2424 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 52, NO. 9, SEPTEMBER 2004
A Parallel Electromagnetic Genetic-Algorithm
Optimization (EGO) Application for
Patch Antenna Design
Frank J. Villegas, Member, IEEE, Tom Cwik, Fellow, IEEE, Yahya Rahmat-Samii, Fellow, IEEE, and
Majid Manteghi, Member, IEEE
Abstract—In this paper, we describe an electromagnetic genetic
algorithm (GA) optimization (EGO) application developed for the
cluster supercomputing platform. A representative patch antenna
design example for commercial wireless applications is detailed,
which illustrates the versatility and applicability of the method. We
show that EGO allows us to combine the accuracy of full-wave EM
analysis with the robustness of GA optimization and the speed of a
parallel computing algorithm. A representative patch antenna de-
sign case study is presented. We illustrate the use of EGO to design
a dual-band antenna element for wireless communication (1.9 and
2.4 GHz) applications. The resulting antenna exhibits acceptable
dual-band operation (i.e., better than 10 dB return loss with 5.3
and 7% operating bandwidths at 1.9 and 2.4 GHz) while main-
taining a cross-pol maximum field level at least 11 dB below the
co-pol maximum.
Index Terms—Genetic algorithms, method of moments (MoM),
microstrip antenna, optimization, parallel computing.
I. INTRODUCTION
A
current trend in electronics technology is the emphasis
on increasingly stringent system requirements in both the
commercial as well as military sectors, in addition to main-
taining low costs in manufacturing, operations and maintenance.
For the military, the new paradigm shift is toward network-cen-
tric warfare, wherein a major emphasis is placed on the com-
plex interaction between the various information subsystems
that comprise a complete military system. Hence, the desire is to
obtain a system that is capable of reconnaissance, data analysis,
ordnance control, communications, etc., all in a real-time setting
via an ad hoc virtual network. Antenna designs for both ground-
and airborne-based subsystems present a unique challenge, in
that they should be as simple as possible and low-cost while at
the same time satisfying the particular electrical requirements.
In the commercial domain, the development of Wireless Fidelity
(WiFi) Internet access systems (IEEE 802.11b), 2.5 G and 3
G wireless technology, broadband cellular technology that han-
dles high-rate voice and data, etc., has also placed a significant
Manuscript received October 19, 2002; revised April 25, 2003.
F. J. Villegas was with Raytheon Electronic Systems, El Segundo, CA
90245-0902 USA. He is now with The Aerospace Corporation, El Segundo,
CA 90245–4691 USA.
T. Cwik is with the High Performance Computing Group, Jet Propulsion Lab-
oratory, Pasadena, CA 91109 USA.
Y. Rahmat-Samii and M. Manteghi are with the Antenna Research and
Measurement Laboratory, Department of Electrical Engineering, University
of California at Los Angeles, Los Angeles, CA 90095-1594 USA (e-mail:
rahmat@ee.ucla.edu; www.ee.ucla.edu).
Digital Object Identifier 10.1109/TAP.2004.834071
burden on the design of low-cost antennas that achieve quite re-
markable specifications in terms of bandwidth, gain, multiband
operation, and physical (e.g., size) constraints.
As a result, designers have had to turn to ever-more inge-
nious methods to achieve these goals. A technique that has be-
come quite popular over the last several years has been the use
of evolutionary optimization strategies for electromagnetic de-
sign. In particular, the use of genetic algorithms (GA) has ex-
ploded onto the research scene with great success, predomi-
nantly due to its particular characteristics that make it an ideal
tool that marries quite well with existing EM analysis tech-
niques [1]–[5], and typically yields results that satisfy the given
requirements in a nonintuitive fashion. A great deal of effort
has already been expended in furthering both the computational
maturity of GA optimization in electromagnetics [3], [6]–[8], as
well as in extending the domain of applications to include quite
ingenious designs [9]–[18]. Two distinct focus areas in which
GA optimization has yielded quite fruitful results are novel pat-
tern synthesis [19]–[30] and broadband (or multiband) opera-
tion [31]–[34]. Another area in which the use of GA designs
shows promise is the development of “smart” antennas [35].
In this article, we describe an electromagnetic GA optimiza-
tion (EGO) application (introduced in [36]) that has been devel-
oped for the cluster supercomputing platform, and is thus quite
powerful and apropos for today’s tough antenna design prob-
lems. A representative patch antenna design example for com-
mercial applications is detailed, which illustrates the versatility
and applicability of the method. We show that EGO allows us to
combine the accuracy of full-wave EM analysis with the robust-
ness of GA optimization and the speed of a parallel computing
algorithm. In Section II, the EGO application software architec-
ture proposed in [36] is presented in greater detail, i.e., the EM
analysis procedure and parallel infrastructure is fully developed.
In particular, we present a more in-depth development of the par-
allel application architecture. The single-program multiple data
model is in essence a key feature that allows us to use the more
accurate full-wave method of moments (MoM) simulations in
conjunction with the evolutionary optimization approach. We
also provide a more detailed treatment of parameter extraction
steps required by the GA’s fitness function after the MoM solu-
tion has been computed. Although not explicitly made use of in
the present study, Section II-A describes a more general func-
tionality implicit in EGO for the design of N-port guided-wave
and radiating structures. Section III then presents a detailed rep-
resentative patch antenna design case study. We illustrate the
0018-926X/04$20.00 © 2004 IEEE