Review of Driver Behavior Characteristics and Hazard Perception Enhancement in Low-Illumination Environments
Abstract
Low-illumination environments significantly affect drivers’ behavior characteristics and hazard perception, making it one of the key factors contributing to the high frequency of traffic accidents in such conditions. Although nighttime driving accounts for a relatively small proportion of total driving mileage, the associated accident risk is significantly higher than during the day. This paper analyze the mechanisms of changes in drivers’ visual behavior, physiological and psychological states, and speed control characteristics under low-illumination conditions and investigated the impact of individual factors (such as age, gender, and driving experience) and external environments on hazard perception. It summarized various existing training methods aimed at improving hazard perception and further discusses the quantification models for hazard perception. Finally, it summarized the limitations of the current research and explores future research directions, with the aim of providing theoretical and technical support for reducing traffic accident risks in low-illumination environments.
1. Introduction
Based on research by the International Commission on Illumination (CIE), vehicles traveling at night account for only 25% of total mileage and nearly half of all traffic fatalities occur during this time. Furthermore, these nighttime accidents often result in more severe consequences compared with those during the day [1]. Studies from various countries have revealed that the risk of accidents at night is 1.5–2 times higher than during daylight hours [2], and the incidence rate of fatal accidents, defined as the number of accidents per mile traveled, is 3–4 times higher at night than during the day [3, 4].
When driving at night in low-illumination environments, drivers face numerous influencing factors. These include their existing knowledge of driving behavior, habitual road conditions, observation methods, attention distribution, and driving experience. These factors can delay reaction times, cause judgment errors, and make it harder to perceive road information and hazards. Moreover, since drivers rely primarily on vision for gathering information at night, reduced visibility and visual illusions can lead to delays, neglect, or errors in perceiving road information and hazards. These issues can alter driver behavior, thereby affecting hazard perception and leading to incorrect driving action. In emergencies, drivers may fail to respond appropriately, ultimately increasing the likelihood of traffic accidents.
Road traffic safety is influenced by four key factors: human, vehicle, road, and environment, with the human factor being the most significant contributor to accidents. Thus, understanding driver behavior and hazard perception is essential. Although many experts have studied the impact of environmental changes on driver behavior, research on driving behavior in low-illumination environments at night remains significantly limited. In particular, the mechanisms behind changes in driver behavior and the effects of low illumination on hazard perception require further investigation. Furthermore, limited research on improving hazard perception in low-illumination nighttime environments hinders the development of effective strategies to reduce accident risk in such conditions.
This paper reviews research on nighttime driving in low-illumination environments, focusing on driving behavior, hazard perception and its influencing factors, strategies to improve hazard perception, and methods to quantify hazard perception under such conditions. Figure 1 illustrates the main content of current study.
2. Driver Behavior in Low-Illumination Environments
Drivers must continuously adjust their behavior in response to changes in the road environment to drive safely. As a result, they often exhibit different behavior characteristics in varying environments. Safe driving largely depends on drivers’ situational awareness and their ability to gather information from the surrounding environment, with vision being the primary means of perception. In low-illumination environments, drivers’ visual behavior is the first to be affected. This, in turn, affects their physiological and psychological states, which then affect their reactions and operational behaviors. These changes manifest in speed control, car-following, and lane-changing behaviors. Extensive research has been conducted in these areas.
2.1. Visual Behavior in Low-Illumination Environments
Research indicated that over 80% of the information drivers gather while driving is visual [5]. Plainis et al. [6] investigated how retinal adaptation influences visual perception under low-illumination driving conditions and the findings highlighted that slower road pathways under dim lighting may hinder the ability to respond to fast-changing visual conditions. Wood [7] found that visual function is reduced under the mesopic lighting conditions at night, with these effects worsening with age and visual impairment. Easa et al. [8] analyzed the impact of ambient light intensity on drivers at night, finding that on straight road sections, higher light intensity improved drivers’ visual recognition ability, enhancing road sign identification; however, on curved sections, excessive light intensity can negatively affect sign recognition, reducing visual capability. Arumi et al. [9] suggested that changes in the nervous system can reduce night vision in drivers. Konstantopoulos et al. [10] found that in poor visibility conditions, especially during rain, drivers’ visual search efficiency decreased. Goodman et al. [11] conducted controlled experiments to study the effects of different lighting conditions on drivers’ visual performance. They proposed an intermediate visual benefit optimization model and verified its effectiveness in improving nighttime driving safety. Jones et al. [12] explored how contrast sensitivity in elder adult drivers affects their ability to detect obstacles at night. Their findings indicated a close relationship between low-contrast visual sensitivity and the ability to detect hazards, which is especially significant for elder drivers.
