AIMS Electronics and Electrical Engineering, 5(4): 315–333.

DOI: 10.3934/electreng.2021016
Received: 23 September 2021
Accepted: 24 November 2021
Published: 02 December 2021

http://www.aimspress.com/journal/ElectrEng

Review

Application of Genetic Algorithm for Binary Optimization of

Microstrip Antennas: A Review

Jeevani W. Jayasinghe*

Department of Electronics, Wayamba University of Sri Lanka, Kuliyapitiya, Sri Lanka

* Correspondence: Email: jeevani@wyb.ac.lk.

Abstract: Researchers have proposed applying optimization techniques to improve performance of

microstrip antennas (MSAs) in terms of bandwidth, radiation characteristics, polarization, directivity
and size. The drawbacks of the conventional MSAs can be overcome by optimizing the antenna
parameters while keeping a compact configuration. Applying a global optimizer is a better technique
than using a local optimizer or a trial and error method for performance enhancement. This paper
discusses genetic algorithm (GA) optimization of microstrip antennas presented by the antenna
research community. The GA optimization procedure, antenna parameters optimized by using GA
and the optimization objectives are presented by reviewing the literature. Further, evolution of GA in
the field of MSAs and its significance are explored. Application of GA optimization to design
broadband, multiband, high-directivity and miniature antennas is demonstrated with the support of
several case studies giving an insight for further developments in the field.

Keywords: cost function; design optimization; genetic algorithms; microstrip antennas

1. Introduction

1.1. Background

Microstrip antennas (MSAs) are compact in size, light in weight and inexpensive. More
importantly, they can be easily integrated with the circuit inside the electronic devices. However, the
conventional MSAs have inherent drawbacks of narrow bandwidth and low gain. Numerous
performance improvement techniques such as modifying the patch geometry, shorting the patch and
 316

the ground, incorporating parasitic patches, stacking multiple substrates and keeping an air gap
between the patch and the ground have been suggested by researchers in order to obtain broadband,
multiband, miniature, high-directivity and polarization properties. Such modifications are applied
after performing several trials for tuning antenna parameters. It is a primary technique, which does
not facilitate designing of multiple parameters in parallel. That limitation reduces the possibility of
obtaining the global optimum. The solution space is also lessened and ultimately a local optimum is
produced. Even though the local optima have better performance than classical MSAs, they do not
perform outstandingly as the global optimum. Therefore, antenna researchers have incorporated
optimization techniques [1‒3] such as genetic algorithms (GAs) [4‒6], particle swarm optimization
(PSO) [7], invasive weed optimization (IWO) [8] and differential evolution (DE) [9] for performance
improvement in the field of electromagnetics. Among the aforementioned techniques, GA
optimization has been widely used for MSA design. It is classified as a global optimizer, having
many advantages over the local optimizers [4]. GA has demonstrated its suitability to solve complex
electromagnetic problems in the last two decades.

1.2. GA antenna design procedure

Applying GAs on MSA optimization starts with a randomly generated group of antenna
topologies, which is called the initial population (Figure 1: step 1). Even though the common
practice is to keep the population size constant throughout the optimization process, some
researchers have reported use of a large population at the initial population with the objective of
creating a pool of designs with better performance from the very beginning [10]. A smaller
population size was kept in the subsequent generations for minimizing the time taken for simulations.
GA implementations with different population sizes such as 20, 32, 40 and 200 [11‒14] are reported
in the literature. However, it is recommended to keep a population size of 30-100 as it needs to be
large enough to perform GA operations effectively and small enough to avoid running antenna
simulations over several days [4]. Larger populations facilitate the achievement of optimization
objectives within a lower number of generations due to the higher diversity of the designs. On the
other hand, smaller populations consume less simulation time per generation.
Pairs of chromosomes are selected randomly and the GA operators (crossover and mutation) are
applied on each pair creating a new pair (Figure 1: step 2 and step 3). After applying GA operators,
an expanded population comprising parent designs and children designs is formed (Figure 1: step 4).
Usually, the size of the expanded population is twice the regular population size. Selection of the best
designs for the next generation from the expanded population is done by evaluating the fitness of
each design. A cost function needs to be derived to select the best performing antenna designs. When
the cost function is applied in the optimization process, the average fitness of the population
improves gradually over the iterations. Cost functions are problem-specific and objective-focused.
When the cost functions are defined well based on the antenna performance expected, they facilitate
designing the globally optimized antenna instead of a local optimum.
Once the performance of the antenna designs is evaluated in terms of the fitness, the least fit
designs are removed and the generation is replaced by the individuals with the best fitness (Figure 1:
step 5). Likewise, the GA process is repeated until the termination criterion is met. The termination
may simply be specified in terms of the simulation time or the number of iterations [16].
Convergence of the fitness over iterations or reaching a pre-defined fitness value can also be
considered for termination.

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 317

Figure 1. GA antenna design procedure.

(a) (b)

Figure 2. GA operations (a) Crossover (a pair of parent chromosomes produce children);

(b) Mutation (only a few genes are changed).

1.3. GA operators

In order to apply GA, antenna parameters are encoded creating a string of bits called a
chromosome. It is a data structure that holds genes, which are usually "1"s and "0"s. During the
crossover operation, genes of a pair of candidate designs (parent chromosomes) are exchanged in
order to form two new designs (children chromosomes) for the next generation. For example, Figure
2a illustrates the crossover operation performed on a pair of MSAs in which the coloured region
represents the conducting patch and corresponds to “1”s in the chromosome. Similarly, the white
colour spaces corresponding to “0”s in the chromosomes represent non-conducting areas of the
antenna. The crossover probability is normally higher than 60% giving a considerable portion of the
population the opportunity to produce better chromosomes [4]. The probability may be 100%
assuring all parent chromosomes involve in generating children chromosomes [15]. In mutation, a
gene "0" is changed to a "1" and vice-versa. Correspondingly, the conducting region turns into a
non-conducting space and vice-versa (Figure 2b). Mutation ensures that no potential solutions are

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 318

lost and prevents repeated mating. Keeping the mutation probability low is recommended in order
to confirm reaching towards the optimum solution without deviating from the gradual
improvement [4].
This paper reviews applications of GA for performance enhancement of MSAs. In Section 2,
the antenna parameters fine-tuned in the GA optimization applications are explored and their impact
is analyzed. Section 3 presents the optimization objectives the antenna researchers tried to achieve
by applying GA on MSAs. The impact of the different fitness functions is also analyzed. In Section
4, GA is compared with some other high performance optimization techniques applied in the field of
MSAs. Moreover, the future directions are envisaged.

