An investigation of road crossing in a virtual environment

https://doi.org/10.1016/S0001-4575(02)00081-7Get rights and content

Abstract

The reported study employed a virtual reality (VR) system, using a head mounted display (HMD), to investigate road crossing behavior in children and young adults. Younger children (aged 5–9 years) made the greatest number of unsafe road crossings and the oldest participants (aged >19 years) the fewest. Overall performance was better (fewer unsafe road crossings) in uniform speed than uniform distance trials, consistent with previous research suggesting that pedestrians base road crossing decisions on inter-vehicle distance rather than vehicle speed. Results are discussed in terms of road crossing behavior and the use of VR simulations in the study of pedestrian behavior.

Introduction

Pedestrian accidents are among the most common causes of death and serious injury to young children in the developed world (Ampofo-Boateng and Thomson, 1990, Connelly et al., 1996, Demetre and Gaffin, 1994, Haight and Olson, 1981, Roberts, 1995, Thomson, 1991). Recent pedestrian injuries figures from New Zealand, broken down into 5 years age groups, indicate that children aged 5–9 years accounted for the highest percentage of injuries (16%), followed by those aged 10–14 (13%) and 15–19 years (12%), with no other age group accounting for more than 7% (20–24 years). Pedestrian deaths, however, were highest for 15–19 years old (17%) and for 75–79 years old (14%) with the 5–9 and 10–14 years old age groups each accounting for only 3% of pedestrian fatalities (Land Transport Safety Authority, 2000). Similar findings are reported from the UK and US (Malek et al., 1990) and from Canada (Jonah and Engel, 1983).

The need to teach children how to judge when it is safe to cross the road is widely accepted. Effective skills training should reduce not only childhood injury but also the incidence of accidents, and especially fatalities, amongst the cohort as they age (Ampofo-Boateng and Thomson, 1990, Christoffel et al., 1986, Connelly et al., 1996, Demetre et al., 1993, Grayson, 1981, Malek et al., 1990, Yeaton and Bailey, 1978). Before effective training programs can be developed, however, children’s road crossing behavior and the nature of road crossing errors need to be fully understood. Only then can educational programs targeted to reduce road crossing errors can be developed. Two major deficiencies in young children’s road crossing have been identified (Thomson, 1991): the selection of appropriate locations to safely cross a road and the uptake and use of information about traffic flow which specifies whether an attempted road crossing will be successful. The present research considers only the latter, although further study of the selection of crossing locations amongst children is also advocated.

The ability to safely cross a road is a perceptual-motor skill, one that involves co-ordination between perception of the oncoming traffic and the action of walking across the road. Judgment of whether a gap in the traffic is sufficient to safely cross requires the determination of the time-to-contact of the nearest vehicle with the planned crossing line and the assessment of whether this time-to-contact exceeds the time required to cross the road, taking into account one’s own locomotive speed. The task is not one of perceiving the size of the traffic gap in absolute terms but the size of the gap in terms of time to act. The size of the gap in traffic needed for safe crossing will differ between pedestrians as a function of their locomotive speed. In addition, for any given individual the size of the gap required may vary as a function of environmental (e.g. strong winds, rough road surface) and personal factors (e.g. fatigue, carrying heavy luggage). For each road crossing attempt, the pedestrian must ask themselves anew whether a gap in the traffic affords safe crossing.

Although research consistently shows young children to be worse at making safe road crossing decisions than older children and adults (Demetre et al., 1993, Foot et al., 1999, Pitcairn and Eldmann, 2000), the visual timing skills required to make safe road crossing decisions are not beyond children even as young as 5 years of age (Lee et al., 1984, Vinje, 1981). That is, the errors made by children in road crossing are the consequence of a failure to apply their perceptual-motor skills appropriately rather than of an inability to make the required judgments accurately.

No clear pattern of the nature of road crossing errors has emerged from the previous investigations of road crossing behavior, primarily due to the employment of different research methodologies across studies and the involvement of different age ranges of participants. Naturalistic observation is, of course, the most realistic setting for data collection (Grayson, 1975, Routledge et al., 1976, van der Molen, 1981). It does not, however, allow for experimental control of different aspects of the road crossing situation (e.g. vehicle speed) and the accurate assessment of their contribution to road crossing decision making and safety. The major difficulty with naturalistic settings is the risk of accidents associated with road crossing errors. An additional difficulty, especially for young children, is that road crossing usually occurs alongside adults. Often decisions regarding when and where to cross the road are made by adults with the children simply accompanying them (van der Molen et al., 1983). It is unclear whether young children can themselves make safe road crossing decisions if called upon to do so.

To completely remove the risk of accidents researchers have employed a number of traffic-free approaches to both measuring and teaching road crossing behavior. These techniques include chalkboard methods, videotapes (Pitcairn and Edlmann, 2000), table-top simulations (Thomson et al., 1998), and computer animations (Foot et al., 1999). Although such methods simulate many features of the road crossing situation, they require only passive involvement by participants; the participants are not required to actually cross a road. As described earlier, judgment of safe crossing gaps requires visuo-motor calibration (Lee et al., 1984), evaluating the size of gaps in the traffic relative to one’s own locomotive abilities and time required to safely cross the road. Having participants make judgments about the safety of crossing the road within various gaps in the traffic without requiring them to actually cross the road does not assess this visuo-motor calibration and raises questions about the generalization of such judgments to real road crossing situations.

