Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214

Contents lists available at ScienceDirect

International Journal of Electronics andCommunications (AEÜ)

j ourna l h omepage: www.elsev ier .com/ locate /aeue

REGULAR PAPER

Design of load-ended spiral antennas for RFID UHF passive tags usingimproved artificial bee colony algorithm

Sotirios K. Goudos a,∗, Katherine Siakavara a, John N. Sahalosb

a Radiocommunications Laboratory, Department of Physics, Aristotle University of Thessaloniki, GR-541 24 Thessaloniki, Greeceb Radio & Telecommunications Laboratory, Department of Electrical and Computer Engineering, University of Nicosia, Cyprus

a r t i c l e i n f o

Article history:Received 13 June 2014Accepted 12 September 2014

Keywords:RFID tag designOptimization methodsEvolutionary algorithmsSwarm intelligenceArtificial bee colony optimization

a b s t r a c t

In this paper, new planar spiral antennas for passive RFID tag application at UHF band are designed andoptimized using a new artificial bee colony (ABC) algorithm. We apply the improved ABC (I-ABC), which isan improved version of the original ABC algorithm. The I-ABC introduces the best-so-far solution, inertiaweight and acceleration coefficients to modify the search process. The optimization goals are antennasize minimization, gain maximization and conjugate matching. The antenna dimensions were optimizedand evaluated using I-ABC in conjunction with commercial EM software. We compare the I-ABC withthe original ABC algorithm. The obtained results show that both algorithms are powerful optimizers thatcan be efficiently applied to tag antenna design problems. I-ABC produces better results than the originalABC algorithm in terms of solution accuracy and success rate. RFID tags with dimensions less than 3 cm,maximum gain that reaches the value of 1.46 dBi and read distance more than 10 m were among thoseobtained by the algorithm.

© 2014 Elsevier GmbH. All rights reserved.

1. Introduction

The radiofrequency identification (RFID) technology is knownfor more than two decades but has been used extensively as faras in the last decade. Nowadays the RFID technology providingautomated wireless identification and tracking capability and beingmore robust than the barcode system, has shown a commercialworldwide deployment following frequency allocation in the UHFband, ranging from 860 MHz to 950 MHz. An ordinary RFID systemcomprises of at least, a reader (interrogator) with a reader antenna[1,2], tags (transponders) which are microchips combined with anantenna in a compact package, a host computer and middlewareincluding software and data base. Tag antennas are crucial ele-ments for the good performance of an RFID network. The basic taskof a tag antenna designer is to obtain high efficiency and effectiveimpedance matching to IC chips with typically capacitive reactance[3–6].

These antenna requirements are essential to optimize the RFIDsystem power performance, especially for passive configurations

∗ Corresponding author at: Radiocommunications Laboratory, Department ofPhysics, Aristotle University of Thessaloniki, GR-541 24 Thessaloniki, Greece.Tel.: +30 2310998392; fax: +30 2310998069.

E-mail addresses: sgoudo@physics.auth.gr, sgoudos@gmail.com (S.K. Goudos),sgoudo@physics.auth.gr, skv@auth.gr, sahalos.j@unic.ac.cy (J.N. Sahalos).

where the only energy source is the incoming reader energy. Theantenna designer is confronted with two major problems. Thefirst is the antenna miniaturization, which opposes to the desiredattribute of relatively high gain and the second is the fact that thetag antenna has to be conjugate matched to the capacitive reactanceof the IC. The first of these problems could be solved by select-ing the shape of the printed structure in a way that it would beimprovable with respect to its size. The second problem could besolved by changing the geometrical parameters of the antenna viathe ordinary trial and error method or by employing an algorithmof optimization.

Various geometrical configurations have been employed for thedesign of antennas of this kind and have been proved effectivein giving to them the required characteristics of operation. In thepresent work, the spiral shaped antenna printed on a dielectricslab was selected because this geometry has the attribute to fillthe space it occupies. Thus, it has small size and at the same timehas long length that is capable of resonating at small frequencies.Moreover, this configuration has several geometrical parameters,which could assist the optimization algorithm to find a good solu-tion. In addition to these parameters a novel one was selected withintend to be employed at the effort to minimize the antenna’s sizeand to conjugate match it with the IC. We consider that the spi-ral is not open-ended, as it is usually found in the literature, butload-ended. The load layout parameters are additional geometricalparameters that also contribute to the design process.

http://dx.doi.org/10.1016/j.aeue.2014.09.0081434-8411/© 2014 Elsevier GmbH. All rights reserved.

