sensors

Article
Radio Wave Propagation and WSN Deployment in
Complex Utility Tunnel Environments †
Mikel Celaya-Echarri 1 , Leyre Azpilicueta 1, * , Peio Lopez-Iturri 2,3 , Imanol Picallo 2,3 ,
Erik Aguirre 2 , Jose Javier Astrain 4 , Jesús Villadangos 4 and Francisco Falcone 2,3
1 School of Engineering and Sciences, Tecnologico de Monterrey, 64849 Monterrey, NL, Mexico;
mikelcelaya@tec.mx
2 Electric, Electronic and Communication Engineering Department, Public University of Navarre,
31006 Pamplona, Navarra, Spain; peio.lopez@unavarra.es (P.L.-I.); imanol.picallo@unavarra.es (I.P.);
erik.aguirre@unavarra.es (E.A.); francisco.falcone@unavarra.es (F.F.)
3 Institute of Smart Cities, Public University of Navarre, 31006 Pamplona, Navarra, Spain
4 Mathematical Engineering and Computer Science Department, Public University of Navarre,
31006 Pamplona, Navarra, Spain; josej.astrain@unavarra.es (J.J.A.); jesusv@unavarra.es (J.V.)
* Correspondence: leyre.azpilicueta@tec.mx; Tel.: +52-8183582000-5362
† This paper is an extended version of our paper published in Celaya-Echarri, M.; Azpilicueta, L.;
López-Iturri, P.; Picallo, I.; Aguirre, E.; Astrain, J.J.; Villadangos, J.; Falcone, F. Wireless Channel
Characterization and System Analysis of Complex Utility Tunnel Environments. In Proceedings of the 6th
International Electronic Conference on Sensors and Applications, 15–30 November 2019; Available online:
https://ecsa-6.sciforum.net.




Received: 29 October 2020; Accepted: 19 November 2020; Published: 24 November 2020


Abstract: The significant growth of wireless communications systems in the last years has led to the
adoption of a wide range of applications not only for the general public but, also, including utilities
and administrative authorities. In this context, the notable expansion of new services for smart
cities requires, in some specific cases, the construction of underground tunnels in order to enable the
maintenance and operation works of utilities, as well as to reduce the visual impact within the city
center. One of the main challenges is that, inherently, underground service tunnels lack coverage
from exterior wireless communication systems, which can be potentially dangerous for maintenance
personnel working within the tunnels. Accordingly, wireless coverage should be deployed within
the underground installation in order to guarantee real-time connectivity for safety maintenance,
remote surveillance or monitoring operations. In this work, wireless channel characterization for
complex urban tunnel environments was analyzed based on the assessment of LoRaWAN and ZigBee
technologies operating at 868 MHz. For that purpose, a real urban utility tunnel was modeled and
simulated by means of an in-house three-dimensional ray-launching (3D-RL) code. The utility tunnel
scenario is a complex and singular environment in terms of radio wave propagation due to the
limited dimensions and metallic elements within it, such as service trays, user pathways or handrails,
which were considered in the simulations. The simulated 3D-RL algorithm was calibrated and verified
with experimental measurements, after which, the simulation and measurement results showed good
agreement. Besides, a complete wireless sensor network (WSN) deployment within the tunnels was
presented, providing remote cloud data access applications and services, allowing infrastructure
security and safety work conditions. The obtained results provided an adequate radio planning
approach for the deployment of wireless systems in complex urban utility scenarios, with optimal
coverage and enhanced quality of service.

Keywords: wireless sensor networks; LoRaWAN; ZigBee; 3D Ray-Launching; tunnel environment;

smart cities

Sensors 2020, 20, 6710; doi:10.3390/s20236710 www.mdpi.com/journal/sensors

Sensors 2020, 20, 6710 2 of 24

1. Introduction
The evolution of wireless communication has led to their adoption in a wide range of applications,
including utilities and administrative authorities. In addition, the growth in population, as well as in
the services offered in cities, has led, in some cases, to the construction of underground tunnels in order
to reduce the visual impact within the city center, as well as facilitating maintenance and operation
works of utilities. In recent years, the underground sector has seen the use of wireless networking
for certain projects, i.e., tunneling construction [1], but they do not make full use of the benefits
offered by wireless communications. The utilization of wireless networking in the underground
environment offers the potential for major benefits in productivity and safety. Currently, most of
the communications infrastructure in operational tunnels such as transport tunnels are provided
by cables. However, if it is an infrequently accessed tunnel, a wired communication infrastructure
will not generally be available; therefore, the maintenance, upgrade and inspection of those tunnels
will rely on wireless communications [2]. Nevertheless, one of the issues that local authorities have
considered is the fact that, inherently, underground service tunnels lack coverage from exterior wireless
systems, such as mobile networks or municipal wireless local area networks (WLAN), which can lead
to potentially dangerous situations for maintenance personnel working within the tunnel. Therefore,
wireless coverage should be deployed within the underground installation in order to guarantee
real-time connectivity.
In emergency situations, such as fires, explosions or earthquakes, the normal wired communication
infrastructure of an underground environment could be inoperable. Due to that, sometimes, it is
necessary for the emergency services to have a wireless communication infrastructure following
a serious incident. Rescue services normally are equipped with very high-frequency (VHF) or
ultra-high-frequency (UHF) radio equipment, which is effective aboveground, but such radios have a
very poor range in tunnels. The use of leaky feeder cables [3] is usually used to extend the coverage
of VHF and UHF systems. However, the use of microwave frequencies provides a much greater
range, without expensive leaky feeder systems, in such environments, and, also, they offer greater
bandwidths. In addition, leaky feeders are also susceptible to damage during a fall or explosion.
In [4], a system including both wired and wireless communications was described that enabled voice
communication, as well as the exchange of photographs, video streams or other data, inside the tunnel.
In [5], a wireless-based personnel tracking system to track people in an underground environment was
presented, showing that the system can reduce the time taken for rescuers to reach any casualties.
The characterization of wireless channel behavior within complex heterogenous underground
environments, such as utility tunnel systems, enables to perform link-balance estimations, which lead
to coverage/capacity estimations. In this sense, an adequate estimation of losses is relevant in order
to characterize the complex operation conditions, such as those given by high-obstacle density,
with a high percentage of non-line-of-sight conditions and multipath propagation buildup owing to
scatterers distributions within the scenarios under test or the effect of diffuse scattering, among others.
This information is relevant in order to perform radio-planning tasks, which are required in order
to design the network topology, node configuration and device-related parameters, such as antenna
location/type. A radio-planning analysis was studied for different types of wireless networks
(e.g., broadcasting systems, trunk radios, public land mobile systems, microwave transport links,
etc.), considering aspects such as multi-variable optimization (energy consumption, node density and
transmission rate as a function of the coverage area under analysis) and the impact of elements such as
user density variation, estimation of cell edge conditions or interference background levels [6–11].
With the progressive adoption of wireless sensor networks, radio-planning tasks have been
adapted to optimize network configuration [12–18]. In this case, the specific considerations given by
the wireless sensor network (WSN) operation have to be taken into account, such as the random node
location, the existence of predominant non-line-of-sight operation conditions or the inherent energy
consumption and processing constraints, among others. In this way, a wireless channel analysis can
be effectively employed in order to assess elements such as cooperative MAC schemes, optimized
Sensors 2020, 20, 6710 3 of 24

routing protocols or specific data aggregator policies in order to optimize the overall system operation.
One of the main challenges in this case is to consider accurate radio channel conditions, given the large
variability of node location and configuration within the wireless network layout, usually leading to the
use of empirical/analytical-based channel estimation models. Radio propagation has been studied for
different wireless systems within the interiors of galleries and tunnels [19–22], as well as for different
tunnel geometries [23]. Other approaches have developed diffraction-based models, such as the
alternate direction implicit parabolic equation method [24], boundary element method [25], numerical
calculation of a vector parabolic equation [26] or application of the perturbation theory, among others.
Another approach is based on the use of deterministic methods, such as ray launching (RL) or ray
tracing (RT). One of the main issues of these approaches is the inherent divergence within the RL
calculation as the distance increases between the transmit and receive antenna. In this case, solutions in
scaled frequency models have been given by reconstructing the initial wave front of the source, achieved
by ray density normalization [27]. One of the approaches adopted to perform wireless characterization
within tunnels has been the adoption of dielectric waveguide models, in which the tunnel is emulated
by a waveguide with dielectric walls [28]. These models, however, do not take into account the
additional complexities, which appear in underground service tunnels, such as ducts, walkways
or distribution trays, among others. Complete modeling can be achieved by employing full-wave
electromagnetic simulation techniques, such as the finite difference time domain (FDTD) or frequency
domain finite element techniques [29,30]. However, the computational cost of these techniques can be
very high, due to the large size of the tunnel structures as compared to the wavelengths considered in
conventional wireless communication systems. Moreover, the tunnels analyzed are usually devoted to
road traffic or trains, with inherently larger cross-sections and lower structural complexity as compared
with utility tunnels, with dense obstacle distributions.
In this paper, an in-house developed three-dimensional ray-launching (3D-RL) code is employed
to analyze the wireless propagation in a real urban utility tunnel scenario at 868-MHz operation
frequency. The aim of the article is a thorough analysis of the real results of measurements in the tunnels
in order to obtain the optimal parameters for the simulation’s 3D-RL algorithm, so that, after their
application, the simulated and experimental results show good agreement. By adequately modeling
the tunnel gallery structure, a precise characterization for the complete extension of the tunnel system
can be obtained, which can aid in the deployment of wireless communication and monitoring systems.
Furthermore, a complete radio wave propagation analysis is performed, and a WSN deployment is
proposed for the considered real underground utility tunnel system gallery. Finally, an infrastructure
security and a work safety monitoring application prototype is developed in order to warrant work
safety legislation.
The paper is organized as follows: the 3D-RL deterministic employed technique, as well as the
considered real and simulated underground scenario are described in Section 2. Section 3 presents
the radio channel analysis and the required convergence analysis due to the particular characteristics
of the scenario. This analysis was performed compared with experimental measurements in the real
utility tunnel in order to obtain the optimal simulation parameters. Once the optimal parameters
were obtained, large- and small-scale simulation results within the tunnel were presented in order to
characterize the radio wave propagation within this particular environment. In Section 4, the WSN
deployment issues are discussed, and the infrastructure security and work safety monitoring application
prototype are presented. Finally, the conclusions are given in Section 5. A schematic view of the
structure of this paper is presented in Figure 1.
Sensors 2020, 20, 6710 4 of 24
Sensors 2020, 20, x FOR PEER REVIEW 4 of 24

1. Schematic view of the structure of this work.

Figure 1. work. WSN: wireless sensor network and 3D-RL:
three-dimensional ray-launching.
three-dimensional ray-launching.

