41

Journal of Computing and Information Technology - CIT 20, 2012, 1, 41–49

doi:10.2498/cit.1002022

Genetic Algorithm Aided Antenna

Placement in 3D and Parameter
Determination Considering
Electromagnetic Field Pollution
Constraints

Tomislav Rolich and Darko Grundler

Faculty of Textile Technology, University of Zagreb, Croatia
This paper presents genetic algorithm based method for sense, Evolutionary Algorithms have
antenna placement in 3D space and parameter
determination satisfying environmental electromagnetic constituted a popular choice [1].
field pollution constraints. The main goal is to find out Radio network design is a fundamental
antenna parameters (power, position in 3D, azimuth and problem in cellular networks for
elevation) in the area of interest so that electromagnetic
field satisfies minimal electromagnetic field strength for
telecommunications. In these networks, the
service availability and, at the same time, be below terrain must be covered by a set of base
prescribed limit in restricted subareas (people populated stations(or antennae), each of which defines a
areas). The proposed method is validated with two real covered area called cell. The problem may be
world antenna types and with seven different terrain
configurations(various restricted areas). Besides finding reduced to figure out the optimal placement of
the most optimal antenna parameters, the method finds antennae out of a list of candidate sites,
“almost” optimal solutions which give certain freedom tryingtosatisfytwoobjectives: tomaximizethe
to choose alternative antenna position if optimal is not
available. The investigation described here is extension area covered by the radio signal and to reduce
of previous 2D research. the number of used antennae [2,3,4,5]. In this
paper we were solving a similar task, but yet a
Keywords: electromagnetic radiation, transmitter, different one.
genetic algorithms, MATLAB, optimization
Excessive electromagnetic fields which can be
dangerous to people is of increased concern
because of exposition of living organisms to
more and more sources of electromagnetic
1. Introduction fields (radio, TV, GSM, WiFi, etc.).
Determination of various antenna parameters
Many of the problems found in (power, position in 3D, azimuth and elevation)
telecommunications can be formulated as is difficult because there is a complex
optimization tasks. Some examples are interaction of various sources of
assigning frequencies in radio link electromagnetic fields and specific area
communications, developing error correcting configuration (residential areas). In previous
codes for transmission of messages, and papers, the authors investigated the problem
designing the telecommunication network. In with uniform and nonuniform transmitter
practice, most of these optimization tasks are radiation diagram in 2D [6,7,8] and with
unaffordable with exact techniques. In this uniform transmitter radiation diagram in 3D
42 Genetic Algorithm Aided Antenna Placement in 3D and Parameter Determination Considering...

[9] to find out suitable antenna placement findoptimalantennaposition,itsdirectioninthe

