Implementation of Particle Swarm Optimization Technique for Enhanced Outdoor Network Coverage in Long Term Evolution Network in Port Harcourt, Nigeria

EJERS, European Journal of Engineering Research and Science
Vol. 2, No. 5, May 2017
DOI: http://dx.doi.org/10.24018/ejers.2017.2.5.346 36
Abstract—This paper describes the development of optimized
model for urban outdoor coverage in Long Term Evolution
(LTE) network at 2300 MHz frequency band in Port Harcourt
urban region, Nigeria. Signal attenuation and fluctuation
remain amongst the major channel impairments for mobile
radio communication systems. This arises as a result of model
incompatibility with terrain and Line of Sight (LOS) obstruction
of the channel signals. Some path loss models such as Okumura-
Hata, COST 231, Ericsson 999, Egli and ECC-33 models were
evaluated for suitability and compared with the modified model
for the environments. The models were based on data collected
from LTE base stations at three geographical locations in Port
Harcourt namely- Rumuokoro, Eneka and Ikwerre roads
respectively. The simulation was implemented using MATLAB
R2014a software. The modified model was further optimized
with some selected parameters such as Root Mean Square Error
(RMSE) and Mean Absolute Error (MAE) using Particle Swarm
Optimization (PSO) technique. The results obtained gave rise to
3.030dB for RMSE and 0.00162dB for MAE respectively. The
results obtained from the PSO optimized model demonstrated a
better performance which is suitable for cell coverage planning
and smooth handoff processes.
Index Terms—Quality of Service; Path Loss; Mean Absolute
Error; Root Mean Square Error Models; Long Term Evolution;
Particle Swarm Optimization.
I. INTRODUCTION
Recently, the use of mobile wireless system became the
most popular technology as a result of lots of services
available in mobile cellular phone [1]. Some of these include
voice call, video call, conference call, data based services and
other multimedia applications. Despite the increase in
number of base stations to augment high rate of subscribers,
the quality of service delivery by the network providers still
remains in poor state [2]. It is observed that the poor quality
of services experienced could be due to several challenges
experienced in the network channel [3]. On that note, it
Published on May 19, 2017.
A. O. Akande is a Lecturer II at the department of Electrical and
Electronic Engineering, Federal University of Technology, Owerri, Nigeria
(e-mail: akandeoluk@ gmail.com). He is currently pursuing his Doctor of
Philosophy at Ladoke Akintola University of Technology, Ogbomoso, Oyo
State, Nigeria.
Dr. O. C. Nosiri is a Lecturer I at department of Electrical and Electronic
Engineering, Federal University of Technology, Owerri (email:
buchinosiri@gmail.com). His area of specialization is wireless and cellular
communication Engineering.
Dr. C. K. Agubor is a senior lecturer at the department of Electrical and
Electronic Engineering, Federal University of Technology, Owerri (e-mail:
aguborcosy@yahoo.com).
C.R. Okpara is a lecturer II at the department of Electrical and Electronic
Engineering, Federal University of Technology, Owerri, Nigeria (e-mail:
chineduokpara@yahoo.com).
becomes paramount to identity these problems and propose
solutions through research analysis.
The communication channel experience Rayleigh fading as
a result of obstruction to Line of Sight (LOS), resulted in
reflection, absorption and scattering of transmitted signal [4],
[5]. The knowledge of the propagation characteristic of a
mobile radio channel is essential for designing any wireless
communication system in a given region [6]. One of the most
important problems of propagation environment is path loss
[5]. Propagation path loss has great influence on the quality
of service of a mobile communication system, the transmitted
signal power get reduced with increased distance from base
station to mobile station. Therefore, accurate estimation of
propagation path loss is important factor for the good design
and planning of mobile system [7]. The quality and coverage
reliability of a wireless network design depends on the
accuracy of the path loss model [4].
There are many propagation path loss models proposed in
previous works to predict coverage. These model are not
sufficient for accurate analysis of the path loss on other
terrain, the suitability of such models differs with different
environment [8]. The objective of optimization is to make
sure the network operates close to the original design in terms
of handoff points, coverage, received signal strength and
Quality of Service (QoS) [2], [3]. In order to overcome the
problem of path loss, the parameters of certain models must
be adjusted with reference to the targeted locations. This will
enable the system achieve minimum error between
Transmitted Signal Strength (TSS) and Received Signal
Strength (RSS) [2], [9].
The performance of Long Term Evolution (LTE) networks
recently deployed in Port Harcourt region, Nigeria, is
investigated in this study. LTE remains one of the latest
technology development from the cellular 3rd Generation
(3G) to 4th Generation (4G) services [10], primarily designed
to meet the requirements of 4G technology. The Technology
Implementation of Particle Swarm Optimization
Technique for Enhanced Outdoor Network Coverage in
Long Term Evolution Network in Port Harcourt, Nigeria
Akande A. Olukunle, Nosiri C. Onyebuchi, Agubor K. Cosmos, and Okpara C. Reginald

