A Multi-role, Multi-user, Multi-technology Virtual Reality-based

Road Tunnel Fire Simulator for Training Purposes

Davide Calandra

1 a

, Filippo Gabriele Prattic

1 b

, Massimo Migliorini

, Vittorio Verda

3 c

and

Fabrizio Lamberti

1 d

Politecnico di Torino, Dipartimento di Automatica e Informatica, Corso Duca degli Abruzzi 24, 10129 Turin, Italy

Fondazione LINKS, Via Pier Carlo Boggio, 61, 10138 Turin, Italy

Politecnico di Torino, Dipartimento Energia, Corso Duca degli Abruzzi 24, 10129 Turin, Italy

Keywords:

Virtual Reality (VR), Serious Game, Emergency Training, Road Tunnel Fire, Fire Dynamics Simulator (FDS).

Abstract:

The simulation of ﬁre emergency scenarios in Virtual Reality (VR) for training purposes has a large number of

advantages. A key beneﬁt is the possibility to minimize the associated risks if compared with live ﬁre training.

Fire in road tunnels is among the most complex and hazardous events to be dealt with in this context, since

the outcome of the incident depends on the actions of both the emergency operators and the involved civilians.

This paper presents a VR-based road tunnel ﬁre simulator designed to support multiple roles (ﬁreﬁghters,

as well as occupants of both light and heavy vehicles) which can be played by multiple networked users

leveraging a broad set of technologies and devices. The simulation tool – named Fr

ejusVR – is developed

as a serious game for training purposes, and includes functionalities to assess the users’ actions which make

it suited to a broad range of applications encompassing not only the training of operators, but also the study

of human behavior during emergencies and the communication of safety prescriptions to tunnel users. The

simulation uses data from a Fire Dynamics Simulator (FDS) to support a realistic visualization of smoke

in VR, whereas for reproducing the spreading of ﬁre a non-physically accurate, yet credible, simulation is

adopted in order to guarantee interactivity.

1 INTRODUCTION

The use of Virtual Reality (VR) technology for train-

ing purposes is deﬁnitely not new. VR allows the

creation of wide and complex Virtual Environments

(VEs), enabling training situations that would be po-

tentially hazardous or very resource-intensive if recre-

ated for real (Engelbrecht et al., 2019).

Numerous works investigated the contribution of

VR in the creation of effective emergency scenarios

for training purposes (Andrade et al., 2018; Lu et al.,

2020). Fire simulation is one of the most explored use

cases (Fathima S J and Aroma, 2019; Corelli et al.,

2020), due to the the number of possible hazards as-

sociated with live-ﬁre training, in which ﬁreﬁghters

are taught how to safely ﬁght ﬁres in a controlled and

https://orcid.org/0000-0003-0449-5752

https://orcid.org/0000-0001-7606-8552

https://orcid.org/0000-0002-4069-2713

https://orcid.org/0000-0001-7703-1372

supervised setting (Engelbrecht et al., 2019). Fire

simulation scenarios may have different aims. Firstly,

they can be used as training tools for professionals;

ﬁreﬁghters, in particular, could be prepared – both

physically and mentally – for real life incidents (En-

gelbrecht et al., 2019). Civilians can be trained too,

by showing them the correct safety behaviours. Sec-

ondly, they can be used for human behaviour investi-

gations. For example, VR simulations can be used to

study the reaction of civilians during the emergency

(Kinateder et al., 2014b). Furthermore, they can be

used for designing and studying the effectiveness of

safety measures in a given scenario.

A particular class of ﬁres, namely tunnel ﬁres,

are the most critical event type for users’ safety

(Ntzeremes and Kirytopoulos, 2019). When a ﬁre

develops inside a conﬁned space such as a road or

railway tunnel and involves heavy vehicles such as

trucks, it can rapidly grow up and become completely

uncontrollable. Hence, the prompt response of the

ﬁre ofﬁcers is essential for evacuating involved peo-

Calandra, D., Pratticò, F., Migliorini, M., Verda, V. and Lamberti, F.

A Multi-role, Multi-user, Multi-technology Virtual Reality-based Road Tunnel Fire Simulator for Training Purposes.

DOI: 10.5220/0010319400960105

In Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2021) - Volume 1: GRAPP, pages

96-105

ISBN: 978-989-758-488-6

ple and then attacking the triggering ﬁre. At the same

time, the conformity of civilian users’ behaviors to

the emergency prescriptions plays a fundamental role

in the incident evolution, since evacuation behaviours

also have a social inﬂuence (Kinateder et al., 2014a)

and can dramatically affect emergency operations.

An effective way to represent and manage the

complexity of such situations is opting for a multi-

user experience, e.g., by introducing virtual charac-

ters in the simulation (Engelbrecht et al., 2019). Al-

ternatively, multi-user networked VEs provide a wide

range of training and operation support possibilities.