2. Materials and
2. Materials and Methods
Methods
In
In this
this section, the in-house
section, the in-housedeterministic
deterministictechnique
techniquebased
based onon the
the 3D-RL
3D-RL approach
approach is described,
is described, as
as well as the considered real and simulated underground utility tunnel scenario used for this analysis.
well as the considered real and simulated underground utility tunnel scenario used for this analysis.
2.1. 3D Ray-Launching Technique
2.1. 3D Ray-Launching Technique
The in-house implemented RL algorithm, based on MATLAB programming environment, was used
The in-house implemented RL algorithm, based on MATLAB programming environment, was
to analyze radio wave propagation in the considered underground tunnel scenario. The main difference
used to analyze radio wave propagation in the considered underground tunnel scenario. The main
between RL and classical RT methods is that, in the RL technique, the principle is that rays are launched
difference between RL and classical RT methods is that, in the RL technique, the principle is that rays
in a solid angle and the rays arriving at the receiver are considered the true path. However, in classical
are launched in a solid angle and the rays arriving at the receiver are considered the true path.
RT techniques, the image of the transmitter or of the receiver is calculated, to be used to find the
However, in classical RT techniques, the image of the transmitter or of the receiver is calculated, to
reflected and refracted paths.
be used to find the reflected and refracted paths.
The RL algorithm is based on geometrical optics (GO) and the geometrical theory of diffraction
The RL algorithm is based on geometrical optics (GO) and the geometrical theory of diffraction
(GTD). To avoid the nonfield regions predicted by GO in the shadow areas, the diffracted rays are
(GTD). To avoid the nonfield regions predicted by GO in the shadow areas, the diffracted rays are
considered with the GTD and its uniform extension, called the uniform GTD (UTD). These diffracted
considered with the GTD and its uniform extension, called the uniform GTD (UTD). These diffracted
rays are introduced to avoid the abrupt field changes and consider the proper field corrections.
rays are introduced to avoid the abrupt field changes and consider the proper field corrections.
Electromagnetic phenomena such as reflection, refraction and diffraction are considered. The principle
Electromagnetic phenomena such as reflection, refraction and diffraction are considered. The principle
of the algorithm is that hundreds or thousands of rays are launched for a fixed position of the transmitter,
of the algorithm is that hundreds or thousands of rays are launched for a fixed position of the
which is an input parameter of the algorithm. The radiation pattern of the transmitter and receiver
transmitter, which is an input parameter of the algorithm. The radiation pattern of the transmitter and
antennas can be introduced, with the aid of the elevation angle θ and the azimuth angle ϕ, considering
receiver antennas can be introduced, with the aid of the elevation angle θ and the azimuth angle φ,
the spherical coordinate system. Each ray has a different path in the considered scenario, and all
considering the spherical coordinate system. Each ray has a different path in the considered scenario,
the parameters related with its path are stored in a matrix. All the rays that achieve the receiver are
and all the parameters related with its path are stored in a matrix. All the rays that achieve the receiver
considered in the calculation of the electromagnetic field at that point. The whole scenario is divided
are considered in the calculation of the electromagnetic field at that point. The whole scenario is divided
into a three-dimensional array of cuboids, with a different spatial resolution in the x-axis, y-axis and
into a three-dimensional array of cuboids, with a different spatial resolution in the x-axis, y-axis and z-
axis, depending on the specific characteristics and dimensions of the considered scenario. The basis is
 Sensors 2020, 20, 6710 5 of 24

Sensors 2020,
z-axis, 20, x FOR PEER
depending REVIEW
on the specific
characteristics and dimensions of the considered scenario. The 5 ofbasis
24

is that, when a ray impacts an obstacle or a wall in the scenario under analysis, reflected and refracted
that, when a ray impacts an obstacle or a wall in the scenario under analysis, reflected and refracted
rays are created, following Snell’s law. When a ray hits an edge, a new family of diffracted rays are
rays are created, following Snell’s law. When a ray hits an edge, a new family of diffracted rays are
created, according to the UTD. The methodology is shown in Figure 2.
created, according to the UTD. The methodology is shown in Figure 2.

Figure
Figure2.2.Principle
Principleofof
operation ofof
operation the in-house
the in-house3D-RL
3D-RLalgorithm.
algorithm.

AsAsinput parameters ofofthe

input parameters the algorithm,
algorithm, the frequency
the frequency of operation,
of operation, numbernumber of multipath
of multipath reflections
reflections
and spatial andandspatial
angular and angular are
resolution resolution are considered.
considered. In the simulations,
In the simulations, the materialthe materialof
properties
properties
all the obstacles within the scenario are also considered, taking into account the conductivitythe
of all the obstacles within the scenario are also considered, taking into account and
conductivity and relative
relative permittivity at thepermittivity at theoffrequency
frequency range range
operation of of operation
the system under of the system
analysis. under
The in-house
analysis. The in-house
implemented 3D-RL implemented
algorithm has3D-RL algorithm
already has already
been validated beenliterature
in the validatedforin different
the literature for
complex
different
scenarios complex scenariosup
for frequencies fortofrequencies
6 GHz. Some up scenarios
to 6 GHz.that
Some scenarios
have that have been
been considered have considered
been typical
have been
office typical[31,32],
scenarios office scenarios [31,32], sports
sports environments environments
[33], [33], vehicular communications
vehicular communications [34] or WSNs in smart[34]
orcities
WSNs in smart cities [35]. The detailed description of the algorithm
[35]. The detailed description of the algorithm can be found in [36]. can be found in [36].

2.2.Scenario
2.2. ScenarioDescription
Description

InInorder
ordertotoperform
performthe theradio
radiowave
wavepropagation
propagationanalysis
analysiswithin
withinthethetunnel,
tunnel,two twoscenarios
scenarioswere
were
consideredfor
considered forthe
the3D-RL
3D-RLsimulations:
simulations:a areal realunderground
undergroundutility utilitytunnel
tunnelscenario
scenariolocated
locatedininthetheOld
Old
TownofofPamplona
Town Pamplona(Pamplona,
(Pamplona,Navarre,
Navarre,Spain)
Spain)and anda atunnel
tunnelrecreation
recreationofofoneoneofofitsits gallery
gallery sections.
sections.
Thedistribution
The distributionofofthe theconsidered
consideredexperimental
experimentalunderground
undergroundutility utilitytunnels
tunnelsgallery
galleryisispresented
presentedinin
Figure3a3awith
Figure withgreen
greenlines
lineson onthe
themap.
map.AsAsit itcancanbebeseen
seenfromfromthe thefigure,
figure,the
thetunnels
tunnelshave havedifferent
different
sections, where
sections, where thethemost
mostremarkable
remarkable characteristics
characteristicsare the
aredifferent straightstraight
the different section lengths
section connected
lengths
to one section
connected to onewith some intersections
section and slight curves.
with some intersections The dashed
and slight curves.yellow line in the
The dashed figure
yellow represents
line in the
the straight
figure tunnel
represents thesection
straight of tunnel
80-m-long where
section the experimental
of 80-m-long where measurements
the experimental were performed in
measurements
orderperformed
were to calibrateinthe 3D-RL
order algorithm,
to calibrate as 3D-RL
the it is explained
algorithm, in Section
as it is3.explained
These tunnels were built
in Section in the
3. These
early 2000s,
tunnels underinsome
were built specific
the early 2000s,streets,
under with
some thespecific
aim of streets,
easing maintenance
with the aim operations. The tunnels
of easing maintenance
consist of an
operations. Theouter layer consist
tunnels of connected
of ancement rings,of
outer layer with the innercement
connected part allocating several
rings, with themetallic trays
inner part
allocating several metallic trays for the corresponding utilities. The longest straight tunnel sectionwas
for the corresponding utilities. The longest straight tunnel section is 327 m. A metallic hallway is
also
327 m.installed
A metallic in hallway
order towas provide access tointhe
also installed maintenance
order to providepersonnel,
access to theas maintenance
can be seen in Figure 3b.
personnel,
asCurrently,
can be seen theyin provide
Figure 3b. a route for thethey
Currently, electrical
provide connections,
a route fortelephone connections,
the electrical connections, waste disposal
telephone
and water distribution
connections, waste disposal systems of the distribution
and water neighborhood. It is worth
systems of the noting that slightly
neighborhood. It is different utility
worth noting
tunnels
that have
slightly been built
different in newer
utility tunnels neighborhoods
have been built of Pamplona, as can be seen of
in newer neighborhoods in Pamplona,
Figure 3c. Specifically,
as can be
this in
seen picture
Figurecorresponds to a recently
3c. Specifically, repaired
this picture tunnel located
corresponds in the suburb
to a recently repaired of Lezkairu,
tunnel locatedwhichinbegan
the
being built
suburb in 2008, which
of Lezkairu, and it has
begana total
beinglength
builtofintunnels
2008, andof 5600
it hasm,a plus
total2300 m of sub-tunnels.
length tunnels of 5600 m,
plus 2300 m of sub-tunnels.
Sensors 2020, 20, 6710 6 of 24

Sensors 2020, 20, x FOR PEER REVIEW 6 of 24

(a)

(b)

(c)
Figure 3. Underground
Underground utility tunnels under analysis. (a) Tunnel location on the map of the Old Town
of Pamplona. (b)(b) Picture of the tunnel where the measurements
measurements were taken.
taken. (c) Picture of a newer
neighborhood of
tunnel built in the neighborhood of Lezkairu
Lezkairu in
in Pamplona.
Pamplona.

Regarding
Regarding security
securityandandsafety, thethe
safety, utility tunnels
utility include
tunnels common
include intrusion
common detection
intrusion and antifire
detection and
systems, with their corresponding local alarms. However, one of the main problems
antifire systems, with their corresponding local alarms. However, one of the main problems of these of these types of
tunnels
types ofistunnels
the lackisofthewireless
lack of coverage within thewithin
wireless coverage tunnel, which
the tunnel,initially
whichwas planned
initially wastoplanned
be provided
to be
by external
provided byinfrastructures, such as municipal
external infrastructures, Wi-Fi or mobile
such as municipal Wi-Fi or networks (3G/4G). However,
mobile networks in reality,
(3G/4G). However,
these technologies
in reality, are unable toare
these technologies provide coverage
unable to all tunnel
to provide coverage sections,
to all astunnel
signal sections,
attenuation as can be
signal
very high in can
attenuation some besections
very highdueintosome
the different
sectionstunnel
due todepths. In thistunnel
the different sense,depths.
one of the mainsense,
In this concerns
one of
municipal authorities
the main concerns has been this
of municipal lack of wireless
authorities has been coverage
this lackwithin the tunnels,
of wireless coverage as coverage
within theistunnels,
needed
in
as order to provide
coverage security
is needed andto
in order safety to the
provide maintenance
security staff.toThus,
and safety a specific indoor
the maintenance staff.coverage system
Thus, a specific
is required,
indoor and therefore,
coverage system isa required,
complete andradiotherefore,
planningaanalysis
complete before
radio the deployment
planning of the
analysis wireless
before the
system is needed in order to secure coverage and provide good performance
deployment of the wireless system is needed in order to secure coverage and provide good at an affordable cost.
As statedatbefore,
performance in ordercost.
an affordable to characterize the variations of received power within the proposed
utility
Astunnel,
stated simulations
before, in orderby means of a 3D-RL
to characterize the code were of
variations performed.
received powerMaterial parameters
within were
the proposed
considered; basically,
utility tunnel, cementbywalls,
simulations means aluminum
of a 3D-RLtrayscode
and PVC
were tubes of different
performed. dimensions.
Material parametersA section
were
of the simulation
considered; scenario
basically, is depicted
cement walls, in Figure 4, which
aluminum corresponds
trays and PVC tubes withofthedifferent
same section where the
dimensions. A
experimental
section of themeasurements were performed
simulation scenario is depicted(see yellow 4,
in Figure dashed
whichline in Figure 3a).
corresponds withThe
thedimensions
same sectionof
where the experimental measurements were performed (see yellow dashed line in Figure 3a). The
dimensions of the tunnel section are the following: width 2 m, height 3 m and length 80 m. Simulation
results can be extended in length; however, due to measurement limitations, they will be considered
Sensors 2020, 20, 6710 7 of 24

the tunnel section are the following: width 2 m, height 3 m and length 80 m. Simulation results
Sensors 2020, 20, x FOR PEER REVIEW
can
7 of 24
be extended in length; however, due to measurement limitations, they will be considered up to the
previously stated distance.
up to the previously statedThe complete
distance. Thetopological and morphological
complete topological characteristics
and morphological of this scenario
characteristics of
were taken into account, prior visual inspection and measurement of dimensions within
this scenario were taken into account, prior visual inspection and measurement of dimensions within the tunnels.
A
theschematic
tunnels. A view of the employed
schematic simulation
view of the employedscenario is shown
simulation in Figure
scenario 4. in Figure 4.
is shown

Figure 4. Schematic representation of the simulation scenario implemented to analyze the radio
Figure 4. Schematic representation of the simulation scenario implemented to analyze the radio
propagation losses within the underground utility tunnels.
propagation losses within the underground utility tunnels.