Some research, notably that conducted by Demetre et al., 1992, Demetre et al., 1993 and Lee et al. (1984), has incorporated actual traffic into road crossing simulations, although safety is maintained as no actual road crossing occurs. Three such methods—the pretend road, the two-step task and the shout task—have been employed. The pretend road is laid out on the pavement alongside a real road, with a barrier beside the actual road. Participants stand beside and cross the pretend road to the barrier when they judge they could safely cross the adjacent real road. The advantage of this method is that the participant is active and their activity must be based on the behavior of real traffic. However, the pedestrian’s viewing perspective and spatial displacement from the traffic differ when standing beside a pretend road and an actual road. These distorting characteristics may affect both the validity of measures taken and the learning acquired from such situations.

The two-step and shout tasks are conceptually similar. They involve participants standing at the roadside kerb (the shout task) or two paces back from the kerb (the two-step task). Participants are asked to indicate, by shouting or by taking two paces onto a pressure pad, respectively when they judge it is safe to cross the road. These techniques again involve real traffic and they overcome the problem of the distorted viewing perspective encountered in the pretend road task. They rely, however, on the pedestrian remembering to allow themselves time to cross the whole road rather than to complete the task at hand (shouting or taking two paces). Roadside simulations are an improvement on classroom techniques but, there is a problem with method validation, assessing whether results from the roadside simulations generalize to crossing actual roads (Demetre et al., 1993).

The present research employed a virtual reality (VR) system, using a head mounted display (HMD), in an attempt to provide participants with a realistic road crossing environment that overcomes many of the difficulties encountered by previously employed methodologies. Within the virtual environment participants can stand beside and cross a road on which virtual traffic is moving. As the participant moves, the view in the HMD changes in a manner consistent with the virtual road being a physical reality and the pedestrian actually walking across it. The visual experience of crossing the road in the virtual environment is the same as the visual experience of crossing an actual road with the same properties, within the limitations (e.g. resolution) of the VR system. Accordingly, the distorting characteristics present in road-side simulations such as the pretend road scenario are eliminated. In the virtual environment, participants are able to see when they have made a mistake and tried to cross the road when it was not safe to do so. They can make corrective action to avoid being hit (e.g. breaking into a jog or stopping) but, if appropriate, participants can be also “hit” by the virtual vehicles. The visual experience of the vehicle looming toward the moment of contact is the same as in an actual collision, but without the physical danger of such a collision. A first aim of the present research was to investigate the feasibility of using a VR system with HMD to investigate road crossing behavior.

Traffic flows, vehicle speeds and inter-vehicle distances can easily be manipulated within the virtual environment and measures of performance taken. In the present study, the speed and inter-vehicle distance of the approaching traffic were varied. In order to correctly assess whether a given gap in the traffic affords safe crossing, pedestrians must determine the time-to-contact of the approaching vehicle. Time-to-contact is a function of the speed and distance to be covered by the approaching vehicle.1Connelly et al., 1996, Connelly et al., 1998 have argued that children base their road crossing decisions on distance information, failing to take into account the speed of the oncoming traffic and, hence, misjudging time-to-contact. They showed children’s distance thresholds for non-safe crossing to remain constant regardless of vehicle approach speeds. Connelly et al.’s participants were a maximum of 12 years of age so it is unclear whether errors made by adults are also a function of an over-reliance on distance rather than speed information. This research also used an unusual measure of road crossing ability, having participants judge when the gap between traffic is no longer large enough for safe crossing. It is unlikely that pedestrians would normally make such a judgment, since what is important in safe road crossing is identifying whether a given gap is large enough, not the point at which it ceases to be large enough.

The present research investigated the impact of speed and inter-vehicle distance on the road crossing ability of both children and young adults in the same, realistic, experimental set-up. Vehicle speed and inter-vehicle distance were systematically varied across experimental trials to assess the extent to which pedestrians of different ages and sexes relied on speed and/or distance information in making road crossing decisions. A number of both timing and outcome measures (based on Demetre et al., 1993; for definitions see Section 2) were employed to assess road crossing behavior.

Section snippets

Participants

There were 24 participants in total, 6 in each of the following age groups: 5–9, 10–14, 15–19 and >19 years. There were three females and three males in each age group. The youngest participant was 5 years old and the oldest participant 30 years old. Participants were recruited via informal social associations.

Virtual environment

The virtual environment was generated by a 800 MHz Pentium III PC with 128MB of RAM and a 32MB Riva TNT2 3D graphics accelerator card. The virtual environment was viewed through a Virtual

Results

Twenty-four participants each completed 12 trials giving a total of 288 trials. Each trial was classified as either a safe or an unsafe road crossing.

Discussion

The reported research represents a first attempt to investigate road crossing behavior using a virtual environment and a head mounted display. The results are discussed in terms of both patterns of road crossing behavior and the use of virtual reality technology.

Acknowledgements

This research was supported by a Child Accident Prevention Foundation of New Zealand Summer Scholarship awarded to the first author.

References (31)

  • J.D Demetre et al.

    The salience of occluding vehicles to child pedestrians

    Brit. J. Educ. Psychol.

    (1994)
  • J.D Demetre et al.

    Errors in young children’s decisions about traffic gaps: experiments with roadside simulations

    Brit. J. Psychol.

    (1992)
  • J.D Demetre et al.

    Young children’s learning on road crossing simulations

    Brit. J. Educ. Psychol.

    (1993)
  • H Foot et al.

    Recognizing the hazards

    Psychologist

    (1999)
  • Grayson, G.B., 1975. Observation of pedestrian behavior at four sites. TRRL Rep....
  • Cited by (143)

    • Effects of a frontal brake light on (automated) vehicles on children's willingness to cross the road

      2023, Transportation Research Part F: Traffic Psychology and Behaviour
    View all citing articles on Scopus
    View full text