optimization method. Based upon promising space and its power, to satisfy electromagnetic
results [6,7,8,9], the authors here use genetic field constraints in the restricted subareas and,
algorithm to optimize electromagnetic field at the same time, to satisfy the reception level
distribution in 3D space for antenna with real- of electromagnetic field in the whole area.
world spatial radiation diagram. There are six Electromagneticfieldhastobelowerthanprescrib
variable antenna parameters: radiation power, ed pollution health hazard limit in restricted
position in 3D (x, y and z), azimuth and subareas and it has to be above prescribed
elevation. The constraints are minimum reception limit in the whole observed area
electromagnetic field strength (minimum (200m × 200m × 40m).
service availability) in the whole area of Strength of the electromagnetic field can be
interest and, at the same time, calculated according to [10]:
maximumallowableelectromagneticfield
strength (health hazard) in the restricted E ·√ )
subareas (buildings or other populated areas).
In this paper, two types of antennas are used
and the results are presented. (1)
where:
2. Investigations EIRP = ERP + 2.15 – effectively isotropic
radiated power,
Presumptions in the investigations described ERP – effectively isotropic radiated power in
here were that the area of interest is a space respect to half wavelength dipole (in the
m investigations described here, ERP is limited
dimension of 200 m × 200 m × 40 (width x to
length x height) and that the transmitter has to 5dbW ≤ ERP ≤ 35dbW),
be positioned within this space. Inside that
space, different shapes of restricted subareas
F(ϕ,ϑ)–dampingfactorofelectricfieldwhich is
are defined. Restricted subareas are the areas
a function of deflection angles from direction
where humans permanently or often reside
(e.g. buildings) and wich are subject to of maximum in azimuth ϕand elevation ϑ, d –
electromagnetic pollution health hazard. It is distance from the antenna.
assumed that restricted areas are transparent In this investigation a damping factor is a
regarding electromagnetic field propagation. horizontal and a vertical radiation diagrams of
an antenna with a nonuniform radiation.
Electromagneticfieldiscalculatedforuniformly
Antenna manufacturers usually supply the
distributed points in this space. Calculated
numerical data of the damping factor for the
points are evenly distributed in the space and
step of onedegree from 0◦ to360◦
the distance between adjacent calculated
intwoperpendicular planes. From these
points in x, y and z axis direction is 1m. This
numerical data of a damping factor,
means that there are 2002×40 = 1,600,000
mathematical functions which best fit the
observed points for which electromagnetic
numerical data for each plane have been set
field strength is calculated. Transmitter
up.
antenna can be positioned only in the limited
range horizontally: In this investigation, one type of antenna with
20 m ≤ x0 ≤ 180 m,20 m ≤ y0 ≤ 180 m and two configurations has been chosen: K752921
vertically: 10 m ≤ z0 ≤ 30 m. The direction of (450 MHz, M=80, D=25; further in the paper
antenna can be positioned also only in the it will be referred to as antenna 1) and
ϕ K752921 (450 MHz, M=100, D=50; further in
limited range for azimuth: 0◦ ≤ ≤ 360◦ and the paper itwillbereferredtoasantenna2)[11].
ϑ Basedon the horizontal and vertical radiation
for elevation: −30◦ ≤ ≤ 30◦. The goal is to diagrams, the following mathematical function
Genetic Algorithm Aided Antenna Placement in 3D and Parameter Determination Considering... 43
has been chosen to analytically represent a
damping factor: fh
i=0

fv
i=0

Parameters from a0 to a6 of the above

functions have been determined using program a. horizontal radiation diagram
MATLAB and Optimization Toolbox [12]. For Figure 1. Horizontal and vertical radiation diagrams based
of a damping factor functions for an
this purpose, a function lsqcurvefit from
Optimization Toolbox has been used for
solving a nonlinear curve-fitting (data-fitting)
problem in the least-squares sense with the
method of preconditioned conjugate gradients.
These functions approximate the numerical
data of a damping factor. Determined
parameters for horizontal radiation diagram
for type of antenna 1 are: a0 = 0.7317, a1 =
0.3595, a2 = −0.0709, a3 = −0.0375, a4 =
0.0142, a5 = −0.0025 and a6 = 0.0009 and for
type of antenna 2 they are: a0 = 0.6522, a1 = a. horizontal radiation diagram
0.0150, a2 = −0.2778, a3 = 0.0398, a4 = 0.0729, Figure 2. Horizontal and vertical radiation diagrams based
of a damping factor functions for an
a5 = −0.0046 and a6 = 0.0066. Determined
parameters for vertical radiation diagram for
type of antenna 1 are: a0 = 0.3964, a1 = 0.3004,
a2 = 0.2953, a3 = 0.0167, a4 = −0.0337, a5 =
0.0007 and a6 = 0.0183 and for type of antenna
2 they are: a0 = 0.5901, a1 = −0.0000, a2 =
0.4395, a3 = −0.0000, a4 = −0.0501, a5 =
−0.0000 and a6 = 0.0271.

a.antenna 1
Figure 3. Spatial radiation diagrams for antenna

Figure1showsthehorizontalandtheverticalradiat
ion diagrams based on an analytical
approximationof damping factor functions for
antenna 1. Figure 2 shows the same data for
antenna 2.
44 Genetic Algorithm Aided Antenna Placement in 3D and Parameter Determination Considering...

