Abstract
Vulnerable road users (VRUs, i.e., pedestrians and bicyclists) have seen an alarming rise in fatalities in recent years. School-age pedestrians in lower-income neighborhoods may be particularly at risk. This paper proposes a data-driven safe-systems approach to develop safety countermeasures for areas near elementary schools serving disadvantaged populations. A review of past literature on child-pedestrian training programs confirms that videos, lectures, and website-based training can provide children with vital information to improve their cognitive abilities relevant to walking safely. However, to improve pedestrian behavior on the road, children need to be safely exposed to traffic environments and practice interactions with traffic. Therefore, the use of virtual reality (VR) is recommended as a platform to introduce children to traffic interactions. Furthermore, the review of existing child-pedestrian training programs showed that most existing training programs (VR-based or otherwise) have an ad-hoc selection of roadway and traffic environment scenarios. Moreover, none of the training programs are designed to address safety issues faced by children in low-income neighborhoods specifically. To address these issues, this study gathered and analyzed crash data for VRUs around schools located in two metropolitan areas: Dallas County in TX and Tampa Bay in FL. Analysis of crash data identified the most prominent factors leading to most crashes as well as disproportionately more severe crashes. A VR-based child-pedestrian training program that involves children interacting with designed environments grounded in local crash data from disadvantaged areas is recommended for a more effective and equitable training platform.
Pedestrian fatalities continue to increase worldwide, with an overrepresentation of children aged 15 years or younger. According to the report published by the World Health Organization on “Child Injury Prevention,” pedestrian injury is among the leading causes of pediatric death in the United States and much of the world (1, 2). The National Safety Council estimated the cost of a pedestrian injury to be around $58,700 per event and the cost of a fatality as $4,538,000 per occurrence ( 3 ). The National Highway Traffic Safety Administration (NHTSA) reported 6,283 fatalities and 75,000 injuries in 2018 for pedestrians from all age groups in the United States ( 4 ). These statistics showed a 3.4% increase in pedestrian fatalities and a 5.4% increase in pedestrian injuries from 2017 and were the highest since 1990. Among these victims, around 17% of the pedestrians killed and 4% of the pedestrians injured were children aged 15 years or younger ( 4 ).
A report on “School-Transportation-Related Crashes” states that 100 school-aged pedestrian deaths that occurred between 2009 and 2018 were attributable to school-transport-related crashes ( 5 ). Past research has also shown that child pedestrians are at risk of severe injuries and fatalities in low-income areas, especially near schools that are located to serve disadvantaged populations (6, 7). The severe injuries and fatalities are higher in such areas, potentially because of socioeconomic issues, which cause children to be more likely to walk than to use any other mode of transportation ( 8 ), even as the surrounding roadway network remains automobile-centric. Low-income neighborhoods often also have poorly designed environments that make them prone to more traffic crashes ( 9 ).
One of the reasons behind children-involved pedestrian crashes is their stature, which makes it difficult for them to be spotted on the streets, especially for drivers in large sport utility vehicles (SUVs). Studies have suggested that most traffic fatalities can be prevented with proper interventions ( 10 ). In the long term, the most effective solution to addressing this issue is to improve the land use and infrastructure to be more vulnerable road user (VRU)-friendly. However, for the short-to-medium term, approaches that involve effective training programs for children to make safe decisions, consistent with their cognitive abilities, can help ameliorate safety issues caused by poorly designed traffic environments for pedestrians and other VRUs. Such training can improve the safety of children walking or cycling to school through educational and infrastructural developments and modifications ( 11 ). The practice of walking or cycling to school will have the added benefit of increased physical activity for the children, promoting better cardiovascular health, and reducing the risk of childhood obesity (10, 12, 13).