Vol. 2, No. 5, May 2017
was based on Universal Mobile Telecommunication System
(UMTS), developed by the 3rd Generation Partnership
Project (3GPP). It is also known as Evolved Universal Mobile
Telecommunication Service Terrestrial Radio Access (e-
UTRA) [11]-[13]. It uses Wideband Code Division Multiple
Access (W-CDMA) as a standard [11]. LTE has downlink
and uplink of 100 Mbps and 50 Mbps respectively, scalable
bandwidth from 1MHz to 20MHz and supports multitude of
user types and high spectra efficiency [10]. LTE uses
multicarrier modulation such as Orthogonal Frequency
Division Multiple Access (OFDMA) and Multiple Input
Multiple Output (MIMO) antennal system at the transmitter
and receiver. This enables 4G to achieve higher spectral
efficiency and higher data rates [11].
Optimization is the process of obtaining the best solution
to a system function efficiently as a result of the modification
of existing models. The optimizations of these existing
models have been carried out by many researchers in other
regions, but some of them are based on Least Square (LS)
method. Other optimization methods such as Genetic
Algorithm (GA), Particle Swarm Optimization (PSO) and
Bee Colony Algorithm (BCA), have not been commonly used
relative to the reviewed literature. The particle swarm
optimization uses the ideas of bird flocking and its survival
of fitness by randomly chosen positions and velocities [2]. In
this paper, we investigated the performance of some
propagation models while appropriate modified model was
developed for the terrain. PSO technique was further
deployed to obtain the best model that suits the environment
and optimize for urban outdoor coverage in LTE network.
II. METHODOLOGY
The drive test routes survey was carried out at each
location before the actual commencement of the
measurements. Accessible routes were selected to avoid
being stranded during measurements in any of the selected
locations. The massive measurements were taking by driving
through the selected road in each location in order to obtain
the path loss and signal strength in those areas. The
measurement taken was on LTE, the base station height is 35
meters, mobile station height is 1.5 meters and the operating
frequency is 2300 MHz. The TEMS phone, Global
Positioning System (GPS), Power Supply Unit (PSU) were
connected to the laptop containing the installed TEMS 11.0
investigation software arranged inside the drive vehicle. The
measurements were made as the vehicle moved in the already
planned routes, the signal strength was observed on the
computer screen showing established calls and end calls. The
RSS and other information were recorded on the log files.
A. Large-Scale Propagation Models
This section discusses some commonly used empirical
models used to predict power attenuation between base
station and mobile station. Their approach in terms of
complexity and accuracy differs from one location to another.
B. Okumura-Hata
Okumura – Hata model is one of the commonly used
propagation path loss models that suit the urban area
propagation. The frequency range is between 150 𝑀𝑀𝑀𝑀𝑀𝑀 −
1500 𝑀𝑀𝑀𝑀𝑀𝑀 the base station antenna height is limited to 30m
- 200 m, and transmission distance (d) is between 1 𝑘𝑘𝑘𝑘 −
20 𝑘𝑘𝑘𝑘. The mobile antenna height ranges from 1 𝑚𝑚 − 10 𝑚𝑚
[14], [15]. Okumura-Hata model is given as;
LP(urban)dB = Ao − 13.82 log10 hbt − a(hcf) + [44.9 −
6.55 log hbt] log10 𝑑𝑑𝑡𝑡 (1)
where
Ao = 69.55 + 26.16 log fc
fc = is the carrier frequency (in MHz)
hbst = base station antenna height (in metres)
hmcf = mobile antennal height (in metres)
a(hmcf) = mobile antenna height correction factor
𝑑𝑑𝑡𝑡 =base station to mobile separation distance (in km)
For a large city, the mobile antenna correction factor
(a(hmb) is given by
a(hcf) = 8.29 [log(1.54hmb)]2
− 1.1 dB (2)
for fc ≤ 300 Mhz
a(hcf) = 3.201 [log(11.75hmb)]2
− 4.97 dB (3)
for fc ≥ 300 Mhz
For a small or medium size city, the mobile antenna
correction factor is
a(hcf) = 1.11[log(fc) − 0.7]hmb − [1.56 log(fc) − 0.8] (4)
Hata formula for suburban area path loss is modified as;
LP(suburan) (dB) = LP(urban)dB − 2[log
fc
28
]2
− 5.4 (5)
For rural area, the formula is modified [8]
LP(rural)(dB) = LP(urban)dB − 4.78[log fc]2
+ 18.33[log fc] −
40.98 (6)
C. Electronic Communication Committee (ECC-33)
ECC-33 is an extended Okumura-Hata model extrapolated
from original measurements by Okumura and modified its
assumptions to have close values on models. The original
experiment of Okumura was carried out at the suburban areas
of Tokyo [16]. The Authors refer to urban areas subdivided
into large city and medium city categories and gave
correction factors to suburban and open areas. The built-up
characteristics of Tokyo city are quite different to those found
in typical European suburban areas. The path loss model is
defined as shown in (7), [17].
PLP(ECC−33)(dB) = Pfsp + Pbm − Gtx − Grx (7)
where,
Pfsp is free space attenuation,
Pbm is basic median path loss,
Gtx is transmitter antenna (BS) height gain factor
Grx is received antenna (MS) height gain factor
The parameters are defined by [2] as;