Spatial radiation diagram for both antenna restricted subareas) and minimal (Emin =
configurations are approximated according to 0.1V/m, for signal reception) strength of the
[13]: electric field. If a GA individual did not satisfy
these criteria then its
F (ϕ,ϑ) ≈ fh (ϕ) · fv (ϑ) (4) fitnessfunctionwasmultipliedbypenaltyfactor
of 1000.
Figure 3 shows spatial radiation diagrams for
both antenna configurations based on equation Representation of an individual is a 6
(4). components vector of real numbers:
Strength of the electromagnetic field in v = [ERP x0 y0 z0 ϕ0 ϑ0]. Genetic algorithm
threedimensional space has been observed, so parameters are shown in Table 1. Parameters
the equation for it is: were obtained empirically (the authors tried a
variety of parameter values and decided that
E these where the most suitable for investigated
problem). In this paper the stochastic universal
5. 5·√EIRP sampling is used as a selection procedure
[15,16,17]. In each generation 4 individuals
F(ϕ,ϑ) have been created with a crossover procedure,
4 individuals have been created with a
(5) mutation procedure, and 2 individuals are elite
where: individuals(individualswiththelowestvalueofaf
itness function from the last generation) in
x0, y0 and z0 are coordinates of the transmitter each of the three subpopulations.
location.
Optimal transmitter parameters (power, Parameter Value / Property
position, azimuth and elevation) have been 30 (3 subpopulations
determined using a genetic algorithm. Program Population size of 10 individuals)
MATLAB version 7.11.0.584 (R2010b) with Number of generation 50
Global Optimization Toolbox version 3.1 has Selection Stochastic uniform
been used for this purpose [14]. Fitness Crossover Heuristic
function which needs to be minimized is: Mutation Adaptive feasible
Fitness scaling Proportional
f = min(E) + E + Ek (6) Number of elite individuals 2
Crossover fraction 0.5
where:
Table 1. Genetic algorithm parameters in Global
min(E) – minimal total strength of the Optimization Toolbox.
electromagnetic field in the whole observed
area,
3. Results and Discussion
E – mean strength of the electromagnetic field
in the whole observed area,
To verify the method, seven cases with a
Ek – mean strength of the electromagnetic field different number, size and shape of restricted
in the restricted subareas. subareas were chosen (maximal strength of the
electric field was Emax = 10V/m and minimal
Fitness function (6) is original and was
strength of the electric field regarding signal
obtained empirically (the authors tried several
covering was Emin = 0.1V/m). Procedure was
different fitness functions and decided that this
performed separately for each type of antenna
one has the best features for investigated
configurations. Optimization computation was
problem). Fitness function mentioned above
performed on Microsoft Windows 7 (32-bits),
has been used for genetic algorithm
MATLAB version 7.11.0.584 (R2010b) with
population’s individuals which satisfied the
Global Optimization Toolbox version 3.1. The
criteria of maximal (Emax = 10V/m, for
Genetic Algorithm Aided Antenna Placement in 3D and Parameter Determination Considering... 45
computer used for this purpose was equipped the last generation) for the 5th and the 7th cases
with Intel Core 2 Duo E6600 2.4 GHz of restricted subarea configuration for antenna
processor and with 3 GB of DDR2 (800 MHz) 1, while Figure 9 shows those for the 5 th and
RAM. the 6th cases of restricted subarea configuration
In Table 2 the best individualsfrom the last for antenna 2. Figures present observing area,
generationarepresentedforallsevenrestrictedsub restricted subareas and valid transmitter
area configurations and for a minimal strength positions. Possible transmitter positions are
of the electric field Emin = 0.1V/m for antenna marked with dark gray circles.
1.
DatarepresentstransmitterpowerERP,transmitte Figures 10 and 11 show examples of different
r location (x0, y0, z0), horizontal direction spatial orientation of antennas 1 and 2
(azimuth) ϕ0, vertical direction (elevation) ϑ0 respectively. Figures also depict approximated
andfitnessfunctionvaluef . InTable 3thesame is spatial radiation diagrams for each antenna.
presented for antenna 2.
Duration of computation for optimization
procedure for antenna 1 was between 171 and
173 minutes for cases from 1 to 7, and for
antenna 2 it was between 167 and 178
minutes.
In followingfigures exampleofsomeresultsare
presented. Contour andvaluesof strengthof the
electromagneticfieldinthewholeobservedarea
for the best individual are presented for
various cases of the restricted subareas
configuration. In Figure 4, contour and values
of strength of the electromagnetic field in the
observed area for the best individual for the
5th case of restricted subarea configuration for
antenna 1 and for 0 m, 20 m and 40 m heights.
In Figure 5, the same is shown for the best
individual for the 7th case of restricted subarea
configuration. In Figure 6, contour and values
of strength of the electromagnetic field in the
observed area for the best individual for the
5th case of restricted subarea configuration for
antenna 2 and for 0 m, 20 m and 40 m heights.
In Figure 7, the same is shown for the best
individual for the 6th case of restricted subarea
configuration.