Walking is a natural form of exercise and can help address childhood obesity by ensuring physical activity during the school week ( 14 ). Inculcating safe walking behaviors such as walking far away from the traffic and carefully looking both ways while crossing the street, among others, will help make this beneficial activity more safe. Schwebel et al. mention some behavioral risk factors for child-pedestrian injuries including developmental factors (cognitive and perceptual development), distraction, temperament and personality, social influences, environmental risks, special populations, and sleep and fatigue ( 2 ). Awareness of these risk factors has encouraged stakeholders to design prevention initiatives to ensure child-pedestrian safety. There are different programs all over the U.S. to train child pedestrians about safety, such as WalkSafe ( 15 ). This is a 2.5 h-long interactive classroom training program standardized by age. The program uses imaginary roads, videos, educational material, and workbooks. Similarly, the Child Pedestrian Injury Prevention Project (CPIPP) is an Australian comprehensive pedestrian safety program that includes both home and school lessons, as well as community-based initiatives ( 16 ).
The introduction of emerging technologies like virtual reality (VR) has opened the door for more interactive and realistic training modules. VR allows children to be immersed in high-risk traffic environments without being exposed to actual danger (2, 17) and identifies risky behaviors exhibited by child pedestrians (17, 18). Several studies have used VR for training purposes and have found that children can grasp safety training better and quicker in a virtual environment as opposed to video-based or paper-based training (19–21). However, similar studies in the context of school zones in the more impoverished urban areas are rare. Also, very few studies have explored child pedestrians’ cognitive and perceptual skill development resulting from training programs. Furthermore, almost none of these studies have used scenarios that are grounded in the analysis of crash data near schools.
In this paper, the researchers have gathered and analyzed crash data for VRUs, both pedestrian and cyclist, around school zones located in two metropolitan areas: Dallas County in TX and Tampa Bay in FL. The inclusion of both pedestrian and cyclist populations helped us to analyze more crash events and to recognize factors with high-risk travel conditions for VRUs. Analysis of the crash data provides a framework to inform training programs that educate children on safe pedestrian–traffic interactions using VR platforms. The roadmap to a comprehensive and effective child-pedestrian training program is provided by considering crash data-based analysis for module and scenario selection and by incorporating VR. The paper is organized as follows: first, a literature review of factors associated with crashes with a focus on behavioral factors and their interaction with the roadway environment is provided, followed by a review of the state of the art in children’s training programs based on emerging technologies. A descriptive analysis of crash data from the two metro areas is then provided to demonstrate a roadmap for more effective VR-based training and intervention programs for child pedestrians. This is followed by a Discussion section and Concluding Remarks.
Literature Review
Child-pedestrian injuries and fatalities are primarily caused by complex interactions between unsafe roadway environments and children’s stature and their limited cognitive abilities ( 1 ). The literature review provided in this section focuses on the risk factors in relation to child pedestrians’ injuries and fatalities near school zones; child pedestrians’ perception of traffic infrastructure and environment (e.g., number of lanes, vehicle type, time of day, weather conditions, etc.); their cognition for gathering information, processing information, and making a decision; and social and environmental influences on children’s pedestrian behavior.
Risk Factors for Pedestrian Behavior
Cognition
As pedestrians are one of the most vulnerable road users, they must have proper cognitive skills for information gathering, information processing, decision making, and decision initiation. To gather information from a traffic environment, pedestrians need strong visual attention. Pedestrians must attend to all relevant information from available visual cues to identify important information. Sometimes distraction can affect pedestrians’ attention and lead to collisions. In the case of school-going children, both talking on phones and talking with companions were found to cause significant increases in risk-taking and significant decreases in safety while crossing a street ( 22 ).
Once pertinent information has been gathered by attentive pedestrians, they need to process this information and simultaneously make an estimation of whether the conditions are safe or unsafe for their specific maneuver. This information processing can become very complex based on factors such as roadway type, traffic flow, and the presence of other road users, among others. If there is a traffic collision, VRUs like pedestrians, who are not protected by vehicle structures, are the ones who will suffer the most severe injuries. Therefore, it is critical for pedestrians to develop proper cognitive skills for information processing especially given that the roadway network in much of the U.S. is not designed with a safe-systems approach.
