Abstract
This study used a speed–space analysis to analyze the distribution of red-light runners at intersection approaches during the all-red interval to assess their risk levels. Additionally, various speed–space diagrams were utilized to assess the performance of the Dynamic All-Red Extension (DARE) system and establish system parameters for field implementation. Two signalized intersections in rural Alabama with speeds of over 55 mph were selected for system implementation and data collection. Analysis results revealed that a significant number of red-light runners at these intersections were not protected by typical all-red intervals calculated based only on the Institute of Transportation Engineers (ITE) red-clearance equation. These red-light runners were detected up to 340 ft upstream of the stop line when the signal turned red. Their time to red-light violations extended up to 3.6 s from the onset of the red indication. The results of the speed–space analysis, combined with the distribution of red-light runners’ times to red-light violation, indicated that enhanced protection for these vehicles could be achieved by incorporating kth-percentile values of red-light runners’ times to red-light violation alongside the red clearance intervals based on the ITE equation. This concept involves safeguarding a certain portion of red-light runners through the red-clearance interval, while the remainder would be protected by DARE. Speed–space analysis also proved to be helpful in determining key parameters for DARE implementation. Overall, the DARE system demonstrated commendable performance in protecting high-risk red-light runners with minimal disruption to traffic operations.
Keywords
Red-light running at high-speed signalized intersection approaches is a severe safety concern that can lead to serious injuries and fatal crashes. It accounts for a staggering 165,000 injury crashes annually and has resulted in the loss of nearly 9,000 lives in the last decade alone ( 1 ). Approximately 16%–20% of intersection crashes are directly attributed to red-light running ( 2 ). This hazardous driving behavior can be intentional or unintentional. Intentional red-light running is typically characterized by aggressive and impatient behavior, while unintentional red-light running is associated with distractions, signal visibility issues, or insufficient yellow interval durations ( 3 ). Understanding the characteristics of red-light running vehicles, particularly at high-speed intersections where crashes can result in severe harm or loss of life, is crucial for the development of effective traffic engineering countermeasures and policies.
The present study utilizes a speed–space diagram to enhance the understanding of red-light runners’ characteristics, as described in Figure 1. The figure shows the speed and location of approaching vehicles at the onset of the all-red indication as well as their subsequent decision—either to proceed or to stop—after the red signal. These data were collected at one of the intersections selected in the study, specifically, US-43 approaches at intersection CR-96 in Alabama. This intersection is in a rural area, and its major road approach (US-43) has a speed limit of 55 mph. The intersection width of the major road approach is 78 ft, and 2.0 s of the minimum all-red period (ARmin) was set for that approach. According to the Alabama Traffic Signal Design Guide & Timing Manual, a minimum all-red clearance time of 1.2 s is recommended for this approach ( 4 ). This recommendation is based on the traditional kinematic equation provided by the Institute of Transportation Engineers (ITE) (see Equation 1) ( 5 ). However, since the intersection is situated on a high-speed road with a speed limit of 55 mph, the state highway agency extended the all-red interval by an additional 0.8 s for safety purposes. As a result, vehicles that run the red light within the first 0.8 s after the all-red indication can be considered to be protected by the ARmin of 2.0 s, as they would clear the intersection before the conflicting phases turn green. The speed–space diagram depicted in Figure 1 offers readers insight into the characteristics of red-light-running vehicles at high-speed intersection approaches, given the all-red interval of 2.0 s calculated based on the ITE equation (1.2 s) and engineering judgment (0.8 s). Further details, including the intersections selected for the present study as well as engineering countermeasures employed to mitigate red-light running, will be discussed in the forthcoming relevant sections.
Speed–space diagram of red-light runners at the onset of the red indication for the US-43 northbound approach to intersection CR-96.
The speed–space diagram presented in Figure 1 shows the result of a 24 h study conducted at the US-43 northbound approach to the intersection. Each dot in the figure represents the speed and location of an approaching vehicle at the onset of the red indication as well as its decision to either go or stop. Hollow green dots (
) represent vehicles that came to a stop. Blue dots (
) represent vehicles that ran the red light but were protected by the ARmin of 2.0 s. These blue-dot vehicles spent the first 0.8 s of the all-red interval traveling to the stop line and then used the remaining 1.2 s to clear the intersection. In the present paper, the red-light runners who completely cross the intersection before the conflicting traffic receives the green signal (i.e., blue-dot vehicles in the figure) are defined as type-A red-light runners. On the other hand, the red triangles (
) in the figure represent vehicles that ran the red light and were not fully protected by the 2.0 s of the all-red interval. These red-light runners were still inside the intersection (i.e., had not cleared the intersection) when the conflicting phase started serving the green signal. The present study classifies such vehicles as type-B high-risk red-light runners and further discusses how to effectively prevent them. As shown in Figure 1, 27 vehicles marked with red triangles were classified as type-B red-light runners because they were unable to clear the intersection before the end of the all-red interval. These red-light runners arrived at the stop line 0.8 s after the all-red signal had started. Although the number of type-B red-light runners may vary at different intersections (depending on the length of the all-red interval installed at each site), it is evident that these aggressive red-light runners are present and pose a significant safety threat. Therefore, type-B red-light runners require additional countermeasures (besides the ARmin) to reduce the chance of crashes associated with them. Equation 1 shows a traditional red clearance interval formula that is provided in the Traffic Engineering Handbook (7th ed.) and also used by the Alabama Department of Transportation. Here, R, W, Lv, and v represent the time required to clear an intersection, the intersection width, the length of a vehicle, and the speed of an approaching vehicle, respectively (
5
).