Multiple human users can visualize and interact with

a shared scenario, practicing teamwork at a geograph-

ical scale (Louka and Balducelli, 2001). A weak-

ness of this type of VR experiences is the lack of

multi-user ﬁdelity, if compared with solitary VR sim-

ulations (Engelbrecht et al., 2019). For example, in

case of low visibility due to the presence of smoke,

a ﬁreﬁghting crew would hugely rely on the sense of

touch, which cannot be easily recreated with the com-

monly available haptic devices. Furthermore, if the

visual representation of the other human users does

not reach an proper level of credibility, factors such as

immersion and perceived realism would be severely

affected too. However, the evolution of commercial

VR technology constantly leads to new systems, de-

vices and functionalities. If correctly integrated, these

elements may help to mitigate the mentioned issue.

The aim of this paper is to present a tool for

multi-role, multi-user, and multi-technology VR ﬁre

simulation set in a road tunnel during a ﬁre event.

Multi-role refers to the possibility to impersonate ei-

ther an emergency operator (i.e., a ﬁreﬁghter) or a

non-professional user, being it a frequent user (truck

driver) or an infrequent visitor (a civilian). Each

of these roles has its own procedures to follow in

an emergency, which are subjected to an assessment

through the devised tool. This feature gives the sim-

ulation a threefold purpose: that of a training tool for

ﬁreﬁghters, of a training tool to teach civilians safety

behaviours, and of an investigation tool to study the

civilians’ behaviour during the emergency. Multi-user

refers to the possibility, for different users in different

places, to play the available roles all together and at

the same time in a shared experience. As a matter

of example, a ﬁreﬁghter trainee may play the oper-

ator role, while other users may be playing as civil-

ians. Finally, multi-technology refers to the charac-

teristic of the scenario of being designed to support

a broad variety of VR device and system conﬁgura-

tions, either common (desktop VR, HMD and con-

trollers, etc.) or less widespread (like locomotion

treadmills, leg sensors and motion capture suits). The

Figure 1: Smoke simulation displayed inside the VR sce-

nario (top) and inside Smokeview, a tool for FDS visualiza-

tion (bottom).

goal is to increase the effectiveness of the VR experi-

ence when beneﬁcial hardware is available, while pre-

serving the compatibility with common VR systems

to prevent a strict technology barrier. In order to ac-

curately represent the physical phenomena of gasses

spreading inside the tunnel and further enhance ex-

perience ﬁdelity, data of a Fire Dynamics Simulator

(FDS) were used to drive the visual representation of

the smoke evolution (Figure 1).

The scenario, labeled Fr

ejusVR, was developed in

collaboration with SITAF S.p.A.

, the Authority of

the Fr

ejus tunnel. Falling within the context of the

PITEM RISK FOR

project, the serious game accu-

rately reproduces a section of the real road tunnel,

around the border between Italy and France, since the

goal is to support the interoperability of procedures

for ﬁre management across the two countries.

2 BACKGROUND

To date, the use of VR to create simulations of emer-

gency situations has been extensively investigated

(Louka and Balducelli, 2001; Kinateder et al., 2014b;

Lovreglio, 2020). The same can be said for simula-

tions involving ﬁre events, being them intended either

for training or human behaviour analysis.

In (St Julien and Shaw, 2003), for instance, a ﬁre-

ﬁghter command training VE in non-immersive VR

was presented. The system, intended to instruct com-

manding ofﬁcer trainees, allows the user to navigate

the VE, visualize a building on ﬁre from any angle,

and command a squad of computer-controlled virtual

ﬁreﬁghters. The correct sequence of commands leads

https://pages.nist.gov/fds-smv/

https://www.sitaf.it/

https://www.pitem-risk.eu/progetti/risk-for

A Multi-role, Multi-user, Multi-technology Virtual Reality-based Road Tunnel Fire Simulator for Training Purposes

to the extinguishing of the ﬁre with the lowest amount

of danger to the squad. The system combines the rep-

resentation of animated ﬁreﬁghters with the integra-

tion of a reasonable – but not physically accurate –

ﬁre and smoke simulation.

In (Ren et al., 2008), a VR system simulating

emergency evacuations during ﬁre was introduced. To

increase the level of ﬁdelity of ﬁre and smoke, these

phenomena were modeled by taking advantage of nu-

merical ﬁre simulations. Then, particle systems were

used to visually represent them in the interactive VE.

The aim was to show which evacuation techniques

could be effective for building safety evacuation.

In (Cha et al., 2012), a real-time processing frame-

work to develop ﬁre training simulators based on FDS

data was proposed, along with a set of data conversion

techniques for the purpose. In this way, many useful

physical quantities like toxic gases, heat, smoke and

ﬂames may be used both to visualize the results of the

simulation and to measure safety achievements levels

in the areas where the trainees pass through. The pro-

posed framework was used to implement a single-user

VR road tunnel ﬁre simulator supporting simple ﬁre-

ﬁghting actions such as evacuation and rescue. Inside

the scenario, a safety level-based visualization map-

ping is used to display risk factors that are invisible

but closely related with the safety of the trainees, such

as toxic gasses and temperature distribution.

