Comprehensive Survey on 5G: Next Generation Mobile Communication

This document presents a comprehensive survey on 5G, the next generation of mobile communication. It begins by highlighting the limitations of 4G networks and the increasing demands of mobile devices, voluminous data, and higher data rates, setting the stage for the necessity of 5G. The survey then outlines the key features of 5G, including ubiquitous connectivity, zero latency, and high-speed Gigabit connections. It delves into the challenges in 5G network development, such as increasing data rates with low power consumption, handling interference, and ensuring security and privacy. The paper explores various proposed architectures, including multi-tier, cognitive radio-based, cloud-based, and energy-efficient designs. Implementation issues like interference management, handoff, QoS, load balancing, and channel access are discussed. The survey also covers methodologies, technologies, applications, and real-world demonstrations and testbeds related to 5G, providing a holistic view of this evolving technology. This survey is designed to provide a detailed overview of the characteristics and advancements of 5G mobile communication.

Accepted Manuscript
A survey on 5G: The next generation of mobile communication
Nisha Panwar, Shantanu Sharma, Awadhesh Kumar Singh
PII: S1874-4907(15)00053-1
DOI: http://dx.doi.org/10.1016/j.phycom.2015.10.006
Reference: PHYCOM 302
To appear in: Physical Communication
Received date: 30 June 2015
Revised date: 11 October 2015
Accepted date: 30 October 2015
Please cite this article as: N. Panwar, S. Sharma, A.K. Singh, A survey on 5G: The next
generation of mobile communication, Physical Communication (2015),
http://dx.doi.org/10.1016/j.phycom.2015.10.006
This is a PDF file of an unedited manuscript that has been accepted for publication. As a
service to our customers we are providing this early version of the manuscript. The manuscript
will undergo copyediting, typesetting, and review of the resulting proof before it is published in
its final form. Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
A Survey on 5G: The Next Generation of Mobile Communication1
Nisha Panwar1, Shantanu Sharma1, and Awadhesh Kumar Singh2
2
1Department of Computer Science, Ben-Gurion University of the Negev, Israel.3
{panwar, sharmas}@cs.bgu.ac.il.4
2Department of Computer Engineering, National Institute of Technology, Kurukshetra, India.5
[e-mail]6
Abstract7
A rapidly increasing number of mobile devices, voluminous data, and higher data rate are pushing to rethink8
the current generation of the cellular mobile communication. The next or fifth generation (5G) cellular networks9
are expected to meet these requirements. The 5G networks are broadly characterized by three unique features:10
ubiquitous connectivity, very low latency, and very high-speed data transfer. The 5G networks will provide novel11
architectures and technologies beyond state-of-the-art architectures and technologies. In this paper, we will find12
an answer to the question: “what will be done by 5G and how?” We investigate and discuss serious limitations13
of the fourth generation (4G) cellular networks and corresponding new features of 5G networks. We identify14
challenges in 5G networks, new technologies for 5G networks, and a comparative discussion of the proposed15
architectures that can be categorized on the basis of energy-efficiency, network hierarchy, and network types.16
Interestingly, implementation issues, e.g., interference, QoS, handoff, security-privacy, channel access, and load17
balancing, hugely effect the realization of 5G networks. Furthermore, our discussion highlights the feasibility18
of these models through an evaluation of existing real-experiments and testbeds.19
Keywords: Cloud radio access networks; cognitive radio networks; D2D communication; dense deployment; multi-tier20
heterogeneous network; privacy; security; tactile Internet.21
1 Introduction22
The evolution of the cellular network generations is primarily influenced by a continuous growth in wireless user23
devices, data usage, and the need for a better quality of experience (QoE). It is expected that more than 50 billion24
connected devices will utilize the cellular network services by the end of the year 2020 [1], and it will result in a25
tremendous increase in data traffic, as compared to the year 2014 [2]. However, state-of-the-art solutions are not26
sufficient for the challenges mentioned above. In short, the increase of 3D (‘D’evice, ‘D’ata, and ‘D’ata transfer27
rate) encourages the development of 5G networks.28
Specifically, the fifth generation (5G) of the cellular networks will highlight and address three broad29
views, as: (i) user-centric (by providing 24 ×7 device connectivity, uninterrupted communication services,30
and a smooth consumer experience), (ii) service-provider-centric (by providing a connected intelligent31
transportation systems, road-side service units, sensors, and mission critical monitoring/tracking services),32
and (iii) network-operator-centric (by providing an energy-efficient, scalable, low-cost, uniformly-monitored,33
programmable, and secure communication infrastructure). Therefore, 5G networks are perceived to materialize34
the three main features as below:35
• Ubiquitous connectivity: In the future, many types of devices will connect ubiquitously and provide an36
uninterrupted user experience. In fact, the user-centric view will be realized by ubiquitous connectivity.37
• Zero latency: 5G networks will support life-critical systems and real-time applications and services with zero38
delay tolerance. Hence, it is envisioned that 5G networks will realize zero latency, i.e, very low latency of the39
order of 1 millisecond [3, 47]. In fact, the service-provider-centric view will be realized by the zero latency.40
• High-speed Gigabit connection: The zero latency property could be achieved using a high-speed connection for41
fast data transmission and reception, which will be of the order of Gigabits per second to users and machines [3].42
1
Manuscript
Click here to view linked References

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
A few more key features of 5G networks are enlisted and compared to the fourth generation (4G) of the cellular43
networks, as below [4, 5, 6]: (i) 10-100xnumber of connected devices, (ii) 1000x higher mobile data volume per44
area, (iii) 10-100xhigher data rate, (iv) 1 millisecond latency, (v) 99.99% availability, (vi) 100% coverage, (vii) x
1045
energy consumption as compared to the year 2010, (viii) real-time information processing and transmission, (ix) x
546
network management operation expenses, and (x) seamless integration of the current wireless technologies.47
5G
Networks
Increased
data rate
&
network
capacity
Densification, FDD,
CRN, mMIMO, D2D
communication, full
duplex radio
Multi-RAT, self-heal,
densification, CRN,
NFV, SDN, C-RAN,
RANaaS, CONCERT,
Low latency
Cache, fast
handoff, D2D
communication,
mobile small-
cells, self-heal Scalability
Environmental
friendly & less
money
QoSSecurity &
privacy
Interference &
handoff
management
NFV, SDN,
C-RAN,
RANaaS,
CONCERT
Delay-bound QoS,
Quality management
equipment, multi-links
with multi-flow and
multi-QoS
C-RAN, VLC,
mmWave,
mMIMO, small-
cells, D2D
communication,
user separation
Monitoring and
encryption-decryption
SIC, CRN,
advance receiver,
joint
detection/decodi
ng
Inter-tier, intra-
tier, and
multi-RAT
handoff,
Figure 1: Requirements and proposed solutions for the
development of 5G networks. The inner, middle, and
outermost layers present requirements, solutions, and
applications of 5G networks, respectively. Two colored
wedges highlight primary features of 5G networks.
