.26 and 0.25, respectively. In contrast, in Scenario 2-A, the maximum fluctuation range in the circular section is 0.43, indicating a better guiding effect on left-turn stability for novice drivers in Scenario 2-A. Comparatively, the fluctuation range of steering angles is significantly larger in Scenario 3, suggesting that the larger the intersection area and the more left-turn lanes, the greater the steering angle fluctuation during left turns.

Speed difference. Figure 12 depicts the speed difference distribution during left turns in Scenario 2 and Scenario 3. In Scenario 2 without guide lines, speed differences transition from negative to positive values. Conversely, when guide lines are present, the transition is from positive to negative, with smaller speed differences in the full guide line scenario compared to the single guide line scenario. This underscores the significant impact of guideline configuration on novice drivers, providing a reference during the circular section, resulting in higher speeds during the circular section and decreased speeds when leaving the intersection. Analyzing speed difference errors, both single and full guide line scenarios exhibit smaller fluctuation ranges compared to the scenario without guide lines, emphasizing the influential role of guide lines on driving behavior. In Scenario 3, under different guide line scenarios, the speed difference follows similar trends, gradually transitioning from positive to negative values. Considering the fluctuation range, all three scenarios show comparable ranges in the intersection mid-section, with maximum fluctuation values of 1.53, 1.56, and 2.02. This suggests that in medium intersections with dual left-turn lanes, the impact of guide line configuration on left-turning vehicle speed is minimal.

3.2.3. Large Intersections

Steering angle. Figure 13 displays steering angle variations for novice drivers in Scenarios 4, 5, and 6 with different guide line configurations. In Scenario 4, the steering angle gradually increases, reaching a constant value in the circular section, and then decreases smoothly as the vehicle exits. Full guide lines exhibit a lower mean steering angle and a smoother transition, followed by a single guide line and no guide line scenarios, highlighting a significant impact of guide line configuration. In Scenario 5, without guide lines, the mean steering angle is smaller in the circular section, ensuring a smooth and stable turning process. With a single guide line, the angle peaks when entering the circular section, gradually decreasing afterward, demonstrating stable steering. Full guide lines result in the maximum angle when leaving the circular section, rapidly decreasing thereafter, emphasizing the impact of guide line configuration on novice drivers. In Scenario 6, steering angle variations are similar to those in Scenario 5. In 6-C, the mean and variation are smaller, indicating better stability. With a single guide line, significant changes occur when entering the circular section, with a higher mean angle. Without guide lines, the changes are relatively gentle. Comparing Scenarios 4, 5, and 6, as left-turn lanes increase, guide lines, especially a single guide line, significantly impact steering angles. Full guide lines show consistent impact features, suggesting their superior guiding effect in larger intersections with more left-turn lanes.

Speed difference. Figure 14 illustrates the speed differences distribution during left turns for novice drivers in Scenarios 4, 5, and 6 with different guide line configurations. In Scenario 4, under various configurations, left-turn speed differences show an unstable trend, with mean differences of -1.05 km/h, 0.69 km/h, and 0.29 km/h. Notably, full guide lines in large intersections with a single left-turn lane have a more pronounced impact, resulting in smaller fluctuation ranges compared to a single guide line. In Scenario 5, left-turn speed differences exhibit a relatively stable trend, with mean differences of 0.72 km/h, 0.41 km/h, and 0.32 km/h, especially stabilizing with a single guide line. This highlights the pronounced impact of a single guide line on novice drivers’ left-turn speed in Scenario 5. In Scenario 6, without guide lines and with one guide line, speed differences show similar trends, with mean differences of 0.24 km/h and 0.25 km/h, respectively. However, full guide lines lead to an opposite trend, decreasing from negative to positive values, with a mean difference of −0.39 km/h. In Scenario 6, full guide lines significantly impact novice drivers’ left-turn speed, causing larger fluctuations.

3.3. Impact Analysis Based on Eye-Tracker Data

Eye-tracker data analysis revealed that guide line configuration influences the attention of novice drivers. In Scenario 6, without guide lines, gaze points are scattered, mainly concentrated in the upper-left area (Figure 15). Introducing one guide line maintains scattered gaze points, now predominantly in the middle-left area. Full guide lines concentrate gaze points in the lower-left area, specifically on the roadway surface between the guide lines, demonstrating a significant guiding effect on drivers’ attention.

Table 6. Statistical results of driver gaze time in different areas of interest.