Please note that the purpose of Equation 1 is to ensure the safe passage of vehicles that are already within the intersection or that have just passed the stop line at the onset of the red indication, as it only considers the width of the intersection in its calculation. Thus, the duration of the all-red interval solely determined based on this equation only protects vehicles already within the intersection at the onset of the red indication, rather than the type-A and type-B red-light runners described in Figure 1. According to a recent study by Hurwitz et al. at an intersection approach with a speed limit of 55 mph in Oregon, approximately 20% of vehicles that violated the red light were located more than 100 ft away from the stop line at the onset of the red indication ( 6 ). The Oregon study showed that a significant number of vehicles even enter the intersection 1.0 s after the onset of the red indication at high-speed intersections, which also supports the arguments and findings of the present study. While Alabama employs the ITE equation to determine minimum all-red intervals of signalized intersections, recent practices, informed by an NCHRP study ( 7 ), recommend subtracting 1.0 s from Equation 1 to accommodate the startup delay of conflicting movements ( 5 , 7 , 8 ). This adjustment significantly reduces the length of the all-red interval, as shown in Equation 2:
However, it should be noted that this suggested decrease in the all-red interval in Equation 2 is based on the study of driver behavior for conflicting movements, not the study of subject drivers (i.e., red-light runners) on major road approaches. If the intersection depicted in Figure 1 were to utilize Equation 2 to calculate the red clearance time, its length would be reduced to 0.2 s (1.2 − 1.0 = 0.2). However, the speed–space diagram in Figure 1 demonstrates that reducing the all-red interval even more would lead to an increase in the number of red-light runners entering the intersection after the conflicting phase turns green, thereby increasing the number of type-B high-risk red-light runners. Taking the speed–space analysis shown in Figure 1 into account, despite having an all-red interval of 2.0 s for vehicles on the US-43 approach, which exceeds the length of the red clearance time (1.2 s) based on Equation 1 by 0.8 s, there are still a notable number of type-B red-light runners present. Reducing the length of the all-red interval using the modified kinematic equation would only exacerbate the situation. It is important to note that the utilization of the traditional or modified ITE equation solely for the determination of the all-red interval may not be sufficient for high-speed signalized intersections where a significant number of type-A and type-B red-light runners are present. Thus, traffic engineers generally opt to provide all-red intervals that are longer than those calculated by the ITE equations, based on their engineering judgment on high-risk, high-speed signalized intersections. For instance, the Florida Department of Transportation (FDOT) mandates an all-red interval of 2.0 s, regardless of the intersection size ( 9 ). This method can ensure a sufficient buffer time to accommodate vehicles that may still be entering the intersection after the onset of the all-red interval, thereby enhancing overall safety. However, traffic patterns and driver behavior vary with the time of day and location. Thus, simply providing a longer all-red interval using engineering judgment may not be an ideal solution that balances traffic safety and operations. Furthermore, instances of extreme red-light running are uncommon, and implementing an excessively long all-red interval in every cycle to account for these aggressive red-light runners would result in operational inefficiencies. Therefore, enhancing safety against red-light running crashes without incurring significant operational losses requires an effective engineering countermeasure (e.g., the Dynamic All-Red Extension [DARE] system) instead of simply adjusting the all-red interval.
The objectives of the present study are threefold: (1) to analyze the distribution of red-light runners along high-speed signalized intersection approaches during the all-red interval using a speed-space diagram; (2) to employ a speed–space analysis to discuss and evaluate the performance of the DARE system; and (3) to utilize the outcome of the speed–space analysis to determine key system parameters for successful field implementation of the DARE system. To meet the objectives, two signalized intersections located in rural Alabama, each with a speed limit of 55 mph or higher, were selected for the implementation of the DARE system and subsequent data collection. Note that the DARE system does not aim to reduce the number of red-light runners but instead focuses on safeguarding them, regardless of the reasons for red-light running. Consequently, the present study does not investigate drivers’ behavior during the yellow light or other factors influencing red-light running, as this is beyond the scope of the study. The authors’ previous studies have provided a comprehensive analysis of the dilemma zone characteristics and driver behavior at the onset of the yellow interval, identifying them as primary contributors to red-light running at high-speed signalized intersections ( 10 , 11 ).
The organization of this study is as follows: after the present introductory section, the next section will review the existing literature to see what countermeasures are available to deal with red-light runners. The following sections will provide a description of the site selection, data collection, and the DARE system proposed in this study to deal with high-risk (i.e., type-B) red-light runners. Afterward, results from the analysis are discussed, and the closing section concludes with study limitations and future works.