Therefore, the revolutionary scope and the48
consequent advantages of the envisioned 5G49
networks demand new architectures, methodologies,50
and technologies (see Figure 1), e.g., energy-efficient51
heterogeneous frameworks, cloud-based52
communication (software-defined networks (SDN)53
and network function virtualization (NFV)),54
full duplex radio, self-interference cancellation55
(SIC), device-to-device (D2D) communications,56
machine-to-machine (M2M) communications, access57
protocols, cheap devices, cognitive networks (for58
accessing licensed, unlicensed, and shared frequency59
bands), dense-deployment, security-privacy60
protocols for communication and data transfer,61
backhaul connections, massive multiple-input and62
multiple-output (mMIMO), multi-radio access63
technology (RAT) architectures, and technologies64
for working on millimeter wave (mmWave) 30–30065
GHz. Interestingly, 5G networks will not be a mere66
enhancement of 4G networks in terms of additional67
capacity; they will encompass a system architecture68
visualization, conceptualization, and redesigning at69
every communication layer [51].70
Several industries, Alcatel-Lucent [7],71
DOCOMO [8], GSMA Intelligence [5], Huawei [9],72
Nokia Siemens Networks [3], Qualcomm [10], Samsung [11], Vodafone,1 the European Commission supported73
5G Infrastructure Public Private Partnership (5GPPP) [4], and Mobile and Wireless Communications Enablers for74
the Twenty-Twenty Information Society (METIS) [6], are brainstorming with the development of 5G networks.75
Currently, the industry standards are yet to be explored about the expected designs and architectures for 5G76
networks.77
Scope of the paper. In this paper, we will review the vision of the 5G networks, advantages, applications, proposed78
architectures, implementation issues, real demonstrations, and testbeds. The outline of the paper is provided in79
Figure 2. In Section 2, we will discuss the vision of 5G networks. Section 3 presents challenges in the development80
of 5G networks. Section 4 address the current proposed architectures for 5G networks, e.g., multi-tier, cognitive81
radio based, cloud-based, device proximity based, and energy-efficient architectures. Section 5 presents issues82
regarding interference, handoff, quality of services, load balancing, channel access, and security-privacy of the83
network. Sections 6, 7, and 8 present several methodologies and technologies involved in 5G networks, applications84
of 5G networks, and real demonstrations and testbeds of 5G networks, respectively.85
We would like to emphasize that there are some review works on 5G networks by Andrews et al. [20],86
Chávez-Santiago et al. [35], and Gavrilovska et al. [50], to the best of our knowledge. However, our perspective87
about 5G networks is different, as we deal with a variety of architectures and discuss several implementation affairs,88
technologies in 5G networks along with applications and real-testbed demonstrations. In addition, we intentionally89
avoid an mmWave oriented discussion in this paper, unlike the current work [20, 35, 50].90
We encourage our readers to see an overview about the generations of the cellular networks (see Table 1) and91
crucial limitations of the current cellular networks in the next section.92
1http://www.surrey.ac.uk/5gic/research
2

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Generations Year Features Limitations
1G 1980s Analog signals for voice only communications Very less security
2G 1990s Digital signals, voice communications, and text
messaging
Very less support for the Internet
3G 1998-99 Voice communications, wireless mobile and fixed
Internet access, video calls, and mobile television (TV)
Less support for high-speed
Internet
4G 2008-09 Higher data rate (hundreds of megabits per second) No support for 50 billion ubiquitous
connected devices
5G 2020 Mentioned in Section 1
Table 1: The generations of the cellular networks.
1. Introduction
An introduction
of the paper and
scope of the paper
2. Desideratum
of 5G Networks
Dramatic upsurge in
device scalability,
massive data
streaming and high
data rate, spectrum
utilization,
ubiquitous
connectivity, and
zero latency
1.1 Limitations of
the current cellular
networks
No support for
bursty data
traffic, inefficient
utilization of
processing
capabilities of a
base-station, co-
channel
interference, no
support for
heterogeneous
wireless
networks, and no
separation of
indoor and
outdoor users
4. Architectures of the
Future/5G Mobile
Cellular Networks
4.1 Two-tier Architectures
How small-cells are
deployed under macrocells?
