Jestr Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

Review Article
JOURNAL OF
Engineering Science
and Technology Review

r
www.jestr.org

Antenna Optimization using Machine Learning Algorithms and their Applications:

A Review

Anil Kumar Pandey* and M. P. Singh

Department of Computer Science & Engineering, National Institute of Technology, Patna, India

Received 28 November 2023; Accepted 31 January 2024

___________________________________________________________________________________________

Abstract

Antenna optimization using machine learning is a rapidly evolving field that leverages the power of artificial intelligence
to design and improve antenna systems. Antenna optimization is a process of modifying antenna parameters to achieve
desired performance metrics, such as gain, bandwidth, radiation pattern, and impedance matching. This paper presents a
review of the most advanced development in antenna design and optimization by using machine learning techniques. The
aim of this survey is to focus on different machine learning optimization techniques and their optimization capability with
efficiency challenges. A deep outline from literature survey on optimization of antennas using machine learning are
presented and listing various optimization algorithms and procedures that are applied to produce desired antenna
characteristics and specifications. Firstly, a brief introduction of machine learning and its algorithms, later a quick
explanation of antenna optimization process followed by an arranged introduction of different types of printed antenna
designs using machine learning algorithm are reported. The methods emphasized in this survey have probably an effect on
the imminent advancement of antennas for a variety of wireless applications.

Keywords: Microstrip Antenna, Optimization, Machine Learning, Evolutionary Algorithm, Wireless Communication.
____________________________________________________________________________________________

1. Introduction with finite details, and the expanded structure of antenna

simulation takes large time in HFSS and CST. The fullwave
In modern wireless communication systems, design of electromagnetic simulation for antenna design and
antenna arrays and their optimization, integration, and optimization is very time consuming and leads to various
fabrication are turn into more complication. Traditionally, limitations. Now days efficiency and optimization ability of
antenna optimization has been done through multiple existing antenna optimization techniques is the main
approaches. One of them has trial-and-error approach, where limitation to report an extensive background of antenna
engineers manually tweak antenna parameters until the design challenges. To overcome the limitation, different
desired performance has been achieved. This approach is algorithms of machine learning (ML) have been considered
time-consuming, resource-intensive, and often limited by for optimizing various parameters of antennas for different
human expertise. Nowadays, a huge requirement of compact, applications [10-13]. Machine learning offers a more efficient
multiband, and wideband for higher data rate with low loss and effective way to optimize antennas by leveraging data-
antenna for advanced applications of wireless driven models and algorithms. Machine learning can
communication. To design and analyze of desired antennas in automatically learn from large amounts of data and find
[1], few methods are applied for numerical validations patterns and correlations between antenna parameters and
specifically: finite difference time domain (FDTD) [2,3], performance metrics. This enables engineers to design better
finite element method (FEM) [4,5], and method of moments antennas in less time and with fewer resources. Machine
(MoM) [6,7]. A modified FDTD method is applicable for learning algorithm for antenna optimization has been broadly
thorough analysis of more than one dielectric interfaces [8]. presented to step up the design procedure of antennas and
The method of moments needs accurate computation of the arrays for different applications. Various ML techniques,
impedance matrix for calculating precise values of currents, such as Gaussian Process Regression (GPR), Support Vector
impedance, and resonance frequency of the rectangular Machine (SVM) and Artificial Neural Networks (ANNs),
shaped and nonrectangular shape patch antennas [9]. Antenna have been employed to develop alternate models of antenna
design and simulation are widely preferred easily available to predict quick outcomes. In [14] a model is generated by
commercial Computational Electromagnetics (CEM) using variable-fidelity electromagnetic (EM) simulations and
software such as Advanced Design System (ADS), high- co-Kriging. A genuine approach of combining sampled
frequency structure simulator (HFSS), Computer Simulation accurate EM data and densely sampled coarse-discretization
Technology (CST), and Integral Equation Three-Dimensional in one model has been applied to design low-cost antenna.
(IE3D). All CEM software tools are very efficient, but due to Another approach has been proposed for low optimization
absence of few significant features, limited performance cost with little updated responses of antenna [15]. In [16], a
achieved. For example, ADS have lack of 3D structures multi objective model of an antenna has been designed by
modelling, IE3D have no facilities of structures modelling using kriging interpolation of coarse-discretization simulation
data and known as multi-objective evolutionary algorithm. A
______________ complete performance evaluation of various miniature UWB
*E-mail address: [email protected]
ISSN: 1791-2377 © 2024 School of Science, DUTH. All rights reserved.
doi:10.25103/jestr.172.14
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

antenna based on optimization techniques has been studied in optimization of microstrip antenna parameters using ANN for
[17]. X band (8-12 GHz) and Ku band (12-18 GHz) applications.
In recent decades, many literatures reported various ML An advanced Machine-learning-assisted optimization method
optimization algorithm such as ANNs, SVM and GPR have has been reported for antenna optimization with maximum
been developed for optimization of antenna parameters. The gain 7.4 dBi in [45]. Research paper [46] proposed a new
implementation and complex problems are easily handled by design of isotropic antenna by using machine learning
gaussian process (GP) than ANN and SVM. As per definition, algorithm. Structured supervised learning, which is an
deep learning algorithms and Convolutional Neural Network alternative of neural networks, proposed in [47] to design
(CNN), includes convolutional calculations and a deep patch antenna for the Ku EM band. In [48], ANN has been
structure. Literature [18] presented a deep GP model by utilized for the design of multi-slot microstrip antenna and
combining the structural CNN with GP. In this paper design compared with simulated results by using IE3D software.
parameters can reduce to decrease the training sample data Optimization using ANN of a frequency reconfigurable
and increase the designing efficiency with steady accuracy. planar antenna with metasurface superstrate has been reported
Letter [19] presents a huge shaped-beam reflectarray design in [49]. A modelling using the Gaussian Process Regression
by using a new technique based on SVR. This technique used (GPR) of electromagnetic band gap two-port multiple-input
to find reflection coefficients for fast designing of reflectarray and multiple-output (MIMO) fractal antenna has been
for direct broadcast satellite application. Paper [20] reported presented in [50].
a multi-objective optimization model of antenna by using Present study provides a broad report on different
nested kriging algorithm with effective. To overcome the optimization process that can be applied on designing of an
limitation of optimization efficiency for the designing of antenna with desired parameters. This report gives a quick
electromagnetic (EM) device and machine, a new technique knowledge of different types of ML techniques and types of
named parallel surrogate model-assisted evolutionary microstrip antenna. It helps to researchers in optimization of
algorithm has been proposed in [21] by using mutation antenna who would like to apply ML algorithm for antenna
operators in a parallel computing environment. In [22] various design. In starting of the paper, a brief overview on ML, types
ML algorithm have been employed to analyse and optimize of ML and different literatures on ML are presented.
the performance of different antennas. A nano antenna has Following section elaborates optimization models, ML
been optimized in [23] to achieve minimum loss and algorithms and evolutionary algorithms for efficient
maximum radiation efficiency. To analyse aperiodic array computation to design an antenna. To attract the attention of
with uniform excitation, a novel multi-objective optimisation reader, next section includes the in-depth outlines on different
technique known as non-dominated sorting and local search printed antenna designs and optimization using ML and
(NSLS) has been proposed [24,25]. To improve efficiency a organized according to the type of the antenna. Next section
three stage multi-fidelity-simulation-model assisted antenna investigates another aspect of the literature, where ML has
design optimization outline has been presented in [26]. This been utilized to improve the optimization characteristics of an
outline is employed on SADEA and creating SADEA-II. A antenna for different applications. Concluding explanations
brief study in literature [27,28] demonstrates various printed has been presented in the last section. All the segments of this
antenna designs by utilizing ANN and other machine learning paper are briefly expressed and listed with recommended
approaches. Paper [29] proposed a technique utilizing neuro- literatures as per requirement. Antenna optimization using
fuzzy networks to evaluate the antenna’s resonance frequency machine learning algorithms presents a promising and
of. A new type of ANN has been described in [30] for the evolving field with significant scope and opportunities.
analysis of antenna parameters. A dual band H-shaped patch Machine learning algorithms can optimize antenna
microstrip antenna has been designed by applying machine parameters to enhance performance metrics such as gain,
learning techniques [31] to achieve compactness. An bandwidth, and radiation pattern. ML algorithms enable
evolutionary algorithm known as Particle Swarm antennas to adapt to changing environmental conditions,
Optimization (PSO) has been used to design stacked patch ensuring optimal performance in dynamic scenarios, like
antenna in [32] with the neural networks approach for the mobile communication or satellite systems. ML can aid in
application of satellite communication. designing multifunctional antennas that can efficiently
Several limitations of the EM simulations and operate across multiple frequency bands and communication
evolutionary algorithms (high computational cost, low standards. ML algorithms can improve the beamforming
efficiency, and large optimization time) restrict the antenna capabilities of antenna arrays, optimizing radiation patterns
usages in many applications. To overcome this, an advance for better signal reception and transmission. It can be applied
method named surrogate model assisted differential evolution to mitigate interference in crowded wireless environments,
for antenna synthesis (SADEA), has been reported in [33]. A ensuring reliable communication in the presence of
new ML technique based on the modified K-nearest competing signals. Antenna optimization is vital for satellite
neighbour (KNN) algorithm incorporates advanced communication systems, and ML can contribute to improving
simulation methods with more features from data sets has link quality, data rates, and overall system efficiency. With
been presented in [34]. A reflect array antenna has been the advent of 5G and future communication technologies, ML
analysed through ANN in [35] and a tunnel-based ANN has can play a key role in optimizing antennas to meet the
been developed for the analysis of antenna parameters in [36]. demanding requirements of high data rates, low latency, and
Nowadays many literatures present a thorough survey on the massive device connectivity. There is an opportunity to
current development in antenna design optimization and develop novel machine learning algorithms tailored
emphasized on various methods that include optimization specifically for antenna optimization, considering the unique
ability and efficiency problems using machine learning [37- challenges and requirements of the domain. Explore the
43]. It provides a brief note on ML algorithm and various integration of deep learning techniques for more complex
types of microstrip antenna. It also supports to researchers optimization tasks, leveraging neural networks to model
having least expertise in ML in the field of antenna and desire intricate relationships in antenna design. Develop algorithms
to utilize ML algorithm in study. Literature [44] reported an that allow antennas to adapt in real-time to changing
129
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

conditions, providing continuous optimization for improved autonomously without any human support or program. Such
communication quality. Contribute to the development of assessment is made by obtaining significant fundamental
standards and best practices for applying machine learning to samples in complex data. Several research have been
antenna optimization, ensuring consistency and reliability surveyed on emerging applications of ML, like antenna
across different applications. Identify and explore design and optimization [51-53]. Classification of ML has
commercial applications of machine learning-optimized been done into three basic categories; supervised,
antennas, such as in consumer electronics, automotive unsupervised and reinforcement learning, according to input
communication systems, and IoT devices. and output data types to solve the various problems. One other
learning methods like Semi supervised [54] has been formed
to apply hybrid approach for solving beyond the original
2. Machine Learning Overview observation range. In [55] a brief comparison between
Supervised and Unsupervised Learning Algorithms has been
Machine learning is the field of computer algorithms explained for Pattern Classification. A review on Feature
combined with dataset, that supports systems to understand Selection of Supervised, Unsupervised, and Semi-Supervised
automatically and improve from past data. It is a part of has been reported in [56, 57]. A detailed study and analysis of
artificial intelligence (AI) that allows software applications to basic learning processes with different ML algorithm are
analyse the data accurately. Machine learning algorithms listed in Figure 1 with diagrammatic representation.
permits the computers to process the data and assess

