Wireless Transmissions

This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2022.3151967, IEEE Access
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number
Wireless Transmissions, Propagation and

Channel Modelling for IoT Technologies:
Applications and Challenges
Haider A.H. Alobaidy1, Student Member, IEEE, J. S. Mandeep1,2, Mehran Behjati1,
Rosdiadee Nordin1, Senior Member, IEEE, and Nor Fadzilah Abdullah1, Member, IEEE
1
Department of Electrical, Electronic & Systems Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia
(UKM), Bangi 43600, Malaysia
2
Space Science Centre (ANGKASA), Institute of Climate Change, Universiti Kebangsaan Malaysia (UKM), Bangi 43600, Malaysia
Corresponding author: Rosdiadee Nordin (e-mail: adee@ukm.edu.my).

This work was supported in part by the Malaysia’s Ministry of Higher Education (MOHE), under Grant ref numberFRGS/1/2019/TK04/UKM/02/8 and the
Air Force Office of Scientific Research: FA2386-20-1-4045 (UKM Ref: KK-2020-007)
ABSTRACT The Internet of Things (IoT) has rapidly expanded for a wide range of applications towards a
smart future world by connecting everything. As a result, new challenges emerge in meeting the requirements
of IoT applications while retaining optimal performance. These challenges may include power consumption,
quality of service, localization, security, and accurate modeling and characterization of wireless channel
propagation. Among these challenges, the latter is critical to establishing point-to-point wireless
communication between the sensors. Channel modeling also varies depending on the features of the
surrounding area, which have a direct impact on the propagation of wireless signals. This presents a difficult
task for network planners to efficiently design and deploy IoT applications without understanding the
appropriate channel model to analyze coverage and predict optimal deployment configurations. As a result,
this challenge has attracted considerable interest in academic and industrial communities in recent years.
Therefore, this review presents an overview of current breakthroughs in wireless IoT technologies. The
challenges in such applications are then briefly reviewed, focusing on wireless channel propagation modeling
and characterization. Finally, the study gives a generalized form of commonly used channel models and a
summary of recent channel modeling developments for wireless IoT technology. The outcome of this review
is expected to provide a new understanding of the propagation behavior of present and future wireless IoT
technologies, allowing network engineers to undertake correct planning and deployment in any environment.
Additionally, the study may serve as a guideline for future channel modeling and characterization studies.
INDEX TERMS Channel Modelling, Channel Characterization, IoT Applications, IoT Challenges,
Wireless IoT Technologies.
I. INTRODUCTION
Wireless Channel Characteristics are essential in any A. Problem Formulation and Background
communication system since they directly affect wireless signals However, choosing an accurate channel model to represent the
traveling from the transmitter (Tx) to receiver (Rx). Wireless actual real-world wireless IoT deployment is challenging due to
transmission, on the other hand, has become the backbone for imperfection surrounding the deployment area. These
enabling wireless IoT applications. Thus, establishing and imperfections might include the varying terrain, large objects
developing wireless networks relies on propagation models that (e.g., buildings and tall trees), and various moving objects’ speed.
consider geographical features, among other factors, that can In other words, the actual wireless IoT performance differs when
contribute to signal loss [1]. As a result, there is a need to used in environments other than where they were developed.
investigate the propagation channel parameters that directly Furthermore, the ongoing expansion of wireless networks needs
impact wireless transmission performance. Failure to do so will further signal propagation studies to assure an effective post-
impact the planning and deployment of any IoT application. planning phase service coverage and efficiency [2].
VOLUME XX, 2021 1
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
H. A.H. Alobaidy et al.: A Review on Wireless IoT Technologies
In this sense, radio network engineers frequently employ PL compared to Okumura-Hata, ITU-R P.1546, and Walfisch-
models to predict coverage, optimize constrained network Ikegami, which achieved 12.5, 4.3, and 10.6 dB RMSE,
resources, and conduct interference feasibility studies [3]. As a respectively. Finally, the proposed model in [1] had a lower
result, numerous studies in recent years have focused on either RMSE in all cases, ranging from 1.67 dB to 4.25 dB, compared
surveying and evaluating current models; as in [2], [4]–[8]; or to the Hata model, which had an RMSE ranging of 3.71 dB to
proposing/improving models for more accurate propagation 13.09 dB. The proposed model could also make suitable
estimation in a specific area; as in [1], [3], [9]–[11]. adaptations for places outside Guajaraa Bay, with an RMSE of
For example, [4] compared PL prediction from nine empirical 1.68 dB, compared to 6.99 dB for the mixed terrain model and
PL models to observations from four television transmitters along 7.61 dB for the parabolic equation model.
five pathways in urban and rural areas. Similarly, [2] compared On the other hand, a statistical propagation model to
the performance of three PL models at a 5.8 GHz frequency band estimate Air-to-Ground (ATG) coverage and PL between a
in 12 cities using RSSI from 335 fixed user locations. In [4], the Low Altitude Platform (LAP) and a terrestrial terminal in an
Hata and Davidson models outperformed Ilorin, CCIR, Cost231- urban region was proposed in [11]. Instead of site-specific 3D
Hata, Ericsson 9999, and ECC-33 in prediction accuracy. models, the estimation relied on the elevation angle between
Meanwhile, the SUI and Okumura models had the lowest terminal and LAP. However, the study’s drawbacks are: 1) the
prediction accuracy, ranging from 15% to 21%. In contrast, the Doppler effect caused by the hypothetical high speed of an
SUI Model performed best in [2], with the lowest root mean aerial transmitter was not examined because the assumption
square error (RMSE) of 7.22 dB and standard deviation (STD) primarily focused on a semi-stationary LAP; 2) the study was
compared to COST231-Hata and ECC-33. based on simulations with no physical verification; and 3)
On the other hand, both [5] and [9] focused on studying the potential urban geometry influences, such as foliage,
foliage impact on the propagated signal. In [5], the propagation lampposts, and moving objects, were also ignored, presuming
experiment was conducted in a tropical outdoor canopy and that the large-scale building geometry and its EM
thickly foliated trees. The study assessed foliage excess loss characteristics will dominate the average PL.
from 2 GHz to 18 GHz and 26.5 to 40 GHz mmWave frequency In turn, the forthcoming 5G-based IoT, with its numerous
bands. The Weissberger Modified Exponential Decay (W complex application scenarios, necessitates more effective
MED) and ITU-R provide the most optimistic estimate of the channel estimates and modeling [6]. Several studies have been
measured data, while COST 235 provides the most pessimistic conducted to evaluate and overview various wireless channel
estimate. FITU-R has the largest attenuation increment as a models for such applications. In [12], the authors conducted a
function of frequency. Thus, results disagreed with current survey and evaluated the applicability of available propagation
empirical models built solely for their temperate measurement models for Industrial IoT (IIoT) applications. The evaluation
area. This study, however, did not consider a few other aspects, utilized worst-, median-, and best-case predictive behaviors. As
such as leaf size, foliage density, tree species, wind effect, and case studies, NB-IoT, Sigfox, LoRa, ZigBee, and MIOTY IoT
precipitation. As for [9], the authors suggested a new foliage technologies’ performance metrics such as cell radius, spectral
empirical propagation model for two morphologies based on efficiency, and outage probability were examined. The models
measurements made in outdoor tropical vegetation at 700-800 considered include Free Space Path Loss (FSPL), 3GPP models
MHz. W MED, COST235, and Chen & Kuo foliage models (indoor open and mixed hotspots, outdoor urban macro- and
were also compared. The comparison revealed that the proposed micro-cell, and outdoor rural), and industrial indoor channel
model differs significantly over larger distances between Tx/Rx. loss model for ISM bands. According to the results, NB-IoT has
The authors of [10] and [1] analyzed an outdoor mixed path the longest cell radius and the lowest outage probability in
tropical Amazon region and developed a propagation model for outdoor scenarios, Sigfox has the highest indoor spectral
Digital TV (DTV) services in the UHF band. The model in [10] efficiency, and ZigBee has the widest working bandwidth. The
was based on Geometrical Optics (GO) and the Uniform authors argued that 3GPP models are the most suitable for
Theory of Diffraction (UTD). In contrast, [1] proposed an estimating the examined communication metrics, with
empirical propagation model based on Machine Learning (ML) estimations frequently close to the median-predicted behavior.
approaches suitable to non-homogeneous paths and various Besides, [6] provided an outline of the ML conceptual
climates. The model was then tested using measurements from framework for future radio propagation modeling and
two DTV stations, considering multiple paths, woodland, and estimation. The authors also introduced 5G channel modeling
freshwater. [10] showed that the diffracted city rays and the and showed current advances in ML-based channel modeling.
rays reflected across the water were key transition variables. Then, they discussed some essential methods in ML, such as
Meanwhile, the attenuation of a forest proved to be greater than regression analysis, which is utilized for data fitting to predict
that of a suburban area. They also found that a dense forest had channel modeling parameters. They also showed how
a lower electric field by 12 dB than a sparse forest. Compared classification determines output data to estimate traffic flow.
to the Okumura-Hata and ITU-R P.1546 models, the proposed They then emphasized that much work is needed to overcome
mixed path model had the lowest average RMSE of 3.43 dB. obstacles and achieve accurate modeling despite significant
Similarly, the proposed model achieved 2.75 dB RMSE investment in ML. Nevertheless, ML techniques showed
2 VOLUME XX, 2021
feasibility by estimating channel parameters and extracting suggests some future work directions. To further facilitate
channel information. Hence, it was concluded that successful reading, Fig. 1 provides a detailed structure of the review.
predictive ML techniques would reduce complexity while
increasing precision compared to traditional channel modeling.
Further, as the number of measurements is reduced, the use of
ML approaches in wireless channel modeling becomes
increasingly important as generalization abilities improve.
Finally, [8] provided a comprehensive overview of existing
ATG channel measurement studies, large- and small-scale
fading channel models, limitations, and future guidelines for
UAV communication scenarios. Different UAV propagation
scenarios were also explored, as well as critical elements for
these measurements.
B. Study Motivations and Contributions

In conclusion, most wireless IoT characterization and modeling
studies are clearly limited to study area settings. Besides that,
only one study, in [12], examined the suitability of current
channel models for wireless IoT technologies, which is in turn
explicitly limited to IIoT applications, considering few channel
models. On the other hand, the remaining studies focus on
performance evaluation and surveying the suitability of existing
models for traditional communication technologies with high
transmission power, high power devices, and high gain antennas
without considering IoT-based application requirements. As a
result, more research is needed to precisely evaluate the
suitability of existing models for various wireless IoT
technologies, particularly in regions known for their harsh
environments, such as tropical climates with irregular terrain.
As such, this work builds on past research and presents a
comprehensive review of recent advances in wireless channel
characterization and modeling for wireless IoT technologies.
Therefore, the contributions of this work can be summarized
as follows:
1) Provide a rigorous survey of modern wireless IoT-based
solutions.
2) Describe the most significant challenges in such
applications, with a particular emphasis on wireless
channel propagation modeling and characterization.
3) The study then provides a brief description of channel
modeling, a generalized form of commonly used channel
models, and a list of 34 well-known and commonly used
channel model equations.
4) Finally, the study reviews recent advances in channel
modeling for wireless IoT technologies, identifies gaps in
existing research, and suggests future directions for
developing a robust wireless IoT solution.
The rest of the paper is structured as follows: Section II
examines wireless IoT technologies and potential IoT solutions.
Section III discusses the challenges related to wireless IoT
technologies. Section IV covers wireless channel propagation
modeling concepts and provides detailed equations for well-
known and commonly used models. Section V reviews the most
recent studies on wireless IoT channel modeling and
characterization. Finally, section VI draws conclusions and FIGURE 1. A detailed structure of the review.
VOLUME XX, 2021 3
II. Wireless IoT Technologies and Potential IoT Solutions electricity usage by 76%, resulting in a 73% cost savings
The Internet of Things (IoT) is a network that connects users compared to traditional lighting systems utilized in the city.
and objects by using information sensing devices and Furthermore, the combined effect of LED lighting and wireless
actuators [13]–[15]. IoT has caught academic and industry remote control saved 224 TEP (a ton of oil equivalent) each
interest for the past few years [16] due to the exponential rise year and avoided 500 tons of CO2 emissions.
of connected devices and the need for new or optimized In a smart city paradigm, many other applications exist,
methods to manage many connected devices [13]. As a result, such as automotive and intelligent transportation systems [25],
the number of connected devices nowadays is expected to be [26], Remote Structural Health Monitoring (RSHM) [27],
between 26 to 50 billion [17]–[21]. This trend is anticipated to smart homes and buildings [22], [28], [29], metering
accelerate further, expected to reach around 75 to 100 billion monitoring [28], concrete surface WSN [30], and level of trash
connected devices by 2025 [13], [16], [21]. monitoring [31]; as shown in Fig. 2.
Various wireless IoT technologies and network topologies
can support IoT [22]. Depending on the application B. Internet of Medical Things (IoMT) and Emergency IoT
requirements, operational restrictions, and coverage needs, (EIoT)
these technologies can be classified into three categories. First The healthcare sector has recently experienced tremendous
is traditional short-range technologies, such as Wi-Fi, ZigBee, expansion and is estimated to be very large due to global aging
and Bluetooth, mostly utilized for high availability and low [32] and corona-virus disease (COVID-19) [33], [34]. The
latency activities, whereas energy consumption is not a key healthcare sector has recognized the huge potential of IoMT
issue [14], [23]. As such, they are unsuitable for long-range technologies, often known as healthcare IoT, due to their
transmission scenarios. Second is cellular-based solutions, effective collection, analysis, and transfer of health data. Thus,
such as 2G-4G and future 5G, which provide greater coverage the IoMT has risen as a combination of medical equipment and
but demand excessive device energy. Finally, as a result of the software platforms to offer full healthcare services linked to
crucial IoT application requirements [13], [20], Low Power healthcare IT systems [35].
Wide Area Networks (LPWANs) are emerging as an exciting In this way, IoMT monitors vital signs, instantly detects
new trend in the growth of wireless communication systems. anomalies, and alerts parties concerned. Multiple sensors are
Many LPWAN developments, such as Sigfox, LoRaWAN, embedded in garments, watches, wearable items, and even
and NB-IoT, have lately emerged in both unlicensed and jewels in such a condition to continually monitor vital signs
licensed spectrum, becoming one of the leading novel such as blood pressure, heart rate, blood glucose level, blood
technologies with numerous technological differences [16], oxygen level, and standing position [36].
[20], [21]. Its main features include large coverage area On the other hand, electrical bioimpedance methods require
support and massive scale networking with low cost, long-life, complex devices and measuring settings in medical and
and restricted data rate EDs [13], [16], [18]. personalized healthcare systems. As a result, various
Together, these wireless IoT technologies support various investigations were carried out in this context, as in [37]. The
IoT solutions for outdoor and indoor scenarios, whether authors identified a standalone bioimpedance analyzer termed
deployed in urban, suburban, or rural areas. Those applications "Zink" with IoT monitoring features. The system allows users
cover many sectors of life, including those reviewed in the to obtain/perform single or multiple bioimpedance readings
below subsections and as shown in Fig. 2. remotely. The device may also calculate bioimpedance
synchronized with an electrocardiogram (ECG) or
A. Smart Cities electromyogram (EMG) readings.
These paradigms are primarily evolving to address the issues Similarly, the authors of [38] proposed an IoT-assisted ECG
faced by the rising complexity of today’s urban environments monitoring system with reliable data transfer for continuous
[22]. As a result, IoT systems are widely employed in smart cities cardiovascular health monitoring, enabling automated
to enable remote device monitoring, management, and control, classification and real-time operation. The proposed system has
along with new views and operational knowledge driven from a high potential for defining clinical acceptance of ECG signals
enormous data streams. The essential characteristics of a smart to improve an unsupervised diagnostic system’s performance,
city include a high level of technological integration and broad accuracy, and reliability.
use of information resources [24]. The IoMT may also be highly effective in emergency
One of these smart city IoT applications is smart lighting, scenarios; for example, during the current COVID-19
which aims to boost the efficiency of the lighting infrastructure pandemic, many innovators, medical authorities, and
by lowering electricity consumption using dimmers and task government agencies concentrated on harnessing IoMT
lighting. Smart lighting can also serve as an access network for resources to relieve pressure on medical systems. As a result,
context-aware services such as smart parking. Such technology several IoMT technologies, such as smart thermometers and
has the potential to have a significant economic impact. In telemedicine (remote patient monitoring), have contributed to
2014, for example, studies revealed that smart lighting monitoring and, eventually, dealing with the COVID-19
infrastructure in Montechiarugolo, Italy [22] helped reduce pandemic impact [35].
4 VOLUME XX, 2021
FIGURE 2. Several innovative IoT solutions from the reviewed literature.
Smart IoT-based systems, such as the IoT-based fire-ground and health data to a remote command unit. As a result, the
and firefighter monitoring system (IoT-FFM) suggested in [39], technology will aid in protecting people’s lives (both firemen
can also assist people in risky situations. These additional and victims) and avoid potentially hazardous situations.
functions include surveillance of fire environment gases, real- Other IoMT and EIoT applications include medical bleeder
time notification of danger, and transfer of firefighter location [40], emergency & disaster monitoring [11], Wearable Body
VOLUME XX, 2021 5
Area Networks (WBAN) [41]–[43], and Device-to-Device neural network approach. This can provide users and coaches
(D2D) & Device-to-Machine (D2M) based communication with a highly matched training plan, generate scientific and
devices to monitor and track devices during Search and routine physical fitness tests for athletes, give players
Rescue (SAR) operations [44], as listed in Fig. 2. maximum competitive potential, and improve physical fitness.
C. Industrial Internet of Things (IIoT) E. Environmental and Agriculture IoT

