Handover-based Load Balancing Algorithm for 5G and Beyond Heterogeneous Networks

Full text

Handover-based Load Balancing Algorithm for
5G and Beyond Heterogeneous Networks
Abdussamet Hatipo˘
glu, Mehmet Bas¸aran, Mehmet Akif Yazici, L ¨
utfiye Durak-Ata
Information and Communications Research Group (ICRG)
Informatics Institute, Istanbul Technical University, Istanbul, Turkey
{hatipoglua, mehmetbasaran, yazicima, durakata}@itu.edu.tr
Abstract—In ultra-dense networks, an enormous number of
small base stations (BSs) are required to be deployed, and the
number of handovers (HOs) in heterogeneous networks (HetNets)
increases as a result. Mobility management with the increasing
number of HOs is a critical issue that requires uninterrupted
connectivity due to user mobility in HetNets. However, some BSs
in the network may be more heavily loaded than their neighbors.
Load balancing (LB) involves load transfer from an overloaded
BS to an under-loaded neighboring BS for the more load-balanced
network. HO and LB are two important issues to improve
network performance. In this paper, we propose an algorithm
that utilizes user speed and received signal reference power to
adapt HO margin (HOM) and time to trigger (TTT). Besides,
the proposed algorithm balances the loads among neighbor BSs.
Additionally, the proposed algorithm aims to reduce ping-pong
HOs (HOPP) and HO failure (HOF) ratios. The simulation results
show that the rates of HOPP and HOF are significantly reduced
by the proposed algorithm, thus improving network performance
under various mobile speeds. Moreover, the results show that the
proposed algorithm minimizes the standard deviation of BS loads
in the network. The proposed algorithm achieves load balancing
and reduces overload during user mobility in HetNets. Therefore,
there is more than 60% improvement in terms of the HOF,
and 63% more balanced network is achieved with the proposed
algorithm in the network.
Keywords—5G, Beyond 5G, handover, handover margin, het-
erogeneous networks, load balancing.
I. INTRODUCTION
In recent years, heterogeneous networks (HetNets) have
played an important role in increasing network performance
due to the densification of mobile users in certain sites.
HetNets are complicated networks due to the incorporation of
small cells in macro cells. Deploying a large number of small
cells within 5G and beyond 5G (B5G) networks are expected
to improve overall system performance by extending coverage
and improving user experience [1]. Mobility management is
crucial for the novel services envisioned in 5G and B5G.
Quality of service (QoS) can be improved through transferring
a portion of the traffic load from macro BSs to smaller cells
in an appropriate manner.
If a user equipment (UE) exits the coverage area of the
serving base station (BS) while it is in motion, the UE
must be switched to a neighboring cell without disrupting the
connection. This cell replacement process is called handover
(HO), where the user’s serving cell must be switched before
terminating the user’s previous connection [2]. HO is one of
the basic processes for enabling users to move freely across the
network. Since successful HO and QoS are important factors
in user satisfaction, it is very important to carry out the HO
process as quickly and smoothly as possible [3].
Densification of the network by increasing the number
of BSs reduces the load per BS, and thus, increases the
QoS. This deployment scenario leads to an increase in the
number of HOs for mobile users [4]. As the number of HOs
increases, the concept of load balancing (LB) becomes critical.
Communication link continuity of the UE with respect to
possible serving BSs is provided according to the received
signal strengths. Therefore, HO and LB are always important
operations to ensure service quality for mobile users [5].
Conditional HO was introduced in a technical document
[6]. HO control parameters (HOCP) include HO margin
(HOM) and time-to-trigger (TTT) to maintain communication
links with minimal service interruption during user movement
[7]. UEs can cope with HO failure (HOF) and HO ping-pong
(HOPP) problems by setting the HOM and TTT parameters
appropriately [8]. If HOM and TTT are picked too low, UEs
can experience HOPP, spurious HOs that increase signaling
load and hurt performance [9]. On the other hand, too high
HOM and TTT values lead to HOF. Adaptive schemes that
try to optimize HOCP based on the reference signal received
power (RSRP) and UE speed have been proposed [10]. In
this study, we combine such approaches with LB to enhance
network HO performance.
A. Related Work
Self-optimization and self-organization in cellular networks
has been attracting great interest from the research community
in recent years. Many studies focused on HO optimization
and introduced algorithms and schemes to improve network
performance [11]-[20].