One of
One ofthe
theparticular
particularaspects
aspectsof of
thisthis scenario
scenario is fact
is the the that
fact the
thattunnel
the tunnel
is not is not a conventional
a conventional tunnel
tunnel
(i.e., free(i.e., free hallway
hallway with
with walls thatwalls that
can be can beorsmooth
smooth or can
can present presentdegree
a certain a certain degree roughness
of surface of surface
roughness
and, hence, and, hence,
diffuse diffusedue
scattering), scattering),
to the factduethattothere
the isfact that
a set of there
serviceistrays
a setplaced
of service
insidetrays placed
the tunnel.
inside
This the to
leads tunnel. This leads to
new interactions newthe
with interactions with the
electromagnetic electromagnetic
waves launched by waves launched
the source, by the
with strong
source, with
multipath strong multipath
component generation component
given the generation given thethe
large ratio between large ratio
length ofbetween
the tunneltheinlength of the
comparison
tunnel in comparison with the width of the tunnels, tray dimension and inter-tray
with the width of the tunnels, tray dimension and inter-tray distance. This condition implies the need distance. This
condition
for including implies
morethe need topological
precise for including more precise
information topological
of the scenario information of the scenario
in order to consider in order
the interaction,
to consider the interaction, specifically for launched rays within a RL
specifically for launched rays within a RL algorithm, such as the one employed and described algorithm, such as the onein
employed
next and described in next subsection.
subsection.

3. Radio Channel Analysis

3.1. 3D-RL Calibration

3.1. 3D-RL Calibration and
and Validation
Validation
The
The underground
underground tunnel
tunnel scenario
scenario hashas distinctive
distinctive characteristics
characteristics that
that convert
convert itit in
in aa particular
particular
scenario for radio wave propagation purposes. These peculiar characteristics are
scenario for radio wave propagation purposes. These peculiar characteristics are that it is an area that it is an area
that
that
is considerably long (80-m-long, in this specific case) and extremely narrow (2-m width and 3-m
is considerably long (80-m-long, in this specific case) and extremely narrow (2-m width and 3-m
height).
height). It can be
It can be found
found in
in the
the literature
literature [36,37],
[36,37], the
the convergence
convergence analysis
analysis of the in-house
of the in-house RL algorithm
RL algorithm
to
to obtain
obtain the
the optimal
optimal input
input parameters
parameters for for the
the best
best performance
performance ofof the
the algorithm
algorithm in in typical
typical indoor
indoor
complex scenarios. However, for this specific environment, those parameters
complex scenarios. However, for this specific environment, those parameters are not the are not the optimal,
optimal,
because
because the
the RL
RL algorithm
algorithm diverges
diverges due
due to
to the
the extremely
extremely narrow
narrow width
width ofof the
the scenario.
scenario. The divergence
The divergence
of
of the algorithm means that there are a lot of cuboids of the scenario that do not have any ray
the algorithm means that there are a lot of cuboids of the scenario that do not have any ray if
if the
the
same simulation parameters are used for this specific scenario. Due to that, the 3D-RL
same simulation parameters are used for this specific scenario. Due to that, the 3D-RL algorithm algorithm needs
to be calibrated
needs according
to be calibrated to the to
according experimental
the experimentalmeasurements to validate
measurements the results
to validate accuracy
the results with
accuracy
with the optimal input parameters. Therefore, a convergence analysis of the algorithm was done for
this underground tunnel environment in order to characterize the 3D-RL convergence as a function
of multiple parameters, such as cuboid dimensions, horizontal angular resolution (Δφ), vertical
angular resolution (Δθ) or the maximum number of reflections. Due to the specific characteristics of
Sensors 2020, 20, 6710 8 of 24

the optimal input parameters. Therefore, a convergence analysis of the algorithm was done for this
underground tunnel environment in order to characterize the 3D-RL convergence as a function of
multiple parameters, such as cuboid dimensions, horizontal angular resolution (∆ϕ), vertical angular
resolution (∆θ) or the maximum number of reflections. Due to the specific characteristics of the utility
tunnel environment (i.e., strong reflections due to the partial conductor material presence and large
obstacle density), the maximum number of reflections and the angular resolution of the launching rays
was analyzed. For that purpose, the different number of multipath reflections and different number of
launching rays were considered. Measurement results were compared with the different simulation
results to find which simulation results achieved the best trade-off between precision and simulation
computational cost, thus obtaining a calibrated 3D-RL technique that could be used for a radio-planning
analysis in this type of complex underground environment. The considered frequency of operation
was 868 MHz, emulating the operation of a typical LoRaWAN or ZigBee wireless communication
system. The different cases for the convergence analysis, which was considered for the simulations of
the underground tunnel scenario, are shown in Table 1.

Table 1. Different cases for the convergence analysis of the algorithm.

∆ϕ (rad) ∆θ (rad) Cuboids Size (m) Number of Reflections Number of Rays

Case 1 π/900 π/900 0.1 5 810,000
Case 2 π/900 π/900 0.1 7 810,000
Case 3 π/360 π/360 0.1 7 129,600
Case 4 π/180 π/180 0.1 7 32,400
Case 5 π/180 π/180 0.1 13 32,400
Case 6 π/900 π/900 0.1 13 810,000

Once the simulation results for the different cases were obtained, the validation and the
corresponding calibration was obtained by performing radio channel measurements inside the
tunnel section. The scenario in which the measurements were performed is a section of the utility
underground tunnels in the city of Pamplona, which is depicted in Figure 3b. Measurements were
carried out with the aid of an Agilent CSA N1996A to generate the radiofrequency (RF) signal and an
Agilent FieldFox N9912A portable spectrum analyzer, both from Keysight Technologies. The transmitter
was connected to a 5-dBi monopole antenna, whereas the receiver was connected to small vertical
0-dBi monopole antenna. The considered frequency of the operation was 868 MHz, with a transmitted
power of −10 dBm. The transmitter antenna was placed at the beginning of the tunnel, located at the
center with a height of 2 m, whereas the receiver points were placed at different distances in front of
the transmitter along the tunnel. Measurements were performed with no users crossing the tunnels or
other potential time-varying scatterers within the scenario. Results, compared with the different cases
performed (shown in Table 1) in the 3D-RL simulations from the implemented scenario, are depicted
in Figure 5. It can be seen from the picture that Cases 1, 3, 4 and 5 did not have good agreement with
the measurements after 50-m distance approximately. This is because Cases 3, 4 and 5 considered
larger angular resolutions than the other cases (one degree and 0.5 degrees of angular resolution) in
comparison with 0.2 degrees in the other cases. In Case 1, the angular resolution was high, but the
number of reflections was the lowest. It was shown that Case 1 did not have good agreement with the
measurements. It was observed that Case 2 and Case 6 (both with angular resolutions of 0.2 degrees
but different numbers of reflections considered) both had a good match with real measurements.
Sensors 2020, 20, 6710 9 of 24
Sensors 2020, 20, x FOR PEER REVIEW 9 of 24

Figure 5. Comparison of different cases of simulation and measurement results for the service tunnel
section under analysis.

Table
Table 22shows
shows thethe
mean errorerror
mean and standard deviation
and standard of the comparison
deviation of the different
of the comparison of thesimulation
different
cases
simulation cases with the real measurements. It is also shown the simulation computational timeone
with the real measurements. It is also shown the simulation computational time of each of
of theone
each considered cases. It can
of the considered beItseen
cases. can that Case
be seen 2 and
that CaseCase
2 and6 have
Case the lowest
6 have mean error,
the lowest meanas stated
error, as
before. In Cases
stated before. In1,Cases
3, 4 and1, 35,4the
andmean
5, theerror increases
mean error considerably. With regards
increases considerably. Withto the simulation
regards to the
computational time, it is observed that Case 6 has an increase of 186,861 s more
simulation computational time, it is observed that Case 6 has an increase of 186,861 s more than than Case 2, with
Casea
slight
2, withdecrease
a slight in the mean
decrease error.
in the meanSo, error.
it can So,
be concluded that, for the
it can be concluded underground
that, tunnel scenario,
for the underground tunnel
the optimal simulation parameters are Case 2, achieving a good trade-off between
scenario, the optimal simulation parameters are Case 2, achieving a good trade-off between accuracyaccuracy results and
simulation
results and computational burden.
simulation computational burden.

Table 2.
Table Comparison between
2. Comparison between different
different simulation
simulation cases.
cases.

Case 1 1 Case
Case Case2 2 Case
Case33 Case
Case4 4 Case
Case55 Case
Case66
MeanError
Mean Error (dB)
(dB) 2.952.95 1.34
1.34 2.74
2.74 5.24
5.24 5.24
5.24 1.06
1.06
Standard
StandardDeviation (dB)(dB) 16.14
Deviation 16.14 5.90
5.90 16.56
16.56 14.29
14.29 14.29
14.29 6.14
6.14
Computational
Computational Time (s) (s) 52,664
Time 52,664 78,459
78,459 4852
4852 883
883 1656
1656 265,320
265,320

3.2. Simulation Results

Once thethecalibration
calibration andandverification of the
verification of3D-RL algorithm
the 3D-RL for thisfor
algorithm singular scenario was
this singular achieved,
scenario was
the input optimal parameters, in terms of angular resolution and number
achieved, the input optimal parameters, in terms of angular resolution and number of reflections, of reflections, were applied
in the applied
were simulations of the
in the underground
simulations utility
of the scenario. Large
underground utilityand small-scale
scenario. Largeradio
andwave propagation
small-scale radio
analyses were performed
wave propagation analyses withweretheperformed
aid of simulations
with thein order
aid to provide in
of simulations a general
order tosignal propagation
provide a general
understanding
signal propagation within this complex
understanding and singular
within environment.
this complex and singular Theenvironment.
tunnel cross-sections
The tunnel have an
cross-
important
sections have impact on the signal
an important impactattenuation in this
on the signal type of environment,
attenuation in this type ofasenvironment,
it is stated inas[38].
it is In this
stated
sense,
in [38].typical
In thistunnel
sense,cross-sections
typical tunnelcan be classifiedcan
cross-sections in rectangular
be classified orincircular shapes,
rectangular orwhere theshapes,
circular tunnel
width
where and the height
tunnelneed
widthto beandconsidered
height needwhentocalculating
be consideredthe radiowhensignal attenuation.
calculating the Nevertheless,
radio signal
it has been shown
attenuation. in the literature
Nevertheless, it has been[39,40] that cross-section
shown in the literature dimensions within
[39,40] that tunnels (i.e.,
cross-section different
dimensions
heights
within tunnels (i.e., different heights and widths of the tunnels) have a higher impact on theshapes,
and widths of the tunnels) have a higher impact on the signal attenuation than distinct signal
such as rectangular
attenuation or circular.
than distinct shapes,Therefore, the methodology,
such as rectangular signal
or circular. propagation
Therefore, analysis and signal
the methodology, WSN
deployment presented
propagation analysis andin WSN
this work could potentially
deployment presented in also
thisbework
applied
could topotentially
cylinder tunnels, as long
also be applied
as tunnel cross-sections
to cylinder tunnels, as long areasof similar
tunnel dimensionsare
cross-sections than the ones
of similar presentedthan
dimensions in this work.
the ones In fact,
presented
the cylindrical
in this work. In shape
fact, of thecylindrical
the tunnel will generate
shape of theslightly
tunnellower multipathslightly
will generate propagation
lower within
multipath the
environment,
propagation withinas the number of reflections
the environment, andnumber
as the diffractions within theand
of reflections surface of the tunnel
diffractions within will
thebesurface
lower.
of the tunnel will be lower.
Large scale propagation analysis was performed by means of the RF power distribution for the
whole volume of the scenario. For that purpose, the transmitter antenna was placed at the beginning
Sensors 2020, 20, 6710 10 of 24

Large
Sensors scale
2020, 20, propagation
x FOR PEER REVIEW analysis was performed by means of the RF power distribution for 10 ofthe
24
whole volume of the scenario. For that purpose, the transmitter antenna was placed at the beginning
of
of the
the tunnel
tunnel model,
model, atat a height
height of 2 m (represented
(represented by by aa red
red triangle
triangle in
in Figure
Figure 6). simulation
6). The same simulation
parameters
parameters as as in
in the
the experimental
experimental measurements
measurements were were considered
considered for
for the
the operational
operational frequency
frequency and and
the transmitted power: 868 MHz and −10 dBm, respectively, emulating
the transmitted power: 868 MHz and −10 dBm, respectively, emulating a ZigBee or LoRaWAN a ZigBee or LoRaWAN system
at this specific
system frequency.
at this specific FigureFigure
frequency. 6 shows the obtained
6 shows RF power
the obtained distribution
RF power estimations
distribution for the
estimations for
XY
the bidimensional
XY bidimensional plane at aatheight
plane of 2ofm
a height 2m(the same
(the same height
heightasas
the
thetransmitter
transmitterantenna).
antenna).ItIt can
can bebe
seen, the
the variability
variability of received
received power
power along
along the
the distance,
distance, showing
showing some
some spatial
spatial points
points with
with lower
lower
received
received power
power in inthe
thefinal
finalpart
partofofthe
thetunnel.
tunnel.The
Theservice
servicetrays
trayson
onthe
thesides
sidesofofthe
thetunnel
tunnelgenerated
generated a
nonhomogeneous variation of the signal distribution levels along the
a nonhomogeneous variation of the signal distribution levels along the tunnel distance. tunnel distance.