Fig. 1. Brief description of various types of Machine Learning

130
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

2.1 Supervised Learning

Supervised learning applies labelled data to train the machine
[58]. The labelled data indicates, a few input data is used as a
training data that provides the accurate output. To resolve the
various problems related to real-world such as fraud
detection, spam filtering, risk assessment, image
classification etc., supervised learning model supported
efficiently. Several literatures have been reported different
applications analysed by supervised learning algorithm such
as, detection of Network Intrusion [59], Recognition of Fig. 2. (a) Supervised Learning Graph, (b) Unsupervised Learning Graph,
Human Activity in Ambient Assisted Living Environment and (c) Reinforcement Learning Process
[60], Earth Science [61] and diagnosis of abnormal voltage
[62]. Based on problem types, supervised learning categories
in two kinds of algorithms, namely classification and 3. Machine Learning Algorithms
regression. If the output variables are categorical, like
positive-negative, yes-no, true false etc. then classification To predict the output, observe the hidden patterns from
algorithms are employed. Decision trees, logistic regression, dataset and maximize the positive feedback from experiences,
random forest and SVM are some common classification machine learning algorithms are applied by programming.
algorithms explained in [63]. If there are some relations Different ML algorithms are used for different problems
between input and output variables are exist, then regression according to the complexity [82-88]. For example, linear
algorithms are employed [64-66]. Various common regression can be used for predictions like sales, salary,
regression algorithms have been explained in literatures like age, stock market prediction, the KNN algorithm can be used
linear regression [67,68], regression trees, non-linear for Text mining, Agriculture, Finance, Medical, Facial
regression, bayesian linear regression, support vector recognition, and Gaussian process regression can be used for
regression [69-71] and polynomial regression. process monitoring and fault detection. Some trendy Machine
Learning Algorithms are listed below.
2.2 Unsupervised Learning
Unsupervised learning applies unlabelled data to train the Ø Artificial neural network [89, 90]
machine. The unlabelled data indicates, the input data Ø Support vector machine [91]
without corresponding output values and applied after finding Ø Gaussian process regression [92-94]
hidden patterns of dataset. Unsupervised learning is mostly Ø Linear regression
used for complicated problems as compared to supervised Ø Least absolute shrinkage and selection operator [95]
learning. Literature [72] reported detection of zero-day Ø Kriging regression [96-98]
attacks using Unsupervised Algorithms. Paper [73]
investigates different techniques and applications of A neural network is a data processing system contains
Unsupervised Machine Learning in Networking. Some other many simple, highly interconnected processing elements in an
applications like early fault detection [74], iron mining [75], architecture encouraged by the brain system. Paper [89]
Network Intrusion Detection [76, 77] and deep convolutional provides a brief discussion on neural networks and defines
neural networks enhancement [78, 79] have been reported and applications to a variety of real-world problems. Article [90]
studied in various papers. Based on dataset types gives some ANN implementations with basic theory of ANN
unsupervised learning is classified into two categories: and its specific applications. Article [91] proposed an
Clustering and Association. In clustering, inherent groups are effective deep learning approach STL-IDS based on the self-
created from given dataset to predict output. Association is taught learning (STL) framework to reduce the training and
applied to discover the links between variables in the big testing time. A method has been proposed in [92] by using
dataset. Several common unsupervised learning algorithms Gaussian process regression to model spatial functions for
are neural networks, principal and independent component mobile wireless sensor networks. An investigation of the
analysis, K-means clustering, K-Nearest Neighbours (KNN), Gaussian process approach for classifying multisource and
hierarchal clustering, Apriori algorithm, anomaly detection hyperspectral remote sensing images has been reported in
and singular value decomposition. [93]. [94] summarize different ML algorithms which are not
based on neural networks for the analysis of accuracy of the
2.3 Reinforcement Learning final model. Paper [95] modify the selection process of
Reinforcement learning applies feedback based automatically LASSO to explicitly leverage combinatorial sparsity models
learning technique. During the learning process, an agent through the combinatorial selection and least absolute
learns to perform in a situation by executing the actions and shrinkage operator. [96] provides progressive tutorial on
seeing the results either positive or negative rewards for the regression, kriging, and stochastic kriging. The co-kriging-
actions. The main objective of reinforcement learning is to based antenna structure optimization algorithm has been
maximize the positive rewards by enhancing the performance. proposed in [97]. Paper [98] proposed a cost-efficient
There are two types of reinforcement learning: Positive and Gradient Enhanced Kriging modeling of dielectric resonator
Negative. The positive reinforcement learning increases the antenna structures for considerable reduction of the model
strength and the frequency of the expected performance. The setup cost.
negative reinforcement learning increases the strength of
specific performance that will stop or avoid negative rewards.
Reinforcement learning mostly utilized in game-playing and 4. Machine Learning for Antenna Design
robotics [80,81]. Figure 2 illustrates different machine
learning approaches to solve the problems.
131
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

The need of antennas for various applications expanding from Another popular error function in antenna optimization is
electromagnetic to thermal going through mechanical or the mean absolute error (MAE), which is the average of the
space limits is increased for recent technology development. absolute differences between the predicted and actual values
Printed antennas become a most favourable class of antennas of the performance metrics.
to build an advanced device for communication systems, due
to their notable merits such as cost-effective, easy fabrication, ∑$!%& abs(y! − x! )
compactness, conformable, and adequate gain. Several MAE = (2)
n
literatures reported the use of machine learning in the design
process of an antenna. The machine learning techniques speed Where y! : Actual value for the ith observation
up the design process of an antenna by providing high x! : Calculated value for the ith observation
accuracy, minimum error, less execution time, accurate n: Total number of observations
prediction of the antenna operation, good computational
efficiency, and minimum number of required simulations. The MAE is more robust to outliers than the MSE and is
Following steps are involved in the process of antenna less sensitive to small changes in the predicted values. Other
optimization using machine learning: error functions that can be used in antenna optimization
include the cross-entropy loss, the hinge loss, and the KL-
1. Data collection: The first step is to collect data on divergence. These error functions are often used in
the performance of the antenna. This can be done classification and regression problems and can be adapted to
through simulations or measurements of the the antenna optimization problem by encoding the constraints
antenna's radiation pattern, gain, and other relevant and objectives of the problem in the loss function. Overall, the
metrics. choice of error function depends on the specific problem and
2. Feature extraction: The data collected in step 1 is the objectives of the optimization. The error function should
then processed to extract relevant features that be chosen carefully to balance accuracy, robustness, and
describe the antenna's performance. These features computational efficiency.
can include parameters such as frequency, This section focuses on optimization process,
bandwidth, and polarization. conventional and evolutionary optimization algorithm of
3. Model training: Once the features have been machine learning for antenna design. The papers reviewed in
extracted, a machine learning algorithm is trained on this section are categorized in three area: optimization
the data to learn the relationship between the modelling methods, machine learning optimization
antenna parameters and the performance metrics. techniques and evolutionary computation algorithms.
4. Optimization: With a trained machine learning
model, the antenna design can be optimized by 4.1 Optimization of Antennas
using the model to predict the performance of Machine learning based antenna optimization prefers specific
different antenna configurations. The optimization and evolutionary algorithm which are based on the
process typically involves exploring the design optimization characteristic, efficiency, simplification, and
space using techniques such as genetic algorithms or robustness. There are various modelling methods are used to
reinforcement learning. optimize the design of an antenna with good optimization
5. Validation: Finally, the optimized antenna design is outcomes. Some models namely, Single-objective
validated using simulations or measurements to optimization, multi-fidelity optimization, multi-objective
ensure that it meets the desired performance metrics. optimization and yield-driven optimization that are used for
different antenna optimization problems. Antenna
The antenna optimization using machine learning optimization modelled as single-objective optimization
algorithms provides a powerful tool for engineers to design problems to maximization of the isotropic gain has been
and optimize antennas with improved performance presented in [99]. To enhance optimization efficiency of an
characteristics. To evaluate the accuracy of the machine antenna, multi-fidelity optimization is used to model the
learning model and the optimization algorithm, it is essential problem of antenna design. This technique uses high-fidelity
to calculate the error or loss function, which measures the expensive and precise models and to remove unwanted
difference between the predicted and actual values of the solutions, low-fidelity, imperfect and inadequate models
performance metrics. There are several types of error [100]. In [101], a conventional multi-objective optimization
functions that can be used depending on the nature of the problem has been illustrated and studied. This paper reported
problem and the objectives of the optimization. a design of compact planar antenna with low value of
One common error function used in antenna optimization reflection coefficient in desired operating bands using multi-
is the mean squared error (MSE), which is the average of the objective optimization. Paper [102] depicts statistical
squared differences between the predicted and actual values analysis-based yield-driven optimization for the analytical
of the performance metrics. The MSE is a widely used metric evaluations of deviations. Due to this probability of assumed
in machine learning because it is differentiable, convex, and and nominal design performance specifications of a
easy to interpret. The Mean Squared Error is calculated as: fabricated prototype is increases.

∑$!%&(y! − y)" )# 4.2 Machine Learning Optimization Techniques

MSE = (1) Design of an antenna must have some important parameters
n
like implementation, usability, and cost-efficient.
Where y! : Actual value for the ith observation Optimization is the process of making a model as good as it
x! : Predicted value for the ith observation can be. Optimization using machine learning is one of the
n: Total number of observations important methodologies to get improved results of designed
model by comparing various solutions using iteration.
Optimization process, design a model with minimum
132
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

production cost and maximum production efficiency. Dipole, ultra-wide band, multiband antennas etc. In following
Antenna optimization consists proper range of functions, section some of the printed antenna types, design, and
design parameters, variables, and limitations. There are optimization based on literature survey are described in brief.
several approaches to using machine learning for antenna Figure 4 listed various printed antenna types, shapes, feeding
optimization, including reinforcement learning, genetic techniques and their applications in different fields.
algorithms, and neural networks. Reinforcement learning
involves training an agent to make decisions based on
feedback from the environment. Genetic algorithms mimic
natural selection to evolve antenna designs over multiple
generations. Neural networks use deep learning techniques to
model the relationship between antenna parameters and
performance metrics. Several literatures are discussed
antenna optimization using machine learning for different
applications. There are different ML techniques are illustrated
which can optimize an antenna design parameter:
• Gradient descent [103]
• Adaptive moment estimation Fig. 4. Printed antenna types, shapes, feeding techniques and
• Levenberg-Marquardt algorithm [104] applications.
• Bayesian regularization [105]
5.1 Rectangular Patch Antenna
4.3 Machine Learning Evolutionary Algorithms The most utilized printed antenna is a rectangular patch
An optimization process is required to calculate nearest antenna as shown in figure 5 that appears like a truncated
optimal values of design parameters of antenna. Use of microstrip transmission line. A CAD model of square shaped
evolutionary algorithms are new approach to applying patch antenna is designed by using Neural network in [113].
machine learning in antenna design. Evolutionary Algorithms To obtain and analysed the exact resonant frequency of a
play an important role in computing sciences. It allows to rectangular patch antenna, ANN has been used in paper
achieve nearest solutions to optimization problem and [114,115]. Another research based on ANN technique has
implicated in real selection. To improve the performance been reported in [116] include calculation of the rectangular
characteristics of an antenna for space communication and shaped patch length and width of an antenna by training
faster computations, evolutionary algorithms are effectively through Bayesian Regularization and Levenberg Marquart
employed in machine learning techniques. Figure 3 illustrates algorithms. Particle Swarm Optimization (PSO) optimization
block diagram of antenna optimization process using ML process has been studied in [117] for the designing of ANN
algorithm. based printed antenna. To calculate the size of slots and air
Some standard evolutionary algorithms like, particle gap of an antenna, an ANN based synthesis model has been
swarm optimization and differential evolution, covariance presented in [118]. Radiation patterns of an antenna has been
matrix adaptation evolution strategy, genetic algorithms and analysed in [119] by using a tunnel-based ANN. In [120], an
simulated annealing have been explained in [106]-[112] for analysis of an antenna has been done by using both ANN and
antenna design and optimization. SVM algorithm. A brief comparison between the results of
two advanced nonlinear ML algorithms depicts superiority of
SVM over the ANN. To calculate the resonant frequency of
rectangular shaped patch antenna designed by ANN in [121],
uses various algorithm like feed forward back propagation,
resilient back propagation, Levenberg Marquardt and radial
basis functions.

Fig. 3. Antenna optimization process using ML algorithm.

