Conception and Performance Analysis of Efficient CDMA-Based Full-Duplex Anti-collision Scheme

Transcription

1 Coceptio ad Performace Aalysis of Efficiet CDMA-Based Full-Duplex Ati-collisio Scheme Xiaohua Cao ad Tiffay Li Ultra-hih-frequecy radio-frequecy idetificatio (UHF RFID) is widely applied i differet idustries. The Frame Slotted ALOHA i EPC C1G2 suffers severe collisios that limit the efficiecy of ta recoitio. A efficiet full-duplex ati-collisio scheme is proposed to reduce the rate of collisio by coordiati the trasmitti process of CDMA UWB uplik ad UHF dowlik. The relevat mathematical models are built to aalyze the performace of the proposed scheme. Throuh simulatio, some importat fidis are aied. The maximum umber of idetified tas i oe slot is /e ( is the umber of PN codes ad e is Euler s costat) whe the umber of tas is equal to m (m is the umber of slots). Ulike the Frame Slotted ALOHA, eve if the frame size is small ad the umber of tas is lare, there are t too may collisios if the umber of PN codes is lare eouh. Our approach with 7-bit Gold codes, 15-bit Gold codes, or 31-bit Gold codes operates 1.4 times, 1.7 times, or 3 times faster tha the CDMA Slotted ALOHA, respectively, ad 14.5 times, 16.2 times, or 18.5 times faster tha the EPC C1 G2 system, respectively. More tha 2, tas ca be processed withi 3 ms i our approach. Keywords: UHF RFID, CDMA, full-duplex, aticollisio, performace aalysis, EPC C1 G2. Mauscript received Mar. 27, 214; revised May 21, 215; accepted Jue 25, 215. This work was supported by Techoloy Support Project of Hubei Provice (214BAA33). Xiaohua Cao (correspodi author, Tomm_cao@163.com) is with the Iformatio Techoloy Ceter, Wuha Uiversity of Techoloy, Hubei, Chia. Tiffay Li (xic213@lehih.edu) is with the Departmet of Electroic Eieeri, Lehih Uiversity, Bethlehem, PA, USA. I. Itroductio Radio-frequecy idetificatio (RFID) techoloy has bee widely applied i may idustry areas. A hue umber of RFID tas miht appear i a readi zoe simultaeously. Whe receivi demads from a reader, all the tas will sed sials almost simultaeously. These sials will collide [1]. This is the reaso why ati-collisio procedures are i existece today. Such procedures are widely used ad are desied to prevet tas from broadcasti their iformatio simultaeously. Almost all existi passive RFID systems are based o half-duplex commuicatio (which uses a arrowbad radio frequecy). A reader seds a RF carrier sial to a umber of tas, ad these tas backscatter the data sial by chai the atea load. I both of the aforemetioed sial directios, a arrowbad RF sial is used [2]. However, this simple implemetatio holds some drawbacks. It is sesitive to iterferece, multipath fadi, multiuser iterferece, ad collisios, as well as bei susceptible to passive ad active attacks. Existi UHF RFID ati-collisio solutios are based o time divisio multiple access (TDMA) [3]. Fiure 1 shows a TDMA method; ote that the tas i the reader s field trasmit their data at differet momets i time. The Electroic Product Code Class-1 Geeratio-2 (EPC C1 G2) stadard is widely used i UHF passive RFID systems. It utilizes Frame Slotted ALOHA [4], which is based o time slots that sychroize at the start of a trasmissio. There are two mai reasos for the aforemetioed stadard s low ETRI Joural, Volume 37, Number 5, October Xiaohua Cao ad Tiffay Li 929

2 Ta1 T T T T Collidi Ta2 Ta3 Fi. 1. TDMA ati-collisio method. T T T T Ta4 Ta3 Ta2 Ta1 Ta5 Ta6 Ta7 Ta8 Ta9 Ta4 Collidi Ta11 Ta1 Fi. 2. CDMA ati-collisio method. Time Time throuhput. Firstly, EPC C1 G2 utilizes backscatter modulatio for uplik (ta-to-reader) commuicatio, which limits the data rate to a few hudred kilobits per secod (kb/s) ad is vulerable to dese multipath pheomeo ad multiuser iterferece. Secodly, thouh EPC C1 G2 adopts the Q- alorithm (a dyamic Framed Slotted ALOHA alorithm [5] [6]) to mitiate ta collisios, it still suffers from severe ta collisios ad lacks multiple access capability. I future UHF RFID applicatios, a