Figure 6.
Figure 6. Radiofrequency
Radiofrequency (RF) power distribution for the XY bidimensional
bidimensional plane at 2-m height
height for
for
868 MHz.
868 MHz.

To
To gain
gain insight
insight into
into the
the multipath
multipath behavior
behavior ofof the
the considered
considered scenario,
scenario, due
due to
to the rough
rough surfaces
surfaces
of the
the walls
walls and
and the
the service
service trays
trays located
located in
in the
the tunnel,
tunnel, thethe power
power delay
delay profile
profile (PDP)
(PDP) isis shown
shown in in
Figure 77 for
for aa specific
specific point
point of
of the
the scenario,
scenario, showing
showing the the high
high multipath
multipath trajectories
trajectories within
within a time
time
span ofof 200
200 and
and 2500
2500 ns.
ns. The high multipath propagation within the environment can lead to to signal
signal
fading in some spatial locations,
locations, which
whichcan
canlead
leadto toareas
areaswith
withaalack
lackofofcoverage
coveragewithin
withinthe
thetunnel
tunnel if
if
thethenetwork
networkdesign
designisisnot
notadequately
adequatelyperformed.
performed.In Inaddition,
addition, another
another parameter
parameter that
that can grossly
quantify
quantify the multipath channel is the root mean square delay spread (RMS DS), DS), referred
referred also as σ
also as 𝜎τ ,
which
which shows
showsthe effects
the of dispersion
effects of dispersionand is determined
and is determinedfrom the PDP.the
from ThePDP.
RMSTheDS for
RMSthe DS
considered
for the
utility tunnelutility
considered environment is depicted in Figure
tunnel environment 8. It isinthe
is depicted square
Figure 8.root
It isofthe
thesquare
secondroot
central moment
of the of
second
the PDP,moment
central definedofasthe
[41]:
PDP, defined as [41]: q
στ = τ2 − (τ)2 (1)
𝜎 𝜏 𝜏̅ (1)
where
ka τ k P(τk )τk
P 2 2 P 2
where τ2 = P k 2 k = P (2)
ka k k P(τk )

and ∑ 𝑎 𝜏 ∑ 𝑃 𝜏 𝜏
𝜏 P 2 (2)
∑
a τ𝑎k ∑ 𝑃 𝜏
k P( τk )τk
P
k
τ= P k2 = P (3)
and k ak k P ( τk )

It must be pointed out that values for∑τ, 𝑎τ2 𝜏and σ∑

τ depend
𝑃 𝜏 𝜏on the choice of the noise threshold to
𝜏̅ P(τ). The noise threshold used to depict the RMS DS shown
process the received power at each delay (3)
∑ 𝑎 ∑ 𝑃 𝜏
in Figure 8 was −120 dBm, chosen to emulate the typical sensitivity of a ZigBee or LoRaWAN WSN
It must
that can be pointed
be deployed out tunnel
in the that values for 𝜏̅, 𝜏 and 𝜎 depend on the choice of the noise threshold
environment.
to process the received power at each delay 𝑃 𝜏 . The noise threshold used to depict the RMS DS
shown in Figure 8 was −120 dBm, chosen to emulate the typical sensitivity of a ZigBee or LoRaWAN
WSN that can be deployed in the tunnel environment.
Sensors 2020, 20, 6710 11 of 24
Sensors
Sensors 2020,
2020, 20,
20, xx FOR
FOR PEER
PEER REVIEW
REVIEW 11
11 of
of 24
24

Figure 7. Power delay profile at a given cuboid located at the center (60 m, 1 m and 2 m) at 868-MHz
Figure 7. Power delay profile at a given cuboid located at the center (60 m, 1 m and 2 m) at
frequency.
868-MHz frequency.

FigureFigure 8. Root
8. Root meanmean square
square delay
delay spread(RMS
spread (RMS DS)
DS) estimation
estimationwithin
withinthe utility
the tunnel
utility for 2-m
tunnel height
for 2-m height
for 868-MHz frequency.
for 868-MHz frequency.

It is observed
It is observed thatthat
thethe RMSDS
RMS DSisis higher
higher ininsome
someareasareasclosest to the
closest to center and the
the center andfinal
thepart of part
final
the tunnel. This is because of the distribution of the service trays and tubes in the utility
of the tunnel. This is because of the distribution of the service trays and tubes in the utility tunnel. tunnel. Each
spatial sample of the RMS DS is associated with a PDP. In this specific case, a multipath propagation
Each spatial sample of the RMS DS is associated with a PDP. In this specific case, a multipath propagation
contribution is higher in the central and final part of the tunnel, but it is necessary to consider all the
contribution is higher in the central and final part of the tunnel, but it is necessary to consider all the
spatial samples of the environment due to the influence of the topology and morphology of the
spatial samples
specific of theunder
scenario environment due toTherefore,
consideration. the influence
all theofobstacles
the topology
withinand
the morphology
tunnel and theofmetallic
the specific
scenario under consideration. Therefore, all the obstacles within the
structure of the service trays play an important role in the multipath propagation.tunnel and the metallic structure
of the service traystoplay
In order anknowledge
have important of rolethe
in frequency
the multipath domainpropagation.
channel description, the coherence
In order to have
bandwidth knowledge
(CB) was obtainedoffrom
the frequency
the RMS DS domain
for the channel description,
considered scenario. Thethe CB
coherence bandwidth
can be defined
as the bandwidth over which the frequency correlation function is above
(CB) was obtained from the RMS DS for the considered scenario. The CB can be defined as the bandwidth0.9; then, the CB is
approximately 𝐵 1 ⁄ 50𝜎 ; see [41].
over which the frequency correlation function is above 0.9; then, the CB is approximately Bc ≈ 1/50στ ;
see [41]. Figure 9 shows the CB for the considered utility tunnel showing values that vary from 0.01 to
0.6 MHz. It can be seen that the lowest values of RMS DS cause higher values of CB. Therefore, a
Figure 9 shows the CB for the considered utility tunnel showing values that vary from 0.01 to
receiver in those areas can encounter higher multipath interference.
0.6 MHz. It can be seen that the lowest values of RMS DS cause higher values of CB. Therefore,
a receiver in those areas can encounter higher multipath interference.
Sensors 2020, 20, 6710 12 of 24
Sensors 2020, 20, x FOR PEER REVIEW 12 of 24

Figure
Figure 9. 9.Coherence
Coherence bandwidth
bandwidthwithin the considered
within utility tunnel
the considered utilityfor a 2-m height
tunnel for a for
2-m868-MHz
height for
frequency.
868-MHz frequency.

4. WSN Deployment
4. WSN Deployment
In section,
In this this section, a generic
a generic WSN
WSN deployment case
deployment case for
for complex
complexheterogeneous
heterogeneous utility tunnels
utility is
tunnels is
presented
presented considering
considering environmentaland
environmental andpropagation
propagation conditions
conditionsbyby
means
meansof aofcalibrated 3D-RL
a calibrated 3D-RL
simulation tool. A complete WSN deployment feasibility study is introduced focusing on the
simulation tool. A complete WSN deployment feasibility study is introduced focusing on the reliability
reliability of wireless communications technologies (Zigbee and LoRaWAN) within the considered
of wireless communications technologies (Zigbee and LoRaWAN) within the considered scenario and,
scenario and, particularly, in terms of security management, assessing the human presence
particularly, in terms
propagation of security
impact management,
over the assessing
system. Finally, the humanWSN
an optimized presence propagation
deployment impact over
is presented
the system.
considering the morphology and topology of the real case study underground tunnel system and
Finally, an optimized WSN deployment is presented considering the morphology
topology of the real
environment andcase study underground
an infrastructure and worktunnel system environment
safety monitoring and an infrastructure and
application prototype.
work safety monitoring application prototype.
4.1. Safety Management and Occupational Risk Control
4.1. Safety Management and Occupational Risk Control
One of the main problems of working in tunnels and service galleries, which is aggravated by
their of
One length and narrowness,
the main problems isofthe safety management
working in tunnels and and the occupational
service risks
galleries, control.
which The health by
is aggravated
their length and narrowness, is the safety management and the occupational risks control.includes
of the working environment of the maintenance staff must always be guaranteed. This The health
of themonitoring
working temperature
environment and humidity levels but, also, monitoring CO and CO2 levels; the presence
of the maintenance staff must always be guaranteed. This includes
of flammable or toxic gases and compounds and the presence of animals (rats, snakes, etc.), among
monitoring temperature and humidity levels but, also, monitoring CO and CO2 levels; the presence of
others. In addition, mobility within the tunnel, the absence of obstacles or elements that hinder
flammable or toxic gases and compounds and the presence of animals (rats, snakes, etc.), among others.
movement must be guaranteed. It should be borne in mind that employees, on many occasions, must
In addition, mobility
move tools to carrywithin themaintenance,
out their tunnel, the repair
absence andofinstallation
obstaclestasks.
or elements that hinder movement
must be guaranteed. It should
Current regulations be borne
related in risk
to labor mind that employees,
prevention state thaton many
work mustoccasions,
be done atmust move
least in pairstools
to carry
andout their maintenance,
desirable in work teamsrepair and installation
to minimize tasks. accidents. In addition, workers must
risks and prevent
notify theregulations
Current coordination centerto
related of labor
the task to prevention
risk be performed; thethat
state starting
work time;
mustthebeestimated
done atduration;
least in pairs
the working conditions (lighting, power supply cut, etc.); the possible risks
and desirable in work teams to minimize risks and prevent accidents. In addition, workers must and any other information
notifythat
themay help to ensure
coordination centersafety at work
of the during
task to the job performance.
be performed; the starting Hence,
time;personnel monitoring
the estimated duration;
implies that the staff will carry their own sensors (located in the inhalation zone and at head height)
the working conditions (lighting, power supply cut, etc.); the possible risks and any other information
and must be able to interconnect to the monitoring network. Thus, the use of a wireless network is
that may help to ensure safety at work during the job performance. Hence, personnel monitoring
particularly advantageous in the latter case, as it offers a greater ease of work (hands-free) and greater
implies that the
mobility for staff willHowever,
workers. carry their theown sensors
presence (located
of human in the
beings inhalation
within zone
the tunnels canand at the
affect head height)
radio
and must be able even
propagation, to interconnect to thetheir
more considering monitoring network.
narrowness. Thus,
Therefore, an the use ofRF
adequate a wireless network is
characterization
particularly advantageous
and a complete in theassessment
risk control latter case,must
as it be
offers a greater
performed in ease
orderoftowork (hands-free)
warrant compliance and greater
with
mobility for workers. However, the presence of human
occupational safety management within the utility tunnels. beings within the tunnels can affect the radio
propagation, even more considering their narrowness. Therefore, an adequate RF characterization
and a4.2. Technology
complete Analysis:
risk control ZigBee 868 MHz must
assessment vs. LoRaWAN Range within
be performed the Tunnels
in order to warrant compliance with
occupational3D-RLsafety management
simulations withintothe
were shown utility tunnels.
be accurate for an optimized radio-planning analysis in this
complex and particular underground environment. Based on these preliminary results, the coverage
4.2. Technology Analysis:
for both Zigbee ZigBee 868wireless
and LoRaWAN MHz vs. LoRaWAN
devices can beRange within
estimated (i.e.,the
theTunnels
number of wireless nodes
required for the whole tunnel) considering their corresponding device sensitivities, which are around
3D-RL simulations were shown to be accurate for an optimized radio-planning analysis in this
−100 dBm and −148 dBm, respectively. In order to provide clear insight, range estimations within the
complex and particular underground environment. Based on these preliminary results, the coverage
for both Zigbee and LoRaWAN wireless devices can be estimated (i.e., the number of wireless nodes
required for the whole tunnel) considering their corresponding device sensitivities, which are around
Sensors 2020, 20, 6710 13 of 24