Scenarios	AOI 1		AOI 2		AOI 3		AOI 4		AOI 5
	T (ms)	P (%)	T (ms)	P (%)	T (ms)	P (%)	T (ms)	P (%)	T (ms)	P (%)
1-A	1585.95	26.27	0.00	0.00	510.57	8.46	1776.51	29.44	2162.97	35.83
1-B	786.09	11.00	92.44	1.29	2449.67	34.27	3226.39	45.13	594.41	8.31
1-C	823.48	12.06	49.95	0.73	4936.74	72.28	574.17	8.41	445.66	6.53
2-A	1931.73	27.49	383.24	5.45	810.52	11.54	1711.41	24.36	2189.10	31.16
2-B	740.51	8.49	0.00	0.00	2034.41	23.33	5136.83	58.91	808.25	9.27
2-C	739.48	10.53	0.00	0.00	4892.81	69.67	1157.48	16.48	233.23	3.32
3-A	1468.11	32.08	459.93	10.05	515.71	11.27	1264.18	27.62	869.07	18.99
3-B	29.64	0.57	134.56	2.60	1448.97	28.02	3336.09	64.50	222.75	4.31
3-C	88.52	1.71	70.64	1.37	4583.97	88.60	283.17	5.47	147.71	2.85
4-A	3572.65	42.08	728.53	8.58	942.60	11.10	1109.28	13.06	2137.93	25.18
4-B	377.50	5.39	190.09	2.72	1610.65	23.01	4574.30	65.36	246.46	3.52
4-C	306.31	3.36	45.58	0.50	8228.72	90.38	372.03	4.09	152.36	1.67
5-A	3212.45	53.78	325.93	5.46	540.03	9.04	1043.39	17.47	851.20	14.25
5-B	223.42	4.81	36.20	0.78	966.65	20.81	3367.58	72.48	52.15	1.12
5-C	203.55	3.25	0.00	0.00	5731.42	91.53	288.28	4.60	38.76	0.62
6-A	2595.00	52.71	466.16	9.47	535.87	10.89	818.61	16.63	507.36	10.31
6-B	1782.31	29.76	215.72	3.60	699.98	11.69	3290.99	54.95	0.00	0.00
6-C	161.59	3.02	0.00	0.00	5002.58	93.65	162.21	3.04	15.62	0.29

presents the statistical results of driver gaze time in different areas of interest, denoted by T (in milliseconds) and P (as a percentage). The proportional distribution of driver gaze points across various scenarios is visually depicted in Figure 16. AOI 2 consistently receives the least attention in all scenarios, with some instances showing gaze points close to or at 0%, such as in Scenarios 1-A, 1-B, 1-C, 2-B, 2-C, 3-C, 4-C, 5-B, 5-C, and 6-C. This aligns with typical driving behavior, where drivers tend to focus more on the left side of the field of view during left turns at intersections.

Upon analyzing sub-scenes with various guide line settings across the six major categories, it is apparent that within each major category, the increase in the number of guide lines leads to a notable decrease in the proportion of driver attention points in AOI 1, AOI 2, and AOI 5. In Scenario 5, for instance, proportions in AOI 1 for no guide line (5-A), one guide line (5-B), and full guide lines (5-C) are 53.78%, 4.81%, and 3.25%, respectively. Corresponding proportions in AOI 2 are 5.46%, 0.78%, and 0.00%, and in AOI 5 are 14.25%, 1.12%, and 0.62%. Conversely, the proportion of driver gaze points in AOI 3 significantly increases, with proportions in Scenario 5 being 9.04%, 20.81%, and 91.53% for no guide line (5-A), one guide line (5-B), and full guide lines (5-C), respectively. AOI 3 emerges as an effective area guided by guide lines, underscoring their substantial impact on the areas commanding driver attention during left turns and providing clear guidance for left-turn driving behavior.

In comparing the six major scenarios, the absence of guide lines results in a relatively small and consistent proportion of driver gaze points in AOI 3, ranging from 8% in 1-A to 12% in 2-A. Introducing one guide line leads to significant variation, ranging from 34% in 1-B to 12% in 6-B. Full guide lines show a substantial increase, ranging from 72% in 1-C to 94% in 6-C, particularly in large intersections with three left-turn lanes (6-C), where the majority of gaze points fall within AOI 3, underscoring the impactful role of guide lines.

Examining the same guide line settings across different intersection sizes, without guide lines, the likelihood of driver gaze points falling in AOI 1 increases with intersection size. With one guide line, the probability is generally small but slightly larger in larger intersections. Full guide lines consistently keep the proportion of attention points in AOI 1 low, gradually decreasing with larger intersections, highlighting the clear inductive effect of guide lines on driver attention points.