5 Printed Antennas Using Machine Learning

This section provided a comprehensive survey of several

papers on antenna design using machine learning. An antenna
fabricated by using lithographic methods on a printed circuit
board is known as Printed antenna or microstrip antenna. It
has several advantages like mechanical durability,
conformability, small sizes, and low manufacturing costs.
They are generally used at microwave frequencies for both
the military and commercial applications. There are various Fig. 5. (a) Rectangular Patch Antenna, (b) Circular Patch Antenna, (c)
size and shapes of printed antennas are available which are Dipole, (d) different shaped patch, (e) Slotted Patch and (f) Array
optimized using different optimization techniques of machine Structure.
learning. Some common types of printed antennas are
rectangular patch, circular patch, fractal, slotted, MIMO,
133
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

5.2 Circular Patch Antenna modern ML techniques including least absolute shrinkage and
A multilayer circular microstrip antenna has been optimized selection operator, ANNs, and KNN have been used to
using training algorithm named Levenberg - Marquardt to optimization process in [139].
train the Multi- Layer Perceptron Neural Networks in [122].
In paper [123] delta-bar-delta, extended delta-bar-delta, 5.6 Slotted Antenna
quick-propagation, directed random search and genetic Literature [140] depicts a model using neural network for
algorithms have been utilized to train the Multi- Layer calculating the various performance parameters like;
Perceptron Neural Networks. This paper proposes electrically resonance frequencies, antenna gains, antenna directivities,
thin and thick circular printed antennas by evaluating the antenna efficiencies for dual-band slotted microstrip
resonant frequency, characteristic impedance and the antennas. Reported model of neural network has been
effective permittivity using only one neural model. Another applicable for all varied dimensions of slot and incorporated
paper [124] developed multilayer perceptron based on neural air-gap. A knowledge-based neural network model proposed
network for circular patch antenna design to determine feed in [141] to obtain desired values of resonance frequency,
position. A circular patch antenna model has been optimized antenna gain, directivity, and radiation efficiency of a dual-
in [125] by using Quasi Newton model of Artificial Neural band antenna for accurate prediction of slot-size, that has been
Network. To determine all performance constraints like S11, presented on the radiating patch. A significant drop in training
Voltage Standing Wave Ratio (VSWR), resonant frequency, attempts has been achieved, due to prior knowledge of neural
impedance bandwidth and antenna gain of circular patch model. In [142], a neural networks prototype developed for
antenna a Neuro-Computational approach has been applied in calculating the desired size of slot on the patch and air-gap
[126]. between the ground plane and substrate. A slot loaded
rectangular microstrip antenna has been reported in [143] and
5.3 Dipole modelled by using support vector regression method.
An input impedance of dipole antenna has been optimized Article [144] illustrates desired calculation using SVM
using synthesis ANN with FDTD method in [127] for formulation for evaluating different characteristics of slotted
practical applications. This ANN model employs a hetero- microstrip antenna like resonant frequency, antenna gain,
associative memory on input and output data for the directivity, and radiation efficiency with modified ground
computation of fault toleration. In addition, a latest plane.
randomization method for the optimization of antenna
parameters has been reported. Another new SYNTHESIS- 5.7 Ultra-wideband Antenna
ANN model in [128] has been generated for solving the A compact monopole antenna with coplanar waveguide feed
intensive problems and optimization of microwave devices. has been analysed and optimized by using machine learning
5.4 Fractal Antenna techniques to speed up the process of antenna design in article
An ANN based method has been applied and projected in [145]. This article approaches to five algorithms for the
[129] to design a double band fractal antenna by calculating designing of band-notch antenna named as decision tree,
return loss, resonance frequency and antenna gain. Paper random forest, XGB regression, KNN, and ANN. In paper
[130] developed, an advanced square shaped fractal antenna [146], ANN has been used to analyse and predict the
using Advanced C and optimized by ANN. A quasi-fractal impedance bandwidth and notch frequency of an ultra-
patch antenna model efficiently synthesized by using wideband (UWB) antenna. A multi-adaptive neuro-fuzzy
multilayer perceptrons ANN with one hidden layer and inference system has been developed in [147] for calculating
resilient backpropagation training algorithm in [131]. A PSO the impedance bandwidth and notch frequencies of proposed
based ANN has been developed in [132] by utilizing slotted UWB antennas. Article [148] proposes an
hybridized algorithm to optimize a miniaturized circular optimization process that was achieved by utilizing a self-
fractal antenna resonate at 2.45 GHz. Another custom-made organizing multi-objective genetic algorithm and apply on a
fractal antenna illustrated in [133]. In this article, proposed ring monopole microstrip antenna design by predicting
antenna has been designed by using ANNs and PSO impedance bandwidth, S11 and central frequency deviation.
technique. For the biomedical applications, a miniaturized This antenna is applicable for ultra-wideband operations. To
giuseppe peano and cantor set fractals-based hybrid fractal improve a multi-objective evolutionary algorithm, a machine
antenna reported in [134]. Also, a relative analysis of three learning application has been reported in [149] that is used to
different ANNs has been done to assess the truly excellent predict the nature of fitness function for the accurate design
kind of ANN for the evaluation of proposed antenna. of UWB antenna.

5.5 Various Shaped Patch Antenna 5.8 Array Antenna

A deep neural network has been applied in [135], to compute An array antenna is defined as a set of multiple linked
the resonance frequency of a printed antenna with E– antennas that will work simultaneously as an individual
shaped resonating patch. Paper [136] proposes a small size antenna. Usually, an array is used to obtain high gain, path
microstrip antenna radiated in frequency range from diversity characteristics, interference reduction, beam
0.4856GHz to 7.8476 GHz. In this article a regression-based steering, and for radio direction finding. A robust GA along
machine learning has been employed to calculate resonant with rapid neural network method has been developed for
frequency, slot dimensions and patch dimensions at dominant optimization of a proposed wideband aperiodic linear phased
mode. A Gaussian process regression-based method antenna array in [150]. This paper presents a modern nature-
developed in [137] for the precise computation of resonant based (GA-NN) technique that can design and estimate the
frequencies of proposed dual-band microstrip antennas. In performance parameters with desired element positions in
[138], optimization of a designed antenna has been described given band of operation and scanning range to reduce the
by using Gaussian process regression with a genetic algorithm element VSWRs and sidelobe levels. A novel beamformer
(GA) structure. To achieve desired performance parameters neural networks-based antenna array has been reported in
of a dual-band double T-shaped monopole antenna, some [151]. A new variant of invasive weed optimization known as
134
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

modified adaptive dispersion has been used to train the large multi-dimensional parameter optimization within an
optimized datasets in neural networks training process. In appropriate Time. A brief investigation on different antenna
paper [152] a simulation-based Machine Learning method for design using machine learning is also listed in Table 1.
an antenna optimization has been reported, that provides a

Table 1. Different design process of antenna using machine learning and evolutionary algorithm.
Ref. Antenna Type ML Results Compared To Evolutionary Algorithm Optimization
Algorithm Parameters/
Performance
Parameters
[108] Microstrip antennas and arrays Differential Simulated and measured results Self-Adaptive DE, DE, Antenna parameters
Evolution Biogeography-based DE,
(DE) modified DE, Improved DE,
Harmonic Search DE
[109] Log-periodic antenna -- Evolutionary algorithms DE, PSO, Taguchi, invasive Dipole lengths and
weed, adaptive invasive weed spacing, dipole wire
diameters /Antenna
gain, VSWR, front-
to-back ratio
[113] Rectangularpatch antennas ANN IE3D electromagnetic -- Dielectric constant,
simulator substrate
thickness/resonant
frequency
[114] Rectangular microstrip antenna ANN Conventional simulations Multilayer perceptron modular Resonant frequency
neural network
[115] Rectangular microstrip antenna ANN -- -- Parameters of
antenna
[116] Rectangular patch antenna ANN Theoretical value Bayesian regularization and Length and width of
levenberg marquart antenna
[117] Rectangular or Circular Microstrip ANN Conventional simulations and ANN Optimized by PSO --
Patch Antenna measured antenna values Algorithm
[118] Rectangular or circular microstrip ANN Measured results -- Slot-Size, Airgap/
patch antenna Resonance
Frequencies, Gains,
Directivities,
Antenna
Efficiencies, and
Radiation
Efficiencies
[119] Multi-slot hole-coupled microstrip Tunnel- IE3D software, GA Radiation patterns,
antenna based ANN analytical and experimental resonant frequency
results
[120] Rectangular patch antenna SVM Theoretical results -- Resonant
and ANN results frequency,
impedance
bandwidth and input
impedance

[121] Rectangular patch antenna ANN Standard formula and -- Dielectric constant,
experimental results thickness of
substrate, patch
width and
length/resonant
frequency

[122] Multilayer circular microstrip anten ANN Measured and calculated result Levenberg – marquart algorith Resonant frequency
na s m
[123] Circular patch antenna Neural Experimental results Delta-bar-delta, extended delta- Resonant
models bar-delta, quick-propagation, frequency,
directed random search and GA characteristic
impedance, and
effective
permittivity
[124] Circular microstrip antenna MLP neural Experimental results -- Feed position
network
Radial basis
function
[125] Circular microstrip antenna ANN Experimental results Quasi newton model Resonant frequency
[126] Circular microstrip antenna ANN Simulated, measured and Neurocomputational models Return loss, VSWR,
theoretical results resonant frequency,
bandwidth, gain and
antenna efficiency
[127] Printed dipole antenna ANN Finite-difference time-domain Synthesis ANN Input impedance
results
[129] Elliptical fractal patch antenna ANN Simulated results obtained -- Resonant
using IE3D software frequency, return
loss and gain
[130] Square fractal antenna ANN Simulated using HFSS -- Antenna
parameters/

135
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

resonance
characteristics
[131] Quasi-fractal patch antenna Multilayer Simulated and measured results Efficient resilient Thickness and size
perceptrons backpropagation algorithm of antennas
ANN
[132] Circular fractal antenna PSO based Simulated, desired, and -- --
selective experimental results
ANN
[133] Custom-made fractal antennas ANN and Simulation and experimental -- fractal antenna
PSO results parameters/
operational
frequencies
[134] Hybrid fractal antenna ANN Simulated, and experimental Firefly algorithm Feed position
results
[135] E–shaped patch antenna Deep neural Deep neuro-computing model Resonant frequency
network
[136] Square patch antenna GPR Simulated and measured -- Slot size and patch
dimensions/
resonant frequency
[137] Dual-band microstrip antennas GPR -- Resonant
frequencies
[138] Dual-band CPW-fed slot antennas GPR Moment-method-based GA Several tunable
simulations geometries
[139] Dual-band double t-shaped Least Results obtained from high- -- --
monopole antenna absolute frequency structure simulator
shrinkage
and
selection
operator,
ANN and
KNN
[140] Slotted microstrip antennas Neural Simulated, predicted, and -- Resonance
network measured results frequencies, gain
model and radiation
efficiencies
[141] Microstrip antennas Knowledge- Measured, and simulated -- Resonance
based neural frequencies, gain
networks and radiation
model efficiencies
[142] Microstrip antenna ANN Simulated values -- Slot-size, air-
gap/ resonance
frequencies, gain
and radiation
efficiencies
[143] Rectangular microstrip antenna Support ANN model -- Slots position, slots-
vector size
regression
[144] Slotted microstrip antennas SVM Simulated and computed values -- Resonant
frequency, gain,
directivity, and
radiation efficiency
[145] Band-notched monopole antenna Decision ANN -- Antenna’s
tree, random dimensions
forest, XGB
regression,
KNN, and
ANN
[146] UWB antenna ANN Simulated and experimental -- Impedance
results bandwidth and
multi-band notch
frequencies
[147] Slotted UWB antenna Multi- HFSS GA and PSO Bandwidth and
adaptive notch frequencies
neuro-fuzzy
inference
system
[148] UWB antenna Multi- HFSS results with the real -- Return loss,
objective prototype antenna bandwidth and
genetic central frequency
algorithm deviation
[149] UWB microstrip antenna -- Simulated result Adaptive evolutionary Return loss,
algorithm bandwidth and
central frequency
deviation
[150] Aperiodic linear arrays -- -- Nature-based design technique Optimal element
(includes robust GA optimizer positions
and rapid neural-network
estimation procedures)
[151] Antenna array Neural -- Modified invasive weed --
networks optimization (modified adaptive
dispersion)