In (Xu et al., 2014), the FDS data were again used

to drive the visual representation of a ﬁre event. In

this case, the purpose was to model and volumetri-

cally visualize the smoke spreading inside buildings,

with the aim to allow trainees to experience a realis-

tic, but not threatening, ﬁre scenario. An assessment

model of the smoke hazard was also developed, in or-

der to evaluate the safety of each different path for

evacuation or rescue in virtual training.

In (Kinateder et al., 2013), a tunnel ﬁre simulation

for ﬁre safety behaviour training was presented. The

study, introducing the use of an immersive VR system

(i.e., a CAVE), investigated the effect of information

provided with or without VR on self-evacuation be-

haviors during the depicted emergency.

Interestingly, as shown in (Lovreglio, 2020), de-

spite the tunnel ﬁre scenario was largely explored in

the past, the number of works that focused on training

is very scarce, and mostly limited to ﬁre safety be-

haviour training. A larger number of works pertained

human behaviour investigations. However, most of

them did not rely on immersive VR, or did not take

advantage of the more recent HMD-based immersive

VR systems. Finally, none of them was designed as

multi-user experience.

3 SYSTEM DESIGN

In this section, the design of the proposed tool is re-

ported, discussing the rationale for choices made.

3.1 Use Case

As mentioned in Section 1, this work has been de-

veloped in collaboration with the Fr

ejus tunnel Au-

thority. A preliminary investigation was performed in

order to identify the most representative use case for

a road tunnel ﬁre scenario. A smouldering ﬁre de-

veloping from a light vehicle can be easily managed

by the vehicle occupant, using one of the many ex-

tinguishers available along the tunnel. The same does

not apply when ﬁre originates from a heavy vehicle.

Once the ﬁre spreads to a ﬂammable load, the situa-

tion can dramatically change, as demonstrated by all

the severe ﬁre events happened to date in road tun-

nels around the world, like the Mont Blanc in 1999

(BBC News, 1999), the Saint Gotthard in 2001 (BBC

News, 2001) and the Fr

ejus itself in 2005 (BBC News,

2005). The tunnel depicted in the simulation is de-

tailedly inspired to the real Fr

ejus tunnel, in particu-

lar to the one kilometer-long section over the Italy-

France border; the represented ﬁre event shares a

number of similarities with the 2005 incident, such

as the location and the initial conditions. The section

layout and its apparatuses are shown in Figure 2.

At the beginning of the simulation, a car is trav-

elling from Italy to France with two occupants on-

board. At the same time, on the opposite lane, a truck

is driving from France towards Italy. Suddenly, the

engine truck catches ﬁre and the heavy vehicle stops.

Inside the car, which is traveling at 70 km/h, the radio

reproduces status messages about the tunnel. Shortly

before the car occupants can gain visual contact on

the vehicle on ﬁre, an emergency message is trans-

mitted by the radio. At the same time, a unit of two

ﬁreﬁghters (stationing in a ﬁxed rescue place inside

the tunnel) gets on a ﬁre truck and starts driving to-

wards the incident; at present, a vehicle belonging to

the Italian National Fire Corps has been considered,

though for the purpose of the said interoperability, a

vehicle managed by Sapeurs-pompiers de la Savoie

(SDIS 73) could be easily integrated. From this point

on, actions of all the involved participants will affect

the outcome of the whole simulation.

3.2 Roles and Procedures

In the above contexts, ﬁve roles can be identiﬁed. The

car driver, the car passenger, the truck driver, and the

unit of two ﬁreﬁghters. The car and truck occupants,

GRAPP 2021 - 16th International Conference on Computer Graphics Theory and Applications

Figure 2: Layout of the simulated tunnel section.

in case of ﬁre, have to follow the information printed

on a double-sided safety brochure that is ordinarily

handed over at the tunnel entrance. The green side of

the brochure refers to normal conditions, whereas the

red side (emergency) invites the user to:

• stop the vehicle, keeping a minimum distance of

100m from the burning vehicle (distances can be

estimated by using the tracking lights inside the

tunnel as reference: a blue pair is placed every

150m, interspersed with yellow pairs every 20m);

• turn off the engine and enable the hazard lights;

• reach the nearest shelter following the indications;

• if possible, ask for help by using the SOS tele-

phone in a niche or pressing a SOS button;

• if willing to do that, use the extinguishers avail-

able in niches and shelters.

The truck driver is provided with a ﬁre extinguisher

under the truck driving seat.