Comparing Scenarios 4, 5, and 6, without guide lines, the proportion of attention points in AOI 1 increases with the number of left-turn lanes, but the increase is not significant, and the proportion remains relatively constant in AOI 3. With one guide line, the proportion in AOI 1 is almost the same for scenarios with one and two left-turn lanes, significantly increasing for three left-turn lanes. Conversely, the proportion in AOI 3 is almost the same for scenarios with one and two left-turn lanes, significantly decreasing for three left-turn lanes. Full guide lines maintain a nearly unchanged proportion in AOI 1 with an increase in the number of left-turn lanes and consistently small proportions, while the proportion in AOI 3 gradually increases and remains consistently large. This suggests that in large intersections with one or two left-turn lanes, guide lines have some inductive effect, but the effect is more significant in scenarios with three left-turn lanes.

4. Discussion

4.1. Discussion of Vehicle Trajectory Characteristic Results

In Scenario 1, analyzing novice drivers’ left-turn trajectories, we observe similar lateral fluctuation ranges across no guide line, one guide line, and full guide lines. This suggests minimal impact on novice drivers in a small intersection with a single left-turn lane. The straightforward route and simple layout of small intersections with one left-turn lane enable drivers to execute turns smoothly without explicit guide lines. Thus, the impact of guide line settings on left-turn behavior is relatively small.

The analysis of trajectory characteristics reveals a significantly larger lateral fluctuation range in Scenario 2-A compared to Scenarios 2-B and 2-C, indicating a notable impact of guide line settings on novice drivers. In Scenario 3, the absence of guide lines leads to inner left-turn lane vehicles occupying the outer lane, resulting in a larger lateral fluctuation range. The presence of guide lines, especially full guide lines, reduces this range significantly, highlighting the impactful role of guide line settings. Medium intersections, with their multiple lanes and turning options, pose complexities. Lack of guide lines may heighten confusion for novice drivers, while guide lines provide clear direction, aiding smooth navigation and promoting compliance with traffic rules. This is particularly crucial for inexperienced drivers facing unfamiliar traffic rules and intersection layouts.

In Scenarios 4, 5, and 6, trajectory analysis reveals significant insights. In Scenario 4, the lateral fluctuation range in 4-A and 4-B is notably larger than in 4-C, showcasing a substantial impact of full guide lines on novice drivers. Scenario 5 displays a decreasing lateral fluctuation range, indicating the impact of guide line settings on novice drivers. Moreover, this impact intensifies with more guide lines. In Scenario 6-A, the evident increase in lateral fluctuation range surpasses 6-B and 6-C, with 6-B being slightly larger than 6-C. This signifies a notable impact of guide line settings on novice drivers, increasing with the number of guide lines. Large intersections with larger radii and complex structures diminish the reference effect of roadside features during left turns, leading to noticeable lateral fluctuations for novice drivers. Guide lines offer precise guidance, reducing confusion and indicating specific lanes. Additionally, large intersections with more traffic and lane choices may cause misunderstandings without clear guide lines, leading to unnecessary lateral movements. Guide lines contribute to standardized driving behavior, providing crucial visual cues for correct and safe left turns, enhancing overall traffic flow stability and safety. The results align with the findings of the research work of [27,32].

4.2. Discussion of Steering Angle Characteristic Results

In small intersections, steering angle fluctuation ranges during left turns are similar, indicating comparable turn stability. However, the full guide line scenario (1-C) exhibits relatively better turn stability. In these intersections, with their small left-turn radii and simple layouts, novice drivers can easily observe the road and maintain stability even without explicit guide lines. The presence of guide lines encourages a smoother turning path, leading to relatively stable changes in steering angles. However, the impact is not significantly pronounced compared to scenarios without guide lines.

In Scenario 2, steering angles during left turns show general similarity across different guide line settings. However, with one guide line and full guide lines, the fluctuation range is relatively small, indicating that guide lines enhance left-turn stability. In Scenario 3, the fluctuation range of steering angles is notably larger, and the maximum fluctuation point varies with guide line settings, suggesting an influence on steering control during left turns. However, no apparent positive impact on turn smoothness is observed. In medium intersections with larger radii and higher traffic flows, guide lines provide clearer guidance, standardizing behavior and reducing steering angle fluctuation at mid-intersection. In scenarios with two left-turn lanes, lane changes during entry and turning phases may lead to larger fluctuations even with guide lines, especially for novice drivers with varying degrees of experience.

