2366 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO.

2, APRIL 2019

LoRaWAN Network: Radio Propagation Models

and Performance Evaluation in Various
Environments in Lebanon
Rida El Chall , Samer Lahoud, and Melhem El Helou, Senior Member, IEEE

Abstract—Recently, LoRaWAN has emerged as a promising diverse requirements. LPWANs typically operate in licensed
technology for the Internet of Things (IoT), owing its ability and unlicensed frequency bands. Many LPWAN technologies
to support low-power and long-range communications. However, have been investigated by different standards and industrial
real-world deployment and network optimization require accu-
rate path-loss (PL) modeling, so as to estimate network coverage, consortia, including LoRa, Sigfox, NB-IoT, ECGSM-IoT, ran-
performance, and profitability. For that reason, in this paper, dom phase multiple access (RPMA), weightless, DASH7
LoRaWAN radio channel is investigated in the 868 MHz band. alliance, etc. An overview and comparison of these emerg-
Extensive measurement campaigns were carried out in both ing LPWAN technologies have been presented in [4]–[7]. In
indoor and outdoor environments at urban and rural locations particular, LoRaWAN is one of the most deployed LPWAN
in Lebanon (Saint Joseph University of Beirut campus, Beirut
city, and Bekaa valley). Based on empirical results, PL mod- technology, gaining greater interest from the research and
els are developed for LoRaWAN communications and compared industrial communities. From theoretical aspects, many stud-
with widely used empirical models. Moreover, the performance ies have focused on the performance and characteristics of
and the coverage of LoRaWAN deployment are evaluated based LoRaWAN communications. An overview of the capabilities
on real measurements. The results show that the proposed PL and the limitations of LoRaWAN has been presented in [8].
models are accurate and simple to be applied in Lebanon and
other similar locations. Coverage ranges up to 8 and 45 km were Theoretical evaluation of the capacity and scalability has also
obtained in urban and rural areas, respectively. This reveals been performed [9], [10]. Moreover, an adaptive configuration
the reliability of this promising technology for long-range IoT of LoRa networks has been proposed for scalable IoT deploy-
communications. ments [11]. The impact of physical settings, such as spreading
Index Terms—Indoor, Internet of Things (IoT), long range, factor, coding rate, and bandwidth on the DR and time on
LoRa, LoRaWAN, low-power wide area networks (LPWANs), air have been investigated [12]. All these works concluded
outdoor, path-loss (PL) model, rural, urban. that LoRaWAN systems should be carefully configured and
dimensioned to achieve a good tradeoff between scalability
and efficiency.
I. I NTRODUCTION On the other hand, physical and link layer performance
of LoRa/LoRaWAN have been evaluated experimentally by
HE INTERNET of Things (IoT) is a promising paradigm
T that is rapidly evolving to make any device part of
the Internet environment. According to Cisco, it is expected
field tests [13]–[15]. The experimental tests have been con-
ducted in various real-world environments ranging from
indoor [16]–[18] and urban/suburban [19]–[23] to mar-
that more than 50 billion devices will be connected through
itime [20], rural [23] and mountain [24] scenarios. The
radio communications by 2020 [1]. IoT devices will be used
experiments show communication ranges from 10 to 30 km
in wide range of applications including security, industrial
in rural areas and 2 to 8 km in urban areas. Furthermore, the
monitoring, smart homes, smart cities, smart agriculture, etc.
impact of environmental factors, such as temperature and veg-
Comprehensive surveys on the emerging IoT technologies and
etation has been investigated [24], [25]. It has been shown
their challenges have been reported in [2] and [3].
that vegetation and higher temperature significantly reduce
The main characteristics and requirements of IoT applica-
communication ranges. In case of indoor operation, results
tions are long range, low data rate (DR), low energy consump-
showed that LoRa can achieve good coverage in the entire
tion, and cost-effectiveness. Low-power wide area networks
buildings of Oulu Campus University [17]. However, connec-
(LPWANs) have been therefore developed to meet these
tivity issues and high packet losses might be encountered in
Manuscript received December 2, 2018; revised February 21, 2019; the basement [16], whereas the best coverage is achieved when
accepted March 10, 2019. Date of publication March 21, 2019; date of cur- the receiver is located on the roof rather than in the base-
rent version May 8, 2019. This work was supported in part by the National ment [18]. Different kinds of applications have been tested
Council for Scientific Research in Lebanon CNRS-L and in part by the Saint
Joseph University of Beirut. (Corresponding author: Rida El Chall.) using LoRa, such as vehicle to grid communications [26],
The authors are with the Ecole Supérieure d’Ingénieurs de Beyrouth, health monitoring [27], [28], and river monitoring [29].
Faculty of Engineering, Saint Joseph University, Beirut 1107 2050, The key parameters to optimize network performance prior
Lebanon (e-mail: [email protected]; [email protected];
[email protected]). to real deployment is to correctly predict the coverage and to
Digital Object Identifier 10.1109/JIOT.2019.2906838 carefully adjust antenna heights of installation sites. Therefore,
2327-4662 c 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 EL CHALL et al.: LoRaWAN NETWORK 2367