136
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

A comprehensive survey on various antenna designs by

using advanced methods of antenna design through machine Overall, while machine learning algorithms have the
learning, deep learning, and ANN listed in table 1 provide potential to optimize antennas, their limitations mean that
better results than conventional methods. After significant applied algorithm may not always be the best approach for
work has been done in the field of antenna optimization using specific antenna. Other optimization techniques may be more
machine learning algorithms, researchers can explore several appropriate in some cases. Here are some ways to improve the
advanced and forward-looking directions to push the process of microstrip antenna optimization using machine
boundaries of knowledge and address emerging challenges. learning algorithms:
Explore the integration of machine learning with other
optimization techniques, such as genetic algorithms, 1. Data Augmentation: To address the issue of
simulated annealing, or reinforcement learning. Hybrid limited data, we can use data augmentation
approaches may provide synergies that outperform individual techniques to increase the size of the training
optimization methods. dataset. This can be done by generating synthetic
data from existing data using techniques such as
rotation, translation, and scaling.
6 Constraints and Overcomes 2. Feature Engineering: Feature engineering
involves selecting or designing features that are
Table 2 listed various parameters of recently proposed relevant to the problem at hand. In the case of
antenna models using different ML algorithm. However, antenna optimization, this might involve selecting
machine learning algorithms have been used to optimize or designing features that capture key antenna
antennas, there are some constraints to this approach. Here are parameters such as frequency, radiation pattern, and
some potential constraints: polarization. Effective feature engineering can help
improve the accuracy of the machine learning
1. Limited Data: The performance of machine algorithm.
learning algorithms depends heavily on the amount 3. Ensemble Methods: Ensemble methods involve
and quality of data used for training. However, for combining multiple machine learning models to
antenna optimization, it may be challenging to improve their performance. For example, we could
collect and label enough data to train the algorithm train multiple neural networks with different
effectively. initializations and combine their outputs to obtain a
2. Complexity: Antenna optimization using machine more robust and accurate solution.
learning algorithms can be a complex process that 4. Transfer Learning: Transfer learning involves
requires expertise in both antenna design and leveraging knowledge from one domain to another.
machine learning. The algorithm must be carefully In the case of antenna optimization, this might
designed to account for the various parameters that involve using a machine learning algorithm that has
impact antenna performance, such as resonance been trained on a related problem, such as image
frequency, polarization, and radiation pattern. classification, and fine-tuning it for antenna
3. Computational Cost: Antenna optimization using optimization.
machine learning algorithms can be 5. Explainability: To improve the interpretability of
computationally expensive, especially for large and the machine learning algorithm, we can use
complex antenna designs. This may limit the techniques such as feature importance analysis and
practical application of the approach in some visualization. These techniques can help us
situations. understand which features are most important for
4. Interpretability: Machine learning algorithms are antenna optimization and how the algorithm is
often criticized for their lack of interpretability. It making its decisions.
can be challenging to understand why a particular
antenna design is optimal, which can limit the Generally, by addressing these issues and incorporating
ability to make design decisions based on the results these techniques, we can improve the process of microstrip
of the algorithm. antenna optimization using machine learning algorithms.

Table 2. List of recently optimized antenna model using different ML algorithm.

References Antenna Types Frequency (GHz) Dataset ML Algorithm Error
[50] Two-Port MIMO 2.43–2.50 GHz 125 GPR RMSE less than
Antenna 0.0001%
[136] Square Patch 0.48–7.84 3822 GPR RMSE are 0.0087%
Antenna
[143] Rectangular 2.15 198 SVR Average Error 0.40%
Microstrip
Antenna
[144] Slotted Antenna 1.75 and 2.50 GHz 198 SVM Average Error 0.79%
[145] Band-Notched 2.9 and 21.6 GHz 766,413 KNN Mean Square Error
Monopole Antenna 0.290%
[147] Slotted RMSA 2–3.17/9.1–14/0– 1195 MANFIS-PSO Error is >1%
4.84/0–4.39/0–3.51
[166] U-shaped Slotted 2-10.75 956 ANN Average Relative
Antenna Error 1.59%
137
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

[167] Square Patch MSA 5.9206–25-7625 125 GPR Average Error is

0.400281%

7 Machine Learning for Antenna Applications These wearable antennas are fabricated with textile materials
Machine learning approaches have been depicting great and essential for emerging wireless textiles communication
outcomes in many antenna applications like 5G, biomedical, systems. In textiles systems, communication has been with
textile, Global Positioning System (GPS), satellite the help of wearable antennas with external devices. Several
communication system and smart homes and cities. Its papers [159-162] are focused on sensors and techniques using
exceptional abilities to process on real problem statements machine learning that have being more useful to design the
and get hands-on experience with accurate data make it more flexible textile system.
appropriate for applications. By applying machine learning,
the antenna is used for many civilian uses and further wireless 7.5 Machine learning for global positioning system
applications. Additional, body-centric communication Antennas are also important in automotive applications, such
systems also utilize machine learning to improve their as GPS, Bluetooth, and cellular communication. Machine
capabilities. learning can be used to optimize the design of these antennas
to improve their performance in challenging environments
7.1 Machine learning for millimeter wave such as urban canyons and tunnels. The GPS is mostly used
Machine learning can be used to optimize the design of to obtain the accurate geographical position in civil and
antennas for wireless communication systems such as 5G, military applications. The microstrip antennas are mostly
IoT. In electromagnetic spectrum the frequency range of used in GPS due to compact size and good radiation
millimeter wave covers 30 to 300 GHz (10 to 1 mm properties. Due to the requirement of circular polarization
wavelength range). The millimeter wave frequency range are property in receiver antennas of GPS, microstrip antennas is
widely used for the data transmission and sensing systems. It needed. The purpose of machine learning in GPS application
provides good flexibility and capability, due to large range of has been illustrated in [163-165] with detailed explanation.
unlicensed bandwidth. Different field of applications uses
millimeter wave technology that will come under the 7.6 Smart homes and cities
millimeter wave range. Nowadays, wireless applications are Antennas are increasingly being used in smart home
utilizing millimeter wave antennas designed by using applications such as home automation, security systems, and
machine learning algorithms to achieve large flexibility. entertainment systems. Machine learning can be used to
Literatures [153,154] reported optimization of antennas using optimize the design of these antennas to improve their range,
Machine learning tools for 5G application. data rate, and energy efficiency. Antennas are also important
components in smart city applications, such as traffic
7.2 Machine learning for body-centric communication management, air quality monitoring, and public safety.
Antennas are increasingly used in medical applications such Machine learning can be used to optimize the design of these
as imaging and wireless sensing. Machine learning can be antennas to improve their performance in challenging
used to optimize the design of these antennas to improve their environments, such as urban canyons and high-rise buildings.
performance, such as reducing the power required for wireless In all these applications, machine learning can help
sensing. Currently, demand of body-centric communication engineers optimize antenna designs more efficiently and
systems has been increased in different areas and applications effectively than traditional methods. By automating the
such as biomedical, security, identification, sports, smart optimization process using machine learning, engineers can
phones etc. Wearable antennas are the most important save time and resources while also exploring a wider range of
elements for the body-centric communication systems. Some design options to find the best antenna configuration for a
of the research papers [155, 156] proposes wearable antennas given application. While the integration of machine learning
and experimentally analysed using machine learning into antenna applications offers numerous benefits, careful
algorithms. consideration of the associated challenges is crucial. Machine
Learning dependent Broadband Millimeter-Wave SIW
7.3 Machine learning for satellite communication Cavity-Backed slot antenna has been proposed. But some
Antennas are critical components in aerospace applications, issues like requirement of substantial amounts of labelled data
such as radar systems and satellite communication. Machine for training, long duration of development, and the cost.
learning can be used to optimize the design of these antennas Antenna optimization based on an ANN method for detecting
to improve their performance and reduce their size and GPS spoofing signals may be faced problem of high error-
weight. Satellite communication consist of two main susceptibility and a lack of capable resources.
modules: first is the ground portion that contains fixed or
mobile transmission/reception system, and supplementary
equipment, second is the space portion, that is the satellite. A 8 Conclusion
satellite travels around the Earth in orbit and transferring
information from one place to another. Nowadays, machine A thorough discussion on machine learning algorithms in
learning based antennas are designed for satellite antenna design optimization has been reported in this paper.
communication to deliver signals to a station. Various Due to fast processing speed of the antenna design with
literatures [157,158] propose different works on satellite minimum errors and high accuracy, machine learning
applications using machine learning. techniques are mostly studied. A concise study has been given
on various machine learning types and evolutionary
7.4 Machine learning for textile communication systems algorithm. In this survey, a thorough investigation has been
Demand of wearable antennas in biomedical applications is done on various antenna designs by recently developed
increases due to its flexibility, durability, and washability. machine learning optimization techniques for improved
138
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

results. This paper has been discussed general approaches to antenna designs to meet the specific requirements of smart
antenna optimization using evolutionary ML algorithm, applications, such as low power consumption, high data rate,
which are appropriate for antenna parameters with high and and reliable communication.
multiple specifications. These algorithms are proposed for This is an Open Access article distributed under the terms of
different modern antenna design due to their simplification, the Creative Commons Attribution License.
robustness, and optimization ability. Applications of antennas
in various fields like telecommunications, aerospace,
automotive, medical and textile are explored. In all these
applications, machine learning can help engineers optimize