The IIoT and Industry 4.0 development goals include Climate change, population expansion, demographic shift,
massively distributed smart computing and networking urbanization, and resource scarcity imply that the world’s
advances in industrial production and manufacturing systems largest cities will need to adapt to survive and grow in the
for automation, quality, and control [45], [46]. IIoT has future decades [56]–[58]. Cutting greenhouse gas emissions to
distinct features and requirements that set it apart from avoid catastrophic global warming and so sustaining or
commercial IoT, like specific types of built-in smart devices, improving living quality, on the other hand, maybe an
network capabilities, QoS, and strict command and control expensive and hard task [56], [57]. Where major aspects
standards [45]. Thus, IIoT helps reduce faults, costs, and directly impacting the quality of life, such as water and air
improves safety and performance in manufacturing processes, quality, are routinely monitored using expensive monitoring
giving the industry a consistent degree of precision, systems [56], [59]. Furthermore, such assessments,
availability, and scalability [46], [47]. particularly for water quality monitoring, typically include on-
Among other features, IIoT allows control commands to be site sample collection and subsequent laboratory analysis,
sent rapidly from a client device to anyplace in a facility, such which adds intense and costly labor [59]. Recent IoT advances
as an oil rig or power plant, without requiring a physical have thus enabled unique methods to address these issues and
presence. This ability to respond swiftly from any location may autonomous real-time monitoring of these factors by
save the firm and the environment from costly damages or integrating low-cost sensor devices, machine-to-machine
disasters like the 2015 oil spill in Santa Barbara, CA, USA [46]. (M2M), and IoT technologies [56], [59].
Another innovative use of IoT is in condition-based Besides, IoT monitoring devices may track environmental
monitoring (CbM) applications to investigate vibration-based impacts on urban underground infrastructure caused by excess
machine health. As such, the objective is to compare observed storm and wastewater volumes entering pipelines, causing
vibration to components that are vulnerable to common wear backups and, as a result, sanitary sewer overflows. As a result,
processes, such as bearings, gears, chains, belts, brushes, these systems, also known as urban underground IoT (UIoT),
shafts, coils, and valves. These sensors can save money by are required for controlling external water flow into pipelines
detecting early warning signs of breakdown in industrial and are enabled by merging underground wireless
machinery and optimizing the maintenance schedule [48]. communication and sensor technologies [60].
Other IIoT applications include industry risk management Similarly, various studies have been conducted in
(to reduce infectious disease outbreaks such as COVID-19) agriculture IoT to digitalize the agricultural industry [61].
[49], smart grid (SG) [50], [51], smart goods tracking and Thus, enabling the implementation of Precision Agriculture
transportation systems [52], [53], smart oilfields [49], and (PA) with increased financial returns, increased product
smart factories [45]; as listed in Fig. 2. quality and yield, and cost reduction [60], [62]. On the other
hand, farming has recently become fairly reliant on processing
D. Sport Activities production area information for various crops, utilizing
Another prominent application of IoT technology is in sporting actuators to remotely operate network equipment,
activities [36]. Where gadgets such as IoT-based monitoring and consequently adopting two-way communication systems [62].
tracking devices could, for example, monitor hikers’ positions, Using IoT in the agricultural process to improve the
health status, and overall safety while hiking [44]. D2D traceability system also improves overall farming practices,
connectivity is favored for such situations with IoT, as small effective land utilization, agricultural quality, and safety. As a
devices may be unable or need to connect with standard cellular result, internal traceability assures support for seedlings used
infrastructure. As a result, such devices may communicate data in major crops across the whole agricultural sector. In this
on the positions of the various participants, allowing each to keep regard, [63] described an IoT-based greenhouse traceability
track of the others for either competition or safety [54]. system for tracking seedlings and other agricultural goods
Another example is presented in [55], which is an IoT-based from germination to harvest. The system monitors luminosity,
smart physical fitness tracking device. The device comprises moisture, temperature, and water usage to indicate water
several portable sensors and RFID to adapt and control the consumption, plant development trends, and product
physical load intensity in real-time, accurately, and efficiently, harvesting timelines. The technology also allows for
based on the physiological load within the human body, to automatic monitoring of the indoor greenhouse environment
meet the predetermined objectives of the training plan. Thus, via an irrigation system or temperature control and a vital
athletes’ physical fitness testing data may be saved and overview of agricultural product internal traceability from
categorized with such a system then evaluated using the BP seed to the final product.
6 VOLUME XX, 2021
Another example is automated irrigation monitoring and monitoring the dairy cows’ nutritional and health status and
control systems, which can help identify the precise timing of predicting anomalies (e.g., metabolic disorders after calving).
watering and the amount of water required to meet crop water In contrast, [76] described an innovative IoT-based system
needs while improving water usage efficiency [64]–[67]. with multiple sensors, wireless data transfer, and self-
Other research looks into solutions like UIoT to enable smooth sustaining power for long-term and robust bee colony
access to data from agricultural fields, including in-situ soil monitoring. The suggested system can detect hive temperature
sensing capabilities, communication through plants and soil, and humidity, bee comb weight, colony sounds, and the
and supplying real-time environmental data, thus delivering number of bees entering the hive. Long-term monitoring trials
useful information to farmers [60]. revealed that the system could run continuously without
For example, [67] offered an open-source IoT-based smart human intervention, and data can reveal bee colonies' activity
irrigation architecture with a hybrid machine-learning-based and growth. Notably, data successfully identified a swarming
strategy to predict soil moisture and watering requirements. The behavior, indicating a high potential for recognizing specific
system uses sensor data such as soil moisture, soil temperature, bee colony activities. As a result, the suggested approach had
ambient conditions, and weather forecast data to predict soil a significant impact on verifying the activities and state of bee
moisture for the next few days. Similarly, [68] uses an colonies, which could contribute to the evaluation of bee
autonomous IoT-enabled WSN system comprising soil behavior and the enhancement of beekeeping quality.
moisture & temperature, environmental temperature &
humidity, CO2, and daylight intensity sensors to obtain real- G. Unmanned Aerial Vehicles (UAVs)
time farm data. Furthermore, farm history is saved to create Unmanned aerial vehicles (UAVs), commonly known as
relevant actions throughout the farming cycle. The study also drones, have witnessed increased adoption in recent years, from
uses neural networks to describe accurate water valve control agriculture to industry, government to private organizations, and
based on soil water demand estimation one hour ahead. smart cities to rural area monitoring. UAVs are becoming more
Accordingly, the structural similarity (SSIM) index-based soil maneuverable and intelligent with recent advances in IoT, cloud
moisture content shortage is calculated to control the specified & edge computing, and wireless communication technologies.
valves and maintain uniform water needs over the entire farm As such, the Internet of Drones (IoD) is emerging as one of the
area. Valve control commands are examined again using a fuzzy promising technologies and use cases for UAVs. Many
logic-based weather condition modeling approach to change industries are projected to adopt IoD-based automation for
control commands in response to changing weather conditions. smart monitoring, surveillance, and search and rescue via
In contrast, the authors of [61] proposed the design of a reliable mobile communications systems.
cloud-based greenhouse crop production solution to provide Drones have recently been used as flying cellular base
various services for the environmental and economic benefits stations (BSs) to deliver reliable and energy-efficient IoT
of agricultural operations, as well as to improve system communications [77], [78]. Such deployment of BSs would
performance by offering recommendations for the use of increase the probability of LOS communication and mitigate
water, pesticides, fertilization, or energy. destructive effects such as shadowing and blockage, thereby
Other agricultural solutions (as listed in Fig. 2) could increasing the reliability of communication links [79].
include tracking cows’ grass grazing behaviors [69], [70], One application of IoD is remote sensing, which allows for
monitoring grass growth on major roadways, residential easier measurements and monitoring in harsh or remote areas.
lawns, outdoor sports centers, and inaccessible areas [70], and The authors of [80], for example, presented an IoD-based
rice field monitoring [71]. In addition, homeowners can utilize system for livestock monitoring on large-scale rural farms, in
IoT devices to track and operate irrigation systems, allowing which a fixed-wing UAV was equipped with a LoRa gateway
for self-management during extended absences [70]. and collected data from IoT-based monitoring equipment
deployed across the farm. Disaster relief and management is
F. Internet of Animals (IoA) another novel application of IoD. [81] proposes a UAV-cloud
In this context, IoT can be useful in a variety of applications. framework for disaster sensing in disconnected, intermittent,
For instance, IoT monitoring and tracking devices could be and resource-constrained environments. Other UAV use cases
used to track and monitor wildlife in rural or mountainous & include aerial entertainment [82], agriculture inspection [83],
rocky areas, like mountain lions [44], pink iguanas [72], and powerline/BS inspection [84], surveillance [85], aerial
zebras [73]. WSN and MEMS advancements have also aided monitoring [86], radiation monitoring [87], logistics and
other solutions, such as remote monitoring of dairy cow health transportation [88], and localization and mapping [89].
issues using IoT-based WBAN, as reported in [74]. Besides
that, [75] described real-time IoT-based tracking systems H. Vehicle-to-Vehicle (V2V) Communication
leveraging sensors to enable large dairy farms to maximize The automotive industry is currently undergoing a technological
revenue and improve cow welfare. Where health indicators, revolution fueled by IoT technologies. As a result, vehicles are
such as ruminal temperature and pH, are critical factors for going through a disruptive shift from manually driven vehicles
to self-driving vehicles with varying degrees of autonomy,
VOLUME XX, 2021 7
depending on the technology on-board and the situations Finally, the underwater wireless sensor network (UWSN)
encountered [90]. Existing Original Equipment Manufacturer has been widely used for data collecting in an underwater
(OEM) embedded systems use standard sensors such as GPS, environment for military and civilian uses. UWSN
camera, and proximity sensors, limiting their use to some basic applications include tsunami and earthquake warning systems,
applications such as sensing adjacent objects and crash undersea military monitoring, ocean exploration, navigation,
detection. Hence, more advanced technologies are needed to environmental, and pollution management [95].
achieve a high level of autonomy, safety, and traffic control.
V2V communication is a game-changing technology that III. Challenges of Wireless IoT Technologies
connects vehicles and provides drivers with critical information With the technological advancements and current
about their vehicle, other nearby vehicles, and the surrounding developments in the wireless IoT paradigm, various
environment, such as weather, roadblocks, and traffic [91]. challenging factors must be considered while deploying
This data will be utilized to warn drivers of potential hazards precise, high-performance, and cost-effective IoT systems. As
via a visual display, seat vibration, or tone. These warnings will such, the following sub-sections summarize the most crucial
assist drivers in responding more quickly and avoiding challenges confronting any IoT solution:
potential accidents [92]. Some examples of V2V-based driver
assistance applications include the following: A. Power Consumption (Reliable and Permanent Power
• Intersection movement application; warns drivers when Source)
it is unsafe to enter an intersection. In many cases, IoT systems are placed in locations with
limited power resources, resulting in most IoT systems being
• Do not pass applications; warns drivers when passing a
battery-powered [96], [97]. Furthermore, the rising
slower-moving car is unsafe.
complexity of IoT data transmission, storage, and processing
• Emergency electronic brake light application; notifies
capabilities requires more power-intensive tasks. In this sense,
the driver when an out-of-sight vehicle several cars
the capacity of IoT devices to operate for long periods without
ahead is braking.
recharging is a significant challenge [96], [98], [99]. Finally,
• Blindspot warning application; enables drivers to the rise of IoT and the emergence of LPWAN IoT
virtually view what is occurring in their blind spots. technologies, e.g., in traffic monitoring, PA and animal health
On the other hand, V2V IoT services can continuously monitoring, posed an additional challenge to power efficiency
monitor the vehicle’s condition and inform the driver in in long-range transmission situations [96], [100].
advance if something is likely to go wrong. Additionally, the As a result, battery life and the ability of end nodes to
IoT services can be used to notify medical personnel or the communicate over long ranges is one of the most crucial and
police in the event of an accident or emergency. Another problematic issues, particularly for off-grid and wearable IoT
application of V2V communication is path optimization, applications [96], [97]. Many studies have focused on
where critical travel-related information can be exchanged addressing this issue to analyze and optimize power
between vehicles to assist drivers in arriving at their performance for optimal IoT applications, considering both
destinations most efficiently. hardware and software perspectives [96]–[99], [101], [102].
Accordingly, low power systems or energy harvesting
I. Military
techniques such as micro-magneto-electric, thermoelectric,
Beyond traditional civilian contexts, IoT has many military
piezoelectric, or photoelectric technologies have lately been
and antagonistic uses. The US Army Research Laboratory
proposed as a feasible solution [99].
(ARL), for example, has created an Enterprise strategy to deal
with the complexities of the Internet of Battlefield Things B. Quality of Service (QoS)
(IoBT). ARL intends to build a new collaborative project (the Conventionally QoS was mainly determined by factors such as
IoBT CRA) to develop IoBT core capabilities for future bandwidth, jitter effect, packet loss, and network delay. To
military operations [93]. Similarly, IoT monitoring and provide superior IoT services, however, various other
tracking systems can enable connectivity during remote characteristics of the IoT network must be considered, including
military operations via device-to-device and device-to- network/server connection time, service level agreement
machine communications [44]. Furthermore, UIoT can be compliance, availability, and reliability. The authors in [103]
used in border surveillance applications such as border identified and classified major QoS metrics based on the main IoT
enforcement and infiltration prevention [60]. components, including communication, things, and computing.
On the other hand, UAVs have been widely used in military It is challenging to deliver guaranteed QoS, particularly for
applications. When outfitted with various communication IoT applications distributed in harsh environments or when IoT
devices, UAVs working in tandem with satellites and BSs nodes are placed near the ground, as in agricultural areas [62],
form a diverse three-tier (space-air-ground) network that [104]. In such conditions, communication suffers from severe
provides seamless coverage and expands the capacity of attenuation due to obstacles [104], thus adding complexity to the
increasingly profitable IoT networks [94]. network design phase [62]. Therefore, such an issue might
8 VOLUME XX, 2021
impact the QoS, causing poor node communication and an D. Security Concerns
unreliable IoT system that could increase the number of packet Despite all-new IoT-enabled capabilities, there is an increased
retransmissions and nodes’ energy consumption, resulting in security risk [46], [47], which could be caused by device failure,
wireless link failure [104]. Consequently, providing the nodes malicious attacks, unauthorized access, or poor key
with adequate resources to operate the system self-sufficiently management [46], [47], [116]. As a result, such security
without affecting QoS is another tricky task [101]. concerns will cause severe disruption to global IoT systems,
potentially outweighing their benefits [47]. A wide-scale
C. Localization security attack on a large IoT network, on the other hand, can be
IoT applications frequently demand location data from end costly and difficult to prevent. For instance, security attacks may
devices [105], [106]. As a result, data obtained for various result in factory shutdown and disruption to public safety [116].
applications, such as animal tracking, surveillance, On the other hand, wearable IoT devices are predicted to
autonomous vehicle guidance, and patient monitoring [43], have weaker security features due to design compromises to
[101], [106]–[108], is only meaningful if the precise location accomplish lightweight and low power consumption.
of sensor nodes is known [105], [106]. Maintaining low complexity in wearables is so challenging to
Typically, the position can be obtained using Global attain with enhanced security standards [99]. In this regard,
Navigation Satellite Systems (GNSS), such as the Global practical and reliable security algorithms must be considered,
Positioning System (GPS), which is viewed as an effective taking both security and reliability into account [116].
outdoor localization solution [106], [109]. These methods,
however, are not always viable, particularly in indoor E. Accurate Wireless Channel Propagation Modeling
environments [105], [106]. Where, for example, such and Characterization
environments may impose numerous multi-path effects due to For wireless systems, the surrounding environment, such as
obstacles between satellites and users [106], [109], resulting in terrain, plant height and density, and obstacles, can easily
significant signal degradation [106]. GPS is also recognized affect the received signal [71]. As a result, these channel
for its poor accuracy, which is limited to a maximum of five imperfections cause signal power loss, which affects QoS,
meters [109] and a solution that requires more energy and causing poor communication and thus a higher number of data
costs [105]. Aside from that, the cost, power, and size packet retransmissions between nodes. Such issues ultimately
limitations of IoT technologies, particularly LPWAN, impose result in higher power consumption in nodes and radio link
further constraints on embedding a GPS receiver into each end failure, leading to an inefficient IoT application [104].
device [110]. Hence, localization is viewed as one of the most Wireless channel propagation modeling and
difficult IoT-related challenges [106], [109], [111]. characterization are thus required for the design and evaluation
Numerous IoT localization techniques have been developed of robust wireless IoT systems [71], [104], [117]. Path loss
recently [105], [110], [112], to address the existing challenges (PL) modeling, in particular, enables an accurate estimation of
and introduce accurate, energy-efficient, low-cost localization IoT system propagation behavior and interference [71], [117].
techniques [105], [113]–[115]. These techniques are typically Thus, this enables the best possible estimation of the range and
classified into three categories [43], [101], [107], as follows: coverage between adjacent IoT terminals, together with an
First is range-based localization techniques, which rely on estimate of the number of nodes required to cover a specific
the range to identify the position of nodes or objects after a area [62], [104], [118], [119]. In agriculture IoT, for example,
series of ranging and computing phases [101], [107]. These the widespread deployment of IoT-based WSNs would be
techniques can thus be based on methods such as Time of reliant on unique factors impacting PL model accuracy in such
Arrival (TOA), RSSI, weighted centroid-locating algorithm, situations, predicting nodes height and coverage range, and
Time Difference of Arrival (TDOA), Angle of Arrival (AOA), enhancing transmission quality [71].
and Phase of Arrival (POA) [101], [106], [107]. In conclusion, wireless channel propagation modeling and
Second is range-free localization techniques; these do not characterization are crucial for sophisticated wireless IoT
employ any measuring techniques and usually necessitate systems and related to the other challenges this section
special hardware to do the ranging, which is subsequently used mentions. As a result, its accuracy is crucial for properly
to compute coordinates [101]. Thus, these approaches assessing and optimizing network efficiency amid deployment
prioritize cost-effectiveness over range-based methods [107] stages. It also contributes to energy savings, precise node
and are frequently utilized in mobile scenarios where the localization (for RSSI-based techniques), reduced network
precise location of nodes is unknown [101]. interference, and enhanced network capabilities, all of which
Finally, the third category combines both range-free and improve QoS in various deployment scenarios [119].
range-based techniques [107]. Among the three categories, Due to the significance of this constraint, several studies
range-based localization techniques are the most extensively have been conducted recently to evaluate the effectiveness of
utilized, particularly the RSSI-based localization technique wireless channel models or propose improved/new models for
[105]–[107], [110], which requires no additional hardware or wireless IoT technologies. As such, the next sections will
synchronization in network end devices [105], [106]. focus on this challenge, providing an overview of wireless
VOLUME XX, 2021 9
channel modeling and characterization concepts, current well- dispersion, leading the signal to be distorted [6], [120]. Time
known and regularly used models, and a survey of the most dispersion may also cause inter-symbol interference (ISI).
recent works aiming at wireless channel modeling. Moreover, distortion can arise due to object movement,
resulting in frequency dispersion due to Doppler spread [6].
IV. Wireless Channel Propagation Modeling and a As a result, wireless channel propagation modeling
Summary of Commonly Used Models commonly referred to as “channel modeling” or “propagation
The wireless channel is the physical link between the modeling,” is essential for the design of wireless
transmitter and the receiver through which the carrier signal communication systems [6], [114], [120], [123]. It employs
transmits data [6]. The signal interacts with various obstacles mathematical parameters to determine the impact of the channel
on its path to the receiver, as shown in Fig. 3, leading to medium on the transmitted signal [6], [31]. These models are
various impacts that cause the signal to be either destructive or classified into two groups: deterministic (theoretical) and
constructive during propagation [6], [120]. These obstacles stochastic channel models [6], [118], [119], which can contain
cause the signal to be reflected, refracted, or diffracted, a variety of randomly varying parameters [6].
resulting in signal attenuation (by absorption) and inducing Others may classify channel models into three groups:
scattering and secondary waves [120], [121]. Additionally, geometry-based deterministic models such as ray-tracing,
when an antenna radiates its signal, the signal can take several non-geometric stochastic models such as empirical models,
paths to the receiver. Each signal may interact chaotically with and geometry-based stochastic models (GBSM), also known
the environment and arrive at the receiver marginally delayed. as semi-empirical models, which combine deterministic and
As a result, depending on whether the delayed signals are in or empirical models [6], [125].
out of phase, they can introduce constructive or destructive The stochastic modeling approach is based on the statistical
interference. The spread of this delay is known as delay spread distributions of channel parameters and can be narrowband or
(DS), and the attenuation associated with it is known as multi- wideband [6]. Meanwhile, deterministic relies on knowledge
path fading [120]. of the physical characteristics of the wireless channel [118],
Subsequently, propagation in wireless channels results in [119] and is based on Maxwell equations [6]. This approach
either large- or small-scale fading based on signal variations has the drawback of increasing computing complexity and
[6], [122]. In turn, large-scale fading can be caused by either requiring costly site-specific designs and extremely accurate
PL as a function of distance or shadowing / slow-fading 3D maps [6], [118], [126]. In contrast, stochastic modeling
caused by big fixed obstacles such as mountains and buildings approaches, such as empirical models, are based on real
[6], [28], [123]. Meanwhile, destructive interference from measurements of wireless channels. These models have the
multi-path effects and small scatterers causes small-scale advantage of being simple to develop and flexible enough to
fading / fast-fading / scattering [6], [120], [122], [123]. Small- incorporate all environmental factors influencing signal
scale fading is frequently used to evaluate link-level propagation during real-world measurements [118], [119].
performance using bit error rates and average fading [6], However, such models only include frequency and distance,
[124]. In small-scale fading, the amplitude distribution of the leaving out reflections, diffraction, and refraction
received signal is typically modeled using a probability characteristics. Another disadvantage of such models is the
distribution, such as Rayleigh, Rician, and m-Nakagami. [6], uniform representation of measurement data acquired in a
[28], [122]–[124]. Signal delays caused by reflections, specific area with varying terrain and climate [4].
scattering, and diffraction, on the other hand, cause time
FIGURE 3. A Basic Wireless Signal Transmission with Various Attenuation Scenarios due to Channel Imperfection.
10 VOLUME XX, 2021
Impulse response within a delay bin in the narrowband can models, such as Okumura, Lee, and Stanford University
result in non-selective flat or frequency fading. On the other Interim (SUI), and deterministic models, such as the Air-to-
hand, Wideband focuses on each channel response ray Ground (ATG) model, as given in Eq. (12) to Eq. (15), are
independently and can be signal dispersion or selective examples of these models. There are also many more highly
frequency fading. As a result, the stochastic narrowband complex but extremely popular models in this category, such
model is commonly used to characterize fading statistics and as the Longley-Rice Irregular Terrain Model (ITM) [31],
the Doppler spectrum. The stochastic wideband model, on the [123], [127], and Irregular Terrain with Obstructions Model
other hand, often focuses on received power, delay, departure (ITWOM) [127], [129].
and arrival angle, and Doppler shift [6]. Despite their accuracy, empirical terrain models have
Having stated that, numerous channel models have been several limits due to low antenna heights, limited prediction
described in past years [28], [31]. However, there is a lack of distances, and other area-specific limitations. Meanwhile, ray-
studies in the literature that summarize and describe available tracing ATG epitomizes deterministic models by employing
channel models suitable for a broad range of wireless IoT- Maxwell’s equations as well as the laws of reflection and
based systems. As such, this section summarizes several well- diffraction. The ATG propagation model is widely used in
known and frequently used PL model equations for various Low Altitude Platforms (LAP) since it offers enhanced cell
communication scenarios and environments, as given in Table capacity and downlink coverage. PL is calculated using a
I. Where 34 equations, from Eq. (1) to Eq. (34), are listed and closed-form approach between Tx and Rx based on two major
classified according to model category (empirical, semi- ATG propagation categories. The first is LOS or near-LOS,
empirical, or deterministic). For simplicity, the models are whereas the second is NLOS but still receives signals with
further categorized into eight sub-categories based on their significant reflections and diffractions [130].
intended use scenario, as follows: The ITM model is used at frequencies ranging from 20
MHz to 20 GHz. It is a two-part system with an ITM core and
A. Channel Model Foundation an input-output module. This model is generally used for
This set of models includes those models driven purely from point-to-point (P2P) communications and uses terrain data to
the idealized theory of electromagnetic propagation, including measure PL in P2P mode. This model, however, is based on
the Friis equation of Free Space Path Loss (FSPL), Eq. (1), traditional diffraction theory, which does not incorporate
and two-Ray (2-Ray) for Plane Earth (PE) ground reflection precise radio wave calculations over irregular terrain. It also
model, Eq. (2). These two models are considered deterministic does not account for environmental factors near the Rx [31],
(theoretical) and are widely used as baseline models in most [127]. ITWOM, on the other hand, is an extension to ITM that
comparison studies and more complex models derivation incorporates high location estimates and better accuracy into
[123]. the ITM core. In addition, unlike ITM, it considers more than
one obstacle in its diffraction calculations [127]. As a result,
B. Basic the latter two models are commonly utilized in ray-tracing and
This set of models includes commonly used models, typically coverage estimation tools like Cloud-RF® and Radio
empirical and based on basic inputs such as distance, Mobile® [31], [127], [129].
frequency, antenna height, and gain, with measurement-based
area-specific tuning [123]. The Okumura-Hata model, given D. Basic/Stochastic Fading and Basic/Terrain
in Eq. (3), is the most popular. It is an empirical model with This set of models consists of two sub-categories of basic
environmental type parameters and is valid for frequencies models. These models comprise Eq. (16) to Eq. (21), with the
ranging from 150 MHz to 1500 MHz, 30 m to 200 m Tx/GW, last equation being the Egli PL model, which combines a basic
and 1 m to 10 m Rx/ED. One drawback of the Okumura-Hata PL model with an empirical terrain loss factor; although,
model is that its estimation does not consider the terrain profile diffraction losses are not explicitly considered [131].
[31], [127]. Due to its popularity, several models adapted the On the other hand, stochastic fading models, as given in Eq.
Okumura-Hata model to expand its supported frequency, (16) to Eq. (20), are similar to basic models except that they
distance, and Tx/Rx height [120], [128]. These models include incorporate one or more random variables to account for
Hata-Davidson, ILORIN (an optimization of the Hata- channel variance [123]. Numerous models fall into this
Davidson model based on measurements in Nigeria), CCIR, category, with the Log-Normal Shadowing PL (LNSPL) model,
COST231-Hata, Extended COST 231, ECC-33, Ericsson as defined in Eq. (16), being the most widely used due to its
9999, and ITU-R P.529-3, as given in Eq. (4) to Eq. (11), simplicity and flexibility to account for a wide variety of
respectively. environmental and propagation scenarios [117]. As a result, it is
commonly applied in the PL modeling of most IoT-based
C. Terrain wireless technologies [117], [114]. LNSPL is a one-slope
In contrast to the basic models, this set of models is commonly model, typically known as a one-slope log-distance model if the
considered more complex and considers the estimation of PL estimation does not include the random variable. This model
diffraction losses due to terrain obstacles [123]. Empirical is also adopted in both WINNER II and 3GPP standards and is
VOLUME XX, 2021 11
referred to as the Floating Intercept (FI) or Alpha-Beta (AB) PL The W MED model was proposed for dense foliage areas in
model [132]–[135]. It is based on floating-intercept and line temperate climates with propagation paths blocked by dry-leaf
slope to provide the best minimum PL error fit. This model’s trees [5], [9], [139]. By contrast, the ITU-R model was
parameters are usually derived through curve fitting and developed utilizing Weissberger’s approach [9] and VHF /
minimum mean square error (MMSE) approaches [132], [134]. UHF frequency band measurements [5], [139]. The latter was
Another common model in the category of stochastic fading then refined further using data from the VHF-mmWave
models is the so-called Close-In (CI) PL model [6], [134], frequency band, comprising trees with and without leaves over
[136], as given in Eq (17). 3GPP adopts this model for Urban a short foliage depth [5]. Similarly, the COST-235 model was
Microcell (Umi), Urban Macrocell (Uma), and Indoor Hotspot proposed utilizing Weissberger’s method and mmWave
(InH) using a free-space reference PL at a reference distance measurements through a small grove of trees [5], [139].
of 1 m [6], [136]. Therefore, it relies on a single parameter Finally, the Chen & Kue model was proposed based on
optimization of a distance-dependent PLE [134], [136]. CI diffraction theory with a four-layer medium to address
model also has an intrinsic PL frequency dependence within propagation impact in a forest environment utilizing
the close-in (1 m) FSPL value [135]. It thus models frequency measurements over frequencies from 1-100 GHz [9].
dependence across a wide range of frequencies and exhibits The latter exponential decay models are purely empirical
good precision and parameter stability across various outdoor and are therefore not constrained by inherent propagation
settings, distances, and frequency ranges [136]. mechanisms. As a result, several semi-empirical models have
Several models have also been suggested to further modify been proposed to address this issue [54]. These models may
the CI model, such as the Close‐In with Frequency-Dependent include the current ITU-R foliage model (P.833-9) [54], also
Exponent (CIF) model [6], [132], [136], given in Eq. (18), and known as the Maximum Attenuation (MA) model [54], [139],
the Close‐In with Hight Dependent Exponent (CIH) model [140], as well as the Non-Zero Gradient (NZG) model [54]
[136], [137], given in Eq. (19). The CIF model was proposed and the Dual Gradient (DG) model [139], as specified in Eqs.
to extend the CI model by modifying the PLE to compensate (27), (28), and (29). The MA model is recommended for
for frequency dependency empirically observed in the terrestrial radio paths in wooded areas [140] and can operate
environment while maintaining a straightforward physical at frequencies ranging from 30 MHz to 100 GHz [54].
basis of frequency dependency due to Friis’ equation at Meanwhile, the NZG and DG models expand the MA model
reference distances of 1 m [6], [137]. On the other hand, the beyond 5GHz [54], [139].
CIH model modifies the CI model by adjusting the PLE to The Tewari model, provided in Eq. (30), is an empirical
account for the Tx antenna height dependency in the PL model proposed for frequencies ranging from 50 MHz to 800
calculation [136], [137]. Finally, Eq. (20) provides the Alpha- MHz, considering both terrain and foliage loss. Thus, this
Beta-Gamma (ABG) PL model [6], [132], [134] that is a model is listed in Table I under the Foliage/Terrain sub-
multi-frequency [132], [134] and the current 3GPP 3D model category. This model was further modified in the Extended
[6]. Due to its three parameters, it always has a lower shadow Tewari model, as given in Eq. (31), by including an antenna
fading STD value than other models [6]. height gain factor to account for the effect of Tx and Rx
antenna height above terrain [54].
E. Foliage and Foliage/Terrain
Another significant constraint in signal propagation is foliage F. Indoor
attenuation. It is shown to be a function of various parameters, Propagation imperfections are primarily caused by reflection
including frequency, foliage depth, tree species, foliage and diffraction from/around building objects, walls, and floors
thickness, leaf density, leaf size, branches, trunks, humidity, for indoor environments. Additional loss happens due to
wind speed, tree height relative to antenna heights, and foliage signals traveling through these objects and moving objects,
length [138]. Furthermore, foliage attenuation is expected to resulting in temporal and spatial signal fluctuation [141]. As a
be greater in tropical areas where foliage is relatively result, estimating and modeling indoor propagation is critical
comparable to or larger than wavelength plus humidity. for indoor applications, gaining significant research and
Interestingly, recent field studies of tropical foliage development interest [142]. Accordingly, several models have
attenuation showed that the presence of a single tree in a link been proposed to compensate for these indoor propagation
could cause the signal to suffer foliage attenuation [5]. imperfections and model the PL in such environments. These
Numerous models exist to predict the excess loss due to models may include the Cost231 Multi-Wall Multi-Floor
foliage attenuation, which is referred to in this sub-category as (Cost231-MWF) model [142], the ITU-R Indoor (P.1238-10)
foliage models. These models may include empirical foliage model [141], and the Multi-Wall Multi-Floor (MWMF) model
models such as Weissberger’s Modified Exponential Decay [142], [143], as represented in Eqs. (32), (33), and (34).
(W MED), ITU-R, Fitted ITU-R (FITU-R), COST-235 [5], Indoor PL is addressed in the Cost231-MWF model by
[9], [54], [139], and Chen & Kue [9], as specified in Eqs. (22), incorporating an FSPL with additional attenuation losses due to
(23), (24), (25), and (26). building walls and floors [142]. Meanwhile, the ITU-R P.1238-
10 is a site-specific model that requires minimal information
12 VOLUME XX, 2021
about the signal path and the site while spanning a broad the 2-Ray model was investigated. According to [145], the FI
frequency range of 300 MHz to 450 GHz. It is typically an model does not physically represent either the LOS or NLOS
empirical basic PL model that considers the various attenuations channel. Also, the small STD difference shows that the CI
of walls and floors and the related shadowing impact under model may be better suited for closed-form analysis than the
certain conditions [141]. On the other hand, the MWMF model FI and ABG models. The proposed model was compared to
was proposed utilizing ray-tracing measurements, considering the CI, FI, and ABG models, which showed that the PL might
the nonlinear relationship between total attenuation and the be modeled more accurately utilizing the proposed model with
number of crossed walls or floors [142], [143]. a single parameter (PLE). According to [147], the PLE of the
Apart from the models discussed in the preceding sub- CI and FI were similar, around 1.8 and 2.9, respectively, for
sections, various models have been introduced over the years. LOS and NLOS scenarios. Conversely, the results showed that
These models include 3GPP TR 38.901 and ITU-R M.2412, the FI model gave the best minimum error fit and, just like the
which support a wide range of transmission scenarios for a CI model, would be the most suitable for indoor PL modeling
wide range of frequencies and bandwidths, such as Outdoor- of a 5G network operating at 40 GHz.
to-Outdoor (O2O), Outdoor-to-Indoor (O2I), and Indoor-to- In contrast, [132] conducted a comparative study for the
Indoor (I2I) [128]. indoor 5G channel, then proposed two models that addressed
loss due to wall edge diffraction and high-frequency band. The
V. Channel Modeling and Wireless Propagation of IoT results showed that the PLE values for the LOS were 1.6 and
Technologies 1.3 at 3.5 and 28 GHz, respectively. However, the power
This section thoroughly reviews the most recent developments received was dropped in the NLOS, with PLE values being 2.7
and studies aimed at wireless channel modeling and and 3.6 at 3.5 GHz and 28 GHz, respectively. The results also
characterization of wireless IoT technologies. For simplicity, indicated that FI and ABG models provided reliable PL
the section will be divided into three sub-sections according to performance in the LOS scenario for single and multi-
wireless IoT technology categories, as follows: frequency models. At 3.5 GHz and 28 GHz, the average
diffraction loss was 11.11 dB and 23.37 dB, respectively.
A. Cellular-based wireless IoT channel modeling and Meanwhile, the frequency-related loss, referred to as
characterization
frequency drop, was 19.73 dB for the LOS and 32 dB for the
The wireless communications sector is undergoing significant
NLOS. The root mean square delay spread (RMS-DS) values
growth. In this regard, the fifth-generation (5G) of mobile
for the LOS and NLOS scenarios were less than 8 ns and 12
communication technology has been designed to address all
ns, respectively. These results indicated that the 5G channel
wireless broadband communication needs and enable massive
has excellent PL performance and a very small DS, enabling
IoT deployments over the next decade [144], [145]. While
future real-world deployments of 5G-based smart city IoT
spectrum in the existing sub-6 GHz band remains constrained,
networks.
new frequency bands are required to enable the efficient
On the other hand, [148] presented an empirical
design of IoT applications in 5G networks [144]. As a result
characterization of the mmWave frequency bands, 6.5, 10.5,
of its vast bandwidth, which spans 30 GHz to 300 GHz, the
15, 19, 28, and 38 GHz, in an indoor corridor scenario. Over
mmWave band is regarded the leading choice for adoption
4,000 power delay profiles (PDP) were observed overall bands
[132], [144]. However, the electromagnetic (EM) wave
using an omnidirectional Tx antenna and a highly directional
characteristics of these bands present several challenges in
horn Rx antenna in co- and cross-polarized setups. After that,
terms of coverage limits, signal attenuation, PL, penetration
a new PL model was developed to account for frequency
loss, diffraction, and scattering [125], [144], [146]. Besides,
attenuation due to distance, termed the frequency attenuation
buildings and other structures may block the mmWave band
(FA) PL model, which included a frequency-dependent
signal [125]. Accordingly, several studies have addressed
attenuation factor. Additionally, a more generic and simpler
these challenges and characterized the 5G frequency bands’
method for estimating the cross-polarization discrimination
wireless channel characteristics [145].
(XPD) factor of close-in reference distances using XPD (CIX)
For instance, the studies in [132], [145], [147], [148]
and ABG with XPD (ABGX) PL models was proposed to
considered a variety of indoor scenarios and a broad range of
avoid the computational complexity associated with the
frequencies to evaluate the channel’s behavior. In [145],
MMSE approach. Small-scale metrics were considered to
various channel models were studied for the 4.5, 28, and 38
describe multipath channel dispersion, including RMS-DS,
GHz frequency bands, including CI, FI, and ABG. Then, a
mean excess delay, dispersion factors, and maximum excess
new hybrid probabilistic PL model was proposed, based on the
delay. Additionally, multiple RMS-DS statistical distributions
CI PL model, for directional and omnidirectional antennas in
were investigated. The results implied that the proposed
LOS and NLOS scenarios. Meanwhile, [147] presented
models are more physically based and simpler than other well-
measurements of an indoor cellular system operating at 40
known models. RMS-DS values ranged from 0.2 to 13.8 ns,
GHz in LOS and NLOS scenarios. A single frequency PL
with dispersion factors less than one for all measured
model was thus proposed, based on the CI and FI models, and
frequencies.
VOLUME XX, 2021 13
TABLE I
WELL-KNOWN PATH LOSS MODELS FOR WIRELESS IOT CHANNEL MODELING UNDER DIFFERENT COMMUNICATION SCENARIOS AND ENVIRONMENTS.
Category/Type/Model PL in dB Eq. Parameters
32.44 + 20 log10(𝑓[𝑀𝐻𝑧]) + 20 log10(𝑑[𝑘𝑚])
[54],[130],
FSPL
𝑓 is the frequency
[149],
[150]
𝑃𝐿𝐹𝑆 (𝑓, 𝑑) = { 32.44 + 20 log10 (𝑓[𝐺𝐻𝑧]) + 20 log10(𝑑[𝑚]) (1)
Deterministic
Foundational
𝑑 is Tx-Rx separation distance

92.45 + 20 log10(𝑓[𝐺𝐻𝑧]) + 20 log10 (𝑑[𝑘𝑚])
𝑃𝐿𝐹𝑆 (𝑓, 𝑑) 𝑓𝑜𝑟 𝑑 < 𝑑𝐶
𝑃𝐿2−𝑅𝑎𝑦 = {
2-Ray
[139]
𝑃𝐿𝑃𝐸 𝑓𝑜𝑟 𝑑 ≥ 𝑑𝐶 , 𝑑 > ℎ𝑇𝑥 , ℎ𝑅𝑥 𝑑 in m; 𝑑𝐶 is Crossing distance; ℎ𝑇𝑥 and ℎ𝑅𝑥
[54],
(2)
𝑃𝐿𝑃𝐸 = 40 log10(𝑑) − 20 log10(ℎ𝑇𝑥 ) − 20 log10 (ℎ𝑅𝑥 ) are Tx and Rx antenna height in m.
{ 𝑑𝐶 = (4𝜋ℎ𝑇𝑥 ℎ𝑅𝑥 )/𝜆
𝑃𝐿𝑂𝐻𝑈 𝑓𝑜𝑟 𝑈𝑟𝑏𝑎𝑛
Okumura-Hata [149]–[152]
𝑃𝐿𝑂𝐻𝑈 − 2(log10(𝑓/28))2 − 5.4 𝑓𝑜𝑟 𝑆𝑢𝑏𝑢𝑟𝑏𝑎𝑛

𝑃𝐿𝑂𝐻 =
𝑃𝐿𝑂𝐻𝑈 − 4.78(log10(𝑓))2 + 18.33 log10(𝑓) − 40.94 𝑓𝑜𝑟
{ 𝑅𝑢𝑟𝑎𝑙
69.55 + 26.16 log10(𝑓) − 13.82 log10 (ℎ𝑇𝑥 ) (3)
𝑃𝐿𝑂𝐻𝑈 = {
−𝑎(ℎ𝑅𝑥 ) + (44.9 − 6.55 log10 (ℎ𝑇𝑥 )) log 10 (𝑑)
8.29(log10(1.54ℎ𝑅𝑥 ))2 − 1.1 𝑈𝑟𝑏𝑎𝑛, 𝑓 ≥ 300 𝑀𝐻𝑧
2
𝑎(ℎ𝑅𝑥 ) = { 3.2(log10(11.75 ℎ𝑅𝑥 )) − 4.97 𝑈𝑟𝑏𝑎𝑛, 𝑓 ≤ 300 𝑀𝐻𝑧
(1.1 log 10 (𝑓) − 0.7)ℎ𝑅𝑥 − (1.56 log10(𝑓) − 0.8) 𝑓𝑜𝑟
{ 𝑆𝑢𝑏𝑢𝑟𝑏𝑎𝑛 𝑎𝑛𝑑 𝑅𝑢𝑟𝑎𝑙
Davidson
[130],
𝑃𝐿𝐻𝐷 = 𝑃𝐿𝑂𝐻 + 𝐶𝐻𝐷