precise modeling of radio propagation characteristics is very The remainder of this paper is organized in the following
crucial for LoRaWAN network planning and optimization. way. Section II provides an overview of LoRa and LoRaWAN
Radio propagation characteristics have been widely studied technology. Section III presents channel modeling principles
over the world. Numerous field measurements have been car- for both indoor and outdoor environments and reviews the
ried out in various indoor and outdoor environments in the most widely used PL models. Section IV describes measure-
context of cellular and wireless sensor networks. Generally, ment campaigns carried out in Lebanon at different environ-
the path-loss (PL) is impacted by many factors, such as dis- ments. In Section V, the results are analyzed and discussed.
tance, frequency band, average antenna heights, geography The PL models are therefore developed and compared with
and terrain in terms of obstacles, buildings, hills, mountains, others. Finally, the conclusions are drawn in Section VI.
people, etc. However, for the indoor environment, additional
factors need to be considered, such as floor plans, walls, and II. L O R A AND L O R AWAN OVERVIEW
type and thickness of building materials. Several PL models This section provides an overview of LoRaWAN technol-
applicable to outdoor environments for the [800–1800] MHz ogy. First, LoRa physical layer is presented followed by a
and [2.5–5] GHz bands are developed by research insti- description of LoRaWAN link protocol and basic network
tutes and standard organizations, e.g., Okumura–Hata, Cost architecture.
231-Hata, Bertoni-Walfisch, International Telecommunications
Union (ITU) Advanced, WINNER II, WINNER+, and 3GPP A. LoRa PHY Layer
spatial channel model [30]–[35]. Similarly, many indoor prop-
agation models are proposed in literature, e.g., ITU-R P1238, LoRa is a physical layer technology developed and com-
IEEE 802.11n, 3GPP, Cost 231 multiwall, and Motley– mercialized by Semtech for long-range and low-power com-
Keenan [31], [35]–[38]. Although many of these PL models munications [42]. It is a derivative of chirp spread spectrum
are widely used today, they are not intended for long-range (CSS) modulation with integrated forward error correction
LoRaWAN network operating at 868 MHz band. For this rea- (FEC) [43]. CSS technique allows to increase the receiver
son, a channel attenuation model has been derived from the sensitivity, enabling long communications ranges. It actually
data measurements in the city of Oulu, Finland [20]. In [39], allows to correctly decode transmissions 19.5 dB below the
an urban PL model is proposed for LoRa links in Dortmund, noise floor (maximum link budget of 150 dB) [43].
Germany. An empirical evaluation of the indoor propagation Generally, LoRa is characterized by five configured param-
performance of LoRa at Glasgow Caledonian University has eters: carrier frequency (CF), bandwidth (BW), transmission
been presented, showing that the multiwall model has the best power (Ptx ), spreading factor (SF), and coding rate (CR).
overall performance [40]. However, other factors have to be These parameters can be tuned for a tradeoff among several
considered, such as end-device antenna heights and harsh envi- features: DR, transmission range, robustness to interference,
ronmental conditions due to the variety of IoT applications. and energy consumption. LoRa operates on the sub-1GHz
Moreover, irregular terrain profile and topography variation bands, e.g., 433, 868, or 915 MHz ISM bands, depending
have to be considered, e.g., hilly Mediterranean and mountain on the region in which it is deployed. In Europe, 433 and
topography in Lebanon. Consequently, more field measure- 868 MHz are available, with 868 MHz being most commonly
ments are required to accurately elaborate and validate radio used. The bandwidth can be 125, 250, and 500 kHz. A higher
propagation models for optimal deployment of LoRaWAN in bandwidth corresponds to a higher DR, but to a lower sen-
both indoor and outdoor environments. sitivity. The transmit power can be configured based on the
Focusing on those important issues, this paper presents an region and the band used for transmissions. The SF represents
in-depth study of the radio propagation characteristics, con- the ratio between symbol rate and chip rate. The supported
sidering different environments and antenna heights in the SF values range from 7 to 12. A higher SF makes the sig-
868 MHz band in Lebanon. Extensive measurement campaigns nal more robust to noise (increase the sensitivity and range)
were carried out in both indoor and outdoor environment but decreases the DR. Note that each of the available SFs are
at rural and urban areas. The set of measurement data is orthogonal, enabling multiple signals to be transmitted on the
publicly available in [41]. In particular, indoor and out- same channel simultaneously [43]. LoRa uses also FEC to per-
door tests were performed in the Saint Joseph University form error detection and correction. Coding rate can be set to
(USJ) of Beirut campus. Urban and rural tests were con- 4/(CR + 4) with CR ∈ {1, 2, 3, 4}. Depending on the config-
ducted in Beirut city and Bekaa valley, respectively. Based uration of the physical layer parameters, the bit rate ranges
on empirical results, we derive PL models for Lora com- from 0.3 to 50 kb/s as shown in Table I. It is worth noting
munications at the 868 MHz band under various parame- that different LoRa configurations are referred to as DRs in
ters. The proposed models are compared with some widely LoRaWAN specification.
used empirical indoor/outdoor PL models, to determine their
accuracy. We show that the proposed PL models fit mea- B. LoRaWAN Link Layer
surements with more accuracy and simplicity compared to LoRaWAN is the upper layer protocol for LoRa, described
other models. Moreover, the performance and coverage of in an open specification and developed by the LoRa
LoRaWAN deployment is evaluated. The results show cov- Alliance [44]. LoRaWAN relies on an ALOHA-based MAC
erage ranges up to 8 and 45 km in urban and rural areas, protocol to reduce the complexity of end-devices in access-
respectively. ing the channel. The network architecture is a star-of-stars

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 2368 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

TABLE I
DR AND S ENSITIVITY OF D IFFERENT L O R A C ONFIGURATION reflection, diffraction, refraction, and scattering components
PARAMETERS FOR THE 868 MHz BAND resulting from buildings, trees, hills, and other obstacles.
Channel measurements aim to understand the channel
behavior and to develop realistic and trustworthy channel
models. Generally, propagation models include deterministic
models and empirical models. Deterministic models are very
complex since they require detailed knowledge of location,
dimension, and physical parameters of every obstacles in the
area. However, in empirical models, the parameter values are
derived by fitting measurement data to an appropriate function
for a particular environment. This gives more generic model
that can be used by systems operating in similar areas. Herein,
empirical propagation models used in this paper are presented.
These models will be adapted according to our measurements
in Lebanon, as discussed in Section V.

A. Free-Space PL Model
Free-space model is a baseline model that provides a mea-
sure of PL when the transmitter and receiver are within
Fig. 1. LoRaWAN architecture. line-of-sight (LOS) range without any obstacles between them.
It is based on the Friis’ free-space transmission equation, given
in the logarithmic domain as follows:
topology and consists of three entities: 1) end-devices (EDs);
PLFS (d)[dB] = 20 log10 (f ) + 20 log10 (d) + 32.44 (1)
2) gateways (GWs); and 3) a network server as illustrated in
Fig. 1. EDs communicate with GWs using single-hop LoRa where d is the distance between the transmitter and the receiver
communication. The GW simply relays received messages to a in km, and f is the frequency in MHz.
central network server via an IP backbone. The central network
server manages the network access and provides mobility, B. Log-Distance PL Model
frame control as well as security functions.
The log-distance propagation model, also referred as one-
LoRaWAN introduces three classes: 1) Class A (the
slop model, is a general PL model that has been used in a
default); 2) Class B; and 3) Class C (both optional). Class A
large number of indoor and outdoor environments. It assumes
supports basic bi-directional communications, where each
that PL varies exponentially with distance according to the
uplink transmission is followed by two short receive win-
following equation:
dows for transferring data to the ED. Class B extends Class A
by adding extra receive windows at scheduled times. The PL(d)[dB] = 10n log10 (d/d0 ) + PL0 + Xσ (2)
GW then periodically broadcasts beacons to maintain the
synchronization of EDs, while EDs of Class C have almost where n is the PL exponent, d is the distance between
continuously receive windows when not transmitting. the transmitter and the receiver, and PL0 is the PL at a
Three default channels (868.1, 868.3, and 868.5 MHz) must reference distance d0 . Shadow fading is represented by a zero-
be implemented in every LoRaWAN network. In practice, a mean Gaussian random variable Xσ with standard deviation σ
physical channel is chosen on a pseudo-random basis, based (in dB). The PL parameters are derived from a regression or
on regularity requirements defined by ETSI [45]. Each sub- fitting curve over the measured data and depend upon the envi-
band has specific requirements regarding maximum effective ronment. For instance, n = 2.32 and PL0 = 128.95 in the city
radiated power (ERP), and duty cycle limits. For the majority of Oulu (Finland) [20], whereas n = 2.65 and PL0 = 132.25
of sub-bands, the ERP is 25 mW (14 dBm) and the duty cycle in the city of Dortmund (Germany) [39].
limits vary between 0.1% and 1%. Furthermore, LoRaWAN
specification defines an adaptive DR (ADR) scheme which C. Multiwall-and-Floor Model
enables the server to set the spreading factor of each node, In order to characterize the PL within buildings, the
maximizing the battery lifetime while optimizing the overall most accurate approach is to consider additional attenuation
network capacity. The maximum MAC payload can range from incurred by walls and floors. Thus, PL is modeled as
59 to 250 bytes depending on LoRa configurations. LoRaWAN
overhead per packet is 13 bytes. PL(d)[dB] = 10n log10 (d/d0 ) + PL0 + WAF + FAF (3)
where WAF and FAF are wall and floor attenuation factors
III. R ADIO P ROPAGATION M ODELS based on the number of traversed walls nw and floors nf
Radio waves take several ways when traveling between between the transmitter and the receiver, respectively. These
transmitter and receiver, resulting in a significant loss in the factors can be extracted by ray tracing techniques or empir-
received signal. This loss may be due to many effects including ical measurements. Actually, the penetration losses of walls

