+ All Categories
Home > Documents > Automobile conformal antenna design using non-dominated sorting genetic algorithm (NSGA)

Automobile conformal antenna design using non-dominated sorting genetic algorithm (NSGA)

Date post: 20-Sep-2016
Category:
Upload: ek
View: 216 times
Download: 2 times
Share this document with a friend
4

Click here to load reader

Transcript
Page 1: Automobile conformal antenna design using non-dominated sorting genetic algorithm (NSGA)

Automobile conformal antenna design usingnon-dominated sorting genetic algorithm (NSGA)

Y. Kim and E.K. Walton

Abstract: The use of conformal antennas for vehicle applications is growing very rapidly due to thedevelopment of modern wireless communication technology and service. One of the popular waysto design automobile conformal antennas is to modify an existing antenna. These modificationprocesses rely on the designer’s intuition and many tedious measurements. The authors propose acomputational conformal antenna design based on the nondominated sorting genetic algorithm(NSGA) modifying existing antenna types to improve performance. Simulations and measurementresults are presented and discussed.

1 Introduction

In the modern automobile industry, the need for newautomobile conformal antennas to satisfy growing newdevelopments and services, such as conformal FM radio,global positioning systems (GPS), satellite digital audioradio systems (SDARS), etc., has been increasing rapidly[1–3]. Automobile antenna designers are interested inchoosing conformal antennas for new automobiles due totheir various advantages. Conformal antennas can reducewind noise and drag, improve the aesthetics of the vehicle,and are safer than typical whip-type antennas with respectto vandalism and damage [4]. As the need for newautomobile conformal antennas is rapidly growing, thedemand for a computational tool to design and optimisethe automobile antennas has also been growing. Mostautomobile antenna design procedures strongly rely onthe designer’s intuition in modifying existing antennatypes. Also, these procedures can involve many tediousmeasurements.

Recently, an antenna design and optimisation techniqueusing a genetic algorithm has been introduced [5–9]. Thispaper proposes a new computational tool to design andoptimise automobile conformal antennas using a nondomi-nated sorting genetic algorithm (NSGA). We apply theNSGA to modify existing antenna shapes to improveperformance. The NSGA can find a set of Pareto-optimalsolutions, instead of finding a single optimal solution [10].In multi-objective optimisation problems like automobileantenna design, one may not find a single best solution.There may be many solutions, which are considered betterwith respect to all objectives. The Pareto-optimum solutionsare a set of compromise solutions based on a comparisonwith each objective, such as low VSWR, omnidirectional

E-mail: [email protected]

Y. Kim is with the Communication Laboratory, Samsung Advanced Instituteof Technology, Mt.14-1, Nongseo-Ri, Giheung-Eup, Yongin-Si, Gyeonggi-Do,Korea

E.K. Walton is with the Department of Electrical Engineering, The Ohio StateUniversity, Columbus, OH 43210, USA

r The Institution of Engineering and Technology 2006

IEE Proceedings online no. 20050055

doi:10.1049/ip-map:20050055

Paper first received 18th March 2005 and in final revised form 13th March 2006

IEE Proc.-Microw. Antennas Propag., Vol. 153, No. 6, December 2006

gain pattern, less complexity, etc. In this paper, a set ofPareto-optimum automobile conformal antenna geometriesfor FM radio is produced. The antenna designers can easilyand reliably apply these optimised antenna geometries tonew automobile body shapes. Also, the many tediousmeasurements can be significantly reduced. Several compu-tational methods have been applied to solve EM properties,including those of automobile antennas [11, 12]. In thispaper the method of moments (MoM) code (ESP5,developed by Dr. Newman at The Ohio State UniversityElectroScience Laboratory) is used as the electromagneticnumerical analysis tool [13].

2 Antenna design and optimisation processes

Let us suppose we already had a conformal antenna.However, the performance of the antenna is no longersatisfactory for to various reasons, such as changingthe automobile body shape and dimensions. Then, weneed to modify this existing antenna to satisfy the newautomobile. In that case, we can modify the existingantenna using the proposed design processes in this paper.An example of existing conformal antenna geometry isshown in Fig. 1. The shape of the existing antenna isrelatively simple.