2. Optimization of antenna parameters

2.1. Outline

GA optimization is used in the field of MSAs mainly to design the characteristics of the patch,
substrate and the feed with the objective of performance enhancement. Further, conventional
performance improvement techniques such as modification of the patch shape, use of shorting pins
or strips and use of an air gap were integrated with GA. In the field of antenna optimization,
fine-tuning a single parameter as well as multiple parameters was proposed. Out of the two
approaches, parallel optimization of multiple antenna parameters is more effective due to the higher
degree of freedom, which enables exploring the entire search space.

2.2. Optimization of the geometry

As per the literature related to GA-based MSA optimization, the mostly reported MSA
parameter optimized by using GA is the patch geometry. This paper reviews only binary
optimization problems where the patch is divided into smaller elements and the final shape is
obtained by turning some of these elements off. Though there are continuous problems as well,
where one or more physical dimensions of the antenna are tuned over a continuous range using GA
to optimize the antenna, no continuous problem is addressed in this paper. In the related research
studies conducted about last two decades ago, only the patch geometry was optimized, while the
substrate parameters and the feed position were kept fixed [4]. Nevertheless, optimization of both
the patch shape and feed position is more effective as it expands the solution space while giving
freedom to match the impedance of the modified shape [16]. When a shorting mechanism such as a
pin, strip or a wall is used for antenna miniaturization, parallel optimization of the patch shape and
the positions of feeding and shorting is the most effective method [11]. For example, the resonant
behavior of a rectangular patch antenna with and without a shorting pin is compared in Figure 3a.
The antenna with a rectangular-shaped patch resonates around 6.4 GHz at the fundamental mode of
operation. The shorted patch exhibits miniature performance resonating around 3 GHz as the
shorting pin alters the classical current pattern. With GA optimization, the resonant frequency of the
miniaturized MSA could be reduced further to 2.45 GHz as a result of proper positioning of the
shoring and feeding pins on the radiating patch with a modified shape (Figure 3b). Such a behavior
proves that the modified patch geometries are helpful to create elongated current paths resulting
longer electrical lengths. More importantly, the bandwidth could be improved due to parallel
optimization of three parameters.

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 319

(a)

(b)

Figure 3. Parallel optimization of the patch shape, feed position and shorting position (a)
Resonant behavior of rectangular antennas (b) Resonant behavior of the optimized antenna.

In the antenna designs with an air gap between the patch and the ground plane, its thickness
and the feed position were optimized with the patch geometry simultaneously [17]. Parallel GA
optimization of more than three antenna parameters such as the patch geometry, feed position,
substrate thickness and relative permittivity was also proposed. However, use of GA to optimize
substrate thickness and dielectric constant is not much effective due to the small solution space of
each parameter. Hence, keeping such parameters fixed at suitable values derived based on related
theory or determined by performing a simple set of simulations can be recommended.

2.3. Different patch topologies

In case of optimizing the patch geometry, researchers proposed dividing the patch area into a
grid of small rectangular [18] or square [12] cells. Each cell is assigned either conducting or
non-conducting property based on the value of the corresponding gene of the chromosome. A gene

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 320

with value "1" may represent a conducting region while that with value "0" may represent a
non-conducting region. As the dielectric substrates are usually Copper plated, a printed circuit board
(PCB) prototyping machine or a chemical process is used to remove Copper in the area
corresponding to the non-conducting genes. When diagonally connected cells remain on the patch, a
proper electrical connection cannot be guaranteed between them. Due to such disconnections, the
expected current path is disconnected making the performance of the antenna highly deviated from
the simulated performance. Therefore, techniques such as overlapping cells with a constant [19] or a
variable overlap [20], array of overlapping sub-patches [13], 2D median filter [21] and amorphous
shapes using ellipses [22] have been proposed to ensure connectivity at the corners of the cells. In
case of using overlapping conducting regions, constant overlaps [13] as well as non-uniform
overlaps [20] were proposed. A patch geometry divided into an array with the size of 4 x 5 and
assigned conducting properties to eleven of them is shown in Figure 4a. The connection between
diagonally connected conducting regions is infinitesimal. In contrast, the adjacent regions connect
well diagonally when constant overlaps are introduced (Figure 4b). Figure 4c illustrates the patch
with non-uniform overlaps, which facilitates a wide range of sizes for the conducting regions.

(a) (b) (c)

Figure 4. Different patch topologies a) A traditional grid with infinitesimal connections.

b) Scheme with constant overlaps. c) Scheme with non-uniform overlaps.