A Survey on 5G: The Next Generation of Mobile Communication
3. Challenges in
the Development
of 5G Networks
Increase data rate
and network
capacity with low
power consumption,
scalability and
flexibility, handling
interference,
environmental
friendly, low latency
and high reliability,
price, high mobility,
self-healing
infrastructures, QoS,
and security and
privacy of the
network and UEs
5.Implementation
Issues in 5G
Networks
6. Methodologies
and Technologies
for 5G Networks
Remaining
methodologies and
technologies are
discussed, e.g., SIC,
DUD, NFV, SDN,
mmWave, M2M
communication,
mMIMO, VLC
7. Applications
of 5G Networks
Personal usages,
virtualized homes,
smart societies,
smart grids, the
tactile Internet,
automation, health-
care systems,
logistics and
tracking, and
industrial usages
8. Real
Demonstrations
of 5G Networks
How industries and
academia are
looking towards
5G? What kind of
real
implementations
and testbeds they
are doing?
5.1 Interference
Management
4.2 CRN-based
Architectures
How CRNs are deployed
under a macrocell?
4.3 D2D Communication
Architectures
How devices communicate
to their close devices
without involving a MBS?
4.4 Cloud-based
Architectures
How the cloud facilitate
communication in 5G
networks?
4.5 Energy-efficient
Architectures
How to save energy in 5G
networks?
5.2 Handoff
Management
5.3 QoS
Management
5.4 Load
balancing
5.5 Channel
Access Control
Management
5.6 Security and
Privacy
Management in
5G Networks
Figure 2: Schematic map of the paper.
1.1 Limitations of the Conventional Cellular Systems93
4G networks are not substantial enough to support massively connected devices with low latency and significant94
spectral efficiency, which will be crucial in the future. In this section, we discuss a few crucial aspects in which95
conventional cellular networks lag behind, thereby motivating the evolution of 5G networks.96
No support for bursty data traffic. There are several mobile applications that send heartbeat messages to their97
servers and occasionally ask for a very high data transfer speed for a very short duration. Such types of data98
transmission may consume more battery life of (mobile) user equipments (UEs) with increasing bursty data in the99
network, and hence, may crash the core network [123]. However, only one type of signaling/control mechanism is100
designed for all types of the traffic in the current networks, creating high overhead for bursty traffic [64, 25].101
Inefficient utilization of processing capabilities of a base-station. In the current cellular networks, the processing102
power of a base-station (BS) can only be used by its associated UEs, and they are designed to support peak time103
traffic. However, a BS’s processing power can be shared across a large geographical area when it is lightly loaded.104
For example: (i) during the day, BSs in business areas are over-subscribed, while BSs in residential areas are105
almost idle, and vice versa [115], and (ii) BSs in residential areas are overloaded in weekends or holidays while106
BSs in business areas are almost idle [92]. However, the almost idle BSs consume an identical amount of power as107
over-subscribed BSs, hence, the overall cost of the network increases.108
Co-channel interference. A typical cellular network uses two separate channels, one as a transmission path from a109
UE to a BS, called uplink (UL), and the reverse path, called downlink (DL). The allocation of two different channels110
for a UE is not an efficient utilization of the frequency band. However, if both the channels operate at an identical111
frequency, i.e., a full duplex wireless radio [27], then a high level of co-channel interference (the interference112
between the signals using an identical frequency) in UL and DL channels is a major issue in 4G networks [86]. It113
3

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
also prevents the network densification, i.e., the deployment of many BSs in a geographical area.114
No support for heterogeneous wireless networks. Heterogeneous wireless networks (HetNets) are composed of115
wireless networks of diverse access technologies, e.g., the third generation (3G), 4G, wireless local area networks116
(WLAN), WiFi, and Bluetooth. HetNets are already standardized in 4G; however, the basic architecture was not117
intended to support them. Furthermore, the current cellular networks allow a UE to have a DL channel and a UL118
channel must be associated with a single BS that prevents the maximum utilization of HetNets. In HetNets, a UE119
can select a UL channel and a DL channel from two different BSs belonging to two different wireless networks for120
performance improvement [29, 42].121
No separation of indoor and outdoor users. The current cellular networks have a single BS installed near the122
center of the cell and interacts with all the UEs irrespective of the indoor or outdoor location of the UEs; while123
UEs stay indoors and outdoors for about 80% and 20% of the time, respectively. Furthermore, the communication124
between an indoor UE and an outside BS is not efficient in terms of data transfer rate, spectral efficiency, and125
energy-efficiency, due to the attenuation of signals passing through walls [107].126
Latency. When a UE receives an access to the best candidate BS, it takes several hundreds of milliseconds in the127
current cellular networks [121], and hence, they cannot support the zero latency property.128
2 Desideratum of 5G Networks129
A growing number of UEs and the corresponding surge in the bandwidth requirement for the huge amount of data130
transmission certainly necessitate the novel enhancement to the current technology. In this section, we highlight131
requirements of the future 5G networks.132
Dramatic upsurge in device scalability. A rapid growth of smart phones, gaming consoles, high-resolution TVs,133
cameras, home appliances, laptops, connected transportation systems, video surveillance systems, robots, sensors,134
and wearable devices (watches and glasses) is expected to continue exponentially in the near future. Therefore, 5G135
networks are perceived to support massively connected devices [107, 1, 15].136
Massive data streaming and high data rate. A vast growth in a number of wireless devices will of course137
result in a higher amount of data trading (e.g., videos, audio, Web browsing, social-media data, gaming, real-time138
signals, photos, bursty data, and multimedia) that will be 100-times more as compared to the year 2014 and would139
overburden the current network. Thus, it is mandatory to have matching data transfer capabilities in terms of new140
architectures, methods, technologies, and data distribution of indoor and outdoor users [61, 15, 60].141
Spectrum utilization. The two different channels (one for a UL and another for a DL) seem redundant from142
the point of view of the spectrum utilization [59]. In addition, the currently allocated spectrums have their143
significant portions under-utilized [12]. Hence, it is necessary to develop an access control method that can144
enhance the spectrum utilization. Furthermore, the spectrum utilization and efficiency have already been stretched145
to the maximum. It definitely requires spectrum broadening (above 3 GHz) along with novel spectrum utilization146