Countermeasure Literature Review
Several countermeasures have been developed to prevent crashes caused by red-light runners. The countermeasures explored to date can be broadly categorized into two groups: enforcement methods and proactive methods. Enforcement methods refer to the rigorous enforcement of laws to ensure drivers’ compliance with traffic regulations and policies, achieved via traffic tickets and citations ( 12 ). This may involve law enforcement officers stationed at intersections or the utilization of automated red-light cameras. Extensive research has been conducted to evaluate the impact of red-light cameras on both driver behavior and the overall safety of intersections ( 13 , 14 ). The research found that camera enforcement generally reduces the number of red-light running violations by about 40%–50%, accompanied by a 25%–30% reduction in overall injury crashes. However, red-light cameras can also lead to unintended consequences, such as abrupt stops and an increase in rear-end crashes ( 14 , 15 ).
Proactive methods are favored over enforcement methods as they focus on mitigating risky driver behaviors and potential crashes before they materialize, rather than merely responding to them ( 16 ). In the context of red-light-running vehicles, proactive measures aim to prevent crashes from occurring in the first instance. One widely adopted proactive countermeasure is the all-red extension system. This system is designed to detect potential red-light runners and extend the duration of the all-red interval, ensuring the safe passage of these risky vehicles. Although the concept of all-red extension is not new, advancements in vehicle detection technology have led to significant improvements in the system over time. Considering its evaluation process, the all-red extension system can be categorized into two types: static and dynamic.
A static all-red extension system offers a predetermined, constant extension time when it detects potential red-light runners using predefined speed and location criteria. This system utilizes advanced loop detectors positioned upstream or downstream from the intersection stop line, along with relevant logic. The implementation of a static all-red extension system was first observed in Portland, Oregon ( 17 ). This system adds additional time for a vehicle to clear the intersection if it is detected by a loop detector within the latter half of the yellow interval or during the all-red interval. The findings from the Oregon study revealed a reduction in the number of right-angle crashes after the system implementation. Another example of using the all-red extension can be found in North Carolina ( 18 ). The North Carolina study examined eight high-speed signalized intersections that used two advanced loop detectors per site for the all-red extension of through traffic. If a vehicle gets detected exceeding a speed threshold, the system calculates its available protection time, which comprises the yellow and all-red durations. If the required time for the vehicle to safely clear the intersection exceeds the available protection time, the system extends the all-red period accordingly. The authors of the North Carolina study claimed that their system is dynamic, as it provides different lengths of the all-red extension for different vehicles based on their times of arrival. However, the system assumes a constant speed when calculating the time required for vehicles to travel to the stop bar. It cannot reflect vehicle dynamics, such as sudden acceleration, deceleration, and abrupt stops of approaching vehicles. Thus, this spot-detection-based all-red extension system can be classified as semi-dynamic, and the likelihood of false alarm cases remains high. The study found that this system yields unnecessary all-red extensions because of false-positive cases and spends up to 8 s of additional all-red time per hour for vehicles that eventually stop at the intersection.
Radar-sensor-based vehicle detection has gained popularity because of its capability to continuously track vehicles ( 19 ). An all-red extension system utilizing radar sensors for vehicle detection is dynamic, enabling continuous tracking of the vehicle speed and location within a specific distance range (rather than at a single point). This dynamic capability offers a high level of precision—as low as 0.001 s—under varying traffic volumes, traffic compositions, and weather conditions ( 20 ). Consequently, its performance should not significantly vary among sites and time periods unless there are sight distance issues ( 21 ). As a result, the utilization of radar sensors for all-red extensions offers improved efficiency and vehicle safety in protecting red-light runners. The Maryland State Highway Administration employed a dilemma-zone protection system that uses microwave radars for vehicle detection ( 22 ). The system continuously extends the all-red interval at a rate of 0.1 s as long as vehicles within the range of the detection zone during the all-red interval exceed a specified speed threshold. The study did not find any false-negative cases (i.e., failures to detect an actual red-light runner) during its brief evaluation period. Another recent study conducted in Maryland utilized the advanced Intelligent Intersection Control System (III-CS), which includes features such as dynamic green extension (DGE) and DARE. The DGE feature was enhanced with an additional algorithm capable of terminating the green signal when the collision risk is at its lowest. The results from two study sites indicated a 100% success rate of the DARE system in detecting red-light running vehicles, with false alarm rates ranging from 7.3% to 11.7% ( 23 ).
Recently, the authors of the present study ( 24 ) conducted a comprehensive field evaluation of a radar-sensor-based dilemma-zone protection system utilizing dynamic green extension (DGE) features in Alabama. This study revealed a significant 50% reduction in the frequency of red-light running violations after the implementation of the DGE system. However, it is important to note that the main purpose of the DGE system is to promote vehicle arrivals on green so as to indirectly decrease vehicle arrivals in yellow and red intervals. The DGE is not a direct countermeasure to protect type-B high-risk red-light runners, as described in Figure 1. To protect drivers caught in the dilemma zone as well as the potential type-B red-light runners, the integration of the DARE feature with the DGE system is desired. The layout and main components of the DARE system are described in the forthcoming relevant section.