In the meantime, the ﬁreﬁghter unit has to follow

its own procedure. In case of ﬁre and smoke hazard,

the ﬁreﬁghting unit stationing inside the closest ﬁxed

rescue place is the ﬁrst to reach the incident site. Once

the driver of the ﬁre truck reaches the distance of 50m

from the ﬁre, the two unit components get off the ve-

hicle and have to perform the following steps:

• verify the safety of the area, excluding the pres-

ence of threats;

• rescue possible victims, helping whoever is un-

able to move; in case of seriously injured victims,

bring them out of the tunnel; help the other civil-

ians to reach a safe place (i.e., a shelter or the ﬁre

truck);

• start attacking the ﬁre once all the civilians are

safe; the driver connects the back of the ﬁre truck

with the closest hydrant by means of a ﬁrst hose;

the other component grabs a second hose, con-

nected to the ﬁre truck and equipped with an han-

dline nozzle, gets close to the truck and attacks the

base of the ﬁre with a mixture of water and foam;

• evacuate the area if the ﬁre is not extinguished in

the ﬁrst few minutes.

In the simulation, the car passenger is intentionally

made unable to move, stuck in the car due to a blocked

seat-belt. The ﬁreﬁghting unit will have to free the

civilian by cutting the seat-belt with ad-hoc scissors.

Although each of the above roles could be, in prin-

ciple, played by a human, it has been noted that some

of them would be characterized by a very limited set

of actions to perform. Thus, the car passenger and the

second ﬁreﬁghter (the driver) were designed as Non-

Player Characters (NPC) only.

The assessment of each role had to take into ac-

count the set of available actions for each role, which

can be mandatory or optional.

3.3 Fire and Smoke

Fire and smoke were managed separately. On the

one hand, ﬁre had to be handled in real-time, in or-

der to guarantee interactivity with the ﬁreﬁghting unit.

Thus, an accurate physical modeling was excluded.

Starting from the truck engine, the ﬁre gradually

has to spread to the other parts of the truck, and, in

few minutes, to the whole vehicle. Before reaching

this point, the ﬁreﬁghting unit should be able to prop-

erly attach it, by aiming the stream of water at its

base. The civilian use of ﬁre extinguishers, as well

A Multi-role, Multi-user, Multi-technology Virtual Reality-based Road Tunnel Fire Simulator for Training Purposes

as the incorrect use of the ﬁre hose (e.g., by aiming at

the top of the ﬁre) should not be considered enough

for a complete extinguishing, but useful to slow down

the ﬁre evolution. Heat was also taken into account,

not only for the spreading phenomenon, but also as a

threat to the tunnel occupants.

Smoke, on the other hand, was designed with a

completely different approach. Initial tests with com-

mon real-time particle systems showed a number of

critical problems. Firstly, they required a dramatically

high amount of particles to completely ﬁll the tunnel.

Secondly, the resulting visual effect and its evolution

over time resulted to be not accurate enough for a ﬁre-

ﬁghting training scenario. Hence, it was decided to

drop the real-time constraint, and opt for a physically

accurate smoke simulation, that will be detailed in the

following section. Along with the effect on the users’

visibility, the smoke had to be a source of intoxication

for tunnel occupants without breathing apparatuses,

eventually leading to a failure for the given user.

4 IMPLEMENTATION

This section illustrates the system implementation,

along with the set of supported VR technologies. The

architecture of the system is depicted in Figure 3.

4.1 Devices and Technologies

The scenario was developed in Unity as a SteamVR

application. This choice guaranteed compatibility

with a very large set of commercial VR systems, such

as the HTC Vive, Samsung Odyssey and Oculus Rift,

etc. However, in case of no HMD detected, the appli-

cation falls back to non-immersive VR.

On the one side, for moving around in the sce-

nario in non-immersive VR, a traditional mouse and

keyboard approach was adopted. Locomotion in VR,

on the other side, required a preliminary investigation.

“Magic” techniques like teleporting were not consid-

ered, in order to guarantee a high level of immersion

(Cannav

o et al., 2020), whereas joystick-based move-

ments showed to induce a signiﬁcant cyber-sickness.

Hence, it was decided to focus on more natural loco-

motion paradigms.

The ﬁrst technique to be integrated was be the

Arm-Swinging (AS), in which the user holds a but-

ton (the grip) and swings the hand controllers back

and forth to generate virtual motion. In addition,

the support for two locomotion treadmills was added,

namely, the Cyberith Virtualizer (CV) (Cakmak and

Hager, 2014) and the KATVR KatWalk

(KW). With

these devices, the user slides his or her feet on a

slippery surface to move in the VE. Finally, with

the HTC Vive, two additional tracking elements (the

Vive Trackers) can be attached to the user’s legs to

enable the Walking-In-Place (WIP) locomotion, in

which motion is generated by marching in place. All

these techniques were also evaluated with a prelimi-

nary single-user and single-role (civilian) version of

the proposed system, considering aspects like usabil-

ity, sense of immersion, presence, and sickness. In

particular, in (Calandra et al., 2018), two techniques

were considered: AS and CV, showing the superiority

of the former. Then, in (Calandra et al., 2019), a fur-

ther evaluation involved AS, KW and WIP. Results

showed that AS outperformed both the other tech-

niques, and the WIP was positioned in the middle.

The support for the Xsens MVN

motion capture

suit was also integrated to enable full-body tracking

and completely control the virtual representation of

the tracked user on the other remote users’ clients.