In Scenario 4, guide line settings exhibit distinct effects on mean steering angles. Full guide lines lead to lower mean angles and smoother transitions, while one guide line and no guide line scenarios show higher mean angles and less smooth transitions, indicating a significant impact on novice drivers in large intersections with a single left-turn lane. In Scenario 5, the mean steering angle is smaller in the mid-curve. In Scenarios 5-B and 5-C, there is significant steering angle variation. In Scenario 6, full guide lines result in the smallest fluctuation range of steering angles, indicating better turning stability. One guide line shows a slightly larger fluctuation, while no guide line exhibits a more gradual fluctuation. These findings suggest that in large intersections, guide lines provide clear guidance, indicating the correct turning path. Full guide lines may eliminate hesitation and confusion, leading to lower mean steering angles. Guide line settings can influence drivers’ behavior, with different settings affecting their decisions and driving manner before and after the turn. Full guide lines may guide vehicles to make left turns more standardized, reducing drastic changes in steering angles, providing clear instructions and boosting driver confidence, and resulting in smoother turns. The results show similar findings with the research work of [27,30,31].

4.3. Discussion of Speeding Characteristic Results

In small intersections, speed differences between scenarios with one guide line and full guide lines are minimal, signifying a significant impact of guide lines on the speed stability of vehicles during left turns. Guide lines provide clear path guidance, aiding accurate lane selection and turning maneuvers, minimizing hesitation, and promoting consistent speeds. Additionally, guide line settings encourage a systematic approach to lane selection and adherence to specific paths and speeds, which are particularly beneficial for novice drivers due to reducing speed fluctuations caused by unfamiliarity with intersection layouts.

In Scenario 2, both one guide line and full guide lines exhibit similar trends in speed differences. However, the speed difference with full guide lines is smaller than that with one guide line, signifying a significant impact of guide line settings on novice drivers. Guide lines, especially full guide lines, provide clear guide and path cues, facilitating drivers in adjusting speeds for turning. They enable more accurate prediction and control of speed changes during turns, with full guide lines offering superior guidance and resulting in smaller speed differences. In the absence of guide lines, drivers may experience confusion and uncertainty, leading to speed fluctuations around the turning point. Conversely, guide lines aid drivers in better planning turning points, reducing uncertainty and fluctuations in speed adjustments for smoother changes.

In Scenario 3, speed differences exhibit similar trends across all three guide line settings. This suggests a minor impact of guide line settings on left-turn speeds in medium intersections with two left-turn lanes. The complexity introduced by two left-turn lanes may diminish the influence of guide lines on speed differences, considering factors such as lane choices, driver behavior, and the inherently restrictive nature of guide lines. Mutual constraints between two left-turn traffic flows lead to few or no lane-changing vehicles, resulting in predetermined paths and minimal fluctuations in speed differences.

In Scenario 4, speed difference trends during left turns are unstable across different guide line settings, with full guide lines showing a relatively smaller fluctuation range. This implies a more pronounced impact of full guide lines on the left-turn speed of novice drivers, while the impact of one guide line is smaller in Scenario 4. Full guide lines may offer a more detailed and accurate path guide, facilitating drivers in understanding the correct turning path and adjusting speeds. In contrast, one guide line may provide less detailed guidance, leading to less accurate and stable speed adjustments during turns.

In Scenario 5, the trend in speed differences during left turns is more stable across different guide line settings. This indicates a more noticeable impact of one guide line on novice drivers in Scenario 5. Mutual influence between the two left-turn lanes limits flow fluctuations, resulting in stable speed differences. With one guide line, a clear guide is provided for two turning lanes, and cautious drivers maintain stable speeds.

In Scenario 6, speed difference analysis results suggest that full guide lines have a significant impact on novice drivers’ left-turn speeds, resulting in larger speed fluctuations in Scenario 6. Full guide lines offer a more detailed and comprehensive path guide, making it easier for drivers to understand the turning path. This reliance on guide lines for speed adjustments significantly impacts novice drivers’ left-turn speeds and leads to larger speed fluctuations. The results show similar findings to the research work of [27,30,31].