Use of fixed overlap sizes as well as non-uniform overlaps on the radiating patch is demonstrated
in [20]. Four MSAs with fixed overlap sizes of 0 mm, 0.5 mm, 1 mm and 2 mm and one MSA
combining all the options were optimized using GA. Both the patch geometry and the feed position of
each antenna were tuned and the variation of the best fitness over the iterations was recorded. The
MSA with non-uniform overlapping cells converged resulting in a penta-band design, while the other
MSAs produced multiband designs with less number of resonances and narrowband performance.
Analysis of the results indicate that the non-uniform overlapping method has resulted better antenna
performance than use of any constant overlap, due to the higher degree of freedom.
Usually, a significant portion of the chromosome is dedicated to define the patch geometry. GAs
with larger chromosome sizes consume more time for simulation of the corresponding designs. Some
researchers have proposed modeling only a half of the patch structure in order to reduce the size of the
chromosome [10]. Antennas with symmetrical geometries exhibit symmetrical radiation patterns,
which is beneficial in mobile communication applications. However, this approach limits the solution
space and the solution may be optimized towards a local optimum. Another method to reduce the
AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.
 321

chromosome size is GA optimization of sub-domains on a patch creating meander edge notches

instead of optimizing the whole patch area [23]. Moreover, design of a switch array on the aperture
facilitating reconfiguration is proposed [24]. Another simplified method for optimizing the shape is
etching a small slot on the patch by using GA to design only the size and location of the slot [25].
When the slots are placed at the opposite corners of a square-shaped patch, the MSA demonstrates
circularly polarized characteristics [26].
Expanding research beyond the shape optimization of a MSA having a patch foot print of a
conventional shape such as rectangular or square, complex shapes such as bow-tie have also been
considered in GA optimization [27]. Moreover, Fractal antennas with the shape of Sierpinski
carpet [27], Sierpinski Bow-Tie [29], Koch [30], Fudgeflake [31] and Gosper island [31], which have
non-conventional complex shapes, have been genetically optimized with the objective of starting the
optimization process with an already modified geometry instead of a conventional shape. However,
only a limited number of research studies were done on GA optimization of non-conventional patch
shapes, which may produce outstanding performance. For example, local optimizers applied on a
Sierpinski carpet fractal antenna of the fourth iteration improved the resonant behavior (Figure 5). Its
performance may be further enhanced by applying a global optimizer. In this sense, research avenues
exist in the field to expand the research.
Another GA optimization approach is the design of antenna parameters on a predetermined patch
shape. One such example is optimizing the locations and widths of the shorting strips between the
rectangular patch and the ground [32]. Optimization can be further simplified by applying GA only to
find the most suitable shorting position on a patch, which has a pre-specified shape [33]. Another
approach is to optimize the properties of the dielectric substrate such as the length, width, thickness
and the dielectric constant [36]. When the design objective is too simple, optimization parameters and
their range can also be simplified. Optimization of the feed position and patch dimensions of a
classical rectangular [34] or circular [35] patch is reported, when the objective is simple as impedance
matching.
Moreover, optimization of the feed network of MSAs has been proposed. In such approaches,
the major objectives were obtaining circular polarization [37] and bandwidth enhancement [38]. In
case of an aperture-coupled rectangular MSA, optimizing the feed parameters such as the aperture
dimensions and stub length is proposed [39].

Feed position 50W

Figure 5. Optimization of a Sierpinski carpet fractal.

In this sense, antenna parameters to be optimized highly depend on the structure of the MSA
and the expected antenna performance. Optimization of multiple antenna parameters and
consideration of a wider range of values demand a chromosome with a higher number of genes.

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 322

Modeling and solving an MSA, which is represented by a larger chromosome, consume more
computation time. As the antenna designs are simulated over iterations until the pre-defined
termination criterion is met, a high performance computer or a cluster of CPUs is required for GA
optimization of MSAs.

2.4. Cost of computation

An MSA optimization problem solved in 2004 reported that the simulation time of an individual
chromosome consisting of 48 binary elements was 13 minutes on a Pentium III-450 MHz CPU with
512 MB RAM and an 8 GB hard drive [10]. As the estimated simulation time for the completion of
the GA process was longer than a month, a cluster of 26 CPUs was used. As a result of using
multiple nodes, the GA optimization process could be completed within two days. Thanks to the
continuous advancements in the performance of computers taken place during the last decade or so,
optimization of complex design problems could be done on a single computer within a reasonable
time. For example, optimization of an MSA represented by a chromosome with a size of 35 bits
reported completion of the simulation process within a day on an Intel Core i7 processor with 2 GHz
speed and a RAM of 6 GB a few years ago [20]. With the availability of high performance computers,
optimization of multiple parameters has also become convenient. Simulation of an MPA on a 5th
generation Core i7 processor consumes only a few seconds nowadays. However, even with high
computational facilities, complex GA optimization problems run over multiple days continuously.
Use of commercially available software such as Ansys HFSS, CST and IE3D for MSA simulations
has been reported by the antenna research community.

3. Optimization objectives

3.1. Introduction

The classical MSAs are etched on a thin dielectric substrate with Copper layers on both sides
resulting in a planar configuration. Even though such an antenna topology is suitable to be integrated
with the circuit in small handheld devices, the MSAs with classical shapes have low performance
and are not suitable for wireless communication applications. Therefore, the application of GA on
MSAs has been proposed with the objective of enhancing the performance by means of the resonant
behaviour, radiation, efficiency and polarization. The antenna size also needs to be optimized as the
half-wavelength or quarter-wavelength of MSAs is not compact enough for some applications,
particularly at low frequency operation. In this context, the design objectives highly depend on the
applications for which an MSA is designed. Different types of GA optimization projects developed to
design broadband, multiband, directive and compact MSAs are found in the literature.

3.2. Improve the resonant behaviour

As the conventional MSAs exhibit narrowband properties and do not show multiband resonance
at a multiple number of frequency bands closer to each other, GA optimization has been applied to
improve the resonant behavior [97]. When the patch shape is conventional, the current flows along a
direct line, resulting in narrowband performance at the fundamental mode as well as at the higher order
modes. Genetically optimized patch shapes alter the direct current flow and allow multiple current

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 323

paths on the patch resulting multiband operation (Figure 6a). Design of multiband MSAs using GA is
reported in the literature [61‒72]. When the patch shape facilitates the current paths in order to make
the MSA resonate at multiple frequencies closer to each other, broadband performance can be
obtained. For example, an MSA resonating at multiple frequencies of 1800 MHz, 2100 MHz, and
2340 MHz operates over a wide band ranging from 1710 MHz to 2500 MHz (Figure 6b). The
bandwidth enhancement of a single patch element resulting genetically optimized UWB [40‒43] and
broadband [43‒61] antennas were reported since late 1990s. Some MSAs optimized by using GA to
obtain UWB performance are summarized in Table 01.
Y

Current path at 1800 MHz

Current path at 2100 MHz

Current path at 2340 MHz

(a)

(b)

Figure 6. A GA optimized broadband antenna (a) Multiple current paths (b)

Multi-frequency broadband behaviour.

