Proc.

of the 2nd International Conference on Electrical, Communication and Computer Engineering (ICECCE)
12-13 June 2020, Istanbul, Turkey

Opportunistic Resource Allocation for Narrowband

Internet of Things: A Literature Review
K.F Muteba K Djouani T.O Olwal
Electrical Engineering(FSATI) Electrical Engineering (FSATI) Electrical Engineering (FSATI)
Tshwane University of Technology Tshwane University of Technology Tshwane University of Technology
Pretoria, South Afirca Pretoria, South Afirca Pretoria, South Afirca
franckdestjean@gmail.com djouanik@tut.ac.za OlwalTO@tut.ac.za

Abstract—Nowadays, the growing adoption of the Internet hardware communication to gather , send and act on data they
of Things (IoT) is reshaping the telecommunication landscape get from the environment. IoT devices exchange the data they
and has penetrated every aspect of our lives with influential gather by connecting to an IoT gateway for the data to be
applications on smart health, home automation, smart logistics,
smart industries and smart cities. These advanced technologies analysed remotely (i.e., cloud) or for the data to be analysed
bring about numerous benefits and has begun to play a major locally by an edge device (i.e., base station) [4]. More often
role in daily lives, particularly data mining applied in precision than not, IoT devices connect with other related devices and
agriculture to discover knowledge. Also in agro-industrial pro- act on the data they share using connectivity, networking and
duction chain, the combination of wireless and distributed specific communication protocols that largely depend on their specific
sensor devices with the simulation of climatic conditions in order
to track the evolution of grapes for wineries is outstanding. deployment applications .
Mobile IoT such as narrowband-IoT (NB-IoT) and the long-term Consequently, research on the IoT and its derivatives has
evolution (LTE) for machines (LTE-M) are significant innovations taken over academia and industry alike with the focus of
of this accelerating development of IoT technologies. In the era increasing the transmission throughput in order for the wire-
of ubiquitous communication where everything is connected to less network to be able to support the huge communication
the internet, NB-IoT systems are expected to offer better quality-
of-services (QoS) to end users than the traditional IoT paradigm. demand. As a result, IoT services can now be classified into
However, offering better QoS satisfaction to end users will two groups, i.e. short-range communications such as Zigbee,
become a great challenge due to the bottleneck caused by the dual WiFi and Bluetooth to name just a few; and wide area com-
problem of increasing IoT use cases and the shortage of wireless munication generally called low power area network (LPWA),
spectrum resources. Whilst discussing the recent innovative such as long range (LoRa), Sigfox and the narrowband IoT
solutions of NB-IoT resource allocation, significant challenges
and open issues related to the real-time implementation of NB- (NB-IoT).
IoT are identified and discussed. Therefore, this paper gives The NB-IoT, which is a LPWA technology developed by
a general overview of the resource allocation solutions in this the Third-generation Partnership Project (3GPP) as a typical
NB-IoT innovation and further suggests and motivates for the application of the traditional IoT envisioned for use in smart
intelligentization of future resource allocation solutions (i.e., the metering, precision agriculture (i.e., crops monitoring), etc.,
use of artificial intelligence (AI) strategies).
Index Terms—Internet of Things, Narrowband-IoT,Long-term [5]. The objective of this initiative is for enabling the use of
evolution, Resource allocation, Intelligentization, Artificial intel- a broad range of IoT devices in the 5G era since it fits the
ligence, Machine learning. requirements of IoT services and is awaited to play a very
important role in the wireless connection of several devices
I. I NTRODUCTION [5].NB-IoT can be expanded into three different operating
The Internet of Things (IoT) is an advanced technological modes - (i) in-band, (ii) stand-alone and (iii) within the guard-
innovation that interconnects devices and enables them to band . As illustrated in Fig. 2 below, in stand-alone operation
communicate and interact autonomously over the Internet [1]. mode,he NB-IoT can use one GSM channel (200 kHz) while
With the adoption of the IoT technology escalating and making for both in-band and guardband modes of operation it uses a
real impacts in the way we live and work, it has been predicted block of LTE physical resources (180 kHz).The NB-IoT’s de-
that by the year 2020 billions of devices will be interconnected sign purposes include devices with low deployment cost, super
and interact with each other on the Internet [2]. Experts also coverage distance (20 dB improvement over GPRS), extended
predicted that more than 75 billion devices will be connected battery lifespan (over 10 years), low power consumption and
to the IoT by 2025, and this volume of connected devices will massive connection devices.The specifications for NB-IoT in
require wireless resources that can support them [3]. IoT is a release 14 enhance its protocol in several ways.
