2020 IEEE REGION 10 CONFERENCE (TENCON)

Osaka, Japan, November 16-19, 2020

Traffic aware sleeping strategies for Small-Cell Base

Station in the Ultra dense 5G Small Cell Networks
Kuna Venkateswararao Pravati Swain
Department of Computer Science and Engineering Department of Computer Science and Engineering
National Institute of Technology Goa National Institute of Technology Goa
Goa 403401, India Goa 403401, India
[email protected] [email protected]

Abstract—The 5G ultra-dense small cell network plays a key meters with 20dBm transmission power and supports 30 UEs.
role in the future generation of mobile networks. It provides high Similarly, microcell covers over 2000 meters of the area which
data rate, seamless coverage and reliable services for wireless can be an indoor or outdoor environment and supports 2000
communication in an ultra-dense network. The dense deployment
of small cells is needed to control forthcoming traffic demands UEs.
which leads to enhance the operational cost and reduces the The Ultra dense small cell networks [5], [6] (UDSCN) are
energy efficiency. One way to improve the energy efficiency is by solution to achieve the demands of current cellular network.
using the sleeping strategy of small cell by transferring the traffic However, the densely deployment of heterogeneous small
load of a small cell to other small cells. This work proposes an cell in the cellular network will increase energy usage and
Initial Connection algorithm for establishing an initial association
between the UEs and small base stations (s-BSs) while considering operational cost of the cellular network. To meet the energy
the UE preference. The Initial Connection algorithm creates a efficiency and coverage capacity demands of the current cel-
connected network. Moreover, the proposed Load Sharing Based lular architecture, we propose a novel small cell deployment
Sleep Approach (LSBSA) algorithm performs small cells sleeping approach. The proposed method is based on load sharing and
on the connected network which results in the deployment of s- the small cell sleeping strategy by considering the dynamic
BSs. The proposed Initial Connection and LSBSA algorithms are
implemented and evaluated in MATLAB for different mobile data data traffic control. The sleeping policy of s-BSs is divided
traffic. Also, various UEs distribution scenarios are considered. into four different modes namely as On mode, Standby mode,
The results are demonstrated that the proposed approaches Sleep mode and Off mode. In the On mode, the s-BSs use
improve network performance in terms of energy efficiency of maximum power supply while standby represents light sleep.
the small cell network by deploying the optimal number of active The s-BSs are in deep sleep during the Sleep mode. In the
s-BSs.
Index Terms—5G, Ultra dense network, Small cell network Off mode, the s-BSs have zero power usage. The power
consumption of the cellular network components is given
I. I NTRODUCTION in [7], [8]. It is mentioned that 60% power is consumed at
The last decade has seen exponential growth in the cellular the base station while 20%, 15% and 5% are used for mobile
access network [1]. The 5G cellular networks are key enabler switching, core transmission and retailer, respectively.
for wireless communication of various devices in different The rest of the paper is arranged as follows: Section II
technologies such as the Internet of Thing (IoT), Unmanned describes the relevant research work. The considered network
Aerial Vehicles (UAV) and Wireless sensor networks. More- model and problem definition are provided in Section III.
over, the number of mobile users and usage of data volume has In Section IV, the proposed algorithm is explained in two
increased due to social networking and e-learning applications. sub-problems. The performance evaluation is elaborated in
Ultra-dense networks (UDNs) [2], [3] where the number of Section V. Finally, the conclusion is discussed in Section VI.
small cells are more than active user equipments (UE) are new
II. R ELATED W ORK
paradigm for providing high data rate, massive connectivity,
and control dynamic traffic in the future generation cellular To deal with dynamic mobile traffic and massive connectiv-
network. ity in the ultra-dense 5G network area, small cells and device
A Small Cell (SC) network [4] is a very promising technique to device (D2D) communication techniques are used in [9],
in the UDN which covers a small geographical area filled with [10]. Dense deployment of the small cells (s-BS) are used to
UEs and provides mobile services to the UEs. The small cells control dynamic mobile traffic, provide massive connectivity
are deployed under indoor (office, home) and outdoor (railway and a high data rate. However, dense s-BSs have higher energy
stations, shopping malls) environments. A small cell has fully usage with a few s-BSs being underutilized. So, to improve
functioning small base station (s-BS). The s-BS consumes the energy usage and s-BSs utilization without compromising
low power and short-range wireless accessed with different the Quality of Service of the UE, the redundant s-BSs are
coverage capacities. These small cells can reduce the burden switched off as proposed in [11].
of the macro base station by providing hotspots to the UEs. Similarly in [12], the random sleeping policy is used which
They are classified as microcell, femtocell, picocell. A picocell is based on the average data rate and coverage probability of
covers 200 meters of range with 33dBm transmission power the s-BSs. The green SC planning (GSCP) in [13] is used to
and supports 100 UEs in an indoor and outdoor deployment control the traffic flow in smart cities with a dynamic traffic
scenario. Femtocell is deployed indoors with a range of 100 pattern. In GSCP, a connected network is formed between the