S.K. Goudos et al. / Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214 207

Several RFID design cases can be found in the literature [7–12].Most of these use a trial and error procedure for selecting theantenna geometry. Evolutionary algorithms (EAs) are suitable opti-mization tools for solving the above-described design problem.Genetic algorithms (GA) and ant colony optimization (ACO) havebeen applied successfully to RFID antenna design [13–15]. The arti-ficial bee colony (ABC) algorithm [16] models the foraging behaviorof honey bee swarm. The ABC algorithm has been applied suc-cessfully to RFID tag design [17]. In this paper, we extend ourearlier work [17–19]. We apply a recently proposed ABC variant, theImproved ABC (I-ABC) [20] to load-ended spiral antennas design.The I-ABC introduces the best-so-far solution, inertia weight andacceleration coefficients to modify the search process. To best of theauthors’ knowledge, this is the first time that the I-ABC algorithmis applied to an electromagnetics design problem. We compare theI-ABC with the original ABC algorithm on two design cases using

two different objective functions. The results show that I-ABC out-performs the original ABC algorithm.

2. Artificial bee colony algorithm

The ABC algorithm models and simulates the honey bee behav-ior in food foraging. In ABC algorithm, a potential solution to theoptimization problem is represented by the position of a foodsource while the nectar amount of a food source corresponds to thequality (objective function fitness) of the associated solution. Forpopulation-based optimization algorithms, both exploration andexploitation are necessary. The exploration refers to the ability toinvestigate the various unknown regions in the solution space todiscover the global optimum, while, the exploitation refers to theability to apply the knowledge of the previous good solutions tofind better solutions [10].

Fig. 1. I-ABC flowchart.

208 S.K. Goudos et al. / Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214

In order to find the best solution the algorithm defines threeclasses of bees: employed bees, onlooker bees and scout bees.The employed bee searches for the food sources, the onlooker beemakes a decision to choose the food sources by sharing the infor-mation of employed bee, and the scout bee is used to determinea new food source if a food source is abandoned by the employedbee and onlooker bee. Each employed bee corresponds to one foodsource (i.e. the number of the employed bees is equal to the num-ber of solutions). The employed bees search for new neighbor foodsource near of their hive.

2.1. Improved ABC (I-ABC)

The ABC algorithm is a very effective technique, which canhandle efficiently arbitrary optimization problems. The mainadvantages of using ABC are not only the fact that it is a high per-formance optimizer but also that it is an easy to understand andimplement algorithm that requires little computational bookkeep-ing. However, ABC could converge slowly and sometimes may trapin local optima. The authors in [20] in order to further improve theABC performance, propose three major changes by introducing thebest-so-far solution, inertia weight and acceleration coefficients tomodify the search process. In addition, the original solution genera-tion equation of ABC is good at exploration but poor at exploitation[20]. Therefore, they improve the exploitation by differentiation ofthe modification forms of the employed bees and the onlooker onesin the second acceleration coefficient. The new proposed improvedABC algorithm is called as I-ABC. The I-ABC differs from the originalABC algorithm in new solution generation equation. A new positionof the xi = (xi,1, . . ., xi,j, . . ., xi,D) solution, where D is the problemdimension, is generated using:

ui,j = xi,jwi,j + 2(rand1[0,1] − 0.5)(xi,j − xk,j)˚1

+ rand2[0,1](xbest,j − xk,j)˚2 (1)

where k ∈ {1, 2, . . ., SN} , k /= i, j ∈ {1, 2, . . ., D} are randomly chosenindices, where SN is the number of food sources, wi,j is the inertiaweight that controls impacts of the previous solution xi,j, xbest,j is thejth parameter of the global best solution so far, rand1[0,1], rand2[0,1]

are uniformly distributed random numbers in [0,1], and ˚1, ˚2 arethe acceleration constants, that is positive control parameters thatcontrol the maximum step size. These along with the inertia weightare given by:

wi,j = ˚1 =1

1 + e−fiti/ap

˚2 =

⎧

⎪

⎨

⎪

⎩

1 for employed bees

1

1 + e−fiti/apfor onlooker bees

(2)

where ap is the fitness of the global best in the first iteration, andfiti is the fitness value of the ith solution. To balance the processesof the exploration and the exploitation, the modification forms ofthe employed bees and the onlooker ones differ in the accelerationcoefficient ˚2. I-ABC uses a greedy selection operator, which forminimization problems is defined by

x′i =

{

ui, if f (ui) < f (xi)

xi, otherwise(3)

where x′iis the new position of the food source.

An onlooker bee chooses a food source depending on the prob-ability value associated with that food source, pi, given by:

pi =fiti

∑SN

m=1fitm

(4)

where the fitness value of the ith solution is proportional to the nec-tar amount of the food source in the ith position. When a food source(solution) cannot be improved anymore through a predeterminednumber of trials, called “limit”, then the scout bee helps the colonyto randomly generate create new solutions:

xi,j = randj(0,1)(xj,U − xj,L) + xj,L j = 1, 2, . . ., D (5)

where xj,L and xj,U are the lower and upper bounds of the jth dimen-sion respectively and randj(0,1) is a uniformly distributed randomnumber within (0,1). The flowchart of the I-ABC algorithm is givenin Fig. 1 The I-ABC algorithm is outlined as follows.

(1) Initialize the I-ABC parameters. That is, set the maximum cyclenumber (MCN), the number SN of food sources (populationsize), and the limit parameter.

(2) Evaluate the objective function value for every vector of thepopulation.

(3) For cycle = 1 set the ap = fit1.(4) Employed bees phase. For each employed bee. Produce a new

solution vector ui using (1) and (2). Evaluate new solutions andapply greedy selection rule for new solutions of (3).

(5) Calculate the probability values pi

(6) Onlooker bees phase. For each onlooker bee. Depending onprobability pi, produce a new solution vector ui using (1) and(2) from the selected solution xi. Evaluate new solutions andapply greedy selection rule for new solutions of (3).

(7) Scout bees phase. If the limit number is reached for a solutionrandomly generate another using (5).

(8) Keep the best solution so far.(9) If the maximum number of cycles MCN is reached, then exit;

otherwise increment cycle number and go to step 4 for the nextcycle.

More details about the I-ABC algorithm can be found in [20].

Fig. 2. RFID tag antenna geometry.

S.K. Goudos et al. / Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214 209

Fig. 3. RFID tag antenna geometry. (a) RFID tag with total size 66.82 mm × 65.82 mm, (b) 3D radiation pattern, (c) radiation patterns for different phi angles and (d) radiationpattern for theta = 90◦ .

3. Problem description

The first difficulty in designing a tag antenna arises from therequirement to be conjugate matched to the IC chip, which hasinput impedance with relatively low real part and high capacitiveimaginary part. Therefore, the antenna’s input impedance musthave an inductive imaginary part of equal value. Moreover, thedemand for the tag antenna to be of small size is also opposed tothe desired attribute of relatively high gain.

Therefore, the two optimization goals for RFID tag design aregain maximization and conjugate matching. We have designed tagsusing two different objective functions. By the first one optimiza-tion with respect to both, gain and matching, is simultaneouslyattempted. It is described by Eq. (1):

F1(x) = −G(x) + � × | max{0, |� | − 0.3}| (6)

where x is the vector of the antenna geometry, G is the antenna gaincalculated, � is a very large number, and � is the (load depend-ent) reflection coefficient of the tag antenna-load system which iscomputed by

� =Zin,chip − Z∗

a

Zin,chip + Za(7)

where Zin,chip = Rin,chip + jXin,chip is the chip input impedance andZa = Ra + jXa the tag input impedance respectively.

The second objective function was

F2(x) = min |� | (8)

By this function, the optimization goal is the minimizationreflection coefficient regardless of the antenna gain. Additionallythe antenna size, in the optimization process for the second objec-tive function was restricted to 40 × 40 mm.