Table 1. GA optimized UWB MSAs.

Reference Antenna dimensions (mm3) S11<-10dB Bandwidth (GHz)
[40] 21.2 x 34.45 x 0.795 3.1‒12
[41] 33 x 28 x 1.57 3.1–10.6
[42] 45 x 55x 0.762 3.3‒10.55
[43] 25 x 25 x 1 2.5‒11
[98] 20 × 15 × 0.508 3.18‒3.75 and 4.785‒14.25

3.3. Miniaturization

Another objective of GA optimization is miniaturization of MSAs to make them suitable for

applications such as biotelemetry and handheld wireless devices, which demand very small

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 324

antennas [74‒76]. Genetically optimized patch geometries create elongated current paths resulting
the MSA resonating at a lower frequency than at the fundamental mode, which ultimately produces a
miniature antenna [11]. For example, a rectangular-shaped patch (Figure 7a) was divided into a
7 × 10 grid and optimized by applying GAs. At the fundamental mode of operation the length of the
current path is equal to the patch length and that of the optimized design is longer along the path
a-b-c-d-e (Figure 7b). As a result, the resonance frequency could be reduced by 62%.

(a) (b)

Figure 7. Genetically elongated current paths (a) current pattern on a rectangular antenna
(b) current pattern on the optimized antenna.

3.4. Improve radiation

Moreover, GA optimization was applied to improve the radiation characteristics as required for
specific applications (Table 02). For example, making the radiation pattern along the broadside
direction and improving the directivity or gain while maintaining the necessary bandwidth
requirements are reported in the literature [77‒80]. Moreover, GA has been used to design the MSAs
with a low radar cross section [81]. In-phase current on the genetically optimized patch (Figure 8a)
has made the antenna highly directive along one direction. As a result, a genetically designed single
patch element which occupies the same area of a 1 × 4 array could replace the array by means of the
bandwidth and the directivity. A measured directivity of 12 dB was obtained with antenna efficiency
of 81% (Figure 8b). Further, the GA optimized MSA has demonstrated better results than fractal
antennas with the same foot print in terms of the directivity [15].
Moreover, GAs were used to optimize the polarization characteristics as required for the
antenna application of interest [82‒86]. Synthesizing MSAs to achieve circular polarization with a
single feed [82] and dual polarization with dual frequency operation [83,84] are among them.

Table 2. GA optimized high-directivity MSAs.

Reference Antenna dimensions (mm3) S11<-10dB Bandwidth (MHz) Gain (dB)
[15] 140 ×60 x1.52 160 12.1
[78] 590 x170 x 98 100 13.8
[78] 80x80x1.52 70 10.5
[15] 120 ×120 x1.52 40 13.2

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 325

(a)

(b)

Figure 8. GA optimized high-directivity MPAs (a) Current pattern (b) Radiation cuts.

3.5. Multi-Objective optimization

As briefed above, the optimization of antenna performance by means of bandwidth, resonant

frequency, directivity/ gain, radiation pattern and polarization has shown an explosive growth in the
field of MSAs. Design exercises related to single-objective optimization as well as multi-objective
optimization are among them. Optimization efforts put only on bandwidth enhancement may generate
MSAs with distorted radiation patterns or low gain. Similarly, optimization with the objective of
directivity improvement may result in narrowband MSAs. For example, when the broadband
performance was the only objective in GA optimization, a fractional bandwidth of 75% could be
achieved with a broadside gain limited to 5.8 dB [58]. When multi-objective optimization was applied
to the same problem, the optimized MSA demonstrates fractional bandwidth of 60% and a broadside
gain increased up to 7 dB.
In this sense, solving the design problems with conflicting objectives is a challenge. It can be
addressed by designing a fitness function, which represents all the objectives with necessary weight on
each of them. Performance of MSAs optimized by using five different sets of weighting coefficients in
the fitness function are compared by means of bandwidth, gain and linear polarization factor in
Figure 9. It indicates that the performance in terms of each criterion varies significantly though the
overall fitness is very high and closer to each other. In this sense, weighting coefficients of the partial
fitness functions need to be set depending on the importance of each objective. Even if the
optimization problem is single-objective, the globally optimized design can’t be obtained unless the
most suitable cost function is used. Use of numerous cost functions to design broadband MSAs is
evident in the field [85]. Analysis of several cost functions shows that the optimum design and the
bandwidth performance highly depend on the cost function used in the optimization process (Table 3).
As per the comparison presented in Table 3, the fitness functions which force to have a
reflection coefficient less than -10 dB give a better bandwidth than those consider the reflection
coefficient at the expected resonant frequency or over a frequency band. When multiple objectives in
addition to the broadband performance are considered, the cost function needs to be modified by
considering all the objectives (Table 4). Circular-polarization, high gain and multiband performance
are some of the objectives considered in the fitness function. This paper reviews only the problems
where the objective function is calculated using high fidelity full-wave solvers though approaches

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 326

where low-fidelity surrogate models used for optimizing antennas are also available [95,96].
Moreover, surrogate models significantly reduce the computation cost.

Figure 9. Performance of GA optimized antennas for different weighting coefficients

Table 3. Comparison of cost functions used to design broadband MSAs.