Finally, the combined use of a backpack PC and an

inside-out VR system, such as the Samsung Odyssey,

was exploited to implement a free-walk locomotion

technique, untied from any external sensor, tracking

area or wire. The only drawback of this technique is

the need of a large walking area, comparable with that

of the simulated tunnel section.

The support for CAVEs in combination with either

affordable (Celozzi et al., 2010) or high-end 6DOF

tracking techniques (like, e.g., motion capture sys-

tems by OptiTrack

or Vicon

) was initially consid-

ered. However, due to the substantially different inter-

action paradigms required in these environments w.r.t.

those used by the already selected technologies, this

option was later discarded.

The combined use of some of the mentioned con-

ﬁgurations within a multi-user session can be seen in

Figure 4.

4.2 Multi-user

For the networking, a slightly modiﬁed version of the

Unity legacy high-level network API (U-NET) was

used. A client-server approach was adopted, in which

the server could be either hosted on a client or on a

dedicated machine. In the second case, the server ap-

plication provides a “control room” mode, in which

the simulation can be visualized through a number of

https://www.kat-vr.com/products/kat-walk-vr-

treadmill

https://www.xsens.com/motion-capture

https://optitrack.com/

https://www.vicon.com/

GRAPP 2021 - 16th International Conference on Computer Graphics Theory and Applications

100

Figure 3: Architecture of the Fr

ejusVR system.

ﬁxed security cameras placed along the tunnel. In

this mode, a “director” user can modify the simula-

tion, e.g., by speeding up or slowing down the ﬁre

evolution, turning off the tunnel lights, etc. Commu-

nications among users in the tunnel is guaranteed by

a VOIP channel with 3D positional audio playback.

For the management of NPCs, the Unity built-in

API (NavMesh) was used. A set of animations for

each role was recorded with an OptiTrack system (in

particular, for the ﬁreﬁghters), in order to maintain

a high level of ﬁdelity when the relative character is

computer-controlled. Other animations which did not

require a particularly high ﬁdelity level were created

in Blender

, using also the add-on presented in (Can-

nav

o et al., 2019) for speeding up the process.

4.3 Interaction with the VE

The interaction with objects in the VE is managed in a

standard way. It is based on contact between the inter-

active object and the VR hand controller. The trigger

button is used to initiate the interaction; in case of ob-

jects that can be grabbed, the user does not need to

keep it pressed (a further click will drop the object).

Extinguishers, once grabbed, require a further inter-

action with the other hand to take the safety off; then,

the content can be sprayed by pressing a second but-

ton (the pad). The handline nozzle of the ﬁre hose

has to be managed with two hands, in order to simu-

late the real-life mode of operation (one hand to grab

it, the other one to regulate the jet). Interaction with

large and small doors will trigger an opening anima-

https://www.blender.org/

tion; for the shelters, doors will automatically close

shortly thereafter. SOS telephones inside niches can

be used to simulate a call with a safety operator;

bringing the handset to the ear is enough to consider

the relative action (help requested) as correctly per-

formed. SOS buttons along the tunnel do not provide

any feedback to the user because, as in the real setting,

they just signal an issue at a given location.

At the beginning of the VR experience, users are

let into a play-test room resembling the emergency

shelter they will have to reach later during the simula-

tion. In this VE they can freely familiarize with the lo-

comotion technique, to visualize the safety brochure,

and to try out some of the available interactions (in

particular, the extinguisher and the SOS button).

Then, the simulation starts, and all the civilian

users have their locomotion technique disabled, as

they are travelling inside a vehicle in opposite di-

rections. Vehicles autonomously accelerate up to

60 km/h, and maintain this speed unless the relative

driver presses any controller button. This interaction

will trigger the brakes, which is the only available ac-

tion when vehicles are travelling. Shortly, the heavy

vehicle will rapidly lose speed as the engine goes on

ﬁre. The car driver will instead travel inside the tun-

nel for few moments, until the truck on ﬁre becomes

visible. From this moment on, brakes will need to

be activated in order to stop the vehicle at least 100m

from the ﬁre with the help of the blue tracking lights.

If the brakes are not used, the vehicle will stop au-

tonomously close to the truck. In order to increase

the sense of immersion, during this initial part of the

experience inside vehicles, users may be seated on

a chair. However, the chair should be promptly re-

A Multi-role, Multi-user, Multi-technology Virtual Reality-based Road Tunnel Fire Simulator for Training Purposes

101

Figure 4: Combined use of different devices and technologies in a multi-user Fr

ejusVR session. On the left: User 1, playing

as a ﬁreﬁghter, equipped with an inside-out VR HMD, a backpack PC and a motion capture suit. On the top right: Point-Of-

View (POV) of User 1, viewing User 2 and User 3. On the right: User 2, playing as a car driver, equipped with an outside-in

VR HMD. In the center: POV of User 2, viewing User 1. User 3 playing as a truck driver from a desktop PC (not shown).

moved from the play-area as soon as user stands up

after exiting the vehicle. In case of treadmills that

support the seated position, the chair is not needed.