network of objects or networked things able to convey data The NB-IoT, as an LTE variant, uses Orthogonal Frequency-
without requiring human intervention [1]. All these objects Division Multiple Access (OFDMA) with a system bandwidth
or things form what is called an IoT ecosystem consisting of 180 kHz, which is equivalent to a Physical Resource Block
of internet connected smart devices that use sensor and other (PRB) in LTE/UMTS transmission. [3]. NB-IoT deeply reuses

978-1-7281-7116-6/20/$31.00 ©2020 IEEE

the LTE design to support flexible d eployment scenarios,i.e., Alternative solutions have been studied using machine learning
OFDM in the DL and SC-FDMA in the UL [6]. New key strategies, but due to the distributed nature of future wireless
design features such as time offset between control and networks, paralleled with heterogeneity in terms of use cases,
data transmission, synchronization sequences, random access resource demands and devices, more intelligent solutions are
preamble, repetition for coverage extension, etc. have also a requisite. From the analysis of the reviewed literature, it
been added to meet the requirements of IoT-based applications. transpired that deep architectures (i.e., deep learning and deep
These modifications a re p rimarily m otivated b y t he f act that reinforcement learning) have not been explored extensively
NB-IoT has to operate on a 180kHz bandwidth, which equals in this regard. Therefore, in order to fill this gap, this paper
to one PRB [7], while many LTE control channels have been undertakes a review to show how deep reinforcement learning
designed to cover multiple PRBs using a wider bandwidth than algorithm can be important for resource allocation in NB-IoT
180 kHz. These design modifications m eet t he requirements systems.
of IoT applications and thus ensure better coexistence perfor-
C. Contributions of this Review
mance with the existing LTE system.
However, given the restricted availability of spectrum re- The main contributions of this review can thus be summarized
sources (i.e., 180 kHz) for NB-IoT systems, which corresponds as follows:
to one physical resource block (PRB) in LTE transmission, one • To provide an overview of the NB-IoT, its founda-
major issue was to effectively and efficiently utilize this limited tional research, architecture, framework structure, physi-
resource [7]. Considering that the newly introduced feature cal channels, connection structure, operating modes and
of repeated transmissions, the traditional resource allocation applications
may not be appropriate for NB-IoT to accommodate the huge • The insufficient allocation of resources is not a recent
number of NB-IoT devices [8], new ways of opportunistically problem in wireless networks. However, due to the mas-
reusing other licensed spectrum bands came as a viable sive connectivity of NB-IoT systems, it is quite challeng-
solution. ing. Therefore, a comprehensive analysis of resource al-
The 3GPP objective is to provide communication protocol location techniques related IoT and NB-IoT is presented.
which are low energy consumption, long coverage distance • The real-time deployment of IoT will require network
and massive device connection. Compared to other wireless intelligence in different ways that can be monitored in
technologies, less spectrum is allocated to NB-IoT. With the several ways, one of which is the effective use of AI
growth in the number of Internet devices on the market, access strategies. For this reason, a review of the literature on
to spectrum resources is becoming an issue that affects the machine learning techniques used for IoT-NB is being
network Quality of Service (QOS). Thus, spectrum allocation conducted and the role of deep reinforcement learning in
becomes a major issue that is presented as the need to optimize the context of intelligent resource allocation is also being
the spectrum.Some works have been done concerning the examined..