In [11], the Mobility Robustness Optimization (MRO) use-
case of LTE is improved by taking into account a so-called
“soft metric” and adjusting HOM dynamically. This scheme
is shown to improve the convergence speed of MRO, which
typically requires handovers in the order of 1000 to achieve
reliable adjustment. In [12], HOM is also adjusted dynamically

with user speed, whose change is detected by an increase
in HOF events. The type of HOF determines whether to
increase (in case of early HO, ping pong, or wrong cell) or
decrease (in case of late HO) HOM. In [13], the number of
HOs and especially “low time-of-stays” experienced by highly
mobile users are reduced by predicting user mobility based
on the mobility patterns observed earlier in the network. This
improvement in HO performance comes at the expense of
spectral efficiency.
In [14], the authors proposed an algorithm that mitigates
frequent HOs for fast-moving users in HetNets. This algorithm
adjusts HOCP based on the ratio of HOPP, where UE is handed
over to the macro BSs in high mobility. In [15], a dynamic
HOCP algorithm was proposed to investigate HO types that
cause HOF. The results show that the dynamic HOCP al-
gorithm achieves lower HOF and HOPP rates compared to
fixed HOCP. As opposed to the studies mentioned earlier, [15]
also adjusts TTT in addition to HOM. However, UE speed, a
significant factor in system performance, is ignored.
In [16], the authors tried to solve the LB problem consid-
ering n-th tier nonadjacent neighbors of the overloaded cell.
The algorithm can load users from an overloaded cell into
suitable neighboring cells. In [17] the authors combined UE
association algorithms with traffic offload using selective sBS.
The working region is divided into center and edge regions.
In [18], the authors propose fuzzy logic-based scheme
exploiting user velocity and a radio channel quality to adapt
HOM for HO decision in a self-optimizing manner. In [19], the
authors propose a weighted fuzzy self-optimization approach
for the optimization of HOCP. In this approach, HO decision
relies on three attributes: signal-to-interference-plus-noise ratio
(SINR), traffic load of the BSs, and UE velocity. The objective
of the proposed algorithms in [18] and [19] is to reduce
the number of redundant HOs and the HOF ratio. In [20],
mobility enhancement based on virtual cell design with dense
deployment is proposed to provide seamless coverage and
improve HO performance. The relationship between HOCP
and HO performance in ultra-dense networks with mobility is
studied.
B. Main Contributions and Outline
In this paper, we propose an algorithm that adjusts HOCP
to improve overall network performance and reduce negative
effects on the network under various UE mobilities. The major
contribution of this algorithm is to minimize the HOF rate
and maintain connection links between serving BS and mobile
UEs. HOCP are adjusted continuously after any change in UE
speed.
Another contribution of this study is that it adaptively
adjusts HOCP while taking into account the loads of the candi-
date BSs. Some of BSs may be heavily loaded, while other BSs
experience relatively light loads. To balance the loads among
neighbor BSs, the proposed algorithm is designed. Thanks to
this proposed algorithm, the load distribution among the BSs
in the network becomes more stable. The performance of the
proposed algorithm is analyzed and evaluated via simulations.
Figure 1: An illustration of a typical HetNet system
The rest of this paper is organized as follows. Section
II describes the main parameters. The system measurements
used for HO are also described. Section III introduces the
algorithm. Section IV discusses the simulation results and the
performance of the proposed algorithm. Finally, Section V
presents the main conclusions of the study.
II. SYSTEM MODEL
This section describes the system model of the HetNets
with numerous mBSs and sBSs, as shown in Fig. 1. Three
sets of parameters are considered for the HO process:
i) system metrics, such as RSRP and the speeds of UEs,
ii) control parameters, HOM and TTT,
iii) performance indicators which show the efficiency of
the proposed method.
These parameters are described below.
A. System Metrics
The propagation environment loss and shadowing effect
are accounted for path loss (PL) while calculating RSRP. The
propagation loss (L) model for sBS, considering a carrier
frequency of 2000 MHz with BS antenna height of 15 meters
is defined as [21]
L(dB) = 128.1 + 37.6 log10(R),(1)
where Ris the distance between the BS and the UE in
kilometers. Similarly, for mBS environment, the propagation
model with a carrier frequency of 900 MHz and BS height of
45 meters is given as
L(dB) = 95.5 + 34.01 log10(R).(2)
Log-normally distributed shadowing propagation model with
standard deviation Xαof 10 dB is assumed in PL determina-
tion as per the 3GPP specification [21] and is given by
PL(dB) =L+Xα.(3)

Figure 2: PL and RSRP according to distance
Finally, the RSRP is calculated as
RSRP(dBm) =TP+TG−PL,(4)
where TPis the transmitter power and TGis the transmitter
antenna gain. In Fig. 2, the effect of distance to the RSRP and
PL is shown.