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 EL CHALL et al.: LoRaWAN NETWORK 2369

and floors depend on several factors, such as frequency, thick-

ness, and material of obstacles. Measurement results show
either a linear or a nonlinear relation between the traversed
walls or floors and the penetration loss. Examples of MWF
models are Cost 231-MWF and Motley–Keenan. In both mod-
els, WAFis proportional to the number of penetrated walls,
WAF = nwi Lwi , where nwi is the number of walls of type
i and Lwi is the penetration loss for the wall of type i. In
Motley–Keenan model, FAF is expressed by a floor loss fac-
tor Lf multiplied by the number of floors (FAF = nf Lf ). Cost Fig. 2. Platform used in the measurement campaigns.
231-MWF model provides a nonlinear function of floor atten-
uation which increases more slowly as per additional floors:
((n +2)/(nf +1)−b) IV. M EASUREMENT C AMPAIGNS
FAF = Lf nf f , where b is an empirical constant
selected to obtain a suitable fit of the measured data. A similar This section describes measurement campaigns carried out
approach is taken by ITU-R model [36], where only the floor in Lebanon in order to study the PL characteristics of the
loss is considered. 868 MHz band under different ED antenna heights and
environment conditions. The experimental platform is first
described followed by a presentation of the different sites.
D. Summary of PL Models
Looking into the literature, a lot of field measurements
were carried out to model the PL in different environments A. Measurement Setup
and frequency bands. The characteristics of the most used PL The experimental platform used in our measurement cam-
models are shown in Table II. Their applicability ranges and paigns is depicted in Fig. 2. Pycom LoPy with PyTrack
their environments are specified. Okumura–Hata model is an expansion board was used as LoRa ED [46], powered by
empirical method, based upon extensive measurements made 3.7-V rechargeable lithium battery. The LoPy has an inte-
in Tokyo city between 200 MHz and 2 GHz [30]. grated LoRa SX1272 transceiver and an additional WiFi
This model specifies four generic categories of environ- transceiver. PyTrack module includes an embedded global
ments: 1) large cities; 2) medium-small cities; 3) suburban positioning system (GPS) used to obtain the location of the
areas; and 4) rural areas. It assumes that there are no dominant ED. Kerlink Wirnet Station was used as the GW which is able
obstacles between the base station and the mobile, and that the to receive LoRa frames from −20 to −141 dBm, depending
terrain profile changes slowly. Cost 231-Hata model extends on the LoRa BW and SF [47]. The GW was connected to
Okumura–Hata model for medium to small cities to cover the the network server provided by an open source LoRa server
band [1500–2000] MHz [31]. The ITU has also defined a set of (https://www.loraserver.io.). Both GW and ED use an omni-
channel models for the evaluation of advanced mobile system directional dipole antenna of 3 dBi gain. Upon the reception
proposals [32]. The ITU advanced models apply the same of each frame, the GW provided the received signal strength
approach as 3GPP/3GPP2 models. These models are primar- indicator (RSSI), the signal-to-noise ratio (SNR), and the pay-
ily based on extensive measurement campaigns mostly done load message. These received parameters were recorded on the
within the framework of the European projects WINNER I server side for further analysis and processing. The quality of
and WINNER II [33]. Scenarios specified in 3GPP include reception can be monitored in real-time with an MQTT Web
indoor hotspot (InH), urban microcell (UMi), urban/suburban client application.
macrocell (UMa/SMa), and rural macrocell (RMa) [35]. The The ED was configured to send a packet every 10 s, with
3GPP developed models can be applied in the frequency range GPS coordinates included in the payload field. The packets
of 2–6 GHz (0.46–6 GHz for RMa) for different antenna also include a sequential number, in order to identify the
heights. These models consider further parameters, such as packet loss. No mechanisms for control and automatic retrans-
the average building height h and the street width W. For missions were used. The transmit power was set to 14 dBm,
indoor environment, ITU-R model and Cost-231 MWF were SF to 12 (to achieve the best receiver sensitivity), and the BW
considered. to 125 kHz, using the three default channels (868.1, 868.3, and
It should be mentioned that these models have been 868.5 MHz). The packet payload was 50 and 37 bytes includ-
proposed in the context of cellular and wireless sensor ing 13 bytes MAC header in case of indoor and outdoor setups,
networks. Their applicability ranges show some limitations in respectively. The measurement setup parameters are listed in
terms of antenna heights and terrain profiles. For instance, the Table III. During all the measurements the position of GWs
base station antenna height has a maximum of 200 m, while was fixed, whereas the ED was moved to different locations.
the user has a minimum antenna height of 1 m, assuming a For urban and indoor scenarios, the GW was installed at the
flat terrain. This motivates additional studies for PL modeling rooftop of the ESIB-USJ engineering building at a height of
to investigate the validity of these models for long-range 12 m above ground level, being at a final altitude of 260 m
LoRAWAN deployment, considering low antenna heights for above sea-level. For the rural area, the GW was placed on the
the user, and irregular terrain profile as presented in the rooftop of Kefraya tower at a height of 12 m above ground
following sections. level, being at a final altitude of 970 m above sea-level (Fig. 3).