−100Sensors and20,−148
dBm2020, x FORdBm, respectively. In order to provide clear insight, range estimations13within
PEER REVIEW of 24 the
tunnel for both wireless technologies were obtained. For that purpose, a transmitter was placed on the
tunnel
ceiling forbeginning
at the both wireless technologies
of the tunnel withwerea obtained. For that
length equal purpose,
to the longest a transmitter was of
tunnel section placed on
the scenario
the ceiling at the beginning of the tunnel with a length equal to the longest
under analysis (327 m, as it is presented in Figure 3a). The transmitted powers were set to 13 dBmtunnel section of the
scenario
and 14 dBm under analysis
for ZigBee and(327 m, as it is presented
LoRaWAN, in Figure
respectively. As can 3a).beThe transmitted
concluded powers
from were setdepicted
the results to
13 dBm and 14 dBm for ZigBee and LoRaWAN, respectively. As can be concluded from the results
in Figure 10, both wireless technologies covered the entire length of the tunnel, although LoRaWAN
depicted in Figure 10, both wireless technologies covered the entire length of the tunnel, although
presented better results due to its better sensitivity (−148 dBm) compared to ZigBee (−100 dBm).
LoRaWAN presented better results due to its better sensitivity (−148 dBm) compared to ZigBee (−100
Thus,dBm).
in terms
Thus,ofinwireless
terms oflink distance
wireless within awithin
link distance straight tunneltunnel
a straight of these characteristics,
of these bothboth
characteristics, wireless
technologies could be employed.
wireless technologies could be employed.

ZigBee 868 MHz range (dBm)

2.0
Y (m)

1.5
1.0 -100
0.5 -80
0.0
-60
0 50 100 150 200 250 300
-40
X (m) -20
0

(a)
LoRaWAN range
(dBm)
2.0
Y (m)

1.5 -160
1.0
0.5 -140
0.0
-120
0 50 100 150 200 250 300
-100
X (m) -80
-60
-40
-20
0

(b)
Figure
Figure 10. 10. Bidimensionalplane
Bidimensional plane of the
theRF
RFpower
powerlevel distribution
level at a height
distribution of 3 m for
at a height of 3(a)mZigBee
for (a)868
ZigBee
MHz and
868 MHz and (b)
(b)LoRaWAN.
LoRaWAN.

Nevertheless,
Nevertheless, in in
thethe realenvironment,
real environment, different
differentphenomena
phenomena cancan
lead to channel
lead impairments,
to channel impairments,
suchsuch as the
as the presence
presence of of peoplewithin
people within the
the tunnels
tunnels or
orworsening
worsening thethe
radio
radiopropagation conditions.
propagation conditions.
Therefore, a further study is required in order to provide safe, robust, reliable andWSN
Therefore, a further study is required in order to provide safe, robust, reliable and efficient efficient
WSNdeployment.
deployment.
4.3. Scatterers Analysis: Human Presence within the Tunnel
4.3. Scatterers Analysis: Human Presence within the Tunnel
An important factor that has to be assessed in terms of radio propagation in small cross-section
An important factor that has to be assessed in terms of radio propagation in small cross-section
utility tunnels is the presence of human beings. It is a well-known fact that human bodies create a
utility tunnels is
shadowing the[42],
effect presence
and inof human
such a narrowbeings. It is a well-known
environment fact thattunnels
like the underground humanpresented
bodies create
in a
shadowing effect [42], and in such a narrow environment like the underground
this work, this effect was expected to be especially significant. In order to have insight into this tunnels presented
in this work, thisdifferent
phenomenon, effect was expected
simulations to be
were especially
performed significant.
using In order
the calibrated 3D-RL to have insight
simulation tool. into
this phenomenon,
Human body models different
are simulations
included in thewere performed
3D-RL software,using the calibrated
emulating occupational 3D-RL simulation
personnel at the tool.
Human location
body of models
interest. are
Thisincluded
human body model
in the 3D-RLwassoftware,
implemented for the simulation,
emulating occupational considering
personnel theat the
dimensions of a standard person and all the dispersive material properties
location of interest. This human body model was implemented for the simulation, considering the of its different body parts
(i.e., skin,
dimensions of bones, muscles,
a standard personblood,
anddifferent organs, etc.).
all the dispersive A further
material description
properties of different
of its the considered
body parts
human body model, as well as a signal attenuation analysis for different parts of the human body,
(i.e., skin, bones, muscles, blood, different organs, etc.). A further description of the considered human
can be found in reference [43]. By including computational human body models within the
bodyconsidered
model, asscenario,
well asthe a signal attenuation analysis for different parts of the human body, can be
results of the human presence effect were obtained, allowing analysis and
found comparison. In the context of the utilitycomputational
in reference [43]. By including tunnel, teams ofhuman
up to threebody models
workers are within
generallythe considered
sent into
scenario, the results
the tunnels of the human
for performing presence
different securityeffect were obtained,
and maintenance allowing
tasks, in order analysis
to minimizeandthe
comparison.
risks
In the andcontext
preventof accidents
the utilitydue tunnel, teams ofauthorities’
to municipal up to threerequirements
workers areregarding
generallyoccupational
sent into therisktunnels
for performing different security and maintenance tasks, in order to minimize the risks and prevent
accidents due to municipal authorities’ requirements regarding occupational risk prevention legislation.
Sensors 2020, 20, 6710 14 of 24
Sensors 2020, 20, x FOR PEER REVIEW 14 of 24
Sensors 2020, 20, x FOR PEER REVIEW 14 of 24
preventionalegislation.
Therefore, Therefore,
complete analysis a complete
of the analysis within
human presence of the human presence
the tunnel withinout
was carried theconsidering
tunnel was
prevention
carried out legislation.
considering Therefore,
different a complete
densities of analysisbody
human of themodels
human inpresence within
different the tunnel was
locations.
different densities
carried of human
out considering body models
different densitiesinofdifferent
human body locations.
models in different locations.
First of
FirstFirst all,
of all, the received
the the
received power
power results of a specific scenario with
with no humans
humans is is presented
presented in
in
of all, received powerresults
results of
of aa specific scenario
specific scenario with nono humans is presented in
Figure
Figure 11, considering
11, considering a transmitter
a transmitter located on the tunnel ceiling at x-axis
x-axis =
=m20 m for reference and
Figure 11, considering a transmitterlocated
located on
on the tunnelceiling
the tunnel ceilingatat
x-axis = 20 20for
m reference
for reference
and and
comparison
comparison purposes.
purposes.
comparison purposes.

Heigth
Heigth == 1.5m
1.5m (dBm)(dBm)
2.0 2.0
Y (m)

Y (m)

1.5 1.5 -100 -100

1.0 1.0
0.5 0.5 -80 -80
0.0 0.0 -60 -60
0 0 10 10 20 20 3030 40
40 5050 60 60 70 70 80 80 -40
-40
-20
X (m) -20
X (m) 0
0

Figure 11. Bidimensional planethe

of RF
the power
RF power
levellevel distribution
at aat1.5-m
a 1.5-m height
for for
the the case
Figure
Figure 11. Bidimensional
11.
without Bidimensional plane
planeof of
human presence. the RF power distribution
level distribution at a height
1.5-m height case without
for the case
human presence.
without human presence.
Then, a single human body model was included in the scenario at different x-axis locations in
Then, a single human body model was included in the scenario at different x-axis locations in
Then,
order toa analyze
single human bodyofmodel
the influence was included
the surrounding in the scenario
environment at different
and the transmitter x-axis
distance locations
along the in
order to analyze the influence of the surrounding environment and the transmitter distance along the
ordertunnel. The considered
to analyze locations
the influence of thewere the following:
surrounding 20 m, 27 m,and
environment 34 m,
the41transmitter
m and 48 m.distance
In Figurealong
12, the
tunnel. The
theThe considered locations
corresponding were the following: 20 m, 27 m, 34 m, 41 m and 48As m. In Figure 12,
tunnel. consideredRFlocations
power levels
weredistribution results
the following: 20atm,1.5-m height
27 m, 34 m,are41presented.
m and 48 m.expected,
In Figure 12,
the corresponding RF power
the major RF distribution levels are
changes distribution results
mainly in the at 1.5-m
area behind theheight
locationare presented.
of the human body,As expected,
due
the corresponding RF power levels distribution results at 1.5-m height are presented. As expected,
the major
to theRF distribution
shielding effect. changes are mainly in the area behind the location of the human body, due to
the major RF distribution changes are mainly in the area behind the location of the human body, due
the shielding effect.
to the shielding effect.

Figure 12. Bidimensional planes of the RF power level distribution at a 1.5-m height for different
human being locations along the tunnel.

Figure 12.
Figure 12. Bidimensional
Bidimensional planes
planes of
of the
the RF
RF power
power level
level distribution
distribution at
at aa 1.5-m height for
1.5-m height for different
different
human being
human being locations
locations along
along the
the tunnel.
tunnel.
Sensors 2020, 20, 6710 15 of 24
Sensors 2020, 20, x FOR PEER REVIEW 15 of 24

In order to gain more insights into these results, Figure 13 shows the differences between the case
In order to gain more insights into these results, Figure 13 shows the differences between the
without human presence (i.e., Figure 11) and the different case studies considering a single human body
case without human presence (i.e., Figure 11) and the different case studies considering a single
model at distinct locations (i.e., Figure 12). As can be clearly observed from Figure 13, the presence
human body model at distinct locations (i.e., Figure 12). As can be clearly observed from Figure 13,
of a single worker within the tunnel in different locations leads to RF power level differences up to
the presence of a single worker within the tunnel in different locations leads to RF power level
18 dB. Although this difference is lower the higher the distance from the transmitter is, it has to be
differences up to 18 dB. Although this difference is lower the higher the distance from the transmitter
taken into account for an accurate range estimation. Thus, a margin of about 20 extra dBs should be
is, it has to be taken into account for an accurate range estimation. Thus, a margin of about 20 extra
added to the wireless communication range estimated in the previous subsection both for ZigBee and
dBs should be added to the wireless communication range estimated in the previous subsection both
LoRaWAN technologies.
for ZigBee and LoRaWAN technologies.

Figure 13. RF
Figure13. RFpower
powerlevel
leveldistribution
distributiondifferences between
differences betweenthethe
case without
case a human
without presence
a human and and
presence the
cases withwith
the cases human beings
human at different
beings locations
at different alongalong
locations the tunnel.
the tunnel.

Finally,
Finally,extra
extrasimulations
simulations were
were carried
carried out
out in
in order
orderto toassess
assessififthe
theobtained
obtained 20
20 dBs
dBs are
areenough
enough
when
when work teams of three people are within the utility tunnel. In Figure 14, a rendered view of
work teams of three people are within the utility tunnel. In Figure 14, a rendered view of the
the
inclusion of the three human body models within the created tunnel scenario for the
inclusion of the three human body models within the created tunnel scenario for the 3D-RL 3D-RL simulation
is presented.is presented.
simulation
Figure 15 presents the obtained estimation results in terms of the RF power level distribution
differences between the case without a human presence and the case study with a three-worker team
at different locations along the tunnel. From the results, a mild growth variation can be observed
when comparing the three-worker team case study with the single-worker case study, and therefore,
it can be stated that the previous estimated 20-dB margin is enough for a proper RF planification when
considering teams of up to three workers within the tunnel.
Sensors 2020, 20, 6710 16 of 24

Sensors 2020, 20, x FOR PEER REVIEW 16 of 24

Figure 14. Inclusion of 3 computational human body models in the considered scenario for the 3D-
RL simulation, emulating a maintenance team working in the utility tunnels.