Methodology
High-speed signalized intersection approaches (where the speed of the approaching vehicle is 55 mph or higher) are often characterized by abrupt stops or rapid acceleration, making them vulnerable to dangerous red-light running ( 25 ). Additionally, dilemma-zone length and location from the intersection stop line exhibit a strong positive correlation with the approaching vehicle speed ( 26 ). Thus, high-speed intersections are prone to red-light running, making them ideal locations for studying driver behavior during the yellow and red clearance intervals. The present study employed the following criteria for site selection:
Signalized intersections with a posted speed limit of 55 mph or higher on the major approach
Signalized intersections that have encountered a significant number of red-light runners
Signalized intersections located in rural isolated areas where there is no signal coordination with adjacent intersections
Signalized intersections without installed red-light cameras
Intersections that met all these criteria were selected to analyze red-light-running vehicles as well as to implement the radar-sensor-based DARE system.
Recall that one of the study objectives of this paper was to assess the effectiveness of the DARE system in addressing type-B, high-risk red-light runners, who are not protected with a fixed all-red interval as described in Figure 1. Thus, the study focus was limited to rural, isolated signalized intersections because the cycle length varies over time on activation of the DARE system, so implementing DARE with signal coordination becomes a complex procedure. Furthermore, the selection of rural, isolated intersections as study sites can minimize potential biases associated with driver adaptation ( 18 ). Accordingly, two signalized intersections in Alabama were selected for the present study: US-231 & SR-109 in Dothan (Figure 2a) and US-43 & CR-96 in Mt Vernon (Figure 2b). These figures illustrate the layout and street view of the selected intersections. The intersections selected for the present study are situated on US-231 and US-43, both of which are multilane divided U.S. routes that serve as major approaches to these intersections. US-231 and US-43 are both major arterials that connect multiple cities in Alabama, and both have speed limits that vary as they pass through populated versus less populated areas. Note that in Alabama, U.S. highways serve as vital freight routes, and their speed limits are typically higher (e.g., ranging from 50 mph to 65 mph) when they traverse less populated rural areas. The posted speed limits of US-231 and US-43 are 65 mph and 55 mph, respectively. Standard six-phase dual-ring signal phasing was employed for US-43 & CR-96, and standard eight-phase dual-ring signal phasing was employed for US-231 & SR-109. Both the US-231 and the US-43 approaches were allotted a 5 s yellow interval. A minimum all-red interval (ARmin) of 2 s was established for both the major road approaches (US-43 and US-231), which is higher than the recommended value from the Alabama Traffic Signal Design Guide & Timing Manual ( 4 ). Average annual daily traffic (AADT) for the US-231 & SR-109 intersection was reported as 20,040 in 2022, while a count of 15,087 was recorded for the US-43 & CR-96 intersection during the same period.
The two selected study sites: (a) US-231 at SR-109 and (b) US-43 at CR-96.
Radar-Sensor-Based DARE System
The radar-sensor-based DARE system provides a dynamic extension of the all-red interval every one-tenth of a second by continuously tracking the speeds and locations of all approaching vehicles, and thus it offers a technological advantage when compared with traditional spot detection methods. The author’s previous publication includes a detailed description of how a radar-sensor-based DGE system works and its impact on intersection safety ( 24 ). The DARE system operates based on a logic similar to the DGE system and incorporates three main parameters: ARmin (the minimum length of the all-red interval), ARext (a unit extension time [i.e., 0.1 s] for vehicles within the detection zone exceeding a speed threshold during the all-red period), and ARmax (the maximum length of the all-red interval allowed). For both the US-231 and US-43 approaches selected in the present study, a 100 ft detection zone stretching upstream from the stop line was established for the DARE system (see Figure 3). In this context, the stop line was considered as the zero line. The detection zone for the DARE system can extend downstream of the stop bar depending on the sensor’s mounting location. For the US-231 approach (where the posted speed limit is 65 mph), the speed threshold for the DARE system was set at 50 mph, with an ARmax of 5 s and an ARmin of 2 s. For the US-43 approach (where the speed limit is 55 mph), the speed threshold was set at 45 mph, with an ARmax of 5 s and an ARmin of 2 s. When vehicles get detected within the defined detection zone at any moment during the all-red interval, the sensor places a call to the signal controller to extend the all-red period. As long as there are vehicles inside the detection zone exceeding the speed threshold, the sensor keeps sending calls to the signal controller to extend the all-red interval. The signal controller can extend the all-red period with a precision of 0.1 s after the end of the minimum all-red interval for each call received from the sensor. This process continues until no further calls are received or the length of the all-red interval reaches the ARmax condition. This continuous vehicle detection and tracking capability is advantageous because it ensures that the conflicting phase does not receive a green signal until all subject vehicles, regardless of their classification, have exited the detection zone. This addresses a key limitation of loop-detector-based all-red extension systems, which provide a fixed extension for both heavy and light vehicles despite their different dynamics within dilemma zones. Currently, there is no scientific consensus on the speed threshold and detection zone length for a radar-sensor-based DARE system. Opting for a lower speed threshold decreases the likelihood of false-negative results (failures to detect an actual red-light runner) but increases the probability of encountering more false positives (misidentifications of a vehicle as a potential red-light runner). Analyzing the characteristics of type-B, high-risk red-light runners can provide valuable insights to determine the ideal speed threshold and detection-zone length for a successful implementation of the DARE system with the minimum number of false-positive and false-negative cases.
