Genetic Algorithm Based Efficient Tag Detection inRFID Reader Networks

Nazish Irfan and Mustapha C.E. YagoubSchool of Electrical Engineering and Computer Science (EECS)

University of Ottawa, Ottawa, Canadanirfan@site.uottawa.ca, myagoub@site.uottawa.ca

Khelifa HettakCommunication Research Center (CRC)3701 Carling Avenue, Ottawa, Canada

Email: khelifa.hettak@crc.gc.ca

Abstract—Radio Frequency Identification (RFID) systems, dueto recent technological advances, have been used for variousadvantages in industry like production facilities, supply chainmanagement etc. However, sometimes this requires a densedeployment of readers to cover the working area. Withoutoptimizing reader’s location and number, many of them will beredundant, reducing the efficiency of the whole RFID system.There are many algorithms proposed to solve this redundantreader problem, but all existing algorithms are based on omni-directional reader antenna pattern, which is not practical. In thispaper, a genetic algorithm is used to optimize the antenna beamto eliminate redundant reader based on real directional readerantenna pattern.

Index Terms—RFID, Antenna, Reader-Coverage, Optimiza-tion, Genetic Algorithm.

I. INTRODUCTION

Radio Frequency Identification (RFID) is based on radiocommunication for tagging and identifying an object [1]. Itconsists of two blocks namely, RFID transceivers (readers)and RFID transponders (tags). Over the last few years, RFIDhas drawn a great deal of attention and it is now widelybelieved that RFID can bring revolutionary changes [2]. Someof the major retailers have already invested significantly inRFID and mandated their manufacturers to place tags on casesand pallets, which resulted in mass production of inexpensiveRFID tags [3].

In recent years, more efforts have been made to implementRFID applications in inventory control and logistics manage-ment. RFID based system leads to significant reduction onprocessing time and labor as inventory in warehouses can betracked more accurately in a simple, timely and more efficientmanner [4]. Applications like inventory detection and auto-mated product receiving in supply chain management requireRFID readers to read tags anywhere within a large geographicarea. Since the range of reader-to-tag communication is verylimited, readers must be deployed in high densities over theentire area [5]. Therefore, the deployment of RFID systems hasgenerated the RFID network planning problem that needs tobe solved for large scale deployment. However, RFID networkplanning is one of the most challenging problem since it hasto meet many requirements for efficient operation of RFIDsystem [6].

This dense deployment of RFID systems in large scale re-sults in unwanted effects. In fact, when multiple readers share

the same working environment and communicate over sharedwireless channels, a signal from one reader may reach otherreaders and can cause interference among readers. Hence,the effect of reader interference on the RFID interrogationrange should be analyzed before any large scale deploymentof readers in a RFID system [7]. Indeed, unnecessary readersin the network may consume power, which can be wasteful.Therefore, finding redundant readers is of great importance foran optimal deployment of a RFID network.

The problem of redundant reader elimination has beenstudied extensively in [8]−[11]. Elimination of redundantreader ensures user that an optimal number of readers areused to cover all the tags in a specified zone. In this paper,we proposed a genetic algorithm based practical approachto find redundant readers based on real directional readerantenna pattern. We have demonstrated that practical mea-sured/simulated data from commercial or in-house antennascan be efficiently utilized to eliminate redundant readers fromRFID networks. The redundant reader elimination techniquescan be then used for efficient deployment of readers in largewarehouse, manufacturing units in production lines and largeretail stores.