techniques [34].147
Ubiquitous connectivity. Ubiquitous connectivity requires UEs to support a variety of radios, RATs, and bands148
due to the global non-identical operating bands. In addition, the major market split between time division duplex149
(e.g., India and China) versus frequency division duplex (e.g., US and Europe) so that UEs are required to support150
different duplex options. Hence, 5G networks are envisioned for seamless connectivity of UEs over HetNets [13].151
Zero latency. The future mobile cellular networks are expected to assist numerous real-time applications, the tactile152
Internet [47, 46], and services with varying levels of quality of service (QoS) (in terms of bandwidth, latency, jitter,153
packet loss, and packet delay) and QoE (in terms of users’ and network-providers’ service satisfaction versus154
feedback). Hence, 5G networks are envisioned to realize real-time and delay-bound services with the optimal QoS155
and QoE experiences [15, 86].156
3 Challenges in the Development of 5G Networks157
The vision of 5G networks is not trivial to achieve. There are several challenges (some of the following challenges158
are shown in Figure 1 with their proposed solutions) to be handled in that context, as mentioned below:159
4

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Data rate and network capacity expansion with energy optimization. The deployment of more BSs in a160
geographical area, use of the higher frequency bands, and link improvement might support the network capacity161
expansion, billions of UEs, high data rate, high volume of data, and efficient backhaul data transfer to the core162
network. However, the implementation of these solutions is a cumbersome task in terms of economy and energy163
intake. Hence, the network capacity is required to be significantly increased, keeping the energy consumption and164
cost under strict control.165
Proposed solutions: Network densification or small-cell deployment [15, 28, 107] (Section 4.1), cognitive166
radio networks (CRNs) [16] (Section 4.2), mMIMO [71, 81, 87] (Section 6), network offload using D2D167
communication [33, 104, 113] (Section 4.3), efficient backhaul networks [51, 88] (Section 4.1.1), energy-efficient168
architectures [62, 83] (Section 4.5), full duplex radios [27] (Section 6), NFV, and SDN based architectures [14, 78,169
97, 119] (Section 6).170
Scalability and flexibility. These are the most prominent features of the future mobile communication. The171
future cellular infrastructures and methodologies must be designed to work in HetNets. Moreover, a vast number172
of potential users might request simultaneously for a set of services. Therefore, 5G networks must be powerful173
enough to support a scalable user demand across the coverage area [78, 94].174
Proposed solutions: NFV- and SDN-based architectures [14, 78, 97, 119] (Section 6).175
Single channel for both UL and DL. A full duplex wireless radio [27] uses only a single channel for transmitting176
and receiving signals at identical time and frequency. Thus, a full duplex system achieves an identical performance177
as having different UL and DL channels, and hence, increases link capacity, saves the spectrum, and cost. However,178
the implementation of full duplex systems is not trivial, because now a radio has to use sophisticated protocols179
for the physical and the data link layers [122], and mechanisms to remove the effects of interference [59]. The180
advantages of a full duplex radio in 5G networks are given in [56, 59, 64].181
Handling interference. Handling interference among communicating devices is a well-known challenge in the182
wireless communication. Due to a growing number of UEs, technologies (e.g., HetNets, CRNs, full duplex, and183
D2D communication) and applications, the interference will also increase in 5G networks, and the state-of-the-art184
technique may not perform well in the future cellular networks [61]. In 5G networks, a UE may receive interference185
from multiple macrocell base-stations (MBSs), various UEs, and small-cell base-stations (SBSs). Hence, it is186
required to develop an efficient (in terms of avoiding network overload) and reliable (in terms of perfect interference187
detection and decoding) interference management technique for channel allocation, power control, cell association,188
and load balancing.189
Proposed solutions: Self-interference cancellation [64, 59], an advance receiver with interference joint190
detection/decoding, and network-side interference management [86]. We will discuss these solutions in Section 5.1.191
Environmentally friendly. The current radio access network (RAN) consumes 70%-80%of the total power [64,192
114]. The wireless technologies consume lots of energy that lead to huge CO 2 emission and inflate the cost. It is193
a serious threat to the environment [107]. Thus, it is required to develop energy-efficient communication systems,194
hardware, and technologies, thereby the ratio between the network throughput and energy consumption is equitable.195
Proposed solutions: Cloud-RAN (C-RAN) [114, 62], visual light communication (VLC) [114], mmWave [114],196
separation of indoor and outdoor users [114], joint investigation of spectral efficiency and energy-efficacy [64, 62],197
multi-tier architectures [62], D2D communication [33, 104, 113], mMIMO architectures [62], and full duplex198
radios [64]. Except the above mentioned solutions, we will discuss some special techniques/architectures in the199
context of energy-efficiency in 5G networks in Section 4.5.200
Low latency and high reliability. Low latency and high reliability are critical in several real-time applications,201
e.g., message transmission by robots monitoring patients, life safety systems, cloud-based gaming, nuclear reactors,202
sensors, drones, and connected transportation systems. However, it is very challenging to have very low latency and203
reliable delivery of data over a large scale network without increasing the network infrastructure cost, as it requires204
the development of techniques providing fast connections, quick handovers, and high data transfer rate.205
Proposed solutions: Caching methods [29, 112], VLC, mmWave, mMIMO (Section 6), fast handover206
techniques [40, 93, 102] (Section 5.2), and D2D communication (Section 4.3).207
Network performance optimization. The performance parameters, e.g., peak data rate, geographical area208
coverage, spectral efficiency, QoS, QoE, ease of connectivity, energy-efficiency, latency, reliability, fairness of209
users, and implementation complexity, are crucial for a cellular network [107]. Hence, a general framework for 5G210
5

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Methodologies/Technologies
Section
Increase data
rate
Increase
network
capacity
Massive
device support
Energy-efficient
Low latency
Economic
Security and
privacy
Interference
Mobility
support
Small-cells 4.1 X X X X P 2
Mobile small-cells 4.1 X X X X P X
CRN 4.2 X X
D2D 4.3 X P X P X X P X P
C-RANs 4.4 X X X X X X X
Full duplex radio 3,5.1 X X
Advance receiver 5.1 X X
SIC 5.1,6 X X P X X P
DUD 6 X P X
mmWave 6 X X X X X
mMIMO 6 X X X X X X
VLC 6 X X X X X
CCN-based caching 6 X
Table 2: Summary of methodologies and technologies for 5G networks.