UHF RFID eeds either a fast recoitio capacity for massive RFID tas (more tha 1, tas) or a shorter recoitio time. Besides a TDMA alorithm, aother solutio is to itroduce code divisio multiple access (CDMA) to RFID commuicatio systems [7]. A CDMA-based RFID system usi semi-passive UHF traspoders was proposed i [8], with the reader providi the recoitio of multiple traspoders simultaeously. This meas that the traspoders are trasmitti data withi the same time rae ad frequecy bad, as show i Fi. 2. The results of previous work i the field [9] [1] have show that CDMA trasmissio ca be very attractive for RFID systems, due to its ood multipleaccess capabilities. However, whe CDMA techoloy is applied i UHF RFID, several difficulties arise, such as the fact that CDMA techoloy is eery hury, ta costs, ad data rates. Recetly, CDMA ultra-widebad (UWB) techoloy was cosidered as oe promisi wireless techique for future RFID systems [11] [12]. UWB pulse-positio modulatio (PPM) uses short-duratio (i the order of oe aosecod or less) pulses with a very low duty cycle for commuicatios ad has the possibility of achievi Mb/s hih throuhput. If CDMA UWB ca be well used as a ta-to-reader lik, the it will eable multiple tas to be detected simultaeously. However, oly basic architectures for such related CDMA UWB RFID systems have bee proposed i previous works [13] [14]. To reduce a ta s eery cosumptio, the recoitio time, which is closely related to the ati-collisio approach, should be as short as possible. So, it is very importat to develop a efficiet ati-collisio approach that is coducive to the likes of the aforemetioed systems; such a approach has ot bee cosidered i ay of the aforemetioed works. I eeral, almost all of the existi ati-collisio approaches are based o semi-duplex commuicatio modes. If CDMA UWB is to be used as a ta-to-reader lik i UHF RFID systems, the there are some challees that eed to be overcome to improve the performaces of CDMA-based ati-collisio approaches. Firstly, RFID ati-collisio commuicatio modes have to be compliat with UHF/UWB dual-bad system structures ad remotely powered tas. Secodly, a reader must be able to accurately estimate the umber of uidetified tas. Thirdly, a preferable pseudo-oise (PN) code alo with its leth must be selected. II. Literature Review May ati-collisio methods for UHF RFID systems have already bee proposed. The Radom ALOHA ad uslotted ALOHA [15] protocols have both bee prove; the priciple of ALOHA forms the basis of all moder ati-collisio protocols. A extesio of the ALOHA protocol, called Slotted ALOHA [16], itroduced time slots i which a trasmitter must sed its data at the beii of a slot. Withi this exteded protocol, collisios oly occur withi a full time slot. Most curret RFID protocols are based o the priciple of Slotted ALOHA ad use the EPC stadard UHF Class-1 Geeratio-2 air iterface protocol. Improvi the curret stadard ati-collisio method by dyamically choosi a appropriate value of Q, for example, was described i [17] [19]. These works show that the riht choice of Q is of reat importace for the overall system performace. However, these Slotted ALOHA methods lack the ability for multiple access withi a sile slot. Therefore, CDMA ati-collisio methods were preseted for arrowbad RFID systems ad are commoly combied with Framed Slotted ALOHA schemes. I particular, Slotted ALOHA CDMA systems ad their correspodi performaces may be foud i [2] [21]. Other works describe certai CDMA systems that 93 Xiaohua Cao ad Tiffay Li ETRI Joural, Volume 37, Number 5, October 215