Figure 15 presents the obtained estimation results in terms of the RF power level distribution
differences between the case without a human presence and the case study with a three-worker team
at different locations along the tunnel. From the results, a mild growth variation can be observed
when comparing the three-worker team case study with the single-worker case study, and therefore,
it can be stated that the previous estimated 20-dB margin is enough for a proper RF planification
Figure 14.
14. Inclusion
Inclusion of
of33computational
computationalhuman
humanbody
bodymodels
modelsinin the
the considered
considered scenario
scenario forfor
thethe 3D-
3D-RL
whenFigure
considering teams of up to three workers within the tunnel.
RL simulation,
simulation, emulating
emulating a maintenance
a maintenance team
team working
working in utility
in the the utility tunnels.
tunnels.

Figure 15 presents the obtained estimation results in terms of the RF power level distribution
differences between the case without a human presence and the case study with a three-worker team
at different locations along the tunnel. From the results, a mild growth variation can be observed
when comparing the three-worker team case study with the single-worker case study, and therefore,
it can be stated that the previous estimated 20-dB margin is enough for a proper RF planification
when considering teams of up to three workers within the tunnel.

Figure 15.
Figure RF power
15. RF power level
level distribution
distribution difference
differencebetween
betweenthethecase
casewithout
withoutpeople
peopleand
andthe
thecases
caseswith
witha
work team
a work teamofof
3 people atat
3 people different locations
different along
locations thethe
along tunnel.
tunnel.

In general, the obtained results regarding the study of the human presence within the tunnel show
that a margin of about approximately 20 extra dBs is required in order to preserve the received RF
power level above the sensitivity level, thus ensuring the wireless communication links between nodes,
even when up to a three-worker team is considered within the tunnel. This extra margin corresponds to
no need of extra range reduction estimated in the previous subsection analysis, as can be seen in Figure 16.
Figure 15. RF power level distribution difference between the case without people and the cases with
a work team of 3 people at different locations along the tunnel.
 In general, the obtained results regarding the study of the human presence within the tunnel
show that a margin of about approximately 20 extra dBs is required in order to preserve the received
RF power level above the sensitivity level, thus ensuring the wireless communication links between
nodes, even when up to a three-worker team is considered within the tunnel. This extra margin
Sensors 2020, 20, 6710 17 of 24
corresponds to no need of extra range reduction estimated in the previous subsection analysis, as can
be seen in Figure 16. However, since the sensitivity value of ZigBee is −100 dBm, in this case, a
However, of
reduction sincethethedistance
sensitivity value ofnodes
between −100 dBm,inin order
ZigBeeisis desirable this case,
to aassure
reduction of thewireless
reliable distance
between nodes is desirable in order to assure reliable wireless communication at
communication at every circumstance. This restriction was applied for the design of the WSNevery circumstance.
This restriction
presented in the was applied
following for the design of the WSN presented in the following subsection.
subsection.

ZigBee 868 MHz adjusted range (dBm)

2.0
Y (m)

1.5
1.0 -100
0.5 -80
0.0
-60
0 50 100 150 200 250 300
-40
X (m) -20
0

(a)
LoRaWAN adjusted range
(dBm)
2.0
Y (m)

1.5 -160
1.0
0.5 -140
0.0
-120
0 50 100 150 200 250 300
-100
X (m) -80
-60
-40
-20
0

(b)
Figure
Figure16.
16.Bidimensional
Bidimensionalplane
planeatata a3-m
3-mheight
heightrepresentation
representationofofthe
theRF
RFpower
powerlevel
leveldistribution
distributionafter
after
applying
applying the 20-dB margin due to the human presence for (a) ZigBee 868 MHz and (b)LoRaWAN.
the 20-dB margin due to the human presence for (a) ZigBee 868 MHz and (b) LoRaWAN.

4.4.Proposed
4.4. ProposedWSN
WSNDeployment
Deployment

InInthis
thissection,
section,an anoptimized
optimizedWSN WSNdeployment
deploymentisisproposedproposedconsidering
consideringthe themorphology
morphologyand and
topologyof
topology of the
the real underground
undergroundutility utilitytunnel
tunnelsystem.
system. InIngeneral,
general,thethe
network
networktopology
topologygains a great
gains a
importance
great importance for anfor
adequate deployment
an adequate and infrastructure
deployment design in this
and infrastructure typeinofthis
design complex
type underground
of complex
environments,
underground since their particular
environments, since their morphology (for instance,
particular morphology tunnels
(for instance,in the presented
tunnels in the case) could
presented
limitcould
case) the use of the
limit thestar
usetopology
of the star(thetopology
typical one (thefortypical
LoRaWAN).one forFrom the previous
LoRaWAN). From results, it can be
the previous
concluded
results, it canthat
beLoRaWAN
concluded technology
that LoRaWAN presents a much higher
technology presents range in terms
a much of distance
higher range in and terms better
of
robustness with human presence due to its better sensitivity. Besides, it
distance and better robustness with human presence due to its better sensitivity. Besides, it provides provides a high tolerance
afor interference,
high tolerance for which is highlywhich
interference, relevant in a confined
is highly relevant scenario with the
in a confined presence
scenario of the
with tubes, grids and
presence of
metal trays. However, considering the wireless system technologies’
tubes, grids and metal trays. However, considering the wireless system technologies’ capabilities andcapabilities and constraints,
and the realand
constraints, utility
the tunnel system
real utility conditions
tunnel systeminconditions
terms of connections, intersections intersections
in terms of connections, and infrastructure and
location possibilities,
infrastructure location there are several
possibilities, therekey arefactors
severalthatkeylead to the
factors election
that lead tooftheZigBee
electiontechnology
of ZigBeeto
be deployed
technology to within the considered
be deployed within the scenario.
consideredThe first LoRaWAN
scenario. limitation
The first LoRaWAN is thelimitation
reduced transfer
is the
capacitytransfer
reduced (up to 242 bytes),
capacity (upbut
to the
242 most
bytes), critical
but theone is the
most network
critical one istopology and its
the network deployment
topology and itsin
underground
deployment inconditions.
underground While the mesh While
conditions. topologytheprovided by ZigBee
mesh topology gives the
provided byoption
ZigBee of gives
deployingthe
option of deploying one single gateway, the star topology of LoRaWAN leads to the deploymentdue
one single gateway, the star topology of LoRaWAN leads to the deployment of several gateways of
to the morphology
several gateways due of the tunnels.
to the Furthermore,
morphology of thetaking
tunnels. intoFurthermore,
account that taking
the considered tunnels
into account thatlack
theof
communication
considered tunnels systems
lack ofinfrastructure
communication and systems
coverageinfrastructure
of external systems (such as of
and coverage theexternal
mobile network),
systems
the deployment
(such as the mobile of gateways
network),presents serious difficulties,
the deployment of gateways sincepresents
they need external
serious communication
difficulties, since theylinks
in order to send the data of the underground tunnel network to the cloud. Therefore, the possibility of
deploying a unique gateway is very advantageous.
For the proposed WSN deployment, a Digi International’s XBee Zigbee Gateway was selected,
allowing GPRS communications. The gateway was designed to be installed in the center of the tunnel
system for network performance optimization and located under a sewer, with the aim of facilitating the
Sensors 2020, 20, x FOR PEER REVIEW 18 of 24

need external communication links in order to send the data of the underground tunnel network to
the cloud. Therefore, the possibility of deploying a unique gateway is very advantageous.
SensorsFor the
2020, 20,proposed
6710 WSN deployment, a Digi International’s XBee Zigbee Gateway was selected,
18 of 24
allowing GPRS communications. The gateway was designed to be installed in the center of the tunnel
system for network performance optimization and located under a sewer, with the aim of facilitating
wireless communication with the mobile network. An efficient node of deployment was implemented
the wireless communication with the mobile network. An efficient node of deployment was
with specific nodes located in key corners and intersections, considering the particular morphology of
implemented with specific nodes located in key corners and intersections, considering the particular
the tunnels and the maximum distance between the nodes obtained in the previous subsection analysis.
morphology of the tunnels and the maximum distance between the nodes obtained in the previous
Digi’s XBee SX 868 RF Modules were employed for the sensor nodes. A schematic view of the WSN
subsection analysis. Digi’s XBee SX 868 RF Modules were employed for the sensor nodes. A schematic
deployment is depicted in Figure 17 representing the location of the ZigBee nodes and the gateway
view of the WSN deployment is depicted in Figure 17 representing the location of the ZigBee nodes
over the real utility tunnel system map.
and the gateway over the real utility tunnel system map.

Figure
Figure 17.
17. Schematic
Schematic view
view of
of the
the proposed
proposed ZigBee
ZigBee 868-MHz
868-MHz WSN
WSN deployment.
deployment.

4.5. Infrastructure
Infrastructure and Work Safety Monitoring Application

In order to guarantee the security and protection of the infrastructure infrastructure against unauthorized
vandalism or sabotage,
access, vandalism sabotage, as well as to ensure occupational safety at work, a surveillance and
monitoring systemsystemprototype
prototype waswasdeveloped
developedto allow
to infrastructure and workand
allow infrastructure safety remote
work monitoring.
safety remote
In general, these types of systems usually include WSN, closed-circuit television
monitoring. In general, these types of systems usually include WSN, closed-circuit television (CCTV) (CCTV) and an antifire
system,
and each of system,
an antifire them with eachits of
ownthemindependent
with its own communication
independent network. Although
communication they could
network. share
Although
the same
they communication
could share the same network, for securitynetwork,
communication and fault for
tolerance
securityreasons, normally,
and fault three different
tolerance reasons,
communication
normally, three networks are deployed, both
different communication wired and
networks arewireless.
deployed, Adequate
both wired precautions,
and wireless.suchAdequate
as the use
of shielded cables
precautions, such asor the
fiber optics,
use shouldcables
of shielded be incorporated when
or fiber optics, installing
should the completewhen
be incorporated wiring system
installing
to ensure
the the immunity
complete wiring systemto theto effects
ensureof electromagnetic
the immunity to interference.
the effects However, sometimes,
of electromagnetic due to cost
interference.
or deployment
However, reasons,due
sometimes, a wired
to costnetwork is not available,
or deployment reasons,requiring a wirelessiscommunication
a wired network network.
not available, requiring
aInwireless
the analyzed real case, anetwork.
communication fiber-optic Innetwork was available
the analyzed real case,for the city’s security
a fiber-optic networkcameras, although
was available for
no cameras were included inside the underground utility tunnel systems.
the city’s security cameras, although no cameras were included inside the underground utility tunnel If they were included,
they could
systems. be connected
If they to the they
were included, fiber-optic
could be network without
connected to thefurther complications.
fiber-optic network without Nevertheless,
further
either special infrared
complications. cameras or
Nevertheless, visible
either spectrum
special cameras,
infrared in conjunction
cameras or visible with presence cameras,
spectrum sensors thatin
activate the lighting system, should be included as well. This second option
conjunction with presence sensors that activate the lighting system, should be included as well. This is more efficient in terms
of infrastructure
second option is installation
more efficient costs due toofthe
in terms difference ininstallation
infrastructure price of thecosts
cameras,
due to but
theit difference
has a higher in
probability
price of the of failure because
cameras, but it has it involves
a higher more elements
probability (sensors
of failure and actuators).
because it involvesWhile,
more currently,
elements
no CCTVand
(sensors cameras are available
actuators). While, in the galleries
currently, of the underground
no CCTV utility tunnel
cameras are available in thesystem, a wireless
galleries of the
infrastructure utility
underground and worktunnelsafety monitoring
system, a wirelessapplication is proposed.
infrastructure and work safety monitoring application
Regarding infrastructure security and work safety monitoring, WSN are devoted to monitoring
is proposed.
possible leaks, zones or areas and personnel. When monitoring leaks (mainly water and gases), sensors
should be installed as close as possible of the possible leakage in order to facilitate that the detection
of the leak, if any, be as quick as possible. When monitoring zones/areas, detectors are regularly
distributed throughout the area according to the potential risk (and its assessment) of leakage. On the
Sensors 2020, 20, x FOR PEER REVIEW 19 of 24