The modification and generation processes of antennageometry processes using NSGA are also explained inFig. 1. Basically, the generated antenna has a symmetricshape. The right part of the antenna is generated by theNSGA process and the left part of the antenna is of asymmetric shape to the right side of the antenna. A total of48 bits are used in NSGA processes. The number ofindividual elements in the initial populations is 80. Themaximum number of generations is 15. The operationfrequency is 100MHz. Only vertical polarisation isconsidered for the directional gain pattern. The totalnumber of wire modes in ESP5 is approximately 1970.The total computational time is approximately 18 hours.Two objective cost functions are used. The first costfunction (G1) is the voltage standing wave ratio (VSWR)value, and the second cost function (G2) is the average gainof the azimuthal gain pattern of the antenna. Table 1 showsthe specific objectives and corresponding cost functions inthis design process. The Gain_dB (y¼ 901, f) in Table 1is a gain value at the elevation angle y¼ 901 and azimuth

579

Page 2: Automobile conformal antenna design using non-dominated sorting genetic algorithm (NSGA)

angle f. Subtract one from G2 when the gain value at eachelevation angle with (y¼ 901,f) is larger than � 3dBi. Thetotal coverage of the azimuthal angle is from f¼ � 601 tof¼ 601 and from f¼ 1201 to f¼ 2401 with 11 stepsbecause we focus on the omnidirectional gain pattern at thefront and back of the car. It is shown that the gain values atthe of the car are typically less than � 3dBi [14]. Becausewe prefer that the maximum gain value is as high aspossible, the maximum gain value is subtracted from G2. Inthese cost functions, the lower the cost value means thebetter antenna the performance. Our goal is to findimproved solutions compared to the performance of theexisting antenna.

Fig. 1 Existing conformal antenna geometry and modification

Table 1: Two cost functions for NSGA process

Objectives Cost functions

G1 VSWR G1¼VSWR at f¼100MHz

G2 Omnidirectionalazimuth gainpattern

Start-Initial G2¼0

For f¼ from01 to 3601 (with 1 degree step)

if (Gain_dB (y¼901, f)4 �3dBi)

G2¼G2� 1 (subtraction)

end

end

if (max(Gain_dB)40dBi)

G2¼G2� 3�max(Gain_dB)

end

if (�3dBio¼max(Gain_dB)o ¼0dBi)

G2¼G2� (3þmax(Gain_dB))

end

580

3 Designed and optimised antennas

Some of the individuals at generation 15 are shown inFig. 2. The individual for the existing antenna is also shownin Fig. 2. The G1 (VSWR) and G2 (gain) cost values of theindividual of the existing antenna are 3.37 and � 55.2.Notice that several populations can generate better costvalues (smaller than those cost values of the individual ofthe existing antenna) after 15 generations. Among thosebetter solutions, three solutions are selected. In the case ofindividual element number 31, the VSWR is higher than3.37, but the gain cost value (G2) is lower than � 55.2. Inthe case of individual number 1, the VSWR is lower (2.57against 3.37), while the gain cost value is almost the same.Table 1 shows cost values of existing antenna geometry andimproved antenna geometries, which are obtained byNSGA processes (Table 2).

The corresponding antenna geometries are shown inFig. 3. Figure 3a shows the antenna geometry for theexisting model and Fig. 3b shows the antenna geometry forthe best gain case (no. 31). Also, Fig. 3c shows the antennageometry for the compromise solution case (no. 2), andFig. 3d shows the lowest VSWR case (no. 1). Notice thatthe antenna grid models of these conformal antennas areconsiderably different.

4 Measurement of the designed and optimisedantennas

In this section, the measurement results of the NSGA-designed automobile FM heater-grid antennas, which areproduced in the previous section are presented. We haveobtained a copper-coated scale model of the automobile.This model was created from a 1 : 24 scale plastic model thatwas coated using a copper plating technique. It has beenshown that the performance of the coated scale model isclose enough for the antenna engineer to evaluate theprototype antennas [14]. A photograph of the copper-coated scale model is shown in Fig. 4, and includes a modelof a quarter wavelength monopole in the centre of the roof.This monopole antenna is used as a reference antenna forgain pattern measurement. It is well known that a quarterwavelength antenna on the same device, such as a car,can be used as a reference antenna [15]. The measuredgain values of the designed and optimised antenna are

Fig. 2 Populations at generation 15

IEE Proc.-Microw. Antennas Propag., Vol. 153, No. 6, December 2006

Page 3: Automobile conformal antenna design using non-dominated sorting genetic algorithm (NSGA)

normalised by the measured gain values of the referencemonopole antenna. Also, notice the heater-grid conformalantenna. Three heater-grid antennas, which are shown inFig. 3, are made of 0.2mm diameter metal wires andmounted on a very thin microscope cover glass.