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 2370 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

TABLE II
S UMMARY OF PL M ODELS

TABLE III
M EASUREMENT S ETUP PARAMETERS Lrx represent the transmitter and receiver losses due to cables,
assumed negligible in this paper. The shadowing is classically
characterized by a zero-mean Gaussian variable with standard
deviation σ . This standard deviation describes the dispersion
between measured and expected PLs.
Furthermore, the performance of LoRaWAN was evaluated
during experiments using additional metrics. The SNR indi-
cates the quality of the received signal. The packet delivery
ratio (PDR) which is the ratio between received and transmit-
ted packets, indicates the reliability of communications. The
coverage range is the measured distance between GW and ED,
when the PDR was above a certain threshold (e.g., 90%).

B. Experimental Environments
Our experiments were performed in three locations in
For outdoor usage, various antenna heights were considered Lebanon, with different environmental characteristics as sum-
for the ED to reflect different envisioned IoT applications. The marized in Table IV.
ED was mounted on a tripod at three heights: near-ground at Indoor and outdoor tests were conducted in the campus of
20 cm, 1.5 m, and 3 m as shown in Fig. 3. However, in the ESIB-USJ. The urban area is located in the city of Beirut,
indoor environment, there is no interest to consider different whereas rural area is situated in Bekaa valley. Indoor mea-
ED antenna heights due to the penetration losses through walls surements were performed in multifloor buildings connected
and floors. Moreover, drive tests were performed in urban and together by indoor passages on the ESIB-USJ campus as
rural areas, where ED was attached to the roof-rack of a car shown in Fig. 4. These buildings, mostly built of concrete
approximately at 1.7 m height of the ground (Fig. 3). During and steel, differ in geometry and usage, with dimensions of
the drive test, the vehicle speed was around 30–40 km/h. 110 m × 50 m. Different conditions and constraints (indoor
In order to calculate the PL, the received signal strength and deep-indoor) were considered, such as offices, classes with
Prx is estimated based on both the SNR and the RSSI [47]. If obstacles, corridor, etc. We considered 70 locations of ED
SNR > 0, Prx = RSSI, otherwise, Prx = RSSI + SNR. The with distance of 5–110 m on four different floors ranging
PL is therefore computed as follows: from the basement to the fourth floor. At each location, 20
PL = Ptx − Prx + Gtx + Grx − Ltx − Lrx (4) measurements were taken. We have used the available floor
plan while conducting our measures to mark the positions of
where Ptx is the transmit power in dBm. Gtx and Grx are the ED as illustrated in Fig. 4. This helps us to compute the dis-
transmitter and the receiver antenna gains, respectively. Ltx and tance and to count the number of floors and walls or obstacles

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 EL CHALL et al.: LoRaWAN NETWORK 2371

(a) (b) (c) (d)

Fig. 3. Locations of Lora gateways and end-devices. (a) ESIB-USJ GW, (b) Kefraya GW, (c) Pycom Lopy EDs mounted on a tripod at three heights, and
(d) ED mounted on the roof-rack of a car.

TABLE IV
C HARACTERISTICS OF THE E XPERIMENTAL S ITES

Fig. 4. Map of the indoor environment at USJ campus showing the

measurement locations of EDs.

between the transmitter and the receiver. A total amount of

Fig. 5. Map of the outdoor environment at USJ campus showing the
1400 measurements was collected. measurement locations of EDs.
The outdoor test in USJ campus was realized by moving
the ED, emitting continuously around the campus as shown
in Fig. 5. The campus area is about 200 m north to south
and 280 m east to west. It is a mixed environment composed topography with some terrain elevation changes. The effective
of a combination of medium building heights, green spaces terrain height is about 60 m. We considered 35 fixed point
with trees, having medium population density. The distance measurements around the GW with a distance up to 9 km as
between ED and GW varied in the range of 5–200 m. In this illustrated in Fig. 6. These locations were selected to study the
experiment, 2200 different measurements were recorded. influence of location and environment on the performance of
The urban measurements were conducted in the city of LoRaWAN, e.g., park areas, commercial centers, main roads,
Beirut in Lebanon, which is a densely populated city (over airport areas, and port areas. On each of these measured points,
360 000) characterized by a high density of buildings, gardens, the aforementioned antenna heights were considered for the
roads, and commercial/industrial facilities. The measurements ED, where 25 packets were sent for each realization. A total
were performed within an area of 60 km2 having diverse amount of 2600 measurements was collected.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 2372 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

(a) (b)

Fig. 6. Map of urban and rural environments showing measurement locations of EDs. (a) Beirut city (urban). (b) Bekaa valley (rural).

The rural environment is Bekaa valley on Joub Jannine-Bar and by considering both ITU-R and Cost 231-MWF models,
Elias road, located at an altitude of 900 m above sea-level. This we proposed our PL model as
area is almost flat with small hills of 20–30 m height differ- n +2
ence, being surrounded by open fields, farmland and small hills ( nf +1 −b)
PL = 10n log10 (d) + PL0 + nw Lw + nf f
Lf (5)
with a lack of buildings, and other obstacles. Measurements
were made at 30 locations in LOS and NLOS conditions up where n = 2.85 is the indoor PL exponent, PL0 = 120.4 is the
to 20 km from the GW located at Kefraya tower as shown in reference PL. nw and nf are the number of walls and floors,
Fig. 3. Similarly to urban experiments, three antenna heights respectively. b was taken equal to 0.47 to obtain a suitable fit.
were used for EDs. A total amount of 2200 measurements was Lf = 10 and Lw = 1.41 represent the loss factor of floors and
collected. walls, respectively. All floors and walls are assumed identi-
cal in this model. The shadowing samples match the Gaussian
V. R ESULTS AND D ISCUSSION distribution with a standard deviation of 8 dB. This shadowing
is due to the variety of obstacles, e.g., desks, offices, etc. It
In this section, we present the results of measurements con-
should be mentioned that the obtained PL exponent is close
ducted in Lebanon as described in Section IV. First, we derive
to that used for the office area in ITU-R model (N = 33) at
PL models for the different environments (indoor, outdoor,
900 MHz. Moreover, we have compared the proposed model
urban, and rural). Next, the proposed PL models are compared
with the most used indoor PL models namely ITU-R, Cost
with widely used empirical models. The performance and
231-MWF, and 3GPP for Cellular-IoT [Fig. 7(a)]. The CDF
the coverage of LoRaWAN are also evaluated. Note that the
of shadowing samples resulting from the difference between
presented results were obtained for the uplink communication.
the measured values and estimated values of considered mod-
els are also compared in Fig. 7(b). The proposed model fit
A. Indoor Results measurements with more accuracy compared to other models.
In the indoor environment, we investigate the impact of Indeed, free-space model underestimates the measured values
wall and floor penetration losses on the received signal. Fig. 7 as expected. Moreover, the ITU-R model is close to the sam-
shows the PL values at different ED-GW distances and the ples, but it is less accurate than the proposed model due to
cumulative distribution function (CDF) of shadowing. the lack of additional wall losses. ITU-R model presents a
It can be noticed that the PL increases almost logarithmi- mean and a standard deviation of error of 0.48 and 8.3 dB,
cally with the distance. We observed, during measurements, respectively. Cost 231-MWF and 3GPP models underestimate
that the loss between floors does 3not increase linearly with the measured values and show a standard deviation of error of
the number of floors. The additional loss per floor decreases 8.7 and 10.2, respectively. This may lead to a high estimation
with the increasing number of floors. For instance, the addi- of received signal and consequently to noncovered areas.
tional penetration loss was found to be around 8.1, 6.7, 4.3 dB Cost 231-MWF considers both floor and wall losses, but it
between floor 2, 3, and 4, respectively. This is mainly due has a PL exponent of 2, leading to the underestimated values.
to the different propagation mechanisms between floors. In Furthermore, we have considered the locations of EDs in the
the first floor, the received signal mainly comes through the corridors, and we have obtained a PL exponent of n = 1.8–2.
floors, while the signal in the higher floors may be composed This range of values is widely considered in the literature for
of diffracted paths. Therefore, by using curve fitting techniques corridor area. The performance of LoRaWAN deployment in