In real world applications, available

transmitter locations are restricted and it may
not be possible to place transmitter on the
positions determined by the best individual
from a GA run. In such a case it is possible to
choose one of the other “good“ individuals
from a GA run. This maynotbethebest
setoftransmitterparameters (a GA
individual),but still good enough regarding
initial conditions. Figure 8 shows examples of
transmitter positions (a GA individuals from
46 Genetic Algorithm Aided Antenna Placement in 3D and Parameter Determination Considering...

Case ERP (dbW) x0 (m) y0 (m) z0 (m) ϕ0[◦] ϑ0[◦] f

1st 21,03 52,70 20,33 24,78 117,38 -21,15 0,73
2nd 19,65 101,09 177,68 21,65 243,32 19,60 0,90
3rd 21,40 22,83 41,25 24,26 116,17 -26,21 0,83
4th 18,96 20,83 100,86 21,86 187,21 0,73 0,84
5th 19,54 150,34 57,26 21,43 77,06 29,98 0,90
6th 18,29 164,85 22,63 20,30 358,14 3,09 0,83
7th 19,74 83,78 179,98 18,75 265,08 -22,09 0,83

Table 2. The best individuals from the last generation for a minimal strength
of the electric field Emin = 0.1V/m for antenna 1.

Case ERP (dbW) x0 (m) y0 (m) z0 (m) ϕ0[◦] ϑ0[◦] f

1st 22,11 179,34 27,21 14,96 265,74 28,11 1,02
2nd 21,50 20,27 33,25 27,36 135,67 7,39 0,96
3rd 20,60 39,00 179,73 21,82 210,36 -22,34 0,85
4th 19,95 163,51 45,72 21,64 77,77 24,37 1,05
5th 20,62 24,13 142,41 23,48 359,72 24,91 1,03
6th 20,77 21,57 118,53 14,17 181,29 -0,16 1,10
7th 19,38 26,48 24,69 18,61 288,46 12,98 0,90

Table 3. The best individuals from the last generation for a minimal strength
of the electric field Emin = 0.1V/m for antenna 2.
a. z = 0 m b. z = 20 m c. z = 40 m
Figure 4. Contour and values of strength of the electromagnetic field in the observed area for the best individual
for the 5th case of restricted subarea configuration, antenna 1, for 0 m, 20 m and 40 m heights.

a. z = 0 m b. z = 20 m c. z = 40 m
Figure 5. Contour and values of strength of the electromagnetic field in the observed area for the best individual
for the 7th case of restricted subarea configuration, antenna 1, for 0 m, 20 m and 40 m heights.

a. z = 0 m b. z = 20 m c. z = 40 m
Figure 6. Contour and values of strength of the electromagnetic field in the observed area for the best individual
for the 5th case of restricted subarea configuration, antenna 2, for 0 m, 20 m and 40 m heights.
Genetic Algorithm Aided Antenna Placement in 3D and Parameter Determination Considering... 47

a. z = 0 m b. z = 20 m c. z = 40 m
Figure 7. Contour and values of strength of the electromagnetic field in the observed area for the best individual
for the 6th case of restricted subarea configuration, antenna 2, for 0 m, 20 m and 40 m heights.