______________________________
References

[1] M. Bailey and M. Deshpande, “Integral equation formulation of means oftrust‐region gradient‐search with sparse jacobian
microstrip antennas,” IEEE Trans. Anten. Propagat., vol. 30, no. updates,” IET Micro., Anten. & amp; Propagat., vol. 13, no. 10,
4, pp. 651-656, Aug. 1982, doi:10.1109/APS.1997.631535. pp. 1646–1652, May 2019, doi:10.1049/iet-map.2018.5879.
[2] P. A. Tirkas and C. A. Balanis, “Finite-difference time-domain [16] S. Koziel and S. Ogurtsov, “Multi-objective design of antennas
method for antenna radiation,” IEEE Trans. Anten. Propagat., vol. using variable-fidelity simulations and surrogate models IEEE
40, no. 3, pp. 334–340, Mar. 1992, doi: 10.1109/8.135478 Trans. Anten. Propagt., vol. 61, no. 12, pp. 5931–5939, Dec.2013,
[3] J. G. Maloney, G. S. Smith and W. R. Scott, “Accurate doi: 10.1109/TAP.2013.2283599.
computation of the radiation from simple antennas using the finite- [17] S. Koziel and A. Bekasiewicz, “Comprehensive comparison of
difference time-domain method,” IEEE Trans. Anten. Propagat., compact UWB antenna performance by means of multiobjective
vol. 38, no. 7, pp. 1059-1068, July 1990, doi: 10.1109/8.55618. optimization,” IEEE Trans. Anten. Propagt., vol. 65, no. 7, pp.
[4] J. L. Volakis, A. Chatterjee, L. C. Kempel, and A. 3427–3436, Jul. 2017, doi: 10.1109/TAP.2017.2700044.
Chatterjee, Finite element method for electromagnetics: antennas, [18] X. Y. Zhang, Y. B. Tian and X. Zheng, “Antenna optimization
microwave circuits, and scattering applications. in IEEE/OUP design based on deep gaussian process model,” Inter. J. Anten. and
series on electromagnetic wave theory. New York, NY: IEEE Propag., vol. 2020, pp. 1–10, Nov. 2020, doi:
Press [u.a.], 1998. 10.1155/2020/2154928.
[5] Z. Lou and J. Μ. Jin, “Modeling and simulation of broad-band [19] D. R. Prado, J. A. López-Fernández, M. Arrebola, M. R. Pino and
antennas using the time-domain finite element method,” IEEE G. Gou, “Wideband shaped-beam reflectarray design using
Trans. Anten. Propagat., vol. 53, no. 12, pp. 4099–4110, Dec. support vector regression analysis,” IEEE Anten. and Wirel.
2005, doi: 10.1109/TAP.2005.859905 Propag. Lett., vol. 18, no. 11, pp. 2287–2291, Nov. 2019, doi:
[6] T. K. Sarkar, R. D. Antonije and M. K. Branko, “Method of 10.1109/LAWP.2019.2932902.
moments applied to antennas in Handbook of Anten. in Wirel. [20] S. Koziel and A. Pietrenko-Dabrowska, “Rapid multi-objective
Commun., pp. 239-279, Oct. 2018. optimization of antennas using nested Kriging surrogates and
[7] W. Rawle, “The method of moments: A numerical technique for single-fidelity EM simulation models,” Engg. Computat., vol. 37,
wire antenna design,” High Freq. Electron., vol. 5, pp. 42-47, Jan. no. 4, pp. 1491–1512, Apr.2020, doi: 10.1108/EC-05-2019-0200.
2006. [21] M. O. Akinsolu, B. Liu. V. Grout, P. I. Lazaridis, M. E. Mognaschi
[8] A. Reineix and B. Jecko, “Analysis of microstrip patch antennas and P. D. Barba, “A parallel surrogate model assisted evolutionary
using finite difference time domain method,” IEEE Trans. algorithm for electromagnetic design optimization,” IEEE Trans.
Antennas Propagt., vol. 37, no. 11, pp. 1361-1369, Nov.1989, doi: on Emerg. Topics in Computat. Intell., vol. 3, no. 2, pp. 93–105,
10.1109/8.43555. Apr. 2019, doi: 10.1109/TETCI.2018.2864747.
[9] E. H. Newman and P. Tulyathan, “Analysis of microstrip antennas [22] C. Maeurer, P. Futter and G. Gampala, “Antenna design
using moment methods,” IEEE Trans. Anten. Propagt., vol. 29, no. exploration and optimization using machine learning,” 14th
1, pp. 47-53, Jan. 1981, doi: 10.1109/TAP.1981.1142532. European Conf. on Anten. and Propag. (EuCAP), Mar. 2020, doi:
[10] P. Testolina, M. Lecci, M. Rebato, A. Testolin, J. Gambini, R. 10.23919/EuCAP48036.2020.9135530.
Flamini, C. Mazzucco and M. Zorzi, “Enabling simulation-based [23] A. K. S.O. Hassan, A. S. Etman and E. A. Soliman, “Optimization
optimization through machine learning: A case study,” Antenna of a novel nano antenna with two radiation modes using Kriging
Des., vol. 1908, p. 11225, Dec. 2019, surrogate models,” IEEE Photo. Jour., vol. 10, no. 4, pp. 1–17,
doi:10.1109/GLOBECOM38437.2019.9013240. Aug. 2018, doi: 10.1109/JPHOT.2018.2848593.
[11] S. Ledesma, J. Ruiz-Pinales, G. C. Villafaña and M. G. G. [24] B. Chen, W. Zeng, Y. Lin, and D. Zhang, “A new local search-
Hernandez, “A hybrid method to design wire antennas: Design and based multiobjective optimization algorithm,” IEEE Tran. on
optimization of antennas using artificial intelligence,” IEEE Anten. Evolut. Comput., vol. 19, no. 1, pp. 50–73, Feb. 2015, doi:
Propagt. Mag., vol. 57, no. 4, pp. 23-31, Aug. 2015, 10.1109/TEVC.2014.2301794.
DOI: 10.1109/map.2015.2453912. [25] C. Zhang, X. Fu, S. Peng and Y. Wang, “New multi‐objective
[12] H. M. E. Misilmani and T. Naous, “Machine learning in antenna optimisation algorithm for uniformly excited aperiodic array
design: An overview on machine learning concept and algorithms” synthesis,” IET Micro., Anten. & amp; Propag., vol. 13, no. 2, pp.
International Conference on High Performance Computing & 171–177, Dec. 2018, doi: 10.1049/iet-map.2018.5173.
Simulation; Dublin, Ireland, Jul. 2019, doi: [26] B. Liu, S. Koziel and N. Ali, “Sadea-II: A generalized method for
10.1109/HPCS48598.2019.9188224. efficient global optimization of Antenna Design,” J. Compu. Desi.
[13] Q. J. Zhang, K. C. Gupta and V. Devabhaktuni, “Artificial neural and Engg., vol. 4, no. 2, pp. 86–97, April 2017, doi:
networks for RF and microwave design-from theory to practice,” 10.1016/j.jcde.2016.11.002.
IEEE Trans. Micro. Theo., vol. 51, pp. 1339-1350, May 2003, [27] V. V. Thakare and P. Singhal, “Microstrip antenna design using
doi:10.1109/TMTT.2003.809179. artificial neural networks,” Inter. J. RF and Micro. Computer-
[14] S. Koziel, S. Ogurtsov, I. Couckuyt and T. Dhaene,“Variable- Aided Engg., vol. 20, no. 1, pp. 76–86, Nov. 2009, doi:
fidelity electromagnetic simulations and Co-kriging for accurate 10.1002/mmce.20414.
modeling of antennas,” IEEE Trans. Anten. Propagt., vol. 61, no. [28] C. Gianfagna, M. Swaminathan, P. M. Raj, R. Tummala, and G.
3, pp. 1301-1308, Mar. 2013, doi:10.1109/TAP.2012.2231924. Antonini, “Enabling antenna design with nano-magnetic materials
[15] S. Koziel and A. Pietrenko-Dabrowska, “Reduced‐cost using machine learning,” in 2015 IEEE Nanotechn. Mater. Devic.
electromagnetic‐driven optimisation of antenna structures by
139
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

Conf. (NMDC), Anchorage, AK, USA: IEEE, Sep. 2015, pp. 1– [46] Saifullah and B. Ahmed, “Machine learning for isotropic antenna
5. doi: 10.1109/NMDC.2015.7439256. design,” in 2018, 22nd Int. Microw. Radar Conf. (MIKON), pp.
[29] G. Angiulli and M. Versaci, "Resonant frequency evaluation of 683-687, May 2018, doi: 10.23919/MIKON.2018.8405325.
microstrip antennas using a neural-fuzzy approach," IEEE Trans. [47] M. Chetioui, A. Boudkhil, N. Benabdallah, and N. Benahmed,
on Magn., vol. 39, no. 3, pp. 1333-1336, May 2003, doi: “Design and optimization of SIW patch antenna for Ku band
10.1109/TMAG.2003.810172. applications using ANN algorithms,” in 2018 4th Inter. Conf.
[30] L. Y. Xiao, W. Shao, F. L. Jin and B. Z. Wang, “Multiparameter Optim. and Applic. (ICOA), Apr. 2018, pp. 1-4. doi:
modeling with ANN for antenna design,” IEEE Trans. Anten. 10.1109/ICOA.2018.8370530.
Propagt., vol. 66, no. 7, pp. 3718–3723, Jul. 2018, doi: [48] K. A. Kumar, R. Ashwath, D. Kumar, and R. Malmathanraj,
10.1109/TAP.2018.2823775. “Optimization of multislotted rectangular microstrip patch antenna
[31] K. Sharma and G. P. Pandey, “Designing a compact microstrip using ANN and bacterial foraging optimization,” in 2010, Asia-
antenna using the machine learning approach,” J. Telecomm. Infor. Pacific Intern. Symposium Electro. Compa., pp. 449-452, Jan.
Tech., vol. 4, no. 2020, pp. 44–52, Dec. 2020, doi: 2010, doi: 10.1109/APEMC.2010.5475810.
10.1016/j.aeue.2021.153739. [49] N. Kaur, J. S. Sivia and Rajni, “Design of frequency
[32] S. K. Jain , A. Patnaik and S. N. Sinha, “Design of custom-made reconfigurable planar antenna using artificial neural network,” Int.
stacked patch antennas: A machine learning approach,” Inter. J. J. Micr. and Wire. Techn., vol. 14, no. 9, pp. 1107–1118, Oct.
Mach. Learn. Cybernet., vol. 4, no. 3, pp. 189–194, Mar. 2012, 2021, doi:10.1017/S1759078721001434.
doi: 10.1007/s13042-012-0084-x. [50] K. Sharma and G. Pr. Pandey, “Efficient isolation modelling for
[33] B. Liu, H. Aliakbarian, Z. Ma, G. A. E. Vandenbosch, G. Gielen two-port mimo antenna by gaussian process regression,” PIER C,
and P. Excell, “An efficient method for antenna design vol. 108, pp. 227–236, Jan. 2021, doi: 10.2528/PIERC20120301.
optimization based on evolutionary computation and Machine [51] M. M. Khan, S. Hossain, P. Mozumdar, S. Akter and R. H.
Learning Techniques,” IEEE Trans. Anten. Propagt., vol. 62, no. Ashique, “A review on machine learning and deep learning for
1, pp. 7–18, Jan. 2014, doi: 10.1109/TAP.2013.2283605. various antenna design applications,” Heliyon, vol. 8, no. 4, Art.
[34] L. Cui, Y. Zhang, R. Zhang and Q. H. Liu, “A modified efficient no. e09317, Apr. 2022, doi: 10.1016/j.heliyon.2022.e09317.
KNN method for antenna optimization and Design,” IEEE Trans. [52] R. Chellappa, S. Theodoridis and A. V. Schaik, “Advances in
Anten. Propagt., vol. 68, no. 10, pp. 6858–6866, Oct. 2020, doi: machine learning and deep neural networks,” Proceed. of the
10.1109/TAP.2020.3001743. IEEE, vol. 109, no. 5, pp. 607–611, May 2021, doi:
[35] P. Robustillo, J. Zapata, J. A. Encinar and J. Rubio, “ANN 10.1109/JPROC.2021.3072172.
characterization of multi-layer reflectarray elements for [53] J. T. Wassan, H. Wang, F. Browne and H. Zheng, “A
contoured-beam space antennas in the ku-band,” IEEE Trans. comprehensive study on predicting functional role of
Anten. Propagt., vol. 60, no. 7, pp. 3205–3214, Jul. 2012, doi: metagenomes using machine learning methods,” IEEE/ACM
10.1109/TAP.2012.2196941. Trans. on Comput. Bio. Bioinfo., vol. 16, no. 3, pp. 751–763, May
[36] D. K. Neog, S. S. Pattnaik, D. C. Panda, S. N. Devi, B. Khuntia 2019, doi: 10.1109/TCBB.2018.2858808.
and M. Dutta , “Design of a wideband microstrip antenna and the [54] S. Sigurdsson, J. S. Larsen and L. K. Hansen, “On comparison of
use of artificial neural networks in parameter calculation,” IEEE adaptive regularization methods,” Neur. Netw. Sig. Proc. X.
Trans. Magn., vol. 47, no. 3, pp. 60–65, Jun. 2005, doi: Proceed. of the 2000 IEEE Signal Process. Soc. Worksh., Nov.
10.1109/MAP.2005.1532541. 2002, doi: 10.1109/NNSP.2000.889413.
[37] M. O. Akinsolu, K. K. Mistry, B. Liu, P. I. Lazaridis and Peter [55] R. Sathya and Annamma Abraham, “Comparison of supervised
Excell, “Machine learning-assisted antenna design optimization: a and unsupervised learning algorithms for pattern
review and the state-of-the-art,” 14th European Conf. Anten. classification,” Int. J. Adva. Res. in Arti. Intel., vol. 2, no. 2, Jan.
Propag. (EuCAP), Mar. 2020, doi: 2013, doi: 10.14569/IJARAI.2013.020206.
10.23919/EuCAP48036.2020.9135936. [56] J. Ch. Ang, An. Mirzal , H. Haron and H. N. A. Hamed,
[38] Q. Wu, Y. Cao, H. Wang and W. Hong, “Machine-learning- “Supervised, unsupervised, and semi-supervised feature selection:
assisted optimization and its application to antenna designs: a review on gene selection,” IEEE/ACM Tran. Compu. Bio.
Opportunities and challenges,” China Comm., vol. 17, no. 4, pp. Bioin., vol. 13, no. 5, pp. 971–989, Sep. 2016, doi:
152–164, Apr. 2020, doi: 10.23919/JCC.2020.04.014. 10.1109/TCBB.2015.2478454.
[39] M. Singhal and G. Saini, “Optimization of antenna parameters [57] Q. Wu, H. Wang and W. Hong, “Multistage collaborative
using artificial neural network: A Review,” Int. J. Comp. Tren. and machine learning and its application to antenna modeling and
Tech., vol. 44, no. 2, pp. 64–73, Feb. 2017, doi: optimization,” IEEE Trans. Anten. Propagt., vol. 68, no. 5, pp.
10.14445/22312803/IJCTT-V44P112. 3397–3409, May 2020, doi: 10.1109/TAP.2019.2963570.
[40] B. Murugeshwari and S. Raghavan, “An investigation on special [58] R. Choudhary and H. K. Gianey, “Comprehensive review on
antennae using artificial neural networks,” Int. J. Engg. Res. & supervised machine learning algorithms,” Int. Conf. Mach. Learn.
Tech. (Ijert) Iconn., Volume 5 – Issue 13, 2017, doi: Data Sci. (MLDS), Dec. 2017, doi: 10.1109/MLDS.2017.11.
10.17577/IJERTCONV5IS13063. [59] N. Aboueata, S. Alrasb, A. Erbad, A. Kassler and D. Bhamare,
[41] M. Mushaib and An. Kumar, “Designing of microstrip patch “Supervised machine learning techniques for efficient network
antenna using artificial neural network: a review,” J. Eng. Sci., vol. intrusion detection,” in 2019 28th Int. Conf. Com. Commu. and
11, Issue 7, 2020. Netw. (ICCCN), Jul. 2019, pp.1-8. doi:
[42] H. M. E. Misilmani and T. Naous, “Machine learning in antenna 10.1109/ICCCN.2019.8847179.
design: An overview on machine learning concept and [60] A. D Patel and J. H. Shah, “Performance analysis of supervised
algorithms,” in 2019 Int. Conf. High Perform. Computin. & machine learning algorithms to recognize human activity in
Simulat. (HPCS), Jul.2019, pp. 600-607.doi: ambient assisted living environment,” in 2019 IEEE 16th India
10.1109/HPCS48598.2019.9188224. Council Inter. Conf. (INDICON), Dec. 2019, pp. 1-4. doi:
[43] A. Hoorfar, “Evolutionary programming in electromagnetic 10.1109/INDICON47234.2019.9030353.
optimization: a review,” IEEE Trans. Anten. Propagt., vol. 55, no. [61] K. Virts, A. Shirey, G. Priftis, K. Ankur, M. Ramasubramanian, H.
3, pp. 523–537, Mar. 2007, doi: 10.1109/TAP.2007.891306. Muhammad, A. Acharya and R. Ramachandran, “A quantitative
[44] U. Ozkaya and L. Seyfi, “Dimension optimization of microstrip analysis on the use of supervised machine learning in earth
patch antenna in X/Ku band Via Artificial Neural Network,” Proc. science,” IEEE Inter. Geo. Rem. Ses. Sympo., Sep. 2020, doi:
- Social a Behav. Sci., vol. 195, pp. 2520–2526, Jul. 2015, doi: 10.1109/IGARSS39084.2020.9323770.
10.1016/j.sbspro.2015.06.434. [62] L. Yifei, W. He, P. Shuai, S. Weiqiong, D. Ning and W. Fang,
[45] L. Y. Xiao, W. Shao, F. L. Jin and B. Zhong, “Multiparameter “Application of supervised machine learning algorithms in
modeling with ann for antenna design,” IEEE Trans. Anten. diagnosis of abnormal voltage,” in 2016 China Inter. Conf. on
Propagt., vol. 66, no. 7, pp. 3718–3723, Jul. 2018, doi: Elec. Distr. (CICED), Aug. 2016, pp. 1-5. doi:
10.1109/TAP.2018.2823775. 10.1109/CICED.2016.7576347.
140
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