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 EL CHALL et al.: LoRaWAN NETWORK 2373

(a) (b)

Fig. 7. (a) PL versus distance and (b) CDF of shadowing in indoor environment.

(a) (b)

Fig. 8. (a) PL versus distance and (b) CDF of shadowing in USJ campus.

TABLE V
L O R AWAN P ERFORMANCE IN I NDOOR E NVIRONMENT the USJ campus. Fig. 8(a) shows the PL as a function of
distance under different ED antenna heights. It can be seen
that increasing ED antenna heights improves the received sig-
nal strength (reduce PL). Indeed, increasing antenna heights
reduces the obstruction of Fresnel zone. In order to derive the
expected PL from the measured data, the linear polynomial fit
terms of PDR and SNR is also evaluated during indoor mea- was used as a function of logarithmic distance and ED antenna
surements. Table V summarizes the obtained results in the height as follows:
corridor and in the multifloor building. The results show an
PL = 10n log10 (d) + PL0 + Lh log10 (hED ) + Xσ (6)
average PDR of 95% and an average SNR of about 9 dB for a
distance up to 110 m from the GW. A minimum PDR of 45% where Lh is the additional loss due to the ED antenna height
was obtained. The minimum SNR and RSSI were observed at hED . The fitting process leads to n = 3.119, PL0 = 140.7,
a distance of 100 m in the basement. In general, the results and Lh = −4.7. This means that reducing hED results in
indicate a good quality of signal reception and reveal the reli- an additional loss of 4.7 dB per decade. The results show
ability of using LoRaWAN for effective communications in a good fit between the proposed model and the measured val-
indoor and deep-indoor deployment. ues. Additionally, the shadowing samples are inline with the
Gaussian distribution with zero mean and a standard deviation
of 9.7 dB, as shown in Fig. 8(b). This large value of shadowing
B. Outdoor Campus Results is mainly due to the topography variability and the large num-
In the following, we study the impact of ED antenna height ber of obstacles in this area, as described in Section IV. The
on the received signal and the performance of LoRaWAN in campus could be considered as a suburban environment but

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 2374 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

TABLE VI
E FFECT OF hED ON L O R AWAN P ERFORMANCE IN U RBAN E NVIRONMENT

distribution with zero mean and σ = 7.2 dB. This error con-
firms the relevant impact of obstructions and reflections in the
urban area. We have also compared the proposed model with
the most used urban PL models namely Okumura–Hata for
metropolitan areas, Cost 123-Hata and 3GPP-UMa as shown
in Fig. 11(a). The free-space PL is plotted with a dashed
black curve as a baseline model. The CDF of shadowing
samples are also compared in Fig. 11(b). We can see that
Fig. 9. PL versus distance in urban environment under different ED antenna Okumura–Hata model and Cost 123-Hata predict higher PL
heights, Beirut city.
values, which is also verified by the obtained mean of 2 and
3.9 dB, respectively, and an error standard deviation of 7.6 dB.
More interestingly, 3GPP-UMa model shows relatively low
with high shadowing effect. The performance of LoRaWAN predicted PL values with −1 dB as a mean error and 7.4 dB
was also evaluated. An average PDR, SNR and RSSI of 80%, as a standard deviation. A good fit of the measured data is
8.5 and −86 dB are, respectively, obtained despite the high observed for distances lower than 5 km. This inaccuracy in
shadowing effect in this area. the 3GPP-UMa model is due to the limitations of the base
antenna height to 150 m as well as user antenna height to
C. Outdoor Urban Results 1 m. However, the proposed model reduces the error and fits
In this section, we focus on studying the channel char- the samples with more accuracy compared to other models.
acteristics and the performance of LoRaWAN in an urban The effect of ED antenna height on LoRaWAN performance
environment. Fig. 9 shows the effect of ED antenna height is also evaluated in Table VI. The results show that increasing
on the PL. Similar to the campus results, we can see that by ED antenna height improves the reliability of the link, i.e.,
increasing the ED antenna height from 20 cm to 3 m, the PL is increases the PDR and the SNR. From the results, we can see
reduced by 8 dB. This is explained by the fact that an impor- that PDR exceeds 0.85 even for a low antenna height (20 cm).
tant part of the Fresnel zone will be obstructed by the ground It is worth to highlight here that an average PDR of 0.9 is
when lowering antenna height to 20 cm. By contrast, higher achieved with a coverage area up to 9 km. One single GW
antennas would lead to more clear space from obstacles. is able to cover the majority of Beirut city. The location of
We also note the variability of NLOS conditions in this area the GW at 260 m above sea-level play a significant role in
that can be classified into moderately and heavily obstructed achieving this coverage. The absence of reception or the low
environments. In moderate NLOS conditions, small obstruc- PDR (0.5) in some locations is due to the presence of high
tions, such as trees or building edges partially block the direct construction density or to the topography area.
path between ED and GW, whereas heavily obstructed NLOS An additional GW may be deployed to achieve higher
conditions have large obstructions that fully block the direct performance and coverage. This shows the importance of care-
path, leading to a higher PL. This high building shadow phe- fully selecting the locations of the GWs to improve system
nomenon may lead to noncovered areas in some locations. It reliability and capacity. We note that the lowest DR was used
should also be mentioned that the topography and the elevation in our experiments, higher DR may be used in case of good
profile influence the reception quality. For example, the pres- communication links to reduce the transmission time and to
ence of hills might block the reception as shown in Fig. 10. increase the system throughput.
It can be observed that LOS link is completely blocked in
Ras Beirut located at 9 km from the GW, leading to no packet D. Outdoor Rural Results
reception, while there is 90% PDR in El Manara located 8.5 m In the rural environment, the radio path was identified as
away due to the partially blocked LOS link. LOS link for an unobstructed path between the ED and GW
The PL was modeled by log-distance model including the antennas, and as NLOS link for the obstructed path. Similarly
effect of ED antenna heights using (6). The fitting procedure to the urban environment, various ED antenna heights were
has led to a PL exponent, a reference PL, an antenna loss considered. Similar log-distance model in (6) is considered to
factor equal to n = 4.18, PL0 = 102.86, and Lh = −6.3, derive the PL model.
respectively. It is interesting to note the high value of n (larger Fig. 12 illustrates the PL versus the distance for different
than 2 in free-space conditions) due to the heavy density of the ED antenna heights under LOS and NLOS conditions. The
Beirut city (high density of buildings and obstructions). The results show that the PL exponent is of 1.95 with ED antenna
shadowing samples in Fig. 11(b) show a well-fitted Gaussian heights of 1.5 and 3 m, which is close to the free-space model.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 EL CHALL et al.: LoRaWAN NETWORK 2375

Fig. 10. Example of elevation profile between USJ GW and ED in Ras Beirut and El Manara locations.