a. the 5th case b. the 7th case

Figure 8. Examples of valid transmitter positions from the last generation for the 5th and the 7th cases
of restricted subarea configuration for antenna 1.

a. the 5th case b. the 6th case

Figure 9. Examples of valid transmitter positions from the last generation for the 5th and the 6th cases
of restricted subarea configuration for antenna 2.

a. 1st case b. 2nd case c. 7th case

Figure 10. Spatial radiation diagram of antenna 1 radiation for 1st, 2nd and 7th case.
 48 Genetic Algorithm Aided Antenna Placement in 3D and Parameter Determination Considering...

a. 1st case b. 3rd case c. 5th case

Figure 11. Spatial radiation diagram of antenna 2 radiation for the 1st, 3rd and 5th cases.
4. Conclusion http://www.lcc.uma.es/∼eat/pdf/nidisc2004.pdf,
Accessed: 23.11.2011.

Baseduponpreviousinvestigations[6,7,8,9]and [3] E. ALBA, G. MOLINA, F. CHICANO, Optimal Placement

of Antennae using
here obtained results for two antenna types Metaheuristics,http://neo.lcc.uma.es/staff/francis/
and various terrain configurations, it can be pdf/nma06.pdf, Accessed: 23.11.2011.
concluded that the method is promising and
can give satisfactory results. Although the [4] S. P. MENDES, J. A. G. PULIDO, M. A. V. RO-
DRIGUEZ, M. D. J. SIMON, J. M. S. PEREZ, A
method is quite computing demanding Differential Evolution Based Algorithm to
comparing to 2D problems [6,7,8], it is still Optimize the Radio Network Design Problem,
quite acceptable for ordinary and inexpensive http://www.icsi.berkeley.edu/∼storn/ deb06.pdf,
Accessed: 23.11.2011.
equipment (office PC). The main reason for
excessive computing requirements is the [5] A. J. NEBRO, E. ALBA, G. MOLINA, F. CHICANO, F.
number of points in the 3D for which LUNA, J. J. DURILLO, Optimal Antenna Placement
electromagnetic field strength has to be Using a New Multi-Objective CHC Algorithm,
calculated. Aside from that, the method is http://neo.lcc.uma.es/staff/guillermo/
indexfiles/files/MOCHC.pdf, Accessed:
simple, straightforward and independent of 23.11.2011.
terrain configuration or antenna type.
Additional method benefit are “almost” [6] T. ROLICH, D. GRUNDLER, Minimizing Environmental
Electromagnetic Field Pollution Adjusting Transmitter
optimal solutions which give certain freedom Parameters Using Genetic Algorithm, 2009IEEE
in antenna Congress on EvolutionaryComputation, Trondheim,
placementandavoidanceofprohibitedplacement Norway, pp. 881–887, (2009).
positionsforthe antenna. Future [7] T. ROLICH, D. GRUNDLER, Managing Electromagnetic
investigationwill include more than one Field Pollution Using Genetic Algorithm, 32nd
antenna, where antennas can be of different International
Communication
Convention on Information and
Technology, Electronics and
types, few antennas can be positioned in the Microelectronics – MIPRO 2009, Opatija, Croatia, pp.
same point in 3D (on the same support pole) 227–232, (2009).
etc. [8] T. ROLICH, D. GRUNDLER, Determining optimal power,
location and direction of transmitters using a genetic
algorithm, 33rd International Convention on
Information and Communication Technology,
5. References Electronics and Microelectronics – MIPRO 2010,
Opatija, Croatia, pp. 161–166, (2010).

[9] T. ROLICH, D. GRUNDLER, Reduction of

[1] E. ALBA, J. F. CHICANO, Evolutionary Algorithms in electromagnetic field pollution in 3D space using a
Telecommunications, http://neo.lcc.uma. genetic algorithm, 2010 IEEE World Congress on
es/staff/francis/pdf/melecon06.pdf, Accessed: Computational Intelligence, Barcelona, Spain, pp.
23.11.2011. 3945–3949, (2010).