2.2. Physiological and Psychological Behavior in Low-Illumination Environments
Drivers’ physiological and psychological states vary across different driving environments, affecting driving safety. Hill et al. [13] explored the relationship between psychological stress and environmental factors, finding that weather conditions, visibility, and task complexity all influence stress levels. Chen et al. [14] studied the effects of color temperatures and illuminance levels on driver alertness and reaction time, finding that medium color temperatures improved alertness and reduce reaction time, while lower illuminance worsened these factors. Wang et al. [15] conducted simulated driving experiments to compare drivers’ physiological characteristics during the day and night, discovering significant differences in heart rate variability, brain load levels, and breathing frequency during nighttime driving.
2.3. Speed Control Characteristics in Low-Illumination Environments
In low-illumination driving environments, drivers often have difficulty accurately estimating their own speed, which can result in unintentional speeding [16]. Moreover, drivers often fail to reduce speed to compensate for limited visibility [17]. Baker [18] recommended imposing speed limits at night, as drivers tend to speed and misjudge their speed in low illumination. Haglund [19] found similar results, with 47%–58% of drivers speeding at night, often perceiving their speed to be lower than it actually is. Liu et al. [20] analyzed questionnaire data and found that factors such as gender, driving experience, weekly nighttime driving frequency, and annual mileage significantly both influence speed choice in low illumination. Plainis et al. [21] examined drivers’ reaction times and speeds at night, establishing a relationship between reaction distance and illumination, and confirmed the impact of illumination and spatial frequency on reaction times. Nygardhs et al. [22] studied the effect of different road sign shapes on drivers’ nighttime speeds. Gilandeh et al. [23] examined the correlation between illumination conditions and bus drivers’ behavior, finding that bus speeds under normal illumination were consistently higher than under low illumination though lateral positioning was unaffected by it. Gao et al. [24] explored the impact of adaptive headlight systems on vehicle dynamics at road curves, testing drivers’ preview distances at different speeds and curve radii to predict subsequent speed and acceleration, and proposed an algorithm for simulating driver behavior at curves. Owens et al. [25] studied the effects of low illumination and age on driver behavior through comparative experiments during day and night, finding that nearly all drivers reduced speed under low illumination, with a decreased ability to recognize objects. Bassani et al. [26] used environmental parameters related to road geometry and lighting conditions to develop a speed distribution model, demonstrating that average speed and position deviation are significantly influenced by lighting parameter changes under various conditions such as sunny, cloudy, and dark days.
3. Driver Hazard Perception and Its Influencing Factors
Hazard perception is closely linked to the occurrence of traffic accidents, with subjective hazard perception during driving playing a crucial role in driving safety [27]. A driver’s hazard perception level dynamically changes within the driver–vehicle–road–environment system interaction. When perceived hazard exceeds an acceptable range, the driver will adjust their speed accordingly to manage the perceived hazard. In essence, a driver’s hazard perception significantly influences their driving behavior. Drivers with higher hazard perception skills are adept at forming effective mental models of the road environment and evaluating predictive cues, facilitating early hazard detection.
Numerous studies have demonstrated that hazard perception is related to personal factors such as age, gender, and driving experience and is closely associated with traffic accident occurrence [28, 29]. Kouabenan [30] found that hazard perception positively impacts protective behavior. Farrand [31] observed that drivers with lower hazard perception are more likely to engage in risky driving behaviors. Hunter [32] explained that such drivers tend to assess road risk levels as lower, making them more prone to high-risk driving behaviors.
Typically, a driver’s hazard perception increases with age before eventually declining [33]. Rundmo et al. [34] suggested that young drivers are slower and less efficient at detecting hazards, often underestimating risks in traffic situations. They have weaker hazard perception and tend to focus on the risk itself without a comprehensive understanding, failing to take timely measures to mitigate risk. For older adults, Bunce and Anna [35] found that cognitive processing ability and physical function decline with age, slowing information processing speed. In addition, Garay et al. [36] have found that under both day and night conditions, novice drivers are less likely than experienced drivers to scan for potential hazards, and all drivers are less likely to scan for potential hazards at night.