The skillset mentioned above is highly dependent on the age of the child pedestrians. In a traffic safety context, children are defined as persons aged from 0 to 15 years. There are considerable differences in the abilities of younger and older members of this 0 to 15-years-old cohort. In the cognitive development literature, Piaget’s theory of development stages has been used to further classify the children’s ability to interact with the roadway environment by age ( 23 ). Specifically, children under 6 years do not have fully developed and differentiated sensory functions. They cannot focus on two different tasks (e.g., finding the correct route and checking for traffic signals). They cannot combine two separated perceptions into one (speed and distance of approaching a vehicle from how fast the vehicle’s perceived size grows). In addition, data on child-pedestrian mobility, especially for children under the age of 6 years, are scarce.
A large body of past research showed that children’s cognitive skills become more developed with age, for example, Piaget ( 23 ), Wang et al. ( 24 ), and Barton et al. ( 25 ). Cognitive skills which allow the processing of multiple visual cues from the traffic environment generally start developing from the age of 7 (24, 25). Prior research with children aged 5 to 12 years suggested that only older children are able to discriminate between more relevant and less relevant visual stimuli when presented with depictions of visually complex pedestrian settings ( 26 ). By age 9, most children can identify safe routes for walking, safe crossing locations, and objects (cars, structures, road users, etc.), impeding the view of oncoming traffic (26, 27).
Perception
To make safe decisions, pedestrians must accurately perceive all the information from the traffic environment around them. Most children develop the physical capacity to see and hear traffic ( 28 ) from the age of 7. However, it is not enough for pedestrians to attend to all the visual and audible cues from surrounding traffic and road users. To find a gap between traffic, pedestrians have to simultaneously measure and judge vehicle size, speed, distance, traffic density, and acceleration/deceleration of all approaching traffic, as well as look for these characteristics in traffic making a turn into their path. At the same time, they have to account for the number of lanes, road structure, traffic signals, presence of emergency vehicles, and road users other than motor vehicles. Past research shows that children under the age of 9 years are not skilled at these estimations; they tend to notice vehicle presence and distance but do not consider the speed and distance needed to directly interact with traffic (e.g., crossing distance) nor the acceleration/deceleration of the oncoming vehicles ( 29 ). Often, traffic infrastructure can make pedestrians’ perception of traffic conditions more challenging. Obstructions (parked cars or trees) and road features (road bends or curves) can block pedestrian vision; inclines can affect vehicle acceleration/deceleration and challenge pedestrians’ perception of those changes.
Social Influence
Children first learn about pedestrian behavior from their parents or family members. Research has documented that children are attentive to their parents’ safety practices and that children notice when parental practices diverge from safe behaviors ( 30 ). Therefore, the behaviors that parents model when walking between traffic have a potential influence on children’s practices as a pedestrian. The study was based on observation/interviews with parents/caregivers of children 4 to 11 years old.
Environmental Influence
The type of environment and society a child-pedestrian lives in can affect their walking behavior. The most critical risk factor can be the population and traffic density of the area. Children are more likely to be hurt near schools, presumably because exposure rates are higher. Greater exposure to traffic leads to greater pedestrian injury risk. Therefore, children in urban, higher population, and higher traffic density areas are more likely to experience a pedestrian injury than those in less-populated areas ( 9 ). A secondary consequence of this is that children from lower socioeconomic status backgrounds tend to have higher injury rates, as poorer urban communities tend to have traffic infrastructure that is not well designed and well maintained ( 9 ). This study defined individuals aged between 5 to 19 years as children.