3 really outperform the TDMA-based systems. The traspoders, each equipped with a uique quasi-orthooal spreadi code (for example, Gold codes) (Gold, 1967a) [22], may use the radio chael wheever the traspoders are ready to trasmit their data (asychroous CDMA). Gold codes were proposed for RFID systems [23] [24] because they have much larer code sets tha other codes, which meas that the umber of collisios ca be further reduced. O the other had, loer Gold codes reduce the effective data rate. Thus, PN code leths must be adapted to other parameters of CDMA-based ati-collisio approaches; oce aai, this has ot bee researched i ay of the aforemetioed works. UWB RFID systems have bee proposed recetly [25] [26]. The potetial advataes of UWB for RFID iclude hih pulse rate i ta-to-reader commuicatio; low power cosumptio at the trasmitter side; ad robustess to multipath ad iterferece. However, a UWB receiver is both area ad power hury, which causes challees for tas to be remotely powered. May studies o the desis of UWB trasceivers show that UWB techoloy is a ood cadidate to achieve low power ad low complexity i implemetatios. I [27], several versios of the ALOHA alorithm are preseted to icrease the feasibility ad efficiecy of UWB trasceivers. A semi-duplex commuicatio procedure called ackowledmet is required to resolve collisios or failed trasmissios. Most of the above methods are improved at either the protocol or alorithmic level of a commuicatio lik; however, they usually result i a icrease i circuit complexity, chip area, implemetatio cost, ad power cosumptio. Dual-bad systems have bee itroduced [28]. They use covetioal UHF radio for the forward reader-to-ta lik ad impulse UWB radio for the backward ta-to-reader lik. The reader trasmits commads to tas via UHF radio lik. Tas are remotely powered by the UHF sials [29]. This solutio avoids usi a UWB receiver o tas ad meets the asymmetric traffic requiremets i the forward lik (with very few reader commads) ad the backward lik (with lare umbers of ta resposes). It maily focuses o how to desi hardware schemes yet hardly o ati-collisio problems. However, if there is o efficiet ati-collisio approach with advisable parameters, eve if a UHF RFID system adopts dualbad, the the recoitio efficiecy will ot be obviously improved ad tas will eed more eery. I additio, the estimati methods of tas used i UHF RFID systems are t suitable for UWB RFID systems. Referece [24] proposed that the reader makes all tas respod with the same patter before requesti ta IDs. The reader uses the maitude of the resposes to estimate the umber of tas ad the adjust the ext frame size. However, the effectiveess of this method is low i realistic scearios because of differeces i the backscatter power, chael oises, multipath iterferece, ad others. I summary, UWB CDMA is a better choice for UHF RFID to solve the ta-collisio problem. This paper is to discuss a ovel full-duplex ati-collisio approach to coordiate the trasmitti process of uplik ad dowlik, ad efficietly idetify massive tas by combii UHF RFID with CDMA. III. Our Approach 1. Improvemet of Dual-Bad RFID The EPC C1 G2 UHF RFID system is built o the basis of a backscatteri commuicatio scheme with semi-duplex mode. It is sesitive to multipath fadi, multiuser iterferece, ad collisios. So, a dual-bad passive RFID architecture was suested [3], as is show i Fi. 3. I this scheme, the reader has a UHF trasmitter, a UHF atea, a CDMA PPM UWB receiver, a UWB atea, ad a micro-cotroller for loic fuctio ad ati-collisio protocol. A ta cosists of a UHF receiver, a UHF atea, a CDMA PPM UWB trasmitter, a UWB atea, ad a basebad loic module. Tas eed either the modules for backscatteri modulatio ad FM/Miller ecodi or a UWB receiver. This scheme adopts two asymmetric liks icludi UHF dowlik ad UWB uplik. The reader trasmits commads to tas via UHF dowlik at a rate of 16 kbps. A ta is remotely powered by the UHF sials with a miimum iput RF power as low as 14.1 μw, ad its circuit has bee implemeted i a.13 μm 1.2 V CMOS process. The ta seds iformatio ecoded by PN code back to the reader via UWB uplik, which adopts a pulse rate of 1 Mp/s i.18 μm CMOS to lower the power cosumptio ad reduce the receiver complexity. This kid of architecture for dual-bad systems is promisi [28], [3]. Defiitely, ta collisios lead to a lare icrease i idetificatio time. So, a ati-collisio approach should correspod to these kids of system structures. We take some steps to improve the efficiecy of ta recoitio with reards to a dual-bad scheme. Firstly, the traditioal semi-duplex ati- Basebad loic module UHF receiver Power atheri CDMA PPM UWB trasmitter UHF trasmitter CDMA PPM UWB receiver Fi. 3. Dual-bad passive RFID scheme. Microcotroller ETRI Joural, Volume 37, Number 5, October 215 Xiaohua Cao ad Tiffay Li 931