While in the vehicles, civilians will be allowed to

interact with the cockpit (to turn off the engine, or to

enable the hazard lights) and the door (to leave them).

After having interacted with the door handle, they will

found themselves in the tunnel, outside the vehicle.

From now on, the locomotion techniques will be en-

abled, and users will be able to freely walk around,

read the safety indications inside the tunnel, and fol-

low the safety brochure. The brochure can always be

displayed by using a controller button (the pad) if no

objects are being grabbed with the relative hand.

The human-controlled ﬁreﬁghter, on the other

hand, is kept inside the play-test room until the tunnel

goes on emergency state. Then, the ﬁre truck enters

the tunnel and stops at 50m from the truck on ﬁre.

At this point, an emergency siren is activated, and the

ﬁreﬁghter is instantly moved near the ﬁre truck along

with the NPC partner. Among the available interac-

tions, the ﬁreﬁghter can open the car door, get the

scissors from the partner, use them to cut the seat-

belt and free the civilian, grab the ﬁre hose, put it on

ground closer to the ﬁre, grab the handline nozzle,

and use it to spray the vehicle on ﬁre. Most of the

discussed interactions are illustrated in Figure 5.

4.4 Fire and Smoke

As said, ﬁre was modeled with a non-physically ac-

curate, but credible, logic. The truck was split in

various parts (engine, doors, wheels, tarpaulin, etc.),

and each of them can burn differently and separately,

controlled by parameters like ignition temperature, re-

maining fuel and wetness.

Extinguishing elements (e.g., the water-foam mix-

ture ejected from the ﬁre hose) were modeled so that

the contact of the generated particles with a burning

object causes a temperature loss. If the temperature

goes below the ignition threshold, the ﬁre is extin-

guished, and the part becomes wet. Further exposure

to heat will eventually dry the part, bringing it back to

the initial state, though with a lower amount of fuel.

To manage the effect of ﬁre on the human body,

it was necessary to introduce a gamiﬁcation element.

Each character is provided with a health value at

the beginning of the simulation, and any exposure to

harmful elements has the effect of lowering this value.

A direct contact with ﬂames or extreme heat, for in-

stance, will cause a moderate damage, and will be sig-

naled by playing a scream sound from the given char-

acter. When the health value reaches zero, the simu-

lation is terminated for the user (who is considered as

dead for purpose of the assessment).

Regarding the smoke, a box roughly approximat-

ing the tunnel shape was used as domain for the FDS

GRAPP 2021 - 16th International Conference on Computer Graphics Theory and Applications

102

Figure 5: Road tunnel ﬁre (ﬁrst picture) and some of the available interactions: hazard lights, safety brochure, SOS button,

extinguisher, vehicle engine, SOS telephone, SOS shelter doors, ﬁre hose, NPC ﬁreﬁghter, scissors, and handline nozzle.

simulation. The ﬂuid ﬂow, provided in Eulerian rep-

resentation, was a matrix of 4000×17×831×17 (time

[s], width [m], length [m], height [m]) elements in-

dicating an intensity value of smoke (in the range 0–

255) in a given space coordinate for each time step.

Data were sub-sampled with a 3×3×3 linear ﬁlter

to obtain a 4000×6×300×12 data structure, in or-

der to lighten the computational load on the overall

VR simulation. Then, the resulting grid was used to

generate a set of billboard elements with their opacity

proportional to the intensity value of each grid point,

which are updated over time as the VR simulation

progresses. The grid’s higher vertical resolution was

adopted to better represent the smoke layering phe-

nomenon.

Smoke intoxication was managed separately, by

periodically testing the presence of users inside the

gas cloud, and triggering the intoxication effect if the

test is positive. This condition, signaled by a cough-

ing sound, causes a slow but continuous health reduc-

tion as the user remains inside the smoke-ﬁlled area.

4.5 User Assessment

The assessment module was designed to provide a

summary of the user’s actions with respect to the se-

lected role. In case of civilians, the system displays

whether or not the user:

• maintained the safe distance with the car;

• turned off the engine;

• enabled the hazard lights;

• pressed the SOS button;

• used the SOS telephone;

• tried to use the extinguisher on the ﬁre;

• survived by reaching the shelter.

Along with the timestamp for each action. For the

ﬁreﬁghter, the system shows whether:

• the presence of civilians was checked;

• the operator kept a safe distance from the ﬁre;

• the ﬁre was attacked correctly (from the base);

• the scissors were used correctly (far enough from

the civilian’s body to avoid causing injuries);

A Multi-role, Multi-user, Multi-technology Virtual Reality-based Road Tunnel Fire Simulator for Training Purposes

103

• the procedure was completed (the ﬁre was extin-

guished).

5 CONCLUSIONS

In this paper, a road tunnel ﬁre simulator in VR with

multi-role and multi-user capacity is proposed. The

system is designed to take advantage of a wide set of

VR technologies and devices with the aim to maxi-

mize the ﬁdelity of the experience when better hard-

ware conﬁgurations are available, without introducing

technological barriers. The multi-role and multi-user

capabilities are complemented with a NPC logic to

easily scale down to smaller user counts.