Regarding gender, female drivers generally exhibit higher sensitivity to risks in traffic situations compared with male drivers. Studies have shown that male drivers often underestimate potential traffic risks and engage in aggressive driving behaviors more frequently. For example, males are more prone to overtake and pass through a light in amber [37]. Female drivers, being more cautious, commit fewer violations than male drivers but may make more operational errors during driving [38].
Cohn et al. [39] assert that hazard perception involves cognitive and judgment abilities and is negatively correlated with risky driving behavior. Uleberg et al. [40] found that personality indirectly influences risky driving behavior by affecting behavioral attitude determinants. Drivers with anxious personalities tend to have higher hazard perception levels and exhibit more cautious driving behaviors, while those with aggressive personalities are more likely to engage in risky driving behaviors. Du et al. [41] discovered that drivers with a higher sensation-seeking tendency have poorer hazard perception in situations where vehicle signals serve as cues, particularly for hazards with cues, and their visual agility is worse than drivers with lower sensory-seeking tendencies.
Furthermore, studies have shown that drivers’ hazard perception is closely related to the driving environment [42]. Zhang et al. [43] found that increasing altitude weakens drivers’ visual sensitivity and cognitive processing ability for dangerous information, reducing their hazard perception. Crundall et al. [44] tested the visual recognition performance of environmental and behavioral hazards using traffic situation videos and driving simulators, finding that different hazard categories significantly impact drivers’ hazard recognition efficiency.
4. Methods to Enhance Driver Hazard Perception
- 1.
Hazard prediction training through videos or images: Pradhan et al. [46] conducted PC-based hazard prediction training for young drivers, analyzing their visual scanning behavior while driving on local residential, feeder, and arterial roads. They found that trained drivers paid more attention to and scanned hazardous road areas compared with untrained drivers. Zhang et al. [47] initiated a video-based training program to explore its effectiveness in simultaneously improving drivers’ hazard anticipation and attention maintenance abilities. The results indicated that the training program effectively reduced the likelihood of collisions among drivers with varying levels of experience.
- 2.
Perception training through active reflection, passive feedback, or a combination of both: Borowsky et al. [48] evaluated the training effects of the aforementioned three different modes 1 week after training. They discovered that both the mixed group and the guided group were able to identify more potential hazards involving pedestrians in residential areas. Horswill et al. [49] developed a six-session online hazard perception training course that teaches drivers to avoid accidents using a wealth of real accident videos and a series of evidence-based strategies. Randomized controlled trials showed that this course significantly improved drivers’ hazard perception response times and accuracy in hazard prediction, as well as notably increased following distances. All trained drivers reported improvements in their real-world driving behavior.
- 3.
Systematic training based on behavioral theory: Zhou et al. [50] designed a hazard perception enhancement method for drivers in nighttime environments based on the “Knowledge–Attitude–Practice” theory. This method intervened in drivers’ risk knowledge and attitudes to alter their hazard perception, and its effectiveness was validated through comparative analysis of driving data in simulated environments. Zhang et al. [51] compared the effectiveness of expert commentary and experiential learning methods, finding that the experiential learning method based on Bandura’s learning theory yielded better results for novice drivers in hazard identification and speed control.
In addition, many specific software for hazard perception training had been developed, such as the Driving Assessment and Training System (DATS) developed in the United States, which integrates driver hazard perception evaluation, education, training, and retraining [52]. Another example is the Driver-ZED training software [53], which utilizes driving simulators to cover various levels of hazardous scenarios, aimed at training novice drivers to recognize and avoid dangerous situations.
5. Quantification of Driver Hazard Perception
Quantification of hazard perception enables deeper analysis of its variations and provides validation tools for enhancement strategies [54, 55]. Hazard perception quantification refers to the use of objective measurement tools and techniques to quantitatively assess a driver’s ability to detect, understand, and respond appropriately to potential hazards in the traffic environment. Quantification of hazard perception during driving can generally be divided into two categories: the first involves single surrogate indicator-based quantification schemes and the second category involves quantification models based on multiple indicators or even multidimensional indicators.
Single indicator quantification methods typically select indicators from aspects such as the driver’s physiology, psychology, and eye movement behavior. Perello-March et al. [56] measured drivers’ hazard perception with cardiac and skin conductance indicators in a series of high fidelity, simulated highly automated driving scenarios, and found that heart rate variability features are superior at capturing arousal variations in long term, low to moderate risk scenarios. Crundall et al. [57] used fixation variance to quantify hazard perception, finding that experienced drivers have a larger visual search range. They also used mean fixation time for hazard perception quantification and found that experienced drivers spent longer on average focusing their gaze compared to novice drivers [44]. However, Borowsky et al. [48] also used mean fixation time for quantification, and their findings contrasted with previous studies, showing that experienced drivers had shorter mean fixation times. Although such single surrogate indicator-based methods are often easier to implement and typically do not interfere with natural driving processes, a key limitation is their inability to capture the multifactorial nature of hazard perception dynamics.