resource spectrum allocation [8], [14], [15] and [20] . The rest of the paper is structured as follows: Firstly, in Section
II, the fundamentals of IoT operation are discussed. Section
A. Objectives and Scope of this Review
III discusses the fundamental design and operation of NB-
The objective of this literature review is to provide a coverage IoT systems . In Section IV, the performance of optimization
of resource allocation strategies in NB-IoT systems. This techniques in terms of traditional resource allocation strategies
review proceeds to cover the recent advancements in solutions and machine learning-based resource allocation strategies are
that involve artificial i ntelligence ( i.e., M achine Learning). discussed. Section V points out some open research challenges
Therefore, the first p art o f t his r eview p resents a n overview and finally section VI gives the conclusion of this review.
of resource allocation in NB-IoT systems and then introduces
the need for intelligent solutions. Then, the second part of this II. F UNDAMENTALS OF I OT O PERATION
review covers resource allocation solutions that use artificial Within this IoT paradigm, objects or things around us
intelligence. can be connected to the internet using Wireless technologies
like Bluetooth, ZigBee, etc. The development of the IoT in
B. Context of the Problem and Research Gap the past few years has been characterized by attribute labels
In the physical layer of the NB-IoT scheme, many issues such as: Anything (any device), to be transferred from/to
such as the effect of interaction between coverage classes, the Anyone (anybody), located Any place (anywhere), at Any
impact of channel selection and scheduling, and channel link time (any context), using the most appropriate physical path
adaptation need to be addressed [9]. In resolving these issues, from Any path (any network) available between the sender
a plethora of research works has been conducted, using solu- and the recipient based on performance and/or economic
tions from mathematical programming techniques from signal considerations, to provide Any service (any business) [1]. This
processing (i.e., queuing theory, game theory, etc.). However, fundamental ideology behind the IoT is shown in Fig. 1 below.
in order to equip NB-IoT systems for making better decision- Fig. 1 illustrates the fundamental ideology behind the IoT
making in future wireless networks, the traditional signal ecosystem hyper-connection, which has been coined as the
processing strategies fall short of addressing this requirement. 6As and 6Cs [10]. Using the 6As and 6Cs ideology, the IoT
 Fig. 2. NB-IoT deployment scenario (i.e., in-band and guard-band) and when
Re-farming the GSM Spectrum (standalone), Adopted from [16].

Fig. 1. The Next Generation IoT Hyper-connected: The 6As and 6Cs, Adopted
from [10]
block (PRB) of 180 kHz in the LTE frequency band for
the DL and the UL. Alternatively, it can be assigned as
innovations has transformed common physical objects in the standalone using the 200 kHz bandwidth by re-farming the
neighbouring environment into information ecosystems that global system for mobile (GSM) communication band. These
greatly improve our lives. The IoT paradigm is developing versatile implementation options facilitate rapid integration
rapidly and the whole IoT ecosystem is now based on and coexistence with cellular bands.NB-IoT systems have two
innervation elements known as 6Cs [11]. This basically UL operation options, (i) SC-FDMA and (ii) FDMA with
means, Collect (the heterogeneity of devices of diverse Gaussian minimum shift keying (GMSK) modulation, and
intelligence and complexity, which increases the real-time two DL options, (i) OFDM with 2.5 kHz subcarrier and (ii)
gathering of data from device and information connections), OFDM with 15kHz subcarrier spacing. To assess the amount
Connect (Pervasive connections of heterogeneous devices and of supported devices in NB-IoT, a study has been done in
information, where connections are the basic component of [3] for the option 2.5 kHz subcarrier spacing FDMA with
the IoT.), Cache (information stored in the IoT computing GMSK and SC-FDMA.