978-1-7281-8455-5/20/$31.00 ©2020 IEEE 102

Authorized licensed use limited to: Ericsson. Downloaded on April 22,2025 at 17:13:48 UTC from IEEE Xplore. Restrictions apply.
UEs and s-BS based on the signal-to-interference-plus-noise
ratio (SINR). Then, the redundant connections between the
UEs and s-BS are removed in the connected network. Here,
authors do not consider the preference of the UEs and s-BSs
to form the small cell (SC). Here, the SC formation is a time
consuming process.
In [14], the initial matching connection algorithm (IMCA)
for SC formation and ”turn-off if possible” algorithm (TIPA)
improves the energy efficiency of the SC network by turning-
off the redundant s-BSs. The TIPA algorithm switch off one
randomly selected s-BS , and reformulate the SC network. If
the network is successfully created1 then the process is con-
tinued until it is possible to switch off the s-BS. If the network
is not successfully generated, the network is recovered. This
is a time consuming process to improve the energy efficiency
of the network.
In this paper, an efficient small cell formation and energy
efficient small cell deployment algorithms are proposed to
address the above limitations. The proposed Initial Association Fig. 1: Network Model
algorithm for small cell formation considers the preference of
the UEs and small cells. Load Sharing Based Sleep Approach
calculate the degree of the ith s-BS and j th UE defined by
(LSBSA) algorithm is proposed to minimize the small cell Mt t
ΣN t
i=1 ai,j and Σj=1 ai,j . The Signal Interference Noise Ratio
deployment in the small cell network. The proposed algo-
(SINR) of the j th UE to the ith s-BS under a given traffic
rithms enhance the energy efficiency, throughput, and spectral t
pattern t is represented by γi,j which is given in Eq. 1. Here,
efficiency of the ultra-dense small cell network.
in Eq.1:
The contributions of this paper are summarized as follows: t
• An Initial Association algorithm is proposed for small • T Pi,j is the assigned transmission power of the j th UE
th
cell formation which establishes the initial connection from i s-BS
between the UEs and s-BSs by considering the UE • Gti,j (Gtjk ) is the channel gain between the j th UE and
preference to choose higher signal small base stations. ith s-BS (k th s-BS)
t
• The Load Sharing Based Sleep Approach (LSBSA) al- • IPi,k is the inference power from the j th UE from k th
gorithm is proposed for minimizing the deployment of s-BS
small base stations to improve the energy efficiency in • σ 2 is the thermal noise power
the small cell network. • ati,k represents j th UE associated to k th s-BS
III. N ETWORK M ODEL t
T Pi,j Gti,j
t
Consider an Ultra dense small cell network area managed γi,j = N t t Gt
(1)
Σk=1,j6=i ai,k IPi,k i,j + σ2
by a macro base station located at the center of the network.
The macro base station manages a number of small base The j th UE is associated with the ith s-BSs when its γi,j
t
value
th
stations (s-BSs) and UEs. We assume that the small cell is greater than the defined SINR threshold (∆ ). Therefore,
network has a set of N = {n1 , n2 , n3 ..nN } small base the Shannon capacity gained by the k th UE is given as
t t t
stations (s-BSs) and M = {m1 , m2 , m3 ..mM } UEs in which Ri,k = log 2(1 + γi,k ), where γi,k is the SINR given in
UEs are distributed under various traffic patterns denoted by Eq. 1. Moreover, we can measure the energy efficiency of the
T. Consider traffic patterns are uniform, random and cluster small cell network (SCN) under the traffic pattern t, which
t
distribution. The s-BSs are required to satisfy the demand of is presented in Eq. 2. The total SCN throughput (SN T ) and
t
the UEs. However, placing more s-BSs increases the operation energy consumption (SN C ) are calculated using Eq. 3 and
cost and energy consumption. Thus, it is necessary to mitigate Eq. 4. In Eq. 4 for a particular traffic pattern t:
the energy consumption through the s-BS sleeping approach. • Cit is the energy consumption of the ith s-BS
The s-BS operates with two modes, namely active mode and • pmax
i is the maximum transmission power of the ith s-BS
sleep mode. In the active mode, the s-BS works with full max
• S is the maximum number of UEs serviced by the
functionality while minimum power is consumed in the sleep ith s-BS
mode. Figure 1 shows the assumed network model of the small • pai /psi is the circuit power consumption of the ith ac-
cell network. tive/sleep s-BS
Under a given traffic pattern t, the state of the s-BS is t St T
represented by S t = [sti ]1∗N where sti = 1 if the s-BS is SEE = N t (2)
SN C
in active mode, otherwise sti = 0. The association between
the j th UE and the ith s-BS is denoted by At = [ati,j ]M ×N , t
SN N t M t t
T = Σi=1 si Σj=1 ai,j Ri,k (3)
where ati,j = 1 if the j th UE is associated with the ith s-
t N t t a t s
BS, otherwise ati,j = 0. Based on the At values, we can SN C = Σi=1 [si (Ci + pi ) + (1 − si )pi ] (4)
ati,j
1 All UEs are associated with the any one s-BS in the small cell network where, Cit = pmax
i ΣM
j=1 S max