DARE and DGE system layout with radar sensors at the US-43 and CR-96 intersection.
Data Collection and Processing
The data available from the radar-sensor-based DARE system include trajectory information for all approaching vehicles within the detection range, including dates and times of arrival, unique vehicle IDs, speeds, distances from the stop line, and the location where each vehicle was initially detected. Twenty-four hours of sensor and signal-controller event data from the major road approach of each selected intersection were collected for the analysis of red-light runners. The analysis began by filtering the signal controller event data for the study phase and further refining it to identify the exact start and end points of the all-red intervals. The total duration of the all-red interval in each cycle allowed for the determination of the frequency that the all-red extension was activated beyond the ARmin as well as the amount of additional time consumed by the all-red extension. The start time of the all-red interval for every cycle, retrieved from the signal-controller event data, was utilized as input for a Python program allowing for the filtering of vehicle IDs and their speed trajectories from sensor data during the red interval. Each vehicle ID and its speed trajectory detected within the red interval were meticulously observed to ascertain whether they could be considered red-light runners or not.
The speed–space diagrams created from this trajectory data illustrate the speeds and locations of all approaching vehicles at the onset of the all-red interval, aiding in the classification of these vehicles into four distinct groups. These classifications include vehicles that come to a complete stop during the all-red interval, type-A and type-B red-light runners, and any false-negative cases. To determine which vehicles ran the red light, a set of criteria was developed. As a vehicle approaches the intersection, its speed and distance from the stop line keep changing. Since the sensor considers the stop line to be the reference point (zero), when the vehicle passes the stop line, its distance from the stop line becomes negative. Thus, any vehicle whose location becomes negative after the onset of the red interval crossed the stop line during the red interval. Note that the sensor also tracked vehicles that changed lanes and made left or right turns during the red signal. As a result, the distance values for the turning vehicles could register as negative, potentially creating the false appearance of red-light-running violations. To ensure the accurate identification of actual red-light runners, two specific criteria based on the idea that red-light runners and turning vehicles have different speed profiles that can be used to differentiate them were established: (1) if a vehicle’s distance from the stop line was negative after the onset of the red interval and (2) if its last recorded location fell within 25 ft upstream of the stop line and its approach speed exceeded 40 mph, then the vehicle was classified as a red-light-running vehicle. Other vehicles that traversed the stop line at a speed of less than 40 mph during the red interval were considered turning movements.
Please note that the distributions of red-light runners shown in the speed–space diagrams in this paper were developed based on data collected from two intersections. While the distribution of red-light runners may vary among intersections and with varying traffic, weather, and time-of-day conditions, a similar trend is expected, with a clear distinction between type-A and type-B red-light runners. Aggressive behavior by type-B red-light runners represents a significant safety threat. Therefore, an effective engineering countermeasure, such as DARE, is needed to mitigate the potential safety risks posed by these high-risk type-B red-light runners.
Results and Discussion
Speed–Space Diagram After DARE Implementation
A speed–space diagram can offer an intuitive way of visualizing the characteristics of drivers who encounter a red signal at signalized intersections. It can provide valuable insights into drivers’ red-light-running behavior, illustrating their speeds and locations at the onset of red indication as well as their decisions (either to stop or go) during the all-red interval. The radar-sensor-based DARE system was implemented on the major road approaches of the two selected intersections in July 2022. Twelve months after the DARE system implementation, sensor and signal event data were collected to assess the effectiveness of the DARE system. A speed–space diagram for each intersection approach was produced with 24 h data collected from the sensor and signal controller. Figure 4 shows the outcomes of the analysis in speed–space diagrams for both intersections: Figure 4a for US-231 approaches at SR-109 and Figure 4b for US-43 approaches at CR-96.
Speed–space diagram of red-light runners at the onset of red indication for (a) US-231 approaches at SR-109 and (b) US-43 approaches at CR-96.
In Figures 4a and 4b, vehicles that stopped at the intersection during the all-red interval are denoted by green hollow dots (). Most vehicles (more than 95%) were classified into this category. Blue dots () represent instances of red-light runners who successfully cleared the intersection before the expiration of the 2 s of ARmin. These were classified as type-A red-light runners and spatially distributed within 100 ft upstream of the intersection stop line at the onset of the red interval. On the other hand, red triangles () in the figure represent vehicles that were not fully protected by the ARmin but were safeguarded by the DARE system. There is, however, one false negative (denoted as a yellow diamond 
) where the DARE system failed to extend the all-red interval despite the presence of a red-light runner. There may be several possible reasons for such a case. One possible reason is a communication error between the sensor and the signal controller. The speed–space diagram for the US-231 approach at SR-109 revealed a notable presence of high-risk (i.e., type-B) red-light runners. Among the 150 observed red-light runners, 87 vehicles were classified as type B as they were not protected by the 2 s of ARmin. Most high-risk red-light runners in this intersection were located between 100 ft to 340 ft from the stop line at the onset of the red interval. Among the 87 high-risk red-light runners, only one false-negative case was observed.