Cost function S11<-10 dB impedance bandwidth (%)
F   ( f i ) 10
 ( f i ) is the reflection coefficient at frequency

f2
16
F     ( fi )
f i  f1

28
f2  ( fi )dB  ( fi ) dB  10dB
F    L( f i ) , L( fi )  
f i  f1   10dB  ( fi ) dB  10dB

f2
21
 ( f  3.4)
F   ( f i ) sin
f i  f1 1.2

  ( fi )  ( f i )  10dB 33

f2
  15dB  15dB   ( f i )  10dB
F    L ( f i ) L( f i )  
f i  f1   30dB   ( fi )  30dB   ( f i )  15dB
 0  ( f i )  30dB


4. Discussion and conclustions

4.1. Discussion

In case of optimizing the geometry, if a patch area is divided into a grid of 100 small cells, it has
a large solution space of 2100 (1.26 × 1030) candidates. Even if the simulation time taken per design
is 1sec, solving the whole solution space takes 4 × 1022 years. Solving such complex design
problems within a few days by using GA optimization is reported.
High performance optimization techniques such as PSO, ant colony optimization (ACO),
Bacterial Foraging Optimization (BFO) and DE have also been applied for performance
enhancement of MSAs. Use of PSO, GA based PSO and ACO to optimize the feed position of a
rectangular patch antenna is presented in [91]. The best feed position could be found within the least

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 327

number of iterations when GA based PSO is used, but it consumes more time. In contrast, PSO is
faster, though it ran over a higher number of iterations. Performance of DE, PSO and BFO in
optimizing the patch length, patch width and the inset feed is presented in [92]. DE is the fastest and
BFO is faster than PSO as per the research findings.
Modification of the GA optimization process and creating hybrid versions of GA incorporating
advanced features of other optimization techniques may be the future direction in the field of MSA
optimization. As per the studies, hybridization of the GA optimization process with particle swarm
optimization, space-mapping and mixed potential based method of moments have been
proposed [89‒92]. The modified GA optimization techniques have resulted in a better design with
accelerated convergence than the classical GA [90]. Moreover, local constrained optimization of GA
optimized MSAs without increasing their footprint has considerably improved antenna performance
due to the utilization of all available degrees of freedom.

4.2. Conclusions

This paper investigates single-objective or multi-objective GA optimization of MSAs. GA

optimization was successfully used to design broadband, multiband, high-gain, directive, miniature
or circularly polarized MSAs. As per the findings, antenna parameters such as the patch geometry,
properties of the substrate and position of feeding or shorting probe have been optimized in order to
achieve the objectives. GA parameters, particularly the cost function, need to be defined carefully, in
order to facilitate global optimization (Table 04). As the wireless communication industry demands
multifunctional devices, multi-objective GA optimization will play a vital role in the field.

Table 4. Cost functions used in multiobjective GA optimization.

Ref. Cost function objectives

87 1 n Circularly-polarization
F  W1  ( )   W  ARi W1 and Wi are weight coefficients, ARi is the ith axial value at the
VSWR i  2 i Conical-beam
elevation angle ϴi, n is the number of elevation angles required.
56 S (f) 1 Broad/dual-band
F   k ( 11   k)
f  100 AR( f )  10 characteris-tics Circular
f  {2.05,2.25,2.35,2.4,2.45,2.55,2.75 GHz} polarizati-on
k ={1 when S11( f )  10 dB and AR( f )  3dB
0 when S11( f )  10 dB and AR( f )  3dB}
AR( f ) is the axial ratio at the sampling freq. f .
N freq
88 
 0, if i  max
Broadband
F  Ai  
2 ,
i 1  i if i  max
 Circularly polarizati-on

Ai is the axial ratio and  is the reflection coefficient

58 f2 f2 Broadband
F  k1F1  k2 F2  K3F3 , F1  G(  00 ) , F2    L( f i ) , F3   ( Ec  E x )
f i  f1 f i  f1 High gain

  ( fi )  ( f i )  10dB Linear polarizati-on


  15dB  15dB   ( f i )  10dB
L( f i )   G is the gain,  ( fi ) is the reflection
 30dB   ( f i )  30dB   ( f i )  15dB
 0  ( f i )  30dB

coefficient, Ec and E x are copolarization and cross polarization, k1 k 2 and k 3 are weighting
,
coefficients

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 328

Conflict of interest

The author declares that there is no conflict of interest.

References

1. Goudos SK, Kalialakis C, Mittra R (2016) Evolutionary algorithms applied to antennas and
propagation: A review of state of the art. International Journal of Antennas and Propagation.
2. Hoorfar A (2007) Evolutionary programming in electromagnetic optimization: a review. IEEE T
Antenn Propag 55: 523‒537.
3. Rahmat-Samii Y, Kovitz JM, Rajagopalan H (2012) Nature-inspired optimization techniques in
communication antenna designs. Proceedings of the IEEE 100: 2132-2144.
4. Johnson JM, Rahmat-Samii Y (1997) Genetic algorithms in engineering electromagnetic. IEEE
Antenn Propag M 39: 7-21.
5. Weile DS, Michielssen E (1997) Genetic algorithm optimization applied to electromagnetics: A
review. IEEE T Antenn Propag 45: 343‒353.
6. Haupt RL, Werner DH (2007) Genetic algorithms in electromagnetic. John Wiley & Sons.
7. Robinson J, Rahmat-Samii Y (2004) Particle swarm optimization in electromagnetic. IEEE T
Antenn Propag 52: 397‒407.
8. Karimkashi S, Kishk AA (2010) Invasive weed optimization and its features in electromagnetic.
IEEE T Antenn Propag 58: 1269‒1278.
9. Rocca P, Oliveri G, Massa A (2011) Differential evolution as applied to electromagnetic. IEEE
Antenn Propag M 53: 38-49.
10. Villegas FJ, Cwik T, Rahmat-Samii Y, et al. (2004) A parallel electromagnetic
genetic-algorithm optimization (EGO) application for patch antenna design. IEEE T Antenn
Propag 52: 2424‒2435.
11. Jayasinghe JM, Uduwawala D (2015) A Novel Multiband Miniature Planar Inverted F Antenna
Design for Bluetooth and WLAN Applications. Int J Antenn Propag.
12. Sun SY (2009) Design of broadband microstrip antenna utilizing genetic algorithm.
In International Symposium on Electromagnetic Compatibility.
13. John M, Ammann M (2006) Design of a wide-band printed antenna using a genetic algorithm
on an array of overlapping sub-patches. IEEE Int Workshop on Antenna Technol.: Small
Antennas and Novel Metamaterials (IWAT2006).
14. Mishra RG, Mishra R, Kuchhal P, et al. (2017) Optimization and analysis of high gain
wideband microstrip patch antenna using genetic algorithm. International Journal of
Engineering & Technology 7: 176‒179.
15. Jayasinghe JW (2014) Optimization of the performance of microstrip patch antennas and arrays
by using genetic algorithms. Ph.D. Thesis, University of Peradeniya.
16. Soontornpipit P, Furse CM, Chung YC (2005) Miniaturized biocompatible microstrip antenna
using genetic algorithm. IEEE T Antenn Propag 53: 1939‒1945.
17. Heidari AA, Dadgarnia A (2011) Design and optimization of a circularly polarized microstrip
antenna for GPS applications using ANFIS and GA. General Assembly and Scientific
Symposium, XXXth URSI, 1‒4. IEEE.