[149]
Hata-
{ (4)
𝐶𝐻𝐷 = 𝐴(ℎ𝑇𝑥 , 𝑑) − 𝑆1 (𝑑) − 𝑆2 (ℎ𝑇𝑥 , 𝑑) − 𝑆3 (𝑓) − 𝑆4 (𝑓, 𝑑) 𝑓 is the frequency in MHz (except for ECC-33
and ITU-R P.529-3, where the frequency is
given in GHz),
ILORIN
𝑃𝐿𝐼𝐿𝑂𝑅𝐼𝑁 = 73.56 + 26.16 log10(𝑓) − 13.82 log10(ℎ𝑇𝑥 )

[149]
𝑑 is Tx-Rx separation in km (except for ITU-R

{ (5)
−𝑎(ℎ𝑅𝑥 ) + 30.5 log10(𝑑) + 𝐶𝐻𝐷 P.529-3, where the Tx-Rx separation distance
is given in m),
𝑎(ℎ𝑅𝑥 ) is the effective Rx antenna height
CCIR
[149]
𝑃𝐿𝐶𝐶𝐼𝑅 = 𝑃𝐿𝑂𝐻𝑈 − (30 − 25 log10(𝑏𝜌 )) for Suburban (6) correction factor in dB,
ℎ𝑇𝑥 & ℎ𝑅𝑥 are Tx & Rx antenna heights in m,
𝐶𝐻𝐷 is Hata-Davidson correction parameter,
𝑃𝐿𝐶231_𝐻 = 46.3 + 39.9 log10(𝑓) − 13.82 log10( ℎ𝑇𝑥 ) − 𝑎(ℎ𝑅𝑥 ) 𝐴(ℎ𝑇𝑥 , 𝑑) and 𝑆1 (𝑑) are Hata-Davidson model
[149], [151],
COST231-
Hata [2],
+(44.9 − 6.55 log10 ( ℎ𝑇𝑥 )) log10( 𝑑) + 𝐶𝑚 correction factors to extend Okumura-Hata

[152]
(1.1 log10(𝑓 ) − 0.7)ℎ𝑅𝑥 − (1.56𝑓 − 0.8)𝑓𝑜𝑟 𝑆𝑢𝑏𝑢𝑟𝑏𝑎𝑛 (7) model maximum supported distance to 300km,
𝑎(ℎ𝑅𝑥 ) = {
3.2(log10(11.75 ℎ𝑅𝑥 ))2 − 4.97 𝑓𝑜𝑟 𝑈𝑟𝑏𝑎𝑛 𝑆2 (ℎ𝑇𝑥 , 𝑑) is Hata-Davidson model correction
Empirical
{ 𝐶𝑚 = 0 𝑑𝐵 𝑓𝑜𝑟 𝑆𝑢𝑏𝑢𝑟𝑏𝑎𝑛 𝑎𝑛𝑑 𝑅𝑢𝑟𝑎𝑙, 3 𝑑𝐵 𝑓𝑜𝑟 𝑈𝑟𝑏𝑎𝑛 factor that extends Okumura-Hata model
Basic
supported Tx antenna height to 2500m,

Extended
𝑃𝐿𝐶231_𝐻 𝑓𝑜𝑟 𝑈𝑟𝑏𝑎𝑛 𝐴𝑟𝑒𝑎 𝑆3 (𝑓) and 𝑆4 (𝑓, 𝑑) are Hata-Davidson model
COST
[150]
231
𝑃𝐿𝐸𝑥_𝐶231 = { (8) correction factors to extend Okumura-Hata

𝑃𝐿𝐶231_𝐻 − 2(log10(𝑓/28))2 − 5.4
model frequency to 1.5GHz,
𝑃𝐿𝐸𝐶𝐶−33 = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑) + 𝐴𝑏𝑚 − 𝐺𝑇𝑥 − 𝐺𝑅𝑥 𝑏𝜌 is the percent of the building covered area,
𝐶𝑚 is an environment correction factor in dB,
ECC-33 [2],
𝐴𝑏𝑚 = 20.41 + 9.83 log10(𝑑) + 7.89 log10(𝑓)

[152], [153]
𝐴𝑏𝑚 is a basic medium loss,

+9.56(log10(𝑓/28))2
(9) 𝐺𝑇𝑥 & 𝐺𝑅𝑥 are Tx and Rx antenna height gain
𝐺𝑇𝑥 = log10(ℎ𝑇𝑥 /200)(13.958 + 5.8 log10(𝑑))2 factor,
(42.57 + 13.7 log10(𝑓))(log10(ℎ𝑅𝑥 ) − 0.585) ∝0 , ∝1 , ∝2 , and ∝3 are different environment
𝐺 ={
{ 𝑅𝑥 0.759(ℎ𝑅𝑥 ) − 1.862 𝑓𝑜𝑟 𝑙𝑎𝑟𝑔 𝑐𝑖𝑡𝑦 correction constants for Ericsson 9999 model,
𝐺(𝑓) is a frequency correction factor for
[151], [153]
9999 [149],
𝑃𝐿𝐸𝑟 =∝0 +∝1 log 10 (𝑑) +∝2 log10(ℎ𝑇𝑥 ) +∝3 log10(ℎ𝑇𝑥 ) log10(𝑑)
Ericsson
Ericsson 9999 model,

{ −3.2(log10(11.75ℎ𝑅𝑥 ))2 + 𝐺(𝑓) (10) ℎ′𝑇𝑥 is a correction factor for Tx antenna
𝐺(𝑓) = 44.49 log10(𝑓) − 4.78(log10 (𝑓))2 height,
𝑏 is a distance correction factor
𝑃𝐿𝑃5293𝑈 𝑓𝑜𝑟 𝑈𝑟𝑏𝑎𝑛
𝑃𝐿𝑃5293𝑈 − 2(log10(𝑓/28))2 − 5.4 𝑓𝑜𝑟 𝑆𝑢𝑏𝑢𝑟𝑏𝑎𝑛
𝑃𝐿𝑃5293 =
𝑃𝐿𝑃5293𝑈 − 4.78(log10(𝑓))2 + 18.33 log 10 (𝑓)
ITU-R P.529-3 [130]
{ −40.94 𝑓𝑜𝑟 𝑅𝑢𝑟𝑎𝑙

69.55 + 26.16 log10(𝑓) − 13.82 log10(ℎ𝑇𝑥 )
𝑃𝐿𝑃5293𝑈 = {
−𝑎(ℎ𝑅𝑥 ) + (44.9 − 6.55 log10(ℎ𝑇𝑥 ))(log10(𝑑))𝑏
(11)
𝑎(ℎ𝑅𝑥 ) = (1.1 log10 (𝑓) − 0.7)ℎ𝑅𝑥 − (1.56 log10(𝑓) − 0.8)
1 𝑓𝑜𝑟 𝑑 ≤ 20𝑘𝑚
𝑏 = { 1 + (0.14 + 1.87 ∗ 10−4 𝑓) + (1.07 ∗ 10−3 ℎ′𝑇𝑥 )
∗ (log10(𝑑/20000))0.8 𝑓𝑜𝑟 20𝑘𝑚 < 𝑑 ≤ 100𝑘𝑚
ℎ′𝑇𝑥 = ℎ𝑇𝑥 /√1 + (7 ∗ 10−6 ∗ ℎ2𝑇𝑥 )

{
14 VOLUME XX, 2021
TABLE I (CONTINUED)
𝑓 is the frequency in MHz,
𝑑 is the Tx-Rx separation distance in km,
Okumura [130],
𝑃𝐿𝑂𝑘𝑚 = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑) + 𝐴𝑚𝑢 (𝑓, 𝑑) − 𝐺(ℎ𝑅𝑥 ) − 𝐺(ℎ𝑇𝑥 ) − 𝐺𝑎𝑟𝑒𝑎 𝐴𝑚𝑢 is a medium attenuation factor relative to FS
𝐺(ℎ𝑇𝑥 ) = 20 log10(ℎ𝑇𝑥 /200)
[149]
30𝑚 < ℎ𝑇𝑥 < 1000𝑚 attenuation,

(12)
20log10 (ℎ𝑅𝑥 /3) 𝑓𝑜𝑟 ℎ𝑅𝑥 ≤ 3 𝐺(ℎ𝑇𝑥 ) & 𝐺(ℎ𝑅𝑥 ) is the Tx/Rx antenna height
𝐺(ℎ𝑅𝑥 ) = { gain factor,
{ 10log10 (ℎ𝑅𝑥 /3) 𝑓𝑜𝑟 3𝑚 < ℎ𝑅𝑥 < 10𝑚
𝐺𝑎𝑟𝑒𝑎 is environment type gain,
ℎ𝑇𝑥 and ℎ𝑅𝑥 are Tx and Rx antenna height in m
𝑑0 is the reference distance,
Lee [150],
𝐴, and 𝐵 are environment correction factors,

[154]
𝑃𝐿𝐿𝑒𝑒 = 𝐴 + 𝐵 log10(𝑑/𝑑0 ) + 10𝑛 log10(𝑓/900) + 𝛼0

Empirical
{ (13) 𝑛 is a distance-dependent PL Exponent (PLE),

𝑑0 = 1.6km
𝛼0 is a correction factor to consider Tx and Rx
antenna heights
𝑑 is the Tx-Rx separation distance in m,
𝑃𝐿𝑆𝑈𝐼 = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑0 ) + 10𝛾 log10(𝑑/𝑑0 ) + 𝑋𝑓 + 𝑋𝑅𝑥 + 𝑋𝜎
Terrain
𝛾 is SUI effective PLE,

SUI [2], [149], [153],
𝑑 > 𝑑0 , 𝑑0 = 100𝑚 𝑜𝑟 1𝑚, 𝛾 = 𝑎 − 𝑏ℎ𝑇𝑥 + (𝑐/ℎ𝑇𝑥 ) 𝑎, 𝑏, 𝑎𝑛𝑑 𝑐 are model parameters for terrain
[155]–[157]
𝑋𝑓 = 6 log10 (𝑓(𝑀𝐻𝑧)/2000) type A (hilly, dense vegetation), B (hilly, rare

−10.8 log10 (ℎ𝑅𝑥 /2) 𝑓𝑜𝑟 𝑡𝑒𝑟𝑟𝑎𝑖𝑛 𝑡𝑦𝑝𝑒 𝐴 & 𝐵 vegetation), and C (flat, rare vegetation),
𝑋𝑅𝑥 = { (14)
−20 log10(ℎ𝑅𝑥 /2) 𝑓𝑜𝑟 𝑇𝑒𝑟𝑟𝑎𝑖𝑛 𝑡𝑦𝑝𝑒 𝐶 𝑋𝑓 is the correction factor for frequency > 2 GHz,
𝑎 = 4.6, 𝑏 = 0.0075, 𝑐 = 12.6, , 𝑓𝑜𝑟 𝑡𝑒𝑟𝑟𝑎𝑖𝑛 𝑡𝑦𝑝𝑒 𝐴 𝑋𝑅𝑥 is the correction factors for Rx antenna height,
𝑎 = 4, 𝑏 = 0.0065, 𝑐 = 17.1, , 𝑓𝑜𝑟 𝑡𝑒𝑟𝑟𝑎𝑖𝑛 𝑡𝑦𝑝𝑒 𝐵 𝑋𝜎 is the shadowing variable in dB, typically zero-
{ 𝑎 = 3.6, 𝑏 = 0.005, 𝑐 = 20, , 𝑓𝑜𝑟 𝑡𝑒𝑟𝑟𝑎𝑖𝑛 𝑡𝑦𝑝𝑒 𝐶 mean Gaussian random variable with standard
deviation 𝜎 in dB
𝜌𝐿𝑂𝑆 and 𝜌𝑁𝐿𝑂𝑆 are the probability of having
𝑃𝐿𝐴𝑇𝐺 =𝜌𝐿𝑂𝑆 𝑃𝐿𝐿𝑂𝑆 + 𝜌𝑁𝐿𝑂𝑆 𝑃𝐿𝑁𝐿𝑂𝑆 LOS/NLOS conditions at an elevation 𝜃;
Deterministic
𝑎−𝑏
ATG [130]
𝑃𝐿𝐿𝑂𝑆 and 𝑃𝐿𝑁𝐿𝑂𝑆 is the PL for LOS/NLOS

𝜌𝐿𝑂𝑆 = 𝑎 − , 𝜌𝑁𝐿𝑂𝑆 = 1 − 𝜌𝐿𝑂𝑆
𝜃−𝑐 2 (15)
condition,
1+( ) 𝜂𝐿𝑂𝑆 and 𝜂𝑁𝐿𝑂𝑆 are average additional loss
𝑑
𝑃𝐿𝐿𝑂𝑆 = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑) + 𝜂𝐿𝑂𝑆 , 𝑃𝐿𝑁𝐿𝑂𝑆 = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑) + 𝜂𝑁𝐿𝑂𝑆 depending on environment type,
{ 3𝑑𝐵 ≤ 𝜂𝐿𝑂𝑆 ≤ 5𝑑𝐵, 8𝑑𝐵 ≤ 𝜂𝑁𝐿𝑂𝑆 ≤ 12𝑑𝐵 𝑎, 𝑏, 𝑐, and 𝑑 are ITU-R parameters for
environment type (rural, suburban, urban)
𝑃𝐿𝐿𝑁𝑆𝑃𝐿 (𝑑) = 𝑃𝐿(𝑑0 ) + 10𝑛 log10(𝑑/𝑑0 ) + 𝑋𝜎 (16) 𝑓 is the frequency in GHz,
LNSPL
[114],
[133],

[158]
Known as Floating Intercept (FI) or Alpha-Beta (AB) PL model (for

[6],
𝑑 ≥ 1𝑚), where: 𝛽 = 𝑃𝐿(𝑑0 ), 𝛼 = 𝑛, 𝑋𝜎𝐹𝐼 = 𝑋𝜎 , 𝑛 is a distance-dependent PLE,

𝐶𝐼𝐹 𝐶𝐼𝐻
𝑃𝐿𝐹𝐼 (𝑑) = 𝛽 + 10𝛼 log10(𝑑/𝑑0 ) + 𝑋𝜎𝐹𝐼 𝑛𝑒𝑓𝑓 , and 𝑛𝑒𝑓𝑓 are effective PLE for CIF and
(17) CIH model,
CI [6],
[132]–
[136],
[159]
𝑏 is the linear frequency-dependent factor of

Basic/Stochastic Fading
𝑃𝐿𝐶𝐼 (𝑓, 𝑑) = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑0 ) + 10𝑛 log10(𝑑/𝑑0 ) + 𝑋𝜎 , 𝑑0 =1m CIF PL over considered reference frequency,
𝑏𝑇𝑥 , ℎ𝑇𝑥 , and ℎ𝑇𝑥0 are linear Tx height
Empirical
𝐶𝐼𝐹 (18) dependent PLE, Tx antenna height, average Tx

𝑃𝐿𝐶𝐼𝐹 (𝑓, 𝑑) = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑0 ) + 10𝑛𝑒𝑓𝑓 log10(𝑑/𝑑0 ) + 𝑋𝜎𝐶𝐼𝐹
[132],
[133]
CIF
[6],
{ , 𝑑0 =1m antennas heights in a set of measurements,

𝐶𝐼𝐹
𝑛𝑒𝑓𝑓 = 𝑛(1 + 𝑏((𝑓 − 𝑓0 )/𝑓0 ) 𝛼, and 𝛾 are constant coefficients used to
𝐶𝐼𝐻 indicate dependencies of frequency and
𝑃𝐿𝐶𝐼𝐻 (𝑓, 𝑑, ℎ𝑇𝑥 ) = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑0 ) + 10𝑛𝑒𝑓𝑓 log10(𝑑/𝑑0 ) + 𝑋𝜎𝐶𝐼𝐻 (19)
distance on PL,
[136]
{ ,
CIH
𝐶𝐼𝐻
𝑛𝑒𝑓𝑓 = 𝑛(1 + 𝑏𝑇𝑥 ((ℎ𝑇𝑥 − ℎ𝑇𝑥0 )/ℎ𝑇𝑥0 ) 𝛽 is offset in PL,
𝑑0 =1m 𝑋𝜎 , 𝑋𝜎𝐹𝐼 , 𝑋𝜎𝐶𝐼𝐹 , and 𝑋𝜎𝐶𝐼𝐻 are zero-mean
Gaussian random variables with standard
[132]–
𝑃𝐿𝐴𝐵𝐺 (𝑓, 𝑑) = 10𝛼 log10(𝑑/𝑑0 ) + 𝛽 + 10𝛾(𝑓/𝑓0 ) + 𝑋𝜎𝐴𝐵𝐺 ,

[135],
ABG
deviation 𝜎 in dB
[159]
[6],
(20)
𝑑0 =1m, 𝑓0 = 1𝐺𝐻𝑧, 𝑓 ≥ 1𝐺𝐻𝑧 𝑋𝜎𝐴𝐵𝐺 is a Gaussian random variable with a
standard deviation 𝜎 𝐴𝐵𝐺 in dB.
Basic/Terrain
𝑓 is the frequency in MHz, 𝑑 is Tx-Rx

Empirical
𝑃𝐿𝐸𝑔𝑙𝑖 = 20 log10(𝑓) + 40 log10 (𝑑) − 20log10(ℎ𝑇𝑥 ) + 𝐶(ℎ𝑅𝑥 )

[131],
[151]
separation in km, ℎ𝑇𝑥 and ℎ𝑅𝑥 are Tx and Rx

Egli
{ 76.3 − 10 log10 (ℎ𝑅𝑥 ) 𝑓𝑜𝑟 ℎ𝑅𝑥 < 10𝑚 (21)

𝐶(ℎ𝑅𝑥 ) = { antenna height in m, and 𝐶(ℎ𝑅𝑥 ) is the Rx
85.9 − 20 log10 (ℎ𝑅𝑥 ) 𝑓𝑜𝑟 ℎ𝑅𝑥 ≥ 10𝑚 antenna height correction
1.33𝑓 0.284 𝑑 0.588 𝑓𝑜𝑟 14𝑚 < 𝑑 ≤ 400𝑚

W MED
[5], [9],
𝑃𝐿 ={
[139]
{ 𝑊_𝑀𝐸𝐷 0.45𝑓 0.284 𝑑 𝑓𝑜𝑟 0𝑚 < 𝑑 ≤ 14𝑚 (22)

230𝑀𝐻𝑧 ≤ 𝑓 ≤ 95𝐺𝐻𝑧
Empirical
𝑓 is the frequency in MHz (except for W

ITU-R
[5], [9],
Foliage
𝑃𝐿𝐼𝑇𝑈−𝑅 = 0.2𝑓 0.3 𝑑 0.6

[139]
{ (23) MED, frequency is given in GHz),

200𝑀𝐻𝑧 ≤ 𝑓 ≤ 95𝐺𝐻𝑧, 𝑑 ≤ 400𝑚
𝑑 is the tree depth in m
0.37𝑓 0.18 𝑑 0.59 𝑓𝑜𝑟 𝑜𝑢𝑡 𝑜𝑓 𝑙𝑒𝑎𝑓
FITU-R
[5], [9],
𝑃𝐿𝐹𝐼𝑇𝑈−𝑅 = {
[139]
{ 0.39𝑓 0.39 𝑑0.25 𝑓𝑜𝑟 𝑖𝑛 𝑙𝑒𝑎𝑓 (24)

10𝐺𝐻𝑧 ≤ 𝑓 ≤ 40𝐺𝐻𝑧, 𝑑 ≤ 400𝑚
VOLUME XX, 2021 15
TABLE I (CONTINUED)
26.6𝑓 −0.2 𝑑0.5 𝑓𝑜𝑟 𝑜𝑢𝑡 𝑜𝑓 𝑙𝑒𝑎𝑓

235 [5],
COST-
[139] 𝑃𝐿 ={
[9],
{ 𝐶235 15.6𝑓 −0.009 𝑑 0.26 𝑓𝑜𝑟 𝑖𝑛 𝑙𝑒𝑎𝑓 (25)

Empirical
9.6𝐺𝐻𝑧 ≤ 𝑓 ≤ 57.6𝐺𝐻𝑧, 𝑑 ≤ 200𝑚 𝑓 is the frequency in MHz,

(0.001𝑓 + 0.2)𝑑 + 0.5𝑓 + 3 𝑣𝑒𝑟𝑡𝑖𝑐𝑎𝑙 𝑝𝑜𝑙𝑎𝑟𝑖𝑧𝑎𝑡𝑖𝑜𝑛 𝑑 is the tree depth in m
& Kue
Chen
𝑃𝐿𝐶−𝐾 ={
[9]
{ (0.002𝑓 + 0.2)𝑑 + 0.03𝑓 + 2 ℎ𝑜𝑟𝑖𝑧𝑎𝑛𝑡𝑎𝑙 𝑝𝑜𝑙𝑎𝑟𝑖𝑧𝑎𝑡𝑖𝑜𝑛 (26)

1𝐺𝐻𝑧 ≤ 𝑓 ≤ 100𝐺𝐻𝑧
𝜉𝑑
−( )
𝑃𝐿𝑀𝐴 = 𝐴𝑚 (1 − 𝑒 𝐴𝑚 )
𝐴𝑚 = 𝐴1 𝑓 ∝1 𝐴𝑚 is the maximum attenuation in dB for a
[54], [139], [140]
30𝑀𝐻𝑧 ≤ 𝑓 ≤ 100𝐺𝐻𝑧 specific type of vegetation,

Foliage
𝜉 is the attenuation for a very short path via a

𝐴1 = 0.18, ∝1 = 0.752 𝑓𝑜𝑟 𝑡𝑟𝑜𝑝𝑖𝑐𝑎𝑙 𝑡𝑟𝑒𝑒𝑠 𝑎𝑡 900 − 1800𝑀𝐻𝑧
MA
(27) specific type of vegetative in dB/m,

@ℎ𝑅𝑥 = 2.4𝑚 𝑤𝑖𝑡ℎ 15𝑚 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑡𝑟𝑒𝑒 ℎ𝑒𝑖𝑔ℎ𝑡
𝑑 is tree depth in m (propagation distance
Semi-Empirical
𝐴1 = 1.15, ∝1 = 0.43 𝑓𝑜𝑟 𝑚𝑖𝑥𝑒𝑑 𝑓𝑜𝑟𝑒𝑠𝑡 𝑎𝑡 900 − 2200𝑀𝐻𝑧 through foliage),

@ℎ𝑅𝑥 = 1.6𝑚 𝑎𝑛𝑑 ℎ𝑇𝑥 = 25𝑚 𝑤𝑖𝑡ℎ 15𝑚 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑡𝑟𝑒𝑒 ℎ𝑒𝑖𝑔ℎ𝑡 𝑅0 is the initial gradient of the attenuation
𝐴1 = 1.37, ∝1 = 0.42 𝑓𝑜𝑟 𝑚𝑖𝑥𝑒𝑑 𝑓𝑜𝑟𝑒𝑠𝑡 𝑎𝑡 105.9 − 2117.5𝑀𝐻𝑧 curve,
{ @ℎ𝑅𝑥 = 1.5𝑚 𝑤𝑖𝑡ℎ 14𝑚 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑡𝑟𝑒𝑒 ℎ𝑒𝑖𝑔ℎ𝑡 𝑅∞ is final attenuation in dB/m,
𝑘 is final attenuation offset in dB,
[139]
NZG
[54],
𝑅0 −𝑅∞
𝑃𝐿𝑁𝑍𝐺 = 𝑅∞ 𝑑 + 𝑘(1 − 𝑒 −( 𝑘
)
) (28) 𝑓 is the frequency in GHz,
𝑤 is the maximum effective width between
𝑐𝑑
Tx-Rx,
𝑅∞ 𝑘 (𝑅 −𝑅∞ )𝑊
[139]
−( 0 𝑘 ) 𝑎, 𝑏, and 𝑐 are estimated constants

DG
𝑃𝐿𝐷𝐺 = 𝑎 𝑏
𝑑 + 𝑐 (1 − 𝑒 ) (29)
𝑓 𝑤 𝑤
𝐴2 𝑒 −(∝2 𝑑) 𝐵2
{𝑃𝐿𝑇𝑒𝑤𝑎𝑟𝑖 = −27.56 + 20 log10(𝑓) − 20 log10 ( 𝑑
+ 2)
𝑑 𝑓 is the frequency in MHz,
Tewari [54]
50𝑀𝐻𝑧 ≤ 𝑓 ≤ 800𝑀𝐻𝑧, 𝑑 ≤ 4000𝑚 𝑑 in m,

Foliage/Terrain
𝑓 [𝑀𝐻𝑧] ∝2 [Horizontal (H), Vertical (V)] 𝐴2 [H, V] 𝐵2 [H, V] (30) ∝2 is attenuation rate in dB/m,
Empirical
50 -, - 0, 0 7.367, 1.917
𝐴2 and 𝐵2 are empirically found constants for
200 0.011, 0.0125 0.8201, 0.4989 5.045, 1.8358
500 0.0138, 0.0135 0.6571, 0.3658 1.4304, 0.904
different frequencies and polarization,
800 0.0152, 0.014 0.4491, 0.2661 0.6291, 0.5331 𝐺(ℎ𝑇𝑥 , ℎ𝑅𝑥 ) is the antenna height gain factor
for both Tx and Rx,
𝑃𝐿𝐸𝑥𝑡_𝑇𝑒𝑤𝑎𝑟𝑖 = 𝑃𝐿 𝑇𝑒𝑤𝑎𝑟𝑖 + 𝐺(ℎ𝑇𝑥 , ℎ𝑅𝑥 )
Extended
ℎ𝑇𝑥 and ℎ𝑅𝑥 are Tx and Rx antenna heights

Tewari
[54]
{𝐺(ℎ𝑇𝑥 , ℎ𝑅𝑥 ) = 14 + 4 log10 (𝑓) − 20 log10(ℎ𝑇𝑥 ℎ𝑅𝑥 ) (31) above terrain

50𝑀𝐻𝑧 ≤ 𝑓 ≤ 800𝑀𝐻𝑧, 𝑑 ≤ 4000𝑚
𝐿𝑐 is a measurement-based constant loss,
Cost231-MWF [142]
usually found using multi-linear regression

method;
𝐼 𝐾𝑓 +2
(𝐾 +1−𝑏) 𝐾𝑤𝑖 and 𝐿𝑤𝑖 are the number and the loss of
𝑓
𝑃𝐿𝐶231_𝑀𝑊𝐹 = 𝑃𝐿𝐹𝑆 (𝑓, 𝑑) + 𝐿𝑐 + ∑ 𝐾𝑤𝑖 𝐿𝑤𝑖 + 𝐾𝑓 𝐿𝑓 (32) traversed walls of type 𝑖, respectively;
𝑖=1 𝐾𝑓 and 𝐿𝑓 are number and loss of penetrated
floors, respectively;
𝑏 is an empirical accuracy adjustment factor,
𝐼 is the number of wall types
𝑃𝐿(𝑑0 ) is the basic transmission loss at the
reference distance 𝑑0 ,
(P.1238-10) [141]
𝑁 is a distance dependent power loss

ITU-R Indoor
𝑃𝐿𝑃123810 (𝑑) = 𝑃𝐿(𝑑0 ) + 𝑁 log10 (𝑑/𝑑0 ) + 𝐿𝑓 (𝑛) coefficient given in [141],

𝑃𝐿(𝑑0 ) = 20 log10 (𝑓) − 28, 𝑓𝑜𝑟 𝑑0 = 1𝑚 𝑑 is the Tx-Rx separation distance in m,
Empirical
(33)
Indoor
𝐿𝑓 = 0𝑑𝐵, 𝑓𝑜𝑟 𝑛 = 0 𝑑0 is the reference distance in m,

{ 300𝑀𝐻𝑧 ≤ 𝑓 ≤ 450𝐺𝐻𝑧, 𝑑 > 1𝑚, 𝑛 ≥ 0 𝐿𝑓 is the floor penetration factor in dB, given
in [141],
𝑛 is the number of floors between Tx and Rx,
𝑓 is the frequency in MHz
𝑑0 is the reference distance,
MWMF [142], [143]
𝐼 𝐾𝑤𝑖 𝑛 is the PLE,