and distributed processing environment), Compute (extended NB-IoT is the most recent technology that has been
computing and treatment of data and information), Cognize acknowledged and normalized within a short space of time
(data analysis, information understanding, retrieval, real-time in reaction to several consumer demands and competitive
AI processing and Create (the creation of new interactions, pressure from other non-proprietary 3GPP technologies. It is
experiences, models and business solutions) [10]. an autonomous radio interface, firmly integrated into the LTE,
which also means that it is integrated into the current LTE
specifications. It can address the needs of massive Machine
III. NB-I OT M ODE OF O PERATION Type Communication (mMTC) based on its features. NB-IoT
requires a global industry standard and operates in licensed
In 2016 NB-IoT has been implemented by the 3GPP in spectrum. NB-IoT and LTE-M meet customer’s expectations
release 13 with the goal of creating affordable devices for international operations, stability, reliability, security and
compatible with LTE band with long range and long battery cost-effectiveness, and high scalability. The advantages of
life-time [12]. Designed to coexist with LTE technology, the NB-IoT compared to other IoT technologies are tabulated in
NB-IoT brings benefits t o I oT s uch a s w ide a rea network Table I below. As shown in Table I, NB-IoT offers advantages
coverage, massive device connection, low power consumption, regarding network coverage, frequency band efficiency, data
inexpensive devices, and low data throughput [13]. NB-IoT throughput, power consumption as well as the cost of the
employs the same frequency spectrum as LTE and uses the module.Therefore, under LPWA scenarios, NB-IoT is a
same modulation which is OFDM in Downlink , SC-FDMA superior choice.
in the uplink. Thus, by reusing LTE bandwidth, NB-IoT
resource allocation is realized in the time domain, whereas
in the LTE system it is done in the time and frequency
domains [14]. Hence, NB-IoT is more reliable and offers IV. P ERFORMANCE O PTIMIZATION IN NARROWBAND
better quality of service (QoS) as it coexists with the LTE I NTERNET- OF -T HINGS
cellular network [15]. Due to their restricted bandwidth
condition, NB-IoT systems can be expanded in any of the The NB-IoT technology is just in its initial development stages
following approaches: -in-band, guard-band, and stand-alone, and there is still more optimization that needs to be carried out
as illustrated in Fig. 2 below [16]. Fig. 2 shows the three in order to improve its performance. This section presents the
deployment modes of NB-IoT systems, in in-band and miscellaneous work that has been carried out in the context of
guard-band modes, the NB-IoT spans one physical resource NB-IoT optimization with regard to resource allocation.
 TABLE I availability thereby with the best coverage distance. In [17],an
OVERVIEW OF I OT T ECHNOLOGIES , Data Source: 3GPP S TANDARD , evaluation of the performance of the different NB IoT channels
H UAWEI T EST R ESULT AND M ARKET DATA . NB: ‡ = PSM, MCL, AND †
= MCL IN D B. is presented by considering a single cell scenario showing that
the coverage performance of NB IoT can achieve a coverage
LTE NB-IoT LTE-M LoRA improvement of up to 20dB compared to LTE technology.
Spectrum Licensed Licensed Licensed Unlicensed In this evaluation inter-cell interference arising from multiple
Bandwidth 20 MHz 180 kHz 1.4 MHz 125 - cells transmission is not considered. A study conducted in [18]
500 kHz
Maximum 144 dB 164 dB 156 dB 157 dB focused on the performance evaluation of the downlink and
Coupling uplink channels, providing their link budget and data channel
Loss analysis using the design of NB-IoT system. The results
Peak Data UL: 10 M UL: 150 K 1M 50 K
Rate (bps) DL: 5 M DL: 100 K (UL/DL) (UL/DL) obtained demonstrated that the design meets the coverage
UL Latency <1 ms 0.89 - 2.34 s ≤13.5 s 0.17 - 8.9 s requirement, i.e., 20dB coverage improvement. In [19], the
(100Bytes) (‡ = 154) (‡ = 154) († = 154) authors describe the target set forth for NB-IoT by the 3GPP in
Mobility >250 km/h 80 km/h >120 km/h 40 km/h
Module >10 USD <5 USD <10 USD 6 - 10 USD
release 13, in term of coverage enhancement,capacity (massive
Cost device connection), battery lifespan and latency; concluding
Expected 3 - 5 yrs 10+ yrs 5 - 10 yrs 10+ yrs that all the target can be met in all deployment scenarios.
