1. Introduction
Autonomous driving has brought a lot of new opportunities for the automotive industry. The Society of Automotive Engineers (SAE) classifies autonomous vehicles into 6 levels (Level 0–Level 5) [
1]. Among them, the advanced driver-assistance system (ADAS) (Level 1 to Level 2) has specific functions, mainly including providing warnings or performing a limited set of lateral and/or longitudinal vehicle motion control actions to help drivers operate the vehicle safer [
1,
2]. The automated driving system (ADS) (Level 3 to Level 5) can continuously perform complete dynamic driving tasks (DDTs) and reduce and avoid collisions. Nowadays, many people believe that autonomous vehicle technologies (such as ADASs and ADSs) can greatly improve driver safety by eliminating collisions caused by human error, which 94% of road traffic accidents are attributed to [
3,
4]. However, if the driving behavior of an autonomous vehicle is significantly different from that of a conventional vehicle, not only will the driver’s trust and acceptance of autonomous vehicles be weakened [
5,
6], but it may also lead to unpredictable risks in traffic accidents [
7,
8]. At the same time, personalized assisted driving can also improve the safety and acceptance of autonomous driving. Therefore, an in-depth understanding of the driving behaviors of conventional vehicles is of great significance for the design of autonomous vehicles and the improvement in the safety and acceptance of autonomous vehicles.
Due to the complexity of the intersection scene, intersection accidents are one of the most common types of urban vehicle accidents for both conventional vehicles and autonomous vehicles [
3,
7,
8,
9]. Intersection areas are shared by drivers from multiple directions, and the diversity of vehicle movements leads to multiple conflict points [
10,
11]. The convergence of traffic flow and various driving tasks forces drivers to rapidly assimilate information from multiple angles, which can significantly increase the complexity of the situation [
12]. Therefore, intersections involve a series of complex information acquisition, reaction judgments, and operational processes. Understanding the driving behaviors of surrounding conventional vehicles and making a human-like driving model based on them is a necessary condition for autonomous vehicle to successfully pass the intersection.
The geometric features and functional design of an intersection significantly affects intersection maneuvers [
13], and successful intersection maneuvering requires the ability to perceive all useful information resources and make correct and safe decisions in this dynamic driving environment, whereas the individual differences in drivers similarly affect their operational decisions. Indeed, Dukic and Broberg [
14] recorded drivers’ visual behaviors at a signalized intersection in real-world traffic and found that younger drivers focus more on dynamic objects, such as other road users, while older drivers pay more attention to static objects, such as road markings. Wu and Xu [
15] analyzed the driving behaviors of right-turning drivers at signalized intersections using naturalistic driving research data. They found that drivers exhibit a high acceleration and low observation frequency during right-turn-on-red (RTOR) maneuvers, which can pose potential dangers to other road users. Pathivada and Perumal [
16] examined the factors influencing driver behaviors in the dilemma zone at signalized approaches under mixed traffic conditions and found that the approach speed and type of vehicle significantly impacts the probability of stopping. More research has focused on drivers’ abilities to perceive information at unsignalized intersections than at signalized intersections. Li et al. [
17] conducted a comparative analysis of the differences in driving behaviors between older and younger drivers in unsignalized intersection conflict scenarios and established driving behavior graphs, which found that older drivers were slower in observing and collecting traffic information based on the changes in behavioral nodes. Additionally, research by Yamani et al. [
18] indicated that older drivers performed worse in visual search tasks compared to middle-aged drivers and required more steering adjustments. Romoser et al. [
19] suggested that when approaching and entering unsignalized intersections, the reason why older drivers are unable to scan for hazards is because some of the difficulties experienced by them in scanning intersections stem from specific attentional deficits that hinder their ability to suppress the primary goal of monitoring the expected path of other vehicles. Werneke and Vollrath [
9] conducted a study on the psychological processes of driving behaviors and attention allocation of drivers at unsignalized T-intersections with varying environmental characteristics. The results revealed that drivers’ attention allocation and driving behaviors systematically depended on the environmental features of the intersection. Lemonnier et al. [
20] investigated the allocation of explicit visual attention at unsignalized intersections with different priority rules under multitasking and dynamic situations. The results showed that visual attention to the intersecting roads varied according to the priority rules and visual attention was influenced by vehicle control subtasks. Subsequently, a field study was conducted to examine the visual attention of intersection drivers [
21]. The study found that the effects of priority rules, the expected traffic density, and familiarity were reliable factors in understanding drivers’ gaze allocations on the road.