𝐿𝑤𝑖𝑘 is the loss in dB of the kth traversed wall of
𝑃𝐿𝑀𝑊𝑀𝐹 (𝑑) = 𝑃𝐿(𝑑0 ) + 10𝑛 log10(𝑑/𝑑0 ) + ∑ ∑ 𝐿𝑤𝑖𝑘 type 𝑖,
𝑖=1 𝑘=1
𝐽 𝐾𝑓𝑗 (34) 𝐾𝑤𝑖 is the number of traversed walls of type 𝑖,
𝐼 is the number of wall types,
+ ∑ ∑ 𝐿𝑓𝑗𝑘 , 𝑑0 = 1m 𝐿𝑓𝑗𝑘 is the loss in dB of the kth traversed floor of
𝑗=1 𝑘=1
type 𝑗,
𝐾𝑓𝑗 is the number of traversed floors of type 𝑗,
𝐽 is the number of floor types
16 VOLUME XX, 2021
In contrast, [125], [134], [160]–[162] focused on the characteristics. Rain attenuation of up to 15 dB was
evaluation of outdoor channel models. For instance, [134] measured for a 300 m path at a rain rate of roughly 125
demonstrated the large-scale propagation characteristics of mm/h, indicating that 0.001% and 0.011% of outages must
5G in an outdoor parking lot scenario with several end users. be considered for 38 GHz. Moreover, the received power
As such, a new CI-based PL model was proposed, including dropped by 33.1 dB for NLOS scenarios. When considering
a parking lot compensation factor. The PL was then rain attenuation, it was also found that received power
evaluated using several models and observations at 28 and dropped by 6.4 dB for Malaysia compared to a temperate
38 GHz for various scenarios. The authors of [160] discussed area communication link due to increased PL at 38 GHz.
measurement campaigns in a tropical climate for an outdoor Thus, the study implies that further research is required to
5G network operating at 32 GHz, considering the distance, model 5G channels in tropical environments.
polarization, and antenna type. For the LOS scenario, highly Machine learning techniques are capable of mining high-
directional horn antennas with Co- and Cross-polarizations dimensional data and extracting the required information to
were used for Tx and Rx. learn the structural relationship between data in complex
Meanwhile, horn and omnidirectional antennas were environments. As a result, ML may be a powerful tool for
utilized on the receiver side for NLOS. Finally, the CI and FI extracting radio wave characteristics and developing channel
PL models were evaluated based on the outdoor data. The models from measurement data. However, integrating
results in [134] indicated that the PLE values were almost channel modeling and ML is challenging, and research in this
similar. Meanwhile, the compensation factor suggested for subject is still in the exploration stage. On the other hand,
28 GHz and 38 GHz was 10.6-23.1 dB and 13.1-19.1 dB, artificial neural network (ANN) is a significant branch of ML
respectively. They did, however, show that additional that benefits from adaptability, self-learning, and robustness
compensation factors are required for more scattered objects, and may be used to predict PL. Recently, [163], [164]
particularly at 28 GHz. According to the findings in [160], introduced PL models based on ANNs and Levenberg-
the PLE for LOS and NLOS scenarios ranged between 3.4 Marquardt backpropagation (BP), respectively. The results
and 6.7. Additionally, the FI PL model was unsuitable for the indicate that when compared to conventional PL models, the
NLOS scenario, shown by the large divergence between the proposed methods can significantly enhance the accuracy of
slope lines of the horn-horn and horn-omnidirectional cases. PL prediction. However, the methods’ input set consists of
Likewise, the results indicated that co-polarization decayed signal parameters that describe the channel’s properties, with
faster than cross-polarization in LOS scenarios. the impact of the environment being ignored.
Similarly, [125], [161] evaluated the potential capability Providing geometrical information is a complicated
of mmWave PL models using measurements taken in procedure. The environment type can be valuable in this
outdoor tropical environments. While both studies evaluated sense, as it can provide useful information for describing the
the CI and FI models at 38 GHz, [161] included the CIF and environment [165]. In [166], a PL prediction model based on
ABG models and extended the evaluation to include 20-38 the multilayer perceptron (MLP) neural network (NN) and
GHz frequency bands. The authors in [161] then proposed a environment types is established, with BS and Rx, digital
new PL model for 20 GHz and 30 GHz frequency bands. images, and satellite maps as inputs. To simplify 3D
Meanwhile, [125] examined network performance in LOS environment modeling, principal component analysis (PCA)
and NLOS scenarios with co- (vertical-vertical) and cross- was utilized to extract low-dimensional environmental
polarization (vertical-horizontal). The results in [161] features from the limited environmental types. An ANN
showed that the proposed model was both reliable and dataset was constructed to train and evaluate the PL
straightforward in terms of frequency and environment prediction model by combining measured information from
signal attenuation, suggesting its usefulness for establishing the BS and receiver, including 3D locations, frequency,
suitable communication links for the scenarios studied. [125] Tx/Rx power, antenna data, feeder loss, and other
showed that the CI PL model predicted significant NLOS environmental features. The measurement campaigns were
outcomes and better network performance in the LOS case. done at 2.5 GHz in Hangzhou, China, covering 20 different
Also, the results showed that the FI PL model was unsuitable environmental types like suburban areas, urban areas, high-
for the NLOS scenario, particularly for V-V polarization. raised buildings, irregular buildings, green land, wetland,
Additionally, the simulation results indicated that user and forest. Three BSs with heights of 30 m, 42 m, and 62 m
throughputs decreased faster with cross-polarization than were considered, with the Rx placed on a vehicle with a
with co-polarization in both LOS and NLOS scenarios. height of 2 m. The CI and A-B models were used to fit the
Finally, in [162], the authors presented a brief simulation measurement data for PL prediction and then compared to
analysis of rain fading using simultaneous one-minute rain predicted results in terms of the absolute value of mean error
rate measurements and its effects on a short 38 GHz (AME), mean absolute error (MAE), STD error, and
experimental link. The PDP was also generated for correlation coefficient (R). The comparison showed that the
omnidirectional and directional scenarios by observing the proposed model could achieve a higher prediction accuracy.
received power and PL using Malaysia’s environmental However, the accuracy of ANN-based prediction methods
VOLUME XX, 2021 17
can be improved at the expense of the ANN architecture’s used to characterize the PL, with model parameters
complexity. This study also examined the impact of ANN statistically characterized from known distributions. The
architectures, dimension, and training sample percentage on shadowing effect was further studied, and a statistically
PL prediction models, finding that 60% of training samples extended model for inter-cell shadowing distribution and
from training sets is sufficient for a high-efficiency and distance correlation was provided. The proposed PL model
stable PL prediction model. was verified across NB-IoT and LTE operators, and the
Convolutional neural networks (CNN) is a well-known findings indicated that it outperforms state-of-the-art NB-
deep learning (DL) model that may be used to solve a wide IoT PL models in terms of estimation accuracy.
variety of classification and regression problems with little To develop a suitable channel model for D2D IoT
pre-processing and feature extraction. [167] proposed a communication technologies, the authors of [169] presented
method for 28 GHz mmWave PL modeling based on CNNs two log-distance-based PL models (frequency-independent
for the suburban scenario. The measurement campaign PLE and frequency-dependent PLE), as well as a new
considered 13 scenarios using a fixed Tx at 3 m and an Rx statistical distribution of the DS for IoT communications,
mounted on the rooftop of a moving car. Four directional based on quasi-simultaneous wideband channel measurements
antennas were utilized on the receiver side to investigate the in the VHF/UHF frequency bands (37.8–370 MHz) in Halifax,
mmWave system’s characteristics in various environmental Canada. The results indicated that both models achieved
scenarios. The method was built assuming that the CNN could similar estimation results, with PLE values ranging from 4.13
generate relevant environmental features from map imagery. to 4.8 and shadow fading standard deviation values ranging
Thus, two-dimensional (2-D) Google map images were used from 8.87 to 10.96 dB. As a result, they emphasize that
to extract geometrical information about buildings and street existing VHF/UHF propagation models are unsuitable for IoT
objects. The Enhanced local area multi-scanning (ELAMS) communications with low Tx and Rx antenna heights.
algorithm was proposed to build a training set for the CNN. By contrast, the authors of [54], [170] concentrated on
The algorithm extracts environmental data between Tx and Rx NB-IoT connectivity in a rural forest with low antenna
from pre-processed map images. A CNN was constructed with heights. Both measurement campaigns were done at a range
four subnetworks and 20 neurons, with each subnetwork of 2.5 km, utilizing LTE band 8 at a frequency of 917.5 MHz,
learning the same propagation and environmental information with a special focus on PL and coverage evaluation in a D2D
from the ELAMS images. A feature sharing layer was communication scenario. [170] noted that previous PL
introduced between convolutional layers to concatenate the studies for 900 MHz near-ground scenarios lag of the 164 dB
activation map of the previous layer as an input to the next PL specified in the NB-IoT standard. As a result,
layer. It was claimed that the addition of feature-sharing layers measurements were made with a custom-built 180 dB
did not increase the model’s complexity but rather aided in the dynamic measurement instrument. According to the
backpropagation of loss from each antenna to the entire measurements, a D2D system with Tx and Rx antennas at 1.5
network. The comparison results showed that the proposed m height could achieve a range of around 2 km when
CNN model outperforms both the CI and ABG models in employing the NB-IoT 164 dB PL limit. Meanwhile, in [54],
accuracy and complexity, with an RMSE of 8.59 dB for PL the authors evaluated the effects of various antenna heights
prediction in the test scenarios. by utilizing similar measurement equipment and placing
Although prior research has demonstrated potential antennas 1.5 m, 2.5 m, and 3.5 m above the ground.
channel characterizations for some 5G bands under various Additionally, measured data were compared to known
setup scenarios, this technology is still in its initial foliage excess loss models and related PL models to determine
deployment in many countries. As a result, most of these which models are suitable for NB-IoT D2D communication.
studies are primarily experimental or simulation-based, The results indicated that the antenna height had no significant
focusing on high-power devices and high-gain antennas, and influence on the received power in the given setting.
do not consider the requirements of IoT-based applications. Additionally, the results indicated that the dominant
Thus, future research and characterization must incorporate propagation path for the first km was through the foliage,
IoT parameters into the modeling process and optimize the resulting in foliage excess loss being the dominant loss factor
performance of such applications. in this area. Thus, the measured received power obeys the
Other studies have evaluated the channel characterization fourth-power law after the first km, implying that the dominant
of traditionally utilized cellular-based IoT technologies, loss factor is the distance-dependent PL. Additionally, the
including narrowband-IoT (NB-IoT) and cellular-based D2D comparison revealed that only two models accurately
communication. In [168], the authors proposed an empirical anticipated the level of foliage excess loss. Thus, they stated
PL model for NB-IoT in urban areas using a large-scale that by combining these excess loss models with PL models,
measurement campaign conducted in Oslo, Norway. Three the estimated total PL might be modeled with an RMSE of less
datasets in the LTE band from two cellular operators were than 10 dB. This was achieved using either the Tewari model
included, considering three different scenarios: outdoor, or combining a 2-Ray PL model with an ITU-R P.2108 clutter
indoor, and deep indoor. The ABG and CI PL models were
18 VOLUME XX, 2021
loss model. However, for all models compared, foliage excess In comparison, the authors of [173] proposed an empirical
losses were modeled only from 200 m onwards. outdoor to deep-indoor PL model for NB-IoT at sub-GHz
In [171], [172], the authors followed DL or ML approaches frequencies. The deep-indoor environment considered is an
to model the PL. Accordingly, [171] incorporated tabular data underground tunnel, where distance-related (Tx-Rx
and images as inputs for CNN to perform PL prediction in distance) and tunnel-related features (such as the closest
urban areas. Hence, the vectors of tabular data were first corridors, the distance and angle to the farthest tunnel corner,
manipulated and transformed into images, and then each and the distances to the tunnel walls and ceiling) were
feature was spread across several pixels, proportional to its characterized to develop the empirical PL model. It was
calculated importance. Then, pseudo images were created by determined that the 2-D indoor distance and the distance to
mixing synthetic images (tabular data transformed) with the tunnel walls are the most relevant parameters for RSRP
images depicting selected regions of the area’s map. The prediction. A linear and a Gaussian process model were also
pseudo images were then used as inputs for a CNN that constructed for the indoor PL prediction. The derived models
predicts the PL value at a certain point within the considered outperformed the 3GPP TR 38.901 model by 1.8 and 4.1 dB.
area. The results indicated that the presented approach In conclusion, few studies have explored the propagation
outperforms models using a single input mode. modeling of such cellular-based IoT technologies. These
In contrast, [172] proposed combining two ML studies were mostly focused on specific areas, utilizing
algorithms, ANN and Random Forests, with three types of experimental hardware equipped with limited
input data to estimate PL for NB-IoT operating at 900 and omnidirectional Tx and Rx antennas. As a result, additional
1800 MHz. Both approaches were trained and validated study is needed to determine the applicability of various
using the same sets of area architectural attributes to allow existing channel models, propose new models that consider
comparison. As a result, they concluded that data inputs are different scenarios and commercially available devices, and
critical for predicting PL using ML approaches. assess the limitations of such cellular-based IoT
Additionally, it was determined that when the transmitter is technologies. To summarize the studies evaluated in this
mounted above the building rooftops, LOS data is more section, Table II includes a detailed description of each
significant than site-specific information, while combining study’s modeling approaches, analysis metrics, key findings,
both types of data results in even higher performance. and limitations.
Additionally, for the 900 MHz scenario, PL predictions were
more precise for all input types and ML approaches. B. Short-range based wireless IoT channel modeling
Finally, developing a PL model for the deep-indoor and characterization
scenario is challenging, and existing PL models are Numerous studies have attempted to provide practical lower
inaccurate in such cases. Studies in such scenarios may also bounds on the accuracy of PL model prediction due to the
be motivated because many critical IoT scenarios, such as high need for baseline performance values. Additionally, a
telemedicine or the monitoring of important assets, require a well-established error bound for more sophisticated PL
robust long-range communication network to perform well. modeling and coverage mapping approaches is required to
Guarantees of QoS and high availability are also necessary verify their accuracy. As a result, the authors in [123]
for such harsh radio environments, where tracked objects or described and implemented 30 different propagation models,
patients may move to difficult-to-locate locations, such as a considering data from networks operating at 2.4 GHz, 5.8
basement. As a result, developing accurate channel models GHz, and 900 MHz in rural and urban areas. They concluded
for these scenarios is a major process toward crucial IoT. In from the results that the PL model environment is uncertain.
this context, the authors of [128] conducted an experimental As a result, these models’ typical best-case performance
measurement campaign to determine the attenuation of NB- accuracy is in the range of 12–15 dB RMSE, and it can be
IoT deep indoor signals at known points. substantially worse in practice. However, adjustable models
Additionally, they showed how deep indoor scenarios and specific data fitting approaches may reduce the RMSE
impacted RSSI and evaluated the effect of indoor distance to to 8-9 dB. These constraints on modeling error remain
the outermost wall on various indoor scenarios. The findings relatively constant over various environments and frequency
indicated that indoor signal transmission varies significantly ranges.
between underground and above-ground scenarios. The As such, the author suggested using a few well-accepted
findings indicated that the 3GPP TR 38.901 indoor PL model and well-performing models, such as Okumura-Hata or Hata-
accurately predicted the above-ground indoor scenario but Davidson, in scenarios requiring priority predictions, and if
mispredicted the underground tunnel measurements. The possible, well-validated, measurement-driven methods. They
latter implied that theoretical models are inapplicable to all emphasize, however, that the most critical concerns for a
indoor scenarios, as the overall link budget is far too researcher are having a reasonable expectation of error and
complicated to be fully characterized by linear dependence. picking a model that allows replication and comparison of
They emphasize, however, that additional experiments are findings.
needed to verify this concept.
VOLUME XX, 2021 19
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3151967, IEEE Access

TABLE II
A Summary of the Reviewed Cellular-based Wireless IoT Channel Modeling and Characterization Studies.
Environment/ Targeted Modeling approach & Comparison
Ref F. RAT Analysis Metrics Key Findings Limitations
Scenario application used input parameters Models
Hybrid probabilistic PL Showed that the PL might be modeled
Single frequency, N/A
Indoor 4.5, 28, General 5G model based on CI. Use more accurately utilizing the proposed Lmited to indoor. Do not consider
[145] 5G CI & FI, and multi- Compare in terms
LOS/NLOS 38 GHz applications spectrum analyzer power model with a single parameter (PLE) than other indoor models for comparison.
frequency, ABG. of best fit to PL
measurements. well-known models.
CI & FI parameters Use MATLAB for Showed that CI & FI are most suitable Lmited to indoor. Do not consider
Indoor General 5G FSPL (baseline),
[147] 40 GHz 5G calculation. Use spectrum analysis. Metrics for indoor PL modeling. FI provides the other indoor models for comparison.
LOS/NLOS applications CI, FI, & 2-Ray.
analyzer power. are not listed. best minimum error fit. Do not consider multiple floors.
The results indicated that the 5G channel
CI, FI, CIF, & ABG RMS-DS, excess
has excellent PL performance and a very Lmited to indoor. Do not consider
Indoor 3.5, 28 5G-based model. Use PDP & Rx Estimated vs.
delay & PDP.
[132] 5G small DS, enabling future real-world other indoor PL models for
LOS/NLOS GHz smart city power from a spectrum measured data.
Diffraction loss &
deployments of 5G-based smart city IoT comparison.
analyzer. frequency drop.
networks.
6.5, 10.5, Lmited to indoor. Different
CI, CIX, FI, FA, ABG, & Modeled vs. PL models for single/multi-frequency,
Indoor 15, 19, General 5G indoor/outdoor & time-varying
[148] 5G ABGX. Use PDP from a measured data & RMS-DS & STD. RMS-DS, & excess delay statistics were
LOS/NLOS 28, 38 applications dynamic environments need to be
spectrum analyzer. previous studies. presented and modeled.
GHz investigated.
CI model with a parking Mean variation in Showed that more compensation factors Further investigation is required,
Outdoor 28, 38 General 5G
[134] 5G lot compensation factor CI, FI, & ABG.
compensation are required for more scattering objects, considering indoor/ underground/
Parking lot GHz applications
added. factor. especially at 28 GHz. proximity parking environments.
Investigate the impacts of PL concerning No statistical analysis. Limited to
Outdoor General 5G CI & FI. Use power from Measurements vs.
[160] 32 GHz 5G N/A distance, polarization, and antenna type. indoor & consider only two
LOS/NLOS applications a spectrum analyzer. CI & FI.
FI is shown to be unsuitable for NLOS. models.
Hybrid probabilistic PL
CI, FI, CIF, ABG Further investigation is required
Outdoor 26, 28, model based on CI, FI,
General 5G models & modeled Showed that the proposed model is to incorporate tropical terrain,
[161] Tropical urban 36, 38 5G CIF, & ABG. Use Rx STD
applications parameters from acceptable for considered scenarios foliage, and weather impact,
LOS/NLOS GHz power from a handheld
previous studies especially rain and fog.
spectrum analyzer.
CI & FI were evaluated under several Simulation-based. Limited for
scenarios (LOS V-V, V-H, NLOS V-V). outdoor with no comparisons. To
Outdoor General IoT
[125] 38 GHz 5G CI & FI. N/A N/A CI predicted significant NLOS outcomes study the influence of various
LOS/NLOS applications
& better network performance in the LOS beam-forming effects & analyze
case. FI is unsuitable for NLOS (V-V). their outcome.
Outdoor The decrease in Rx
Rain Attenuation. Use Showed that further research is needed to Simulation-based with limited
[162] Urban 38 GHz 5G N/A N/A power for different
PDP from simulation. model 5G channels in tropical areas. analysis.
LOS/NLOS scenarios.
Outdoor Environmental features cannot
CI & A-B [using PRx]. The impact of environmental features,
Urban, suburban, Measured PL vs. AME, MAE, STD, accurately reflect propagation
ANN [use environmental ANN architectures, dimension, & training
[166] forest, green- 2.5 GHz 4G General CI, A-B, & ANN R, & training characteristics without height
features combined with sample percentage on PL prediction
/wet-land. model. efficiency. information, affecting the PL
Tx/Rx information]. models was investigated.
LOS/NLOS model prediction accuracy.
CNN. Extract features A feature sharing layer between
Outdoor The PL model was developed using
from image data sets. convolutional layers was added to benefit
[167] Suburban 28 GHz N/A General CI & ABG. RMSE a 2-D street map with no regard
Empirical PL is backpropagation without increasing
LOS/NLOS for object heights or vegetation.
calculated from RSS. complexity.
20 VOLUME XX, 2021

TABLE II (CONTINUED)
A SUMMARY OF THE REVIEWED CELLULAR-BASED WIRELESS IOT CHANNEL MODELING AND CHARACTERIZATION STUDIES.
Environment/ Targeted Modeling approach & Comparison
Ref F. RAT Analysis Metrics Key Findings Limitations
Scenario application used input parameters Models
Outdoor Okumura-Hata, RMSE, empirical They claim their models can simulate NB-
Site-specific, i.e., PL model is
Urban 800-900 General IoT ABG & CI models. Use 3GPP TR 45.820, PDF, & statistical IoT PL in multi-cell urban deployments
[168] NB-IoT limited to Oslo-like urban
LOS/NLOS & MHz applications. RSRP. & UMa. FSPL distributions like like Oslo, improving PL estimation
environments.
O2I. (baseline). GEV & Weibull. accuracy.
PL / DS models based on PL, DS, RMS-DS, Points out that existing VHF/UHF
Outdoor Other models are not considered
37.8 - 370 General D2D log-distance PL model & Statistical analysis mean excess delay, propagation models are unsuitable for
[169] Urban city N/A to compare the accuracy of
MHz IoT DS statistical distribution. & best fit. & cross-correlation IoT communications with low Tx & Rx
LOS/NLOS developed models
Use Rx power and APDP. coefficient. antenna heights.
Outdoor Sports (forest No PL modeling. Calculate Measured PL vs. Showed that a D2D system with 1.5 m Limited to max distance & PL
917.5
[170] Rural forest NB-IoT hiking tracking PL from spectrum analyzer FSPL & 2-Ray N/A Tx/Rx antennas could reach 2 km using evaluation. Further research is
MHz.
Terrain & monitoring). power readings. models the NB-IoT 164 dB PL limit. needed using physical NB-IoT EDs.
2-Ray, Tawari, The antenna height had no significant effect
FSPL. Above 1km on the Rx power. The dominant propagation
Comparison only & (reasonable loss path for the 1st km was through the foliage,
Outdoor Sports (forest All models compared only
917.5 combine PL and foliage prediction): ITU-R resulting in the dominance of foliage excess
[54] Rural forest/ NB-IoT hiking tracking RMSE modeled foliage excess losses
MHz excess loss models, P833-9, P2108-0, loss. This suggests a 4th-power law for
terrain. & monitoring). from/above 200 m.
Tewari model. FITU-R. Combined measured Rx power beyond 1 km. Two
PL between 2-Ray models had RMSE less than 10dB: Tewari
& ITU-R P.1546. or 2-Ray combined with ITU-R P.2108.
DL approach. Use a mix
Proposed compound pseudo images as a
of tabular & images data CNN-based Simulation-based. There are no
Outdoor General IoT fusion of transformed tabular & images
[171] 900 MHz N/A as two diverse inputs models with MAE comparisons with conventional
Urban applications (area map). Outperformed DL models
types for DL CNN to single-mode input PL models. No statistical analysis.
with single-mode inputs.
predict PL.
Both ML models performed very similarly.
No models are considered for
MAE, RMSE, & Conclusions: 1) data inputs are critical in
900 & ANN & Random Forests comparison. No experimental data
Outdoor mean absolute predicting PL using ML. 2) LOS data is
[172] 1800 NB-IoT General (RF) based PL models N/A (ray-tracing simulations only).
Urban area percentage error more crucial than site-specific data when the
MHz using three input types. Limited description of simulation
(MAPE) Tx is located on a building roof, although
parameters.
integrating both improves performance.
Showed the impact of deep indoor on
Limited to RSSI measurement
RSSI & analyzed the impact of indoor
with no calibration. Considers one
Deep Indoor Remote Modeled based on a distance to the outermost wall. 3GPP TR
820.5 Power heatmap & model for comparison, and there
[128] Indoor to NB-IoT metering or simple I2O model. Use 3GPP TR 39.901 38.901 well-matched the above-ground
MHz RSSI vs. distance. is no statistical analysis. Do not
Outdoor (I2O) asset tracking RSSI. indoor scenario but not the underground
provide Tx power, antenna specs,
tunnel. Thus, theoretical models may be
and height of Tx & Rx.
unsuitable for all indoor scenarios.
Need more data in diverse
Linear regression & Showed that the 3GPP distance-based
scenarios to improve model
Gaussian process (GPR). model fits fairly well NB-IoT
Underground 3GPP TR 38.901 accuracy, including more
[173] Deep-indoor 868 MHz NB-IoT Use RSRP measurements RMSE & MAE. measurement above the ground while
scenarios [UMa]. environment features, obstacle
with a set of distance & performs poorly in underground deep-
distribution, & detailed tunnel
tunnel-related features. indoor scenarios.
geometry.
VOLUME XX, 2021 21
As previously noted, most of the PL models currently used in channel transmission environment deteriorated concerning the
WSN were originally developed to provide signal prediction in developmental stage of the rice field. As a result, the optimal
conventional high-power wireless systems, such as satellite and node antenna height was found to be 2 m above ground. While
personal communication systems, which differ significantly FSPL was shown to be inappropriate, the 2-Ray model may be
from WSN environments. As a result, the authors in [30], [119] used if the antenna height exceeds 1.2 m, as height
used RF measurements to characterize the propagation behavior significantly impacts the 2-Ray model prediction. On the other
of WSNs in outdoor scenarios and subsequently developed an hand, the one-slope log-distance model performed better,
LNSPL-based model. The focus of [30] was on WSN deployed although its estimated relative error (RE) was greater than 3%.
on concrete surfaces, whereas the focus of [119] was on PL Finally, the modified two-slope log-distance model
modeling of WSN nodes scattered across a sand terrain outperformed the one-slope log-distance model in all heights,
environment. Both studies compared observed and predicted PL with an estimated RE of less than 2%, and was more
values to those obtained from FSPL, 2-Ray, and models applicable to the complicated rice field environment. As a
developed for WSNs deployed in long-grass and sparse-tree result, the authors claim that the latter will aid in developing
environments. [30] showed a significant difference in the PL efficient rice field WSNs while also improving the quality of
and model parameters between the proposed model, previous wireless transmission.
studies, FSPL, and 2-Ray models. Additionally, the MAPE was The results in [139] indicated that the most important
36% and 22% for the FSPL and 2-Ray models, confirming their vegetation effects occur at 1.5 m tree height. Likewise, the
inaccuracy in predicting PL in concrete surface environments. results proved that PL prediction using FSPL and 2-Ray models
Similar outcomes were reported in [119], indicating that the is inaccurate in certain environments due to their simplistic and
compared models were inaccurate in predicting PL in sand optimistic nature. Additionally, results indicated that the
terrain. However, the proposed model does not consider the combined COST-235 and FSPL models provided the best
electrical ground properties of sandy terrain or the placement of results compared to other foliage models. However, the latter
Tx/Rx at various heights. model is not ideal, as the MAPE was 10.69%. In contrast, the
In contrast, [62], [71], [139] analyzed the influence of proposed model MAPE has a MAPE of 2.75 %, suggesting that
foliage attenuation on wireless signal and the applicability of it is the most efficient model for representing greenhouse
existing PL models in a variety of scenarios. The authors of vegetation loss. Finally, the authors noted in [62] that a proper
[71] studied signal attenuation in a rice field environment at propagation model, such as those proposed, that enables
node antenna heights of 0.8 m, 1.2 m, 1.6 m, and 2.0 m during network performance evaluation enables QoS optimization and
the tillering, joining, and grain filling stages of rice fields. can be used as part of a QoS guarantee management platform
Also, they evaluated PL fluctuation over a range of distances for native Irish trees and dense woodland applications.
and compared the measured PL to the predicted PL using The authors in [104] proposed a channel model for rural
FSPL and a 2-Ray model. Then, a one-slope log-distance smart agricultural WSNs operating in near-ground conditions
model was established using regression analysis, and a in the soil, short grass, and tall grass fields. The measurements
modified two-slope log-distance model was proposed. considered Tx and Rx antennas mounted at 0.2 and 0.4 m
The authors of [139] analyzed and developed a new linear above ground and utilized three frequency bands, 868 MHz,
regression model for predicting PL in greenhouse 2.4 GHz, and 5.8 GHz. The PL was then calculated and
environments. A combination of foliage effects, reflections adapted using a three-slope log-normal PL model. Based on
(from the ground or tree canopy), diffraction, and traveling RSSI measurements, the second experiment used commercial
wave scattering were used to model propagation loss. ZigBee nodes at 2.4 GHz to predict link performance.
Numerous empirical measurements were done at 2.425 GHz Accordingly, two sensor nodes were mounted at identical
utilizing an IEEE 802.15.4-based WSN to determine the heights to those in the previous experiment, but only in a short
influence of growing tree components on PL at various Tx/Rx grass field scenario. The QoS efficiency was estimated using
heights. Finally, [62] analyzed experimental data from a real- theoretical BER values for various digital modulations. It was
world ECOMESH test-bed in a native woodland environment, indicated that the Tx-Rx separation could be divided into three
emphasizing propagation issues in a dense foliage area and regions, defined by two critical points and a break or cross-
along a pathway over four vegetation growth seasons. Several over point. According to the QoS analysis, the near-ground
theoretical and empirical foliage models were evaluated, scenario was more constrained than an obstructed LOS. As a
including the FSPL, Fresnel, Plane Earth (PE) PL (2-Ray), W result, it may be necessary to use more robust digital
MED, COST-235, and FITU-R models. Then, two empirical modulation schemes or error correction codes to suit the
models were developed to predict the performance of network performance of LOS-based networks. The consequence of
attenuation in two different scenarios. more accurate characterization and formulation of near-
The results in [71] indicated that node height significantly ground systems would be a significant improvement in their
impacted channel propagation characteristics and feasible applications, particularly in the near future of 5G-IoT.
transmission distance, with RSSI decreasing monotonically as The focus of [70] was on the effective deployment of wireless
antenna height was changed. Additionally, the wireless sensors for applications such as tracking cow grazing behavior
22 VOLUME XX, 2021
on grass or track sporting events. As such, an LNSPL-based the difficulty of precisely calculating the location of a mobile
model was proposed, considering experimental data from node due to channel impairments caused by multipath, NLOS,
natural, short, and tall-grass fields. The empirical PL models fading, or other interference issues, they consider two methods
were then compared to theoretical PL models such as FSPL and for estimating distance. The first method was based on
2-Ray. The findings indicated that theoretical models differ by LNSPL, whereas the second method used a proposed hybrid
12% to 42% from proposed models. Compared to the proposed particle swarm optimization–artificial neural network (PSO–
model in the short grass field, measured data from comparable ANN) algorithm to improve the accuracy of distance estimate.
environments revealed a MAPE of 1.1% and a Ts of 5.6. As a The LNSPL parameters were estimated using RSSI
result, they concluded that theoretical models are inappropriate measurements in outdoor and indoor scenarios. The distance
for estimating the PL for WSN application in dense grass fields between the mobile and coach locations was then computed
with a height of less than 1 m. Additionally, comparisons to using the hybrid PSO-ANN and LNSPL algorithm to optimize
similar past studies indicated a significant difference in PL and the precision of the predicted distance. The results indicated
the parameters of empirical models. that when compared to LNSPL and previous works, the hybrid
In addition to prior work, the authors of [44] highlight that PSO-ANN algorithm significantly improved distance estimate
there are currently no accurate and reliable propagation models accuracy. Additionally, the hybrid PSO-ANN algorithm
that ensure the successful and practical deployment of IoT achieved an MAE of 0.022m and 0.208m for outdoor and
devices in mountainous terrain. As a result, they examined the indoor scenarios, respectively.
effect of mountain terrain on the efficiency of tracking devices As in the previous sub-section, we summarize the reviewed
and wireless sensor nodes, conducting measurement campaigns studies in this sub-section, as shown in Table III, with a
involving IoT devices operating at 900 MHz and 2.4 GHz. They detailed description of each study’s modeling approaches,
then proposed and compared log-distance-based PL models to analysis metrics, key findings, and limitations. In conclusion,
FSPL, 2 Ray, and log-distance PL models. A variance of 8 dB it is worth noting that most of the research covered in this sub-
to 38 dB was observed, indicating that such models significantly section either evaluates the performance of these wireless
underestimated the performance of IoT systems in such systems in terms of coverage and propagation limits or
environments. Additionally, tests revealed that mountains and proposes channel models based on relatively simple models
rocks cause an average signal loss of 8 dB. As a result, it was such as 2-Ray, FSPL, and LNSPL. These studies were mostly
concluded that existing models are unsuitable, and some are limited to near-ground deployment scenarios utilizing low-
proposed based on FSPL and 2-Ray models to support high- power omnidirectional antennas. Furthermore, none of the
power systems in tree or wooded areas. studies reviewed considered more complex propagation
In order to analyze and enhance RSSI-based PL models, the scenarios or evaluated other models available in the literature,
authors in [117] identified several factors that have an such as deterministic ray-tracing-based models. As a result, it
undeniable negative influence on measured RSSI and, thus, on can be concluded that more study is needed to assess the
the performance of derived PL models. They classify these suitability of alternative models for diverse implementation
factors as intrinsic, resulting from transceivers, connectors, scenarios, such as the influence of different Tx/Rx deployment
antennas, or extrinsic, resulting from the number of heights and varied, challenging climate conditions, to examine
measurements, coexisting interfering devices, packet length, the full boundaries of such technologies and propose more
ED battery level, temperature, and environmental impacts. As accurate models.
a result, a set of adjustment models and computational
procedures was developed to resolve the disturbances above C. LPWAN based wireless IoT channel modeling and
for three well-known technologies: MicaZ, Iris, and characterization
Waspmote. They then compared two RSSI-derived models (a In recent years, the rapid expansion of ICT and the rapid
base model, LNSPL, with no adjustments to raw RSSI development of new technologies have increased the
measurements and an adjusted model, using proposed tools importance of rapid and accurate planning and deployment of
and procedures) to an electrical substation ground-truth VNA- emerging LPWAN wireless IoT technologies [154]. However,
based PL model. The average prediction error was reduced by such technologies’ success relies on signal propagation
around 91.76 %, indicating a significant improvement in robustness, especially in complex terrain and irregular
network simulation accuracy. As a result, operational elevation profiles [52], [174]. For example, LoRa offers a
expenses are reduced and estimating quality and network wide range of coverage options; it can span hundreds of meters
planning for critical industrial environments are improved. or tens of kilometers, depending on its surroundings and the
Overall, the findings indicated that when intrinsic and factors influencing its performance directly [118], [126],
extrinsic RSSI-affecting variables are included, RSSI-based [175]. In contrast, widely used LoRa channel models do not
models can be useful, precise, and cost-effective for WSNs. identify this high variability, and typical on-site measurement
Finally, [114] aimed to determine the distance between a options are inaccurate given the huge geographic areas
bicycle and a coach using a mobile ZigBee sensor node and a covered [118], [126]. As a result, several studies have been
ZigBee anchor node in outdoor and indoor scenarios. Due to undertaken in this domain.
VOLUME XX, 2021 23