[10] CEI/IEC 61566: 1997 International Standard:

[2] E. ALBA, Evolutionary Algorithms for Optimal
Measurement of exposure to radiofrequency
Placement of Antennae in Radio Network Design,
electromagnetic fields – Field strength in the frequency
range 100 kHz to 1 GHz, 1997. [11] Kathrein Scala
 Genetic Algorithm Aided Antenna Placement in 3D and Parameter Determination Considering... 49
Division – professional antenna and filter products for His research activity is in technical sciences, computer engineering
broadcast, land mobile and wireless communication and artificial intelligence and his area of special research interest is
evolutionary algorithms, neural networks and fuzzy logic. His current
applications, Available: http://www.kathrein- position is that of an associated professor at the Faculty of Textile
scala.com/,2011. [12] The MathWorks – MATLAB and Technology, University of Zagreb. He has published 39 papers,
Simulink for delivered 15 presentations at international conferences; he is co-author
Technical Computing, Available: of several university handbooks and two university textbooks.
http://www.mathworks.com/,2011.

[13] F. MIKAS, P. PECHAC, The 3D Approximation of

Antenna Radiation Patterns, The Institute of Electrical DARKO GRUNDLER was born and lived in Zagreb (Tresnjevkaˇ ) until
Engineers, The IEE, Michael Faraday House, Six Hill the age of 25. In 1973 started to work in Kutina as a secondary school
Way, 2003. teacher. From 1976 he was employed in Kutina in the company SELK
on various positions, mainly as technical manager. From 1990 he was
[14] Genetic Algorithm and Direct Search Toolbox User’s a lecturer, assistant professor and professor at the University of
Guide, Version 2, The MathWorks, Inc., 2006. Zagreb, Faculty of Textile Technology. Among other things, his
professional experience includes: electronic watches assembly,
optoelectronics assembly, single board computers design and
[15] J. E. BAKER, Reducing Bias and Inefficiency in the application, microprocessor based system design and application,
Selection Algorithm, Proceedings of the Second CAM, PC based industrial control, genetic algorithms in general and
International Conference on Genetic Algorithms and specially applied to process control. He is the author or coauthor of
their Application, pp. 14–21, 1987. more than thirty books and textbooks (one awarded) in the field of
microprocessors, personal computers, MATLAB, evolutionary
[16] T. BACK¨ , Evolutionary Algorithms in Theory and computation, etc. Field of recent interest and research: genetic
Practice: Evolution Strategies, Evolutionary algorithms applied to process control and textile design.
Programming, Genetic Algorithms, Oxford University
Press, 1996.

[17] D. E. GOLDBERG, Genetic Algorithm in Search,

Optimization and Machine Learning, Addison-Wesley,
Reading MA, 1989.

Received: December, 2011

Accepted: January, 2012
Contact addresses:
Tomislav Rolich
Faculty of Textile Technology
University of Zagreb, Croatia
e-mail: [email protected]
Darko Grundler
Faculty of Textile Technology
University of Zagreb, Croatia
e-mail: [email protected]

TOMISLAV ROLICH was born on 22nd April 1971 in Zagreb. He

graduated from the Faculty of Electrical Engineering and Computer
Science, University of Zagreb in 1995, under the mentorship of Prof.
Neven Mijat.
In January 1995, during the study, he got a job in the Technical School
“Ruder Boskoviˇ c”, first as an associate and later as a teacher. In
October´ 1998 he got a job at the Faculty of Textile Technology,
University of Zagreb as a junior assistant for the courses: Information,
Computing and Applied Computing. At the same time he enrolled in
postgraduate studies at the Faculty of Electrical Engineering and
Computer Science, University of Zagreb.
In November 2001, he got the Master’s degree from the Faculty of
Electrical Engineering and Computer Science, University of Zagreb,
with the theme “Evaluation of the Application of Evolutionary
Algorithms in Achieving Optimal Control” under the mentorship of
Prof. Darko Grundler. In April 2005 he got the Ph.D. degree from the
Faculty of Electrical Engineering and Computer Science, University
of Zagreb, with the theme “Searching the Space of Aesthetic
Evaluation for Matching Weave and Color in Fabrics Design” under
the mentorship of Prof. Darko Grundler.