Maximum However, these studies did not take into account the impact
Battery Life
of the repetition factor on the evaluation of performance [3].
Moreover, they do not consider the inter-cell interference
which is one of the major factors of network degradation [8].
A. Traditional NB-IoT Resource Allocation Solutions Other algorithms have been proposed to optimize spectrum
Some preliminary studies evaluate the NB-IoT capacity [6], allocation in NB-IoT where many characteristics are taken
where the authors considered the UL transmission mode and into account in the process (e.g. dynamic data slicing) but
suggest an algorithm which determines a suitable resource unit the impact of inter-cell interference and the repetition factor
to allocate the spectrum to every single device by minimising are not taken into account. Similarly, the study in [5] did not
the number of sub-frames consumed. The result shows that take into account the effect of the number of repetitions and
effectively the number of consumed sub-frames was reduced inter-cell interference but proposed an algorithm with the aim
by 50% comparatively to the baseline. of minimizing the number of subframes consumed to allocate
In terms of scheduling, the authors in [8] suggested a link and determine the resource unit for each device.
adaptation algorithm to ebhance the allocation of resource and Regarding resource management, a contribution from [20]
time for NB-IoT devices. Here, an algorithm that selects a examined the effect of intercellular interference by propos-
number of repetitions and a modulation coding scheme has ing an iterative algorithm based on a cooperative approach.
been designed for each NB-IoT device using an internal and The authors formulated an analytical model on a realizable
external loop to ensure transmission reliability and enhance throughput of NB-IoT in a single cell scenario. This approach
the data rate of the NB-IoT system. As a result obtained presents the effects on the repetition factor on the NB-IoT
the proposed method outperforms the straight forward method data rate in a multi-cell scenario. Next, a multi-cell scenario
and the repetition-dominated method by saving more active is explored to study the impact of inter-cell interference on
time and resource consumption. A review done by [14] con- the network, improving spectrum allocation to allow efficient
sidered scheduling techniques for resource allocation in NB- allocation to users.The result is an 8% improvement in data
IoT devices. A comprehensive analysis of the NB-IoT DL throughput and a 17% reduction in power consumption.
scheduling structure is presented with with specific details
involving physical channels and signals. Also, the design of B. Machine Learning-based NB-IoT Resource Allocation So-
the physical layer for downlink communication was described lutions
and the problems related to allocation of resources in NB-IoT The NB-IoT offers a wide range of flexible configurations
that satisfies the time constraints. Finally, a NB-IoT algorithm, (i.e., amount of radio resources allocated to a cluster of
which uses a pre-divided resource allocation to satisfy the user devices) for massive IoT radio access from clusters with
in order to maximize spectrum use in various circumstances, heterogeneous requirements for data transmission. The
has been suggested. online determination of the configurations that maximizes
Having studied capacity and scheduling, issues regarding the long-term average number of devices to be served
the coverage of NB-IoT devices still pose challenges in their at each transmission time interval without knowing the
deployment. Having said that, the authors in [15] has proposed traffic statistics poses a huge challenge. To address this
an algorithm based on a reinforcement learning for dynamic complexity of optimal configurations search, the authors
spectrum access with the aim of improving coverage while in [30] proposed a reinforcement learning-based real-time
reducing the number of repetitions - and thus minimizing configuration. They developed their algorithms using tabular
power consumption. As result obtained the proposed algorithm Q-learning, linear approximation-based Q-learning, and
efficiently selects the best channels with highest probability of a deep Q-network (DQN) in a single-parameter single-
group scenario. The simulation results obtained indicate learning is actually the best technique to use for classification
that the proposed approaches considerably outperforms tasks such service and device differentiation.
the conventional heuristic-based approaches that use load
estimation in terms of the number of served IoT devices. Also, B. Circumventing Network Congestion
the results indicated that linear approximation Q-learning
and DQN perform better than tabular Q-learning in terms of Another important issue that needs to be addressed in IoT
convergence time. services is the issue of network congestion. The deployment of
In another contribution in [31] by the same authors a large number of IoT services (i.e., devices and applications),
in [30], they used the same algorithms to develop real- either performing the same function or different applications,
time optimization approaches for configuration selection with each device performing the same or different activity
to compare with the load estimation based conventional transports information to a service center simultaneously.