4 Query Carrier wave ACK PN1 PN2 ACK PN3 PN4 ACK ACK PN5 PN6 Ta1 PN1 Ta5 PN5 Ta2 PN2 Ta6 PN5 Ta9 PN8 Ta11 PN9 Ta3 PN3 Ta7 PN6 Ta1 PN8 Ta12 PN1 Ta4 PN4 Ta8 PN7 Slot 1 Slot 2 Slot 3 Slot 4 Slot 5 Fi. 4. Semi-duplex commuicatio process i TDMA ati-collisio approaches. collisio mode is replaced by a full-duplex oe. I this ew mode, a reader ca coordiate the trasmitti process of uplik ad dowlik ad efficietly idetify multiple tas. Secodly, a uique PN code is solidified i a sile ta to reduce the computatio of the ati-collisio approach. The uique PN code is stored i the ta s memory, ad the eerati process of the PN code is removed from all CDMA ati-collisio processes. Fially, a accurate estimati method is proposed to advace the performace of the ati-collisio approach. 2. Full-Duplex Ati-collisio Process Covetioal TDMA ati-collisio approaches are based o the semi-duplex mode featured i the EPC C1 G2 stadard. A time slot separately icludes a ta s respose time ad a reader s ackowledemet time, as show i Fi. 4. However, because our proposed RFID system has siificat asymmetry betwee the dowlik ad the uplik (UWB data rate is much hiher tha the arrowbad radio data rate), if a covetioal TDMA ati-collisio approach is used, the the hih data ratio ad multiple access ability of the UWB chael will be wasted. The ackowledemet from the reader to tas becomes a bottleeck that decreases the etwork throuhput. To improve etwork throuhput, we propose a more efficiet full-duplex ati-collisio commuicatio scheme to overcome the bottleeck ad reduce the recoitio time. It trasmits uplik ad dowlik data simultaeously ad coordiates commads ad respodi data with rhythm. I additio, it adopts CDMA techoloy with appropriate parameters i the UWB uplik to ehace multiple access ability. The choice of a appropriate set of PN codes is a key issue whe desii CDMA systems. To describe the full-duplex ati-collisio approach, we firstly defie a status fla for each ta, ACK, which idicates whether a ta has bee idetified. Because CDMA techoloy is used i our approach, the respodi processes will sped the same duratio eve if more tha oe ta trasmits data simultaeously. The leth of a slot is determied by which is larer the duratio of ACK or the duratio of a ta uploadi data. The duratio of a ta respose is fixed withi a frame because all tas have the same leth of ID ad data. The tas use the same data rate to sed iformatio. Geerally, the duratio for a ta to upload data is set as a slot because it is loer tha reader ackowledemet time. The proposed fullduplex ati-collisio process is ive i Fi. 5. 1) Query a commad from the reader throuh UHF dowlik. A Query commad uses a Q value to set the umber of slots (frame size). 2) Tas receive a Query commad ad respod immediately. A ta chooses oe of the 2Q slots to sed its ID ad data back to the reader with the CDMA UWB. A ta seds data i oe slot ad receives the ACK i oe of the remaii slots. 3) The reader will sed a ACK ad the PN codes of idetified tas to allow the simultaeous ackowledmet of multiple tas if some tas were idetified i the previous slot. If a ta trasmitted its ID ad data i the previous slot ad detects its PN code from the ACK commad i the curret slot, the it meas that this ta has bee successfully idetified ad ackowleded. 4) A idetified ta will keep silet i the followi frames of this ivetory roud, whereas a uidetified ta will choose a ew slot for respose i the ext frame. 