[63] J. Nayak, B. Naik and H. S. Behera, “A comprehensive survey on [82] S. Ray, “A quick review of machine learning algorithms,” in 2019
support vector machine in data mining tasks: applications & Inter. Conf. on Machine Learning, Big Data, Cloud and Par. Com.
challenges,” Inter. Jou. of Data. The. and App., vol. 8, no. 1, pp. (COMITCon), Feb. 2019, pp. 35-39. doi:
169–186, Feb. 2015, doi: 10.14257/ijdta.2015.8.1.18. 10.1109/COMITCon.2019.8862451.
[64] S. Weisberg, Applied Linear Regression. John Wiley & Sons, [83] S. Angra and S. Ahuja, “Machine learning and its applications: A
2005. doi: 10.1002/0471704091. review,” in 2017 Inter. Conf. on Big Data Ana. and Compu. Intelli.
[65] D. C. Montgomery, E. A. Peck, and G. G. Vining, Introduction to (ICBDAC), Mar. 2017, pp. 57-60. doi:
Linear Regression Analysis. Wiley: A John Wiley & Sons, Inc., 10.1109/ICBDACI.2017.8070809.
Hoboken, NJ, 2012. [84] R. B. Shrestha, M. Razavi, and P. W. C. Prasad, “An Unsupervised
[66] G. A. Seber and A. J. Lee, Linear regression analysis. John Wiley Machine Learning Technique for Recommendation Systems,” 5th
& Sons, 2012. Inter. Confer. on Inno. Techn. in Intel. Syst. and Indus. Applica.
[67] M. Huang, “Theory and Implementation of linear regression,” in (CITISIA), Nov. 2020.
2020 Intern. Conf. on Comp. Vi., Image and Deep Lear. (CVIDL), [85] P. Vikhar, “Evolutionary algorithms: A critical review and its
Jul. 2020, pp. 210-217. doi: 10.1109/CVIDL51233.2020.00-99. future prospects,” Intern. Con. on Global Trends in Signal
[68] H. L. Lim, “A linear regression approach to modeling software Processing, Infor. Com. and Commun. (ICGTSPICC), Dec. 2016.
characteristics for classifying similar software,” in 2019 IEEE [86] R. R. Halde, “Application of machine learning algorithms for
43rd An. Com. Softw. and Applic. Confe. (COMPSAC), Jul. 2019, betterment in education system,” in 2016 Inter. Confe. on Autom.
pp. 942-943. doi: 10.1109/COMPSAC.2019.00152. Con. and Dyn. Optimiz. Tech. (ICACDOT), Sep. 2016, pp. 1110-
[69] H. Drucker, D. Wu and Vl. Vapnik, “Support vector machines for 1114. doi: 10.1109/ICACDOT.2016.7877759.
spam categorization,” IEEE Trans. on Neur. Netw., vol. 10, no. 5, [87] B. Alić, L. Gurbeta, and A. Badnjević, “Machine learning
pp. 1048–1054, Jan. 1999, doi: 10.1109/72.788645. techniques for classification of diabetes and cardiovascular
[70] Alex J. Smola and Bernhard Schölkopf, “A tutorial on support diseases,” in 2017 6th Mediter. Conf. on Embed. Com. (MECO),
vector regression,” Stati. and Comp., vol.14, pp. 199–222, Aug. Jun. 2017, pp. 1-4. doi: 10.1109/MECO.2017.7977152.
2004, doi: 10.1023/B:STCO.0000035301.49549.88. [88] M. Ferdous, J. Debnath, and N. R. Chakraborty, “Machine
[71] M. Awad and R. Khanna, “Support vector regression”, Eff. Lear. Learning Algorithms in Healthcare: A Literature Survey,” in
Mac., pp. 67–80, 2015, doi: 10.1007/978-1-4302-5990-9_4. 2020 11th Intern. Conf. on Com., Comm. and Netwo. Tech.
[72] T. Zoppi, A. Ceccarelli and A. Bondavalli, “Unsupervised (ICCCNT), Jul. 2020, pp. 1-6. doi:
algorithms to detect zero-day attacks: strategy and 10.1109/ICCCNT49239.2020.9225642.
application,” IEEE Access, vol. 9, pp. 90603–90615, Jan. 2021, [89] O. Obulesu, M. I. Mahendra, and M. ThrilokReddy, “Machine
doi: 10.1109/ACCESS.2021.3090957. Learning Techniques and Tools: a survey,” in 2018 Intern. Conf.
[73] M. Usama, J. Qadir, A. Raza, H. Arif, K. A. Yau, Y. Elkhatib, A. on Inven. Rese. in Comp. Applic. (ICIRCA), Jul. 2018, pp. 605-
Hussain and A. Al-F, “Unsupervised machine learning for 611. doi: 10.1109/ICIRCA.2018.8597302.
networking: techniques, applications and research [90] R. E. Uhrig, “Introduction to artificial neural networks,” in 2002
challenges,” IEEE Access, vol. 7, pp. 65579–65615, Jan. 2019, Proceedings of IECON ’95 - 21st Ann. Confe. on IEEE Indus.
doi: 10.1109/ACCESS.2019.2916648. Elec., Nov. 2002, pp. 33-37. doi: 10.1109/IECON.1995.483329.
[74] N. Amruthnath and T. Gupta, “A research study on unsupervised [91] W. B. Hudson, “Introduction and overview of artificial neural
machine learning algorithms for early fault detection in predictive networks in instrumentation and measurement applications,” in
maintenance,” in 2018 5th Intern. Conf. on Indus. Engg. and Appl. 2002 IEEE Instrum. and Measu. Techn. Conf., Dec. 2002, pp. 623-
(ICIEA), Apr. 2018, pp. 355-361. doi: 626. doi: 10.1109/IMTC.1993.382568.
10.1109/IEA.2018.8387124. [92] M. Al-Qatf, Y. Lasheng, M. Al-Habib and K. Al-Sabahi, “Deep
[75] L. S. B. Pereira, R. N. Rodrigues and E. A. C. A. Neto, learning approach combining sparse autoencoder with SVM for
“Unsupervised machine learning in industrial applications: a case network intrusion detection,” IEEE Access, vol. 6, pp. 52843–
study in iron mining,” in 2020 IEEE Bombay Sec. Sign. Confe. 52856, Jan. 2018, doi: 10.1109/ACCESS.2018.2869577.
(IBSSC), Dec. 2020, pp. 204-207. doi: [93] D. Gu and H. Hu, “Spatial Gaussian process regression with
10.1109/IBSSC51096.2020.9332174. mobile sensor networks,” IEEE Trans. on Neural Netw. and Lea.
[76] M. Verkerken, L. D’hooge, T. Wauters, Br. Volckaert and F. De Syst., vol. 23, no. 8, pp. 1279–1290, Aug. 2012, doi:
Turck, “Unsupervised machine learning techniques for network 10.1109/TNNLS.2012.2200694.
intrusion detection on modern data,” in 2020 4th Cyber Sec. in [94] Y. Bazi and F. Melgani, “Gaussian Process approach to remote
Netw. Conf. (CSNet), Oct. 2020, pp. 1-8. doi: sensing image Classification,” IEEE Tran. on Geosci. and Rem.
10.1109/CSNet50428.2020.9265461. Sens., vol. 48, no. 1, pp. 186–197, Jan. 2010, doi:
[77] P. Mehta, H. Shah, V. Kori, V. Vikani, S. Shukla, M. Shenoy, 10.1109/TGRS.2009.2023983.
“Survey of unsupervised machine learning algorithms on precision [95] D. Kinaneva, G. Hristov, P. Kyuchukov, G. Georgiev, P. Zahariev,
agricultural data,” in 2015 Int. Conf. Innov. Infor., Emb. Comm. and R. Daskalov, “Machine learning algorithms for regression
Sys. (ICIIECS), Mar. 2015, pp. 1-8. doi: analysis and predictions of numerical data,” in 2021 3rd Inter.
10.1109/ICIIECS.2015.7193070. Con. on Human-Computer Interaction, Optimi. and Rob. Appl.
[78] K. Nguyen, Cl. Fookes and S. Sridharan, “Improving deep (HORA), Jun. 2021, pp. 1-6. doi:
convolutional neural networks with unsupervised feature 10.1109/HORA52670.2021.9461298.
learning,” in 2015 IEEE Inter. Conf. on Image Processing (ICIP), [96] A. Kyrillidis and V. Cevher, “Combinatorial selection and least
Sep. 2015, pp. 2270-2274. doi: 10.1109/ICIP.2015.7351206. absolute shrinkage via the Clash algorithm,” in 2012 IEEE Inter.
[79] R. B. Shrestha, M. Razavi and P. W. C Prasad, “An unsupervised Symp. on Infor. Theory Proc., Jul. 2012, pp. 2216-2220. doi:
machine learning technique for recommendation systems,” in 10.1109/ISIT.2012.6283847.
2020 5th Inter. Conf. on Inno. Techn. in Inte. Syst. and Indus. [97] J. Staum, “Better simulation metamodeling: The why, what, and
Applic. (CITISIA), Nov. 2020, pp. 1-9. doi: how of stochastic kriging,” in 2009 Proceedings of the 2009
10.1109/CITISIA50690.2020.9371817. Winter Simu. Confer. (WSC), Dec. 2009, pp. 119-133. doi:
[80] M. A. Samsuden, N. M. Diah and N. A. Rahman, “A review paper 10.1109/WSC.2009.5429320.
on implementing reinforcement learning technique in optimising [98] S. Koziel, S. Ogurtsov, I. Couckuyt, and T. Dhaene, “Efficient
games performance,” in 2019 IEEE 9th Inter. Conf. on System simulation-driven design optimization of antennas using co-
Engg. and Tech. (ICSET), Oct. 2019, pp. 258-263. doi: kriging,” in 2012 IEEE Intern. Symp. on Anten. and Propag., Jul.
10.1109/ICSEngT.2019.8906400. 2012, pp. 1-2. doi: 10.1109/APS.2012.6348754.
[81] W. Qiang and Z. Zhongli, “Reinforcement learning model, [99] S. Ulaganathan, S. Koziel, A. Bekasiewicz, I. Couckuyt, E.
algorithms and its application,” in 2011 Intern. Conf. on Mecha. Laermans, and T. Dhaene, “Cost-efficient modeling of antenna
Scie., Elect. Engg. Com. (MEC), Aug. 2011, pp. 1143-1146. doi: structures using Gradient-Enhanced Kriging,” in 2015 Loughbor.
10.1109/MEC.2011.6025669.
141
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