Vol. 2, No. 5, May 2017
Pfsp = 92.4 + 20 log(d) + 20 log(f) (8)
Pbm = 20.41 + 9.83 log(d) + 7.894 log(f) + 9.56 [log(f)]2
(9)
Gtx = log �
htx
200
� [13.98 + 5.8(log d)2
] (10)
For medium sized cities
Grx = [42.57 + 13.7 log (f) ] [ log (hr) − 0.585] (11)
For large cities
Grx = 0.759 (hr) − 1.862 (12)
d = distance base station and mobile station (km)
hbs = base station antenna height (m)
hms = mobile antenna height (m)
D. Egli Model
The model was introduced by Egli John in 1957 to predict
the path loss in higher terrain environment. Egli model is
suitable for cellular communication scenarios where one
antenna is fixed and other is mobile [19]. The model is
applicable to scenarios where the transmission has to go over
an irregular terrain at frequency from 40 MHz to 900 MHz
[20]. Egli model is not applicable to the area where some
vegetative obstruction, such as trees or shrubbery is in the
middle of the link [19]. The Eglis model expressed by [1],
[20] as
Lp(dB) = K0 − 20 log hb + 76.3 − 10 log hm , (13)
for hm ≤ 10
Lp(dB) = K0 − 20 log hb + 85.9 − 10 log hm , (14)
for hm ≥ 10
Where,
K0 = 20 log 𝑓𝑓𝑐𝑐 + 40 log R
hb = height of the base station antenna (m)
hm = height of the mobile station antenna (m)
R =distance from base station antenna (km)
𝑓𝑓𝑐𝑐 = frequency of transmission
E. Ericsson 9999 Model
The model stands on the modified Okumura-Hata model to
give room for changing in parameters according to the
propagation environment [17]. This model is used for
frequency up to 1900 MHz. Path loss according to Ericsson
9999 model is evaluated by [19] as;
LP (ERIC)(dB) = a0 + a1 log(d) + a2 log hb +
[a3 log(hb) log(d)] − 𝑃𝑃0 + g(fc) (15)
Where,
𝑃𝑃0 = 3.2 [log(11.75 hr)2
]
g(fc) = 44.49 log(f) − 4.78[log(f)2
]
f = frequency in MHz,
hb = transmitter antenna height (m) and
hr = receiver antenna height (m)
1
Source: [17]
The default values of these parameters a0 , a1 , a2 , a3 for
different terrain are contained in Table I, it shows the asterisk
(*) values of parameter a0 , a1 based on the Least Square (LS)
method.
TABLE I: THE PARAMETER VALUE FOR DIFFERENT TERRAIN FOR
ERICSSON 9999 MODEL1
Environment a0 a1 a2 a3
Urban 36.20 30.20 12.00 0.10
Suburban 43.20* 68.93* 12.00 0.10
Rural 45.95* 100.6* 12.00 0.10
F. COST 231- Hata model
The European Co-operative for Scientific and Technical
research (EURO-COST) formed the COST-231 working
committee to develop COST-231- Hata model. The COST-
231 is the extension of the Hata model restricted to the
following range of parameters: frequency of1500 −
2000 𝑀𝑀𝑀𝑀𝑀𝑀, base station antenna height of 30 𝑚𝑚 − 200 𝑚𝑚,
mobile antenna height is between1 𝑚𝑚 − 10 𝑚𝑚, and separation
distance between transmitter antenna and receiver antenna of
1 𝑘𝑘𝑘𝑘 − 20 𝑘𝑘𝑘𝑘. It is designed to be used in the operating
frequency band between 1500 MHz to 2000 MHz. The basic
equation for path loss in dB is given in [4], [8] as;
LP(COST) dB = A0 − a(hmb) + [44.9 − 6.55 log hms] log d +
CAC (16)
Where,
A0 = 46.3 + 33.9 log fc − 13.82 log hbt
CAC = Area correction factor
LP = path loss in (dB)
𝑓𝑓𝑐𝑐 = carrier frequency (MHz)
hbt = base station antenna height (m)
hms = mobile antennal height (m)
a(hmb) = mobile antenna height correction factor
d = mobile antenna separation distance (km)
CM = 0 𝑑𝑑𝑑𝑑 for medium sized city and suburban areas, and
3 𝑑𝑑𝑑𝑑 for urban centres
For a large city, the mobile antenna correction factor is
given by
a(hmb) = 8.29 [log(1.54hmb)]2
− 1.1 dB (17)
for fc ≤ 300 Mhz
a(hmb) = 3.2 [log(11.75hmb)]2
− 4.97 dB (18)
for fc ≥ 300 Mhz
For a small or medium size city, the mobile antenna
correction factor is
a(hmb) = 1.1[log(fc) − 0.7]hm − [1.56 log(fc) − 0.8] (19)
G. Log-Distance Propagation Model
The mobile radio propagation models are derived by
employing a combination of empirical and analytical
methods. The empirical approach is based on analytical
expressions created by a set of measured data [8]. The