US-43 & CR-96 is a lower-volume and lower-speed intersection as compared with US-231 & SR-109. As a result, the number of red-light runners observed at this intersection during the 24 h study period was lower: 48, including 18 high-risk (i.e., type-B) red-light runners. For this intersection, most type-B red-light runners were located between 95 ft to 200 ft from the stop line at the onset of the red interval. Notably, no false-negative case was observed at this intersection since the DARE system was activated for every high-risk red-light runner encountered.
The speed–space diagrams presented in Figures 4a and 4b provide valuable insights into the speed and location profiles of vehicles at the onset of the red interval. However, it is important to note that this diagram does not offer information on the time it takes for the red-light runners to cross the stop line after the red signal starts. Examining the distribution of elapsed times between the onset of the all-red interval and the moment that red-light runners cross the stop line can assist in determining the ideal lengths of ARmin and ARmax for red-light runners. Figure 5 illustrates the distribution of vehicle time to the stop line after the red signal starts.
Distribution of vehicle time to red-light violation from the onset of red.
Figure 5 displays the travel times of red-light runners, which can be divided into two parts. The first part, depicted by a boxplot, represents the driver’s travel time to the stop line after the onset of the red indication. Here, the x-axis of the boxplot represents the stop line of the intersection, while the y-axis shows the distribution of travel times taken by the red-light runners to reach the stop line after the onset of the red indication. This time can vary from one vehicle to another, with type-B, high-risk red-light runners typically taking longer to enter the intersection. The second part of the travel time is the time required to clear the intersection (denoted as R), typically calculated based on the ITE equation (Equation 1). Note that the red clearance times calculated for the intersections selected for the present study were approximately 1.2 s and 1 s for the US-43 and US-231 approaches, respectively.
The box plot shown in Figure 5 reveals that red-light runners can enter the intersection several seconds after the red signal starts. However, extending the all-red interval excessively to cover all (i.e., 100%) of the red-light runners would lead to an increased delay at the intersection. Thus, it would be appropriate to allocate the length of the ARmin based on (1) the red clearance interval and (2) the distribution of time taken by the red-light runner after the red signal starts. It is reasonable to protect a substantial portion (e.g., 50% to 75%) of red-light runners through the utilization of ARmin, calculated based on the characteristics of the red-light runners in conjunction with site-specific engineering judgment. The remaining portion (e.g., 50% to 25%) can then be further safeguarded using the DARE system at high-speed signalized intersections. This can be achieved by introducing an additional time factor k when calculating ARmin, as shown in Equation 3, where R represents the red clearance time calculated based on the traditional ITE method and k is the 25th, 50th, or 75th percentile of the travel times of red-light runners after the onset of the red indication.
From the boxplot shown in Figure 5, the two study intersections exhibit significantly different distributions of vehicle time to the stop line after the onset of the red indication. This variability arises because of different site-specific characteristics, such as the intersection speed limit, traffic volume, and other relevant parameters. For the US-43 approach at CR-96, the 50th percentile for the time to red-light violation (k50) was determined to be 0.8 s. Meanwhile, k50 for US-231 at SR-109 was measured at 1.5 s. US-43 at CR-96 exhibited 1.3 s for the 75th percentile of travel time to the red-light violation (k75), whereas k75 for the US-231 at SR-109 notably increased to 2.4 s. Based on the findings from the present study, it was recommended that k values (e.g., k50) should be incorporated along with the red clearance interval (R) when allocating ARmin for the two selected intersection approaches. As such, 2.0 (1.2 + 0.8) and 2.5 (1.0 + 1.5) s of ARmin were recommended for the US-43 and US-231 approaches, respectively. One could argue that setting an ARmin that is long enough to ensure the safe clearance of all red-light runners would eliminate the need for a DARE system. However, this approach would result in substantial operational losses because aggressive (e.g., type-B) red-light runners are relatively infrequent, yet a fixed long ARmin would be applied during every cycle. Moreover, drivers familiar with the intersection might adapt to the longer all-red interval, potentially increasing the likelihood of red-light running. Therefore, implementing the DARE system and setting the all-red interval based on the k value for high-speed signalized intersections would minimize operational loss while effectively protecting high-risk aggressive red-light runners.
Figure 6 provides a conceptual depiction of two categories of red-light runners (type A and type B) across varying all-red intervals, utilizing the same dataset as depicted in Figure 4. Figure 6a illustrates the situation where the all-red period is set at 1 s according to the conventional ITE equation (Equation 1). In this case, all red-light runners fall into the type-B category, indicating a significant safety concern. As previously discussed, the traditional ITE equation focuses solely on the time necessary for vehicles to traverse the width of the intersection. However, since these red-light runners are positioned upstream of the stop line at the onset of the red signal, these vehicles will still be inside or entering the intersection when conflicting traffic receives the green signal. It can be argued that even if a vehicle runs a red light and remains within the intersection when the green signal serves for conflicting movements, they would still be safe since the conflicting traffic typically experiences starts-up delays and does not immediately enter the intersection. However, as discussed earlier, a significant number of red-light runners would still be inside the intersection 1 s after the start of the conflicting green signal. Moreover, recent practices advocate for the removal of 1.0 s from the traditional ITE equation (see Equation 2), which ultimately eliminates 1.0 s of the safety buffer. Thus, the all-red interval for the US-231 approach becomes zero (1.0 − 1.0) if Equation 2 is utilized for it, which is highly undesirable.