II. RELATED WORK

During the last decade, the RFID collision problem hasbeen extensively covered in literature. It can be categorizedas reader-to-reader interference or reader-to-tag interference.Reader-to-reader interference occurs when the interrogationzones of two readers intersect and can interfere with eachother. Two readers may also interfere each other even if theirinterrogation zones do not overlap. This interference is dueto the use of wireless radio frequencies for communication.Reader-to-tag interference occurs when more than one readertry to read the same tag simultaneously. In this type of interfer-ence, each reader may believe that it is the only reader commu-nicating with the tag while the tag, in fact, is communicatingwith multiple readers at the same time. The reader collisionproblem not only results in incorrect operation but also resultsin reduction of overall read rate of the RFID system [7],[12], [13]. To separate the individual participant signal fromone another, many procedures have been developed. Basically,there are four main procedures namely, the Carrier SenseMultiple Access (CSMA), the Frequency Domain Multiple

978-1-61284-925-6/11/$26.00 ©2011 IEEE

Access (FDMA), the Time Domain Multiple Access (TDMA)and the Code Division Multiple Access (CDMA) [14]. Thereare many algorithms in literature, which cover reader collisionproblem [5], [12], [13], [15].

Redundant reader elimination techniques mentioned aboveare all based on omni-directional reader antenna, which isnot the case for real antennas designed for RFID systems.Moreover, in [16], the authors mentioned that antennas havingan omni-directional radiation pattern should be avoided andwherever possible directional antennas should be used becausedirectional antennas have advantages of less disturbance to theradiation pattern and to the return loss. Also, omni-directionalantennas have short range and no particular advantage in termsof signal-to-noise ratio (SNR). Therefore, for more realisticapplications, redundant reader elimination should be based ondirectional reader antenna pattern.

In antenna, tilt is another important design parameter fornetwork planning when considering coverage vs. capacity forcell planning as well as tuning network. Antenna tilting ofbase station is a very common technique for improving cellisolation and / or increasing covers in cellular networks. Tiltangle optimization can be achieved electrically, mechanicallyor by a combination of both [17]. It is reported that bothcoverage and quality performances are very sensitive to an-tenna tilt variations [18]. There are many papers [17], [19],[20] in literature, which have studied the effect of antenna tiltcontrol on radio network planning and cellular network. Theirauthors have shown that significant reductions in path-lossdelay spread, transmitted power and system interference can beachieved when suitable height and antenna tilt are selected. Itis reported that in a dense network, the importance of antennadown tilt becomes more significant. Therefore, in this paperwe have presented an redundant reader elimination approach,based on an antenna tilt optimization using genetic algorithm,which can be effectively utilized with any commercial or in-house directional antenna radiation pattern data.

III. GENETIC ALGORITHM

A genetic algorithm (GA) is an optimization and searchtechnique based on the principles of genetics and naturalselection. Initially the population is generated randomly. Thefitness values of all chromosomes are evaluated by calculatingthe objective function in a decoded form (phenotype). Fromthe population, a particular group of chromosomes (parents)are selected to generate the offspring by the defined geneticoperations. The fitness of the offspring is evaluated in a similarfashion to their parents. Based on certain replacement strategy,the chromosomes in the current population are replaced bytheir offspring. The above GA cycle is repeated until a desiredtermination criterion is reached. Finally, the best chromosomesin the final population can become a highly evolved solution ofthe population. There are various techniques that are employedin the GA process for encoding, fitness evaluation, parentselection, genetic operation and replacement.Encoding Scheme: In GA algorithm, encoding scheme is oneof the key issues because it can severely limit the window of

information that is observed from the system. In general, theGA evolves a multi-set of chromosomes and each chromosome𝑥𝑖(𝑖 = 1, 2, ⋅ ⋅ ⋅𝑁) represents a trial solution to the problemsetting. The chromosome is usually expressed in a string ofvariables, each element of which is called a gene. The variablecan be represented by a binary real number and its range isusually defined by the problem specified.

Fitness Techniques: In GA, the status of each chromosomeis evaluated using an objective function. The objective functiontaking chromosome as an input produces a list of numbersas a measure to the performance of the chromosome. Thefitness function is needed to map the objective value to afitness value to maintain uniformity over various problemdomains. Linear normalization, Roulette wheel weighting andTournament selection are some of the commonly used fitnesstechniques.

Genetic Operations

∙ Crossover: This operation combines subparts of two par-ent chromosomes to produce offspring that contain someparts of both parents genetic material. There are mainlytwo types of crossover, i.e. single and multi-point.