Although previous research has analyzed driver behaviors at intersections, we found only two studies that investigated the differences in driver behaviors between signalized and unsignalized intersections. Specifically, Liu and Ozguner [
22] used simulation to analyze driver decision-making and operational responses. Li et al. [
23] examined the differences in visual scanning behaviors between signalized and unsignalized intersections using naturalistic driving data. However, these studies only focused on a single aspect of driver behavior and did not provide a comprehensive understanding of the differences, as driving tasks involve perception, cognition, and operation. The type of intersection can influence driver behavior, as drivers adopt different driving strategies at signalized and unsignalized intersections, leading to variations in safety levels. Existing research is not yet sufficient to fully elucidate the performance of drivers under different traffic conditions, and there are even fewer studies in China. Considering potential cultural and regional differences, there may also be differences in the conclusions regarding driving behaviors among different countries. To explore the underlying mechanisms behind this situation, there is a pressing need to determine the existing differences in driver behaviors. Therefore, to fully understand driver behaviors at intersections, it is necessary to study the impacts of external and internal factors on driver behaviors.
Thoroughly studying the driving behaviors at intersections can improve the performance of ADS driving, improve the interaction between ADSs and conventional vehicles, and enhance the acceptance and safety of ADSs. Therefore, the study aims to examine the impact patterns of intersection types on driver behaviors by driving simulation experiments. And the contributions of this article can be summarized as follow:
- (1)
The type of intersection affects a driver’s behavioral performance, but there are fewer studies on the differences in driving behaviors between signalized and unsignalized intersections, and the research in this paper fills the gap that exists in this area.
- (2)
In this paper, the driver’s behavior when crossing an intersection is studied in stages, and the stages of the driving behavior pattern can reflect the different characteristics of the driving process, and this pattern is the inherent behavioral pattern of the driver. Therefore, it can help researchers to deeply understand the driving behaviors at intersections and provide theoretical behavior support for the research and design of safety systems afterwards.
- (3)
Individual differences in driving behaviors based on intersection characteristics were assessed, i.e., the influence of personal characteristics such as gender and age on drivers’ physiological, psychological and operational performances. By addressing the interactive effects of personal characteristics as well as intersection characteristics on a driver’s intersection performance, the preferred behavioral patterns of different drivers are investigated. This case study of driving behavior patterns can inform personalized driving behavior modeling as well as vehicle system safety designs.
4. Discussion
The study results showed that, compared to the physiological and psychological characteristics, the differences in the driving performances were more pronounced at intersections. From the changes observed throughout the entire process, depicted in
Figure 2 and
Figure 3, it is evident that speeds steadily decreased at unsignalized intersections, while the deceleration effect was more noticeable at signal-controlled intersections. This indicates that drivers tended to maintain a more consistent speed when crossing unsignalized intersections, without coming to a complete stop. This behavior may be attributed to the prevalent culture of risky driving practices in Chinese traffic conditions [
30]. Furthermore, the results of the acceleration indicators also revealed that under unsignalized conditions, the acceleration during the right-turn task while entering the intersection was significantly higher compared to signal-controlled intersections. Conversely, the acceleration during the left-turn task while exiting the intersection was significantly lower at unsignalized intersections. These findings are related to the driving rules in China, which require drivers to drive on the right side of the road, resulting in a higher demand for left-turn maneuvers at unsignalized intersections compared to right-turn and straight-through tasks, particularly during the exit stage. Drivers making a left turn must pay more attention to opposing traffic and the surrounding traffic environment to safely navigate the intersection without conflicts with other vehicles [
9,
31]. Therefore, when drivers make a left turn at unsignalized intersections, they tend not to manipulate their speed significantly for safety. However, at signal-controlled intersections, where traffic lights organize orderly vehicle movements during the green phase, it is a subconscious reaction of most drivers to slow down and pass through instinctively. The variations in the physiological and psychological characteristics also demonstrated significant differences in the heart rates between the entering and exiting stages (
Figure 4). The fluctuations and noticeable peak in the heart rate values during the exiting stage indicated that drivers were more tense, experienced higher mental workload, and were more prone to be involved in traffic accidents during this stage. Meanwhile, the results regarding the gaze duration indicated that drivers tended to have slightly longer gaze durations at signal-controlled intersections compared to unsignalized intersections, although the difference was not significant (
Figure 5). TAY et al. [
32] suggested that the higher collision risk at signal-controlled intersections could be attributed to differences in scanning behaviors between signalized and unsignalized intersections.