Regarding infrastructure security and work safety monitoring, WSN are devoted to monitoring
possible leaks, zones or areas and personnel. When monitoring leaks (mainly water and gases),
Sensors
sensors 2020, 20, 6710
should be installed as close as possible of the possible leakage in order to facilitate that 19 ofthe
24

detection of the leak, if any, be as quick as possible. When monitoring zones/areas, detectors are
regularly
one hand, distributed
some combustiblethroughout gases,thesucharea accordingortomethane,
as hydrogen the potential risk (and
are lighter its assessment)
than air, so the detector of
leakage. On the one hand, some combustible gases, such as hydrogen or methane,
should be placed above the possible leak origin point or on the ceiling, where there is a high probability are lighter than
air,accumulations
of so the detectorof should
these be placed
gases. Onabove the possible
the other hand, otherleak origin pointas
gases such orpropane
on the ceiling, where
or butane there
tend to
is a high probability of accumulations of these gases. On the other hand, other
move through the lower parts and can cover large areas. Therefore, in order to detect this type of gas gases such as propane
or the
in butane tend
fastest to move
possible way, through the lower
it is necessary parts the
to place anddetectors
can coverinlarge areas.parts
the lower Therefore, in order to
of the installation,
detect this
avoiding thetype of gas
contact with inwater.
the fastest possible
Finally, personnelway,monitoring
it is necessary to place
is pivotal the detectors
in order to provideinworkthe safety
lower
parts of the installation, avoiding the contact with water. Finally, personnel
conditions, especially in underground environments. By means of a wireless network, surveillance monitoring is pivotal in
and maintenance teamwork can be controlled and monitored along the galleries using wearables ora
order to provide work safety conditions, especially in underground environments. By means of
wireless
worn bodynetwork,
sensors,surveillance
allowing free and maintenance
mobility and work teamwork
service. can be controlled and monitored along
the galleries using wearables or worn body sensors,
The development of an infrastructure-monitoring tool implies allowing free mobility
an adequateand work service.
analysis and design
The development of an infrastructure-monitoring tool implies
of the whole system but, also, an adequate choice of the types of sensors, their locations, an adequate analysis and design
their
of the whole system but, also, an adequate choice of the types of sensors,
configurations and their communication frequencies. In addition, the tool must allow the definition their locations, their
configurations and their communication frequencies. In addition, the
of alarm levels and the processes and procedures to be activated in case of alarm confirmation. tool must allow the definition
of alarm levelsmonitoring
Infrastructure and the processes
will focusand procedures
on both temperatureto beand activated
humidity in case
controlof to
alarm confirmation.
prevent mold and
Infrastructure monitoring will focus on both temperature and humidity
bacteria buildup and to control potential sources of contamination. To do this, the humidity control to prevent moldmustand
bacteria buildup and to control potential sources of contamination. To
be kept below 50% (desired values 30–50%). The same applies to dangerously high levels of carbon do this, the humidity must be
kept belowor50%
monoxide (desired
dioxide in thevalues 30–50%).compounds
air or nitrogen The same applies
(monoxide, to dangerously
dioxide and high other levels
nitrogenof oxides)
carbon
monoxide or dioxide in the air or
to prevent poisoning of the maintenance staff.nitrogen compounds (monoxide, dioxide and other nitrogen oxides)
to prevent poisoning
Considering the of
realthecase
maintenance
utility tunnelstaff.system and the proposed WSN deployment presented
Considering the real case utility
in the previous subsection, an infrastructure and tunnel system
work and the proposed
safety monitoring WSN deployment
application presented
prototype was
in the previous subsection, an infrastructure and work safety monitoring
developed in order to collect the data obtained by the sensors located in the gallery in real time application prototype was
developed
and in order
store them in a to collect the
database data obtained
allowing by the sensors
cloud computing. located inview
A schematic the gallery in real time tool
of the monitoring and
store them in a database allowing cloud computing. A schematic view
architecture is depicted in Figure 18. The prototype was developed using open-source tools such as the of the monitoring tool
architecture
Bootstrap is depicted
framework in for
[44] Figure 18. The prototype
the development of thewasweb developed
application using open-source
Grafana [45] for tools such as
the analytics
the Bootstrap
and interactiveframework
visualization [44]
andforthethe development
D3.js [46] JavaScriptof the webfor
library application
the production Grafana [45] for and
of dynamic the
analytics and interactive visualization and the D3.js [46] JavaScript library for the production of
interactive data visualizations in web browsers. The system also includes a web service in charge of
dynamic and interactive data visualizations in web browsers. The system also includes a web service
data collection, developed in Java, and a relational database (MySQL) [47].
in charge of data collection, developed in Java, and a relational database (MySQL) [47].

Figure 18. Architecture of the monitoring tool.

The design premises are security and use simplicity. The monitoring tool is responsive,
which allows its use on mobile devices (smartphones and tablets), workstations or video walls.
 allows its use on mobile devices (smartphones and tablets), workstations or video walls. It embeds
Grafana’s panels using iframes. On the one hand, there is a Grafana dashboard for the direct
monitoring of the gallery’s health parameters, which allows the visualization of the data provided by
the tunnel’s sensor network, which is depicted in Figure 19. On the other hand, the monitoring tool
(see Figure
Sensors 20), in addition to integrating the information provided by Grafana, includes
2020, 20, 6710 20 ofthe
24
monitoring of the activities performed inside the gallery. These activities include dead man
functionalities, monitoring of the task progress, access and exit points of the gallery, evacuation
Itprotocol
embedsand Grafana’s panelsin
instructions using
caseiframes. On the emergency
of emergency, one hand, there is a Grafana
acoustic and lightdashboard for the direct
signals activation, etc.
monitoring of the gallery’s health parameters, which allows the visualization
The tool includes tree-based and sunburst visualization to describe the workflows. of the data provided by
the tunnel’s
The data sensor network,
collected by which
the WSNis depicted
is stored,in Figure 19. On
processed andthe other hand,
analyzed in athe monitoring
private cloud. tool
For (see
that
Figure
purpose, an Apache CloudStack was implemented over hypervisors VMware ESXI managed byofa
20), in addition to integrating the information provided by Grafana, includes the monitoring
the activities
VMware performed
vSphere. inside
The the gallery.
resulting These activities
infrastructure include
allows the dead
agileman functionalities,
and monitoring
easy deployment and
of the task progress,
management of theaccess and exit
services. points of
Besides, it the gallery,
allows evacuationorchestration,
computing protocol and instructions in case
user and account
of emergency, and
management emergency
provides acoustic and light
data storage andsignals activation,
data analytics. Theetc. The toolcomputing
proposed includes tree-based and
cloud is easily
sunburst visualization
scalable and allows hottomigrations
describe the workflows.
according to the availability of the servers.

Figure 19. Time series monitoring by means of Grafana (iFrames).

Figure 19. Time series monitoring by means of Grafana (iFrames).

The data collected by the WSN is stored, processed and analyzed in a private cloud. For that
purpose, an Apache CloudStack was implemented over hypervisors VMware ESXI managed by a
VMware vSphere. The resulting infrastructure allows the agile and easy deployment and management
of the services. Besides, it allows computing orchestration, user and account management and provides
data storage and data analytics. The proposed computing cloud is easily scalable and allows hot
migrations according to the availability of the servers.
Sensors 2020, 20, 6710 21 of 24

Sensors 2020, 20, x FOR PEER REVIEW 21 of 24

Figure 20. Web-based monitoring tool.

Figure 20. Web-based monitoring tool.

5.
5. Conclusions
Conclusions
In
In this
this work,
work, thethe radio
radio wave
wave propagation
propagation assessment
assessment and and wireless
wireless system
system deployment
deployment for for real
real
urban
urban utility tunnels was described. The impact of large obstacle density, given by the inclusion of
utility tunnels was described. The impact of large obstacle density, given by the inclusion of
elements
elements such
such as as service
service trays
trays or
or handrails,
handrails, as as well
well as
as the
the presence
presence ofof occupational
occupational staff
staff within
within the
the
underground
underground utilityutilitytunnels,
tunnels,was
wasanalyzed
analyzedbyby means
means of of
anan
in-house 3D-RL
in-house code.
3D-RL The The
code. calibration and
calibration
validation of the 3D-RL algorithm for this singular scenario were presented
and validation of the 3D-RL algorithm for this singular scenario were presented by comparison with by comparison with
experimental
experimental measurements
measurements in in the
the real
real scenario,
scenario, showing
showing good
good agreement between them.
agreement between them. Once
Once the
the
optimal
optimal simulation
simulation parameters
parameters were
were obtained
obtained from
from the
the experimental
experimental measurements,
measurements, the the large
large and
and
small-scale
small-scale simulation
simulation results
results were
were presented
presented to to characterize
characterize thethe radio
radio wave
wave propagation
propagation within
within the
the
tunnel.
tunnel. A A complete
complete volumetric
volumetric analysis
analysis waswas performed
performed in in order
order to
to consider
consider thethe location
location of of wireless
wireless
transceivers
transceivers andand enable
enable them
them to
to be embedded within
be embedded within the elements of
the elements of the
the tunnels.
tunnels. The system-level
The system-level
analysis considering the conventional transceiver parameters (ZigBee
analysis considering the conventional transceiver parameters (ZigBee and LoRaWAN) was performed and LoRaWAN) was
performed
for an 868-MHzfor anoperational
868-MHz operational
frequency, frequency,
obtaining the obtaining the initial deployment
initial deployment planning
planning results. Theresults.
use of
The use of higher frequencies could lead to a higher signal attenuation, and then, the
higher frequencies could lead to a higher signal attenuation, and then, the coverage could be affected coverage could
be affected
within within the underground
the underground tunnel. Additionally,
tunnel. Additionally, an infrastructure
an infrastructure security andsecurity and work
work safety safety
monitoring
monitoring
application application
were developed were in
developed
which the ininformation
which the information
was storedwasin astored
privateinCloud.
a private
The Cloud. The
proposed
proposed methodology enables to consider specific propagation losses in complex
methodology enables to consider specific propagation losses in complex urban utility tunnels in order urban utility
tunnels in order to deploy wireless communication systems with optimal coverage, minimum
interference and an enhanced quality of service.
Sensors 2020, 20, 6710 22 of 24

to deploy wireless communication systems with optimal coverage, minimum interference and an
enhanced quality of service.

Author Contributions: Conceptualization, L.A. and P.L.-I.; methodology, L.A.; software, M.C.-E. and I.P.;
validation, M.C.-E.; formal analysis, M.C.-E., L.A. and P.L.-I.; investigation, M.C.-E. and L.A.; data curation,
M.C.-E. and I.P.; writing—original draft preparation, L.A. and M.C.-E.; writing—review and editing, P.L.-I., J.J.A.
and F.F.; visualization, E.A. and M.C.-E.; supervision, F.F.; wireless network planning, P.L.-I. and I.P.; cloud design
and development, J.J.A., monitoring system design and development J.J.A. and J.V. and funding acquisition, F.F.
All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded in part by the School of Engineering and Sciences at Tecnologico de Monterrey,
in part by Project RTI2018-095499-B-C31, funded by Ministerio de Ciencia, Innovación y Universidades, Gobierno
de España (MCIU/AEI/FEDER, UE) and, in part, by the European Union’s Horizon 2020 research and Innovation
program under grant agreement N◦ 774094 (Stardust-Holistic and Integrated Urban Model for Smart Cities).
Acknowledgments: The authors would like to acknowledge the support and collaboration of the School of
Engineering and Sciences and the Focus Research Group of Telecommunications and Networks at Tecnologico
de Monterrey.
Conflicts of Interest: The authors declare no conflict of interest. The statements made herein are solely the
responsibility of the authors.