5) The reader estimates the umber of uidetified tas at the ed of a frame ad adjusts the Q value for the ext frame. I this process, the PN codes of idetified tas are stored to oe buffer of reader memory i the receivi sequece, as show i Fi. 5. These PN codes will be take out from the buffer i a first-i first-out sequece. Fiure 5 shows a example of these statuses (assumi h = 2). I Slot 1, Ta1, Ta2, Ta3, ad Ta4 trasmit their IDs ad data to the reader with four differet PN codes, PN1, PN2, PN3, ad PN4, respectively. The reader successfully receives the iformatio from the tas ad stores the PN codes to the buffer. I Slot 2, the reader takes out PN1 ad PN2 from the buffer ad seds a ACK commad, PN1 ad PN2 to ackowlede 932 Xiaohua Cao ad Tiffay Li ETRI Joural, Volume 37, Number 5, October 215

5 Query Carrier wave ACK PN1 PN2 ACK PN3 PN4 ACK ACK PN5 PN6 Ta1 PN1 Ta2 PN2 Ta5 PN5 Ta6 PN5 Ta9 PN8 Ta11 PN9 PN1 PN9 PN8 PN7 Ta3 PN3 Ta7 PN6 Ta1 PN8 Ta12 PN1 PN6 Ta4 PN4 Ta8 PN7 PN5 PN4 Slot 1 Slot 2 Slot 3 Slot 4 Slot 5 PN3 Fi. 5. Full-duplex ati-collisio process. the tas. Ta1 ad Ta2 detect their desired PN codes, respectively, ad keep silet i the rest of this ivetory roud. Meawhile, Ta5, Ta6, Ta7, ad Ta8 trasmit their IDs ad data i Slot 2. Amo them, Ta5 ad Ta6 choose the same PN code, PN5, ad Ta7 ad Ta8 have differet PN codes, PN6 ad PN7. The reader ecouters a collisio but receives PN6 ad PN7 correctly ad stores them to the buffer. I Slot 3, the reader takes out PN3 ad PN4 from the buffer ad seds a ACK, PN3, ad PN4 to ackowlede the tas. Ta3 ad Ta4 detect their desired PN codes, respectively, ad keep silet i the rest of this ivetory roud. Meawhile, Ta9 ad Ta1 trasmit their IDs ad data i Slot 3. They choose the same PN code, PN8. The reader ecouters a collisio ad does ot receive ay ta. Slot 4 is a empty slot without ay ta resposes ad seds oly ACK to tas for sychroizi the ext slot. Because the lobal clock is scalable ad cotrolled by the reader, it provides a possibility to shorte idle slots. By detecti the icomi sials at the beii of each slot, the reader ca determie if there is ay trasmissio i this time slot. Therefore, there is a limitatio, as follows: T ht T, (1) ACK PN Ta where T ACK is the duratio of ACK, T PN is the duratio of PN, ad T Ta is the duratio of ta respose. So, the umber of fetched PN codes, h, is determied by TTa TACK h. (2) TPN I the full-duplex ati-collisio process, oe of the followi four kids of statuses could appear i ay ive slot: 1) All the respodi tas have differet PN codes or there is oly oe ta to respod ad there is o collisio to happe. All the tas ca be idetified correctly. 2) All the respodi tas have the same PN codes ad there are collisios to happe. The tas caot be idetified correctly. 3) Some tas have the same PN codes i the respodi tas ad the others have differet PN codes, so collisios will happe but some tas with differet PN codes ca be idetified correctly. 4) There is o ta to respod. It is a idle slot. 3. Estimati Number of Tas A estimati alorithm that is able to estimate the umber of tas ad is suitable to our ati-collisio approach is ecessary to achieve the maximum throuhput. I [31], a rouh estimator of tas was ive, but it performed well for oly a small umber of tas. The estimator caot be applied directly to a CDMA-RFID system due to its lare error. So, a alterative approach is preseted. The receiver cotais pipes of matched filters as show i [32]. Each pipe correspods to a sile PN code. A empty pipe does ot detect a ta respodi with its relevat PN code i a slot; a successful pipe detects a ta with a uique PN code that is differet from the codes used by other tas i the same slot; ad a collisio pipe detects more tha oe ta respodi with its relevat PN code simultaeously. At first, we estimate the expected umber of tas per collisio pipe. Let m be the umber of slots i oe frame, the umber of tas, ad the umber of PN codes. The probability that a ta chooses a certai slot ad a uique PN code is ive by 1 P 1. (3) m Therefore, the umber of tas that choose the same slot ad the same PN code is biomially distributed as 1 i i 1 Ps () i 1. (4) i m m Let P / m ; s the maximum idetifyi probability ca be achieved whe = m. If i is larer tha 1 (2 i ), the there are collisios ad the i tas caot be recoized correctly. ETRI Joural, Volume 37, Number 5, October 215 Xiaohua Cao ad Tiffay Li 933