Comparison of type-A and type-B red-light runners with different all-red periods for US-231 approaches at SR-109: (a) 1.0 s, (b) 2.0 s, and (c) 2.5 s of all-red periods.
Figure 6b is identical to Figure 4a, where the all-red interval was set to 2.0 s, including an additional 1.0 s based on engineering judgment by the state traffic engineer. This extra 1.0 s offered an additional safety buffer for the drivers and resulted in approximately 40% of the type-B red-light runners from Figure 6a being converted to type-A red-light runners in Figure 6b. Please note that Figure 4a was reproduced in Figure 6b to allow for a good and comprehensive comparison with Figure 6, a and c . Figure 6c depicts a scenario where the recommended 2.5 s all-red period is implemented based on the traditional ITE equation (Equation 1) and the k50 value. In this scenario, 50% of the red-light runners become type A, and they receive full protection through the recommended 2.5 s of the all-red interval. The remaining 50% of the red-light runners (type B) shown in Figure 6c will be protected by the DARE system. This method would be more defendable as well as effective than the other two methods, as it combines the result (i.e., k values) of the speed–space analysis for red-light runners and the ITE equation. It is important to note that relying solely on the traditional or modified ITE equation when setting the all-red intervals is not desirable for high-speed signalized intersection approaches, as shown in Figures 4–6. Similarly, implementing longer all-red intervals based solely on engineering judgment lacks reliability. Instead, integrating the k value with the ITE equation presents a more logical approach for determining the all-red interval at high-speed signalized intersections.
In regard to ARmax (i.e., the maximum length of the all-red interval allowed), the k value for the furthest red-light runner’s time to the stop line (i.e., k100) can be used. As shown in Figure 5, the k100 values for the US-43 and US-231 approaches were found to be 2.4 s and 3.6 s, respectively. Thus, 3.6 (1.2 + 2.4) and 4.6 (1.0 + 3.6) s of ARmax were recommended for the US-43 and US-231 approaches, respectively. Note that the DARE system extends the all-red interval dynamically based on the speeds and locations of red-light runners. Thus, the all-red extension beyond ARmin varies with each incidence of DARE activation, and the length of the all-red interval does not reach ARmax in most cases. Note that the accuracy of detecting red-light runners and successfully extending the all-red period depends not only on ARmin and ARmax but also on the combination of detection zone length and speed threshold. To ensure that every possible red-light runner is captured, it is essential to set an ideal speed threshold that is lower than the speed profile of red-light runners within the detection zone. Additionally, the ideal length of the detection zone should be long enough to allow even the furthest red-light runners to enter the detection zone before the termination of ARmin.
Figure 7 displays the speed trajectories of 14 randomly selected type-B, high-risk red-light runners at the US-231 approach to the intersection SR-109 following the start of the red indication. As shown in the figure, it is observed that every red-light runner maintained a speed exceeding 50 mph within the last 150 ft and a speed exceeding 55 mph within the last 100 ft upstream of the intersection stop line. This type of analysis can help traffic engineers understand appropriate speed thresholds for corresponding DARE detection-zone lengths. For example, as shown in Figure 6, if a 100 ft detection zone (measured from the stop line) is used for DARE, then the corresponding speed threshold should be about 55 mph for the DARE system to protect all possible red-light runners. However, if a red-light runner enters the detection zone after the termination of the all-red interval, the system will not extend the all-red interval, and the safety risk will still be there. So, it is important to ensure that the detection zone length is long enough for the furthest red-light runners to enter the detection zone before the termination of the all-red interval. Let us examine the worst instances of red-light running vehicles found at the US-231 approach by referring to the speed–space diagram shown in Figure 4a. Two of the furthest-located type-B red-light runners at the onset of the red indication were 335 and 340 ft away from the stop line. Assuming 2.5 s of ARmin for the US-231 approach, calculated based on R and k50 values of 1.0 and 1.5 s, respectively, from Equation 3, it was observed from the sensor data that these two red-light runners would be located 75 and 85 ft upstream of the stop line 2.5 s after the onset of the all-red period. Therefore, it can be inferred that the detection zone length established at this intersection approach should be more than 85 ft from the stop line to protect the high-risk (type-B) red-light runners for the corresponding 2.5 s of ARmin. Incorporating such information with speed trajectory data is important for determining the appropriate detection zone length and corresponding speed threshold to ensure effective performance of the DARE system. Similar methods to ascertain the detection zone length, as well as the determination of the speed threshold discussed in the present study, can be employed for implementing the DARE system at other high-speed intersection approaches.
Speed trajectories of red-light runners after the start of the red indication on US-231 approaches at SR-109.