Vol. 2, No. 5, May 2017
wireless communication signal propagation may be
characterized by some factors such as signal power
attenuation, shadowing and fading. Hence, the path loss
model (LP (𝑑𝑑𝑖𝑖)) as a function of distance is given by [18].
LP (𝑑𝑑𝑖𝑖) 𝛼𝛼 (
𝑑𝑑
𝑑𝑑0
)𝑛𝑛𝑛𝑛
(20)
LP (𝑑𝑑𝑖𝑖) = LP (𝑑𝑑𝑜𝑜) (
𝑑𝑑
𝑑𝑑0
)𝑛𝑛𝑛𝑛
(21)
If we taking logarithm of (21)
LP (𝑑𝑑𝑖𝑖) = LP (𝑑𝑑𝑜𝑜) + 10(𝑛𝑛𝑛𝑛)𝑙𝑙𝑙𝑙𝑙𝑙(
𝑑𝑑
𝑑𝑑0
) (22)
Where, np is path loss exponent, 𝑑𝑑 is the distance
between the transmitter and receiver, 𝑑𝑑0 is the reference
distance.
Therefore,
𝑛𝑛𝑛𝑛 =
LP �𝑑𝑑𝑖𝑖�− LP (𝑑𝑑 𝑜𝑜)
10𝑙𝑙𝑙𝑙 𝑙𝑙(
𝑑𝑑
𝑑𝑑0
)
(23)
To compensate for random shadowing effect that can result
from terrain, the modified power expression is given by [18]
as;
LP (𝑑𝑑𝑖𝑖) = LP (𝑑𝑑𝑜𝑜) + 10(𝑛𝑛𝑛𝑛)𝑙𝑙𝑙𝑙𝑙𝑙 �
𝑑𝑑
𝑑𝑑0
� + 𝑋𝑋𝜎𝜎𝑥𝑥
(24)
Where, 𝑋𝑋𝜎𝜎 is a zero- mean Gaussian distributed random
variable (in dB) with standard deviation 𝜎𝜎𝑥𝑥 in (dB). Path
loss exponent(𝑛𝑛𝑛𝑛), can be determine in a mean square sense
using regression analysis [20] as;
𝑛𝑛𝑛𝑛 =
∑ [LP (𝑑𝑑 𝑜𝑜)−LP �𝑑𝑑𝑖𝑖�]2N
i=1
∑ 10 log (di d0⁄ )N
i=1
(25)
where, LP (𝑑𝑑𝑜𝑜) is the measured path loss and LP (𝑑𝑑𝑖𝑖) is
the predicted path loss at any distance 𝑑𝑑𝑖𝑖, N is the number of
data points. The standard deviation is given as;
𝜎𝜎𝑥𝑥 = (
1
N
∑ [LP (𝑑𝑑𝑜𝑜) − LP (𝑑𝑑𝑖𝑖)]2N
i=1 )
1
2 (26)
The expression LP (𝑑𝑑𝑜𝑜) − LP (𝑑𝑑𝑖𝑖) is an error term with
respect to 𝑛𝑛𝑛𝑛, from (25) and the sum of the mean square error,
𝑒𝑒(𝑛𝑛𝑛𝑛) is given as;
𝑒𝑒(𝑛𝑛𝑛𝑛) = ∑ [LP (𝑑𝑑𝑜𝑜) − LP (𝑑𝑑𝑖𝑖)]2N
i=1 (27)
The derivative of (27) can be equated to zero while
minimizing the mean square error to solve for 𝑛𝑛𝑛𝑛:
𝜕𝜕𝜕𝜕(𝑛𝑛𝑛𝑛)
𝜕𝜕𝜕𝜕
= 0
H. Model Data Analysis and Measurements
The path loss measured in Port Harcourt urban area has
been estimated using the propagation models. These models
were selected for the fact that they can predict the mean path
loss as a function of various parameters. The regression
analysis between the measured and the predicted path loss
data against distance for Rumuokoro road (location 1), Eneka
road (location 2) and Ikwerre road (location 3) respectively
are presented in Table II.
TABLE II: PORT HARCOURT URBAN AREA DATA ANALYSIS
Dist. (m)
Average
RSS (dBm)
Measured PL
(dB) location1
Measured PL
(dB) location 2
Measured PL
(dB) location 3
Average PL
𝐿𝐿𝑝𝑝(𝑑𝑑0) 𝑑𝑑𝑑𝑑
Predicted PL
𝐿𝐿𝑝𝑝(𝑑𝑑𝑖𝑖) 𝑑𝑑𝑑𝑑
100 -76.43 90.20 100.5 99.8 96.83 96.83
200 -79.48 113.1 128.0 141.5 127.5 96.83+3.01n
300 -85.07 113.5 123.5 124.1 120.4 96.83+4.77n
400 -85.03 112.5 120.36 123.3 118.7 96.83+6.02n
500 -81.94 99.60 129.15 158.0 128.9 96.83+6.99n
600 -86.24 104.4 120.5 158.0 127.6 96.83+7.78n
700 -84.17 107.8 125.5 158.0 130.4 96.83+8.45n
800 -79.76 140.5 122.5 116.0 126.3 96.83+9.03n
900 -82.60 115.0 118.8 127.0 120.3 96.83+9.54n
1000 -88.01 145.1 121.3 124.5 130.3 96.83+10.00n
1100 -88.55 117.6 120.95 129.5 122.7 96.83+10.41n
1200 -89.55 158.0 154.7 119.5 144.1 96.83+10.79n
1300 -86.55 130.5 123.9 116.6 123.7 96.83+11.14n
1400 -84.25 113.5 148.0 127.6 129.7 96.83+11.46n
1500 -80.65 120.5 123.9 117.8 120.7 100.5+11.76n
I. LTE data analysis based on measurement locations
The regression analysis was carried out on the measured
path loss for locations 1-3, the reference path loss, exponent,
and the measured data are presented in Table III.
TABLE III: AVERAGE PATH LOSS EXPONENT, REFERENCE PATH LOSS AND
MEASURED DATA FOR URBAN AREA OF PORT HARCOURT
Measurement
Locations
Exponent
(n= np)
𝐿𝐿𝑃𝑃 (𝑅𝑅𝑅𝑅𝑅𝑅) Measured data
Location 1 3.48 90.2 16804.43 - 7943.442n +1139.919𝑛𝑛2
Location 2 3.26 100.5 14242.92 – 7436. 3734n +1139.919𝑛𝑛2
Location 3 3.16 99.8 17179.763 -7211.73 n +1139.919𝑛𝑛2
Average 3.30 96.83 16075.704 – 7530.516 n +1139.919𝑛𝑛2
The path loss exponent (𝑛𝑛𝑛𝑛) was obtained by equating the
derivatives of (27) to zero as
𝑒𝑒(𝑛𝑛𝑛𝑛) = 1139.919𝑛𝑛2
− 7530.516 n + 16075.704 = 0
𝜕𝜕𝜕𝜕(𝑛𝑛𝑛𝑛)
𝜕𝜕𝜕𝜕
= 2(1139.919𝑛𝑛) − 7530.516 = 0
𝑛𝑛𝑛𝑛 = 3.30