103

Authorized licensed use limited to: Ericsson. Downloaded on April 22,2025 at 17:13:48 UTC from IEEE Xplore. Restrictions apply.
A. Problem Formulation • Firstly, calculate the SINR (γij ) between the UEs
The small cell network (SCN) comprises a set of s-BSs and and s-BSs by using Eq. 1. If the SINR of the j th
UEs where the UEs are connected to the s-BSs. To reduce UE to ith s-BS is greater than the predefined SINR
the communication cost and energy consumption of the SCN, threshold value(∆th ) then the j th UE is connected
sleeping small cell technique is used without compromising to the ith s-BS (aij = 1). The node degree of the
the UE requirements. Here, the problem can be treated as ith s-BS ( Dibs ) and j th UE (Djue ) is incremented
a multi objective optimization problem where the energy by one.
efficiency is maximized and the deployment of s-BSs in the • As illustrated in the above step, the connections
SCN is minimized. The energy efficiency and deployment of between UEs and s-BSs are established and the
s-BSs in the SCN under various traffic patterns are given in node degrees of the UEs and s-BSs are obtained.
Eq. 5 and Eq. 6. However, to minimize the number of deployed • If the j th UE is connected to the ith s-BS (aij =
s-BSs can enhance the energy efficiency of the SCN, because 1) and the degree of the UE is greater than one
a large portion of energy is consumed at the base station. then the s-BSs with the least SINR from the j th
UE are identified and the connection between the
M aximize O1 = ΣTt=1 SEE
t
(5) j th UE to ith s-BS is kept. Thereafter, the j th UE
connections are removed from the other s-BSs. The
M inimize O2 = ΣTt=1,iN Sit (6)
node degree of the j th UE and s-BSs are updated.
Based on the given Eq. 5 and Eq. 6, the problem can be defined • If the j th UE is connected to the ith s-BS (aij = 1)
as follows: and the degree of the UE (U Ejd ) is equal to one, it
is added to the connected network (Conn N et).
• Moreover, if the node degree of ith s-BS (BSid ) ex-
F1 = [Obj1 , Obj2 ]
ceeds the upper limit value (S max ) then the UEs in
s.t C1 : sti {0, 1}, ∀t, i the s-BS are sorted in ascending order with respect
C2 : atij {0, 1}, ∀t, i, j to their SINR from s-BS. Then, the S max UEs
C3 : ΣN t from the sorted list is connected to the s-BS and
i=1 aij = 1, ∀t, j (7)
t the remaining (U E r ) UEs are connected to the next
C4 : ΣM t
j=1 aij ≤ S
M ax
, ∀t, i choice of the s-BSs. Later, U Ejd and Conn N et
t
C5 : ΣM t t
j=1 aij ≤ si , ∀t, i
are updated. The above process is repeated until
t the network is connected successfully.
C6 : ΣN M t
i=1 ((Σi=1 aij )/S
M ax
) < Tload , ∀t
2) The s-BS sleeping Approach: The output of the proposed
where C1 and C2 are the binary values for UE association
Algorithm 1 is to establish the initial connection between
and s-BS deployment respectively. The C3 value ensures that
the UEs and s-SBs, resulting in a connected network.
each UE is associated with a single s-BS and C4 defines the
Here, the proposed Load Sharing Based Sleeping Ap-
degree of the s-BS which is less than or equal to the S max
proach (LSBSA) algorithm 2 provides a solution to
value. C5 ensures that the UE associates with any s-BS when
the second subproblem which operates the states of
it is active state. C6 ensures the traffic load the s-BS is less
the s-BSs. The connected network along with the set
than the predefined load threshold. The Tload represents the
of UEs and s-BS are given as input to the LSBSA
maximum traffic load of the small cell.
algorithm while its output is the deployment of s-BSs.
IV. P ROPOSED S YSTEM The algorithm 2 is described as follows:
The main objective of the proposed approach is to maximize • Firstly, the states (S) of the s-BSs are identified.