(a) (b)

Fig. 11. Comparison of proposed PL model with other models in (a) urban environment and (b) CDF of shadowing Beirut city.

(a) (b)

Fig. 12. PL model versus distance in rural environment. (a) LOS and (b) NLOS, Bekaa valley.

The standard deviation of shadowing is 2–3 dB. However, for Compared to the urban area, the loss factor Lh is slightly
hED = 20 cm, the LOS condition cannot be achieved due to the higher. Hence, the impact of ED antenna height on improving
obstruction of the Fresnel zone by the ground as previously the received signal is higher in the rural area due to fewer
discussed. In the case of NLOS, similar behavior is observed obstructions. The shadowing follows a Gaussian distribution
compared to the urban case, where lowering the antenna height with zero mean and σ = 6.43 dB [Fig. 13(b)].
results in increasing the PL. We note that, in the case of rural The proposed model was also compared to some exiting
environments, obstructions are mainly due to high vegetation, rural models namely Okumura–Hata model for open area and
trees, mountains, etc. The presence of high density of vege- 3GPP-RMa model considering NLOS, as shown in Fig. 13. In
tation can lead to a bad quality reception. The obtained PL case of 3GPP-RMa model, two cases were considered. The
exponent is n = 3.033, PL0 = 111.75, and Lh = −6.65 dB. first considers a street width W of 25 m, and the second

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 2376 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

(a) (b)

Fig. 13. Comparison of (a) proposed PL model with empirical models and (b) CDF of shadowing in rural environment, Bekaa valley.

(a) (b)

Fig. 14. Map of the drive test in urban and rural environment, coverage range up to 9 km (urban) and 47 m (rural). (a) Beirut City (urban).
(b) Bekaa Valley (rural).

TABLE VII
has the maximum street width W of 50 m. We observed L O R AWAN P ERFORMANCE IN RURAL E NVIRONMENT
that Okumura–Hata model underestimates the PL values and
presents a mean and a standard deviation of −17 and 6.9 dB,
respectively. By adjusting the street width to 50 m, the 3GPP-
Uma model is close to the samples with error mean and
standard deviation of 1 and 6.9 dB, respectively. The proposed
model shows lowest error of zero mean and standard deviation
of 6.45 dB, thus indicating that the proposed model matches
the measurements more accurately compared to other mod- the heights of both ED and GW. Control and retransmission,
els. Next, the performance of LoRaWAN in terms of PDR as well as ADR mechanisms, could also be used for further
and SNR was evaluated. Table VII summarizes the average improvement. An average PDR above 0.9 is achieved, mak-
and the minimum obtained values. These results validate those ing LoRaWAN a promising technology for IoT applications in
obtained in the urban scenario, where a similar impact of the rural environments.
antenna height on the reliability of LoRaWAN is observed. Driving tests were also conducted in urban and rural envi-
For instance, a minimum PDR of 0.25, 0.48, and 0.68 was ronments (Fig. 14). The results show that LoRaWAN EDs can
observed with hED = 0.2, 1.5, and 3 m, respectively. This relia- communicate up to a distance of 9 and 47 km in urban and
bility can be improved by carefully adjusting the locations and rural environments, respectively.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 EL CHALL et al.: LoRaWAN NETWORK 2377

TABLE VIII
S UMMARY OF PL C HARACTERISTICS AND S TANDARD D EVIATION OF can include other aspects, such as adaptivity and scalability
S HADOWING FOR VARIOUS T YPES OF E NVIRONMENTS of LoRaWAN systems, network planning, and optimization of
energy consumption.