TABLE III
A SUMMARY OF THE REVIEWED SHORT-RANGE BASED WIRELESS IOT CHANNEL MODELING AND CHARACTERIZATION STUDIES.
Environment/ Targeted Modeling approach &
Ref F. RAT Comparison Models Analysis Metrics Key Findings Limitations
Scenario application used input parameters
RMSE, SC-RMSE, Implemented 30 common models.
Outdoor Thirty models,
2.4 & 5.8 Comparative (no modeling). competitive Showed that a limited no of readings
Rural, urban, including foundational, Do not consider ray-tracing
[123] GHz, 900 Wi-Fi General Use RSS with some success, accuracy, can be used to tune & fit basic empirical
& suburban. basic, supplementary, models or indoor scenarios.
MHz calibration. skewness, rank models. Overall, Okumura-Hata &
PTP & mesh. & terrain models.
correlation Hata-Davidson are top performers.
Average, STD, The PL and empirical models’ Modeling is based on simple
WSN deployed FSPL, 2-Ray, & two
Outdoor 1.925 LNSPL model. absolute percentage parameters differed significantly. linear fitting. Do not
[30] N/A on a concrete empirical models for
LOS/NLOS GHz Use RSSI. error (APE), R2, FSPL & 2-Ray proved to be consider large distances.
surface. long grass / sparse tree.
MAPE. inaccurate for concrete surfaces. Mostly LOS.
Outdoor General WSN FSPL, 2-Ray, & two Showed the inaccuracy of FSPL & LOS only. No details on PTx
1.925 Linear regression to model R2, APE, &
[119] Sand terrain. N/A applications in empirical models for 2-Ray in predicting PL between & Tx/Rx height. Neglect sandy
GHz LNSPL. Use RSS. MAPE.
LOS (daytime). sand terrain. long grass / sparse tree. WSNs deployed in sand terrain. electrical ground properties.
One-slope log-distance & Node height impacted the propagation. Neglect obstacles, terrain, &
Outdoor ZigBee FSPL, 2-Ray, & one
Agriculture propose a modified two- The modified two-slope log-distance weather impact. The study
[71] Rice field. 2.4 GHz IEEE slope log-distance RMSE & RE.
monitoring slope log-distance model. performed better & claimed to be was conducted in a flat rice
Partial LOS. 802.15.4 model.
Use RSSI. usable for WSNs in rice fields. field during sunny days only.
ZigBee Mango FSPL, 2-Ray, log- Foliage models used in conjugation Few considered scenarios.
Greenhouse Linear regression curve-
[139] 2.4 GHz IEEE greenhouse distance, W MED, APE & MAPE. with FSPL & 2-Ray are inaccurate Use a simple linear curve
environment. fitting. Use RSSI.
802.15.4 monitoring ITU-R, & COST 235. in predicting PL in a greenhouse. fitting method.
Outdoor IEEE Empirical foliage models. FSPL, Fresnel Zone, States that a proper propagation Need further evaluation,
PA IoT RMS power, R2, &
[62] Dense 2.4 GHz 802.11 & Use Rx power from a PE, W MED, COST model that enables network performance considering other signal
applications P-value statistics.
vegetation. 802.15.4 spectrum analyzer. 235, & FITU-R. evaluation enables QoS optimization. influence factors.
Outdoor Three slop Log-Normal PL Indicates that Tx-Rx separation Terrain, weather, and NLOS
868 MHz, Analyze the first
Rural (soil, Smart model. Use spectrum analyzer FSPL & few previous could be divided into three regions, impact are not considered.
[104] 2.4 & 5.8 ZigBee Fresnel zone
short & tall agriculture power & RSSI for 2.4 GHz studies PL model. defined by two critical points and Only a few models are
GHz obstruction.
grass field). (commercial ZigBee). the break or cross-over point. compared.
Outdoor Monitor grass R2, P-value, F- 2-Ray is unsuitable for WSN. Need to consider the casing,
Measured vs. proposed,
Short & tall growth. Smart value, & α-level Theoretical models differ by 12-42% of electrical properties of grass,
[70] 2.4 GHz ZigBee LNSPL. Use RSSI. FSPL, 2-Ray, & a few
grass. Mostly irrigation statistics. MAPE, proposed models and cannot estimate terrain, ground, & weather
similar studies.
LOS. system. Ts, MSE, RMSE. PL in dense grass fields (≤1 m). effect on propagation.
Having precise PL models saves The electrical properties of
Outdoor ZigBee IoA (wildlife),
900 MHz Mean, MSE, power, improves node localization, mountains and rocks were
Rural (rocky & IEEE D2D/D2M Log-distance PL model. Use FSPL, 2-Ray, & log-
[44] & 2.4 RMSE, MAPE, reduces interference, and increases ignored. More study is needed
mountainous 802.15.4, & SAR, military, RSSI. distance PL model.
GHz and P-values. network capacity. Mountains and on different empirical/ray-
terrain). LoRaWAN hiking sports.
rocks reduce the signal by 8 dB. tracing models and IoT RAT.
RSSI-based models could be useful, Limited to the outdoor
Compared against a MSE, RMSE, precise, & cost-effective when scenario & do not consider
LNSPL. Use raw &
Outdoor IEEE IIoT (smart ground truth LNSPL mean network including intrinsic & extrinsic RSSI- ray-tracing models. Do not
[117] 2.4 GHz corrected RSSI to derive two
LOS/NLOS 802.15.4 grid). derived from high-end delay, & mean affecting variables. The proposed provide details of mounting
models for each platform.
VNA measurements. network lifetime. scheme improves real-world SG height of used equipment
characterization accuracy. and Tx power.
Outdoor & ZigBee WPAN Compare distance R2, RMSE, MAE Used hybrid PSO-ANN & LNSPL to Do not consider other PL
[114] Indoor. 2.4 GHz IEEE (localization of LNSPL. Use RSSI. estimation accuracy to assess predicted model the outdoor/indoor propagation models. Highly dependent
Cycling fields. 802.15.4 cycling sports) against related studies. distance accuracy & improve the estimated distance. on RSSI.
24 VOLUME XX, 2021
For instance, [22] described two smart city testbeds and spatial resolution. The focus was on the empirical
developed in Italy for public lighting based on IEEE 802.15.4 Okumura-Hata model, showing that its predictions are close
and smart buildings based on LoRa. The latter was then to their observations and that their method can automatically
investigated using measurement campaigns and simulations to select and configure its parameters. Additionally, the results
assess LoRa’s coverage and performance in an urban scenario. showed that their approach closely estimates expected signal
Similarly, in [25], the authors analyzed the feasibility of using power (ESP) within a 10 dB error, compared to a 20 dB to 40
LoRa-based WSNs in smart public transports to collect dB errors for widely used channel models. Nonetheless, the
pollution data and meteorological parameters. They also model ignores the impact of walls when GWs are placed
investigated propagation and network architecture for possible indoors and ignores other signal-influencing parameters (e.g.,
practical network realization. physical parameters).
According to [22], LoRa’s maximum coverage in dense In contrast, [176] evaluated the RSSI accuracy of two LoRa
urban areas was 1-2 km, far less than the 15 km claimed by chipsets in a laboratory environment. The Longley-Rice
manufacturers and vendors of LoRa. This record was achieved Irregular Terrain Model (ITM) was then simulated and
in favorable conditions, with a GW height of 71 m AGL and evaluated using SPLAT, a software tool, with real-world
the highest SF. Coverage is expected to be reduced even further Digital Elevation Map (DEM) resolution of 1 arc-second (~30
if such criteria are not fulfilled. Finally, the findings indicated m) to predict PL. An extensive measurement campaign was
that it is wise to deploy several LoRa GWs or equip a single conducted in Germany’s outdoor suburban environment to
GW with multiple receivers, operate on multiple channels, and verify ITM prediction accuracy and compare it to FSPL as a
use a larger SF to cover larger areas. The analysis in [25], on baseline model and the log-distance-based models proposed in
the other hand, showed that the proposed WSN is suitable for [177] and [26]. The result indicated that there is no perfect
integration in a city’s public transportation network since model for all environments. Further, the results showed that
theoretical propagation performance, based on the Okumura- the chipsets reported significantly different RSSI and terrain
Hata model, shows that LoRa delivers adequate outdoor urban data improves prediction accuracy. Finally, they concluded
coverage areas. As a result, they concluded that the that conventional terrain data-based PL models,
experimental results would provide significant indications predominantly used for mobile or TV broadcasting, cannot be
concerning the deployment of LoRa and testing on propagation used for LoRa wireless technology.
and networking while accounting for the Doppler effect. Researchers in [31], [52], [118] provided a detailed
In contrast, the authors of [26] used commercial LoRa evaluation of the LoRaWAN channel at 868 MHz for various
devices to determine and analyze LoRa coverage in Oulu, scenarios. Accordingly, [118] conducted detailed
Finland. Measurements were made utilizing a node on the measurement campaigns for indoor and outdoor scenarios in
ground, mounted to a car’s roof rack, or on the water, Lebanon’s urban and rural environments. They then proposed
mounted to a boat’s radio mast, transmitting data to a GW. a set of PL models and evaluated their accuracy against
They then presented an LNSPL-based model for the 868 commonly used empirical PL models. These models include:
MHz band at 14 dBm Tx power and maximum SF. Results 1) ITU-R, Cost 231-MWF, 3GPP’s Cellular-IoT models for
showed that the PLE for the on-ground scenario was greater indoor scenarios and 2) Okumura–Hata, Cost 231-Hata, and
than the PLE for the FSPL scenario due to the presence of 3GPP-UMa/RMa for outdoor scenarios. They observed that
buildings and other obstacles in the path between the end the proposed PL models are better, accurate, and simple to
device and GW. Meanwhile, for the boat scenario, it was apply in the study area or similar places.
1.76, below that of FSPL. Meanwhile, results showed that a coverage distance of up
Interestingly, they achieved higher coverage ranges, to 9 km and 47 km could be obtained in urban and rural areas,
reaching over 15 km on the ground and nearly 30 km on water. respectively. The ITU-R model showed lower precision for
However, they found that for on-ground, the PDR tends to be indoor scenarios with a mean error and standard deviation
high, over 80%, for ranges up to 5 km then degrades for higher error of 0.48 dB and 8.3 dB, respectively. Similarly, Cost 231-
ranges, being lowest for more than 10 km. Meanwhile, PDR MWF and 3GPP models underestimated the measured PL and
was nearly 70% at distances under 15 km for the water reported a standard deviation error of 8.7 dB and 10.2 dB,
scenario. Finally, they claimed that network operators could respectively.
utilize the derived model to estimate the needed GW density In [52], the authors provided a detailed performance
and enable precise LoRa performance analysis. analysis in urban, suburban, and rural environments. Various
In [126], the authors presented an automated method for PHY layer settings were studied to evaluate the most suitable
estimating post-deployment coverage of LoRa GWs in one based on propagation conditions. Next, they took a
outdoor environments without on-site measurements by different approach, assessing the predicted signal strength in
integrating free multi-spectral images from remote sensing these scenarios using an RF planning tool (Cloud-RF®) that
with the correct channel model. The method automatically uses topographic maps and the Okumura-Hata model. Then,
classifies the type of environment (such as buildings, trees, or an extensive measurement campaign validated the theoretical
open fields) penetrated by a signal with high precision (~90%) findings. In urban and suburban scenarios, coverage ranges of
VOLUME XX, 2021 25
about 6 km were obtained, while a long transmission range of the derived PL model based on a maximum-likelihood
over 18 km with the lowest data rates (DRs) was obtained in approach outperformed the least-square method and existing
the rural scenario. Hence, it was concluded that there is a clear empirical models for all studied environments.
trade-off between link reliability and DR (and therefore packet Authors in [175] performed an experimental study to
time-on-air); thus, the LoRaWAN configuration parameters evaluate the performance of LoRa by considering various
must be adjusted appropriately, based on the propagation topographical areas on-campus sites and discussing impacting
conditions and the range between GW and EN. Finally, they factors such as the Doppler effect, Fresnel zone,
examined the same scenarios but in a stationary condition to environmental factors, and interference. The results indicated
evaluate the mobility impact on performance, which indicated that the GW location, environmental scenarios, and topology
that LoRaWAN poses a significant Doppler-related must be considered during deployment to improve
vulnerability if using high DRs; however, this impact was far performance. Furthermore, an outdoor experimental study was
less evident when using low DRs. It was concluded that it is performed to test LoRa’s coverage and performance, based on
essential to review the deployment scenario’s propagation RSSI and PDR, by applying different SF in different terrain
conditions before actual implantation to reach a compromise and environmental scenarios. The findings indicated that
between network reliability and transmission DR. terrain impacts LoRa signal propagation, limiting coverage
In comparison, [31] examined the impact of seasonal and reducing PDR by half.
weather changes on the signal-to-noise ratio (SNR), RSSI, and In [127], the authors evaluated the RSSI measurements of
SF usage while enabling adaptive data rate (ADR). the LoRaWAN network implemented in Skelleftea, Sweden.
Additionally, they evaluated Okumura-Hata and ITM models Then, they compared it to data obtained from Cloud-RF®
using Cloud-RF® to see how they fit real-world based on ITM, Irregular Terrain with Obstructions Model
measurements. They observed that results gain from cold (ITWOM), and the Okumura-Hata model. These models were
weather since low temperatures increase noise intensity, used since Okumura-Hata is a well-known model for large
resulting in a higher SNR. In contrast, the battery drains more cellular networks, while ITM and ITWOM are recommended
quickly in cold temperatures. Further, the findings indicated by the planning tool. The data were obtained from five sensors
that the combination of long distances and snow had a mounted along a wooden bridge with various SF
negative impact on LoRaWAN performance. As a result, configurations (7, 10, and 12). Hence, results showed that
snow must be considered while designing the deployment of ITWOM accuracy was nearest to measured values due to
sensors and GWs to reduce the effect of snow on signal ITWOM improvements that include the adoption of Radiative
transmission. When measured RSSI was compared to Cloud- Transfer Engine (RTE) rather than the classical theory of
RF® predictions, Okumura-Hata was the best, whereas the diffraction. They, therefore, concluded that ITWOM
ITM model seemed to overestimate RSSI. As a result, they performed the best, followed by ITM and then Okumura-Hata.
indicated that propagation models should be optimized by However, the study considered only outdoor scenarios and did
taking weather into account when tuning the models. It was not consider SNR values to infer channel conditions. The
also concluded that temperature should be considered while study also indicated that terrain profile, environment, and
designing a LoRaWAN implementation strategy due to its transmission distance are significant when selecting a
effect on battery life, propagation conditions, and noise planning tool’s specific propagation model.
behavior, all of which impact coverage and transmission rates. To run WSN all year round and provide a comprehensive
In contrast, [178] evaluated the P2P LoRa coverage and communication range of several tens of kilometers, authors of
proposed a log-distance-based PL model based on [179] used LoRa technology in Antarctica’s outdoor polar area
measurements conducted in urban, forest, and coastal areas, to implement these nodes with minimal power consumption.
utilizing low-height nodes of 1.5 m at 868MHz. The PER was They studied the system propagation channel conditions
then computed based on the predicted PL parameters to within 30 km of the Belgian Princess Elisabeth Antarctic
quantify P2P link quality. In all scenarios, 80% of the data Station. Hence, they verified the usefulness of LoRa
were successfully received at around 200 m. While, for the technology in polar regions and found that installing
LOS scenario, a range of over 4 km was achieved. Given the directional antennas at the BS would achieve a maximum
urban scenario’s unfavorable radio propagation and heavily range of 30 km in both the 434 MHz and 868 MHz bands for
forested areas, a 1 km distance was achieved. They observed the LOS scenario. Besides, the dominant factor affecting
relatively high variation around the standard deviation (σ) of propagation was varying terrain elevation, often obstructing
shadowing samples in the coastal area due to handheld device the LOS path. Finally, they suggested that NLOS links are
usage, resulting in higher PER at shorter distances. They also possible but require field measurements or ray-tracing
indicated that the Okumura–Hata model significantly simulation to determine optimal antenna locations.
overestimates PL, whereas the modified Cost 231-Hata model In [154], the authors explored the Lee propagation model’s
and the model derived in [26] underestimated PL. They also potential use and optimization for 868MHz LoRa network
observed that the aligning trees acted as waveguides yielding planning, design, and management. The analysis was based on
a PLE near the PLE of FSPL for the forest scenario. Finally, RSSI field measurement comparison with model calculations.
26 VOLUME XX, 2021
Hence, based on results, RMSE averaged 6.71731 dB for Different buildings are known to have different
Tokyo, 15.0949 dB for Philadelphia, and 19.5495 dB for communication constraints due to varying sizes, shapes, and
Newark. These analyses showed the accuracy of the Lee PL structures. Therefore, the authors in [28] conducted a detailed
model for urban areas. Nonetheless, they point out that further study to investigate the large-scale fading characteristics,
measurements and comparisons with the Lee model should be temporal fading (TF) characteristics, coverage, and energy
carried out. consumption of LoRa technology in four types of multi-floor
In contrast, the authors in [180] evaluated foliage buildings. They also conducted an RSSI adjustment
attenuation and its overall contribution to PL and link budget experiment and observed a constant 2 dB shift between the
estimates. Accordingly, the attenuation impact of five tree measured RSSI of the LoRa node (mDot) and spectrum
types was studied for different path crossings (e.g., trunk, tree- analyzer-measured RF power. The PL characterization results
top, and branches). For this evaluation, the Okumura-Hata, showed that a non-fixed intercept model has significantly
LNSPL, and foliage models were used as references. Both larger intercept PL(d0) and smaller PLE than the fixed
LOS and NLOS (across tree) tests are considered through intercept model. Also, standard deviations for non-fixed
RSSI measurements for horizontal and slant paths. After that, intercept and fixed intercept fit well, although it is somewhat
both horizontal and slant paths RSSI reading over a tree are larger for a fixed intercept. Therefore, concluding that using a
taken. Various measurements also determined LoRa non-fixed intercept PL model is better than using a fixed-
communication coverage in a campus area. They found that intercept model. A Kolmogorov-Smirnov (K-S) goodness-of-
the Mimusops Elengi tree, characterized by its large size and fit test was conducted on one-slope shadow fading samples to
leaf density, provided the highest foliage attenuation of up to determine log-normality.
20 dB. Trunks showed higher attenuation than tree-tops and They concluded that a one-slope PL model can estimate PL
branches. The study also showed that the Okumura-Hata in indoor scenarios but can only be used as a first-order
model failed to capture the foliage effect compared to prediction. Hence, they indicate that an attenuation factor (AF)
measurements. However, the study did not consider the effect model is needed to attain more accurate PL prediction in a
of thick foliage, especially in a dense jungle area, where multi-floor building. A site-specific model should also be
attenuation is expected to exceed 20 dB. Also, it does not considered since building materials, structure, and other
consider other impacting factors such as weather and being a factors could significantly affect the PL. Also, the
limited evaluation based purely on RSSI measurements. decorrelation distance in a multi-floor building is small,
On the other hand, the authors in [181] evaluated suggesting that the large-scale fading is almost independent of
LoRaWAN propagation in an outdoor-indoor scenario based one area to another. Results also showed that TF met a Rician
on RSSI measurements and compared it to commonly used distribution with Rician K-factors of 12 dB to 18 dB. Hence,
propagation models such as log-distance and indoor link budget analysis should consider a fade margin of 9 dB and
COST231-MWM models. They also adjusted the COST231- 7 dB for tested scenarios. The study also showed that energy
MWM model for better accuracy. However, these models did consumption would vary up to 145 times, using different
not accurately estimate outdoor–indoor propagation parameter configurations. Such results highlight the
characteristics. Thus, a novel hybrid propagation estimation importance of selecting parameters and enabling LoRa’s
method was developed and examined. This hybrid model adaptive data rate feature in energy-limited applications.
consists of ANN and an Optimized COST231-MWM, thus Moving forward, the authors in [40] presented
showing higher predictive accuracy and reduced initial measurement campaign results to evaluate empirical
COST231-MWM MSE from 21 to 11.23. characterization and mathematical modeling of the radio
Similarly, for LoRa channel characterization and link channel for a wearable LoRaWAN node for different
performance analysis, an autonomous LoRa-compatible node operating ranges across different environments such as urban,
was presented in [182] for both 434 MHz and 868 MHz bands. suburban, and rural. Furthermore, anechoic measurements
They showed that the LoRa node provided a significantly were also carried out to evaluate body shadowing effects for
more extensive dynamic range by applying stepped this technology. Findings showed that the best fit model for all
attenuators controlled by a dynamic attenuation adjustment measured received signal strength, using the Akaike
algorithm. The node was calibrated to accurately measure the information criterion, is the Nakagami distribution with mu =
received signal power in dBm based on SNR measurements. 0.52 and Ω = 662.13. Furthermore, anechoic measurement
Findings showed a correlation between a sudden drop in signal showed typical additional effects regarding the user
and an event of rainfall, whereas outdoor temperature orientation concerning the GW location.
fluctuation showed no correlation with measured signal levels. As mentioned in previous sections, advances in wireless
Finally, indoor measurements showed that people's presence sensor technology and MEMS have enabled dairy cow health
in a building also has a measurable influence on the LoRa link conditions to be monitored remotely using the IoT and
quality, where 2 dB and 3 dB difference in standard deviation WBAN. While on-cow measuring devices are energy-
values were found between day and night measurements, constrained, adequate characterization of the wireless off-
respectively. body link between the on-cow sensor nodes and the back-end
VOLUME XX, 2021 27
GW is needed for the improved operation of these networks in role in designing a WUC system. In [184], the authors proposed
barns. Thus, authors in [74] characterized the 868 MHz off- a WUC PL model for precision agriculture, in which, first, the
body wireless channel for dairy cows in three different barns. underground communication was simplified to a generic model.
LoRa motes investigated both PL and TF. RSSI calibration The developed model was then integrated with an accurate
was also performed, resulting in a constant 6dB adjustment prediction of the complex dielectric constant (CDC), called the
between measured RSSI and actual RF power. Results showed Mineralogy-Based Soil Dielectric Model (MBSDM). The
that a one-slope LNSPL model could be suitable for the large- integrated method can predict PL in different scenarios,
scale fading characterization. It was also shown that the underground-to-underground (UG-UG), underground-to-
maximum PL increase of around 4 dB resulted from the cow aboveground (UG-AG), and aboveground-to-underground
body wearing the sensor node. Meanwhile, other cows had (AG-UG). The PL model considers reflective and refractive
less influence of about 1 dB. The TF was statistically wave attenuation according to the sensor node burial depth in
characterized by Rician distributions with an average K-factor each scenario. To further validate the reliability of the proposed
of 8 dB. Finally, the authors claim that study findings could method, intensive experiments were conducted in a real
enable reliable IoT cow monitoring systems with optimized environment with two different pairs of wireless transverses,
network planning and energy consumption. nRF905 and LoRa SX1278. The results showed that the
In contrast, the authors in [75] characterized the in-to-out proposed model outperforms the existing PL models in different
body PL between an antenna inside the cows' rumen and a communication types and soil conditions, where the proposed
specific GW at 433 MHz. Measurements were conducted on model can be used on a real cheap sensor with 87.13% precision
seven different fistulated cows using a signal generator and a and 85% balanced accuracy.
spectrum analyzer. Later free space antenna measurement was On the other hand, the PL analysis of underground wireless
conducted to evaluate PL increase due to the cow body. Results communications in urban UIoT for wastewater tracking was
showed an average PL increase of 45.5 dB (all cows), with 39.7 presented in [60]. They demonstrated that an underground
dB and 51.1 dB variability. Also, an LNSPL model matched the transmitter, communicating through a 10 cm thick asphalt
measured PL as a transmitter-receiver distance function in a layer, could achieve a range of up to 4 km with PL less than
dairy barn. The observed models were then used to evaluate a 100 dB and 10 km with PL of 107 dB. The propagation loss
LoRa-based network range. Therefore, using the highest LoRa was also observed to be less than 5 dB with a layer thickness
transmit power of 20 dBm, the range reached up to 100 m with below 1 m. However, as the layer thickness increases, it can
the least DR, whereas using the highest DR, the range was reach up to 15 dB for the 4 m asphalt layer. It was also
limited to 11 m. Therefore, they concluded that coverage could observed that RSSI drops with distance, with dramatic drops
be increased by using a lower DR with a higher transmit power for distances below 2 km. It decreases gradually afterward.
while reducing the battery life and data collected. Finally, at a communication distance of 4 km, the -80 dBm
The authors in [183] proposed a LoRa-based positioning RSSI demonstrates that underground devices in urban UIoT
algorithm for search and rescue operations in the mountain could communicate effectively with urban roadside wireless
environment. The positioning algorithm was developed based communication infrastructure.
on the PL measurement. Hence, a PL model was developed Finally, for Sigfox devices, their location is predicted and
according to the measured SNR and RSSI. The measurements stored on the Sigfox cloud platform. However, the location
were conducted in three relevant mountain scenarios: canyon prediction accuracy is unsatisfactory, where, in some cases,
(with maximum width, length, and depth of 40 m, 8 km, and the distance error between predicted and actual location may
400, respectively), LoRa Tx over the snow, and Tx buried exceed 20 km or 30 km, with average distance error ranging
under 1 m of snow. In measurements, Tx was fixed and placed from 2 km to 10 km, as indicated by authors of [185].
on/under the ground (depends on the considered scenario), and Accordingly, they proposed a new ML-based localization
Rx was mobile and put inside a volunteer’s jacket. For the method to estimate the Sigfox device location. The method
measurements, two LoPy-4 expansion boards were used that divides the area around each BS into a few sectors and trains
operated in the 868 MHz frequency band with a bandwidth of a more precise PL model, based on LNSPL, for each sector to
125 kHz, spread factor of 7, the coding rate of 4/5, and Tx represent their site-specific multipath propagation
power of 14 dBm. The results showed that although the environment. Hence, the proposed method used RSSI
communication range of LoRa decreased from kilometers to measurements and a multi-sector training and predicting
hundreds of meters, the operation range of the proposed method with four 90-degree sectors. The experimental data
method is at least five times greater than the golden standard observed in a big city utilizing 30 Sigfox devices showed that
technologies such as ARVA. the proposed method maximum distance errors are often
Wireless underground communication (WUC) has numerous smaller compared to the official localization service of Sigfox.
applications in ecology, agriculture, health care, and As in the previous sub-section, we summarize the reviewed
environment preservation. However, the propagated signal is studies in this sub-section, as shown in Table IV, with a
severely attenuated as it travels across the ground due to the soil detailed description of each study’s modeling approaches,
composition. Hence, a PL prediction model would play a vital analysis metrics, key findings, and limitations.
28 VOLUME XX, 2021