Vol. 2, No. 5, May 2017
Therefore, the standard deviation, 𝜎𝜎𝑥𝑥 (𝑑𝑑𝑑𝑑), about a mean
value, was determine from (26) as
𝜎𝜎𝑥𝑥 = �
1
15
(1139.919(3.30)2
− 7530.516(3.30) +
16075.704 �
1
2
𝜎𝜎𝑥𝑥 = 12. 93 𝑑𝑑𝑑𝑑
The standard deviation, 𝜎𝜎𝑥𝑥 (𝑑𝑑𝑑𝑑) was added to (25) to
compensate for the shadowing error of the terrain. Also,
𝐿𝐿𝑃𝑃𝑃𝑃(𝑑𝑑0)𝑅𝑅𝑅𝑅𝑅𝑅 and 𝑛𝑛𝑛𝑛 were substituted to produce (28).
Therefore, modified shadowing empirical model for Port
Harcourt urban area is presented in as;
𝐿𝐿𝑃𝑃𝑃𝑃(𝑑𝑑𝑖𝑖) = 96.83 + 10(3.30) log �
𝑑𝑑1…𝑁𝑁
𝑑𝑑 𝑅𝑅𝑅𝑅𝑅𝑅
� + 12.93 𝑑𝑑𝑑𝑑 (28)
J. Optimization Technique
PSO algorithm is a population based optimization method
for solution of non-linear problems [21]. This is a global
optimization algorithm relevant to the social behavior of
organisms such as swarm birds within a flock, schools of fish
and flying particles in space. The particles shift its position
and velocity according to the best previous experience in
order to reach the best possible solution. The solution reached
by each particle is called personal or local best (𝑃𝑃𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥 ). The
final best solution obtained among all swarm of particle in the
search space is called global best (𝑃𝑃𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 ). In the iteration
stage, each particle in search space updates its velocity and
position using (29) and (30).
Velocity of the particle is given by [21] as;
𝑣𝑣𝑖𝑖
( 𝑗𝑗+1 )
= 𝑣𝑣𝑖𝑖
( 𝑗𝑗 )
+ 𝑐𝑐1. 𝑟𝑟𝑟𝑟𝑛𝑛𝑛𝑛𝑑𝑑1�𝑃𝑃𝑖𝑖
( 𝑗𝑗 )
− 𝑋𝑋𝑖𝑖
( 𝑗𝑗 )
� +
𝑐𝑐2. 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑑𝑑2[𝑃𝑃𝑔𝑔
( 𝑗𝑗 )
( 𝑗𝑗 )
] (29)
Position of individual particle is updated as follows:
𝑋𝑋𝑖𝑖
( 𝑗𝑗+1 )
= 𝑋𝑋𝑖𝑖
( 𝑗𝑗 )
+ 𝑣𝑣𝑖𝑖
( 𝑗𝑗+1 )
∆(𝑡𝑡) (30)
where
𝑣𝑣𝑖𝑖
( 𝑗𝑗+1 )
is the modified velocity of individual 𝑖𝑖 at iteration
( 𝑗𝑗 + 1 ),
𝑋𝑋𝑖𝑖
( 𝑗𝑗+1 )
is the modified position of swarm particle 𝑖𝑖 at
iteration ( 𝑗𝑗 + 1 ),
𝑣𝑣𝑖𝑖
( 𝑗𝑗 )
is the initial velocity of particle
𝑋𝑋𝑖𝑖
( 𝑗𝑗 )
is initial position of particle
𝑃𝑃𝑖𝑖
( 𝑗𝑗 )
is the best individual particle position
𝑃𝑃𝑔𝑔
( 𝑗𝑗 )
is the global best swarm position
𝑐𝑐1 and 𝑐𝑐2 are acceleration constant (cognitive and social
coefficients), 𝑐𝑐1 = 1, 𝑐𝑐2 = 4 − 𝑐𝑐1
𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑑𝑑1 and 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑑𝑑2 are random numbers between 0 and 1.
∆(𝑡𝑡) is time step, it is usually chosen as 1 𝑠𝑠𝑠𝑠𝑠𝑠
PSO Constriction Factor (𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)
PSO with constriction factor (cf), provides a balance
between global and local particle searches. It also maintains
better stability and good convergence for optimization. In
this, the velocity in (30) is changed as;
𝑣𝑣𝑖𝑖
( 𝑗𝑗+1 )
= 𝑐𝑐𝑐𝑐 ∗ 𝑣𝑣𝑖𝑖
( 𝑗𝑗 )
+ 𝑐𝑐1. 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑑𝑑1�𝑃𝑃𝑖𝑖
( 𝑗𝑗 )
( 𝑗𝑗 )
� +
𝑐𝑐2. 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑑𝑑2[𝑃𝑃𝑔𝑔
( 𝑗𝑗 )
( 𝑗𝑗 )
] (31)
where, 𝑐𝑐𝑐𝑐 =
2
�2−𝑘𝑘−� 𝑘𝑘2−4𝑘𝑘�
, 𝑘𝑘 = 𝑐𝑐1 + 𝑐𝑐2
𝑗𝑗 = 1, 2, 3, 4 … … 𝑁𝑁 − 1
The simulation of the existing models and modified model
were carried out using MATLAB tools. The velocity and
position of individual particles in the search space are updated
using (31) and (30) to reach maximum iteration.
K. Performance analysis and validation of models
This section deals with the validation and performance
analysis of existing and modified model. The performance
metric such as RMSE and MAE are used to calculate the
quantity of error between the received signal and the
predicted value. The best fit model for loss prediction in Port
Harcourt urban area will be decided using (32) and (33) given
by [22], [23] as;
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 = �
1
𝑛𝑛
∑ (𝑝𝑝𝑚𝑚𝑚𝑚 − 𝑝𝑝𝑟𝑟𝑟𝑟)2𝑛𝑛
𝑖𝑖=1 (32)
𝑀𝑀𝑀𝑀𝑀𝑀 =
1
𝑛𝑛
∑ �
𝑝𝑝 𝑚𝑚𝑚𝑚 −𝑝𝑝𝑟𝑟𝑟𝑟
𝑝𝑝 𝑚𝑚𝑚𝑚
�𝑛𝑛
𝑖𝑖=1 (33)
where, 𝑝𝑝𝑚𝑚𝑚𝑚(𝑑𝑑𝑑𝑑) is the measured path loss, 𝑝𝑝𝑟𝑟𝑟𝑟 (𝑑𝑑𝑑𝑑) is the
predicted propagation path loss and 𝑚𝑚 is the number of data
points. The RMSE and MAE analysis for PSO Optimized
model, Modified model and existing models are presented in
Table IV.
TABLE IV: PERFORMANCE ANALYSIS OF THE MODELS
Error
Metrics
Okumura-
Hata (dB)
COST-
231 (dB)
Ericsson
(dB)
Egli
(dB)
ECC-33
(dB)
Modified
(dB)
PSO-
Optimized(dB)
MSE 3.84 9.11 35.43 97.87 196.40 3.57 3.030
MAE 0.00206 0.00488 0.0190 0.0524 0.1051 0.00191 0.00162
III. RESULTS AND DISCUSSION
The results obtained through the measured values from
LTE base station in three different locations are presented in
Fig. 1 to 6. The path loss (dB) as contained in Table V are
obtained every 100 m and plotted against distance (km).