R EFERENCES
[1] “Embracing the Internet of everything,” San Jose, CA, USA, Cisco,
White Paper, 2016. [Online]. Available: www.cisco.com
[2] L. Atzori, A. Iera, and G. Morabito, “The Internet of Things: A survey,”
Comput. Netw., vol. 54, no. 15, pp. 2787–2805, Oct. 2010.
[3] G. A. Akpakwu, B. J. Silva, G. P. Hancke, and A. M. Abu-Mahfouz,
“A survey on 5G networks for the Internet of Things: Communication
In general, we see from the results that LoRaWAN network technologies and challenges,” IEEE Access, vol. 6, pp. 3619–3647, 2018.
with one GW gives satisfactory performance. The quality of [4] U. Raza, P. Kulkarni, and M. Sooriyabandara, “Low power wide area
reception and the communication range depends on different networks: An overview,” IEEE Commun. Surveys Tuts., vol. 19, no. 2,
pp. 855–873, 2nd Quart., 2017.
factors, such as the density of obstructions (buildings and veg- [5] J. de Carvalho Silva, J. J. P. C. Rodrigues, A. M. Alberti, P. Solic,
etation) as well as the terrain profiles in the outdoor area. In and A. L. L. Aquino, “LoRaWAN—A low power WAN protocol for
the indoor environment, the wall and floor penetration affect Internet of Things: A review and opportunities,” in Proc. 2nd Int.
Multidisciplinary Conf. Comput. Energy Sci. (SpliTech), Split, Croatia,
the transmission. Jul. 2017, pp. 1–6.
A good selection of LoRaWAN GW locations achieves good [6] K. Mekkia, E. Bajica, F. Chaxela, and F. Meyerb, “A comparative study
coverage. The absence of reception in some locations, may be of LPWAN technologies for large-scale IoT deployment,” ICT Exp.,
due to the presence of high construction density or elevation vol. 5, no. 1, pp. 1–7, 2018.
[7] J. Finnegan and S. Brown. (2018). A Comparative Survey of LPWA
profile, which can be resolved by the installation of additional Networking. [Online]. Available: https://arxiv.org/abs/1802.04222
GWs. Table VIII summarizes the PL parameters and shadow- [8] F. Adelantado et al., “Understanding the limits of LoRaWAN,” IEEE
ing standard deviation for different tested environments. These Commun. Mag., vol. 55, no. 9, pp. 34–40, Sep. 2017.
[9] K. Mikhaylov, J. Petaejaejaervi, and T. Haenninen, “Analysis of capacity
proposed models are accurate and more simple to be used for and scalability of the LoRa low power wide area network technology,”
estimating the communication range and for enabling more in Proc. 22nd Eur. Wireless Conf., Oulu, Finland, May 2016, pp. 1–6.
analysis of LoRaWAN performance in the areas similar to [10] M. C. Bor, U. Roedig, T. Voigt, and J. M. Alonso, “Do LoRa low-power
wide-area networks scale?” in Proc. 19th ACM Int. Conf. Model. Anal.
Lebanon. Simulat. Wireless Mobile Syst. (MSWiM), 2016, pp. 59–67.
[11] M. Stabicki, G. Premsankar, and M. D. Francesco, “Adaptive config-
uration of LoRa networks for dense IoT deployments,” in Proc. 16th
VI. C ONCLUSION IEEE/IFIP Netw. Oper. Manag. Symp. (NOMS), Taipei, Taiwan,
Apr. 2018, pp. 1–9.
LoRaWAN has recently emerged as an attractive solution [12] U. Noreen, A. Bounceur, and L. Clavier, “A study of LoRa low power
for low-power and long-range IoT communications. In this and wide area network technology,” in Proc. Int. Conf. Adv. Technol.
paper, an in-depth study of the radio propagation characteris- Signal Image Process. (ATSIP), Fes, Morocco, May 2017, pp. 1–6.
tics using LoRaWAN in several realistic environments under [13] A. Augustin, J. Yi, T. Clausen, and W. M. Townsley, “A study of LoRa:
Long range & low power networks for the Internet of Things,” Sensors,
various ED antenna heights has been presented. The radio vol. 16, no. 9, p. 1466, 2016.
channel characterization is an essential issue in the design [14] O. Khutsoane, B. Isong, and A. M. Abu-Mahfouz, “IoT devices
and deployment of communication systems. Therefore, exten- and applications based on LoRa/LoRaWAN,” in Proc. 43rd Annu.
Conf. IEEE Ind. Electron. Soc. (IECON), Beijing, China, Oct. 2017,
sive measurement campaigns were conducted in Lebanon in pp. 6107–6112.
three different environments namely indoor, urban, and rural. [15] J. M. Marais, R. Malekian, and A. M. Abu-Mahfouz, “LoRa and
Indoor tests were carried-out in USJ campus. Urban tests LoRaWAN testbeds: A review,” in Proc. IEEE AFRICON, Sep. 2017,
pp. 1496–1501.
were realized in Beirut city, whereas rural tests were real- [16] P. Neumann, J. Montavont, and T. Noël, “Indoor deployment of low-
ized in Bekaa valley. Based on the empirical measurements, power wide area networks (LPWAN): A LoRaWAN case study,” in Proc.
PL models were further derived. Additionally, we compared IEEE 12th Int. Conf. Wireless Mobile Comput. Netw. Commun. (WiMob),
New York, NY, USA, Oct. 2016, pp. 1–8.
the proposed models with several well used models. It was [17] J. Petäjäjärvi, K. Mikhaylov, R. Yasmin, M. Hämäläinen, and J. Iinatti,
shown that the proposed models fit measurements with more “Evaluation of LoRa LPWAN technology for indoor remote health
accuracy and are much simple to be used in areas similar and wellbeing monitoring,” Int. J. Wireless Inf. Netw., vol. 24, no. 2,
pp. 153–165, Jun. 2017.
to Lebanon. Moreover, the performance of LoRaWAN was [18] L. Gregora, L. Vojtech, and M. Neruda, “Indoor signal propagation of
evaluated in terms of PDR and SNR. The reported results LoRa technology,” in Proc. 17th Int. Conf. Mechatronics Mechatronika
show the reliability of LoRAWAN communications in real- (ME), Dec. 2016, pp. 1–4.
life environments for long distances. In a dense urban area, a [19] K. E. Nolan, W. Guibene, and M. Y. Kelly, “An evaluation of low power
wide area network technologies for the Internet of Things,” in Proc. Int.
coverage range up to 9 km was attained, whereas in the rural Wireless Commun. Mobile Comput. Conf. (IWCMC), Paphos, Cyprus,
case a coverage range up to 47 km was reached using a single Sep. 2016, pp. 439–444.
deployed GW. The quality of transmission in high shadowing [20] J. Petajajarvi et al., “On the coverage of LPWANs: Range evaluation and
channel attenuation model for LoRa technology,” in Proc. 14th Int. Conf.
and blocking environments (buildings and elevation profile) ITS Telecommun. (ITST), Copenhagen, Denmark, Dec. 2015, pp. 55–59.
can be improved by the installation of additional GWs and [21] T. Petrić, M. Goessens, L. Nuaymi, L. Toutain, and A. Pelov,
the optimization of GW locations. The performance results “Measurements, performance and analysis of LoRa FABIAN, a real-
world implementation of LPWAN,” in Proc. IEEE 27th Annu. Int.
revealed the reliability of LoRaWAN for several IoT applica- Symp. Pers. Indoor Mobile Radio Commun. (PIMRC), Valencia, Spain,
tions, such as smart cities, smart agriculture, etc. Future work Sep. 2016, pp. 1–7.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.
 2378 IEEE INTERNET OF THINGS JOURNAL, VOL. 6, NO. 2, APRIL 2019