networks should substantially optimize these parameters. However, there are some tradeoffs among all parameters,211
which further emphasize the need of a joint optimization algorithm.212
Economical impacts. A revolutionary change in the future mobile communication techniques would have drastic213
economical impacts in terms of deployment and motivation for user participation. It is critical to provide an entirely214
new infrastructure due to economical stretch. Therefore, the cost of deployment, maintenance, management, and215
operation of an infrastructure must be affordable from the perspective of governments, regulating authorities, and216
network operators. Also, the cost of using D2D communication should be feasible, so that devices involved in D2D217
communication should not charge more than using the services of a BS [15, 46]. Further, the projected revenue218
growth is much lower than the traffic growth [14]; hence, it is required to develop 5G networks in a manner that219
both network operators and users get honey in their hands.220
High mobility and handoff. The 5G wireless UEs are meant for retaining an active service connection while221
frequently moving from one cell to another or from one RAT (e.g., 3G, 4G, 5G, WiFi, Bluetooth, and WLAN) to222
another. The mobility adaptation for the wireless services should not back-off even at a very high speed as a UE223
inside a moving vehicle. Moreover, during a particular interval, many UEs move from one place to another; for224
example, moving to offices from residential areas in the morning. As a result, 5G networks are envisioned to use225
the spectrum in the best manner and to cope up with pace of the device movement.226
Proposed solutions: Inter-tier, intra-tier, and multi-RATs handoff mechanisms, and a mechanism for secure227
handoff [40, 93, 102, 52], which we will discuss in Section 5.2.228
Self-healing infrastructures. A self-healing infrastructure finds a failed macrocell or small-cell (i.e., a cell that is229
unable to work because of hardware failures, software failures, or misconfigurations) with the help of neighboring230
cells and provides a way for communication to the affected users by adjusting the transmission power and231
operating channels in the neighboring cells [41, 111]. The design of a self-healing network insists on the frequent232
communication among cells; hence, it brings in the following challenges, as: (i) develop an efficient algorithm233
that can detect and reconfigure a failed cell with insignificant communication and computational overheads in the234
minimal detection time, and (ii) reconfiguration of a failed cell should not lead to degradation of nearby cells’235
services.236
Proposed solutions: A small-cell network with self-healing property is suggested in [111], which we will237
discuss in Section 4.1.2.238
QoS. QoS guarantee in 5G networks has inherent difficulties, e.g., node mobility, multi-hop communication,239
resource allocation, and lack of central coordination. In addition, in 5G networks, a huge amount of bursty and240
2X P : Partial support
6

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
multimedia data, multi-RATs, and low latency bound for different applications and services are major hurdles in241
achieving the desired QoS. Hence, it is challenging to design fast and efficient algorithms to maintain real-time242
QoS without overloading a BS [123, 120].243
Proposed solutions: Delay-bound QoS [62, 120], intelligent equipment [123], and multi-link with multi-flow244
and multi-QoS [69] have been suggested, which we will discuss in Section 5.3.245
Security and privacy of the network and UEs. The promising features of 5G networks bring in hard challenges246
in the design of security and privacy oriented 5G networks. For example, a huge number of new types of social247
(all-time connected) devices may originate several types of attacks like impersonation, denial-of-services (DoS),248
replay, eavesdropping, man-in-the-middle, and repudiation attacks. Also, the transfer of a huge volume of data in249
secure and high speed manners is critical while preventing malicious files to penetrate. In addition, the network250
densification needs to be secure and requires fast-secure handoff of UEs. We further highlight challenges in security251
and privacy of the network and UEs in Section 5.6.252
Proposed solutions: Physical layer security [118], monitoring [105, 48, 76], secret adaptive frequency253
hopping [76], encrypted- [79], and policy-based communications [67], which we will discuss in Section 5.6.254
All the above mentioned methodologies and technologies are comparatively studied in Table 2.255
4 Architectures for the Future/5G Mobile Cellular Networks256
In this section, we elaborate on the existing architectures for 5G networks, namely multi-tier, CRN-based, D2D257
communication based, and the cloud-based architectures. These proposed 5G architectures will be explained in the258
light of relevant advantages, disadvantages, and the challenges that are yet to be resolved.259
4.1 Two-tier Architectures260
A mobile small-cell
on a train
DR-OC
DC-OC
Relay device
Destination
Source
DR-DC
DC-DC
A SBS in
a home
CRN
The core
networkA mobile-small-
cell on a bus
A SBS on a factory
Car
communication
MBS
Figure 3: A multi-tier architecture for 5G networks with small-cells, mobile small-cells,
and D2D- and CRN-based communications.
Several two-tier261
architectures have262
been proposed for 5G263
networks, where a MBS264
stays in the top-tier265
and SBSs work under266
the supervision of the267
MBS in the lower tier.268
A macrocell covers269
all the small-cells of270
different types, e.g.,271
femtocell, picocell,272
and microcell (see273
Table 3), and both the274
tiers share an identical275
frequency band. The276
small-cell enhances the277
coverage and services278
of a macrocell, and the advantages of small-cells are mentioned at the end of this section. In addition, D2D279
communication and CRN-based communication enhance a 2-tier architecture to a multi-tier architecture; see280
Figure 3. Note that in this section, we confine ourselves on the deployment of small-cells under the cover of a281
macrocell; the discussion of CRN-based and D2D communications will be carried out in Sections 4.2 and 4.3,282
respectively.283
Cells Range Users
Femtocell 10-20 meters A few users
Picocell 200 meters 20 40
Microcell 2 kilometers > 100
Macrocell 30-35 kilometers Many
Table 3: Classification of the cells.