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 329

18. Gupta R, Gurjar S, Kumar A (2013) Using genetic algorithms reduction of rectangular
microstrip patches. International Journal of Advanced Research in Computer and
Communication Engineering 2: 1643‒1646.
19. Jayasinghe JW, Anguera J, Uduwawala DN (2012) A simple design of multi band microstrip
patch antennas robust to fabrication tolerances for GSM, UMTS, LTE, and Bluetooth
applications by using genetic algorithm optimization. Progress In Electromagnetics Research M
27: 255‒269.
20. Jayasinghe JM, Anguera J, Uduwawala DN, et al. (2015) Nonuniform Overlapping Method in
Designing Microstrip Patch Antennas Using Genetic Algorithm Optimization. Int J Antenn
Propag.
21. Choo H, Hutani A, Trintinalia LC, et al. (2000) Shape optimisation of broadband microstrip
antennas using genetic algorithm. Electron Lett 36: 2057‒2058.
22. Griffiths LA, Furse C, Chung YC (2006) Broadband and multiband antenna design using the
genetic algorithm to create amorphous shapes using ellipses. IEEE T Antenn Propag 54:
2776‒2782.
23. Polivka M, Drahovzal M, Rohan J, et al. (2006) Multiband patch antenna with perturbation
elements generated by genetic algorithm. In 2006 First European Conference on Antennas and
Propagation, 1‒4.
24. Xiao S, Wang BZ, Yang XS, et al. (2003) Reconfigurable microstrip antenna design based on
genetic algorithm. IEEE Antennas and Propagatopm Society International Symposium 1:
407‒410.
25. Wyant A, Venkataraman J (2008) Optimization of Reduced Size Microstrip Patch with slots
using a Genetic Algorithm. Dept. of Electrical Engineering, Rochester Institute of Technology.
26. Alatan L, Aksun MI, Leblebicioglu K, et al. (1999) Use of computationally efficient method of
moments in the optimization of printed antennas. IEEE T Antenn Propag 47: 725‒732.
27. Kerkhoff AJ, Rogers RL, Ling H (2004) Design and analysis of planar monopole antennas using
a genetic algorithm approach. IEEE T Antenn Propag 52: 2709‒2718.
28. Koziel S, Jayasinghe JW, Saraereh O, et al. (2017) Local Optimization of a Sierpinski Carpet
Fractal Antenna, IEEE 12th International Conference on Industrial and Information Systems,
1‒5.
29. Behera BR (2017) Sierpinski Bow-Tie antenna with genetic algorithm. Engineering Science and
Technology 20: 775‒782.
30. Adelpour Z, Mohajeri F, Sadeghi M (2010) Dual-frequency microstrip patch antenna with
modified Koch fractal geometry based on genetic algorithm. 2010 Loughborough Antennas and
Propagation Conference (LAPC), 401‒404.
31. Spence TG, Werner DH (2004) Genetically optimized fractile microstrip patch antennas.
In IEEE Antennas and Propagation Society International Symposium 4: 4424‒4427.
32. Ozgun O, Mutlu S, Aksun MI, et al. (2003) Design of dual-frequency probe-fed microstrip
antennas with genetic optimization algorithm. IEEE T Antenn Propag 51: 1947‒1954.
33. Namkung J, Hines EL, Green RJ, et al. (2007) Probe‐fed microstrip antenna feed point
optimization using a genetic algorithm and the method of moments. Microw Opt Techn Lett 49:
325‒329.
34. Lu Q, Edward K, Sean D, et al. (2009) Optimum design of a probe fed dual frequency patch
antenna using genetic algorithm. Proceedings of the ARMMS RF and Microwave Society.