In addition, the driving tasks also had a significant impact on the behaviors of drivers at intersections. The left-turn task exhibited the most noticeable decrease in speed while crossing the intersection, and during both the exiting and entering stages, the acceleration during the straight-through task was significantly lower than that of the left-turn and right-turn tasks. This can be attributed to the higher operational demands and workloads associated with turning tasks, especially during left turns. For safety reasons, drivers need to exhibit more noticeable speed changes during turning tasks compared to tasks with lower requirements, such as going straight. Regarding the physiological and psychological characteristics, the heart rate showed a declining trend throughout the entire stage of crossing the intersection for the right-turn task, and the gaze duration was also the lowest. This indicates that drivers experienced the lowest driving workload when performing the right-turn task.
Furthermore, the driver characteristics were also considered as influential factors in this study. And the results indicated that the age, gender, and driving experience had significant influences on intersection driving behaviors, albeit with some differences across different scenarios. Overall, middle-aged drivers were relatively the safest group. And for elderly drivers, the impact of age on driving was negative. The speed decreased with age, and elderly drivers had higher heart rates at signal-controlled intersections, as well as longer gaze durations during left turns at unsignalized intersections and the highest heart rates overall. This suggests that elderly drivers experienced higher levels of anxiety while crossing intersections, particularly during tasks that required higher demands. To ensure safe crossing, compensatory strategies such as reducing the speed, maintaining larger following distances, avoiding specific traffic conditions, or shortening driving distances are often adopted [
33,
34]. On the other hand, for young drivers, at signal-controlled intersections, they exhibited lower speeds during left-turn tasks and higher heart rates during straight-through tasks. This may be attributed to a lack of confidence resulting from their limited experience, making young drivers more nervous and inclined to maintain lower speeds. Feng et al. [
35] demonstrated that as information and task demands increased, the differences between different age groups became more apparent. Regarding driving experience, it was found that driving behaviors also followed certain patterns among different levels of driving experience. Drivers with little experience showed a “nervous” type of behavior, those with a moderate level of experience exhibited an “impulsive” type, and those with a high level experience showed a “steady” type, with the latter being the safest group. This is because drivers with little experience had higher heart rates, longer gaze durations during left turns at unsignalized intersections, and larger acceleration standard deviations at signalized intersections. This indicates that drivers with little experience experienced higher levels of anxiety and workload while crossing intersections, particularly during tasks with higher demands, and exhibited less stable driving behaviors. Drivers with a moderate level of experience, due to their self-perceived rich driving experience, became less cautious when performing critical driving tasks, such as lane changes or turning. They exhibited higher speeds during straight-through tasks at unsignalized intersections and larger acceleration standard deviations during left-turn tasks, potentially showing unstable and unsafe driving tendencies. Drivers with a high level of driving experience not only possessed proficient vehicle control but were also capable of handling various situations with ease. They exhibited a more relaxed and calm driving behavior throughout the driving process, as indicated by their lowest gaze durations during left turns at unsignalized intersections. Nabatilan et al. [
36] also found that experienced drivers had lower error rates compared to inexperienced drivers. In terms of gender, male drivers exhibited lower speeds during left turns at unsignalized intersections and lower heart rates during straight-through tasks. This suggests that male drivers, compared to female drivers, demonstrated more confidence and stability while driving. This finding is consistent with reports indicating a higher involvement of women in crashes involving injury and higher rates of all reported accidents by the police [
37].
5. Conclusions
This study not only examined the differences in driver behaviors between signalized and unsignalized intersections but also investigated their correlations with driver factors, providing a comprehensive understanding of the characteristics of driver behaviors at intersections. We found that both the intersection type and driver factors have an impact on a driver’s behavior. These findings suggest that future developments in advanced driver assistance systems and intersection design should take into account the behavioral characteristics and needs of different drivers at different intersections, in order to provide better services for drivers.
Traffic environments at intersections are diverse and complex, and this study only introduces the signal control and driving task variables, with a limited sample size. Further research should include more factors (such as road geometry, car following behavior, and the presence of other traffic participants) as well as more critical situations (such as dilemma zones and yellow lights). This will help to increase the understanding of driving behaviors at intersections. Considering the design, evaluation, and implementation of future autonomous vehicles, it is also necessary to model and predict the driving behaviors at intersections.