Wang et al. [107] suggested a way for separating indoor and284
outdoor users and using a mobile small-cell on a train or a bus.285
For separating indoor and outdoor users, a MBS holds large286
antenna arrays with some antenna elements distributed around the287
macrocell and connected to the MBS using optical fibers. A288
SBS and large antenna arrays are deployed in each building for289
7

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
communicating with the MBS. All UEs inside a building can have a connection to another UE either through the290
SBS or by using WiFi, mmWave, or VLC. Thus, the separation of users results in less load on a MBS.291
Wang et al. [107] also suggested to use a mobile small-cell that is located inside a vehicle to allow292
communication among internal UEs, while large antenna arrays are located outside the vehicle to communicate293
with a MBS. Thus, all the UEs inside a vehicle (or a building) appear to be a single unit with respect to the294
corresponding MBS, and clearly, the SBS appears as a MBS to all these UEs.295
In [28], a two-tier architecture is deployed as a process of network densification that is a combination of spatial296
densification (increasing the number of antennas per UE and MBS, and increasing the density of BSs) and spectral297
aggregation (using higher frequency bands > 3 Ghz). A tradeoff between the transmission power of a macrocell298
and the coverage area of small-cells is presented in [28]; i.e., on the one hand, if the transmission power of a299
macrocell is high, then many adjacent UEs to a small-cell may find themselves in the service area of the macrocell,300
and hence, it will decrease the coverage area of that small-cell. On the other hand, if the transmission power of301
a macrocell is low, then the coverage area of the small-cell will increase. Therefore, cell range expansion (i.e.,302
a biased handoff in the favor of small-cells) is carried out to serve more UEs by small-cells to which they are303
closer. Moreover, SBSs, deployed in offices or homes, can be used to serve outdoor users, e.g., pedestrians and304
low-mobility vehicles, in their neighborhoods, and the approach is called indoor-to-outdoor user service [28].305
Hossain et al. [61] presented a multi-tier architecture consisting of several types of small-cells, relays, and306
D2D communication for serving users with different QoS requirements in spectrum-efficient and energy-efficient307
manners. Interestingly, all of these architectures consider that UEs spontaneously discover a SBS. Zhang et308
al. [121] proposed a centralized system in which a MBS assists UEs to have connections to particular SBSs, thereby309
interference between UEs and SBSs is reduced. However, this approach overburdens the MBS.310
Advantage of the deployment of small-cells.311
• High data rate and efficient spectrum use: The small physical separation between a SBS and UEs (served by312
the same SBS) leads to a higher data rate and a better indoor coverage. Also, the spectrum efficiency increases313
due to fewer UEs in direct communication with a MBS [111].314
• Energy saving: The use of small-cells reduces the energy consumption of the network (by not involving MBSs)315
and of UEs (by allowing UEs to communicate at a shorter range with low signaling overhead) [51].316
• Money saving: It is more economical to install a SBS without any cumbersome planning as compared to a MBS,317
and also the operational-management cost is much lower than the cost associated with a MBS [41, 28].318
• The plug-and-play utility of small-cells boosts the on-demand network capacity [123].319
• Less congestion to a MBS: SBSs offload UEs from a MBS so that the MBS is lightly loaded and less congested,320
and hence, improve the system capacity [41].321
• Easy handoff : Mobile small-cells also follow the advantages of small-cells. Moreover, they provide an attractive322
solution to highly mobile UEs by reducing handoff time overheads, since a mobile small-cell is capable to do323
the handoff on behalf of all related UEs [28].324
Disadvantage of small-cells. Despite numerous prominent benefits as mentioned above, there are a few realistic325
issues such as implementation cost and operational reliability. The small-cells indeed impose an initial cost to the326
infrastructure, but less than the cost associated with a MBS. Moreover, a frequent authentication is mandatory due327
to frequent handoff operations. Further, an active or passive (on/off) state update of any small-cell would definitely328
result in frequent topological updates.329
Open issues in the deployment of 2-tier architectures using small-cells.330
• Interference management: The deployment of small-cells results in several types of interferences, as: inter-tier331
interference (i.e., interference from a MBS to a SBS, interference from a MBS to a SBS’s UEs, and interference332
from a SBS to a MBS’s UEs), and intra-tier interference (i.e., interference from a SBS to other SBSs’ UEs).333
Hence, it is also required to develop models and algorithms to handle these interferences [41, 38].334
• Backhaul data transfer: Though we have models to transfer data from a SBS to the core network, which we335
will discuss next in Section 4.1.1, an extremely dense-deployment of small-cells requires a huge amount of data336
transfer, and certainly, requires cost efficient architectures.337
4.1.1 Backhaul data transfer from small-cells338
Data transfer from a SBS to the core network is a challenging task, and in general, there may be three approaches339
to transfer (backhaul) data to the core network, as follows:340
8

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
• Wired optical fiber: by establishing a wired optical fiber link from each SBS to a MBS; however, it is341
time-consuming and expensive.342
• Wireless point-to-multipoint (PTMP): by deploying a PTMP-BS at a MBS that communicates with SBSs and343
transfers data to the core network.344
• Wireless point-to-point (PTP): by using directional antennas in line-of-sight (LOS) environments; hence, it345
provides high capacity PTP links (same as with wired optical fibers), at a significantly lower cost.346
Ge et al. [51] presented two architectures based on the wireless PTMP approach. In the first (centralized)347
architecture, all SBSs send data using mmWave to a MBS that eventually aggregates the received data and forwards348
the same to the core network using fiber. In the second (distributed) architecture, all small-cells cooperatively349
forward data using mmWave to a specified SBS that transfers data to the core network using fiber without the350
explicit involvement of a MBS.351