Vol. 2, No. 5, May 2017
TABLE V: SHOWS THE DISTANCE, MEASURED PATH LOSS, EMPIRICAL MODELS AND MODIFIED MODEL FOR PORT HARCOURT URBAN AREA
Distance
(km)
Measured PL
(dB)
Okumura-Hata
(dB)
Cost231-Hata
(dB)
Ericsson 999
(dB)
Egli model
(dB)
ECC-33
(dB)
Modified model
(dB)
PSO
Optimized
(dB)
0.1 96.83 100.28 102.79 135.45 190.1 296.84 109.76 109.27
0.2 127.5 110.76 116.22 144.59 202.14 305.82 119.69 119.16
0.3 120.4 116.88 122.35 149.93 209.19 311.08 120.51 119.97
0.4 118.7 121.23 126.69 153.73 214.18 314.8 121.63 121.09
0.5 128.9 124.6 130.06 156.67 218.06 317.69 123.83 123.28
0.6 127.6 127.35 132.82 159.07 221.23 320.06 125.44 124.88
0.7 130.4 129.68 135.15 161.1 223.91 322.05 126.65 126.09
0.8 126.3 131.7 137.16 162.86 226.23 323.78 128.56 127.99
0.9 120.3 133.48 138.94 164.42 228.27 325.31 131.25 130.67
1.0 130.3 135.07 140.54 165.8 230.1 326.67 132.76 132.17
1.1 122.7 136.51 141.98 167.06 231.76 327.91 134.13 133.53
1.2 144.1 137.83 143.29 168.21 233.27 329.04 135.37 134.77
1.3 123.7 139.03 144.5 169.26 234.66 330.07 136.52 135.91
1.4 129.7 140.15 145.62 170.24 235.95 331.03 137.58 136.97
1.5 120.7 141.2 146.66 171.15 237.15 331.93 138.57 137.95
Fig. 1 shows the plot of existing model with measured data
and Fig. 2 and 3 illustrate a comparative plot of measured
data, existing models data and modified model data. Fig. 4
shows the PSO optimized performance with the measured
and existing models while Fig. 5 and 6 depict the plots of
measured data, existing model, modified model and PSO
Optimized model data. It can be observed from Fig. 1,3-6 that
ECC-33, Egli, Ericsson models overestimates the
propagation path loss values. At 0.1 km away from the LTE
base station, the path loss values are 96.83 dB, 100.28 dB,
102.79 dB, 135.45 db, 190.1 dB, 296.84 dB, 109.76 dB and
109.27 dB for Measured, Okumura-Hata, COST-231,
Ericsson, Egli, Modified and PSO Optimized models
respectively. At 1.5 km away the values obtained are 120.7
dB, 141.2 dB, 146.66 dB, 171.15 dB, 237.15 dB, 331.93 dB,
138.57 dB and 137.95 dB. Therefore, the results showed that
the PSO Optimized model is the well suited model to be
implemented for accurate prediction of mobile radio channel
in the selected urban regions in Port Harcourt.
Fig. 1. Plot of existing models with the measured data
Fig. 2. Plot of existing models with measured data and modified model
Fig. 3. Comparative plot of measured data, existing models and modified
Model data
Fig. 4. Plot of PSO Optimized model with measured data and existing
models