In summary, the literature search in relation to risk factors shows that low-income school zones with poor traffic infrastructure may lack walkability for child pedestrians, resulting in the need for children to be especially attentive in such neighborhoods. The lack of school-provided transportation and adult supervision means that many of these children from low-income families walk alone to school. Although walking is a good practice for physical activities, poor walking infrastructure, unsupervised walking, and exposure to school-zone traffic, added to the limited cognitive and perception skills of children, can increase child-pedestrian injury and fatality risks in low-income school zones. Thus, it is an equity concern as well. Therefore, this research intends to investigate crash data in school-zone areas for VRU-involved crashes and develop a roadmap to child-pedestrian training programs for elementary school children located in low-income areas.
Child-Pedestrian Training Programs
There has been much research done on the efficacy of various training interventions (20, 31–35). The training programs reviewed in this section were primarily focused on elementary school children, as that is the focus of this research. Parents play a vital role in teaching children how to exhibit safe pedestrian behavior, but research has shown that parental training alone is not enough to teach children safe road-crossing skills ( 2 ). One study designed a gamified e-learning platform that provides learners incentives similar to those used in games to look at different criteria of pedestrian training such as traffic knowledge, situation awareness, risk detection, and risk management ( 36 ). This is self-learning computer-based training that requires minimal supervision and uses gamification elements and context-related footage to train young pedestrians. The researchers found that participants’ skills improved in each of the four modules and confirmed this as an efficient mode of training ( 36 ). Another study looked at a school-based intervention method that involves both theoretical and practical aspects of traffic safety education. The researchers compared this approach with a strenuous but expensive method of training in Iran’s traffic-park, a training complex designed to create a traffic environment for elementary school students ( 37 ). Results showed that the school-based intervention was more effective than the traffic park-based intervention ( 37 ).
Video training has been a popular method of training child pedestrians (31, 32). When Arbogast et al. compared training through interactive video games with traditional didactic studies, they found that participants who were trained by playing video games performed similarly to those who were trained in a more conventional, labor-intensive setting ( 31 ). However, the video game group exhibited more appropriate behavior on specific behaviors such as exiting a parked car, signaling to a car that was backing up, signaling to a stopped car, and crossing streets. Similarly, Hammond et al. designed an interactive hazard perception video that teaches children the skill of crossing safely between parked cars ( 32 ). They found the kind of training that focuses more on awareness skills rather than knowledge and acquisition alone to have a more positive impact on children’s behavior. Their results show that interactive hazard identification could improve the on-street behavior of child pedestrians. However, Schwebel and McClure found that widely available videotape and website training tools that require minimal to no adult support were ineffective in improving children’s pedestrian route selection ( 34 ). They suggested that this intervention did not improve children’s pedestrian behavior compared with children with individualized training in the control group. This result is consistent with a past study outcome that reported videos and lecture-based training programs on traffic safety can successfully improve children’s perception of safety, their attention, and information-processing skills, but cannot improve their behavior as pedestrians ( 38 ). These results pose an important concern and highlight the need to find more effective intervention methods to train children on safe pedestrian behaviors.
Recent researchers have indicated interest in incorporating VR-based training as an effective intervention method for individualized street-side training (32, 34, 35). Some advantages of using VR training include creating a complex, real-world traffic environment in the virtual world that eliminates the need to put the child in real danger and provides an authentic context to identify dangerous pedestrian behaviors (2, 39). Moreover, VR not only allows the identification of risky behaviors, but also enables recording and further analysis for feedback and training module updates. It also allows the children to practice repeatedly with minimal adult supervision. Lastly, these training approaches can be “gamified,” making them interactive and enabling the children to gain understanding while enjoying the game (33–36).
A case study in elementary schools looked at the efficacy of implementing VR-based training and found that VR increased pedestrian performance both during and after the intervention ( 33 ). The study consisted of different locations in both urban and rural regions. Results indicated no significant difference in the performance between the two groups (urban versus rural), suggesting that VR training can be universally applicable. Researchers also looked at having a mobile virtual environment that can be used to train children in the community ( 20 ). The advantage of having a mobile environment is that it can be moved around to different schools and community centers and provide intense training to multiple groups of children over a few weeks. Comparing their pre and post-study results, the researchers found pedestrian behavior improved modestly. In one of the previous studies, researchers compared different training interventions in knowledge gained and behaviors changed in the children ( 38 ). Children who received VR training exhibited safe behavior but did not gain knowledge, and children trained via videos/software/internet gained knowledge but did not change their behavior. Children who received a theoretical background about safe pedestrian behavior followed by VR-based training gained knowledge and safe behavior. These results suggest that although VR is an effective tool to improve pedestrian behavior, other platforms might be needed as supplements to enhance safety training.