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 330

35. Raychowdhury A, Gupta B, Bhattacharjee R (2000) Bandwidth improvement of microstrip

antennas through a genetic‐algorithm‐based design of a feed network. Microw Opt Techn Lett
27: 273‒275.
36. Chattoraj N, Roy GS (2006) Application of Genetic Algorithm to the Optimization of
Microstrip Antennas with and without Superstrate. Mikrotalasna revija, 32‒35.
37. Aljibouri B, Lim EG, Evans H, et al. (2000) Multiobjective genetic algorithm approach for a
dual-feed circular polarised patch antenna design. Electron Lett 36: 1005‒1006.
38. Raychowdhury A, Gupta B, Bhattacharjee R (2000) Bandwidth improvement of microstrip
antennas through a genetic‐algorithm‐based design of a feed network. Microw Opt Techn Lett
27: 273‒275.
39. Mukherjee P, Gupta B (2005) Genetic algorithm-based design optimisation of aperture-coupled
rectangular microstrip antenna. Defence Sci J 55: 487.
40. Ding M, Jin R, Geng J (2007) Optimal design of ultra wideband antennas using a mixed model
of 2‐D genetic algorithm and finite‐difference time‐domain. Microw Opt Techn Lett 49:
3177‒3180.
41. Silva CRM, Lins HWC, Martins SR, et al. (2012) A multiobjective optimization of a UWB
antenna using a self organizing genetic algorithm. Microw opt Tech let 54: 1824‒1828.
42. Kim J, Yoon T, Kim J, et al. (2005) Design of an ultra wide-band printed monopole antenna
using FDTD and genetic algorithm. IEEE Microw Wirel Co 15: 395‒397.
43. Pourahmadazar J, Shirzad H, Ghobadi C, et al. (2010) Using a MAM and genetic algorithm to
optimize UWB microstrip monopole antenna with FEM and HFSS. 5th International
Symposium on Telecommunications (IST), 115‒119.
44. John M, Ammann MJ (2007) Wideband printed monopole design using a genetic algorithm.
IEEE Antenn Wirel Pr 6: 447‒449.
45. Sun S, Yinghua LV, Zhang J (2010) The application of genetic algorithm optimization in
broadband microstrip antenna design. Antennas and Propagation Society International
Symposium (APSURSI), 1‒4.
46. Michael K, Kucharski AA (2006) Genetic algorithm optimization for broadband patch antenna
design. International Conf. on Microwaves, Radar & Wireless Communications, 748‒751.
47. Sun S, Yinghua LV, Zhang J (2010) The application of genetic algorithm optimization in
broadband microstrip antenna design. Antennas and Propagation Society International
Symposium (APSURSI), 1‒4.
48. Deshmukh AA, Kumar G (2005) Compact broadband U‐slot‐loaded rectangular microstrip
antennas. Microw Opt Techn Lett 46: 556‒559.
49. Johnson JM, Rahmat-Samii Y (1999) Genetic algorithms and method of moments (GA/MOM)
for the design of integrated antennas. IEEE T Antenn Propag 47: 1606‒1614.
50. Noor AOA (2015) An Automated Optimization Technique of a MSPA for Broadband
Communication. Eng Tech J 33.
51. Lech M, Mitchell A, Waterhouse R (2000) Optimization of broadband microstrip patch
antennas. Asia-Pacific Microwave Conference, 711‒714.
52. Lotfi AA, Kashani FH (2004) Bandwidth optimization of the E-shaped microstrip antenna using
the genetic algorithm based on fuzzy decision making. Antennas and Propagation Society
International Symposium 3: 2333‒2336.

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 331

53. Bulla G, de Salles AA, Vuong TP (2010) PIFA bandwidth optimization using genetic algorithm
and capacitive feeding. IEEE International Conference on Wireless Information Technology and
Systems (ICWITS), 1‒4.
54. Nguyen TD, Duroc Y, Vuong TP (2011) Genetic algorithm for optimization of L-shaped PIFA
antennas. Int J Microw Wirel T 3: 691‒699.
55. Chandran PP, Viswasom S (2014) Gain and bandwidth optimization of a novel microstrip patch
antenna. Fourth International Conference on Advances in Computing and Communications
(ICACC), 315‒318.
56. Tseng LY, Han TY (2010) An evolutionary design method using genetic local search algorithm
to obtain broad/dual-band characteristics for circular polarization slot antennas. IEEE T Antenn
Propag 58: 1449‒1456.
57. Spence TG, Werner DH, Groff RD (2004) Genetic algorithm optimization of some novel
broadband and multiband microstrip antennas. Antennas and Propagation Society International
Symposium 4: 4408‒4411.
58. Jayasinghe JW, Anguera J, Uduwawala DN, et al. (2017) A multipurpose genetically engineered
microstrip patch antennas: Bandwidth, gain, and polarization. Microw Opt Techn Lett 59:
941‒949.
59. Choo H, Ling H (2003) Design of broadband and dual-band microstrip antennas on a
high-dielectric substrate using a genetic algorithm. IEE Proceedings-Microwaves, Antennas and
Propagation 150: 137‒142.
60. Fertas K, Kimouche H, Challal M, et al. (2015) Design and optimization of a CPW-fed tri-band
patch antenna using genetic algorithms. ACES Journal 30: 754‒759.
61. Wakrim L, Ibnyaich S, Hassani M (2016) Novel Miniaturized Multiband and Wideband PIFA
Antenna for Wireless Applications. 3rd International Conference on Automation, Control,
Engineering and Computer Science, 3‒6.
62. Choo H, Ling H ((2002) Design of multiband microstrip antennas using a genetic algorithm.
IEEE Microw Wirel Co 12: 345‒347.
63. Ohira M, Deguchi H, Tsuji M, et al. (2004) Multiband single-layer frequency selective surface
designed by combination of genetic algorithm and geometry-refinement technique. IEEE T
Antenn Propag 52: 2925‒2931.
64. Sathi V, Taherizadeh S, Lotfi A, et al. (2010) Optimisation of multi-frequency microstrip
antenna using genetic algorithm coupled with method of moments. IET Microw Antennas P 4:
477‒483.
65. Michael K, Kucharski AA (2006) Genetic algorithm optimization for Multiband patch antenna
design. First European Conference on Antennas and Propagation, 1‒4.
66. Werner DH, Werner PL, Church KH (2001) Genetically engineered multiband fractal antennas.
Electron Lett 37: 1150‒1151.
67. Ozgun O, Mutlu S, Aksun ML, et al. (2003) Design of dual-frequency probe-fed microstrip
antennas with genetic optimization algorithm. IEEE T Antenn Propag 51: 1947‒1954.
68. Griffiths L, Chung YC, Furse C (2005) Integrated dual band GSM microstrip monopole using
GA and FDTD. Antennas and Propagation Society International Symposium 4: 48‒51.
69. Lu Q, Korolkiewicz E, Danaher S, et al. (2009) Optimum design of a probe fed dual frequency
patch antenna using genetic algorithm.