Multi-indicator approaches often incorporate mathematical or machine learning frameworks. Li et al. [58] constructed a crash probability estimation model based on the driver’s braking behavior and time-to-collision (TTC) in near-collision scenarios, applying it to evaluate the effectiveness of forward collision avoidance technologies. Xie et al. [59] presents a multisource data-driven self-attention-based bidirectional long short-term memory (Att-Bi-LSTM) model for predicting drivers’ hazard perception, achieving superior performance, thus supporting advancements in driver assistance systems. Gonçalves et al. [60] proposed a conceptual model and methodology for objectively estimating drivers’ hazard perception, contributing to DSM development by allowing individualized readiness thresholds based on experimental data without subjective assessments. Feng et al. [61] developed a quantitative model for assessing hazard perception in human-machine codriving, demonstrating that physiological and eye movement metrics correlate with increased hazard perception, thus supporting safer human–machine interaction in driving. Rui et al. [62] identified stressful hazard perception scenarios in intelligent driving systems using subjective and objective measures, highlighting key factors such as EEG connectivity and EOG data, and provided insights for improving system safety and accuracy. Avetisyan et al. [63] developed a predictive model using multimodal data to assess driver situation awareness in conditionally automated vehicles, achieving promising performance and contributing to safer driver-AV interactions. Song et al. [64] classified drivers’ driving risk levels using a random forest classifier based on self-reported sensation seeking, hazard perception, and risky driving behavior.
6. Conclusion
The findings of this review on drivers’ driving behaviors and hazard perception in low-illumination conditions have important practical implications for road safety, vehicle design, and driver training. A deeper understanding of how drivers detect and respond to potential hazards in nighttime environments can inform the development of more effective lighting systems. In addition, insights from these studies can contribute to the enhancement of driver education programs by incorporating training modules that simulate low-illumination scenarios, thereby improving drivers’ hazard anticipation skills. Furthermore, the results can aid policymakers in refining traffic regulations and infrastructure planning to mitigate nighttime driving risks.
- 1.
While existing research has explored the effects of low illumination on drivers’ visual, physiological, psychological, and behavioral responses, the underlying mechanisms driving changes in driving characteristics remain unclear. In addition, the integration of these effects across different driving aspects is incomplete.
- 2.
Most studies assess hazard perception based on external behaviors and subjective reports, with limited efforts to quantify this ability through a comprehensive analysis of objective driving behaviors.
- 3.
Although various training methods have been implemented to improve hazard perception, the internal mechanisms and the extent of their effectiveness remain insufficiently examined. Moreover, the temporary significant changes in hazard perception after training are likely to be false positives. Therefore, it is necessary to assess the sustainability of the training effects.
- 4.
Research on hazard perception quantification and enhancement primarily focuses on well-lit environments, with limited studies addressing these aspects in nighttime low-illumination conditions, despite their high accident risk.
Future research on drivers’ hazard perception in low-illumination environments should expand in several critical directions. First, interdisciplinary approaches combining neuroscience, psychology, and artificial intelligence (AI) could provide deeper insights into how drivers process and respond to hazards under reduced visibility. Second, advancements in real-time monitoring technologies, such as eye tracking, physiological sensors, and vehicle telemetry, should be leveraged to develop more precise and dynamic models of hazard perception. Third, personalized training programs using virtual reality (VR) and augmented reality (AR) could offer immersive, adaptive learning experiences tailored to individual drivers’ needs. In addition, the integration of intelligent driver-assistance systems, such as adaptive lighting and AI-powered hazard detection, could enhance drivers’ awareness and reaction times in low-illumination conditions. Finally, large-scale field studies should be conducted to validate laboratory findings and ensure practical applicability in real-world driving scenarios.
Conflicts of Interest
The authors declare no conflicts of interest.
Funding
This work was supported by the National Natural Science Foundation of China (no. 52402415), the Open Project of Laboratory of Traffic Information and Safety in Higher Education Institutes of Anhui Province (no. JTX202302), and the Project of Talent Introduction and Doctoral Startup Fund of Anhui Jianzhu University (no. 2023QDZ36).
Open Research
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