Morrongiello et al. looked at the cognitive and perceptual aspects of VR training in different traffic conditions ( 35 ). They provided feedback to the children on their crossing behavior to help study participants with cognitive learning. With repeated practice, children developed the allocation of visual attention for interpreting vehicle movement. Results showed that this training not only improved pedestrian behavior, but also advanced their conceptual learning. These results bridge the gap that was found in another study where children did not seem to gain cognitive knowledge from VR interventions ( 40 ). Like VR, Cave Automatic Virtual Environment (CAVE) also provides a safer environment to conduct pedestrian behavior studies. CAVE is an arena surrounded by projector screens that can create a 3D environment for virtual exposure. Dommes and Cavallo used CAVE to train the elderly population by utilizing repeated practices in simulated environments, providing personalized feedback, and having educational discussions ( 41 ). The intervention seemed to improve their street-crossing behavior; however, the improvement was no longer observed after 6 months. The participants also failed to judge the speed of approaching vehicles while making their decisions, showing their struggle with perception and cognition.
Lastly, in VR training, children should not only have no adverse effects from VR ( 42 ), they should also feel the realistic immersion within VR environments. The literature has shown that VR headsets do not pose any risk of photo-induced seizure from the 3D view, even in children with known photo-sensitive epilepsy. Studies have also found VR to be effective for the realistic immersion of children aged 4 years and up ( 43 ). Table 1 summarizes past literature on pedestrian training.
Summary of Intervention Programs
Note: CAVE = Cave Automatic Virtual Environment; VR = virtual reality.
Conclusions from the Background Research
The following conclusions in relation to an effective training program may be drawn from this detailed review of the behavioral research and existing child-pedestrian training programs.
These studies have created child-pedestrian training modules for all income-group populations. However, most of the more affluent children of this age group are driven to schools by their parents. Not focusing these training programs on low-income neighborhoods may perpetuate an equity issue. The authors believe that the training programs can be more effective if tailored for children in low-income neighborhood schools where a larger number of children walk to school, unsupervised, on the street networks that involve higher crash risk because of the state of the infrastructure.
This research focuses on elementary school-age children (5–12 years). As the past research has shown VR to be safe for children as young as 4 years old, VR-based pedestrian training is a safe way to encourage safe behavior by the children.
The previous studies have included many scenarios considered high-risk situations for pedestrians; for example, crossing with obstructions blocking the visibility of approaching traffic (parked vehicles, blind curve, or blind hill, etc.) ( 35 ), crossing while car making left or right turns ( 33 ), crossing at wrong lights ( 31 ), crossing at unsignalized intersections ( 34 ), and so forth. However, these scenarios have been chosen by the researchers of past studies in an ad-hoc way, and may not represent all possible high-risk conditions, or even the conditions children are most likely to encounter.
To develop training programs that can address safety problems for children walking to schools and equity issues with child-pedestrian injuries in the low-income areas, we need to rely on insights from crash data near school zones to identify infrastructural factors as well as behavioral factors that lead to crashes. These factors can then be used to design VR-based scenarios to create effective training programs.