Vol. 2, No. 5, February 2017
Fig. 5 Plot of PSO Optimized, modified model with measured and existing
models
Fig. 6 Comparative plot of PSO Optimized with measured, Modified
model and existing models
VI. CONCLUSION
In this paper, the RSS from the LTE base stations at three
different cell sites up to a distance of 1.5 km were measured
through drive test. The differences between the mean
measured path loss values and modified model values were
optimized using Particle Swarm Optimization Algorithm for
suitability of the terrain. The results obtained showed through
error metrics that PSO technique gave a better result
compared to the modified model, Okimura-Hata model and
other existing models for the LTE system at 2300 MHz.
Hence, the PSO Optimized model could be suitably deployed
for signal attenuation improvement for LTE network in Port
Harcourt, South-South, Nigeria.
REFERENCES
[1] B. Sridhar and Z.A.K. Mohammed, “A Novel LMMSE Based
Optimized Perez-Vega Zamanillo Propagation Path Loss Model in
UHF / VHF Bands for India”, Progress In Electromagnetics Research
B, vol. 63, pp. 17–33, 2015.
[2] Z.K. Adeyemo., O.K. Ogunremi, A.O. Akande, “Genetic Algorithm
Based Pathloss Optimization for Long Term Evolution in Lagos,
Nigeria”, International Journal of Applied Science and Technology
Vol. 6, No. 2, pp.(79-88), 2016
[3] D.S. Nyitamen, M. Ahmed and T.A. Danladi, “Path loss Propagation
Model Prediction for GSM Mobile Network Planning in Kaduna
Town”, International Journal of Engineering Sciences & Research
Technology (IJESRT), vol. 4, no. 4 pp.345-352, 2015.
[4] G.L. Stuber, “Principles of Mobile Communication”, second edition,
Kluwer Academic publishers, Boston, 2001
[5] Z.K. Adeyemo, O. Fawole, A.O. Akande, “Comparative Analysis of
CMA and MMSE in MIMO-OFDM system”, Innovative System
Design and Engineering, vol. 7, no.5, pp. 35-42, 2016.
[6] N.C. Huan, N.T. Tung and N.T. Duc, “Bit and Packet Error Rates in
Rayleigh Channel with and without diversity”, Mobile Communication
Group project, Department of Communication Technology Institute of
Electronic Systems, Aalborg University, 2003.
[7] S. Sharma, Mobile and Wireless Communications, fourth edition,
Kataria & Sons publishers, New Delhi, 2013.
[8] T.S. Rappaport, Wireless Communications Principles and Practice,
second edition, Prentice Hall. New Jersey, 2002
[9] K.M. Aied. and A.J. Ahmed, “Performance Evaluation of Path Loss in
Mobile Channel for Karada District in Baghdad City”, Engineering and
Technology journal vol. 30 no. 17, pp. 3023-3037, 2012.
[10] A.N. Jadhav and S. Kale-Sachin, “Suburban Area Path loss
Propagation Prediction and Optimization Using Hata Model at 2375
MHz”, International Journal of Advanced Research in Computer and
Communication Engineering”, Vol 3, no. 1, pp. 5004 -5008, 2014.
[11] C. Cox, An Introduction to LTE, John Wiley & Sons Ltd, second
edition, United Kingdom, 2012.
[12] R. Mardeni and K.F. Kwan, “Optimization of Hata Propagation
Prediction Model in Suburban Area in Malaysia”, Progress
In Electromagnetics Research C, vol. 13, pp. 91-106, 2010.
[13] S. Noman, T.S. Muhammad, K. Hasnain and U. Rizwan, “Comparison
of Radio Propagation Models For Long Term Evolution (LTE)
Network”, International Journal of Next-Generation Networks
(IJNGN), vol.3, no.3, pp. 27-41, 2011.
[14] S. Sharma, Mobile Computing, first Edition, Kataria & Sons
publishers, Katson Books. New Delhi, 2014.
[15] M. Hata, “Empirical Formula for Propagation Loss in Land Mobile
Radio Services”, IEEE Transactions on Vehicular Technology, vol.29,
no. 3, pp. 318-325, 1980.
[16] Y. Okumura, E. Ohmori, T. Kawano, and K. Fukuda, “Field strength
and its variability in VHF and UHF land mobile radio service. Review
of Electrical and Communication Laboratory, vol. 16, no 9-10, pp.
(825–873), 1968.
[17] M. Josip, S. Rimac-Drlje and K. Bejuk, “Comparison of propagation
model accuracy for WiMAX on 3.5GHz” 14th IEEE International
conference on electronic circuits and systems, Morocco, pp. 111-114,
2007.
[18] P.K. Sharma and R.K. Singh “Comparative study of path loss model
depends on various parameter,” International Journal of Engineering
Science and Technology, vol. 3, pp. 4683–4690, 2011
[19] B. Sridhar and M.Z.A. Khan, “RMSE comparison of path loss models
for UHF/VHF bands in India”, IEEE Region 10 Symposium, pp. 330–
335, Kuala Lumpur, Malaysia, 2014.
[20] G.C. Nwalozie, S.U. Ufoaroh, C.O. Ezeagwu and A.C. Ejiofor, “Path
loss Prediction for GSM Mobile Networks for Urban Region of Aba,
South-Eastern Nigeria”, International Journal of Computer Science and
mobile Computing, vol. 3, no. 2, pp. 83-90, 2014
[21] J. Kennedy and R. Eberhart “Particle Swarm Optimization,” IEEE
International Conference on Neural Networks, vol. 4, pp. 1942-1948,
1995
[22] H. Park, “Forecasting Three-Month Treasury Bills Using ARIMA and
GARCH Models”, Econ 930, Department of Economics, Kansas State
University, 1999.
[23] L.J. Cao and E.H.T. Francis, “Support Vector Machine with Adaptive
Parameters in Financial Time Series Forecasting”, IEEE Transaction
on Neural Networks, vol. 14, no. 6, pp. 1506-1518, 2003.
Olukunle Akinyinka Akande obtained his
Bachelor of Technology from the department of
Electronic and Electrical Engineering Ladoke
Akintola University of Technology, Ogbomosho,
Nigeria. Also his Master of Engineering in
Electrical and Electronic Engineering from the
University of Ilorin, Kwara State, Nigeria.
Currently he is pursuing his Doctor of Philosophy
at Ladoke Akintola University of Technology,
Ogbomoso, Oyo State, Nigeria. He is a member of
Teacher’s Registration Council of Nigeria (TRCN), Nigeria Society of
Engineers (NSE). He is a lecturer at the department of Electrical and
Electronic Engineering, Federal University of Technology, Owerri,
Nigeria.
Onyebuchi Chikezie Nosiri obtained his Ph.D,
M.Eng and B.Eng Degrees from Nnamdi Azikiwe
University, Awka, Anambra State, Nigeria. He was
employed as an academic staff in the Department
of Electrical and Electronic Engineering, Federal
University of Technology, Owerri, in 2011. He has
taught several courses for undergraduate and
postgraduate students including project
supervisions. His current area of research includes
developing adaptive noise /interference mitigation
techniques for wireless networks, Cognitive Radio Technology, TV
White space, Wireless Sensor Networks, Call Admission Control
algorithm development for wireless networks and Criminal Community
Network detection in Mobile Phone Metadata. He is a member of IEEE