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 332

70. Villegas FJ, Cwik T, Rahmat-Samii Y, et al. (2002) Parallel genetic-algorithm optimization of a
dual-band patch antenna for wireless communications. IEEE Antennas and Propagation Society
International Symposium 1: 334‒337.
71. Castellana F, Bilotti F, Vegni L (2001) Automated dual band patch antenna design by a genetic
algorithm based numerical code. IEEE Antennas and Propagation Society International
Symposium 4: 696‒699.
72. Fertas K, Kimouche H, Challal M, et al. (2015) An optimized shaped antenna for multiband
applications using Genetic Algorithm. 4th International Conference on Electrical Engineering
(ICEE), 1‒4.
73. Saxena NK, Khan MA, Pourush PKS, et al. (2011) GA optimization of cutoff frequency of
magnetically biased microstrip circular patch antenna. AEU-Int J Electron C 65: 476‒479.
74. Herscovici N, Osorio MF, Peixeiro C (2002) Miniaturization of rectangular microstrip patches
using genetic algorithms. IEEE Antenn Wirel Pr 1: 94‒97.
75. Lamsalli M, El Hamichi A, Boussouis M, et al. (2016) Genetic algorithm optimization for
microstrip patch antenna miniaturization. Prog Electrom Res LE 60: 113‒120.
76. Gupta VR, Gupta N (2006) Realization of a compact microstrip antenna: An optimization
approach. Int J RF and Microw C E 16: 367‒373.
77. Chattoraj N, Roy JS (2007) Application of Genetic Algorithm to the Optimization of Gain of
Magnetized Ferrite Microstrip Antenna. Eng let 14.
78. Kim D, Ju J, Choi J (2012) A mobile communication base station antenna using a genetic
algorithm based Fabry-Perot resonance optimization. IEEE T Antenn Propag 60: 1053‒1058.
79. Jayasinghe JW, Anguera J, Uduwawala DN (2013) A high-directivity microstrip patch antenna
design by using genetic algorithm optimization. Progress In Electromagnetics Research C 37:
131‒144.
80. Chattoraj N, Roy JS (2006) The optimization of gain of patch antennas using Genetic Algorithm.
ACTA TECHNICA CSAV 51: 279.
81. Zhu X, Shao W, Li JL, et al. (2012) Design and optimization of low RCS patch antennas based
on a genetic algorithm. Progress In Electromagnetics Research 122: 327‒339.
82. Mittra R (1997) Genetic algorithm: the last word for solving all of your design problems.
Antennas and Propagation Society International Symposium 3: 1672‒1675.
83. Lu Q, Edward K, Sean D, et al. (2010) Optimum design of a probe fed dual frequency patch
antenna using genetic algorithm. ARMMS RF and Microwave Society, 20‒21.
84. Kim Y, Yun W, Yoon Y (1999) Dual-frequency and dual-polarisation wideband microstrip
antenna. Electron Lett 35: 1399‒1400.
85. Jayasinghe J, Anguera J, Uduwawala D (2015) On the behavior of several fitness functions for
genetically optimized microstrip antennas. Int J Sci World 3: 53‒58.
86. Lee D, Lee S (2000) Design of a coaxially fed circularly polarized rectangular microstrip
antenna using a genetic algorithm. Microw Opt Techn Lett 26: 288‒291.
87. See CH, Abd-Alhameed RA, Zhou D, et al. (2006) A new design of circularly-polarised
conical-beam microstrip patch antennas using a genetic algorithm. First European Conference
on Antennas and Propagation, 1‒4.
88. Zentner R, Sipus Z, Bartolic J (2001) Optimization synthesis of broadband circularly polarized
microstrip antennas by hybrid genetic algorithm. Microw Opt Tech Lett 31: 197‒201.

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.

 333

89. Grimaccia F, Mussetta M, Pirinoli P, et al. (2006) Genetical swarm optimization (GSO): A class
of population-based algorithms for antenna design. ICCE'06 First International Conference on
Communications and Electronics, 467‒471.
90. Günel T, Erer I (2002) Application of fuzzy genetic algorithm to the problem of synthesizing
circular microstrip antenna elements with thick substrates. AEU-Int J Electron C 56: 215‒217.
91. Zhang YJ, Gong SX, Wang X, et al. (2010) A hybrid genetic-algorithm space-mapping method
for the optimization of broadband aperture-coupled asymmetrical U-shaped slot antennas. J
Electromagnet Waves 24: 2139‒2153.
92. Pissoort D, Rogier H, Olyslager F, et al. (2003) Optimization of a microstrip antenna with a
genetic algorithm for use as a ground penetrating radar. J Electromagnet Waves 17: 1197‒1216.
93. Akila M, Anusha P, Sindhu M, et al. (2017) Examination of PSO, GA-PSO and ACO
algorithms for the design optimization of printed antennas. IEEE Applied Electromagnetics
Conference, 1‒2.
94. Mishra S, Chattopadyay S, Gangopadhyaya M (2015) A comparative study of DE, PSO and
BFO for optimisation of Rectangular Microstrip Patch Antenna with inset feed parameter.
International Conference and Workshop on Computing and Communication (IEMCON), 1‒7.
95. Boada DFM, Sakomura ES, Nascimento DC (2021) Cavity model surrogate-based optimization
for electrically thick circularly polarized rectangular microstrip antennas. AEU-Int J Electron C
131: 153597.
96. Koziel S, Bekasiewicz A (2016) Multi-objective design of antennas using surrogate models.
World Scientific.
97. Lee W, Jung M, Kim TH, et al. (2020) Patch antenna design using the GA and the binary
method for dual-band applications. J Electromagnet Waves 34: 1691‒1704.
98. Awan WA, Zaidi A, Hussain M, et al. (2021) The Design of a Wideband Antenna with
Notching Characteristics for Small Devices Using a Genetic Algorithm. Mathematics 9: 2113.

© 2021 the Author(s), licensee AIMS Press. This is an open access

article distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/4.0)

AIMS Electronics and Electrical Engineering Volume 5, Issue 4, 315–333.