Figure 5 shows the VSWR comparison between theexisting antenna geometry (shown in Fig. 3a) and theimproved VSWR case (shown in Fig. 3d). In this VSWRmeasurement, the actual measured frequency range is1–3GHz with a 10-MHz step. Then, the data are convertedto the UHF range to compare with the simulation data.Notice that the average VSWR in the FM frequency rangeof the improved VSWR case is approximately 2, whilethe average VSWR over the same frequency range ofthe existing antenna is approximately 4.5. It is clear that theVSWR can be improved when the modified antenna isused. Azimuth gain pattern measurements were also

Fig. 3 Corresponding heater-grid conformal antenna geometriesand existing conformal antenna geometry modificationa Existing antenna geometryb No. 31–lowest gain costc No. 2–compromise solutiond No. 1–lowest VSWR

Table 2: Cost values of the selected new antenna geometryand existing conformal antenna geometry

Individual number G1 cost(VSWR)

G2 cost(gain cost)

Comment

Existing geometry 3.37 � 55.2

No. 31 7.72 � 70.54 best gain case

No. 2 3.07 � 62.28 compromise solution

No. 1 2.57 � 58.22 best VSWR case

Fig. 4 Copper-coated small scale automobile model formeasurement

IEE Proc.-Microw. Antennas Propag., Vol. 153, No. 6, December 2006

performed. The result is shown in Fig. 6. Note that thepattern shapes of the existing antenna case and the best gaincase are close. However, the average gain value of the bestgain case is approximately 3dB higher. The gain perfor-mance of the best gain antenna is better than that of theexisting antenna. Since the ESP5 gain calculation is basedon an assumption of perfect input match, this gainimprovement is not due to the VSWR improvement.

5 Conclusions

It has been shown that an existing antenna shape canbe modified to generate improved antenna performance(VSWR is improved from 4.5 to 2, and average azimuthgain is 3dB higher) using the NSGA process. The

Fig. 5 VSWR comparisons between an existing antenna and amodified antenna in the FM frequency range

Fig. 6 Azimuthal gain pattern comparisons between existingantenna and a modified antenna at f¼ 100 MHzdBr: dB with respect to a reference monopole antenna

581

Page 4: Automobile conformal antenna design using non-dominated sorting genetic algorithm (NSGA)

theoretical expectations were verified using the copper-coated small scale (1 : 24 subscale) model measurements.Therefore, it has been demonstrated that the NSGA can beused effectively to design and optimise FM frequency bandconformal antennas.

6 References

1 Hall, S.W.: ‘Vehicle window glass antenna arrangement’, U.S. Patentno. 5936585, Aug. 1999

2 Nagayama, Y., and Maegawa, M.: ‘Glass antenna for side windshieldof automobile vehicle’, U.S. Patent no. 6437749, Aug. 2002

3 Oka, H.: ‘Glass antenna and glass system using the same’, U.S. Patentno. 6906671, June 2005

4 Abou–Jaoude, R.N.: ‘Design and development of conformal auto-mobile antennas using numerical modeling and experimental techni-ques’. PhD Dissertation, The Ohio State University, Columbus, OH,USA, 1997

5 Villarroel, W.: ‘Automated design and optimization of VHF/UHFautomotive conformal antennas’. PhD Dissertation, The Ohio StateUniversity, Columbus, OH, USA, 2002

6 Rahmat–Samii, Y., and Michielssen, E.: ‘Electromagnetic optimiza-tion by genetic algorithm’ ( John Wiley & Sons Inc., NewYork, 1999)

582

7 Boag, A., Boag, A., Michielssen E., and Mittra, R.: ‘Design ofelectrically loaded wire antennas using genetic algorithm’, IEEE Trans.Antennas Propag., 1996, 44, pp. 687–695

8 Linden D.S.: ‘Wire antenna optimized in the presence of satellitestructures using genetic algorithm’. Proc. IEEE Radio and WirelessConf., 2000, pp. 91–99

9 Chung, Y.C., and Haupt, R.: ‘Log-period dipole array optimization’.IEEE Aerospace Conf., March 2000, pp. 449–455

10 Deb, K.: ‘Multi-objective optimization using evolutionary algorithm’( John Wiley and sons, Ltd, 2001)

11 Notaros, B.M., Djordjevic, M.L., Popovic, B.D., and Popovic, Z.:‘Rigorous EM modeling of cars and airplanes’. Radio and WirelessConf., 1999, pp. 176–180

12 Ruddle, A.R., Sarantidis, A., and Ward, D.D.: ‘Modeling theinstalled performance of vehicle antennas using TLM’. Proc.IEE Conf. Antennas Propag., IEE Conf. Proc. no. 361, 1999,pp. 5–7

13 Newman, E.H.: ‘A users manual for electromagnetic surface patchcode: Version V (ESP5)’ (The Ohio State University ElectroscienceLaboratory, 1998)

14 Walton, E., Newman, E., Kim, Y., and Jamil, K.: ‘Experimentalmeasurements and computational modeling for automobile antennas’.Rep. 740016-2, Electroscience Lab., Ohio State Univ., Columbus, OH,USA, August 2001

15 PROCOM International Website, ‘Antennas and antennaconcepts’, http://www.procom-com/techinfo/e-0507-antennas-measuring-gain

IEE Proc.-Microw. Antennas Propag., Vol. 153, No. 6, December 2006


Recommended