B. Control Parameters
Each UE periodically measures the RSRP values of neigh-
boring cells. It sends measurement reports to the BS only when
the RSRP of a neighbor cell is at a better signal level than the
signal of the serving cell. Based on these measurement reports,
the serving BS makes a decision whether to perform HO. If
the RSRP value of one of the neighboring cells is greater than
the RSRP value of the serving cell by a margin of HOM for
a duration of TTT, the BS sends the HO command to the UE.
The condition that triggers HO is expressed as
Pneigh > Pserv +HOM,(5)
where Pserv indicates the RSRP value of the connected cell,
while Pneigh represents the RSRP value of the neighbor cell
with the strongest signal. If the condition given in (5) continues
to be satisfied for a duration of TTT, HO is performed. TTT
helps to avoid disconnection by mitigating HOPP events.
C. Performance Indicators
1) HOPP Rate: If a UE reverts back to the source cell
within one second after making a decision for transferring
to the neighboring cell, the HOPP effect is said to occur.
HOPP rate is calculated by the ratio of the number of HOPP
(NHOP P )to the total number of HOs (NHO ). Then the HOPP
rate is defined as [22]
HOPP =NHOP P /NHO .(6)
2) HOF Rate: When an unsuccessful HO operation occurs
or the connection is lost, this is noted and this is called HOF.
HOF is the ratio of the number of HO failures (NHOF )to the
total number of HOs (NHO )and defined as [22]
HOF =NHOF /(NH O +NH OF ).(7)
Algorithm 1: LB-Aware HO Algorithm
1Inputs: Vc,Vth1,Vth2;
2Outputs: HOM, TTT;
3if Vc≤Vth1then
4Update HOM, TTT acoording to Table II;
5else if Vth1< Vc≤Vth2then
6Update HOM, TTT acoording to Table II;
7else if Vth2> Vcthen
8Update HOM, TTT acoording to Table II;
9end
10 // Before HO occurs, the capacity status of the target
BS is examined. //;
11 if C1< αCth then
12 Perform HO to the 1st neighbor BS;
13 else if αCth ≤C1< Cth then
14 // LB Algorithm for the 2nd neighbor BS //;
15 if C2< αCth then
16 Perform HO to the 2nd neighbor BS;
17 else
18 Perform HO to the 1st neighbor BS;
19 end
20 else if C1=Cth then
21 Perform HO to the mBS with best RSRP;
22 end
III. PROPOSED ALGORITHM
The proposed algorithm is given in Algorithm 1 and can
accommodate different mobility modes in daily life such as
walking, cycling, and driving. The user is classified with
respect to its speed using pre-determined threshold values,
Vth1, and Vth2. The proposed algorithm consists of two stages.
In the first stage, the HOCP are adaptively adjusted according
to the speed of UEs. Afterward, before HO occurs, the capacity
status of the target BS is examined. The proposed algorithm
is described in detail below.
The proposed algorithm enables the UE to select the HOM
and TTT parameters required for HO from the associated speed
group. This algorithm is run periodically to ensure HOCP is
adaptively set appropriately based on the instantaneous speed
of the UE, Vc. Thereby, the algorithm aims to successfully
perform the HO process. Furthermore, average speeds of
walking, cycling, and driving can be assumed to be around 5,
25, and 60 km/h [23], and thus, the threshold values of Vth1
and Vth2are picked as 10 km/h and 45 km/h, respectively. In
this study, it is assumed that the instantaneous speed of users
is known.

Figure 3: An illustration of a sample HetNet topology (UEs
with different colors correspond to different mobile speeds)
After completing the first stage, which determines HOCP
according to the instantaneous speed of the user, the system
directs BS to the second stage to control the capacity of the
network, before HO occurs. Here, C1refers to the current
load in terms of the number of users being served by the
1st neighbor BS, and C2refers to the current load of the
2nd neighbor BS. While the 1st neighbor BS represents the
neighboring cell with the strongest signal, the 2nd neighbor BS
represents the neighboring cell with the next strongest signal.
Cth is the maximum user capacity of that BS. The αvalue
plays a role in balancing the load of the BSs. In the proposed
algorithm, the load control of BS is performed as follows:
•HO to 1st neighbor is performed if C1is less than
αCth.