Crash Data Collection
We have collected crash data from two metropolitan areas (Dallas County, TX and Tampa, FL). Data were obtained from the respective States’ Departments of Transportation databases for years 2015 through 2019. The aggregated data were filtered based on Person Type (pedal cyclist and pedestrian) and were spatially joined to elementary school districts in the selected areas. To obtain a larger sample size we used crashes involving either pedestrians or bicyclists, as both groups are considered VRUs in the traffic environment dominated by automobiles. We made half-mile buffers around each elementary school point and then selected crashes that occurred in these half-mile buffers around each school location. As we analyze the crash data, it is important to remember that both frequency and injury severity of crashes are crucial for pedestrian training scenario development. From the VR-based scenario design standpoint, any situation with more crashes is worth considering in the training module, even if the number of crashes is high only as a result of increased exposure. This allows us to make sure that children are trained for scenarios they are most likely to encounter. In addition to the most likely to encounter scenarios, we also looked for factors that led to disproportionately more severe crashes (involving fatal and incapacitating injury to the VRUs). Although some of these scenarios may not be the most numerous, they still represent higher risk and vulnerability to the children. The factors which may be most relevant for developing scenarios based on both frequency and severity criteria are described in the next section.
The specific process used for the collection of crash data is as follows:
Five-years crash data were collected and aggregated through TxDOT’s Crash Records Information Systems (CRIS). These data included 605,244 crashes from 2015 to 2019 in Dallas County.
The aggregated 5-year crash data were filtered based on pedal cyclist and pedestrian involvement, resulting in 5,388 crashes.
We made half-mile buffers around each elementary school located in Dallas County and then selected unique crashes in those half-mile buffers around the school locations. Accordingly, the final crash data included 2,077 unique crash records.
The exact process was followed for the data from the Tampa Bay area.
Results: Crash Data Analysis
This section provides a descriptive analysis of crash data from the roadway network within a 0.5-mile radius surrounding the elementary schools located in two metropolitan areas (Dallas County, TX and Tampa, FL). Results are presented based on critical factors for VRU-involved collisions. The first factor investigated was the number of lanes. According to the past literature, the majority of the crashes happen in two-way road conditions ( 44 ). Previous research also shows that two to three lanes with no physical median is a critical location characteristic for urban traffic crashes involving VRUs ( 45 ). As our data involved multiple jurisdictions in Dallas and Tampa Bay regions and relied on the state DOT databases, crashes on locally managed surface streets were missing key attributes (Texas data in particular), including the number of lanes. Thus, we could not analyze the context-specific crash data with respect to the number of lanes. Therefore, future VR experiments for this work would be informed by the number of lane information for the roads most commonly encountered in the network surrounding the school. In the paper’s Discussion section, we further elaborate on this limitation.
Crashes by Traffic Control
Three traffic control scenarios were predominant in the crash data (see Figure 1). Most injury crashes happened in no traffic control scenarios, followed by signalized intersections and intersections with stop signs, respectively. It has been documented in the literature that children feel comfortable crossing roads at a stopped-controlled intersection near school zones where vehicles are required to come to a complete stop (17, 46). In the case of signalized intersections, it can be overwhelming for children to process information from both traffic flow and traffic signals to make a safe decision. When there is no traffic control, children only have to depend on visual cues from the traffic and get confused with their limited cognitive abilities and lack of practice.
Frequency and relative severity of crashes involving vulnerable road users corresponding to different traffic control scenarios.
Crashes in all three traffic control conditions are numerous enough to warrant consideration within a VR-based training platform. However, no-control and signalized intersection control scenarios should be prioritized because not only do they experience most crashes, but they also have a disproportionately higher share of severe injury crashes.