4.4. Discussion of Driver Gaze Characteristic Results

Analyzing eye-tracking data across the six scenarios reveals that as the number of guide lines increases, the proportion of driver fixations in AOI 1, AOI 2, and AOI 5 significantly decreases, while the proportion in AOI 3 significantly increases. AOI 3, representing the effective area guided by the guide lines, indicates that the presence of guide lines significantly influences the areas where drivers focus during left turns. This suggests a clear inductive effect of guide lines on left-turn driving behavior. Without guide lines, novice drivers tend to have dispersed attention, focusing more on distant areas. However, with guide lines, especially full guide lines, drivers concentrate their gaze on the road surface between the two guide lines, resulting in most fixations falling within AOI 3.

Comparing the six scenarios, in the absence of guide lines, the proportion of driver fixations in AOI 3 remains consistently small. Introducing one guide line shows noticeable variation in fixation proportions in AOI 3, and with full guide lines, there is a significant fluctuation as well. This emphasizes that as the intersection area increases, the inductive effect of guide lines on driver fixations in AOI 3 becomes more prominent. Particularly in Scenario 6, the majority of driver fixations concentrate within AOI 3, underscoring the substantial impact of guide lines. This result suggests that with a larger intersection, especially in the case of three left-turn lanes, the visual guidance and decision-making benchmarks provided by guide lines become more crucial. This heightened inductive effect likely directs drivers’ attention more within the effective area of the guide lines, reducing uncertainty, enhancing confidence, and improving judgment regarding the correct turning path.

Analyzing the eye-tracking data, it is evident that in large intersections with one or two left-turn lanes, guide lines exhibit a certain inductive effect, though it is not very significant. However, in scenarios with three left-turn lanes, the inductive effect of guide lines becomes pronounced. In large intersection scenarios, the complexity increases with the number of left-turn lanes. With one or two lanes, the inductive effect is less apparent due to fewer turning choices. However, with three left-turn lanes, the complexity rises, making guide lines crucial for providing direction. This intensified inductive effect likely prompts drivers to focus more on the guide line area, reducing uncertainty, enhancing judgment, and increasing confidence in the correct turning path.

In summary, the eye-tracking data analysis indicates that guide lines play a significant role in directing drivers’ attention during left turns. The inductive effect of guide lines becomes more pronounced as the intersection size increases, especially in scenarios with a larger number of left-turn lanes, where the complexity of traffic choices necessitates clearer guides for drivers. The results show the similar findings with the research work [22,27].

5. Conclusions

This study conducted an on-site survey at 21 intersections within the Lingang area of Shanghai, with a specific emphasis on analyzing five representative intersections. These intersections included small, medium (featuring two left-turn lanes), and large intersections (with two and three left-turn lanes). Six simulation scenarios were designed using a driving simulator. A series of driving simulator experiments were then conducted to collect and analyze vehicle trajectories, steering angles, driving speeds, and eye-tracking information. The primary objective of the study was to explore and evaluate the impact of left-turn guide line settings on the driving behavior of novice drivers.

Combining the analyses of vehicle trajectories, steering wheel angles, turning speed variations, and eye-tracking data, this study draws the following conclusions: in small intersections with a single left-turn lane, the impact of guide line settings on the driving behavior of novice drivers is relatively minimal. However, as the intersection size and the number of left-turn lanes increase, the guiding effect of guide lines become significantly more pronounced. In medium-sized intersections, guide lines positively influence stability and steering control, which is especially evident in scenarios with two left-turn lanes, showcasing a notable effect on the distribution of steering wheel angle changes. In large intersections, guide line settings significantly affect novice drivers’ behavior in single left-turn lane scenarios, and the impact becomes even more prominent with two or three left-turn lanes, particularly when employing full guide lines. Overall, the application of guide lines can effectively improve the stability of drivers. Furthermore, this study analyzes intersections of different sizes and driver’s visual attention in more detail, leading to the further conclusion that the presence and effectiveness of guide lines are positively correlated with the size of the intersection and the number of left-turn lanes in enhancing the stability of novice drivers’ behavior.

The results of this study provide a theoretical support and guide for the design and application of intersection guide lines. Nevertheless, there are some limitations in this study. For example, the experimental results need further validation with real-world vehicle turning trajectory data, which could be collected on-site using drones or GPS data. Larger-scale experiments involving a more diverse sample of drivers will be conducted to enhance the accuracy of the results. Additionally, the experimental scenarios were relatively simplified, and future research will involve more detailed designs of different types of intersections. In addition, more extensive comparative analysis of the indicators should be conducted in future research based on the existing literature to more accurately assess and reflect the reliability of the findings from this study. Furthermore, the study did not examine the interactions between the test vehicle and surrounding vehicles, which have to be considered in future studies.