the energy efficiency and minimize the deployment of s-BSs If any ith s-BS is active (Si == 1) then the load
in a SCN. For minimizing the deployment of s-BSs, apply of the s-BS (BS load ) is calculated using Eq.8 and
the s-BSs sleeping approach, without compromising the QoS it is added to the deployment of the s-BSs set
of the UEs. Thus, the problem can be divided into two sub (Dep BSs). Then, the calculated s-BS loads are
problems such as initial association of UEs and s-BS sleeping sorted in ascending order as BS load .
approach. The macro base station recognizes and classifies • The least loaded s-BS (BS least ) is identified from
the traffic patterns using pattern matching algorithms. Once a the sorted list (BS load ) and local search is per-
traffic pattern is identified by the base station, the proposed formed within the transmission range of BS least .
algorithms can be applied for the deployment of s-BSs. The The found s-BSs are stored in the neighbor s-BSs
proposed Initial Association algorithm provides the initial list (N B list ).
connection between the UEs and s-BSs, and Load Sharing • The N B list list is sorted in ascending order
Based Sleeping Approach (LSBSA) algorithm for small cell with respect to their loads and the first pair s-
pair
sleeping. BSs in the sorted list is taken as BSab (
pair pair
1) Initial Association of UEs: To establish the initial con- BSa and BSb ). Thereafter, the load of the
pair
nection between the UEs and s-BSs which is shown in BS least is shared to the BSab s-BSs using Eq.9.
least pair
Algorithm 1. In the Algorithm 1, the number of UEs • If the load of the BS is shared with BSab
least
and number of s-BSs are given as inputs M and N, s-BSs successfully, then the state of the BS
respectively. The output is a fully connected network. s-BS is changed to sleep and the Dep BSs list
The algorithm is described in the following steps: is updated. Else, the process is aborted and the

104

Authorized licensed use limited to: Ericsson. Downloaded on April 22,2025 at 17:13:48 UTC from IEEE Xplore. Restrictions apply.
 Algorithm 1: Initial Association() Algorithm 2: LBSA()
Input: The UEs M , the s-BSs N , SINR threshold Input: The UEs M , the s-BSs N , Connected Network
(∆th ) (Conn N et)
Output: Connected Network (Conn N et) Output: Deployment of s-BSs (Dep BSs)
pair
Initialization: U E d = ∅, BS d = ∅, U E r = ∅, Initialization: Dep BSs = ∅, BSab =
load least
Conn N et = ∅ ∅, BS = ∅, BS =
while Conn N et == empty do ∅, N B list = ∅
for i=1 to N do while Dep BSs == empty do
for j=1 to M do for i=1 to N do
γi,j = Calculate the SINR between the Ni if (Si == 1) then
and Mj using Eq. 1. BSiload =Calculate the load of the Ni s-BS
if (γi,j > ∆th ) then using Eq. 8
aij = 1, U Ejd = U Ejd + 1, Add Ni s-BS to Dep BSs
BSid = BSid + 1 end
end end
end Sort BS load
end BS least =Identify the least load of the s-BS from
for i=1 to N do BS load
for j=1 to M do for i=1 to N do
if (aij == 1&U Ejd >1) then Do the local search from BS least s-BS within
Keep the connection of Mj UE with TR
smallest SINR s-BSi if (Si == 1) then
Remove connections from other s-BSs N B list = Ni
Update the U Ejd , BSid end
else end
Add aij to Conn N et BSab pair
= Sort N Blist and get first pair of s-BSs
end from N B list
end Share the load of the BS least to the BSab pair
s-BSs
if (BSid >S max ) then using Eq. 9
Sort the BSid based on SINR of the UEs if (BS least load shared successfully) then
U Eir = S max − BSid Switch the state of BS least s-BS to sleep state
U E r are connected next choice of the Update the Dep BSs list
s-BSs else
Update the BSid and Conn N et Cancel the process and recover the network
end end
end end
end