An important parameter of the RFID system performance, asmentioned before, is the read range that is the maximum distanceat which RFID reader can detect the backscattered signal from thetag. As reader sensitivity is typically high in comparison with tag,the read range is defined by the tag response. The read range isexpressed using Friis free-space formula as [2]

R =�

4�

√

PEIRPrdeDtagploss�

Pin,chip(9)

where PEIRPrd is the effective isotropically radiated power by thereader, Dtag is the tag directivity, e is the tag efficiency, ploss is thepolarization loss factor (represents the loss of EM power because ofpolarization mismatch 0 ≤ ploss ≤ 1), Pin,chip is the power absorbedby the chip given by

Pin,chip = (1 − |� |2)Ptag (10)

210 S.K. Goudos et al. / Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214

Table 1

Load-ended RFID tag case 1 best geometry obtained by I-ABC.

SL (mm) D0 (mm) m1 m2 Per1 (%) Per2 (%) hl (mm) wl (mm)

4.26 3.74 8 2 20.9 1 1.54 7.71

Table 2

Comparative results of the best tag characteristics found by the two algorithms for the first objective function. The values are for 867 MHz.

Algorithm Gain (dBi) Maximum read distance (m) Za (Ohm) |� |

I-ABC 2.15 14.19 49.708 +j209.43 0.24ABC 2.00 9.36 23.463 +j196.85 0.29

where Ptag is the available power at the input of the tag antenna,and � is the power transmission coefficient given by

� =4RaRin,chip

|Zin,chip + Za|2

(11)

3.1. Antenna geometry

In the present work, as it was above-mentioned, the load termi-nated spiral shape was selected for the RFID tag implementation.The loading of the ends of the spiral contributes to the conjugatematching of the chip impedance. Therefore, we have designed a spi-ral antenna type with loaded ends. Fig. 2 shows the spiral geometry.The optimization parameters are the first spiral length SL, the spac-ing between the two spirals D0, the first branch number m1 (i.e.the number of the horizontal and vertical sections of each spiralbranch), the second branch number m2, the horizontal load lengthhl, and the vertical load additional width wl. The last horizontallengths are cropped arbitrary. Therefore, we need two additionaldesign parameters, which represent the percentage length reduc-tion of the last horizontal length of the first per1 and the secondspiral per2 respectively. Thus, a total of eight design parameters.

4. Numerical results

In this paper, we design printed tag antennas for the IC chipHiggs 3 of Alien Com, with size, small as possible and gain greaterthan zero for the 865–869 MHz frequency band. All the RFID tagspresented are designed for operation in Europe, where the min-imum effective isotropically radiated power (EIRP) requirementis PEIRPrd = 3.28 W. The power threshold for the chip to run on isPin,chip = −18 dBm = 15.84 �W. All of the spirals are structured withcooper strips of 1 mm width, printed on dielectric substrate withεr = 1.046 and thickness h = 3.17 mm. The tags are assumed to bein free space. All algorithms are compiled using the same com-piler (MS Visual C++ 2010) on a PC with Intel Core i5-3470 at3.20 GHz with 8 GB RAM running Windows 7. The computation ofthe objective functions requires the use of a full-wave numericalmethod. The RFID antenna is modeled in FEKO. In order to inte-grate the in-house source code of the ABC algorithms with FEKO,a wrapper program is created. All results from antenna charac-teristics are obtained by FEKO. The reader antenna characteristicsare: emitted power 1 W, gain 7 dBiC and circular polarization. Thereceiver sensitivity is −70 dBm and ASK modulation is used. The

Fig. 4. Input impedance versus frequency of the best spiral case 1.

optimization is performed on a single frequency (867 MHz) withisotropic gain.

We compare I-ABC with the original ABC algorithm on the twoobjective functions presented in the previous section. Both algo-rithms ran for 20 independent trials. A population of 20 vectorswas selected. The total number of iterations was set to 300. Thestopping criterion was the iteration number. The limit parameteris set to 100 for both algorithms. The algorithms’ statistical resultsare compared. The best value, the worst value, the mean, and thestandard deviation of the last generation computed by each algo-rithm are presented here. Additionally, if each algorithm has notimproved the objective function value for 200 iterations then itstops to avoid stagnation. We also obtain the minimum FEmin, andthe average FEavg number of objective-function evaluations for eachalgorithm and the success rate (Sr) which is the number of eachalgorithm successful runs divided by the total runs number.