Table 1 presents a summary of the field assessment of the DARE system conducted for the two intersections selected in this study. As shown in the table, the system performed well when the DARE feature was activated for all the red-light-running vehicles detected. It is important to note that there were several occasions when a single DARE activation safeguarded multiple red-light runners. Therefore, the number of DARE system activations is fewer than the number of red-light runners detected. The outcome of these results suggests that red-light runners no longer pose a significant safety threat at these intersections, as they were fully protected by the DARE system. One limitation of the DARE system would be an increase in the lost time resulting from the additional time spent on all-red extensions and the potential occurrence of false-positive cases. However, it was found that only 3.6 and 1.6 min per day were spent because of the all-red extensions by the DARE system at the US-231 & SR-109 and US-43 & CR-96 intersections, respectively. Given the substantial safety benefits offered by the DARE system, the extra time spent on protecting high-risk red-light runners can be considered negligible. Finally, a false alarm rate of about 2%–3% was observed in the field study.
Field Assessment of the DARE System at Two Intersections in Alabama
Note: DARE = Dynamic All-Red Extension.
Conclusions
Red-light running has consistently remained a critical safety concern despite years of research and experimentation. This is particularly true at high-speed signalized intersections, where a significant number of high-risk red-light runners are often observed. While countermeasures such as red-light cameras and all-red intervals have led to some safety improvements, they come with limitations.
The primary objective of this study was to deepen our understanding of the characteristics of red-light runners at high-speed signalized intersections, which was achieved through a speed–space analysis. Furthermore, the present study applied this understanding to evaluate the effectiveness of the DARE system as well as to determine appropriate system parameter values (the speed threshold, detection zone length, etc.) for its field implementation, aiming to safeguard high-risk red-light runners with minimal operational loss.
A speed–space diagram can assist in visually capturing vehicle speeds and locations at the onset of red indication as well as in depicting subsequent decisions of red-light runners about whether to stop or proceed through the red light. Furthermore, it can help in identifying high-risk red-light runners (referred to as type B in this paper) based on their location and the time they take to violate the red light after the onset of red. The analysis revealed a noteworthy presence of high-risk (type-B) red-light runners. These aggressive drivers are different from typical (type-A) red-light runners who are protected by the all-red interval. The type-B red-light runners identified in this study are positioned between 100 ft to 350 ft upstream of the intersection stop line at the onset of red indication, presenting a significant safety threat. This study showed that employing additional countermeasures beyond the minimum all-red interval (ARmin) is essential to reduce the likelihood of crashes associated with those drivers. An engineering countermeasure known as the DARE system was implemented at two high-speed signalized intersections in Alabama to safeguard these high-risk vehicles and those potentially involved in crashes with them. The outcomes of the speed–space analysis, alongside the distribution of red-light runner times to the stop line, revealed that improved protection for red-light runners could be achieved not only by employing the minimum all-red (ARmin) intervals based on the red clearance time but also by considering the 25th-, 50th-, or 75th-percentile value of the red-light runners’ times to the stop line. Based on the findings, the study proposes the utilization of speed–space analysis, vehicle time to red-light violation, and speed trajectory analysis to determine key system parameters to ensure the optimal performance of the DARE system. With these analyses, site-specific detection-zone lengths, speed thresholds, and other system parameters can be determined to maximize the safety benefit of the DARE system. Although the study findings are based on only two intersections, the methodology employed in this study can be used to explain site-specific variations in the distribution of red-light runners along the approaches of other intersections during all-red intervals. Additionally, the accuracy of microwave radar sensors remains consistent across different locations unless there are sight distance issues. Therefore, the methodology and recommendations of this study are reasonably applicable to any high-speed signalized intersection.
Field assessment of the DARE system demonstrated that all red-light runners (including the type-B runners) were effectively protected by all-red extensions. It was found that only 3.6 and 1.6 min per day were spent as a result of the all-red extensions, with a 2%–3% false alarm rate of the DARE system at the two intersections selected in the study. Considering the significant safety advantages provided by the DARE system, the additional time spent protecting high-risk red-light runners can be deemed negligible.
Study Limitations and Future Work
The scope of the present study is limited to rural high-speed signalized intersections where there is no signal coordination with nearby traffic signals. The characteristics of red-light runners in urban signalized intersections may differ to some extent. Also, retaining signal coordination in urban corridors with DGE and DARE systems can be technically complex. Thus, future work would include examining the behavior of red-light runners in urban settings and implementing and evaluating the DGE and DARE systems at coordinated signalized intersections in urban areas, based on the findings from a speed–space analysis of red-light runners. It should also be noted that the assessment of the DARE system presented in this study is not exhaustive. A more comprehensive field assessment incorporating additional data obtained under varied traffic, weather, and time-of-day conditions would provide more conclusive insights into the distribution of red-light runners and the performance of the DARE system. Future research may also include a comprehensive sensitivity analysis of DARE system performance with different system parameters and varying detection zone lengths.
Footnotes
Author Contributions
The authors confirm their contribution to the paper as follows: study conception and design: M.-W. Kang, R. Hossain; data collection: M.-W. Kang, R. Hossain, M. Rahman; analysis and interpretation of results: R. Hossain, M.-W. Kang; draft manuscript preparation: R. Hossain, M.-W. Kang, M. Rahman, P. Biswas. All authors reviewed the results and approved the final version of the manuscript.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Alabama Department of Transportation; grant number: SPR-0001(062).
Data Accessibility Statement
Data used in this study are available on request.