∙ Mutation: This is an operator that introduces variationsinto the chromosome. This variation can be global orlocal.

IV. PROPOSED ALGORITHM AND PROBLEM DEFINITION

In any arbitrary RFID network it is desirable that anyreader covers maximum number of tags in its coverage zoneto minimize the number of readers in a network. Therefore,proposed work uses antenna tilt to optimize the antenna beamplacement so that a reader can cover maximum number oftags in its coverage zone. The proposed algorithm assumesthat the readers and the tags coordinates are easily available.The readers and the tags coordinates are then used to optimizeantenna tilt by using genetic algorithm. The reader is thenadjusted with the optimized tilt angle to maximize its coverage.

The cost function for genetic algorithm is defined in termsof both forward link i.e. successful reading of a tag by a readerand backward link i.e. successful signal received from a tagby a reader and given by

𝐶𝑜𝑠𝑡 = 𝐹 (𝑃𝑟𝑒𝑎𝑑𝑒𝑟−𝑡𝑎𝑔, 𝑃𝑡𝑎𝑔−𝑟𝑒𝑎𝑑𝑒𝑟) (1)

where 𝑃𝑟𝑒𝑎𝑑𝑒𝑟−𝑡𝑎𝑔 the total power received by a tag froma reader (forward link) and 𝑃𝑡𝑎𝑔−𝑟𝑒𝑎𝑑𝑒𝑟 the reflected powerreceived by a reader from a tag (reverse link).

The simplest way to determine the power received by thetag from the reader in free space path loss environment canbe given by the Friis equation [21].

𝑃𝑟 = 𝑃𝑡𝐺𝑡𝐺𝑟

(𝜆

4𝜋𝑑

)2

(2)

where, 𝑃𝑟 = received power by tag, 𝑃𝑡 = transmitted powerfrom reader, 𝐺𝑡 = transmitter antenna gain, 𝐺𝑟 = receiverantenna gain, while 𝑑 determines the length of the direct path

between the transmitter and the receiver antenna. The factor(𝜆

4𝜋𝑑

)2determines the free space path loss.

Practical deployment of RFID systems in indoor line-of-sight environment involves the in-building path loss. In thiswork, we used the indoor propagation model given by [21].

𝑃𝐿(𝑑𝐵) = 𝑃𝐿(𝑑0) + 10𝑛𝑙𝑜𝑔

(𝑑

𝑑0

)(3)

where 𝑛 is the path loss exponent that depends on surroundingsand building types and 𝑑0 the reference distance.

V. SIMULATION SETUPS AND RESULTS

In this section we present a proposed technique to eliminateredundant readers using real directional reader antenna. Figure1 shows the typical scenario where a reader can optimize thecoverage by adjusting its tilt angle. This scenario is commonto redundant reader elimination setup in RFID networks withantenna readers having directional radiation patterns. From thisfigure, we see that reader 1 after optimizing its tilt angle cancover tags 1-5 and similarly reader 2 covers tags 3-5. Sincereader 1 covers all the tags which reader 2 also covers thenreader 2 can be eliminated safely without compromising thenetwork coverage.

Fig. 1. Typical redundant reader topology

To evaluate the performance of the proposed technique, weimplemented an experimental setup having an area of 20𝑚×20𝑚, 3 readers and 32 tags as shown in Figure 2. As the testdevice, the Intermec RFID reader IA33A circularly polarizedpanel was adopted and the Intermac UHF tag was used [22].

In this work, a continuous genetic algorithm encodingscheme was adopted and the Roulette wheel fitness techniquewas used to evaluate the objective function. Genetic algorithmparameters used for optimization are mutation rate = 0.2 andselection = 0.5.