[22] S. Hosseinzadeh, M. Almoathen, H. Larijani, and K. Curtis, “A neural [46] Pycom Documentation. Accessed: Jul. 2018. [Online]. Available:
network propagation model for LoRaWAN and critical analysis with https://docs.pycom.io/
real-world measurements,” Big Data Cogn. Comput., vol. 1, no. 1, p. 7, [47] Kerlink Wirnet Station. Accessed: Jul. 2018. [Online]. Available:
2017. https://www.kerlink.com/product/wirnet-station/
[23] R. Sanchez-Iborra, J. Sanchez-Gomez, J. Ballesta-Vinas, M.-D. Cano,
and A. F. Skarmeta, “Performance evaluation of LoRa consid-
ering scenario conditions,” Sensors, vol. 18, no. 3, p. 772,
2018.
[24] O. Iova et al., “LoRa from the city to the mountains: Exploration
of hardware and environmental factors,” in Proc. Int. Conf.
Embedded Wireless Syst. Netw. (EWSN), Uppsala, Sweden, 2017,
pp. 317–322.
[25] G. del Campo, I. Gomez, S. Calatrava, R. Martinez, and A. Santamaria,
“Power distribution monitoring using LoRa: Coverage analysis in subur-
ban areas,” in Proc. Int. Conf. Embedded Wireless Syst. Netw. (EWSN),
Rida El Chall received the B.E. degree in
Madrid, Spain, 2018, pp. 233–238.
telecommunications and computer engineering from
[26] P. Ferrari, A. Flammini, S. Rinaldi, M. Rizzi, and E. Sisinni, “On the
the Faculty of Engineering, Lebanese University,
use of LPWAN for EVehicle to grid communication,” in Proc. AEIT Int.
Beirut, Lebanon, in 2012, and the M.S. degree
Annu. Conf., Sep. 2017, pp. 1–6.
in micro-technologies, architecture, networks, and
[27] N. Hayati and M. Suryanegara, “The IoT LoRa system design
communication systems and the Ph.D. degree in
for tracking and monitoring patient with mental disorder,” in Proc.
telecommunication and electronic engineering from
IEEE Int. Conf. Commun. Netw. Satellite (Comnetsat), Oct. 2017,
the Institut National des Sciences Appliquées,
pp. 135–139.
Rennes, France, in 2015 and 2012, respectively.
[28] A. Mdhaffar, T. Chaari, K. Larbi, M. Jmaiel, and B. Freisleben, “IoT-
From 2016 to 2017, she was a Research Engineer
based health monitoring via LoRaWAN,” in Proc. IEEE 17th Int. Conf.
with CEA-Leti, Grenoble, France. She joined ESIB-
Smart Technol. (EUROCON), Jul. 2017, pp. 519–524.
USJ, Beirut, as a Research Assistant for six months, during which time she
[29] W. Guibene et al., “Evaluation of LPWAN technologies for smart cities:
was involved with wireless communication for the Internet of Things. Her cur-
River monitoring use-case,” in Proc. IEEE Wireless Commun. Netw.
rent research interests include digital communications, wireless networking,
Conf. Workshops (WCNCW), San Francisco, CA, USA, Mar. 2017,
and Internet of Things.
pp. 1–5.
[30] Y. Okumura, E. Ohmori, T. Kawano, and K. Fukuda, “Field strength and
its variability in VHF and UHF land mobile radio service,” Rev. Elect.
Commun. Lab., vol. 16, nos. 9–10, pp. 825–873, 1968.
[31] “Digital mobile radio towards future generation systems, final report,”
Eur. Communities, Brussels, Belgium, Rep. COST Action 231, 1999.
[32] “Guidelines for evaluation of radio interface technologies for
IMT-advanced,” ITU-Recommendations, Geneva, Switzerland,
Rep. M.2135-1, 2009.
[33] P. Kyösti et al., WINNER II Channel Models, D1.1.2 V1. 2, document
IST-4-027756, WINNER II, 2008.
[34] P. Heino et al., WINNER+ Final Channel Models, D5.3, V1.0, document Samer Lahoud received the Ph.D. degree in
CELTIC CP5-026, WINNER+, 2010. communication networks from Telecom Bretagne,
[35] “Further advancements for E-UTRA physical layer aspects, v9.0.0,” Rennes, France, in 2006.
3GPP, Sophia Antipolis, France, Rep. TR 36.814, 2010. He is an Associate Professor with Saint-Joseph
[36] “Propagation data and prediction methods for the planning of indoor University, Beirut, Lebanon, where he lectures on
radio communication systems and the radio local area networks in computer networking courses with the Faculty of
the frequency range 900 MHz to 100 GHz,” Int. Telecommun. Union, Engineering (ESIB). He spent one year with Alcatel-
Geneva, Switzerland, ITU-Recommendation P.1238-3, 2001. Lucent Bell Labs Europe, Boulogne-Billancourt,
[37] V. Erceg et al., “TGn channel models: IEEE P802.11wireless LANs,” France. From 2007 to 2016, he was with the
IEEE, Piscataway, NJ, USA, Rep. 802.11-03/940r4, 2004. University of Rennes 1, Rennes, and with IRISA
[38] M. Lott and I. Forkel, “A multi-wall-and-floor model for indoor radio Labs Rennes, Rennes, as an Associate Professor. He
propagation,” in Proc. IEEE Veh. Technol. Conf. (VTC), Rhodes, Greece, has co-authored over 80 papers published in international journals and confer-
May 2011, pp. 64–468. ence proceedings. His current research interest includes routing and resource
[39] P. Jörke, S. Böcker, F. Liedmann, and C. Wietfeld, “Urban channel mod- allocation algorithms for wired and wireless communication networks.
els for smart city IoT-networks based on empirical measurements of
LoRa-links at 433 and 868 MHz,” in Proc. IEEE 28th Annu. Int. Symp.
Pers. Indoor. Mobile Radio Commun. (PIMRC), Montreal, QC, Canada,
Oct. 2017, pp. 1–6.
[40] S. Hosseinzadeh, H. Larijani, K. Curtis, A. Wixted, and A. Amini,
“Empirical propagation performance evaluation of LoRA for indoor
environment,” in Proc. IEEE 15th Int. Conf. Ind. Inf. (INDIN), Emden,
Germany, Jul. 2017, pp. 26–31.
[41] LoRaWAN Measurement Campaigns in Lebanon. Accessed:
Nov. 26, 2018. [Online]. Available: https://zenodo.org/record/1560654
[42] “SX1272/3/6/7/8: LoRa modem designer guide AN1200.13, rev.1,” Melhem El Helou (S’08–M’15–SM’19) received
Semtech, Camarillo, CA, USA, Rep., Jul. 2013. the engineer’s and master’s degrees in telecom-
[43] B. Reynders and S. Pollin, “Chirp spread spectrum as a modulation munications and networking engineering from the
technique for long range communication,” in Proc. Symp. Commun. Veh. Ecole Supérieure d’Ingénieurs de Beyrouth (ESIB),
Technol. (SCVT), Mons, Belgium, Nov. 2016, pp. 1–5. Faculty of Engineering, Saint Joseph University,
[44] N. Sornin, M. Luis, T. Eirich, T. Kramp, and O. Hersent, LoRaWAN Beirut, Lebanon, in 2009 and 2010, respectively,
Specification Version 1.0, LoRa Alliance, Beaverton, OR, USA, and the Ph.D. degree in computer and telecommuni-
Jan. 2015. cations engineering from IRISA Research Institute,
[45] “Short range devices (SRD) operating in the frequency range 25 MHz to University of Rennes 1, Rennes, France, and
1 000 MHz; part 2: harmonised standard covering the essential require- Saint Joseph University, in 2014.
ments of article 3.2 of directive 2014/53/EU for non-specific radio equip- In 2013, he joined ESIB, where he is currently an
ment, v3.1.1,” ETSI, Sophia Antipolis, France, Rep. EN 300 220-2, Assistant Professor (fr: Maître de conférences). His current research interests
2017. include wireless networking, Internet of Things, and quality of service.

Authorized licensed use limited to: UNIVERSIDADE FEDERAL DA PARAIBA. Downloaded on September 21,2020 at 17:39:37 UTC from IEEE Xplore. Restrictions apply.