UL resource configuration a pproach i n s ingle coverage Regardless of the amount of data that each device is offering
enhancement group single scenario. The results obtained to the network, due to the large number of devices themselves,
in this study indicated that tabular Q-learning, linear this simple operation could cause network congestion. In
approximation Q-learning and DQN algorithms considerably such situations, mobile and wireless networks need to provide
outperform the conventional heuristic approaches based on several mechanisms that will remotely manage such devices
load estimation in terms of the number of served IoT devices and applications and allow for intelligent scheduling in order
in the single coverage enhancement group scenario. The to protect and better utilise their capabilities for delivering
algorithms were further advanced via an action simplification such IoT services. Such mechanisms would help facilitate an
approach and via via centralized multi-agent cooperation appropriate application development and reduce the vulner-
approach for the real-time multi-parameters optimization ability of the network to application misbehaviour. Such an
in multiple coverage enhancement groups scenario, thereby intelligent mechanism could be offered through the use of
solve the problem that Q-learning agents do not converge in deep architectures (i.e., deep learning and deep reinforcement
high-dimensional configurations. T he a dvanced Q-learning learning).
based approaches’ superiority over the conventional heuristic
based approaches improves significantly w ith t he i ncrease in VI. C ONCLUSIONS
the configuration s ettings i n b oth t he n umber o f s erved IoT
devices and the convergence time. The objective of this literature review was to cover, in detail,
the most recent research works in opportunistic resource
allocation for NB-IoT system. The recent research works
V. O PEN C HALLENGES discussed thus cover the IoT and NB-IoT fundamentals .
For furure research this section presents some issues that It then proceeds to cover the resource allocation techniques
need to be addressed: using traditional signal processing optimization techniques and
then covers the few recent works that use machine learning
A. Addressing Differing Resource Demands techniques for resource allocation. Since NB-IoT systems have
The increase in the range if IoT services, each varying re- just been standardized, there are few contributions related to
quirement levels in terms of bandwidth, mobility and latency machine learning, which necessitate more research work to
has presented itself as a huge challenge. For example, real- be done in the intelligentization of the IoTs. However, it is
time services that are related to public safety specifically also crucial to address the energy-efficient operation of NB-
require low transmission latency, but not high bandwidth. Such IoT systems using the machine learning intelligent approaches.
alert services would also require higher priority compared to With the integration of cognitive radio networks (CRNs) with
normal metering services. Alternatively, services that provide IoT system already under immense research consideration
surveillance require high bandwidth, but do not have stringent around the globe, the cooperation mechanism between CRNs
delay requirements. Therefore, due to this varying require- and cloud computing services to provide sensing as a service
ments in terms of service demand, future mobile and wireless has also attracted huge research interest. Therefore, for a better
networks may need to be equipped with superior service operation of the IoTs under these proposed mechanisms, the
identification a nd c lassification. Id entifying th e se rvice which issues presented in section V need to be addressed.
is generating the network traffic and classifying it accordingly
in order to meet its specific demands is a requisite. The varying VII. ACKNOWLEDGMENT
levels of device mobility, which entails the degree to which
mobile devices and applications need to be nomadic, is another This work is supported in part by the National Research
important characteristic of service demand in the IoT era. Foundation of South Africa (Grant Number: 90604). Opinions,
Here, location information and geo-fencing become crucial findings and conclusions or recommendations expressed in
enablers in the IoT ecosystem. In such contexts, the mobile any publication generated by the NRF supported research are
network faces the challenge of being able to recognise different those of the author(s) alone, and the NRF accepts no liability
service types as well as different types of devices. Deep whatsoever in this regard.
 R EFERENCES Things, Journal of Networking and Computing Applications, vol. 89, pp.
14 - 25, Jul. 2017.