Crashes by the Speed Limit, Traffic Volume, and Vehicle Type
The speed limit is a critical factor for child-pedestrian training. Although areas close to schools are, in general, lower speed-limit zones, drivers not complying with posted speed limits is a concern ( 47 ). In addition, sudden speed reductions may worsen drivers’ control of their vehicles and result in collisions ( 48 ). Figure 2 shows crash frequency and relative injury severity for both Dallas and Tampa Bay school-zone areas by speed limits of the locations where the collisions occurred. For Dallas school-zone areas, most crashes occurred at locations where the speed limit was 30 to 45 mph, with 1,741 total crashes, including 1,080 moderate-to-severe injuries. Crash counts for this speed-limit range were substantially higher than other speed-limit categories. This may be caused by high exposure, as most streets in the school zones likely had the posted speed limits in this range. For locations in Dallas with a speed limit of more than 45 mph, although the number of crashes was low, severe injuries resulted in 55.4% of those crashes. In Tampa Bay, there were only seven crashes in 45+ mph zones, again emphasizing the need for programs tailored based on local crash experiences. For the Tampa Bay region, both frequency and severity criteria would lead to the inclusion of the speed-limit range between 30 to 45 mph in the VR-based training modules.
Frequency and relative severity of crashes involving vulnerable road users corresponding to speed limits (in mph).
Our analysis showed significant observations with missing annual average daily traffic (AADT, a measure of traffic volume) in the crash dataset (both Dallas and Tampa Bay) because of the same issue that resulted in the missing information about number of lanes. Traffic volume in the form of AADT, however, may not be a critical variable for VR-based experiment design. As traffic volumes are not evenly distributed throughout the day, the training modules should be designed to expose children to conditions prevailing at the time they are more likely to be out on the roadway network as a pedestrian (e.g., within 15–30 min of school before and after school start and end time). In both regions, vehicles involved in most collisions were passenger cars, followed by SUVs and pickup trucks. These findings are consistent with the national statistics from NHTSA ( 4 ).
Crashes by Weather Conditions
Past literature shows that most VRU-involved crashes occur during clear weather conditions. This may be because more pedestrians are likely to walk in clear weather rather than when there are unfavorable weather conditions like rain or snow ( 44 ). In both Dallas and Tampa Bay areas, most of the crashes occurred in clear weather. For the Dallas area, out of 1,616 clear-weather crashes, 1,031 crashes (~63.8%) involved moderate-to-severe injuries. In the case of the Tampa Bay area, out of 497 clear-weather crashes, 279 (~56.1%) involved moderate-to-severe injuries. As children and other VRUs are more likely to be on the road in clear weather as compared with adverse weather conditions, it is logical to include those conditions in the training scenario. If a school serves a high number (or share) of children that do not have any other available mode of transportation and are forced have to walk (or bike) to schools even during adverse weather, then the training scenarios should include inclement weather conditions as well.
Crashes by VRU Location and Action
Das et al. found that mid-block crossings had the highest number of crashes ( 44 ). In our data, marked and unmarked crossings were found to be the most predominant locations, and the crossing maneuver most frequently preceded the collisions involving injuries. This information about mid-block crossing was only available from the Tampa Bay region crash data.
In the next section, we summarize the results from crash data analysis (see Table 2) and propose potential scenarios to be included in a comprehensive VR-based child-pedestrian training program. This strategic method of selecting scenarios will provide more effective training for elementary school children. With VR-based training, we can leverage the low-cost and interactive platform to (i) mimic exposure to realistic high-risk scenarios without causing actual threats to safety, (ii) have opportunities with repeated trials, and (iii) provide immediate feedback from the training platform for corrective action.
Summary of Critical Factors
Note: SUV = sports utility vehicle.
Roadmap for Training Program
Emerging technologies like VR provide platforms to safely expose young children to potentially high-risk traffic environments. VR-based training may also be useful for introducing children to future vehicle technologies that are not currently available. In the coming years, the transportation system is expected to go through many changes, including the introduction of automated and connected vehicle technologies, the development of supporting infrastructure, and the demonstration of user acceptance. Advances in VR technology would allow us to prepare for a truly multimodal future by helping us better adapt to these changes and prepare the future generation to interact with connected and autonomated vehicles.
Therefore, we propose using VR platforms to develop a child-pedestrian training program that can be adapted to future changes in the transportation system. To achieve the most equitable impact through this safety improvement program, the authors propose to target elementary school-going children living in low-income areas (areas with income lower than 50% of the state median income). Based on the data collected from two metropolitan regions, we have identified prominent factors that should be used in the child-pedestrian training modules. Table 2 summarizes the factor associated with the highest frequency and highest injury severity of crashes.