References
1. Mahmoud, S.F. Wireless Transmission in Tunnels with Non-Circular Cross Section. IEEE Trans.
Antennas Propag. 2010, 58, 613–616. [CrossRef]
2. Hrovat, A.; Kandus, G.; Javornik, T. A Survey of Radio Propagation Modeling for Tunnels. IEEE Commun.
Surv. Tutor. 2014, 16, 658–669. [CrossRef]
3. Wu, Y.; Zheng, G.; Saleem, A.; Zhang, Y.P. An Experimental Study of MIMO Performance Using Leaky
Coaxial Cables in a Tunnel. IEEE Antennas Wirel. Propag. Lett. 2017, 16, 1663–1666. [CrossRef]
4. Forooshani, A.E.; Bashir, S.; Michelson, D.G.; Noghanian, S. A Survey of Wireless Communications and
Propagation Modeling in Underground Mines. IEEE Commun. Surv. Tutor. 2013, 15, 1524–1545. [CrossRef]
5. Moridi, M.A.; Sharifzadeh, M.; Kawamura, Y.; Jang, H.D. Development of wireless sensor networks for
underground communication and monitoring systems (the cases of underground mine environments).
Tunn. Undergr. Space Technol. 2018, 73, 127–138. [CrossRef]
6. Athanasiadou, G.E.; Tsoulos, G.V.; Zarbouti, D.A.; Valavanis, I.K. Optimizing Radio Network Planning
Evolution towards Microcellular Systems. Wirel. Pers. Commun. 2019, 106, 521–534. [CrossRef]
7. Xu, B.; Li, J.; Yang, Y.; Postolache, O.; Wu, H. Robust modeling and planning of radio-frequency identification
network in logistics under uncertainties. Int. J. Distrib. Sens. Netw. 2018, 14, 1550147718769781. [CrossRef]
8. Athanasiadou, G.E.; Fytampanis, P.; Zarbouti, D.A.; Tsoulos, G.V.; Gkonis, P.K.; Kaklamani, D.I.
Radio Network Planning towards 5G mmWave Standalone Small-Cell Architectures. Electronics 2020,
9, 339. [CrossRef]
9. González González, D.; Mutafungwa, E.; Haile, B.; Hämäläinen, I.; Poveda, H. A Planning and Optimization
Framework for Ultra Dense Cellular Deployments. Mobile Inf. Syst. 2017, 2017, 9242058. [CrossRef]
10. Hoppe, R.; Wölfle, G.; Jakobus, U. Wave Propagation and Radio Network Planning Software WinProp
added to the Electromagnetic Solver Package FEKO. In Proceedings of the 2017 International Applied
Computational Electromagnetics Society Symposium—Italy (ACES), Firenze, Italy, 26–30 March 2017.
11. Araujo, W.; Fogarolli, R.; Seruffo, M.; Cardoso, D. Deployment of small cells and a transport infrastructure
concurrently for next-generation mobile access networks. PLoS ONE 2018, 13, e0207330. [CrossRef]
12. Iqbal, M.; Naeem, M.; Anpalagan, A.; Ahmed, A.; Azam, M. Wireless Sensor Network Optimization:
Multi-Objective Paradigm. Sensors 2015, 15, 17572–17620. [CrossRef] [PubMed]
13. Sanchez-Iborra, R.; Sanchez-Gomez, J.; Ballesta-Viñas, J.; Cano, M.-D.; Skarmeta, A.F. Performance Evaluation
of LoRa Considering Scenario Conditions. Sensors 2018, 18, 772. [CrossRef] [PubMed]
14. Sulaiman BenSaleh, M.; Saida, R.; Hadj Kacem, Y.; Abid, M. Wireless Sensor Network Design Methodologies:
A Survey. J. Sens. 2020. [CrossRef]
15. Priyadarshi, R.; Gupta, B.; Anurag, A. Deployment techniques in wireless sensor networks: A survey,
classification, challenges, and future research. J. Supercomput. 2020, 76, 7333–7373. [CrossRef]
Sensors 2020, 20, 6710 23 of 24

16. Gao, Y.; Wang, J.; Wu, W.; Sangaiah, A.K.; Lim, S.-J. Travel Route Planning with Optimal Coverage in Difficult
Wireless Sensor Network Environment. Sensors 2019, 19, 1838. [CrossRef]
17. Iliodromitis, A.; Pantazis, G.; Vescoukis, V. 2D wireless sensor network deployment based on Centroidal
Voronoi Tessellation. AIP Conf. Proc. 2017, 1836, 020010.
18. Kanaris, L.; Sergiou, C.; Kokkinis, A.; Pafitis, A.; Antoniou, N.; Stavrou, S. On the Realistic Radio and
Network Planning of IoT Sensor Networks. Sensors 2019, 19, 3264. [CrossRef]
19. Delogne, P. EM Propagation in Tunnels. IEEE Trans. Antennas Propag. 1991, 39, 401–406. [CrossRef]
20. Mariage, P.; Lienard, M.; Degauque, P. Theoretical and Experimental Approach of the Propagation of High
Frequency Waves in Road Tunnels. IEEE Trans. Antennas Propag. 1994, 42, 75–81. [CrossRef]
21. Chen, S.; Jeng, S. SBR Image Approach for Radiowave Propagation in Tunnels with and without Traffic.
IEEE Trans. Antennas Propag. 1996, 45, 570–578.
22. Celaya-Echarri, M.; Azpilicueta, L.; López-Iturri, P.; Picallo, I.; Aguirre, E.; Astrain, J.J.; Villadangos, J.;
Falcone, F. Wireless Channel Characterization and System Analysis of Complex Utility Tunnel Environments.
In Proceedings of the 6th International Electronic Conference on Sensors and Applications, 15–30 November
2019. Available online: https://www.mdpi.com/2504-3900/42/1/53 (accessed on 23 November 2020).
23. Sun, Z.; Akyildiz, I. Channel modeling and analysis for wireless networks in underground mines and road
tunnels. IEEE Trans. Commun. 2010, 58, 1758–1768. [CrossRef]
24. Martelly, R.; Janaswamy, R. An ADI-PE Approach for Modeling Radio Transmission Loss in Tunnels.
IEEE Trans. Antennas Propag. 2009, 57, 1759–1770. [CrossRef]
25. Sakai, K.; Koshiba, M. Analysis of electromagnetic field distribution in tunnels by the boundary-element
method. IEE Proc. 1990, 137, 202–208. [CrossRef]
26. Bernardi, P.; Caratelli, D.; Cicchetti, R.; Schena, V.; Testa, O. A Numerical Scheme for the Solution of the
Vector Parabolic Equation Governing the Radio Wave Propagation in Straight and Curved Rectangular
Tunnels. IEEE Trans. Antennas Propag. 2009, 57, 3249–3257. [CrossRef]
27. Didascalou, D.; Schäfer, T.M.; Weinmann, F.; Wiesbeck, W. Ray-Density Normalization for Ray-Optical
Wave Propagation Modeling in Arbitrarily Shaped Tunnels. IEEE Trans. Antennas Propag. 2000, 48,
1316–1325. [CrossRef]
28. Zhang, Y.P.; Hwang, Y. Enhancement of rectangular tunnel waveguide model. In Proceedings of the Asia
Pacific Microwave Conference, Hong Kong, China, 2–5 December 1997.
29. Mohammed, F.H.; Samir, F.M. A High-Order Compact-FDTD Algorithm for Electrically Large Waveguide
Analysis. IEEE Trans. Antennas Propag. 2008, 56, 2589–2598.
30. Rana, M.; Sanagavarapu, A. Segmented-Locally-One-Dimensional-FDTD Method for EM Propagation Inside
Large Complex Tunnel Environments. IEEE Trans. Magn. 2012, 48, 223–226. [CrossRef]
31. Lopez-Iturri, P.; Aguirre, E.; Azpilicueta, L.; Astrain, J.J.; Villadangos, J.; Falcone, F. Challenges in Wireless
System Integration as Enablers for Indoor Context Aware Environments. Sensors 2017, 17, 1616. [CrossRef]
32. Azpilicueta, L.; Astrain, J.J.; Lopez-Iturri, P.; Granda, F.; Vargas-Rosales, C.; Villadangos, J.; Perallos, A.;
Bahillo, A.; Falcone, F. Optimization and Design of Wireless Systems for the Implementation of Context Aware
Scenarios in Railway Passenger Vehicles. IEEE Trans. Intell. Transp. Syst. 2017, 18, 2838–2850. [CrossRef]
33. López-Iturri, P.; Aguirre, E.; Azpilicueta, L.; Astrain, J.J.; Villadangos, J.; Falcone, F. Implementation
and Analysis of ISM 2.4GHz Wireless Sensor Network System in Judo Training Venues. Sensors 2016,
16, 1247. [CrossRef]
34. Córdoba, A.; Astrain, J.J.; Villadangos, J.; Azpilicueta, L.; López-Iturri, P.; Aguirre, E.; Falcone, F. SesToCross:
Semantic Expert System to Manage Single-Lane Road Crossing. IEEE Trans. Intell. Transp. Syst. 2017, 18,
1221–1233. [CrossRef]
35. Azpilicueta, L.; López-Iturri, P.; Aguirre, E.; Martínez, C.; Astráin, J.J.; Villadangos, J.; Falcone, F.
Evaluation of Deployment Challenges of Wireless Sensor Networks at Signalized Intersections. Sensors 2016,
16, 1140. [CrossRef] [PubMed]
36. Azpilicueta, L.; Rawat, M.; Rawat, K.; Ghannouchi, F.; Falcone, F. Convergence Analysis in Deterministic 3D
Ray Launching Radio Channel Estimation in Complex Environments. ACES J. 2014, 29, 256–271.
37. Azpilicueta, L.; López-Iturri, P.; Aguirre, E.; Falcone, F. An accurate UTD Extension to a Ray Launching
Algorithm for the Analysis of Complex Indoor Radio Environments. J. Electromagn. Waves Appl. 2016,
30, 43–60. [CrossRef]
Sensors 2020, 20, 6710 24 of 24

38. Changsen, Z.; Yan, M. Effects of cross section of mine tunnel on the propagation characteristics of UHF radio
wave. In Proceedings of the 7th International Symposium on Antennas, Propagation EM Theory ISAPE ’06,
Guilin, China, 26–29 October 2006; pp. 1–5.
39. Wang, S. Radio wave attenuation character in the confined environments of rectangular mine tunnel.
Mod. Appl. Sci. 2010, 4, 65–70. [CrossRef]
40. Cheng, L.; Zhang, P. Influence of dimension change on radio wave propagation in rectangular tunnels.
In Proceedings of the 5th International Conference on Wireless Communications, Networking and Mobile
Computing, Beijing, China, 24–26 September 2009; pp. 1–3.
41. Rappaport, T. Wireless Communications: Principles and Practice, 2nd ed.; Prentice Hall PTR: Upper Saddle
River, NJ, USA, 2002.
42. Obayashi, S.; Zander, J. A body-shadowing model for indoor radio communication environments. IEEE Trans.
Antennas Propag. 1998, 46, 920–927. [CrossRef]
43. Aguirre, E.; Arpon, J.; Azpilicueta, L.; de Miguel Bilbao, S.; Ramos, V.; Falcone, F.J. Evaluation of
electromagnetic dosimetry of wireless systems in complex indoor scenarios with human body interaction.
Prog. Electromagn. Res. B 2012, 43, 189–209. [CrossRef]
44. Bootstrap. Available online: https://getbootstrap.com/ (accessed on 30 July 2020).
45. Grafana. Available online: https://grafana.com/ (accessed on 30 July 2020).
46. D3 Data-Driven Documents. Available online: https://d3js.org/ (accessed on 30 July 2020).
47. MySQL. Available online: https://www.mysql.com/ (accessed on 30 July 2020).

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional
affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).