[1] I. V. Ngonadi, Implementing Internet of Things in a Remote Patient [23] S. Li, Q. Ni, Y. Sun, G. Min, and S. Al-Rubaye, Energy-efficient
Medical Monitoring System, DOI: 10.23956/ijermt.v6i8.132, June 2018. Resource Allocation for Industrial Cyber-physical IoT Systems in 5G Era,
[2] Vodacom, Narrowband Internet of Things Pushing the Boundaries of IoT, IEEE Transactions in Industrial Information, vol. 14, no. 4, pp. 2618 -
Vodacom, 2017. 2628, June 2018.
[3] Ericsson, NB-IoT - System Level Evaluation and Comparison - Stan- [24] X. Luo, Z. He, Z. Zhao, L. Wang, W. Wang, H. Ning, J-H. Wang,
dalone, Technical Report, 2015. W. Zhao, and J. Zhang,Resource Allocation in the Cognitive Radio
[4] V. R. Konduru and M. R. Bharamagoudra, Challenges and Solutions of Network-Aided Internet of Things for the Cyber-Physical-Social System:
Interoperability on IoT: How Far Have We Come in Resolving the IoT An Efficient Jaya Algorithm, Sensors, vol. 18, no. 3649, pp. 1 - 20, 27
Interoperability Issues, in Proceedings of the International Conference October 2018.
On Smart Technologies For Smart Nation (SmartTechCon), pp. 572 - [25] A. N. Akhtar, F. Arif, and A. M. Siddique, Spectrum Decision Frame-
578, 2017. work to Support Cognitive Radio Based IoT in 5G, Book Chapter, 05
[5] Y. Miao, W. Li, D. Tian, M. S. Hossain, and M. F. Alhamid, Narrowband December 2018.
Internet of Things: Simulation and Modeling, IEEE Internet of Things [26] A. A. Khan, M. H. Rehmani, and A. Rachedi, When Cognitive Radio
Journal, vol. 5, no. 4, pp. 2304 - 2314, 2018. Meets the Internet of Things?, in Proceedings of the International
Wireless Communications and Mobile Computing Confer-ence (IWCMC),
[6] Y. Yu and J. Wang, Uplink Resource Allocation for Narrowband Internet
pp. 469 - 474, 2016.
of Things (NB-IoT) Cellular Networks, in Proceedings of the Asia-
[27] P. Rawat, K. D. Singh, and J. M. Bonnin, Cognitive Radio for M2M
Pacific Signal and Information Processing Association Annual Summit
and Internet of Things: A Survey, Computer Communications, vol. 94,
and Conference (APSIPA ASC), pp. 466 - 471, 2018.
pp. 1 - 29, 2016.
[7] A. Hoglund, X. Lin, O. Liberg, A. Behravan, E. A. Yavuz, M. Van Der
[28] M. Haddad, Y. Hayel, and O. Habachi, Spectrum Coordination in
Zee, Y. Sui, T. Tirronen, A. Ratilainen, and D. Eriksson, Overview of
Energy-efficient Cognitive Radio Networks, IEEE Transactions on Ve-
3GPP Release 14 Enhanced NB-IoT, IEEE Network, vol. 31, no. 6, pp.
hicular Technology, vol. 64, no. 5, pp. 2112 - 2122, 2015.
16 - 22, 2017.
[29] J. Xu, J. Yao, L. Wang, Z. Ming, K. Wu, and L. Chen, Narrowband
[8] C. Yu, L. Yu, Y. Wu, Y. He, and Q. Lu, Uplink Scheduling and Link Internet of Things: Evolutions, Technologies, and Open Issues, IEEE
Adaptation for Narrow-band Internet of Things Systems, IEEE Access, Internet of Things Journal, vol. 5, no. 3, 2018.
vol. 5, pp. 1724 - 1734, 2017. [30] N. Jiang, Y. Deng, A. Nallanathan, and J. A. Chambers, Reinforcement
[9] K.F Muteba, K. Djouani, and T.O Olwal, A comparative Survey Study Learning for Real-Time Optimization in NB-IoT Networks, IEEE Journal
on LPWA IoT Technologies: Design, considerations, challenges and on Selected Areas in Communications, vol. 37, no. 6, pp. 1424 - 1440,
solutions,Procedia Computer Science vol.155, pp. 636 - 641, 2019. June 2019.