•If C1is between αCth and Cth, the UE tries to
connect to the the 2nd neighbor BS. If C2is lower
than αCth, HO is performed to the 2nd neighbor BS.
Otherwise, HO is performed to the 1st neighbor BS.
•If C1equals Cth, BS cannot serve more UEs. In this
case, HO is attempted to be performed to the mBS
with the best RSRP in that region.
As an example, consider the following scenario. A UE
is moving at a speed of 30 km/h and is considered to have
cycling speed, and the appropriate HOCP are assigned to the
UE. When the RSRP value of one of the neighbors BSs is
greater than the RSRP value of the serving BS by at least the
HOM value and this condition is met for the TTT period, the
load status of the target BS is examined. Suppose that αvalue
is selected as 0.8. In this case, HO is performed to the 1st
neighbor BS if its load is no more than 80% of its capacity.
Otherwise, the algorithm looks for a 2nd neighbor BS as a
candidate for HO. If there is an eligible neighbor BS, then the
Table I: Simulation parameters
Parameters Value
mBS sBS
Number of BSs 3 9
Capacity of BS (users) 200 50
Capacity multiplier (α) 0.5 0.8
Cell radius (m) 500 200
Carrier frequency (MHz) 900 2000
Antenna height (m) 45 15
Trans. power (dBm) 46 30
Trans. antenna gain (dB) 15
Number of simulations 100
Simulation time (s) 600
Number of users 1000
Vth1,Vth2(km/h) 10, 45
Simulation area (m) 1750×1750
UE speed (km/h) 0, 5, 25, 60
HOM (dBm) 0, 1, 2, 3, 4, 5, 6, 7, 8
TTT (ms) 32, 64, 128, 256, 512, 1024
load of that 2nd neighbor BS is checked. If the load of the
2nd neighbor BS does not exceed αCth, HO is performed to
the 2nd neighbor BS. But, if the load of the 2nd neighbor BS
is also above αCth, HO is performed to the 1st neighbor BS,
unless its load has reached full capacity, in which case HO is
performed to the mBS with best RSRP in that region.
IV. SIMULATION RESULTS
In this section, we perform simulations of a HetNet using
a stand-alone simulator written in Python. Then, we evaluate
the performance achieved by the proposed algorithm through
simulation results.
A. Simulation Parameters and Environment
We consider a HetNet that consists of 3 mBS and 9 sBS
in an area of 1750 ×1750 m2. The mBS are deployed in a
hexagonal layout with 500-meter inter-site distance, and the
sBS are deployed inside the macro cell coverage areas, as
shown in Fig. 3. We set the transmission power to 46 dBm
for the mBS and 30 dBm for the sBS [24].
One thousand UEs are distributed uniformly over the
topology. In terms of mobility, we model 40% of the UEs to be
stationary, and the rest are assumed mobile (20% walking, 20%
cycling, 20% driving). A random motion type is selected for
the mobile UEs. When mobile UEs reach the edge of the cell,
they continue to stay in the cell by changing their directions.
UE speed is set to 5, 25, and 60 km/h, respectively, for walking,
cycling, and driving users [23]. Other simulation parameters
are summarized in Table I.

Figure 4: HOPP performance related to TTT and HOM
Table II: System parameters for different velocity
Vc≤Vth1Vth1< Vc≤Vth2Vc> Vth2
HOM (dBm) 6 4 2
TTT (ms) 512 128 32
For an acceptable load threshold, we set the αvalue to 0.5
for mBS and 0.8for sBS. The algorithm initiates LB if at least
one cell exceeds its specified capacity. Once LB is triggered,
the algorithm calculates other BS’s loads.
In this study, the first step is the determination of appro-
priate HOM and TTT values for each speed category from the
possible values listed in Table I. The best values for each speed
category found through this procedure are listed in Table II.
These values are obtained via extensive simulations for various
HOM and TTT value pairs.
B. Performance Analysis
We evaluate the overall performance of the proposed algo-
rithm using the HOPP ratio, HOF ratio, and load distribution.
Fig. 4 illustrates the performance in terms of the HOPP ratio.
The received signals fluctuate more at low speed, and therefore,
the HOPP rate is relatively higher at low speed. However, low
HOPP rates are achieved, especially in high-speed scenarios.
The best HOCP values are determined for each speed group
using these results, and are given in Table II.