Discussion
As this study relied on state DOT-maintained datasets, some of the crash data from locally maintained roads were missing the number of lane and AADT information. We concluded that the absence of AADT data is not critical to designing training programs. However, the missing data on the number of lanes is critical. It meant that our scenarios for the future VR-based evaluation can not be informed by relevant collision data on the number of lanes (at the crash sites) and would have to rely on the examination of the local network near the schools corresponding to the target student population. Although that may be acceptable given the current state of practice for such training programs, the ideal situation would be to have collision data inform VR-experiment scenarios on the number of lanes.
Results from two different regions also highlight the differences in regional roadway networks (e.g., speed limits) that can and should be reflected in the scenario development. Scenarios included in past studies developing intervention programs with high-risk scenarios are also listed in Table 2. It may be observed that several crash data-based factors have never been included in the scenarios tested in the literature. The proposed systematic approach to integrating these crash data-based factors within VR-based training scenarios is a more effective way to achieve the desired outcomes for the training programs.
The next step for this work is to develop a VR-based education program and estimate its effectiveness for improving children’s interaction with the roadway environment. In the next phase of this research, we are working with lower-income school districts in the Dallas area to design and conduct relevant experiments.
Conclusions
An extensive review of literature on existing child-pedestrian training modules led the authors to the following conclusions:
None of the pedestrian training modules are specifically designed for low-income neighborhoods, even though poor infrastructure and the lack of transportation options pose unique safety challenges for school-age children in these neighborhoods.
The scenarios developed by most studies for child-pedestrian training were selected on an ad-hoc basis with no grounding on the crash data.
Recent advances in VR technology have created an opportunity to develop more effective child-pedestrian training programs, especially if combined with supplemental programs (e.g., gamified e-learning).
In light of these conclusions, we gathered VRU-involved crash data from the vicinity of schools in Dallas County and Tampa Bay. Based on the crash data analysis and past literature in this area, the researchers have created a comprehensive list of factors that need to be considered for the scenario development in the child-pedestrian training programs. Previous research focused on pedestrians’ behaviors (safe/unsafe), traffic rules, and visual obstructions as critical factors to design this type of intervention program. We propose to add road infrastructure, vehicle speeds and speed limits, pedestrian location, and pedestrians’ actions leading to crash as part of the training program. The differences in crash patterns between the two regions signified the importance of looking at local crash data to inform child-pedestrian training programs and avoid a one-size-fits-all approach.
Although the details are beyond the scope of this work, we also observed, as expected, that the crash patterns from poorer neighborhoods, especially with regards to factors affecting injury severity and crash frequency, differed significantly between the richest and poorest communities even within the same metropolitan areas. This reinforced the need to address the equity issues around child-pedestrian safety. Toward that end, future research should include developing VR-based comprehensive child-pedestrian training programs for elementary school-going children from low-income neighborhoods. We also invite researchers working on child-pedestrian training modules to augment the development of their programs with a simple but systematic analysis of contextual crash data. In conclusion, we want to emphasize that such education programs should not be construed as a replacement for infrastructure/land-use improvements needed to create walkable neighborhoods and improve safety for all road users.
Footnotes
Author Contributions
The authors confirm contribution to the paper as follows: study conception and design: S. Deb, M. Islam, A. Pande; data collection: M. Islam, R. Etminanighasrodashti; analysis and interpretation of results: M. Islam, R. Etminanighasrodashti, S. Deb, A. Pande, A. Rimu; draft manuscript preparation: A. Rimu, S. Deb, A. Pande, M. Islam. 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 a grant (Project #021-06) from the Center for Transportation Equity, Decisions and Dollars (CTEDD) funded by the U.S. Department of Transportation Research and Innovative Technology Administration (OST-R) and housed at The University of Texas at Arlington.