Ni et al. [88] proposed an adaptive backhaul architecture based on the wireless PTP approach and frequency352
division duplex for UL and DL channels. A tree structure is used, where the root node is connected to the core353
network using fiber, the leaf nodes represent UEs, and other nodes represent SBSs. The data is transferred from the354
leaf nodes to the root node that transfers the same to the core network. The bandwidth is selected dynamically for355
backhaul links, as per the current bandwidth requirements, interference conditions, and the network topology. A356
similar approach is also presented in [28].357
4.1.2 Two-tier architectures with self-healing property358
An automatic detection and recovery of a failed cell is an important issue in densely deployed multi-tier359
architectures. Wang and Zhang [111] provided three approaches for designing a self-healing architecture such360
as below:361
1. Centralized approach: A dedicated server is responsible for detecting a failed cell by measuring and analyzing362
abnormal behavior of users, e.g., received signal strengths (RSSs) at users and handoff by several users at any363
time from a particular cell. The server collects global information and reconfigures the failed cell. However, the364
approach suffers with a high communication overhead and a high computational cost.365
2. Distributed approach: Each SBS detects failed small-cells in neighborhoods by measuring and analyzing users’366
handoff behavior and the neighboring small-cells’ signals. Consequently, on detecting a failed cell, a SBS might367
increase the transmission power in order to incorporate users of the failed cell. However, the approach might368
not work efficiently in case that users are scattered sparsely.369
3. Local cooperative or hybrid approach: This approach combines the benefits of both the previous approaches,370
and therefore, minimizes the drawback. Essentially, two steps are utilized, namely distributed trigger and371
cooperative detection. In the distributed trigger, each SBS collects information about users’ behavior.372
Subsequently, a trigger message is sent to a dedicated server in case the received information thrives below373
a certain threshold. Hence, it does not require communication among small-cells. In the cooperative detection,374
the dedicated server takes the final decision based on the information received from several small-cells, resulting375
in a high accuracy and lower latency.376
4.2 Cognitive Radio Network based Architectures377
A cognitive radio network (CRN) [16] is a collection of cognitive radio nodes (or processors), called secondary378
users (SUs) that exploit the existing spectrum opportunistically. The SUs have the LEIRA (learning, efficiency,379
intelligence, reliability, and adaptively) property for scanning and operating on multiple heterogeneous channels380
(or frequency bands) in the absence of the licensed user(s), termed as primary user(s) (PUs), of the respective381
bands [98]. Each PU has a fixed bandwidth, high transmit power, and high reliability; however, the SUs work on a382
broad range of bandwidth with low transmit power and low reliability.383
A CRN in 5G networks is used for designing multi-tier architectures, removing interference among cells, and384
minimizing energy consumption in the network [63, 41, 110, 60, 72, 83].385
4.2.1 CRN-based architectures for 5G networks386
A CRN creates a 2-tier architecture, similar to architectures discussed in Section 4.1; however, it is assumed that387
either a MBS or a SBS has cognitive properties for working on different channels.388
9

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Hong et al. [60] presented two types of CRN-based architectures for 5G networks, as: (i) non-cooperative and389
(ii) cooperative CRNs. The non-cooperative CRN establishes a multi-RATs system, having two separate radio390
interfaces that operate at the licensed and temporary unoccupied channels by PUs, called cognitive channels.391
The SUs work only on cognitive channels and form a CRN, which overlays on the existing licensed cellular392
network. The two networks can be integrated in the upper layers while must be separated in the physical layer.393
This architecture can be used in different manners, as: (i) the cognitive and licensed channels are used by users394
near a MBS and users far away from the MBS, respectively, (ii) the cognitive and licensed channels are used for395
relaxed QoS and strict QoS, respectively. The cooperative CRN uses only a licensed channel, where SUs access the396
channel in an opportunistic fashion when the PU of the channel is absent. This architecture can be used in different397
manners, as: (i) a SBS communicates with a MBS using the licensed channel and provides service to its UEs via398
an opportunistic access to the licensed frequency band, (ii) a licensed channel is used to serve UEs by a SBS and399
the opportunistic access to the licensed channel is used to transfer backhaul data to the MBS.400
In short, the cooperative CRN [60] provides a real intuition of incorporating CRNs in 5G networks, where a401
SBS works as a SU, which scans activities of a macrocell and works on temporarily unoccupied frequency bands402
(known as spectrum holes [17]) by a PU to provide services to their UEs with minimally disrupting macrocell403
activities.404
A dynamic pricing model based on a game theoretic framework for cognitive small-cells is suggested in [65].405
Since in reality SBSs’ operators and MBSs’ operators may not be identical and small-cells’ UEs may achieve a406
higher data rate as compared to macrocells’ UEs, the pricing model for both UEs must be different.407
4.2.2 Interference Management using CRNs408
Huang et al. [63] provided an approach for avoiding inter-tier interference by integrating a cognitive technique409
at a SBS. The cognitive technique consists of three components, as: (i) a cognitive module, which senses the410
environment and collects information about spectrum holes, collision probability, QoS requirements, macrocell411
activities, and channel gains, (ii) a cognitive engine, which analyzes and stores the collected information for412
estimating available resources, and (iii) a self-configuration module, which uses the stored information for413
dynamically optimizing several parameters for efficient handoff, interference, and power management. Further,414
the channel allocation to a small-cell is done in a manner to avoid inter-tier and intra-tier interferences, based415
on Gale-Shapley [49] spectrum sharing scheme, which avoids collisions by not assigning an identical channel to416
neighboring small-cells.417