The devised assessment features make the simu-

lator suitable for a number of different uses, from the

training and evaluation of ﬁreﬁghting operators, to the

study of the behaviours of individuals during the de-

picted emergency, as well as to communicate the se-

curity prescriptions to civilian users, being them pro-

fessionals (e.g., truck drivers) or not. Furthermore,

the use of gamiﬁcation elements such as the concept

of survival/victory opposed to wrong behaviour/loss

are expected to lead to a positive attitude of the users

towards the tool. A serious game, in fact, can be per-

ceived as a challenge, that could push users into im-

proving themselves by learning from their mistakes.

Future works will be initially devoted to the val-

idation of the presented tool, by involving a relevant

number of tunnel ﬁreﬁghting operators and by col-

lecting their subjective feedback. Afterwards, the tool

will be used to investigate the effectiveness of the

training experience when compared with traditional

training methods, as well as to study the contribution

of each conﬁguration (in terms of devices and tech-

nologies) with respect to the training purposes.

Moreover, the scenario will be also exploited to

evaluate the effectiveness of the safety equipment re-

produced in the VE, with the aim to constantly im-

prove the overall tunnel safety. In fact, since civil-

ians do not have any strict or mandatory objective in

the simulation (except to survive), it would be possi-

ble to analyze the different trends regarding, e.g., the

adherence to particular prescriptions, the use of spe-

ciﬁc equipment, and the order in which actions are

performed. Still regarding non-professional roles, the

effects of the VR-based training on long-term knowl-

edge retention will be investigated by comparing it

with the sole use of the safety brochure.

Furthermore, the multi-role feature could be ex-

ploited to study the beneﬁts of role rotation in two

different ways. Firstly, by putting users with a given

role in the perspective of all the other roles, with the

aim to increase the awareness of everyone’s goals, im-

prove individual behaviors and, consequently, the ef-

fectiveness of the shared experience; for instance, a

civilian could be more prone to rapidly reach the SOS

shelter when knowing that the ﬁreﬁghters cannot be-

gin the extinguishing if there are still users scattered

inside the tunnel. Secondly, a rotation within sim-

ilar roles could be studied with the aim to increase

preparation for any contingency; in this case, further

playable roles would have to be developed (e.g., for

the operators only one ﬁreﬁghter role out of the two

depicted ones can presently be controlled by human

players).

Among the future technical developments, there

is certainly some room for improvement for both the

ﬁre and smoke representations. Regarding the for-

mer, a more physically accurate ﬁre spreading model

could be developed, still ensuring, however, real-time

performance. For the latter, ways to introduce a cer-

tain degree of interactivity could be studied, e.g.,

by interpolating different simulation results to imple-

ment events of interest (like changes in the ventila-

tion, etc.). Moreover, the multi-user, multi-role expe-

rience could be expanded into a completely customiz-

able tool. In this way, a ﬁreﬁghting trainer could be

able to deﬁne new use cases and scenarios on-the-ﬂy,

by adding and removing roles, vehicles and other el-

ements, without requiring software modiﬁcations. In

this context, NPC roles such as the ﬁre truck driver

could be made human-controllable, and methods to

effectively simulate all the possible interactions be-

tween ﬁreﬁghters (e.g., the touch, useful in case of

low visibility) could be investigated too.

ACKNOWLEDGEMENTS

This work has been supported by the Interreg V-A

Francia-Italia ALCOTRA 2014-2020 PITEM RISK

FOR project. The authors want to thank Michele Billi,

Beatrice Fofﬁ, and Andrea Ton for their contribution

to the development phase. They also want to acknowl-

edge the support provided by SITAF S.p.A. in the de-

sign and validation phases.

REFERENCES

Andrade, M., Souto Maior, C., Silva, E., Moura, M., and

Lins, I. (2018). Serious games & human reliabil-

ity. The use of game-engine-based simulator data for

studies of evacuation under toxic cloud scenario. In

Proc. Probabilistic Safety Assessment and Manage-

ment PSAM 14, pages 1–12.

GRAPP 2021 - 16th International Conference on Computer Graphics Theory and Applications

104

BBC News (1999). Europe Tunnel ﬁre kills 30. http:

//news.bbc.co.uk/2/hi/europe/304219.stm. [accessed

03 October 2020].

BBC News (2001). Swiss tunnel ablaze after head-on

crash. http://news.bbc.co.uk/onthisday/hi/dates/

stories/october/24/newsid 2488000/2488557.stm.

[accessed 03 October 2020].

BBC News (2005). Two killed by Alpine tunnel ﬁre. http:

//news.bbc.co.uk/2/hi/europe/4610559.stm. [accessed

03 October 2020].

Cakmak, T. and Hager, H. (2014). Cyberith virtualizer: A

locomotion device for virtual reality. In ACM SIG-

GRAPH 2014 Emerging Technologies.