The first design example is that of a spiral designed with thefirst objective function of gain maximization. The design parame-ters of the best result found by the I-ABC algorithm are presentedin Table 1. We have obtained a tag gain of 2.15 dBi and |� |=0.24 at867 MHz. The tag size is 66.82 mm × 65.82 mm and the maximumread range between 865 and 869.8 MHz is 14.19 m. The conjugate

Table 3

Comparative results of the algorithms for the first objective function. The smaller values are in bold.

Algorithm Best Worst Mean Standard deviation Avg. computation time (h)

I-ABC −2.15 −1.05 −1.60 0.37 4.3ABC −2.00 −0.88 −1.51 0.51 4.1

S.K. Goudos et al. / Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214 211

Fig. 5. First objective function tag using IABC and ABC. (a) Chip input power and (b) reflection coefficient versus frequency.

matching bandwidth is about 2.4 MHz. The chip input impedanceat 867 MHz according to vendor datasheets is 30.46−j211.58 Ohm.

Fig. 3a depicts the surface current distribution of the best spi-ral found and Fig. 3b shows the 3D radiation pattern. We noticethat the direction of the maximum radiation is perpendicular tothe tag surface. Fig. 3c and d shows the radiation patterns at dif-ferent phi angles and at theta = 90◦ planes respectively. The besttag found by I-ABC is compared with the best tag found by ABC.Table 2 reports the tag characteristics found by both algorithms. Itis obvious that the tag found by I-ABC outperforms the one foundby ABC. The input impedance plot for both tags is depicted inFig. 4. Fig. 5a shows the variation of the Pin,chip versus frequencyfor the maximum read range. There are two sets of data for everytag where the Pin,chip is calculated for two different read ranges,which are the maximum of IABC tag R = 14.19 m and the maxi-mum of tag ABC R = 9.36 m. It is obvious that the chip of tag I-ABCoperates above the required power threshold from 865 MHz to869 MHz at distance R = 14.19 m. However, tag ABC at the samedistance operates above the threshold in a bandwidth less than2 MHz. Furthermore, tag I-ABC operates at a higher power levelthan tag ABC for R = 9.36 m. The comparison of the reflection coef-ficient graphs is depicted in Fig. 5b. It is obvious that tag I-ABC hasa larger bandwidth than tag ABC. Additionally, the maximum read

Table 4

Number of objective function evaluations comparative results for the first objectivefunction. The smaller values are in bold.

Algorithm Sr FEmin FEavg

I-ABC 0.8 4455 5788ABC 0.7 4557 5609

range plot for both tags is shown in Fig. 5c. Again IABC outperformsABC.

Table 3 reports the algorithms comparative results. The I-ABC algorithm seems to perform better than the ABC algorithm.Regarding computational time for this case both algorithms requireabout 4 hours for each algorithm run. ABC is slightly faster than I-ABC Table 4 holds the number of objective function evaluationsresults with the success rate. We notice that the I-ABC algorithmhas achieved the highest success rate. The ABC obtained the smalleraverage number of objective-function evaluations.

The second design case presented has been obtained using thesecond objective function of reflection coefficient minimizationwithout taking into account the gain value. Table 5 shows the geom-etry of the best result obtained by I-ABC. The best RFID tag obtained

212 S.K. Goudos et al. / Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214

Fig. 6. RFID tag antenna geometry. (a) RFID tag with total size 28.17 mm × 34.25 mm, (b) 3D radiation pattern, (c) radiation patterns for different phi angles and (d) radiationpattern for theta = 90◦ .

is with a tag gain of 1.46 dBi and |� |=0.0011 at 867 MHz. The tag sizeis 28.17 mm × 34.25 mm and the maximum read range between865 MHz and 869.8 MHz is 10.5 m. The conjugate matching band-width is about 2.2 MHz. The geometry of the best spiral found isshown in Fig. 6a. The 3D radiation pattern is depicted in Fig. 6b. Theradiation patterns at different phi angles and at theta = 90◦ plane isdepicted in Fig. 6c and d respectively. Again the best tags found by

both algorithms are compared. Table 6 holds the tag characteris-tics found by both algorithms. I-ABC tag again outperforms the onefound by ABC. Fig. 7 shows the input impedance plots for both tags.Fig. 8a shows the variation of the Pin,chip versus frequency for themaximum read range. Again as in the previous case, there are twosets of data for every tag where Pin,chip is calculated for two differ-ent read ranges, which are the maximum of IABC tag R = 10.5 m and

Table 5

Load-ended RFID tag case 2 best geometry obtained by I-ABC.

