Content uploaded by Simone Morgagni
Author content
All content in this area was uploaded by Simone Morgagni on Jun 08, 2022
Content may be subject to copyright.
1
Using virtual reality to evaluate the effects of two safety systems at pedestrian track crossings on
human behaviour
Elise GRISON1, Samuel AUPETIT2, Sara ESCAICH2, Simone MORGAGNI1
1SNCF, La Plaine Saint-Denis, France
2Ergocentre, Orléans, France
Corresponding Author: Elise Grison (elise.grison@sncf.fr)
Abstract
Pedestrian track crossings (PTC) allow passengers to cross tracks in small train stations in the absence of a
footbridge or underground passage. In France, despite the addition of a flashing pictogram indicating
approaching trains in the early 70s by SNCF, some people cross the PTC without considering the sign, leading to
few accidents every year. The objective of the research presented in this paper is to test the effectiveness of
two safety systems added to the existing flashing pictogram: a chicane and a ringing alarm. To test the
effectiveness of the two systems in a controlled and safe environment, we simulated a virtual train station
equipped with a PTC. 100 participants were asked to take different routes in various contexts in the train station
(with luggage, wearing headphones, in the presence or not of other people, etc.), some of which led them to
cross the PTC. All pedestrian kinematic measures (speed, head movement, etc.) were recorded as well as gaze
behaviour. The number of dangerous crossings and participants’ understanding of the safety system were also
recorded. The physiological and behavioural data principally demonstrated that the ringing alarm led to safer
behaviour than the chicane or when PTC were only equipped with a flashing pictogram. This study, conducted
using virtual reality various kinds of measures, provided us with reliable evidence to select the more appropriate
safety system to prevent accidents at PTC. It also proves the pertinence of using virtual reality for questions that
cannot be easily and safely studied in ecological contexts.
Keywords: safety behaviour; chicane; ringing alarm
1. Introduction
Since the 1970s, safety at pedestrian track crossings (PTC, see Figure 1) in train stations has been a major subject
of interest for the French National Railway Company (SNCF). Thus, to reduce accidents and make them safer,
PTC have been equipped with 6 flashing red pictograms announcing approaching trains (one on each platform
and four at the middle of the PTC; two in each direction, see Figure 1). Despite this system, a number of accidents
are observed annually, questioning the way in which such a system alerts and explains the danger to pedestrians,
and how safety at PTC could be improved.
Figure 1: Two examples of PTC equipped with flashing pictograms: on the left, activated to signal an
approaching train, and on the right, deactivated in the absence of any approaching trains.
Studies developed over the past two decades on human factors and cognitive psychology behaviour, along with
2
analyses of accidents, help understand why pedestrians might cross the PTC when the pictogram is flashing.
Indeed, moving and safely crossing tracks involve many high-level cognitive processes to (Dommes & Cavallo,
2011; Salmon et al., 2016; SNCF, 2018):
- acquire information (schedules, weather, presence of a train),
- form a global representation of the situation and the crossing track situation,
- evaluate the risks (collision, fall, delay), and
- make a decision (to cross or not).
All these steps necessary for safely crossing tracks might be disrupted by other factors (Aghabayk, Esmailpour,
Jafari & Shiwakoti, 2021). These factors can be related to heuristics to minimize cognitive effort (habits,
delegation to another person or automatism) or distractions (music, smartphone, talking). They can also be
linked to a poor understanding of the risks or of the system itself. Finally, contextual factors may also be involved
(Stefanova et al., 2018), may they be due to what can be called “comfort pressure” (optimisation of path in train
station to avoid detours or bad weather conditions by staying inside the station until the last minute), or
production pressure (not to miss the train).
From a behavioural perspective it appears, therefore, that to improve safety at PTC, actions might be taken to
bring back passengers’ cognitive attention to the actual task at hand – move into the train station and cross the
PTC safely – by facilitating information gathering, for example, or through alerts to minimize inattention. Other
possible levers concern facilitating the understanding of the situation (using a more understandable sign or
alert), and of the decision-making (help passengers to make the right decision by telling them when it is possible
or not to cross, for example).