Calandra, D., Billi, M., Lamberti, F., Sanna, A., and

Borchiellini, R. (2018). Arm swinging vs treadmill:

A comparison between two techniques for locomotion

in virtual reality. In Proc. Eurographics 2018, pages

53–56.

Calandra, D., Lamberti, F., and Migliorini, M. (2019). On

the usability of consumer locomotion techniques in

serious games: Comparing arm swinging, treadmills

and walk-in-place. In Proc. 9th International Confer-

ence on Consumer Electronics (ICCE-Berlin), pages

348–352.

Cannav

o, A., Calandra, D., Prattic

o, F. G., Gatteschi, V.,

and Lamberti, F. (2020). An evaluation testbed for

locomotion in virtual reality. IEEE Transactions on

Visualization and Computer Graphics, page (in press).

Cannav

o, A., Demartini, C., Morra, L., and Lamberti, F.

(2019). Immersive virtual reality-based interfaces for

character animation. IEEE Access, 7:125463–125480.

Celozzi, C., Paravati, G., Sanna, A., and Lamberti,

F. (2010). A 6-DOF ARTag-based tracking sys-

tem. IEEE Transactions on Consumer Electronics,

56(1):203–210.

Cha, M., Han, S., Lee, J., and Choi, B. (2012). A virtual

reality based ﬁre training simulator integrated with ﬁre

dynamics data. Fire Safety Journal, 50:12 – 24.

Corelli, F., Battegazzorre, E., Strada, F., Bottino, A., and

Cimellaro, G. P. (2020). Assessing the usability of

different virtual reality systems for ﬁreﬁghter training.

In Proc. 15th International Joint Conference on Com-

puter Vision, Imaging and Computer Graphics Theory

and Applications, pages 146–153.

Engelbrecht, H., Lindeman, R. W., and Hoermann, S.

(2019). A SWOT analysis of the ﬁeld of virtual re-

ality for ﬁreﬁghter training. Frontiers in Robotics and

AI, 6:101.

Fathima S J, S. and Aroma, J. (2019). Simulation of ﬁre

safety training environment using immersive virtual

reality. International Journal of Recent Technology

and Engineering, 7:347–350.

Kinateder, M., Pauli, P., M

uller, M., Krieger, J., Heim-

becher, F., R

onnau, I., Bergerhausen, U., Vollmann,

G., Vogt, P., and M

uhlberger, A. (2013). Human

behaviour in severe tunnel accidents: Effects of in-

formation and behavioural training. Transportation

Research Part F: Trafﬁc Psychology and Behaviour,

17:20 – 32.

Kinateder, M., Ronchi, E., Gromer, D., M

uller, M., Jost, M.,

Nehﬁscher, M., M

uhlberger, A., and Pauli, P. (2014a).

Social inﬂuence on route choice in a virtual reality

tunnel ﬁre. Transportation Research Part F: Trafﬁc

Psychology and Behaviour, 26:116 – 125.

Kinateder, M., Ronchi, E., Nilsson, D., Kobes, M., M

uller,

M., Pauli, P., and M

uhlberger, A. (2014b). Virtual

reality for ﬁre evacuation research. In Proc. Feder-

ated Conference on Computer Science and Informa-

tion Systems, pages 313–321.

Louka, M. N. and Balducelli, C. (2001). Virtual reality

tools for emergency operation support and training. In

Proc. International Conference on Emergency Man-

agement Towards Co-operation and Global Harmo-

nization, pages 1–10.

Lovreglio, R. (2020). Virtual and augmented reality for

human behaviour in disasters: A review. In Proc.

Fire and Evacuation Modeling Technical Conference,

pages 1–14.

Lu, X., Yang, Z., Xu, Z., and Xiong, C. (2020). Scenario

simulation of indoor post-earthquake ﬁre rescue based

on building information model and virtual reality. In

Toledo Santos, E. and Scheer, S., editors, proc. 18th

International Conference on Computing in Civil and

Building Engineering, pages 1217–1226.

Ntzeremes, P. and Kirytopoulos, K. (2019). Evaluating the

role of risk assessment for road tunnel ﬁre safety: A

comparative review within the eu. Journal of Traf-

ﬁc and Transportation Engineering (English Edition),

6(3):282 – 296.

Ren, A., Chen, C., and Luo, Y. (2008). Simulation of emer-

gency evacuation in virtual reality. Tsinghua Science

& Technology, 13(5):674 – 680.

St Julien, T. U. and Shaw, C. D. (2003). Fireﬁghter com-

mand training virtual environment. In Proc. Confer-

ence on Diversity in Computing, page 30–33.

Xu, Z., Lu, X., Guan, H., Chen, C., and Ren, A. (2014). A

virtual reality based ﬁre training simulator with smoke

hazard assessment capacity. Advances in Engineering

Software, 68:1 – 8.

A Multi-role, Multi-user, Multi-technology Virtual Reality-based Road Tunnel Fire Simulator for Training Purposes

105