Vol. 2, No. 5, February 2017
Nigeria Society of Engineers (NSE), registered member of Council for
the Regulation of Engineering in Nigeria (COREN).
Agubor Cosmas Kemdirim has obtained B.Eng,
M.Eng and PhD degrees. He is a registered
member of the Nigerian Society of Engineers
(NSE) and the Council for the Regulation of
Engineering in Nigeria (COREN). He was an
Engineer with the Nigerian Telecommunications
Limited (NITEL) and rose to the rank of an
Assistant Manager. He is now a lecturer in the
Department of Electrical and Electronic
Engineering, Federal University of Technology,
Owerri, Imo State, Nigeria. His research area is in wireless
communication with special interest in Antenna systems and diversity
techniques in 3G/4G wireless communication. He practices Taekwondo
and loves watching movies.
Okpara Chinedu Reginald has obtained his
B.Eng and M.Eng and is currently pursuing his
PhD at the Nnamdi Azikiwe University, Awka,
Nigeria. He is a COREN Registered Engineer, He
is currently a Lecturer II in Electrical/ Electronic
Engineering Department, Federal University of
Technology, Owerri. His research interest includes
Smart Grid Communications, Mobile Wireless
Communications Systems, Digital and Wireless
Communications, Wireless Sensor Networks,
Digital Signal Processing, and LTE network.

Implementation of Particle Swarm Optimization Technique for Enhanced Outdoor Network Coverage in Long Term Evolution Network in Port Harcourt, Nigeria

More Related Content

What's hot (18)

Similar to Implementation of Particle Swarm Optimization Technique for Enhanced Outdoor Network Coverage in Long Term Evolution Network in Port Harcourt, Nigeria (20)

More from Onyebuchi nosiri (20)

Recently uploaded (20)

Implementation of Particle Swarm Optimization Technique for Enhanced Outdoor Network Coverage in Long Term Evolution Network in Port Harcourt, Nigeria