Ant. & Amp; Pro. Conf. (LAPC), Nov. 2015, pp. 1-5. doi: [117] B. D. Sarkar, S. Shankar, and H. Chaurasiya, “Prediction of length
10.1109/LAPC.2015.7366125. & width of a rectangular patch antenna using ANN,” in 2015
[100] V. Grout, M. O. Akinsolu, B. Liu, P. I. Lazaridis, K. K. Mistry and Annual IEEE India Confe. (INDICON), Dec. 2015, pp. 1-4. doi:
Z. D. Zaharis, “Software Solutions for Antenna Design 10.1109/INDICON.2015.7443343.
Exploration: A comparison of packages, tools, techniques, and [118] I. Vilović, N. Burum, and M. Brailo, “Microstrip antenna design
algorithms for various design challenges,” IEEE Trans. on Magn., using neural networks optimized by PSO,” ICECom 2013, Oct.
vol. 61, no. 3, pp. 48–59, Jun. 2019, doi: 2013, pp. 1-4. doi: 10.1109/ICECom.2013.6684759.
10.1109/MAP.2019.2907887. [119] T. Khan, A. De and M. Uddin, “Prediction of Slot-Size and
[101] S. Koziel and A. T. Sigurðsson, “Multi‐fidelity EM simulations inserted Air-GAP for improving the performance of rectangular
and constrained surrogate modelling for low‐cost multi‐objective microstrip antennas using artificial neural networks,” IEEE Anten.
design optimisation of antennas,” IET Micro. Ante. & Propag., and Wirel. Propag. Let., vol. 12, pp. 1367–1371, Jan. 2013, doi:
vol. 12, no. 13, pp. 2025–2029, Jul. 2018, doi: 10.1049/iet- 10.1109/LAWP.2013.2285381.
map.2018.5184. [120] D. Kr Neog, S. S. Pattnaik, D. C. Panda, S. Devi, B. Khuntia and
[102] A. Alieldin, Y. Huang, S. J. Boyes, M. Stanley, S. D. Joseph, Q. M. Dutta, “Design of a wideband microstrip antenna and the use
Hua and D. Lei, “A Triple-Band Dual-Polarized indoor base of artificial neural networks in parameter calculation,” IEEE
station antenna for 2G, 3G, 4G and sub-6 GHz 5G Anten. and Propag. Mag., vol. 47, no. 3, pp. 60–65, Jun. 2005, doi:
applications,” IEEE Acc., vol. 6, pp. 49209–49216, Jan. 2018, doi: 10.1109/MAP.2005.1532541.
10.1109/ACCESS.2018.2868414. [121] N. T. Tokan and Filiz Güneş, “Support vector characterisation of
[103] S. Koziel and A. Bekasiewicz, “Sequential approximate the microstrip antennas based on measurements,” PIER B, vol. 5,
optimisation for statistical analysis and yield optimisation of pp. 49–61, Jan. 2008, doi: 10.2528/PIERB08013006.
circularly polarised antennas,” IET Micro. Ante. & Propa., vol. 12, [122] B. K. Singh, “Design of rectangular microstrip patch antenna
no. 13, pp. 2060–2064, Jul. 2018, doi: 10.1049/iet-map.2018.5343. based on artificial neural network algorithm,” in 2015 2nd
[104] P. Baldi, “Gradient descent learning algorithm overview: a general Inte.Conf. on Signal Pro. and Inte. Networks (SPIN), Feb. 2015,
pp. 6-9. doi: 10.1109/SPIN.2015.7095291.
dynamical systems perspective,” IEEE Trans. on Neural Netw.,
[123] P. Malathi and R. Kumar, "On the design of multilayer circular
vol. 6, no. 1, pp. 182–195, Jan. 1995, doi: 10.1109/72.363438.
microstrip antenna using artificial neural networks," IJRTE, vol.2,
[105] J. Arif, N. R. Chaudhuri, S. Ray, and B. Chaudhuri, “in 2009
no. 5, Nov. 2009.
Online Levenberg-Marquardt algorithm for neural network based
[124] C. Yildiz, S. Gultekin, K. Guney and S. Sagiroglu, “Neural models
estimation and control of power systems,” Inter. Joint Conf. on
for the resonant frequency of electrically thin and thick circular
Neural Net., Jun. 2009, pp. 199-206. doi:
microstrip antennas and the characteristic parameters of
10.1109/IJCNN.2009.5179071.
asymmetric coplanar waveguides backed with a conductor,” AEU
[106] R. Sujatha, V. Mareeswari, Jyotir Moy Chatterjee, Abd Allah A.
– Inter. Jou. of Elect. and Commu., vol. 56, no. 6, pp. 396–406,
Mousa and Aboul Ella Hassanien, “A Bayesian regularized neural
Jan. 2002, doi: 10.1078/1434-8411-54100128.
network for analyzing bitcoin trends,” IEEE Acce., vol. 9, pp.
[125] I. Vilović and N. Burum, “Design and feed position estimation for
37989–38000, Jan. 2021, doi: 10.1109/ACCESS.2021.3063243.
circular microstrip antenna based on neural network model,” in
[107] J. Kennedy and R. C. Eberhart, “Particle swarm
2012 6th Euro. Confe. on Ante. and Prop. (EUCAP), Mar. 2012,
optimization,” Proceedings of ICNN’95 – Inter. Con. on Neural
pp. 3614-3617. doi: 10.1109/EuCAP.2012.6206281.
Networks, Nov. 2002, pp. 1942-1948. doi:
[126] A. Mishra, G. B. Janvale, B. V. Pawar and P. M. Patil, “The design
10.1109/ICNN.1995.488968.
of circular microstrip patch antenna by using Quasi-Newton
[108] R. M. Storn and K. Price, “Differential evolution: A simple and
algorithm of ANN,” J. Electrom. Anal. Appli., vol. 02, no. 07, pp.
efficient heuristic for global optimization over continuous spaces,”
444–449, Jan. 2010, doi: 10.4236/jemaa.2010.27058.
J. Global Optim., vol. 11, no. 4, pp. 341–359, Dec. 1997, doi:
[127] J. S. Sivia, A. P. S. Pharwaha and T. S. Kamal,
10.1023/A:1008202821328.
“Neurocomputational models for parameter estimation of circular
[109] A. Deb, J. S. Roy and Bhaskar Gupta, “A Differential evolution
microstrip patch antennas,” Procedia Com. Scie., vol. 85, pp. 393–
performance comparison: Comparing how various differential
400, Jan. 2016, doi: 10.1016/j.procs.2016.05.178.
evolution algorithms perform in designing microstrip antennas and
[128] H. J. Delgado and M. H. Thursby, “A novel neural network
arrays,” IEEE Anten. and Propag. Mag., vol. 60, no. 1, pp. 51–61,
combined with FDTD for the synthesis of a printed dipole
Feb. 2018, doi: 10.1109/MAP.2017.2774146.
antenna,” IEEE Trans. Anten. Propagt., vol. 53, no. 7, pp. 2231–
[110] P. I. Lazaridis, E. N. Tziris, Z. D. Zaharis, Tj. D. Xenos, J. P.
2236, Jul. 2005.
Cosmas, Ph. B. Gallion, V. Holmes and I. A. Glover, “Comparison
[129] H. J. Delgado, M. H. Thursby, and F. M. Ham, “A novel neural
of evolutionary algorithms for LPDA antenna
network for the synthesis of antennas and microwave
optimization,” Radio Sci., vol. 51, no. 8, pp. 1377–1384, Aug.
devices,” IEEE Tran. on Neu. Netwo., vol. 16, no. 6, pp. 1590–
2016, doi: 10.1002/2015RS005913.
1600, Nov. 2005, doi: 10.1109/TNN.2005.852973.
[111] N. Hansen, “The CMA evolution strategy: A comparing review,”
[130] P. Arora and B. S. Dhaliwal, “Parameter Estimation of Dual Band
in Studies in fuzz. and soft comp., 2007, pp. 75–102. doi:
Elliptical Fractal Patch Antenna Using ANN,” in 2011 Inte.
10.1007/3-540-32494-1_4.
Confe. on De. and Comm. (ICDeCom), Feb. 2011, pp. 1-4. doi:
[112] J. H. Holland, “Genetic Algorithms and Adaptation,” in Springer
10.1109/ICDECOM.2011.5738516.
eBooks, 1984, pp. 317–333. doi: 10.1007/978-1-4684-8941-5_21.
[131] S. Suganthi and S. Raghavan, “ANN based pattern generation,
[113] Sc. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, “Optimization by
design and simulation of broadband fractal antenna for wireless
simulated annealing,” Science, vol. 220, no. 4598, pp. 671–680,
applications,”in 2016 Inte. Confe. on Emer. Trends in Engg.,
May 1983, doi: 10.1126/science.220.4598.61.
Techn. and Science (ICETETS), Feb. 2016, pp. 1-4. doi:
[114] R. K. Mishra and A. Patnaik, “Neural network-based CAD model
10.1109/ICETETS.2016.7603037.
for the design of square-patch antennas,” IEEE Trans. Anten.
[132] P. H. Da F Silva, E. E. C. De Oliveira, and A. G. D’Assunção,
Propagt., vol. 46, no. 12, pp. 1890–1891, Jan. 1998, doi:
“Using a multilayer perceptrons for accurate modeling of quasi-
10.1109/8.743842.
fractal patch antennas,” in 2010 Inte. Workshop on Anten. Tech.
[115] M. N. Moghaddasi and P. D. Barjoei, “A heuristic artificial neural
(iWAT), Mar. 2010, pp. 1-4. doi: 10.1109/IWAT.2010.5464782.
network design of resonant frequency of rectangular
[133] S. Pattnaik, S. S. Pattnaik and B. S. Dhaliwal, “Modeling of
microstrip/patch antennas,” in 2008 3rd Intern. Conference on
circular fractal antenna using BFO‐PSO–based selective ANN
Infor. and Commu. Techn.: From Theory to Appl., Apr. 2008, pp.
ensemble,” Int. J. Nume. Modelling: Electronic Networks, Devices
1-5. doi: 10.1109/ICTTA.2008.4530138.
and Fields, vol. 32, no. 3, Jan. 2019, doi: 10.1002/jnm.2549.
[116] N. Türker, F. Güneş and T. Yildirim, “Artificial neural design of
[134] Anuradha, A. Patnaik and S. N. Sinha, “Design of Custom-Made
microstrip antennas,” Turkish Jou. of Elec. Engg. and Com.
Fractal Multi-Band antennas using ANN-PSO,” IEEE Anten. and
Sciences, vol. 14, no. 3, pp. 445–453, Mar. 2006, doi:
10.212.2597070/2881.
142
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