SL (mm) D0 (mm) m1 m2 Per1 (%) Per2 (%) hl (mm) wl (mm)

3.72 3.36 2 4 12.5 25.1 1.29 5.52

Table 6

Comparative results of the best tag characteristics found by the two algorithms for the second objective function. The values are for 867 MHz.

Algorithm Gain (dBi) Maximum read distance (m) Za (Ohm) |� |

I-ABC 1.46 10.5 30.398 +j211.58 0.0011ABC 1.21 7.23 33.825 +j211.69 0.0523

Table 7

Comparative results of the algorithms for the second objective function. The smaller values are in bold.

Algorithm Best Worst Mean Standard deviation Avg. computation time (h)

I-ABC 0.001 0.505 0.173 0.165 4.7ABC 0.052 0.506 0.283 0.177 4.6

S.K. Goudos et al. / Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214 213

Fig. 7. Input impedance versus frequency of the best spiral case 1.

the maximum of tag ABC R = 7.23 m. It is evident that the chip of tagI-ABC operates above the required power threshold from 865 MHzto 869 MHz at distance R = 10.5 m. However, tag ABC at the samedistance operates above the threshold in a bandwidth about 3 MHz.

Table 8

Number of objective function evaluations comparative results for the second objec-tive function. The smaller values are in bold.

Algorithm Sr FEmin FEavg

I-ABC 0.8 5117 5859ABC 0.5 4376 5550

Additionally, tag I-ABC operates at a higher power level thantag ABC for R = 7.23 m. Fig. 8b holds the comparison of the reflec-tion coefficient graphs for both tags. Tag I-ABC presents a largerbandwidth than tag ABC. Furthermore, Fig. 8c depicts the maxi-mum read range versus frequency plots for both tags. It is clearthat I-ABC outperforms ABC.

Table 7 shows the objective function value comparative results.It is evident that I-ABC obtains better results than the ABC. AgainABC is slightly faster than I-ABC. Table 8 presents the number ofobjective function evaluations results with the success rate for thiscase. In this case the feasible or success results are those with|� |<0.3. Although ABC obtained the smaller number of objectivefunction evaluations, I-ABC obtained the best success rate (Table 8).

We notice that the first objective function obtained results withhigher gain and larger dimensions. The second objective as it couldbe expected obtained smaller designs as the gain was not optimizedfor this case.

Fig. 8. Second objective function tag using IABC and ABC. (a) Chip input power and (b) reflection coefficient versus frequency.

214 S.K. Goudos et al. / Int. J. Electron. Commun. (AEÜ) 69 (2015) 206–214

5. Conclusion

Various spiral antennas with loaded ends for passive UHF tagswere designed and optimized using two different ABC algorithms.Both were proven efficient for the design of this type of printedantennas as they lead to configurations with small size and simul-taneously offer the potential to control the input impedance of thestructure by modifying the spiral’ s number of the antenna andthe loads’ size. The designed tags exhibit satisfactory large readdistances and the antennas ensure the necessary input power tothe chip for a frequency range of about 5 MHz. The tag designwith the higher gain is that with 2.15 dBi gain and 14.19 m readrange. The smaller design has dimensions less than 35 mm. TheI-ABC algorithm is an improved ABC variant that uses the best-so-far solution, inertia weight and acceleration coefficients to modifythe search process. The results obtained with I-ABC are better thanthose obtained with the original ABC algorithm. Both algorithmshave proven to be quite efficient and powerful for RFID tag design.

Acknowledgements

This research has been co-financed by the European Union(European Social Fund – ESF) and Greek National Funds throughthe Operational Program “Education and Lifelong Learning” of theNational Strategic Reference Framework (NSRF) – Research Fund-ing Program: THALES: Investing the knowledge society through theEuropean Social Fund, MIS 377271.