In Fig. 5, we see the impact of LB on load distribution
accross the 12 BSs. Fig. 5a illustrates the loads of each
BS within 5 minutes of network operation without an LB
algorithm. BS1, BS5, and BS9, which happen to be the macro
BSs, are extremely overloaded with loads reaching 98% on
occasion. However, BS6 and BS8, which are neighbors of
these BSs, are not adequately used, with loads of 26% and
22%, respectively. The difference between the maximum and
minimum capacity ratio of the network is 76%. On the other
hand, when the proposed algorithm is applied, overloaded BSs
release some of their loads to low-load neighboring BSs as
shown in Fig. 5b. The load of the densest BS, which happens to
be BS2 in this scenario, reaches a maximum value of 88%, and
the load of BS6, which is the most lightly loaded among BSs,
Figure 5: Load distribution of the network: (a) without an LB
algorithm and (b) with the proposed LB algorithm
Figure 6: Standard deviation of load among BSs in network.
turns out to be 49% after 5 minutes. Thus, the gap between
the maximum and minimum loads observed in the network is
reduced from 76% to 39%, which means the variance in the
load among the BSs in the network is reduced and the network
becomes much more balanced.
Fig. 6 shows the performance of the proposed algorithm
in terms of the standard deviation of BS loads in the network.
The standard deviation of loads with the proposed algorithm
in the network becomes smaller. This situation is observed
from the difference of standard deviation values which are
0.19 and 0.07, respectively, after 10 minutes. As a result, the
load distribution among the BSs in the network has become
63% more balanced.
Fig. 7 depicts the HOF rates under the setting given in
Table II. HOF can occur due to various reasons, resulting in the
failure of connection between UE and the target BS. According
to the results, the proposed algorithm reduces the HOF ratio
from 9% to 3.5% at walking speed, from 12% to 4% at cycling
speed, and from 13% to 4.5% at driving speed. Therefore, there
is more than 60% improvement in HOF performance.

Figure 7: HOF rate for the mobility of 5, 25, 60 km/h.
V. CONCLUSION
In this paper, an algorithm is proposed for HetNets in order
to improve network performance, in terms of the HOF ratio,
HOPP ratio, and load distribution. In the proposed algorithm,
to avoid HOPP among the BSs, HOCP are adaptively adjusted
according to the speed of UEs. The proposed algorithm is
shown to be effective in evening out the loads of the BSs,
improving HOF performance by avoiding heavily-loaded BSs.
Thanks to the proposed algorithm, the load among the BSs
in the network becomes more uniform. This proposed algo-
rithm can be considered as a promising solution to the HO
management in 5G and B5G networks.
ACKNOWLEDGEMENT
This work is supported by Istanbul Technical University
(ITU) Scientific Research Projects Coordination Unit with
Project number MYL-2019-42135.
This work is supported in part by Istanbul Technical
University (ITU) Vodafone Future Lab under Project No.
ITUVF20180701P06.
REFERENCES
[1] M. Polese, M. Giordani, S. Rangan and M. Zorzi, “Improved handover
through dual connectivity in 5G mmWave mobile networks,” IEEE J.
Sel. Areas Commun., vol. 35, no. 9, pp. 2069-2084, Sep. 2017.
[2] R. Arshad, H. Elsawy, L. Lampe and M. J. Hossain, “Handover
rate characterization in 3D ultra-dense heterogeneous networks,” IEEE
Trans. Veh. Technol., vol. 68, no. 10, pp. 10340-10345, Oct. 2019.
[3] D. Castro-Hernandez and R. Paranjape, “Optimization of handover
parameters for LTE/LTE-A in-building systems,” IEEE Trans. Veh.
Technol., vol. 67, no. 6, pp. 5260-5273, Jun. 2018.
[4] M. Shafi et al., “5G: a tutorial overview of standards, trials, challenges,
deployment, and practice,” IEEE J. Sel. Areas Commun., vol. 35, no.
6, pp. 1201-1221, Jun. 2017.
[5] P. Mu˜
noz, R. Barco and S. Fortes, “Conflict resolution between load
balancing and handover optimization in LTE networks,” IEEE Commun.
Lett., vol. 18, no. 10, pp. 1795-1798, Oct. 2014.
[6] Performance evaluation of conditional handover, document 3GPP TSG
RAN WG2 meeting 103bis, Tdoc R2-1814052, Chengdu, China, Intel
Corporation, Oct. 2018.
[7] R. P. Ray and L. Tang, “Hysteresis margin and load balancing for
handover in heterogeneous network,” Int. J. Future Comput. Commun.,
vol. 4, no. 4, pp. 231-235, 2016.