Taken together, previous findings led to the idea of completing the flashing pictogram with either a ringing alarm
(multimodal alert), or a chicane placed before the PTC “forcing” pedestrians to look for potential approaching
trains. Thus, the main objective of the present study was to evaluate the effectiveness of both security
improvements on PTC and the impact on human behaviour in multiple contexts (distraction, good or poor
visibility, with luggage, etc.).
Due to the fortunately low frequency of accidents, and for safety as well as ethical reasons, the precise
evaluation of the effects of both new systems on safety behaviour were hard to obtain by truly ecological
research. We chose instead to rely on an experiment using virtual reality to simulate a lifelike situation providing
a safe and ethical environment to study human behaviour. It has indeed been demonstrated repeatedly that
behaviour observed in virtual reality can be translated to real life (Burkhardt, 2003), especially in the context of
safety research (Deb et al., 2017). Virtual reality is moreover a better means to record various reliable
physiological and behavioural data (pedestrian kinematics, eye-tracking) that are more sensitive measures of
the effects of each of the safety systems than the number of risky crossings which might not be great enough to
draw valid conclusions.
2. Methodology
2.1 Participants
100 participants took part in the study (average age = 38.95, SE = 15.07; 51 women and 49 men). They were
French citizens, from different parts of the country and socio-economic levels. All participants were users of
French train stations, and approximately 20% of them knew what a PTC was.
2.2 Material
The virtual environment
A virtual environment of a small train station equipped with a PTC was created (see Figure 2) in three different
versions, one for each safety system to be tested (i.e., flashing pictogram, flashing pictogram + ringing alarm and
flashing pictogram + chicane). Participants were immersed in the virtual environment through a virtual reality
3
headset (HTC Vive pro eye) equipped with an eye-tracking system and were able to move and walk in a physical
80m2 environment.
Figure 2: The virtual train station used in the study equipped with a PTC in the “pictogram and chicane”
condition.
Scenarios
To encourage participants to walk naturally into the train station and cross the PTC without revealing the real
purpose of the study (thereby minimizing potential bias), three scenarios in which participants had to review
new information systems in train station, new services or new sanitary equipment were built. In these scenarios,
participants had to cross the PTC 6 or 7 times. The PTC was either active (announcing an approaching train) or
not. Finally, other risk factors were integrated, such as visibility (approaching trains masked by a tunnel),
distraction (listening to music), overload (luggage), and social context (avatars crossing the PTC in a dangerous
situation). A few trials were also done in the context of time pressure.
2.3 Procedure and measures
Each participant was confronted with the three safety systems and the three scenarios. Safety systems and
scenarios were counterbalanced across participants. In total, each participant crossed the PTC 20 times, made
three PTC crossings under time pressure, and was confronted with one of the other 4 risk factors.
Various behavioural and physiological measures were recorded during the crossings, especially gaze behaviour
(fixation number and duration, pupilar diameter), pedestrian kinematics (speed, head orientation on horizontal
and vertical axes), and heart rate. These measures are indicators, among others, of the cognitive or physical load
(Benedetto et al., 2011; Meyer, 1996). They were collected especially in the PTC area (at the entrance, on the
PTC and at the exit) and when safety signages were activated. Finally, the participants’ subjective points of view
were addressed through questionnaires and interviews at the end of the experiment.
3. Results
Given the large quantity of data collected, we decided to present in this section only the major effects of the
study.
3.1 Safety systems main effect
Results revealed that both the ringing alarm and the chicane led to less dangerous PTC crossings, but the
difference wasn’t significant (8.75% vs. 7% for both chicane and ringing alarm conditions). In the additional
questionnaire, participants reported a better understanding of the “message” spread by the ringing alarm telling
them to be attentive of potential danger, than by the chicane.