TABLE IV
A SUMMARY OF THE REVIEWED LPWAN BASED WIRELESS IOT CHANNEL MODELING AND CHARACTERIZATION STUDIES.
Environment/ F. RAT Targeted Modeling approach & Comparison Analysis
Ref Key Findings Limitations
Scenario (MHz) BW (kHz) SF application used input parameters Models Metrics
LoRa Smart City No comparison LoRa’s maximum coverage is ≤ 2 Do not consider terrain,
Outdoor (Smart One-slope log-distance Performance km in dense urban, only achievable weather impact & other
[22] 868 RMSE & PDR
Urban R.W. 12 lightning & model. Use RSSI. evaluation (practical at favorable conditions (max SF & models for comparison. It is
125 buildings). & simulation data). GW at ≥71m). purely based on RSSI.
Sim. 7-12
LoRa Smart City Evaluation of Demonstrated LoRa as an ad-hoc
Outdoor Okumura-Hata Based on a theoretical
[25] 868 (Pollution & N/A maximum network solution for public
Urban N/A N/A (theoretical). evaluation only.
meteorological). range. transport.
Outdoor LoRaWAN Automotive & LNSPL (derive parameters Measured vs. Claim that the derived model can be Do not compare to other PL
[26] Urban (city & 868 intelligent using linear polynomial predicted PL. FSPL N/A used to estimate needed GW density models. Based on simple
over water). 125 12 transportation. fit). Use RSSI & SNR. (baseline). & analyze LoRa performance. linear fitting.
FSPL for the balloon tests.
LoRaWAN Automated coverage estimation Do not consider other
Outdoor LNSPL & Okumura-Hata Prediction vs. data Estimation
General IoT method utilizing free remote impacts (e.g., physical
[126] Urban & rural. 868 for remaining. Use ESP extracted from TTN, error (average
applications. sensing multi-spectral images with parameters) or comparison
LOS/NLOS. 125 7 derived from RSSI & FSPL, & [186]. & STD).
a channel model. with other known models.
SNR.
Measured vs. ITM,
LoRa RSSI accuracy evaluation showed a RSSI-based. Low-resolution
Simulated ITM using FSPL (baseline), RSSI false
Outdoor General IoT significant difference. No model fits DEM, ignoring the reflection
[176] 868 SPLAT with DEM of 1 arc- models in [177] & positive
Suburban applications. all environments, & terrain data & diffraction impact of
125 12 second (~30m) resolution. [26] (based on log- prediction.
increases prediction accuracy. buildings or small terrain.
distance).
FSPL (baseline). Outdoor modeling uses
Extensive measurements in three
LoRaWAN Indoor: ITU-R, Cost simple linear fitting. Do not
environments. The proposed PL
Outdoor/indoor 231-MWF, & 3GPP. CDF, mean compare to ray-tracing
Indoor: ITU-R & Cost231- models are better, accurate, &
Rural, urban, & General IoT Outdoor: Okumura– error, STD models. More study is
[118] 868 MWF. Outdoor: LNSPL. simple to apply in the study area or
suburban. applications. Hata (urban/rural), error, & PDR needed on LoRaWAN
Use RSSI & SNR. similar places. Evaluated LoRaWAN
LOS/NLOS 3GPP-UMa, 3GPP- analysis. adaptability and scalability,
125 12 PDR/SNR & confirmed its reliability
RMa & Cost231- network planning, and
for many IoT applications.
Hata (urban). energy efficiency.
Outdoor LoRaWAN Smart city, Examine expected Range & PDR Present a detailed LoRaWAN No model was proposed,
No modeling
[52] Rural, urban, & 868 farm, & goods signal using evaluation at a performance analysis in three limited to outdoor scenarios
125 7-12 Uses RSSI data
suburban. monitoring. Okumura-Hata. varied payload. typical outdoor environments. and RSSI-based evaluation.
Examine how seasonal weather Do not consider building

changes impact SNR, RSSI, & SF impact, power consumption,
LoRaWAN SF histogram,
usage while enabling ADR. Low PDR, & PL modeling. The
Measured vs. RSSI CDF vs.
Outdoor temperatures increased noise simulation lacked clutter &
No modeling. Okumura-Hata & Cloud-RF®
[31] Urban 868 Smart city. intensity, leading to better SNR. knife-edge features, resulting
Use RSSI & SNR. ITM, simulated in estimates, RSSI
LOS/NLOS Long distances & snow reduce in lower accuracy. Need
Cloud-RF®. vs. SNR, SNR
7, 10, 12 & LoRaWAN performance. further research on how snow
125 CDFs & ADR.
ADR Okumura-Hata was the best, while influences signal transmission
ITM seems to overestimate RSSI. at different distances.
Outdoor LoRa P2P Tx/Rx height is limited at
Okumura-Hata, Cost Okumura-Hata significantly
Urban (campus General IoT Log-distance PL. Use RMSE & 1.5m. Do not consider
[178] 868 231-Hata, & [26] vs. overestimated PL. Modified Cost
/ city) & rural applications. RSSI & SNR. MAE. irregular terrain or ray-
N/A N/A derived PL models. 231-Hata & [26] underestimated PL.
(forest / coastal). tracing models.
VOLUME XX, 2021 29

TABLE IV (CONTINUED)
Indicated that terrain impacts LoRa Suggest further research on
LoRa RSSI & PDR signal propagation, limiting coverage optimal coverage, tropical
Outdoor General IoT N/A for coverage & & reducing PDR by half. Also, GW regions, e.g., Malaysia, rain
[175] 923 N/A
Mixed. applications. Use RSSI. performance location, environmental scenarios, & impact on LoRa performance,
125 7-10 analysis. topology must be considered during & required mitigation strategies
deployment to improve performance. to increase performance.
Outdoor LoRaWAN Measured RSSI vs. Terrain, environment & distance are
Pollution & Do not consider SNR, TF,
City of mixed No modeling. predicted from Cloud- significant factors for PL model selection
[127] 868 dust bin Error in dB. weather & other models.
Suburban/rural Use RSSI for comparison RF® (Okumura-Hata, of planning tool. ITWOM was the best,
N/A 7, 10, & 12 sensors. Limited to RSSI comparison.
areas. ITM, ITWOM). followed by ITM, & Okumura-Hata.
Outdoor LoRa No modeling (propagation No comparison SNR & PR Environment factors did not impact
434 & Environment Do not consider propagation
[179] Antarctic limits). Use calibrated Consider log- signal & terrain was the dominant
868 125 12 monitoring. calibrated. modeling or comparison.
LOS/NLOS SNR to calculate PR(dBm). distance & FSPL. factor influencing the propagation.
LoRa RSSI-based. Limited to Lee
Outdoor MSE & Based on experimental
[154] 868 Smart city. Lee PL Model. Use RSSI. N/A model without comparisons
Urban N/A N/A RMSE. measurements.
& statistical analysis.
LoRa It was shown that Okumura-Hata Few measurements. Purely

Outdoor
General IoT Foliage excess attenuation Okumura-Hata & fails to capture the effect of foliage RSSI-based. Other foliage
[180] Foliage 915 N/A
applications. loss. Use RSSI. LNSPL. in the tropical areas under models & signal influencing
LOS/NLOS 125 10 consideration. factors are not considered.
Outdoor-to-Indoor propagation has
Outdoor-to- LoRaWAN ANN, adjusted COST231, Log-distance, a shorter range. COST231 & log-
General IoT
[181] indoor 868 & optimized COST231- COST231, & MSE distance are inaccurate. With the Purely RSSI-based.
applications.
Urban 125 9 MWM. Use RSSI. Cost231-MWM. proposed model, the COST231-MWM
MSE dropped from 21 to 11.23.
A sudden drop in signal correlated
Propagation limits &
LoRa with a rain event, while temperature
people presence impact. Do not consider modeling
Outdoor/Indoor Mean & STD fluctuations had no impact. The
434 & General IoT Use calibrated SNR to PL, terrain impact, foliage
[182] Urban N/A [day & night indoor study showed a 2-3 dB
868 applications. calculate PR(dBm). Outdoor impact, & comparison with
LOS/NLOS data]. difference in STD between day &
125 12 coverage simulated in other models.
night readings when people are
Radio Mobile.
present in a building.
Investigate large-scale fading, TF,
LoRaWAN Large-scale fading: One Mean/standard
Indoor 915, FSPL & compare coverage, & energy consumption in Do not compare with other
Gas meter, slop or LNSPL. TF: Rician Error. Auto-
Obstructed 919, large-scale calculation four multi-floor building types. Indoor PL models. Limited
[28] smart home & distribution. Floor impact: correlation.
LOS/ LOS/ 923, & methods (non-fixed There was a constant 2 dB difference to indoor propagation
125, 250, buildings. AF model. Use calibrated PDR & power
NLOS 928 7-10 & fixed intercept). between the LoRa node’s RSSI & modeling.
& 500. RSSI & Rx power. consumption.
the spectrum analyzer’s RF power.
The anechoic measurement
LoRaWAN Based on CDF, maximum revealed typical user orientation Further investigation and
Outdoor Medical
[40] 868 likelihood (ML), & N/A N/A effects concerning GW location. comparison with other
LOS/NLOS bleeder.
N/A N/A Nakagami. Use RSSI. They show LoRaWAN as a viable models are required.
wearable wireless technology.
30 VOLUME XX, 2021

TABLE IV (CONTINUED)
A constant 6dB shift was found. TF
LoRa was statistically described by
Indoor LNSPL & TF to evaluate Measured PL vs. Based on a simple one-slope
IoA (health Rician distribution with an average
Rural (three cows’ movement impact modeled PL, FSPL empirical model. Do not
[74] 868 tracking of R2 & mean. K-factor of 8 dB. They argue that
dairy cattle on signal strength. Use (baseline), & TF consider other models for
dairy cows). their results may enable reliable
brans). N/A N/A calibrated RSSI. evaluation. comparison.
dairy cow IoT, optimizing network
planning and energy consumption.
They characterized PL between a
LoRa GW & an antenna inside a cow’s
Compare PL rumen. The average PL increase Do not use actual LoRa
Agriculture IoA (health LNSPL. Use measured
between with & was 45.5 dB (all cows), with 39.7- hardware. Based on a
[75] with/without 433 tracking of signal power from a R2 & STD.
without cow case. 51.1 dB variability. PL well fits simple PL model with linear
Cow. dairy cows). spectrum analyzer.
N/A N/A Consider FSPL. LNSPL, & depending on the Tx fitting.
power & DR, 100 m range was
achieved.
Outdoor LoRa Developed a PL model for LoRa in Need further experiments in
Measured vs.
Rural Search and LNSPL. Use RSSI & a harsh mountain environment and different scenarios. Only
[183] 868 estimated PL & STD
(mountain) Rescue SNR. considered the effect of human compared with FSPL &
125 7 FSPL.
LOS/NLOS body shadowing on wearable Tx. lacks statistical analysis.
LoRa Based on the Friis model

The model covers different
Underground Precision where the α and β are Conventional/NC RMSE, MAE, Further studies are required
[184] 433 scenarios, UG-UG, UG-AG, & AG-
WSN agriculture modified based on modified Friis. & MAPE. in different types of soil.
N/A N/A UG.
MBSDM.
Showed that an underground Tx
Evaluate the
N/A Statistical model, could communicate up to 4 km It does not provide a
temperature &
Urban considering losses caused across a 10 cm thick asphalt layer. detailed description of the
asphalt
[60] Underground 433 UIoT by two stages: FSPL & N/A The PL is less than 5dB for layers measurement setup; limited
thickness
communication through asphalt losses. Use under 1m thick. It rises with layer to signal loss analysis, with
N/A N/A impact on
RSSI. thickness, reaching 15dB for 4m no comparison.
signal loss.
asphalt.
Proposes using ML with RSSI to
Sigfox ML localization model improve Sigfox localization accuracy.
Do not provide details on
based on LNSPL. Hata & ML model prediction Improves localization accuracy by
antenna specs & Rx height.
Outdoor Smart meter & Okumura-Hata were vs. the location Average using a multi-sector (four 90°)
[185] N/A Limited to the usage of
Urban city assets tracking. trained to obtain dummy reported by Sigfox distance error. training & prediction design. This
uncalibrated RSSI & outdoor
data to increase ML cloud. method’s maximum distance errors
N/A - scenarios.
accuracy. Use RSSI. are often lower than the official
Sigfox localization service.
VOLUME XX, 2021 31
In conclusion, it can be noted that among the several Interestingly, it was noted that, among the numerous
LPWAN technologies, LoRa/LoRaWAN has gotten the most existing LPWAN technologies, LoRa/LoRaWAN had
interest from researchers because of its unique characteristics. attracted substantial interest from the research community due
As a result, most studies evaluated this technology’s to its unique features. As a result, the majority of LPWAN
performance and proposed channel models for various research focused on the performance evaluation of this
deployment scenarios. However, these studies mostly used technology and channel model proposals for various
simple models like LNSPL and Okumura-Hata models and implementation scenarios. However, most of this research
primarily focused on urban and suburban implementation used simple models like LNSPL and Okumura-Hata models
scenarios. As a result, we anticipate a significant gap in and concentrated on urban and suburban implementation
channel modeling and characterization for these technologies, scenarios. As a result, we anticipate a significant gap in
particularly in areas with complex terrain and harsh climate channel modeling and characterization for these technologies,
conditions, such as tropical regions. Although some studies particularly in areas with complex terrain and harsh climate
considered utilizing more accurate ray-tracing methodologies conditions, such as tropical regions. Although several studies
or simulation tools, these did not explore the entire limitations have considered utilizing more accurate ray-tracing
of LPWAN technologies. On the other hand, the low- methodologies or simulation tools, these did not explore the
resolution maps available limited most of these studies, entire limitations of LPWAN technologies. On the other hand,
making them site-specific and impractical for reuse in other the low-resolution maps available limited most of these
deployment areas. studies, making them site-specific and impractical for reuse in
In this regard, many future work directions may be possible other deployment areas.
in wireless channel modeling and characterization for LPWAN The recent trend of using low-altitude platforms (LAP) and
technologies. Among these, it is believed that measurement- high-altitude platforms (HAP) such as UAVs have
based and comparative analysis studies are of utmost desire to tremendous future opportunities, particularly for rural and
accurately understand such technologies’ behavior difficult-to-reach IoT deployments. Furthermore, with recent
environments. Other modeling approaches may incorporate improvements in IoT, cloud & edge computing, and wireless
new ML-based and hybrid prediction approaches that combine communication technologies, UAVs are becoming more
well-known stochastic or deterministic modeling techniques maneuverable and smarter. As a result, the IoD is emerging as
with ML. Nonetheless, any of these approaches may provide a one of the promising technologies and use cases for UAVs.
tremendous opportunity, with the potential to have a significant Despite recent work on developing channel models for UAV
impact on a wide variety of critical applications. communications, more comprehensive models for air-to-air
(A2A) and air-to-ground (A2G) scenarios are still required.
VI. Challenges and Opportunities in Wireless IoT The communication channel utilized by UAVs has
Channel Modelling characteristics that are significantly different from those used
Choosing an accurate channel model to represent the actual by conventional models. The most notable characteristics are:
real-world wireless IoT deployment is a difficult task due to (i) highly dynamic propagation channel characteristics due to
imperfections in the deployment area, such as varied terrain, high UAV velocity, (ii) extreme temporal and spatial channel
the presence of large objects such as tall trees, and the varying variation due to UAV mobility, (iii) different LOS propagation
speeds of moving objects. In other words, the actual wireless probability than terrestrial communication due to different
IoT performance varies depending on the surroundings. flight heights, and (iv) additional shadowing effects from the
Although many academics have considered addressing the aircraft body and propellers. As a result, the characteristics of
problem of channel modeling, it can be concluded that a UAV channels must be thoroughly investigated across a wide
significant gap remains, and additional research is required to range of propagation environments, altitudes, and flying
resolve the crux of the problem. On the other hand, the speeds. Thus, more accurate analytical models are required to
proposed models primarily focus on empirical and site- characterize large-scale fading, direct and multipath
specific modeling methodologies that may not apply to other components, and spatial-temporal characteristics in non-
locations or environments. Hence, despite the high demand for stationary channels. Additionally, developing empirically
wireless channel modeling, numerous challenging issues must based channel models is critical for validating or disproving
be resolved to achieve more accurate modeling. theoretical models.
Other limiting factors in certain studies include used On the other hand, having a reliable A2A communication
equipment and deployment scenarios that are far from ideal or link is critical in multi-hop UAV networking. Although the
commercial use scenarios. The research was primarily focused A2A channel appears to be like the free space channel due to
on specific areas for cellular-based IoT technologies, utilizing its high LOS conditions, the dynamic environment and ground
experimental hardware with low height Tx and Rx reflections have a different effect. As a result, issues such as
omnidirectional antennas. Meanwhile, the research has mostly antenna orientation and Doppler spectrum must be
focused on near-ground deployment scenarios for short-range investigated in various A2A propagation scenarios.
IoT technologies utilizing low gain omnidirectional antennas.
32 VOLUME XX, 2021
Finally, ML techniques have advanced significantly in understand this issue, the study begins by briefly reviewing
wireless channel modeling, and studies have demonstrated several innovative wireless IoT-based applications and then
that ML-based methods can significantly improve the highlighting the most crucial challenges associated with them.
accuracy of PL prediction compared to conventional PL The study also includes a brief description of channel
models. However, most proposed methods rely on small input modeling, a generalized form of commonly used channel
sets and ignore the impact of multiple KPIs and other critical models, and a list of the equations for 34 well-known and
parameters such as topological information. A rich training set commonly used channel models, together with a thorough
covering a wide range of required information with high description of each model’s modeling parameters,
resolution plays a vital role in exploiting the potential of ML limitations, and operating conditions. Finally, the study
techniques and extracting the structural relationship between reviews recent advances in channel modeling for wireless
collected data from complex environments. Given the IoT technologies, describes gaps in existing research, and
diversity and dynamic nature of environmental conditions in suggests future research directions towards addressing these
IoT communications, providing rich measurement campaigns gaps.
for various environments and scenarios is considered a crucial Although many researchers have considered addressing the
step toward developing accurate DL-based channel models. problem of channel modeling, it can be concluded that a
Many future work directions in wireless channel modeling significant gap remains, and additional research is required to
and characterization of IoT technologies, in general, and resolve the crux of this problem. Moreover, to improve
LPWAN in particular, may be viable in this regard. Among modeling accuracy, many challenging issues must be
these, it is believed that measurement-based and comparative resolved. On the other hand, it was observed that, among the
analysis studies are of utmost desire to accurately understand numerous existing LPWAN technologies, LoRa/LoRaWAN
such technologies’ behavior in real-world environments. had gained significant research attention due to its unique
Other modeling approaches may integrate emerging machine features. While much research has been done on LPWANs, it
learning-based and hybrid prediction techniques, combining can be observed that there is still a big gap in channel modeling
popular stochastic or deterministic modeling techniques with and characterization of these technologies, especially in areas
ML. Nonetheless, any of these approaches may represent a with complex terrain and harsh climates like tropical regions.
significant opportunity, with the potential to have a significant Further, the use of more precise ray-tracing methodologies or
impact on a broad range of critical applications. Figs. 4 and 5 simulation tools in some research has not investigated the full
summarize the challenges and opportunities in wireless IoT potential of LPWAN technology. Meanwhile, the low-
channel propagation modeling, respectively. resolution maps available constrained most, making such
studies site-specific and impractical for usage in other
VII. Conclusions deployment areas.
For any communication system, the wireless channel Among many directions forward to tackle the channel
characteristics are a critical parameter that directly affects the modeling and characterization for wireless IoT technologies,
wireless signal traveling from the transmitter to the receiver especially LPWANs, it is believed that measurement-based
antenna through the channel. As a matter of fact, wireless and comparative analysis studies are crucial for fully
transmission has recently become the core for enabling understanding the behavior of such technologies in the real-
wireless IoT applications. As a result, it is crucial to world. Other modeling approaches may integrate emerging
investigate the propagation channel characteristics that machine learning-based and hybrid prediction techniques,
directly impact wireless transmission performance. Failure to combining popular stochastic or deterministic modeling
do so will have a negative impact on the planning and techniques with ML. Nonetheless, any of these approaches
deployment of any IoT application. As such, this review may represent a significant opportunity, with the potential to
addresses wireless channel characterization and modeling for have a significant impact on a broad range of critical
wireless IoT technologies, including a comprehensive review applications.
of recent advancements and studies in this area. To properly
FIGURE 4. Challenges and limitations in wireless IoT channel modeling.
VOLUME XX, 2017 33
FIGURE 5. Opportunities in wireless IoT channel modeling.
REFERENCES ground path loss for low altitude platforms in urban environments,” in
[1] I. R. Gomes, C. R. Gomes, H. S. Gomes, and G. P. dos S. Cavalcante, 2014 IEEE Global Communications Conference, Dec. 2014, pp. 2898–
“Empirical radio propagation model for DTV applied to non- 2904, doi: 10.1109/GLOCOM.2014.7037248.
homogeneous paths and different climates using machine learning [12] W. Wang, S. L. Capitaneanu, D. Marinca, and E.-S. Lohan,
techniques,” PLoS One, vol. 13, no. 3, p. e0194511, Mar. 2018, doi: “Comparative Analysis of Channel Models for Industrial IoT Wireless
10.1371/journal.pone.0194511. Communication,” IEEE Access, vol. 7, pp. 91627–91640, 2019, doi:
[2] B. S. L. Castro, M. R. Pinheiro, G. P. S. Cavalcante, I. R. Gomes, and 10.1109/ACCESS.2019.2927217.
O. de O. Carneiro, “Comparison between known propagation models [13] S.-H. Hwang and S.-Z. Liu, “Survey on 3GPP Low Power Wide Area
using least squares tuning algorithm on 5.8 GHz in Amazon region Technologies and its Application,” in 2019 IEEE VTS Asia Pacific
cities,” J. Microwaves, Optoelectron. Electromagn. Appl., vol. 10, no. Wireless Communications Symposium (APWCS), 2019, pp. 1–5, doi:
1, pp. 106–113, Jun. 2011, doi: 10.1590/S2179-10742011000100011. 10.1109/VTS-APWCS.2019.8851631.
[3] S. I. Popoola, A. A. Atayero, O. D. Arausi, and V. O. Matthews, “Path [14] N. Poursafar, M. E. E. Alahi, and S. Mukhopadhyay, “Long-range
loss dataset for modeling radio wave propagation in smart campus wireless technologies for IoT applications: A review,” in 2017 Eleventh
environment,” Data Br., vol. 17, pp. 1062–1073, Apr. 2018, doi: International Conference on Sensing Technology (ICST), Dec. 2017,
10.1016/j.dib.2018.02.026. pp. 1–6, doi: 10.1109/ICSensT.2017.8304507.
[4] A. A. Jimoh, N. T. Surajudeen-Bakinde, N. Faruk, A. A. Ayeni, O. O. [15] M. Shahidul Islam, M. T. Islam, M. A. Ullah, G. Kok Beng, N. Amin,
Obiyemi, and O. W. Bello, “Performance analysis of empirical path loss and N. Misran, “A Modified Meander Line Microstrip Patch Antenna
models in VHF & UHF bands,” 2015 6th Int. Conf. Inf. Commun. Syst. With Enhanced Bandwidth for 2.4 GHz ISM-Band Internet of Things
ICICS 2015, pp. 194–199, Apr. 2015, doi: 10.1109/IACS.2015.7103226. (IoT) Applications,” IEEE Access, vol. 7, pp. 127850–127861, 2019,
[5] H. M. Rahim, C. Y. Leow, T. A. Rahman, A. Arsad, and M. A. Malek, doi: 10.1109/ACCESS.2019.2940049.
“Foliage attenuation measurement at millimeter wave frequencies in [16] W. Ayoub, A. E. Samhat, F. Nouvel, M. Mroue, and J.-C. Prevotet,
tropical vegetation,” in 2017 IEEE 13th Malaysia International “Internet of Mobile Things: Overview of LoRaWAN, DASH7, and
Conference on Communications (MICC), Nov. 2017, pp. 241–246, doi: NB-IoT in LPWANs Standards and Supported Mobility,” IEEE
10.1109/MICC.2017.8311766. Commun. Surv. Tutorials, vol. 21, no. 2, pp. 1561–1581, 2019, doi:
[6] S. M. Aldossari and K.-C. Chen, “Machine Learning for Wireless 10.1109/COMST.2018.2877382.
Communication Channel Modeling: An Overview,” Wirel. Pers. [17] R. S. Sinha, Y. Wei, and S.-H. Hwang, “A survey on LPWA
Commun., vol. 106, no. 1, pp. 41–70, May 2019, doi: 10.1007/s11277- technology: LoRa and NB-IoT,” ICT Express, vol. 3, no. 1, pp. 14–21,
019-06275-4. Mar. 2017, doi: 10.1016/j.icte.2017.03.004.
[7] A. Habib and S. Moh, “Wireless Channel Models for Over-the-Sea [18] K. Mekki, E. Bajic, F. Chaxel, and F. Meyer, “A comparative study of
Communication: A Comparative Study,” Appl. Sci., vol. 9, no. 3, p. LPWAN technologies for large-scale IoT deployment,” ICT Express,
443, Jan. 2019, doi: 10.3390/app9030443. vol. 5, no. 1, pp. 1–7, Mar. 2019, doi: 10.1016/j.icte.2017.12.005.
[8] W. Khawaja, I. Guvenc, D. W. Matolak, U.-C. Fiebig, N. [19] Y. Moon, S. Ha, M. Park, D. Lee, and J. Jeong, “A methodology of NB-
Schneckenberger, and N. Schneckenburger, “A Survey of Air-to- IoT mobility optimization,” in 2018 Global Internet of Things Summit
Ground Propagation Channel Modeling for Unmanned Aerial (GIoTS), 2018, pp. 1–5.
Vehicles,” IEEE Commun. Surv. Tutorials, vol. 21, no. 3, pp. 2361– [20] L. Wan, Z. Zhang, Y. Huang, Y. Yan, and J. Wang, “Performance
2391, 2019, doi: 10.1109/COMST.2019.2915069. Analysis of NB-IoT Technology for Indoor IoT Applications,” in 2017
[9] J. C. Silva, G. L. Siqueira, and P. V. G. Castellanos, “Propagation International Conference on Computer Technology, Electronics and
Model for Path Loss Through Vegetated Environments at 700 – 800 Communication (ICCTEC), Dec. 2017, pp. 1365–1369, doi:
MHz Band,” J. Microwaves, Optoelectron. Electromagn. Appl., vol. 10.1109/ICCTEC.2017.00299.
17, no. 1, pp. 102–120, 2018, doi: 10.1590/2179-10742018v17i11183. [21] H. A. H. Alobaidy, J. S. Mandeep, R. Nordin, N. F. Abdullah, C. G.
[10] L. E. C. Eras, D. K. N. Silva, F. B. Barros, L. M. Correia, and G. P. S. Wei, and M. L. S. Soon, “Real-World Evaluation of Power
Cavalcante, “A Radio Propagation Model for Mixed Paths in Amazon Consumption and Performance of NB-IoT in Malaysia,” IEEE Internet
Environments for the UHF Band,” vol. 2018, 2018. Things J., pp. 1–1, 2021, doi: 10.1109/JIOT.2021.3131160.
[11] A. Al-Hourani, S. Kandeepan, and A. Jamalipour, “Modeling air-to-
34 VOLUME XX, 2021
[22] G. Pasolini et al., “Smart city pilot projects using LoRa and Technologies,” C. X. Mavromoustakis, G. Mastorakis, and J. M.
IEEE802.15.4 technologies,” Sensors (Switzerland), vol. 18, no. 4, pp. Batalla, Eds. Cham: Springer International Publishing, 2016, pp. 3–24.
1–17, 2018, doi: 10.3390/s18041118. [37] L. M. Vela, H. Kwon, S. B. Rutkove, and B. Sanchez, “Standalone IoT
[23] H. A. H. Alobaidy, J. S. Mandeep, R. Nordin, and N. F. Abdullah, “A Bioimpedance Device Supporting Real-Time Online Data Access,”
Review on ZigBee Based WSNs: Concepts, Infrastructure, Applications, IEEE Internet Things J., vol. 6, no. 6, pp. 9545–9554, Dec. 2019, doi:
and Challenges,” Int. J. Electr. Electron. Eng. Telecommun., vol. 9, no. 3, 10.1109/JIOT.2019.2929459.
pp. 189–198, doi: 10.18178/ijeetc.9.3.189-198. [38] G. Xu, “IoT-Assisted ECG Monitoring Framework With Secure Data
[24] T. Kim, C. Ramos, and S. Mohammed, “Smart City and IoT,” Futur. Transmission for Health Care Applications,” IEEE Access, vol. 8, pp.
Gener. Comput. Syst., vol. 76, pp. 159–162, Nov. 2017, doi: 74586–74594, 2020, doi: 10.1109/ACCESS.2020.2988059.
10.1016/j.future.2017.03.034. [39] M. Cicioglu and A. Calhan, “Internet of Things based Firefighters for
[25] S. Bertoldo, L. Carosso, E. Marchetta, M. Paredes, and M. Allegretti, Disaster Case Management,” IEEE Sens. J., pp. 1–1, 2020, doi:
“Feasibility Analysis of a LoRa-Based WSN Using Public Transport,” 10.1109/JSEN.2020.3013333.
Appl. Syst. Innov., vol. 1, no. 4, p. 49, 2018, doi: 10.3390/asi1040049. [40] P. A. A. Catherwood, S. Mccomb, M. Little, and J. A. D. A. D.
[26] J. Petajajarvi, K. Mikhaylov, A. Roivainen, T. Hanninen, and M. McLaughlin, “Channel Characterisation for Wearable LoRaWAN
Pettissalo, “On the coverage of LPWANs: range evaluation and channel Monitors,” Loughbrgh. Antennas Propag. Conf. (LAPC 2017), pp. 6–
attenuation model for LoRa technology,” in 2015 14th International 9, 2017, doi: 10.1049/cp.2017.0273.
Conference on ITS Telecommunications (ITST), Dec. 2015, pp. 55–59, [41] H. Tan and I. Chung, “Secure Authentication and Group Key
doi: 10.1109/ITST.2015.7377400. Distribution Scheme for WBANs Based on Smartphone ECG Sensor,”
[27] M. Sidorov, P. V. Nhut, Y. Matsumoto, and R. Ohmura, “LoRa-Based IEEE Access, vol. 7, pp. 151459–151474, 2019, doi:
Precision Wireless Structural Health Monitoring System for Bolted 10.1109/ACCESS.2019.2948207.
Joints in a Smart City Environment,” IEEE Access, vol. 7, pp. 179235– [42] P. Vijayakumar, M. S. Obaidat, M. Azees, S. H. Islam, and N. Kumar,
179251, 2019, doi: 10.1109/ACCESS.2019.2958835. “Efficient and Secure Anonymous Authentication With Location
[28] W. Xu, J. Y. Kim, W. Huang, S. S. Kanhere, S. K. Jha, and W. Hu, Privacy for IoT-Based WBANs,” IEEE Trans. Ind. Informatics, vol. 16,
“Measurement, Characterization, and Modeling of LoRa Technology no. 4, pp. 2603–2611, Apr. 2020, doi: 10.1109/TII.2019.2925071.
in Multifloor Buildings,” IEEE Internet Things J., vol. 7, no. 1, pp. 298– [43] Z. Munadhil, S. K. Gharghan, A. H. Mutlag, A. Al-Naji, and J. Chahl,
310, Jan. 2020, doi: 10.1109/JIOT.2019.2946900. “Neural Network-Based Alzheimer’s Patient Localization for Wireless
[29] A. Heller, “The Sensing Internet—A Discussion on Its Impact on Rural Sensor Network in an Indoor Environment,” IEEE Access, vol. 8, pp.
Areas,” Futur. Internet, vol. 7, no. 4, pp. 363–371, Sep. 2015, doi: 150527–150538, 2020, doi: 10.1109/ACCESS.2020.3016832.
10.3390/fi7040363. [44] T. O. Olasupo, “Wireless Communication Modeling for the
[30] A. Alsayyari, I. Kostanic, and C. E. Otero, “An empirical path loss Deployment of Tiny IoT Devices in Rocky and Mountainous
model for Wireless Sensor Network deployment in a concrete surface Environments,” IEEE Sensors Lett., vol. 3, no. 7, pp. 1–4, Jul. 2019,
environment,” in 2015 IEEE 16th Annual Wireless and Microwave doi: 10.1109/LSENS.2019.2918331.
Technology Conference (WAMICON), Apr. 2015, pp. 1–6, doi: [45] H. Xu, W. Yu, D. Griffith, and N. Golmie, “A Survey on Industrial
10.1109/WAMICON.2015.7120311. Internet of Things: A Cyber-Physical Systems Perspective,” IEEE Access,
[31] N. S. Bezerra, C. Åhlund, S. Saguna, and V. A. de S. Jr., “Temperature vol. 6, pp. 78238–78259, 2018, doi: 10.1109/ACCESS.2018.2884906.
Impact in LoRaWAN—A Case Study in Northern Sweden,” Sensors, [46] M. W. Condry and C. B. Nelson, “Using Smart Edge IoT Devices for
vol. 19, no. 20, p. 4414, Oct. 2019, doi: 10.3390/s19204414. Safer, Rapid Response With Industry IoT Control Operations,” Proc.
[32] J. Guo, “Smartphone-Powered Electrochemical Biosensing Dongle for IEEE, vol. 104, no. 5, pp. 938–946, May 2016, doi:
Emerging Medical IoTs Application,” IEEE Trans. Ind. Informatics, 10.1109/JPROC.2015.2513672.
vol. 14, no. 6, pp. 2592–2597, Jun. 2018, doi: [47] J. Huang, L. Kong, G. Chen, M.-Y. Wu, X. Liu, and P. Zeng, “Towards
10.1109/TII.2017.2777145. Secure Industrial IoT: Blockchain System With Credit-Based
[33] A. H. Mohd Aman, W. H. Hassan, S. Sameen, Z. S. Attarbashi, M. Consensus Mechanism,” IEEE Trans. Ind. Informatics, vol. 15, no. 6,
Alizadeh, and L. A. Latiff, “IoMT amid COVID-19 pandemic: pp. 3680–3689, Jun. 2019, doi: 10.1109/TII.2019.2903342.
Application, architecture, technology, and security,” J. Netw. Comput. [48] R. Ajith, A. Tewari, D. Gupta, and S. Tallur, “Low-Cost Vibration
Appl., vol. 174, p. 102886, Jan. 2021, doi: 10.1016/j.jnca.2020.102886. Sensor for Condition-Based Monitoring Manufactured From
[34] N. Naren et al., “IoMT and DNN-Enabled Drone-Assisted Covid-19 Polyurethane Foam,” IEEE Sensors Lett., vol. 1, no. 6, pp. 1–4, Dec.
Screening and Detection Framework for Rural Areas,” IEEE Internet 2017, doi: 10.1109/LSENS.2017.2773652.
Things Mag., vol. 4, no. 2, pp. 4–9, Jun. 2021, doi: [49] T. Nguyen, R. G. Gosine, and P. Warrian, “A Systematic Review of
10.1109/IOTM.0011.2100053. Big Data Analytics for Oil and Gas Industry 4.0,” IEEE Access, vol. 8,
[35] V. Chamola, V. Hassija, V. Gupta, and M. Guizani, “A Comprehensive pp. 61183–61201, 2020, doi: 10.1109/ACCESS.2020.2979678.
Review of the COVID-19 Pandemic and the Role of IoT, Drones, AI, [50] R. Morello, C. De Capua, G. Fulco, and S. C. Mukhopadhyay, “A
Blockchain, and 5G in Managing its Impact,” IEEE Access, vol. 8, pp. Smart Power Meter to Monitor Energy Flow in Smart Grids: The Role
90225–90265, 2020, doi: 10.1109/ACCESS.2020.2992341. of Advanced Sensing and IoT in the Electric Grid of the Future,” IEEE
[36] B. Mathieu, C. Westphal, and P. Truong, “Towards the Usage of CCN Sens. J., vol. 17, no. 23, pp. 7828–7837, Dec. 2017, doi:
for IoT Networks BT - Internet of Things (IoT) in 5G Mobile 10.1109/JSEN.2017.2760014.
VOLUME XX, 2017 35
[51] S. Chen et al., “Internet of Things Based Smart Grids Supported by Networks,” IEEE Access, vol. 6, pp. 25686–25697, 2018, doi:
Intelligent Edge Computing,” IEEE Access, vol. 7, pp. 74089–74102, 10.1109/ACCESS.2018.2836185.
2019, doi: 10.1109/ACCESS.2019.2920488. [66] R. Khan, M. Zakarya, V. Balasubramaniam, M. A. Jan, and V. G.
[52] R. Sanchez-Iborra, J. Sanchez-Gomez, J. Ballesta-Viñas, M.-D. Cano, Menon, “Smart Sensing-enabled Decision Support System for Water
and A. Skarmeta, “Performance Evaluation of LoRa Considering Scheduling in Orange Orchard,” IEEE Sens. J., pp. 1–1, 2020, doi:
Scenario Conditions,” Sensors, vol. 18, no. 3, p. 772, Mar. 2018, doi: 10.1109/JSEN.2020.3012511.
10.3390/s18030772. [67] A. Goap, D. Sharma, A. K. Shukla, and C. Rama Krishna, “An IoT
[53] S. Proto et al., “REDTag: A Predictive Maintenance Framework for based smart irrigation management system using Machine learning and
Parcel Delivery Services,” IEEE Access, vol. 8, pp. 14953–14964, open source technologies,” Comput. Electron. Agric., vol. 155, pp. 41–
2020, doi: 10.1109/ACCESS.2020.2966568. 49, Dec. 2018, doi: 10.1016/j.compag.2018.09.040.
[54] J. Hejselbaek, J. Odum Nielsen, W. Fan, and G. F. Pedersen, “Empirical [68] B. Keswani et al., “Adapting weather conditions based IoT enabled
Study of Near Ground Propagation in Forest Terrain for Internet-of- smart irrigation technique in precision agriculture mechanisms,”
Things Type Device-to-Device Communication,” IEEE Access, vol. 6, Neural Comput. Appl., vol. 31, no. S1, pp. 277–292, Jan. 2019, doi:
pp. 54052–54063, 2018, doi: 10.1109/ACCESS.2018.2871368. 10.1007/s00521-018-3737-1.
[55] Z. Huang, Q. Chen, L. Zhang, and X. Hu, “Research on Intelligent [69] T. Natori, N. Ariyama, S. Tsuichihara, H. Takemura, and N. Aikawa,
Monitoring and Analysis of Physical Fitness Based on the Internet of “Study of Activity Collecting System for Grazing Cattle,” in 2019 34th
Things,” IEEE Access, vol. 7, pp. 177297–177308, 2019, doi: International Technical Conference on Circuits/Systems, Computers
10.1109/ACCESS.2019.2956835. and Communications (ITC-CSCC), Jun. 2019, pp. 1–4, doi:
[56] P. Silva Girão, O. Postolache, and J. D. Pereira, “Wireless sensor 10.1109/ITC-CSCC.2019.8793451.
network-based solution for environmental monitoring: water quality [70] T. Olasupo, C. E. Otero, K. O. Olasupo, and I. Kostanic, “Empirical
assessment case study,” IET Sci. Meas. Technol., vol. 8, no. 6, pp. 610– Path Loss Models for Wireless Sensor Network Deployments in Short
616, Nov. 2014, doi: 10.1049/iet-smt.2013.0136. and Tall Natural Grass Environments,” IEEE Trans. Antennas Propag.,
[57] S. N. H. Al-Hussaini, A. H. M. J. Al-Obaidy, and A. A. M. Al- pp. 1–1, 2016, doi: 10.1109/TAP.2016.2583507.
Mashhady, “Environmental assessment of heavy metal pollution of [71] Z. Gao et al., “Wireless Channel Propagation Characteristics and
Diyala River within Baghdad City,” Appl. Water Sci., vol. 8, no. 3, p. Modeling Research in Rice Field Sensor Networks,” Sensors, vol. 18,
87, Jun. 2018, doi: 10.1007/s13201-018-0707-9. no. 9, p. 3116, Sep. 2018, doi: 10.3390/s18093116.
[58] I. M. Jasim, A. A. Al-kubaisi, and A. H. M. J. Al-Obaidy, “Test the [72] P. Loreti, A. Catini, M. De Luca, L. Bracciale, G. Gentile, and C. Di
Efficiency of some Plants in the Tolerant of Air Pollution within the Natale, “The Design of an Energy Harvesting Wireless Sensor Node for
City of Baghdad. Iraq,” Baghdad Sci. J., vol. 15, no. 1, pp. 9–15, Mar. Tracking Pink Iguanas,” Sensors, vol. 19, no. 5, p. 985, Feb. 2019, doi:
2018, doi: 10.21123/bsj.2018.15.1.0009. 10.3390/s19050985.
[59] Y. Wang, S. M. S. M. Rajib, C. Collins, and B. Grieve, “Low-Cost [73] A. Naureen, N. Zhang, S. Furber, and Q. Shi, “A GPS-Less
Turbidity Sensor for Low-Power Wireless Monitoring of Fresh-Water Localization and Mobility Modelling (LMM) System for Wildlife
Courses,” IEEE Sens. J., vol. 18, no. 11, pp. 4689–4696, Jun. 2018, doi: Tracking,” IEEE Access, vol. 8, pp. 102709–102732, 2020, doi:
10.1109/JSEN.2018.2826778. 10.1109/ACCESS.2020.2997723.
[60] A. Salam and S. Shah, “Urban Underground Infrastructure Monitoring [74] S. Benaissa et al., “Internet of animals: characterisation of LoRa sub-
IoT: The Path Loss Analysis,” in 2019 IEEE 5th World Forum on GHz off-body wireless channel in dairy barns,” Electron. Lett., vol. 53,
Internet of Things (WF-IoT), Apr. 2019, pp. 398–401, doi: no. 18, pp. 1281–1283, Aug. 2017, doi: 10.1049/el.2017.1344.
10.1109/WF-IoT.2019.8767358. [75] S. Benaissa et al., “Experimental characterisation of in-to-out-body
[61] M. Munoz, J. L. Guzman, J. A. Sanchez, F. Rodriguez, M. Torres, and path loss at 433 MHz in dairy cows,” Electron. Lett., vol. 55, no. 7, pp.
M. Berenguel, “A New IoT-based Platform for Greenhouse Crop 422–424, Apr. 2019, doi: 10.1049/el.2018.8150.
Production,” IEEE Internet Things J., pp. 1–1, 2020, doi: [76] W. Hong, B. Xu, X. Chi, X. Cui, Y. Yan, and T. Li, “Long-term and
10.1109/JIOT.2020.2996081. Extensive Monitoring for Bee Colonies based on Internet of Things,”
[62] J. Stewart, R. Stewart, and S. Kennedy, “Internet of Things— IEEE Internet Things J., pp. 1–1, 2020, doi: 10.1109/JIOT.2020.2981681.
Propagation modelling for precision agriculture applications,” in 2017 [77] H. Bello, Z. Xiaoping, R. Nordin, and J. Xin, “Advances and
Wireless Telecommunications Symposium (WTS), 2017, pp. 1–8. Opportunities in Passive Wake-Up Radios with Wireless Energy
[63] C. A. Gonzalez-Amarillo et al., “An IoT-Based Traceability System for Harvesting for the Internet of Things Applications,” Sensors, vol. 19,
Greenhouse Seedling Crops,” IEEE Access, vol. 6, pp. 67528–67535, no. 14, p. 3078, Jul. 2019, doi: 10.3390/s19143078.
2018, doi: 10.1109/ACCESS.2018.2877293. [78] M. Mozaffari, W. Saad, M. Bennis, and M. Debbah, “Mobile
[64] R. Godoi Vieira, A. M. da Cunha, L. B. Ruiz, and A. P. de Camargo, Unmanned Aerial Vehicles (UAVs) for Energy-Efficient Internet of
“On the Design of a Long Range WSN for Precision Irrigation,” IEEE Things Communications,” IEEE Trans. Wirel. Commun., vol. 16, no.
Sens. J., vol. 18, no. 2, pp. 773–780, Jan. 2018, doi: 11, pp. 7574–7589, Nov. 2017, doi: 10.1109/TWC.2017.2751045.
10.1109/JSEN.2017.2776859. [79] M. A. Zulkifley, M. Behjati, R. Nordin, and M. S. Zakaria, “Mobile
[65] R. Khan, I. Ali, M. Zakarya, M. Ahmad, M. Imran, and M. Shoaib, Network Performance and Technical Feasibility of LTE-Powered
“Technology-Assisted Decision Support System for Efficient Water Unmanned Aerial Vehicle,” Sensors, vol. 21, no. 8, p. 2848, Apr. 2021,
Utilization: A Real-Time Testbed for Irrigation Using Wireless Sensor doi: 10.3390/s21082848.
36 VOLUME XX, 2021
[80] M. Behjati, A. B. Mohd Noh, H. A. H. Alobaidy, M. A. Zulkifley, R. 2019, doi: 10.1109/JIOT.2018.2875493.