[8] B. K. S. Fatty and Po-Chiang Lin, “Mobility robustness optimization
in wireless mobile networks,” Int. Conf. Ubiquitous Future Netw., pp.
624-629, Aug. 2015.
[9] I. de la Bandera, P. Mu ˜
noz, I. Serrano and R. Barco, “Adaptive cell
outage compensation in self-organizing networks,” IEEE Trans. Veh.
Technol., vol. 67, no. 6, pp. 5231-5244, Jun. 2018.
[10] K. Vasudeva, M. Simsek, D. L´
opez-P´
erez and ˙
I. G¨
uvenc¸ , “Analysis of
handover failures in heterogeneous networks with fading,” IEEE Trans.
Veh. Technol., vol. 66, no. 7, pp. 6060-6074, Jul. 2017.
[11] G. Hui and P. Legg, “Soft metric assisted mobility robustness optimiza-
tion in LTE networks,” Int. Symp. Wireless Commun. Syst. (ISWCS), pp.
1-5, Aug. 2012.
[12] K. Kitagawa, T. Komine, T. Yamamoto and S. Konishi, “A handover
optimization algorithm with mobility robustness for LTE systems,”
IEEE Int. Symp. Personal, Indoor and Mobile Radio Commun., pp.
1647-1651, 2011.
[13] B. Sas, K. Spaey and C. Blondia, “A SON function for steering users
in multi-layer LTE networks based on their mobility behaviour,” IEEE
81st Veh. Technol. Conf. (VTC Spring), Glasgow, pp. 1-7, 2015.
[14] M. M. Hasan, S. Kwon and S. Oh, “Frequent-handover mitigation in
ultra-dense heterogeneous networks,” IEEE Trans. Veh. Technol., vol.
68, no. 1, pp. 1035-1040, Jan. 2019.
[15] A. Alhammadi, M. Roslee, M. Y. Alias, I. Shayea and S. Alraih,
“Dynamic handover control parameters for LTE-A/5G mobile commu-
nications,” Adv. Wireless Optical Commun. (RTUWO), pp. 39-44, Dec.
2018.
[16] N. O. Tuncel and M. Koca, “Joint mobility load balancing and inter-cell
interference coordination for self-organizing OFDMA networks,” IEEE
81st Veh. Technol. Conf. (VTC Spring), Glasgow, pp. 1-5, 2015.
[17] F. Muhammad, Z. H. Abbas and F. Y. Li, “Cell association with load
balancing in nonuniform heterogeneous cellular networks: coverage
probability and rate analysis,” IEEE Trans. Veh. Technol., vol. 66, no.
6, pp. 5241-5255, Jun. 2017.
[18] K. Da Costa Silva, Z. Becvar and C. R. L. Frances, ”Adaptive hysteresis
margin based on fuzzy logic for handover in mobile networks with
dense small cells,” IEEE Access, vol. 6, no. 9, pp. 17178-17189, Mar.
2018.
[19] A. Alhammadi, M. Roslee, M. Y. Alias, I. Shayea, S. Alriah and A.
B. Abas, “Advanced handover self-optimization approach for 4G/5G
hetNets using weighted fuzzy logic control,” Int. Conf. Telecommun.
(ConTEL), pp. 1-6, Sep. 2019.
[20] N. Meng, H. Zhang and H. Lu, “Virtual cell-based mobility enhance-
ment and performance evaluation in ultra-dense networks,” Wireless
Commun. Netw. Conf., Doha, pp. 1-6, 2016.
[21] 3GPP, “Evolved universal terrestrial radio access (E-UTRA); require-
ments for support of radio resource management”, Release 15, tech.
specification (TS) 36.814, 2017.
[22] A. B. Ozyurt, M. Basaran and L. Durak-Ata, “Impact of self-
configuration on handover performance in green cellular networks,”
Advances Wireless Optical Commun. (RTUWO), pp. 194-197, Dec.
2018.
[23] “Typical everyday speeds”. [Online, Accessed: 17 Sep 2020] Available:
https://www.bbc.co.uk/bitesize/guides/zq4mfcw/revision/1
[24] M. Ardanuc, M. Basaran and L. Durak-Ata, “Energy efficient base
station deployment in ultra dense heterogeneous networks via poisson
hole process,” Signal Process. Commun. Appl. Conf. (SIU), pp. 1-4,
2019, in Turkish.