Figure 2 shows the number of tags covered by each readerwhen reader tilt angle is not optimized i.e. the initial setup ofthe RFID network. It can be seen that initially, when readers

0 5 10 150

5

10

15

x (meters)

y (m

eter

s)

Tags covered by reader 1Tags covered by reader 2Tags covered by reader 3Readers in netwrok

Reader 3

Reader 1

Reader 2

Fig. 2. RFID network topology for tilt angle optimization

tilt angle is not optimized, reader 1 covers 12 tags, reader 2covers 8 tags and reader 3 covers 12 tags, respectively. Afteroptimizing the tilt angle of reader 2 and reader 3, it can beobserved that reader 2 and reader 3 cover 16 tags each. Reader1 having no tag under its coverage is therefore removed asredundant reader. Figure 3 shows the coverage of reader 2and similarly Figure 4 presents the coverage of reader 3 afteroptimization. It can be worth mentioning that initially, the useris not aware of the optimum tilt positioning of reader antenna.Table I summarizes the results of the experiment.

0 5 10 150

5

10

15

x (meters)

y (m

eter

s)

Tags covered by reader 2Readers in network

Reader 3

Reader 1

Reader 2

Fig. 3. Tags covered by reader 2 with tilt angle optimization

Simulation results prove that the optimal placement of thebeam angle of a directional reader antenna can efficientlyimprove the tag detection by increasing the reader coverage.The optimization helps in deciding on the optimal numberof reader placement thereby reducing the overall cost of theRFID system. This technique can be used in pre-planning ofoptimal number of reader placement with optimized antenna

TABLE INUMBER OF TAGS COVERED BEFORE AND AFTER OPTIMIZING THE TILT

ANGLE

Reader Tags covered Tags covered Optimized Statusbefore after tilt angle

1 12 NILL NILL Eliminated2 8 16 -16.17 ∘ IN3 12 16 -22.75 ∘ IN

beam angle for warehouse and logistics applications.Authors in [23] have a presented algorithm for efficient tag

detection based on omni-directional reader antenna pattern,whereas the practical reader antenna for RFID applicationshas directional antenna pattern. Therefore, our approach suitsmore to practical applications of RFID systems. Moreover,the algorithm proposed in [23] requires many read-write oper-ations therefore suitable only for tags which support both readand write operations. Write operations require more resourceswhich eventually add to the cost of operation of RFID systems.On the other hand, our approach do not requires read or writeoperation therefore suitable to any type of tags.

0 5 10 150

5

10

15

x (meters)

y (m

eter

s)

Tags covered by reader 3Readers in network

Reader 3

Reader 1

Reader 2

Fig. 4. Tags covered by reader 3 with tilt angle optimization

VI. CONCLUSION

In this paper we proposed a Genetic algorithm to optimizethe tilt angle of a directional RFID reader antenna to efficientlyimprove the tag detection by removing redundant reader fromthe network. Simulation results demonstrate that the proposedapproach can effectively works as a redundant reader elimina-tion tool for RFID readers with practical directional radiationpatterns. This work can be used in pre-planning of warehouseand logistics applications to assess the number of RFID readersused with optimal position of directional antenna beam.

REFERENCES

[1] M. Jo, C. G. Lim and E. W. Zimmers, ”RFID tags detection on watercontent using a back-propagation learning machine,” KSII Transactionson Internet and Information Systems, vol. 1, no. 1, pp. 19-32, 2007.

[2] Y. Bendavid, S. F. Wamba and L. A. Lefebvre, ”Proof of concept ofand RFID-enabled supply chain in a b2b e-commerce environment,” InProceedings of the 8th International Conference on Electronic Commerce(ICEC06), pp. 564-568, 2006.

[3] S. E. Sarma, ”Towards the five-cent tag,” Technical Report MIT-AUTOAID-WH-006, MIT Auto ID Center, 2001.

[4] Q. Wang, R. McIntosh and M. Antony, ”A RFID-based automated ware-house design,” In 2nd International Conference on Computer Engineeringand Technology (ICCET), vol. 6, pp. v6-359 - v6-363, 2010.

[5] J. Waldrop, D. W. Engels and S. E. Sarma, ”A MAC for RIFD readernetworks,” In IEEE Wireless Communications and Networking (WCNC)),vol. 3, pp. 1701-1704, 2003.