TABLE I: Simulation parameters

network is recovered. This process is repeated until Parameter Parameter
it is possible to share the load of the s-BS with Bandwidth for s-BS 20 MHz
other s-BSs. pmax 40 dBm
loadbs = ΣN M max S max 30
i=1 ((Σi=1 aij )/S ) (8)
pa 39 dBm
BSapair ps 35 dBm
SLa = [ BS least ] σ -174 dBm/Hz
BSapair
+ BSbpair (9) ∆th -5 dB
SLb = SLa − BSapair
Where SLa and SLb are the shared percentage of traffic by
BSapair and BSbpair , respectively. A. Evaluation Scenario
V. P ERFORMANCE E VALUATION For the evaluation scenario, random traffic pattern is
This section provides the evaluation scenario and results selected as an example which is appropriate for real-time
obtained from the proposed algorithms. Consider the eval- traffic pattern of UEs. Figure 3a represents the evaluation
uation area as 1000 × 1000m2 with uniform distribution of scenario which consists of 200 UEs and 24 s-BSs. The
24 s-BSs under different traffic patterns. Here, adopted traffic primary goal of the proposed work is to improve energy
patterns are uniform, random and cluster which are shown in efficiency and minimize the deployment of s-BSs in small
Figure 2. Table I represents the simulation parameters of the cell network (SCN). By diminishing the s-BSs, we can
evaluation scenario which are referred from [15]. The proposed enhance the energy efficiency of the network since more
algorithms are simulated in Matlab and executed on Windows power consumption takes place at the s-BSs.
7 system (Intel i5 process and 8GB RAM).

105

Authorized licensed use limited to: Ericsson. Downloaded on April 22,2025 at 17:13:48 UTC from IEEE Xplore. Restrictions apply.
 1 1 1

0.9 0.9
0.9
0.8 0.8
0.8
0.7 0.7
0.7
0.6 0.6

Distance

Distance
Distance

0.5 0.5 0.6

0.4 0.4
0.5
0.3 0.3
0.4
0.2 0.2
0.3
0.1 0.1

0 0 0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Distance Distance Distance

(a) Uniform traffic pattern (b) Random traffic pattern (c) Cluster traffic pattern
Fig. 2: Traffic patterns

1 1 1
UE UE UE
s-BS Active s-BS
0.9 0.9 Active s-BS 0.9
Sleep s-BS
0.8 0.8 0.8

0.7 0.7 0.7

0.6 0.6 0.6

Distance

Distance

Distance
0.5 0.5 0.5

0.4 0.4 0.4

0.3 0.3 0.3

0.2 0.2 0.2

0.1 0.1 0.1

0 0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Distance Distance Distance

(a) Distribution of 200 UEs and 24 s-BSs. (b) Connected network between the UEs (c) No.of active s-BS of the connected
and s-BSs using Initial Assocation Algo- network after applying LSBSA algorithm.
rithm.
Fig. 3: Proposed approach

The proposed Initial Association and LSBSA algorithms

24
are used for s-BS deployment. The distribution of UEs and Proposed Approach
s-BSs are input for the Initial Assocation algorithm which TIPA
22
establishes the initial connection between the UEs and s-BSs
as shown in the Figure 3b. The result of the algorithm 1 20
is a fully connected network (active 24 s-BSs) without any
No.of active s-BSs

redundant connections. In Figure 3c, some of the s-BSs 18

are working in underload condition which will raise the
communication cost and power usage. The LSBSA algorithm 16
is used for sinking the power consumption of the s-BSs by
using the sleep approach. Here, the load of each s-BS is 14