Figure 3 presents the walking speed at the PTC. We observe a simple effect of the safety system on the walking
speed through the entire crossing of the PTC (F(2, 1009) = 16.07 ; p < .001; η² = 0.03) meaning that participants
4
walk faster in the pictogram and chicane condition compared to others (all p<.05). This effect is significant for
all parts of the PTC crossing (entrance: F(2, 1009) = 43.15 ; p < .001, η² = 0.08; on the PTC : F(2, 1009) = 18.71 ; p
< .001, η² = 0.04; and at the exit : F(2, 1009) = 16.67 ; p < .001, η² = 0.03). These results might be interpreted in
the sense that participants in the pictogram and chicane condition took less time to pay attention to the situation
during the PTC crossing than in other conditions.
Figure 3: Speed at PTC crossing in metre/seconds. From the right, the results are presented for all the PTC,
then at the entrance, on the PTC and at the Exit. [*** = < .001 ; ** = p < .01 ; * = p < .05]
The results concerning head orientation on the vertical axis are presented in Figure 4. We observe a main effect
of the safety system on the general orientation of the head (F(2, 1009) = 45.03 ; p < .001, η² = 0.08) meaning
that participants looked towards the ground more in the pictogram and chicane condition compared to the other
conditions (all p>.05). This effect is significant for all parts of the PTC (entrance: F(2, 1009) = 53.09 ; p < .001, η²
= 0,10 ; on the PTC : F(2, 1009) = 42.84 ; p < .001, η² = 0.08; at the exit : F(2, 1009) = 41.25 ; p < .001, η² = 0.08).
This result might be interpreted in the sense that in the pictogram and chicane condition participants paid less
attention to the general environment around them and more to the immediate environment.
Figure 4: Head orientation on the vertical axis in degrees. From the right, the results are presented for all the
PTC, then at the entrance, on the PTC and at the Exit. [*** = < .001 ; ** = p < .01 ; * = p < .05]
3.2 Risk factors effects
The number of risky crossings is significantly higher in the social context (with other people) than with other
tested factors (see Table 1; χ²(3, N = 1200) = 12,6 ; p< .01, ɸ = .11).
Condition
Social context
Distraction
Overload
Visibility
Risky crossing
39,18 %
24,74 %
18,56 %
17,53 %
Table 1: Percentage of risky crossing for all risk factors condition.
5
Some of the other tested risk factors also had an influence on our participants’ behaviour. For example,
participants looked at the approaching train less when carrying luggage (F(1, 262) = 17.26 ; p < .001, η² = 0.05)
and they crossed the PTC faster when listening to music (t(237) = 2.31 ; p < .05, d = 0.30). However, we did not
observe any interaction with the safety system, meaning that none of the systems tested showed a safety
improvement considering these measures.
When analysing the crossings at PTC in the time pressure condition, we observed a larger horizontal angle in the
ringing alarm condition, meaning that participants oriented their head further towards the approaching train in
the ringing alarm condition than in the other conditions (see Figure 5 (F(2, 1478) = 3.52 ; p < .05, η² = 0.01).
Figure 4: Head orientation on the horizontal axis in degrees.
4. Discussion and conclusion
This study was conducted to test in a safe, ethical, and controlled environment, the effectiveness on human
behaviour of two new safety systems elaborated from the human factors literature. Between the two tested
improvements, it appears that the ringing alarm as a complement to the flashing pictogram led to safer
behaviour and is better understood by participants. Indeed, experimental findings show that in the ringing alarm
condition, participants tended to walk slower when crossing the PTC, and to look higher than in the pictogram
or in the pictogram and chicane conditions. These results show that participants paid more attention to the
situation around them so they might be in better condition to understand it and adopt safer behaviour. The
ringing alarm also had an impact on participants’ behaviour when they were under time pressure as they looked
further in the direction of the approaching train. Moreover, participants’ responses in questionnaires and
interviews reinforce this finding as they declared that the purpose of the ringing alarm was to alert them of a
danger and that it was easier to understand than the chicane.
Taken together, these results lead to the conclusion that adding a second and complementary alert mode to the
visual flashing pictogram, like the ringing alarm, should help to reduce unsafe behaviour from passengers and
consequently reduce the number of accidents. However, in our results, we did not observe any improvement
linked to either the ringing alarm or the chicane in the risk factors conditions, meaning that these two systems
are not enough to prevent accidents when people are distracted by music, overloaded with luggage, with other
people or when the visibility is low. As these situations are significant in the accidental contexts, new systems
and tests are needed, to fully improve safety at PTC Existing literature on human factors provides levers that
could be tested using the same methodology, such as focusing on the understanding of the situation or
facilitating the decision-making process.