Wang et al. [110] suggested an approach for mitigating inter-tier interference based on spectrum sensing,418
spectrum sharing, and cognitive relay, where links between a MBS and its UEs are considered as PUs and links419
between a SBS and its UEs are considered as SUs. Cognitive techniques are used for detecting interference from420
a MBS to a SBS and vice versa, and a path loss estimation algorithm is provided for detecting interference from a421
small-cell’s UEs to a macrocell’s UEs. After detecting inter-tier interference, a small-cell shares spectrum with a422
macrocell using either overlay spectrum sharing scheme (i.e., SUs utilize unoccupied channels, and it is applicable423
when a MBS and a SBS’s UEs are very close or no interference is required by a macrocell’s UEs) or underlay424
spectrum sharing scheme (i.e., SUs and PUs transmit on an identical channel while restricting transmit power of425
SUs, and hence, resulting in a higher spectrum utilization).426
Note that a CRN can be used to support D2D communication and mitigate interferences caused by D2D427
communication, which we will see in Section 4.3.428
Advantages of CRNs in 5G networks.429
• Minimizing interference: By implementing a CRN at small-cells, cognitive small-cells can avoid interference430
very efficiently by not selecting identical channels as the channels of neighboring small-cells.431
• Increase network capacity: The spectrum holes can be exploited for supporting a higher data transfer rate and432
enhancing bandwidth utilization [83].433
Open issues. Usually, cellular networks are not energy-efficient as they consume energy in circuits, cooling434
systems, and also radiate in air. Hence, an energy-efficient deployment of a CRN in a cellular network is of435
utmost importance [60]. Further, there is a tradeoff between the spatial frequency reuse and the outage probability,436
which requires the selection of an efficient spectrum sensing algorithm [41].437
10

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
4.3 Device-to-Device Communication Architectures438
Device-to-Device (D2D) communication allows close proximity UEs to communicate with each other on a licensed439
cellular bandwidth without involving a MBS or with a very controlled involvement of a MBS. The standards440
and frameworks for D2D communication are in an early stage of research. In this section, we will review D2D441
communication networks in short. For a detailed review of D2D communication, interested readers may refer442
to [24].443
Challenges in D2D Communication.444
• Interference management: UEs involved in D2D communication, say D-UEs, face (or create) interference from445
(or to) other UEs, or from (or to) a BS, based on the selection of a DL or UL channel, respectively. The following446
types of interferences are investigated in [113]:447
– When using a DL channel: (i) interference from BSs in the same cell, (ii) interference from other co-channel448
D-UEs in the same cell, and (iii) interference from BSs and co-channel D-UEs from other cells.449
– When using a UL channel: (i) interference from all co-channel C-UEs 3 in the same cell and other cells, and450
(ii) interference from all co-channel D-UEs in the same cell and other cells.451
Proposed solutions: A simple solution may exist by implementing CRNs in D2D communication, as: D-UEs452
are considered as SUs and C-UEs are considered as PUs that should not be interfered. Consequently, any453
mechanism of CRNs can be implemented in D2D communication for interference removal.454
• Resource allocation: When UEs involved in D2D communication, it is required to allocate a sufficient amount of455
resources, particularly bandwidth and channels. However, the allocation of optimum resources to D-UEs must456
be carried out in a fashion that C-UEs do not have interference from D-UEs, and D-UEs can also communicate457
and exchange data efficiently [113, 77].458
Proposed solutions: SARA [33], frame-by-frame and slot-by-slot channel allocation methods [74], and a459
social-aware channel allocation [77].460
• Delay-sensitive processing: Audio, video streaming, and online gaming, which are natural in close proximity461
UEs, require real-time and delay-sensitive processing. Hence, it is required to consider delay-sensitive and462
real-time processing in D2D communication [109].463
Proposed solutions: Solutions based on channel state information (CSI) and QoS are provided in [109].464
• Pricing: Sometimes a D-UE uses resources (e.g., battery and data storage) of other UEs for relaying information,465
where the other UE may charge for providing its resources. Hence, the design of a pricing model is needed,466
thereby a D-UE is not charged more money than that involved to communicate through a MBS.467
Proposed solutions: Some solutions based on game theory, auction theory, and bargaining are suggested468
in [104].469
D2D communication types. D2D communication can be done in the following four ways [104], as follows:470
1. Device relaying with operator controlled link establishment (DR-OC): A UE at the edge of a cell or in a poor471
coverage area can communicate with a MBS by relaying its information via other UEs, which are within the472
stronger coverage area and not at the edge; see Figure 3.473
2. Direct D2D communication with operator controlled link establishment (DC-OC): Source and destination UEs474
communicate directly with each other without involving a MBS, but they are assisted by the MBS for link475
establishment; see Figure 3.476
3. Device relaying with device controlled link establishment (DR-DC): Source and destination UEs communicate477
through a relay without involving a MBS, and they are also responsible for link establishment; see Figure 3.478
4. Direct D2D communication with device controlled link establishment (DC-DC): Source and destination UEs479
communicate directly with each other without involving a MBS, and they are also responsible for link480
establishment; see Figure 3.481
Note that DR-OC and DC-OC involve a MBS for resource allocation and call setup, and hence, prevent interference482
among devices to some extent.483
Two types of coding schemes (or communication types) are described in [113]: (i) two-way relay channel484
(TRC), where a source and a destination communicate through a relay, and (ii) multiple-access relay channel485
(MRC), where multiple sources communicate to a destination through a relay with direct links. Note that the486
workings of DR-OC and MRC, and DR-DC and TRC are identical. Two types of node discovery and D2D487
3A UE that is not involved in D2D communication and communicates to a MBS, we call it a cellular user equipment (C-UE) in this section.
11