calculated and the s-BS having the least load among them is
selected i.e., BS least . Then, local search is performed within 12

the transmission range of BS least and the neighbours of the

10
BS least s-BSs are found. 0 5 10 15 20 25 30
No.of iterations

Thereafter, the first pair of s-BSs from the neighbour’s list Fig. 4: No.of active s-BSs of the Proposed Approach and TIPA
is selected and the BS least load is transferred to them. The algorithm
state of the BS least s-BS is changed to a sleep state and this
same process is repeated until it is possible to share the load of
the s-BSs. After execution of algorithm 2, there are 11 active
s-BSs out of the 24, the remaining 13 s-BSs goes under sleep In TIPA, s-BSs are randomly switched off one at a time
state, as given in the Figure 3c. We observe that the proposed and the network (Connected graph between the UEs and s-
algorithm 2 can reduce the deployment of s-BSs and improve BSs) is reformulated until it is not possible to switch off
the energy efficiency of the SCN. any more s-BSs in the network. During the switch-off, there
is no constraint for selection of the s-BS and it takes n-1
B. Evaluation Results iterations to select the one s-BS where n is the number of
The performance of the proposed approach is evaluated in base stations in the network. To overcome the limitations of
terms of the number of active s-BSs, network energy efficiency TIPA, the proposed approach (Initial Association and LSBSA)
and convergence time. Moreover, we compare with Turn-off if considers the load of the s-BS for switch-off/sleep mode.
possible to switch-off s-BS algorithm (TIPA) [14] algorithm. Then, transfers the traffic load to the neighbouring s-BSs with

106

Authorized licensed use limited to: Ericsson. Downloaded on April 22,2025 at 17:13:48 UTC from IEEE Xplore. Restrictions apply.
specific condition which are discussed in the Section IV. VI. C ONCLUSION
To deal with dynamic mobile traffic and massive con-
nectivity in the ultra-dense 5G network by using the dense
4
Proposed Approach
deployment of a heterogeneous small cell. Dense deployment
TIPA of small cells will reduce the operational cost and energy
Network Energy Efficiency (bps/Hz/Joule)