Propag. Mag., vol. 53, no. 4, pp. 94–101, Aug. 2011, doi: [151] Z. D. Zaharis, Ch. Skeberis, Th. D. Xenos, P. I. Lazaridis and J.
10.1109/MAP.2011.6097296. Cosmas, “Design of a novel antenna array beamformer using
[135] M. Kaur and J. S. Sivia, “Giuseppe Peano and Cantor set fractals neural networks trained by modified adaptive dispersion invasive
based miniaturized hybrid fractal antenna for biomedical weed optimization based data,” IEEE Trans. on Broa., vol. 59, no.
applications using artificial neural network and firefly 3, pp. 455–460, Sep. 2013, doi: 10.1109/TBC.2013.2244793.
algorithm,” Int.. J. RF Micro. Computer-Aided Engg., vol. 30, no. [152] M. Lecci, P. Testolina, M. Rebato, A. Testolin, and M. Zorzi,
1, Nov. 2019, doi: 10.1002/mmce.22000. “Machine learning-aided design of thinned antenna arrays for
[136] D. Ustun, A. Toktas and A. Akdagli, “Deep neural network–based optimized network level performance,” in 2020 14th Eur. Conf. on
soft computing the resonant frequency of E–shaped patch Anten. and Prop. (EuCAP), Mar. 2020, pp. 1-5. doi:
antennas,” AEU – Int. J. Electro. Comm., vol. 102, pp. 54–61, Apr. 10.23919/EuCAP48036.2020.9135310.
2019, doi: 10.1016/j.aeue.2019.02.011. [153] Q. Wu, H. Wang, and H. Wang, “Broadband millimeter-wave SIW
[137] K. Sharma and G. P. Pandey, “Efficient modelling of compact cavity-backed slot antenna for 5G applications using machine-
microstrip antenna using machine learning,” AEU - Int. J. Electro. learning-assisted optimization method,” in 2019 iWAT, Mar.
and Comm., vol. 135, p. 153739, Jun. 2021, doi: 2019, pp. 9-12. doi: 10.1109/IWAT.2019.8730801.
10.1016/j.aeue.2021.153739. [154] A. Jafarieh, M. Nouri, and H. Behroozi, “Optimized 5G-MMW
[138] J. P. Jacobs, “Efficient resonant frequency modeling for Dual- compact Yagi-Uda antenna based on machine learning
Band microstrip antennas by Gaussian process regression,” IEEE methodology,” in 2021 29th Iranian Confe. on Elect. Engine.
Anten. and Wire. Propa. Let., vol. 14, pp. 337–341, Jan. 2015, doi: (ICEE), May 2021, pp. 751-756. doi:
10.1109/LAWP.2014.2362937. 10.1109/ICEE52715.2021.9544194.
[139] J. P. Jacobs and Johan Pieter De Villiers, “Gaussian-process- [155] P. Hall and Y. Hao, “Antennas and propagation for body centric
regression-based design of ultrawide-band and dual-band CPW- communications,” in 2006 First European Conf. on Ant. and
fed slot antennas,” J. Electro. Wa. and App., vol. 24, no. 13, pp. Propa., Nov. 2006, pp. 1-7. doi: 10.1109/EUCAP.2006.4584864.
1763–1772, Jan. 2010, doi: 10.1163/156939310792486629. [156] R. Bharadwaj, A. Alomainy and S. K. Koul, “Experimental
[140] Y. Sharma, H. H. Zhang and H. Xin, “Machine learning investigation of Body-Centric indoor localization using compact
techniques for optimizing design of double T-Shaped monopole wearable antennas and machine learning algorithms,” IEEE Trans.
antenna,” IEEE Trans. Anten. Propagt., vol. 68, no. 7, pp. 5658– Anten. Propagt., vol. 70, no. 2, pp. 1344–1354, Feb. 2022, doi:
5663, Jul. 2020, doi: 10.1109/TAP.2020.2966051. 10.1109/TAP.2021.3111308.
[141] T. Khan and A. De, “Estimation of different performance [157] D. R. Prado, J. A. López-Fernández, M. Arrebola and G.
parameters of slotted microstrip antennas with air-gap using neural Goussetis, “Support vector regression to accelerate design and
networks,” ISRN Elect. (Online), vol. 2014, pp. 1–6, Mar. 2014, crosspolar optimization of Shaped-Beam reflectarray antennas for
doi: 10.1155/2014/296105. space applications,” IEEE Trans. Anten. Propagt., vol. 67, no. 3,
[142] T. Khan and A. De, “Prediction of slot shape and slot size for pp. 1659–1668, Mar. 2019, doi: 10.1109/TAP.2018.2889029.
improving the performance of microstrip antennas using [158] Q. Liu, J. Yang, C. Zhuang, A. Barnawi and B. A. Alzahrani,
Knowledge-Based Neural Networks,” Intern. Scholarly Res. “Artificial intelligence based mobile tracking and antenna pointing
Notices, vol. 2014, pp. 1–9, Oct. 2014, doi: 10.1155/2014/957469. in Satellite-Terrestrial Network,” IEEE Access, vol. 7, pp.
[143] T. Khan, A. De, and M. Uddin, “Prediction of Slot-Size and 177497–177503, Jan. 2019, doi:
inserted Air-GAP for improving the performance of rectangular 10.1109/ACCESS.2019.2956544.
microstrip antennas using artificial neural networks,” IEEE Anten. [159] D. Kan, S. D. Ridder; D. Spina, I. Couckuyt, F. Grassi, T. Dhaene,
and Wirel. Propag. Lett., vol. 12, pp. 1367–1371, Jan. 2013, doi: H. Rogier and D. Vande, “Machine learning-based hybrid
10.1109/LAWP.2013.2285381. random-fuzzy modeling framework for antenna design,” in 2020
[144] T. Khan and Ch. Roy, “Prediction of slot‐position and slot‐size of 14th Eur. Conf. on Ante. and Prop. (EuCAP), Mar. 2020, pp. 1-5.
a microstrip antenna using support vector regression,” Int. J. RF doi: 10.23919/EuCAP48036.2020.9135513.
and Micro. Comp.-Aid. Engg., vol. 29, no. 3, p. e21623, Dec. 2018, [160] D. Patron, W. Mongan, T. P. Kurzweg, A. Fontecchio, G. Dion, E,
doi: 10.1002/mmce.21623. K. Anday , Philadelphia PA, and K. R. Dandekar, “On the use of
[145] C. Roy, T. Khan, and B. K. Kanaujia, “Performance parameters knitted antennas and inductively coupled RFID tags for wearable
prediction of slotted microstrip antennas with modified ground applications,” IEEE Trans. on Biom. Circ. and Systems, vol. 10,
plane using support vector machine,” Int. J. Micr. and Wire. no. 6, pp. 1047–1057, Dec. 2016, doi:
Techn., vol. 9, no. 5, pp. 1169–1177, Nov. 2016, 10.1109/TBCAS.2016.2518871.
doi:10.1017/S1759078716001264. [161] H. He, X. Chen, A. Mehmood and L. Raivio, “Cloth Face: a battery
[146] P. Ranjan, A. Maurya, H. Gupta, S. Yadav and A. Sharma, “Ultra- less RFID-Based textile platform for handwriting
wideband CPW fed band-notched monopole antenna optimization recognition,” Sensors, vol. 20, no. 17, p. 4878, Aug. 2020, doi:
using machine learning,” PIER M, vol. 108, pp. 27–38, Jan. 2022, 10.3390/s20174878.
doi:10.2528/PIERM21122802. [162] I. Couckuyt, F. Declercq, T. Dhaene, H. Rogier, and L. Knockaert,
[147] D. Sarkar, T, Khan and R. Laskar, “Multi-parametric ANN “Surrogate-based infill optimization applied to electromagnetic
modelling for interference rejection in UWB antennas,” Intern. J. problems,” Inter. J. RF and Micro. Computer-Aided Engg., vol.
Electr., vol. 107, no. 12, pp. 2068–2083, May 2020, doi: 20, no. 5, pp. 492–501, Jul. 2010, doi: 10.1002/mmce.20455.
10.1080/00207217.2020.1756449. [163] M. Orabi, J. Khalife, A. A. Abdallah, Z. M. Kassas, and S. S. Saab,
[148] D. Sarkar, T. Khan, and F. A. Talukdar, “Multi‐adaptive neuro‐ “A machine learning approach for GPS code phase estimation in
fuzzy inference system modelling for prediction of band‐notched multipath environments,” in 2020 IEEE/ION Position, Location
behaviour of slotted‐UWB antennas optimised using evolutionary and Navigation Symposium (PLANS), Apr. 2020, pp. 1224-1229,
algorithms,” IET Mic. Anten. & Prop., vol. 14, no. 12, pp. 1396– doi: 10.1109/PLANS46316.2020.9110155.
1403, Jul. 2020, doi: 10.1049/iet-map.2020.0055. [164] M. R. Manesh, J. Kenney, H. Wen, V. K. Devabhaktuni, and N.
[149] C. R. M. Silva, H. W. De Castro Lins, S. R. Martins, E. L. F. Kaabouch, “Detection of GPS spoofing attacks on unmanned
Barreto, and A. G. D’Assunção, “A multiobjective optimization of aerial systems,” in 2019 16th IEEE Ann.Consu. Commu. & Amp;
a UWB antenna using a self-organizing genetic Netw. Conf. (CCNC), Jan. 2019, pp. 1-6, doi:
algorithm,” MOTL, vol. 54, no. 8, pp. 1824–1828, May 2012, doi: 10.1109/CCNC.2019.8651804.
10.1002/mop.26945. [165] M. I. M. Ghazali, S. Karuppuswami, and M. H. Jamaluddin,
[150] C. DeLuccia and D. H. Werner, “Nature-based design of aperiodic “Machine learning based design optimization of a GPS antenna on
linear arrays with broadband elements using a combination of a flexible substrate,” in 2021 IEEE Asia-Pacific Conf. on App.
rapid neural-network estimation techniques and genetic Elect.s (APACE), Dec. 2021, pp. 1-3, doi:
algorithms,” IEEE Anten. and Propag. Mag., vol. 49, no. 5, pp. 10.1109/APACE53143.2021.9760562.
13–23, Oct. 2007, doi: 10.1109/MAP.2007.4395292. [166] D. Sarkar, T. Khan, and F. A. Talukdar, “Multi-Parametric
synthesis modeling of slotted UWB antennas using artificial neural
143
 Anil Kumar Pandey and M. P. Singh/Journal of Engineering Science and Technology Review 17 (2) (2024) 128 - 144

network,” in 2020 IEEE 7th UPCON, Nov. 2020, pp. 1-4, doi: Learning,” in 2019 IEEE InCAP, Dec. 2019, pp. 1-4, doi:
10.1109/UPCON50219.2020.9376439. 10.1109/InCAP47789.2019.9134523.
[167] K. Sharma and G. P. Pandey, “Prediction of the resonant frequency
of square patch microstrip antenna with DGS using Machine

144