Nordin, and N. F. Abdullah, “LoRa Communications as an Enabler for [95] I. Ullah, Y. Liu, X. Su, and P. Kim, “Efficient and Accurate Target
Internet of Drones towards Large-Scale Livestock Monitoring in Rural Localization in Underwater Environment,” IEEE Access, vol. 7, pp.
Farms,” Sensors, vol. 21, no. 15, p. 5044, Jul. 2021, doi: 101415–101426, 2019, doi: 10.1109/ACCESS.2019.2930735.
10.3390/s21155044. [96] V. M. Suresh, R. Sidhu, P. Karkare, A. Patil, Z. Lei, and A. Basu,
[81] C. Luo, J. Nightingale, E. Asemota, and C. Grecos, “A UAV-Cloud “Powering the IoT through embedded machine learning and LoRa,” in
System for Disaster Sensing Applications,” in 2015 IEEE 81st 2018 IEEE 4th World Forum on Internet of Things (WF-IoT), Feb.
Vehicular Technology Conference (VTC Spring), May 2015, pp. 1–5, 2018, pp. 349–354, doi: 10.1109/WF-IoT.2018.8355177.
doi: 10.1109/VTCSpring.2015.7145656. [97] M. Molefi, E. D. Markus, and A. Abu-Mahfouz, “Wireless Power
[82] S. J. Kim, Y. Jeong, S. Park, K. Ryu, and G. Oh, “A Survey of Drone Transfer for IoT Devices - A Review,” in 2019 International
use for Entertainment and AVR (Augmented and Virtual Reality),” Multidisciplinary Information Technology and Engineering
2018, pp. 339–352. Conference (IMITEC), Nov. 2019, pp. 1–8, doi:
[83] D. C. Tsouros, S. Bibi, and P. G. Sarigiannidis, “A Review on UAV- 10.1109/IMITEC45504.2019.9015869.
Based Applications for Precision Agriculture,” Information, vol. 10, no. [98] J. Corral-García, F. Lemus-Prieto, and J.-L. G.-S. and M.-Á. Pérez-
11, p. 349, Nov. 2019, doi: 10.3390/info10110349. Toledano, “Analysis of Energy Consumption and Optimization
[84] Y. Zhang, X. Yuan, W. Li, and S. Chen, “Automatic Power Line Techniques for Writing Energy-Efficient Code,” Electronics, vol. 8, no.
Inspection Using UAV Images,” Remote Sens., vol. 9, no. 8, p. 824, 10, p. 1192, Oct. 2019, doi: 10.3390/electronics8101192.
Aug. 2017, doi: 10.3390/rs9080824. [99] F. John Dian, R. Vahidnia, and A. Rahmati, “Wearables and the Internet
[85] H. Kim, L. Mokdad, and J. Ben-Othman, “Designing UAV of Things (IoT), Applications, Opportunities, and Challenges: A
Surveillance Frameworks for Smart City and Extensive Ocean with Survey,” IEEE Access, vol. 8, pp. 69200–69211, 2020, doi:
Differential Perspectives,” IEEE Commun. Mag., vol. 56, no. 4, pp. 98– 10.1109/ACCESS.2020.2986329.
104, Apr. 2018, doi: 10.1109/MCOM.2018.1700444. [100] F. Maroto-Molina et al., “A Low-Cost IoT-Based System to Monitor
[86] S. Sudhakar, V. Vijayakumar, C. Sathiya Kumar, V. Priya, L. Ravi, and the Location of a Whole Herd,” Sensors, vol. 19, no. 10, p. 2298, May
V. Subramaniyaswamy, “Unmanned Aerial Vehicle (UAV) based 2019, doi: 10.3390/s19102298.
Forest Fire Detection and monitoring for reducing false alarms in [101] F. Al-Turjman, “Positioning in the Internet of Things Era: An overview,”
forest-fires,” Comput. Commun., vol. 149, pp. 1–16, Jan. 2020, doi: in 2017 International Conference on Engineering and Technology
10.1016/j.comcom.2019.10.007. (ICET), Aug. 2017, pp. 1–5, doi: 10.1109/ICEngTechnol.2017.8308188.
[87] M. I. Ahmad, M. H. Ab. Rahim, R. Nordin, F. Mohamed, A. Abu- [102] H. Jawad, R. Nordin, S. Gharghan, A. Jawad, M. Ismail, and M. Abu-
Samah, and N. F. Abdullah, “Ionizing Radiation Monitoring AlShaeer, “Power Reduction with Sleep/Wake on Redundant Data
Technology at the Verge of Internet of Things,” Sensors, vol. 21, no. (SWORD) in a Wireless Sensor Network for Energy-Efficient
22, p. 7629, Nov. 2021, doi: 10.3390/s21227629. Precision Agriculture,” Sensors, vol. 18, no. 10, p. 3450, Oct. 2018, doi:
[88] A. Gupta, T. Afrin, E. Scully, and N. Yodo, “Advances of UAVs 10.3390/s18103450.
toward Future Transportation: The State-of-the-Art, Challenges, and [103] M. Singh and G. Baranwal, “Quality of Service (QoS) in Internet of
Opportunities,” Futur. Transp., vol. 1, no. 2, pp. 326–350, Sep. 2021, Things,” in 2018 3rd International Conference On Internet of Things:
doi: 10.3390/futuretransp1020019. Smart Innovation and Usages (IoT-SIU), Feb. 2018, pp. 1–6, doi:
[89] F. Schiefer et al., “Mapping forest tree species in high resolution UAV- 10.1109/IoT-SIU.2018.8519862.
based RGB-imagery by means of convolutional neural networks,” [104] H. Klaina, A. Vazquez Alejos, O. Aghzout, and F. Falcone,
ISPRS J. Photogramm. Remote Sens., vol. 170, pp. 205–215, Dec. “Narrowband Characterization of Near-Ground Radio Channel for
2020, doi: 10.1016/j.isprsjprs.2020.10.015. Wireless Sensors Networks at 5G-IoT Bands,” Sensors, vol. 18, no. 8,
[90] S. Zeadally, J. Guerrero, and J. Contreras, “A tutorial survey on vehicle- p. 2428, Jul. 2018, doi: 10.3390/s18082428.
to-vehicle communications,” Telecommun. Syst., vol. 73, no. 3, pp. [105] R. Pita, R. Utrilla, R. Rodriguez-Zurrunero, and A. Araujo,
469–489, Mar. 2020, doi: 10.1007/s11235-019-00639-8. “Experimental Evaluation of an RSSI-Based Localization Algorithm
[91] F. Lyu, M. Li, and X. Shen, “Characterizing Urban V2V Link on IoT End-Devices,” Sensors, vol. 19, no. 18, p. 3931, Sep. 2019, doi:
Communications,” 2020, pp. 77–99. 10.3390/s19183931.
[92] H. A. Ameen, A. K. Mahamad, S. Saon, D. M. Nor, and K. Ghazi, “A [106] W. Njima, I. Ahriz, R. Zayani, M. Terre, and R. Bouallegue, “Deep
review on vehicle to vehicle communication system applications,” CNN for Indoor Localization in IoT-Sensor Systems,” Sensors, vol. 19,
Indones. J. Electr. Eng. Comput. Sci., vol. 18, no. 1, p. 188, Apr. 2020, no. 14, p. 3127, Jul. 2019, doi: 10.3390/s19143127.
doi: 10.11591/ijeecs.v18.i1.pp188-198. [107] Y. Bae, “Robust Localization for Robot and IoT Using RSSI,”
[93] A. Azmoodeh, A. Dehghantanha, and K.-K. R. Choo, “Robust Malware Energies, vol. 12, no. 11, p. 2212, Jun. 2019, doi: 10.3390/en12112212.
Detection for Internet of (Battlefield) Things Devices Using Deep [108] J. Rezazadeh, K. Sandrasegaran, and X. Kong, “A location-based smart
Eigenspace Learning,” IEEE Trans. Sustain. Comput., vol. 4, no. 1, pp. shopping system with IoT technology,” in 2018 IEEE 4th World Forum
88–95, Jan. 2019, doi: 10.1109/TSUSC.2018.2809665. on Internet of Things (WF-IoT), Feb. 2018, pp. 748–753, doi:
[94] J. Wang, C. Jiang, Z. Wei, C. Pan, H. Zhang, and Y. Ren, “Joint UAV 10.1109/WF-IoT.2018.8355175.
Hovering Altitude and Power Control for Space-Air-Ground IoT [109] S. Sadowski and P. Spachos, “RSSI-Based Indoor Localization With
Networks,” IEEE Internet Things J., vol. 6, no. 2, pp. 1741–1753, Apr. the Internet of Things,” IEEE Access, vol. 6, pp. 30149–30161, 2018,
VOLUME XX, 2017 37
doi: 10.1109/ACCESS.2018.2843325. of Path Loss Models,” Int. J. Antennas Propag., vol. 2012, pp. 1–21,
[110] H. Sallouha, A. Chiumento, S. Rajendran, and S. Pollin, “Localization 2012, doi: 10.1155/2012/754158.
in Ultra Narrow Band IoT Networks: Design Guidelines and [124] J. Zhao, Q. Wang, Y. Li, J. Zhou, and W. Zhou, “Ka-band Based
Tradeoffs,” IEEE Internet Things J., vol. 6, no. 6, pp. 9375–9385, Dec. Channel Modeling and Analysis in High Altitude Platform(HAP)
2019, doi: 10.1109/JIOT.2019.2931628. System,” in 2020 IEEE 91st Vehicular Technology Conference
[111] G. Wang, W. Zhu, and N. Ansari, “Robust TDOA-Based Localization (VTC2020-Spring), May 2020, pp. 1–5, doi: 10.1109/VTC2020-
for IoT via Joint Source Position and NLOS Error Estimation,” IEEE Spring48590.2020.9129385.
Internet Things J., vol. 6, no. 5, pp. 8529–8541, Oct. 2019, doi: [125] F. Qamar et al., “Investigation of Future 5G-IoT Millimeter-Wave
10.1109/JIOT.2019.2920081. Network Performance at 38 GHz for Urban Microcell Outdoor
[112] L. Zhao, C. Su, H. Huang, Z. Han, S. Ding, and X. Li, “Intrusion Environment,” Electronics, vol. 8, no. 5, p. 495, May 2019, doi:
Detection Based on Device-Free Localization in the Era of IoT,” 10.3390/electronics8050495.
Symmetry (Basel)., vol. 11, no. 5, p. 630, May 2019, doi: [126] S. Demetri, M. Zúñiga, G. Pietro Picco, F. Kuipers, L. Bruzzone, and
10.3390/sym11050630. T. Telkamp, “Automated estimation of link quality for LoRa,” in
[113] F. Zafari, A. Gkelias, and K. K. Leung, “A Survey of Indoor Proceedings of the 18th International Conference on Information
Localization Systems and Technologies,” IEEE Commun. Surv. Processing in Sensor Networks, Apr. 2019, pp. 145–156, doi:
Tutorials, vol. 21, no. 3, pp. 2568–2599, 2019, doi: 10.1145/3302506.3310396.
10.1109/COMST.2019.2911558. [127] N. S. Bezerra, C. Ahlund, S. Saguna, and V. A. de Sousa, “Propagation
[114] S. K. Gharghan, R. Nordin, M. Ismail, and J. A. Ali, “Accurate Wireless Model Evaluation for LoRaWAN: Planning Tool Versus Real Case
Sensor Localization Technique Based on Hybrid PSO-ANN Algorithm Scenario,” in 2019 IEEE 5th World Forum on Internet of Things (WF-
for Indoor and Outdoor Track Cycling,” IEEE Sens. J., vol. 16, no. 2, IoT), Apr. 2019, pp. 1–6, doi: 10.1109/WF-IoT.2019.8767299.
pp. 529–541, Jan. 2016, doi: 10.1109/JSEN.2015.2483745. [128] K. M. Malarski et al., “Investigation of Deep Indoor NB-IoT
[115] S. K. Gharghan, R. Nordin, A. M. Jawad, H. M. Jawad, and M. Ismail, Propagation Attenuation,” in 2019 IEEE 90th Vehicular Technology
“Adaptive Neural Fuzzy Inference System for Accurate Localization of Conference (VTC2019-Fall), Sep. 2019, pp. 1–5, doi:
Wireless Sensor Network in Outdoor and Indoor Cycling 10.1109/VTCFall.2019.8891414.
Applications,” IEEE Access, vol. 6, pp. 38475–38489, 2018, doi: [129] S. Kasampalis, P. I. Lazaridis, Z. D. Zaharis, A. Bizopoulos, S. Zettas,
10.1109/ACCESS.2018.2853996. and J. Cosmas, “Comparison of ITM and ITWOM propagation models
[116] Y. Jiang, L. Peng, A. Hu, S. Wang, Y. Huang, and L. Zhang, “Physical for DVB-T coverage prediction,” in 2013 IEEE International
layer identification of LoRa devices using constellation trace figure,” Symposium on Broadband Multimedia Systems and Broadcasting
EURASIP J. Wirel. Commun. Netw., vol. 2019, no. 1, p. 223, Dec. 2019, (BMSB), Jun. 2013, pp. 1–4, doi: 10.1109/BMSB.2013.6621780.
doi: 10.1186/s13638-019-1542-x. [130] F. A. Almalki and M. C. Angelides, “A machine learning approach to
[117] R. M. Sandoval, A.-J. Garcia-Sanchez, and J. Garcia-Haro, “Improving evolving an optimal propagation model for last mile connectivity using
RSSI-Based Path-Loss Models Accuracy for Critical Infrastructures: A low altitude platforms,” Comput. Commun., vol. 142–143, pp. 9–33,
Smart Grid Substation Case-Study,” IEEE Trans. Ind. Informatics, vol. Jun. 2019, doi: 10.1016/j.comcom.2019.04.001.
14, no. 5, pp. 2230–2240, May 2018, doi: 10.1109/TII.2017.2774838. [131] T. Jawhly and R. C. Tiwari, “Characterization of Path Loss for VHF
[118] R. El Chall, S. Lahoud, and M. El Helou, “LoRaWAN Network: Radio Terrestrial Band in Aizawl, Mizoram (India),” in Engineering
Propagation Models and Performance Evaluation in Various Vibration, Communication and Information Processing, K. Ray, S. N.
Environments in Lebanon,” IEEE Internet Things J., vol. 6, no. 2, pp. Sharan, S. Rawat, S. K. Jain, S. Srivastava, and A. Bandyopadhyay,
2366–2378, Apr. 2019, doi: 10.1109/JIOT.2019.2906838. Eds. Singapore: Springer, Singapore, 2019, pp. 53–63.
[119] A. Alsayyari, I. Kostanic, C. Otero, M. Almeer, and K. Rukieh, “An [132] A. M. Al-Samman et al., “Comparative Study of Indoor Propagation
empirical path loss model for wireless sensor network deployment in a Model Below and Above 6 GHz for 5G Wireless Networks,”
sand terrain environment,” in 2014 IEEE World Forum on Internet of Electronics, vol. 8, no. 1, p. 44, Jan. 2019, doi:
Things (WF-IoT), Mar. 2014, pp. 218–223, doi: 10.1109/WF- 10.3390/electronics8010044.
IoT.2014.6803162. [133] S. Sun, G. R. MacCartney, and T. S. Rappaport, “Millimeter-wave
[120] C. Phillips, D. Sicker, and D. Grunwald, “A Survey of Wireless Path distance-dependent large-scale propagation measurements and path
Loss Prediction and Coverage Mapping Methods,” IEEE Commun. loss models for outdoor and indoor 5G systems,” in 2016 10th
Surv. Tutorials, vol. 15, no. 1, pp. 255–270, 2013, doi: European Conference on Antennas and Propagation (EuCAP), Apr.
10.1109/SURV.2012.022412.00172. 2016, pp. 1–5, doi: 10.1109/EuCAP.2016.7481506.
[121] K. S. Muttair, O. A. Shareef, and M. F. Mosleh, “Outdoor to Indoor [134] A. Al-Samman, T. Rahman, M. Hindia, A. Daho, and E. Hanafi, “Path
Wireless Propagation Simulation Model for 5G Band Frequencies,” Loss Model for Outdoor Parking Environments at 28 GHz and 38 GHz
IOP Conf. Ser. Mater. Sci. Eng., vol. 745, p. 012034, Mar. 2020, doi: for 5G Wireless Networks,” Symmetry (Basel)., vol. 10, no. 12, p. 672,
10.1088/1757-899X/745/1/012034. Nov. 2018, doi: 10.3390/sym10120672.
[122] V. Lavanya, G. S. Rao, and B. Bidikar, “Fast Fading Mobile Channel [135] S. Sun et al., “Propagation Path Loss Models for 5G Urban Micro- and
Modeling for Wireless Communication,” Procedia Comput. Sci., vol. Macro-Cellular Scenarios,” in 2016 IEEE 83rd Vehicular Technology
85, pp. 777–781, 2016, doi: 10.1016/j.procs.2016.05.265. Conference (VTC Spring), May 2016, pp. 1–6, doi:
[123] C. Phillips, D. Sicker, and D. Grunwald, “Bounding the Practical Error 10.1109/VTCSpring.2016.7504435.
38 VOLUME XX, 2021
[136] G. R. MacCartney and T. S. Rappaport, “Rural Macrocell Path Loss Communication Networks,” PLoS One, vol. 11, no. 9, p. e0163034,
Models for Millimeter Wave Wireless Communications,” IEEE J. Sel. Sep. 2016, doi: 10.1371/journal.pone.0163034.
Areas Commun., vol. 35, no. 7, pp. 1663–1677, Jul. 2017, doi: [149] G. Castellanos and G. Teuta, “Path loss model in amazonian border
10.1109/JSAC.2017.2699359. region for VHF and UHF television bands,” 2017 IEEE-APS Top. Conf.
[137] G. R. MacCartney and T. S. Rappaport, “Study on 3GPP rural Antennas Propag. Wirel. Commun. APWC 2017, vol. 2017-Janua, pp.
macrocell path loss models for millimeter wave wireless 137–140, 2017, doi: 10.1109/APWC.2017.8062262.
communications,” in 2017 IEEE International Conference on [150] A. V., S. K.M.S., M. N., and F. T., “Propagation Models and Their
Communications (ICC), May 2017, pp. 1–7, doi: Applications in Digital Television Broadcast Network Design and
10.1109/ICC.2017.7996793. Implementation,” in Trends in Telecommunications Technologies,
[138] H. M. Rahim, C. Y. Leow, and T. A. Rahman, “Millimeter wave InTech, 2010.
propagation through foliage: Comparison of models,” in 2015 IEEE [151] J. J. Popoola, A. A. Ponnle, Y. O. Olasoji, and S. A. Oyetunji,
12th Malaysia International Conference on Communications (MICC), “INVESTIGATION ON NEED FOR SPECIFIC PROPAGATION
Nov. 2015, pp. 236–240, doi: 10.1109/MICC.2015.7725440. MODEL FOR SPECIFIC ENVIRONMENT BASED ON
[139] A. Raheemah, N. Sabri, M. S. Salim, P. Ehkan, and R. B. Ahmad, “New DIFFERENT TERRAIN CHARACTERISTICS,” IIUM Eng. J., vol.
empirical path loss model for wireless sensor networks in mango 19, no. 2, pp. 90–104, Dec. 2018, doi: 10.31436/iiumej.v19i2.886.
greenhouses,” Comput. Electron. Agric., vol. 127, pp. 553–560, Sep. [152] H. K. Hoomod, I. Al-Mejibli, and A. I. Jabboory, “Analyzing Study of
2016, doi: 10.1016/j.compag.2016.07.011. Path loss Propagation Models in Wireless Communications at 0.8
[140] J. Ko et al., “Measurements and Analysis of Radio Propagation at 28 GHz,” J. Phys. Conf. Ser., vol. 1003, p. 012028, May 2018, doi:
GHz in Vegetated Areas of Typical Residential Environments,” IEEE 10.1088/1742-6596/1003/1/012028.
Trans. Antennas Propag., vol. 68, no. 5, pp. 4149–4154, May 2020, [153] O. Artemenko, A. H. Nayak, S. B. Menezes, and A. Mitschele-Thiel,
doi: 10.1109/TAP.2020.2968801. “Evaluation of Different Signal Propagation Models for a Mixed
[141] ITU-R, “Recommendation ITU-R P.1238-10, Propagation data and Indoor-Outdoor Scenario Using Empirical Data,” in Ad Hoc Networks,
prediction methods for the planning of indoor radiocommunication N. Mitton, M. E. Kantarci, A. Gallais, and S. Papavassiliou, Eds.
systems and radio local area networks in the frequency range 300 MHz Springer, 2015, pp. 3–14.
to 450 GHz,” 2019. [154] D. Dobrilovic, M. Malic, D. Malic, and S. Sladojevic, “Analyses and
[142] A. Ben Zineb and M. Ayadi, “A Multi-wall and Multi-frequency Indoor optimization of Lee propagation model for LoRa 868 MHz network
Path Loss Prediction Model Using Artificial Neural Networks,” Arab. deployments in urban areas,” J. Eng. Manag. Compet., vol. 7, no. 1, pp.
J. Sci. Eng., vol. 41, no. 3, pp. 987–996, Mar. 2016, doi: 55–62, 2017, doi: 10.5937/jemc1701055D.
10.1007/s13369-015-1949-6. [155] M. A. Abu-Rgheff, “mmWave Propagation Modelling: Atmospheric
[143] M. Lott and I. Forkel, “A multi-wall-and-floor model for indoor radio Gaseous and Rain Losses,” in 5G Physical Layer Technologies, 1st ed.,
propagation,” in IEEE VTS 53rd Vehicular Technology Conference, JohnWiley & Sons Ltd, 2019, pp. 241–288.
Spring 2001. Proceedings (Cat. No.01CH37202), vol. 1, pp. 464–468, [156] A. I. Sulyman, A. Alwarafy, G. R. MacCartney, T. S. Rappaport, and
doi: 10.1109/VETECS.2001.944886. A. Alsanie, “Directional Radio Propagation Path Loss Models for
[144] T. Yilmaz, G. Gokkoca, and O. B. Akan, “Millimetre Wave Millimeter-Wave Wireless Networks in the 28-, 60-, and 73-GHz
Communication for 5G IoT Applications BT - Internet of Things (IoT) in Bands,” IEEE Trans. Wirel. Commun., vol. 15, no. 10, pp. 6939–6947,
5G Mobile Technologies,” C. X. Mavromoustakis, G. Mastorakis, and J. M. Oct. 2016, doi: 10.1109/TWC.2016.2594067.
Batalla, Eds. Cham: Springer International Publishing, 2016, pp. 37–53. [157] V. S. Abhayawardhana, I. J. Wassell, D. Crosby, M. P. Sellars, and M.
[145] M. B. Majed, T. A. Rahman, O. A. Aziz, M. N. Hindia, and E. Hanafi, G. Brown, “Comparison of Empirical Propagation Path Loss Models
“Channel Characterization and Path Loss Modeling in Indoor for Fixed Wireless Access Systems,” in 2005 IEEE 61st Vehicular
Environment at 4.5, 28, and 38 GHz for 5G Cellular Networks,” Int. J. Technology Conference, vol. 1, pp. 73–77, doi:
Antennas Propag., vol. 2018, pp. 1–14, Sep. 2018, doi: 10.1109/VETECS.2005.1543252.
10.1155/2018/9142367. [158] S. Aldossari and K.-C. Chen, “Predicting the Path Loss of Wireless
[146] Weite Zhang, Lei Tian, Zhongyuan Wu, Jianhua Zhang, Yi Zheng, and Channel Models Using Machine Learning Techniques in MmWave
Jingmin Cao, “Dynamic Range Impact on 3D MIMO Channel Urban Communications,” in 2019 22nd International Symposium on
Characteristics in Rural-Macro Scenario at 3.5 GHz,” in 12th European Wireless Personal Multimedia Communications (WPMC), Nov. 2019,
Conference on Antennas and Propagation (EuCAP 2018), 2018, pp. 47 pp. 1–6, doi: 10.1109/WPMC48795.2019.9096057.
(5 pp.)-47 (5 pp.), doi: 10.1049/cp.2018.0406. [159] S. Sun, T. A. Thomas, T. S. Rappaport, H. Nguyen, I. Z. Kovacs, and I.
[147] A. M. Al-Samman, T. A. Rahman, M. H. Azmi, A. Sharaf, Y. Yamada, Rodriguez, “Path Loss, Shadow Fading, and Line-of-Sight Probability
and A. Alhammadi, “Path loss model in indoor environment at 40 GHz Models for 5G Urban Macro-Cellular Scenarios,” in 2015 IEEE
for 5G wireless network,” in 2018 IEEE 14th International Colloquium Globecom Workshops (GC Wkshps), Dec. 2015, pp. 1–7, doi:
on Signal Processing & Its Applications (CSPA), Mar. 2018, pp. 7–12, 10.1109/GLOCOMW.2015.7414036.
doi: 10.1109/CSPA.2018.8368676. [160] A. M. Al-Samman, M. N. Hindia, and T. A. Rahman, “Path loss model
[148] A. M. Al-Samman, T. A. Rahman, M. H. Azmi, M. N. Hindia, I. Khan, in outdoor environment at 32 GHz for 5G system,” in 2016 IEEE 3rd
and E. Hanafi, “Statistical Modelling and Characterization of International Symposium on Telecommunication Technologies (ISTT),
Experimental mm-Wave Indoor Channels for Future 5G Wireless Nov. 2016, pp. 9–13, doi: 10.1109/ISTT.2016.7918076.
VOLUME XX, 2017 39
[161] M. N. Hindia, A. M. Al-Samman, T. A. Rahman, and T. M. Yazdani, rural hydrological monitoring in UNESCO biosphere environment,” in
“Outdoor large-scale path loss characterization in an urban environment 2017 IEEE 4th International Conference on Smart Instrumentation,
at 26, 28, 36, and 38 GHz,” Phys. Commun., vol. 27, pp. 150–160, Apr. Measurement and Application (ICSIMA), Nov. 2017, pp. 1–5, doi:
2018, doi: 10.1016/j.phycom.2018.02.011. 10.1109/ICSIMA.2017.8311981.
[162] A. A. Budalal, M. R. Islam, M. H. Habaebi, and T. A. Rahman, [175] O. Elijah, T. A. Rahman, H. . Saharuddin, and F. . Khairodin, “Factors
“Millimeter Wave Channel Modeling – Present Development and that Impact LoRa IoT Communication Technology,” in 2019 IEEE
Challenges in Tropical Areas,” in 2018 7th International Conference 14th Malaysia International Conference on Communication (MICC),
on Computer and Communication Engineering (ICCCE), Sep. 2018, Dec. 2019, pp. 112–117, doi: 10.1109/MICC48337.2019.9037503.
pp. 23–28, doi: 10.1109/ICCCE.2018.8539324. [176] H. Linka, M. Rademacher, O. G. Aliu, K. Jonas, B. Rhein-sieg, and S.
[163] B. W. Domae, R. Li, and D. Cabric, “Machine Learning Assisted Augustin, “Path Loss Models for Low-Power Wide-Area Networks :
Phase-less Millimeter-Wave Beam Alignment in Multipath Channels,” Experimental Results using LoRa,” VDE ITG-Fachbericht
arXiv Prepr. arXiv2109.14689, 2021. Mobilkommunikation, no. May, pp. 10–14, 2018.
[164] N. Xie, L. Ou-Yang, and A. X. Liu, “A Machine Learning Approach to [177] P. Jorke, S. Bocker, F. Liedmann, and C. Wietfeld, “Urban channel
Blind Multi-Path Classification for Massive MIMO Systems,” models for smart city IoT-networks based on empirical measurements
IEEE/ACM Trans. Netw., vol. 28, no. 5, pp. 2309–2322, Oct. 2020, doi: of LoRa-links at 433 and 868 MHz,” in 2017 IEEE 28th Annual
10.1109/TNET.2020.3008287. International Symposium on Personal, Indoor, and Mobile Radio
[165] L. Wu et al., “Received Power Prediction for Suburban Environment Communications (PIMRC), Oct. 2017, pp. 1–6, doi:
based on Neural Network,” in 2020 International Conference on 10.1109/PIMRC.2017.8292708.
Information Networking (ICOIN), Jan. 2020, pp. 35–39, doi: [178] G. Callebaut and L. Van der Perre, “Characterization of LoRa Point-to-
10.1109/ICOIN48656.2020.9016532. Point Path Loss: Measurement Campaigns and Modeling Considering
[166] L. Wu et al., “Artificial Neural Network Based Path Loss Prediction for Censored Data,” IEEE Internet Things J., vol. 7, no. 3, pp. 1910–1918,
Wireless Communication Network,” IEEE Access, vol. 8, pp. 199523– Mar. 2020, doi: 10.1109/JIOT.2019.2953804.
199538, 2020, doi: 10.1109/ACCESS.2020.3035209. [179] J. Gaelens, P. Van Torre, J. Verhaevert, and H. Rogier, “Lora mobile-
[167] H. Cheng, S. Ma, and H. Lee, “CNN-Based mmWave Path Loss to-base-station channel characterization in the antarctic,” Sensors
Modeling for Fixed Wireless Access in Suburban Scenarios,” IEEE (Switzerland), vol. 17, no. 8, 2017, doi: 10.3390/s17081903.
Antennas Wirel. Propag. Lett., vol. 19, no. 10, pp. 1694–1698, Oct. [180] N. A. B. Masadan, M. H. Habaebi, and S. H. Yusoff, “LoRa LPWAN
2020, doi: 10.1109/LAWP.2020.3014314. Propagation Channel Modelling in IIUM Campus,” in 2018 7th
[168] G. Caso, O. Alay, L. De Nardis, A. Brunstrom, M. Neri, and M.-G. Di International Conference on Computer and Communication
Benedetto, “Empirical Models for NB-IoT Path Loss in an Urban Engineering (ICCCE), Sep. 2018, pp. 14–19, doi:
Scenario,” IEEE Internet Things J., vol. 8, no. 17, pp. 13774–13788, 10.1109/ICCCE.2018.8539327.
Sep. 2021, doi: 10.1109/JIOT.2021.3068148. [181] S. Hosseinzadeh, M. Almoathen, H. Larijani, and K. Curtis, “A Neural
[169] E. Bedeer, J. Pugh, C. Brown, and H. Yanikomeroglu, “Measurement- Network Propagation Model for LoRaWAN and Critical Analysis with
based path loss and delay spread propagation models in VHF/UHF Real-World Measurements,” Big Data Cogn. Comput., vol. 1, no. 1, p.
bands for IoT communications,” IEEE Veh. Technol. Conf., vol. 2017- 7, 2017, doi: 10.3390/bdcc1010007.
Septe, pp. 1–5, 2018, doi: 10.1109/VTCFall.2017.8287901. [182] T. Ameloot, P. Van Torre, and H. Rogier, “A compact low-power LoRa
[170] J. Hejselbæk, J. O. Nielsen, C. Drewes, Wei Fan, and G. F. Pedersen, IoT sensor node with extended dynamic range for channel
“Propagation measurements for device-to-device communication in measurements,” Sensors (Switzerland), vol. 18, no. 7, pp. 10–13, 2018,
forest terrain,” in 12th European Conference on Antennas and doi: 10.3390/s18072137.
Propagation (EuCAP 2018), 2018, pp. 30 (5 pp.)-30 (5 pp.), doi: [183] G. M. Bianco, R. Giuliano, G. Marrocco, F. Mazzenga, and A. Mejia-
10.1049/cp.2018.0389. Aguilar, “LoRa System for Search and Rescue: Path-Loss Models and
[171] S. P. Sotiroudis, P. Sarigiannidis, S. K. Goudos, and K. Siakavara, Procedures in Mountain Scenarios,” IEEE Internet Things J., vol. 8, no.
“Fusing Diverse Input Modalities for Path Loss Prediction: A Deep 3, pp. 1985–1999, Feb. 2021, doi: 10.1109/JIOT.2020.3017044.
Learning Approach,” IEEE Access, vol. 9, pp. 30441–30451, 2021, doi: [184] D. Wohwe Sambo, A. Forster, B. O. Yenke, I. Sarr, B. Gueye, and P.
10.1109/ACCESS.2021.3059589. Dayang, “Wireless Underground Sensor Networks Path Loss Model for
[172] S. P. Sotiroudis, S. K. Goudos, and K. Siakavara, “Neural Networks Precision Agriculture (WUSN-PLM),” IEEE Sens. J., vol. 20, no. 10,
and Random Forests: A Comparison Regarding Prediction of pp. 5298–5313, May 2020, doi: 10.1109/JSEN.2020.2968351.
Propagation Path Loss for NB-IoT Networks,” in 2019 8th [185] S.-Y. Wang, Y.-H. Cheng, and J.-H. Tarng, “Improving the
International Conference on Modern Circuits and Systems Localization Accuracy for Sigfox Low-Power Wide Area Networks,”
Technologies (MOCAST), May 2019, pp. 1–4, doi: in 2019 IEEE Symposium on Computers and Communications (ISCC),
10.1109/MOCAST.2019.8741751. Jun. 2019, pp. 1–6, doi: 10.1109/ISCC47284.2019.8969734.
[173] K. M. Malarski, J. Thrane, H. L. Christiansen, and S. Ruepp, [186] M. C. Bor, U. Roedig, T. Voigt, and J. M. Alonso, “Do LoRa Low-
“Understanding Sub-GHz Signal Behavior in Deep-Indoor Scenarios,” Power Wide-Area Networks Scale?,” in Proceedings of the 19th ACM
IEEE Internet Things J., vol. 8, no. 8, pp. 6746–6756, Apr. 2021, doi: International Conference on Modeling, Analysis and Simulation of
10.1109/JIOT.2020.3027829. Wireless and Mobile Systems, Nov. 2016, pp. 59–67, doi:
[174] R. Nordin et al., “The world-first deployment of narrowband IoT for 10.1145/2988287.2989163.
40 VOLUME XX, 2021
Haider A.H. Alobaidy received the M.Sc. degree