[6] H. Chen and Y. Zhu, ”RFID networks planning using evolutionary algo-rithms and swarm intelligence,” In 4th International Conference on Wire-less Communications, Networking and Mobile Computing (WiCOM)),pp. 1-4, 2008.

[7] D. W. Engels,”The reader collision problem,” White Paper MIT-AUTOID-WH-007, MIT Auto ID Center, 2001.

[8] B. Carbunar, M. K. Ramanathan, M. Koyuturk, C. Hoffmann andA. Grama, ”Redundant-reader elimination in RFID systems,” In SecondAnnual IEEE Communications Society Conference on Sensor and AdHoc Communications and Networks (SECON), pp. 176-184, 2005.

[9] C. H. Hsu, Y. M. Chen and C. T. Yang, ”A layered optimization approachfor redundant reader elimination in wireless RFID networks,” In IEEEAsia-Pacific Services Computing Conference, pp. 138-145, 2007.

[10] K. M. Yu, C. W. Yu and Z. Y. Lim, ”A density-based algorithm forredundant reader elimination in a RFID network,” In Proceedings of theSecond International Conference on Future Generation Communicationand Networking, vol. 1, pp. 89-92, 2008.

[11] N. Irfan and M. C. E. Yagoub, ”Efficient algorithm for redundantreader elimination in wireless RFID networks,” International Journal ofComputer Science Issues, vol. 7(3), pp. 1-8, 2010.

[12] D. H. Shis, P. L. Sun, D. C. Yen and S. M. Huang,”Taxonomy andsurvey of RFID anti-collision protocols,” J. Computer Communications,vol. 29, pp. 2150-2166, 2006.

[13] S. M. Birari and S. Iyer,” Mitigating the reader collision problemin RFID networks with mobile readers,” In 13th IEEE InternationalConference on Networks, vol. 1, pp. 463-468, 2005.

[14] K. Finkenzeller,RFID Handbook Fundamentals and applications incontactless smart cards and identification. John Wiley and SonsLtd.,Chichester, 2003.

[15] K. I. Hwang, K. T. Kim and D. S. Ecom,”DiCa: Distributed tagaccess with collision-avoidance among mobile RFID readers,” In: EUCWorkshops, pp. 413-422, 2006.

[16] P. R. Foster and R. A. Burberry, ”Antenna problems in RFID systems,”IEEE Colloquium on RFID Technology, pp. 3/1-3/5, 1999.

[17] F. Athley and M. N. Johansson, ”Impact of electrical and mechanicalantenna tilt on LTE downlink system performance,” IEEE 71st VehicularTechnology Conference (VTC 2010), pp. 1-5, 2010.

[18] E. Dinan and A. Kurochkin, T̈he effects of antenna orientation errorson UMTS network performance,” IEEE 17th International Symposium onPersonal, Indoor and Mobile Radio Communications), pp. 1-5, 2006.

[19] J. Niemela, T. Isotalo and J. Lempiainen, ”Optimum antenna downtiltangles for macro cellular WCDMA network,” EURASIP Journal onWireless Communications and Networking, vol. 2005 (5), pp. 816-827,2005.

[20] M. Wadhwa, M. Song, V. Rali and S. Shetty ”The impact of antennaorientation on wireless sensor network performance,” In 2nd IEEE Inter-national Conference on Computer Science and Information Technology,pp. 143-147, 2009.

[21] T. S. Rappaport, Wireless communications principles and practice,second edition, Prentice Hall PTR, New Jersey, 2002.

[22] Intermec, ℎ𝑡𝑡𝑝 : //𝑤𝑤𝑤.𝑖𝑛𝑡𝑒𝑟𝑚𝑒𝑐.𝑐𝑜𝑚/𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠/𝑟𝑓𝑖𝑑/𝑎𝑛𝑡𝑒𝑛𝑛𝑎𝑠.[23] B. Carbunar, M. K. Ramanathan, M. Koyuturk, C. Hoffmann and

A. Grama,”Efficient tag detection in RFID systems,” Journal of Paralleland Distributed Computing, vol. 69, pp. 180-196, 2009.