From a methodological standpoint, our study is valuable as it collects, for the first time, objective effects of
safety systems at PTC on passengers’ behaviour. Indeed, studies on PTC remain sparse in the literature, while
those done in similar contexts (at level crossings for example, see Freeman et al., 2013 or Stefanova et al., 2015)
were generally based on observations in natural environments or on interviews, and consequently did not
investigate passenger behaviour at this level of detail. Our research proposed to fill this gap in the literature with
a new methodology that allows to observe, quantify and analyse the effects of different safety systems on
6
human behaviour by simplifying and optimizing data collection all the while avoiding the dangers associated
with studying such behaviour in ecological contexts.
In conclusion, the methodology developed here could be adapted to various safety rails contexts and to test
numerous situations (level crossing, intervention of agent of rail, etc.). In future studies based on the
understanding of human factors, the same methodology could be developed to test new methods to ensure
safer crossings at PTC. Furthermore, more than simply validating the effectiveness of one of the two new safety
systems tested, all the measures collected in this study (gaze, kinematics, heart rate) are extremely informative
to conceive new security systems since they also provide information on general pedestrian behaviour when
crossing PTC, and therefore, their use should be systematized in future research conducted for this purpose.
References
[1] Aghabayk, K., Esmailpour, J., Jafari, A., & Shiwakoti, N. (2021). Observational-based study to explore
pedestrian crossing behaviors at signalized and unsignalized crosswalks. Accident Analysis & Prevention,
151, 105990.
[2] Burkhardt, J.-M., Bardy, B., & Lourdeaux, D. (2003). Immersion, réalisme et présence dans la conception et
l’évaluation des environnements virtuels. Psychologie Française, 48, 35–42.
[3] Benedetto, S., Pedrotti, M., Baccino, T., et Re, A. (2011). Driver workload and eye blink duration.
Transportation research, 14, 199-208. DOI : 10.1016/j.trf.2010.12.001
[4] Deb, S., Carruth, D.W., Sween, R., Strawderman, L.J., et Garrison, T.M. (2017). Efficacy of virtual reality in
pedestrian safety research. Applied ergonomics, 65, 449-460.
[5] Dommes, A., & Cavallo, V. (2011). The role of perceptual, cognitive, and motor abilities in street‐crossing
decisions of young and older pedestrians. Ophthalmic and physiological optics, 31(3), 292-301.
[6] Freeman, J., Rakotonirainy, A., Stefanova, T., & McMaster, M. (2013). Understanding pedestrian behaviour
at railway level crossings: Is there a need for more research?. Road & Transport Research: A Journal of
Australian and New Zealand Research and Practice, 22(3), 72-82.
[7] Meyer, J. (1996). La fréquence cardiaque, un indice d'astreinte physique ancien servi par une métrologie
moderne. INRS Documents pour le médecin du travail TL, 20 (8).
[8] Salmon,P. M., Lenne, M. G., Read, G. J. M., Mulvihill, C. M., Cornelissen, M., Walker, G. H., Young, K. L.,
Stevens, N. & N. A. Stanton (2016) More than meets the eye: using cognitive work analysis to identify
design requirements for safer rail level crossing systems. Applied Ergonomics, 53, 312-322
[9] Stefanova, T., Burkhardt, J. M., Filtness, A., Wullems, C., Rakotonirainy, A., & Delhomme, P. (2015).
Systems-based approach to investigate unsafe pedestrian behaviour at level crossings. Accident Analysis &
Prevention, 81, 167-186.
[10] Stefanova, T., Oviedo-Trespalacios, O., Freeman, J., Wullems, C., Rakotonirainy, A., Burkhardt, J. M., &
Delhomme, P. (2018). Contextual factors explaining risk-taking intentions at Australian level crossings.
Safety science, 110, 145-161.
[11] SNCF (2018). Sécurité des piétons aux PN & TVP [Rapport interne]. SNCF Réseau.