in Electrical Engineering/ Electronics &
Communication from the Faculty of Engineering,
Al-Mustansiriyah University, Iraq, in 2016. He is
currently a Ph.D. student in the Department of
Electrical, Electronic and Systems Engineering,
Faculty of Engineering and Built Environment,
Universiti Kebangsaan Malaysia (UKM). His
research interest includes wireless sensor
networks, wireless communication, IoT, and
channel propagation modeling and estimation. He
is a Student Member of IEEE.
Mandeep Jit Singh is a Professor at the

Department of Electrical, Electronic and Systems
Engineering of the Universiti Kebangsaan
Malaysia (UKM). He is author and co-author of
more than 240 research articles in antenna and
microwave RF. His research interests include
communication antenna design, radio wave
propagation, satellite antennas, and IoT. He
received several International and National Medal
awards for his research and innovation. He is
currently the advisor of the Engineering Education
Technical Division (E2TD), Institute of Engineers Malaysia.
Mehran Behjati received a B.Sc. degree in

Electrical and Electronic Engineering from the
Azad University of Iran, in 2009, his M.Sc. and
Ph.D. degree in Communication Engineering from
the National University of Malaysia (UKM), in
2013 and 2017, respectively. He is currently a
postdoctoral researcher in the Department of
Electrical, Electronics, and System Engineering in
UKM. His main research interests include
advanced Multiple-Input Multiple-Output
(MIMO) systems, interference management, limited-feedback techniques,
aerial wireless communications, IoT, Internet of Drone (IoD), and green
communications.
Rosdiadee Nordin is currently an Associate

Professor of Electrical, Electronic & Systems
Engineering at Universiti Kebangsaan Malaysia
(UKM). Earlier, he graduated with a B.Eng. in
Electrical. Electronic & Systems Engineering from
Universiti Kebangsaan Malaysia (2001) and then
became an engineer in Malaysia’s
telecommunication industry for almost five years.
He later pursued his Ph.D. on wireless engineering
at the University of Bristol, United Kingdom
(2010). He is a Senior Member of IEEE.
Nor Fadzilah Abdullah received the B.Sc.

degree in electrical and electronics from Universiti
Teknologi Malaysia, in 2001, the M.Sc. degree
(Hons.) in communications engineering from The
University of Manchester, U.K., in 2003, and the
Ph.D. degree in electrical and electronic
engineering from the University of Bristol, U.K.,
in 2012. She is currently an Associate Professor
with the Department of Electrical, Electronic, and
Systems Engineering, Universiti Kebangsaan
Malaysia. Her research interests include 5G,
millimeter wave, LTE-A, vehicular networks, MIMO, space time coding,
fountain code, and channel propagation modeling and estimation. She is a
Member of IEEE.
VOLUME XX, 2017 41

Wireless Transmissions

Uploaded by

Copyright:

Available Formats

Wireless Transmissions

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Wireless Transmissions

Uploaded by

Copyright:

Available Formats

This article has been accepted for publication in a future issue of this journal, but has not been

Wireless Transmissions, Propagation and

Corresponding author: Rosdiadee Nordin (e-mail: adee@ukm.edu.my).

VOLUME XX, 2021 1

2 VOLUME XX, 2021

B. Study Motivations and Contributions

VOLUME XX, 2021 3

4 VOLUME XX, 2021

FIGURE 2. Several innovative IoT solutions from the reviewed literature.

VOLUME XX, 2021 5

C. Industrial Internet of Things (IIoT) E. Environmental and Agriculture IoT

6 VOLUME XX, 2021

VOLUME XX, 2021 7

8 VOLUME XX, 2021

VOLUME XX, 2021 9

10 VOLUME XX, 2021

VOLUME XX, 2021 11

12 VOLUME XX, 2021

VOLUME XX, 2021 13

𝑑 is Tx-Rx separation distance

𝑃𝐿𝑂𝐻𝑈 − 2(log10(𝑓/28))2 − 5.4 𝑓𝑜𝑟 𝑆𝑢𝑏𝑢𝑟𝑏𝑎𝑛

𝑃𝐿𝐻𝐷 = 𝑃𝐿𝑂𝐻 + 𝐶𝐻𝐷

𝑃𝐿𝐼𝐿𝑂𝑅𝐼𝑁 = 73.56 + 26.16 log10(𝑓) − 13.82 log10(ℎ𝑇𝑥 )

𝑑 is Tx-Rx separation in km (except for ITU-R

+(44.9 − 6.55 log10 ( ℎ𝑇𝑥 )) log10( 𝑑) + 𝐶𝑚 correction factors to extend Okumura-Hata

supported Tx antenna height to 2500m,

𝑃𝐿𝐸𝑥_𝐶231 = { (8) correction factors to extend Okumura-Hata

𝐴𝑏𝑚 = 20.41 + 9.83 log10(𝑑) + 7.89 log10(𝑓)

𝐴𝑏𝑚 is a basic medium loss,

Ericsson 9999 model,

{ −40.94 𝑓𝑜𝑟 𝑅𝑢𝑟𝑎𝑙

ℎ′𝑇𝑥 = ℎ𝑇𝑥 /√1 + (7 ∗ 10−6 ∗ ℎ2𝑇𝑥 )

14 VOLUME XX, 2021

30𝑚 < ℎ𝑇𝑥 < 1000𝑚 attenuation,

𝐴, and 𝐵 are environment correction factors,

𝑃𝐿𝐿𝑒𝑒 = 𝐴 + 𝐵 log10(𝑑/𝑑0 ) + 10𝑛 log10(𝑓/900) + 𝛼0

{ (13) 𝑛 is a distance-dependent PL Exponent (PLE),

𝛾 is SUI effective PLE,

𝑋𝑓 = 6 log10 (𝑓(𝑀𝐻𝑧)/2000) type A (hilly, dense vegetation), B (hilly, rare

𝑃𝐿𝐿𝑂𝑆 and 𝑃𝐿𝑁𝐿𝑂𝑆 is the PL for LOS/NLOS

𝑑 is the Tx-Rx separation distance in m,

Known as Floating Intercept (FI) or Alpha-Beta (AB) PL model (for

𝑑 ≥ 1𝑚), where: 𝛽 = 𝑃𝐿(𝑑0 ), 𝛼 = 𝑛, 𝑋𝜎𝐹𝐼 = 𝑋𝜎 , 𝑛 is a distance-dependent PLE,

𝑏 is the linear frequency-dependent factor of

𝐶𝐼𝐹 (18) dependent PLE, Tx antenna height, average Tx

{ , 𝑑0 =1m antennas heights in a set of measurements,

𝑃𝐿𝐴𝐵𝐺 (𝑓, 𝑑) = 10𝛼 log10(𝑑/𝑑0 ) + 𝛽 + 10𝛾(𝑓/𝑓0 ) + 𝑋𝜎𝐴𝐵𝐺 ,

𝑓 is the frequency in MHz, 𝑑 is Tx-Rx

𝑃𝐿𝐸𝑔𝑙𝑖 = 20 log10(𝑓) + 40 log10 (𝑑) − 20log10(ℎ𝑇𝑥 ) + 𝐶(ℎ𝑅𝑥 )

separation in km, ℎ𝑇𝑥 and ℎ𝑅𝑥 are Tx and Rx

{ 76.3 − 10 log10 (ℎ𝑅𝑥 ) 𝑓𝑜𝑟 ℎ𝑅𝑥 < 10𝑚 (21)

1.33𝑓 0.284 𝑑 0.588 𝑓𝑜𝑟 14𝑚 < 𝑑 ≤ 400𝑚

{ 𝑊_𝑀𝐸𝐷 0.45𝑓 0.284 𝑑 𝑓𝑜𝑟 0𝑚 < 𝑑 ≤ 14𝑚 (22)

𝑓 is the frequency in MHz (except for W

𝑃𝐿𝐼𝑇𝑈−𝑅 = 0.2𝑓 0.3 𝑑 0.6

{ (23) MED, frequency is given in GHz),

{ 0.39𝑓 0.39 𝑑0.25 𝑓𝑜𝑟 𝑖𝑛 𝑙𝑒𝑎𝑓 (24)

VOLUME XX, 2021 15