References

[1] Djeddou M, Khelladi R, Benssalah M. Improved RFID anti-collision algorithm.AEU – Int J Electron Commun 2013;67:256–62.

[2] Tu Y, Guo X, Wang C, Jin J, Du S, Sun J, et al. An improved 860–960 MHz fullyintegrated CMOS power amplifier designation for UHF RFID transmitter. AEU– Int J Electron Commun 2013;67:574–7.

[3] Lehto A, Nummela J, Ukkonen L, Sydanheimo L, Kivikoski M. Passive UHF RFIDin paper industry: challenges, benefits and the application environment. IEEETrans Autom Sci Eng 2009;6:66–79.

[4] Rao KVS, Nikitin PV, Lam SF. Antenna design for UHF RFID tags: a review and apractical application. IEEE Trans Antennas Propag 2005;53:3870–6.

[5] Raida Z, Kolka Z, Marsalek R, Petrzela J, Prokes A, Sebesta J, et al. Commu-nication subsystems for emerging wireless technologies. Radioengineering2012;21:1036–49.

[6] Tang Z-J, He Y-G, Wang Y. Broadband UHF RFID tag antenna with quasi-isotropicradiation performance. AEU – Int J Electron Commun 2011;65:859–63.

[7] Li Y, Rongwei Z, Staiculescu D, Wong CP, Tentzeris MM. A novel conformalRFID-enabled module utilizing inkjet-printed antennas and carbon nanotubesfor gas-detection applications. IEEE Antennas Wirel Propag Lett 2009;8:653–6.

[8] Mahmoud KR. Design optimization of a bow-tie antenna for 2.45 GHzRFID readers using a hybrid BSO-NM algorithm. Progr Electromagn Res2010;100:105–17.

[9] Abdulhadi AE, Abhari R. Design and experimental evaluation of miniatur-ized monopole UHF RFID tag antennas. IEEE Antennas Wirel Propag Lett2012;11:248–51.

[10] Zi Long M, Li Jun J, Jingtian X, Ye TT. A single-layer compact HF-UHF dual-bandRFID tag antenna. IEEE Antennas Wirel Propag Lett 2012;11:1257–60.

[11] Panda JR, Saladi ASR, Kshetrimayum RS. A compact printed monopole antennafor dual-band RFID and WLAN applications. Radioengineering 2011;20:464–7.

[12] Perret E, Hamdi M, Vena A, Garet F, Bernier M, Duvillaret L, et al. RF and THzidentification using a new generation of chipless RFID tags. Radioengineering2011;20:380–6.

[13] Marrocco G. Gain-optimized self-resonant meander line antennas for RFIDapplications. IEEE Antennas Wirel Propag Lett 2003;2:302–5.

[14] Calabrese C, Marrocco G. Meandered-slot antennas for sensor-RFID tags. IEEEAntennas Wirel Propag Lett 2008;7:5–8.

[15] Randall M, Lewis A, Galehdar A, Thiel D. Using ant colony optimisation toimprove the efficiency of small meander line RFID antennas. In: 3rd IEEE Inter-national Conference on E-Science and Grid Computing; 2007. 2007. p. 345–51.

[16] Karaboga D, Basturk B. A powerful and efficient algorithm for numericalfunction optimization: artificial bee colony (ABC) algorithm. J Global Optim2007;39:459–71.

[17] Goudos SK, Siakavara K, Sahalos JN. Novel spiral antenna design using artificialbee colony optimization for UHF RFID applications. IEEE Antennas Wirel PropagLett 2014;13:528–31.

[18] Goudos SK, Siakavara K, Sahalos JN. Synthesis of miniaturized load-ended spiralantennas for UHF passive tags. In: IEEE-APS Topical Conference on Anten-nas and Propagation in Wireless Communications, IEEE APWC 2013. 2013. p.927–30.

[19] Goudos SK, Siakavara K, Sahalos JN. Modified spiral RFID tag antenna optimaldesign using artificial bee colony optimization. In: 43rd European MicrowaveConference, EuMC 2013. 2013. p. 1255–8.

[20] Li G, Niu P, Xiao X. Development and investigation of efficient artificial beecolony algorithm for numerical function optimization. Appl Soft Comput J2012;12:320–32.