3.5
efficiency of the small cell network. To address these issues,
3 we propose the Initial Association and Load Sharing Based
Sleep Approach (LSBSA). The Initial Association algorithm
2.5 establishes an initial association between the UEs and small
base stations which results in the formation of a connected
2
network without any redundant connections. The results of the
1.5 Initial Association algorithm is given as input for the LSBSA.
The LSBSA performs the small cell sleeping approach based
1
on load sharing, to minimize the small cell deployment in the
0.5
ultra-dense small network. The proposed method obtains high
data rate, enhances the energy efficiency in less number of
0
0 5 10 15 20 25 30
iterations as compare to the existing methods.
No.of iterations
R EFERENCES
Fig. 5: NEE of the Proposed approach and TIPA algorithm [1] O. G. Aliu, A. Imran, M. A. Imran, and B. Evans, “A survey of self
organisation in future cellular networks,” IEEE Communications Surveys
& Tutorials, vol. 15, no. 1, pp. 336–361, 2012.
The primary assessment of the number of the proposed [2] S. Chen, Xiang Ji, C. Xing, Z. Fei, and Hualei Wang, “System-level
performance evaluation of ultra-dense networks for 5G,” in TENCON
approach and TIPA is performed on 200 UEs with 24 s-BSs 2015 - 2015 IEEE Region 10 Conference, 2015, pp. 1–4.
as shown in Figure 3a. Figure 4 represents the number of [3] M. Kamel, W. Hamouda, and A. Youssef, “Ultra-dense networks: A
active s-BSs with the respective iterations of the proposed survey,” IEEE Communications Surveys & Tutorials, vol. 18, no. 4, pp.
2522–2545, 2016.
approach and TIPA. The number of Active s-BSs by the [4] X. Ge, S. Tu, G. Mao, C.-X. Wang, and T. Han, “5G Ultra-Dense
proposed approach, gradually decreases from 24 to 11 with Cellular Networks,” IEEE Wireless Communications, vol. 23, no. 1, pp.
less number of iterations as compare TIPA. UE outage 72–79, 2016.
[5] C. Yang, J. Li, and M. Guizani, “Cooperation for spectral and energy
probability is guaranteed in the TIPA algorithm. It will efficiency in ultra-dense small cell networks,” IEEE wireless communi-
minimize the small cell network coverage capacity. cations, vol. 23, no. 1, pp. 64–71, 2016.
[6] K. Venkateswararao, P. Swain, C. Christophorou, and A. Pitsillides,
“Dynamic selection of virtual small base station in 5G Ultra-Dense
Figure 5 shows the comparison of the proposed approach Network using Initializing Matching Connection Algorithm,” in 2019
and TIPA algorithm in terms of energy efficiency. It is ob- IEEE International Conference on Advanced Networks and Telecommu-
served from Figure 5 that the energy efficiency of the proposed nications Systems (ANTS). IEEE, 2019, pp. 1–6.
[7] S. Vadgama and M. Hunukumbure, “Trends in green wireless access
approach increases as compared to the TIPA due to less networks,” in 2011 IEEE International Conference on Communications
number of active s-BSs. Workshops (ICC). IEEE, 2011, pp. 1–5.
[8] W. Hashim and N. M. Anas, “Power adjustment in dense deployment
network: An empirical study,” in TENCON 2014-2014 IEEE Region 10
TABLE II: Convergence Times of the TIPA and Proposed Conference. IEEE, 2014, pp. 1–5.
Approach [9] C. Christophorou, A. Pitsillides, and I. Akyildiz, “Celec framework for
reconfigurable small cells as part of 5G ultra-dense networks,” in IEEE
Number of UEs Proposed Approach TIPA International Conference on Communications (ICC). IEEE, 2017, pp.
1–7.
Convergence Time (sec) Convergence Time (sec)
[10] P. Swain, C. Christophorou, U. Bhattacharjee, C. M. Silva, and A. Pit-
100 3.617331 5.136465 sillides, “Selection of UE-based virtual small cell base stations using
200 5.129713 7.810696 affinity propagation clustering,” in 14th International Wireless Commu-
300 6.476149 9.424599 nications & Mobile Computing Conference (IWCMC). IEEE, 2018, pp.
500 7.624871 11.70388 1104–1109.
1000 8.124158 13.45278 [11] M. Feng, S. Mao, and T. Jiang, “Base station ON-OFF switching
in 5G wireless networks: Approaches and challenges,” IEEE Wireless
Communications, vol. 24, no. 4, pp. 46–54, 2017.
[12] C. Liu, B. Natarajan, and H. Xia, “Small cell base station sleep strategies
Table II presents the convergence time of the proposed for energy efficiency,” IEEE Transactions on Vehicular Technology,
approach and TIPA algorithm [14] on different size of network vol. 65, no. 3, pp. 1652–1661, 2015.
[13] L. Zhou, X. Hu, C. Zhu, E. C.-H. Ngai, S. Wang, J. Wei, and V. C.
such as 100 UEs, 200 UEs, 300 UEs, 500 UEs and 1000 UEs. Leung, “Green small cell planning in smart cities under dynamic traffic
The convergence time of the proposed approach is reduced by demand,” in 2015 IEEE Conference on Computer Communications
35% to 40% for 100 to 300 UEs. and 50% to 60% for 300 to Workshops (INFOCOM WKSHPS). IEEE, 2015, pp. 618–623.
[14] Y. Xu, J. Chen, D. Wu, and W. Xu, “Toward 5G: a novel sleeping
500 UEs as compare with TIPA. From Table II, It is observed strategy for green distributed base stations in small cell networks,”
that the proposed method has a higher convergence rate as in 2016 12th International Conference on Mobile Ad-Hoc and Sensor
compared to TIPA. The convergence rate increases due to less Networks (MSN). IEEE, 2016, pp. 115–119.
[15] D. S. Baum, J. Hansen, J. Salo, G. Del Galdo, M. Milojevic, and
number of iterations needed for deployment of s-BSs. From the P. Kyösti, “An interim channel model for beyond-3G systems: extending
above discussion, we noticed that the proposed approach will the 3GPP spatial channel model (SCM),” in 2005 IEEE 61st Vehicular
improve the energy efficiency, convergence rate and minimize Technology Conference, vol. 5. IEEE, 2005, pp. 3132–3136.
the s-BS deployments in the UDSCN.

107

Authorized licensed use limited to: Ericsson. Downloaded on April 22,2025 at 17:13:48 UTC from IEEE Xplore. Restrictions apply.