[10] O. Vermesan, M. Eisenhauer, M. Serrano, P. Guillemin, H. Sundmaeker, [31] N. Jiang, Y. Deng, A. Nallanathan, and J. A. Chambers, Deep Rein-
E. Z. Tragos, J. Valino, B. Copigneaux, M. Presser, A. Aagaard, R. forcement Learning for Real-Time Optimization in NB-IoT Networks, ,
Bahr, and E. C. Darmois, The Next Generation Internet of Things - pp. 1 - 30,
Hyperconnectivity and Embedded Intelligence at the Edge, Book Chapter
- Chapter 3, 83 pages.
[11] Q. Wu, G. Ding, Y. Xu, S. Feng, Z. Du, J. Wang, and K. Long, Cognitive
Internet of Things: A New Paradigm Beyond Connection, IEEE Internet
of Things Journal, vol. 1, no. 2, pp. 129 - 143, 2014.
[12] 3GPP, Understanding the Standards for LTE-Advanced Enhancements,
Inside 3GPP Release 13, 2016 Update, June 2016.
[13] U. Raza, P. Kulkarni, and M. Sooriyabandara, Low Power Wide Area
Networks: An Overview, IEEE Communications Surveys & Tutorials, vol.
19, no. 2, pp. 855 - 873, 2017.
[14] R. Boisguene, S. Tseng, C. Huang, and P. Lin, A Survey on NB-IoT
Downlink Scheduling: Issues and Potential Solutions, in Proceedings of
the 13th International Wireless Communications and Mobile Computing
Conference (IWCMC), pp. 547 - 551, 2017.
[15] M. Chafii, F. Bader, and J. Palicot, Enhancing Coverage in Narrow-
band-IoT Using Machine Learning, in Proceedings of the IEEE Wireless
Communications and Networking Conference (WCNC), pp. 1 - 6, 2018.
[16] C. B. Mwakwata, H. Malik, M. M. Alam, Y. Le Moullec, S. Parand,
and S. Mumtaz, Narrowband Internet of Things (NB-IoT): From Physical
(PHY) and Media Access Control (MAC) Layers Perspectives, Sensors
(Basel), vol. 19, no. 11, pp. 2613, June 2019.
[17] A. Adhikary, X. Lin, and Y. E. Wang, Performance Evaluation of
NB-IoT Coverage, in Proceedings the 84th IEEE Vehicular Technology
Conference (VTC-Fall), pp. 1 - 5, 2016.
[18] R. Ratasuk, N. Mangalvedhe, J. Kaikkonen, and M. Robert, Data Chan-
nel Design and Performance for LTE Narrowband IoT, in Proceedings of
the 84th IEEE Vehicular Technology Conference (VTC-Fall), pp. 1 - 5,
2016.
[19] R. Ratasuk, B. Vejlgaard, N. Mangalvedhe, and A. Ghosh, NB-IoT
System for M2M Communication, in Proceedings of the IEEE Wireless
Communications and Networking Conference Workshops(WCNCW), pp.
428 - 432, 2016.
[20] H. Malik, H. Pervaiz, M. M. Alam, Y. Le Moullec, A. Kuusik, and
M. A. Imran, Radio Resource Management Scheme in NB-IoT Systems,
IEEE Access, vol. 6, pp. 15051 - 15064, 2018.
[21] M. Nitti, M. Murroni, M. Fadda, and L. Atzori, Exploiting Social
Internet of Things Features in Cognitive Radio, IEEE Access, vol. 4,
pp. 9204 - 9212, 2016.
[22] F. F. Qureshi, R. Iqbal, and M. N. Asghar, Energy Efficient Wireless
Communication Technique Based on Cognitive Radio for Internet of