6 If oe pipe i a slot is observed as a collisio pipe, the the expected value of the umber of coflicti tas i oe pipe is E( Nctas ) ip ( i) C (5) i2 m m The expected value of coflicti probability i a collisio pipe is PC i (6) i2 m m m E( PC) ( ) The, the expected value of the umber of tas i the collisio pipe, ω, is ive by E( Nctas ). (7) E( PC ) Whe the maximum recoitio capacity is achieved (that is, = m), ω ca be represeted as follows: I practice, a reader ca detect collisio pipes. Let p be the umber of collisio pipes i oe frame, the the umber of uidetified tas amo the curret frame, N e, ca be estimated by N 1 1 p 11 ωp e 1 I our ati-collisio process, the value of is ukow before the first frame. Assumi that the umber of iitial tas is lare, it may be supposed to be ifiite; therefore, the estimated value of the umber of uidetified tas after the first frame ca be ca calculated by 1 (1 e ) lim Ne roud p roud (2.4 p). 1 (1) 1 2e The values calculated from (1) are compared with the statistical results of the umber of coflicti tas i simulatio, as show i Fi. 6. We set the leth of a PN code as 15. The statistical values of the umber of tas were obtaied by taki the mea of 5 statistical values, which were obtaied uder the same coditios. The results of our two simulatios show that the fiures obtaied by our estimati approach are very close to the simulatio statistics whe the umber of tas is smaller tha 2, ad m is 128 or 256; however, there are obvious errors whe the umber of tas is larer tha 1,2.. (8) (9) Number of uidetified tas 1,8 1,6 1,4 1,2 1, m=64, estimated value m=64, statistical value i simulatio m=128, estimated value m=128, statistical value i simulatio m=256, estimated value m=256 statistical value i simulatio , 1,2 1,4 1,6 1,8 2, Number of tas i read rae, Fi. 6. Compariso of our estimated values with statistical values from simulatio. ad m is 64. So, a loer PN code or a reater umber of slots ca cotribute to the accuracy of our estimati method whe the umber of tas is lare (more tha 1,). Accordi to the estimated umber of tas, we ca adjust the Q value i the ext frame. The Q value is set as IV. Performace Aalysis 1. PN Code Aalysis Ne Q lo 2. (11) Gold codes appear to be oe of the better codes to be used with RFID systems. Assumi that there are Gold codes with l-level liear feedback shift reisters, the these codes ca produce the maximum umber of differet codes as = 2 l. Suppose a ta uses oe of the Gold codes i a slot to sed its ow ID umber. If there are other tas usi the same spreadi code to sed their IDs i this slot, the coflicts will occur. The umber of tas, i, i a slot is biomially distributed as 1 i i 1 Pi () 1. i (12) m m Whe i tas are i a slot, the probability of a ta havi a Gold code that is differet from those of the remaii (i 1) tas ca be computed by i1 1 Pd 1. (13) Let k be the umber of idetified tas from amo i tas i a slot. The, the umber of idetified tas i this slot ca be calculated as follows: 934 Xiaohua Cao ad Tiffay Li ETRI Joural, Volume 37, Number 5, October 215

7 Number of idetified tas i oe slot Number of slots take for idetifyi tas m=8, =1 m=8, =7 m=8, = Number of tas i read rae Fi. 7. Number of idetified tas i sile slot =1 =5 =1, Leth of Gold code, Fi. 8. Number of slots eeded for recoizi differet umbers of tas. i1 1 k ipd i1. (14) Whe i varies from 1 to, the expected value of idetified tas i a slot is ive by 1 1 E( k) k p( i) 1. i1 m m (15) Let E( k) /. The, the maximum umber of idetified tas ca be achieved. We ca obtai the coditio required such that the umber of idetified tas reaches a maximum. It is as follows:. (16) m After substituti (16) ito (14), we ca obtai the expected value of the maximum umber of idetified tas i a sile slot as follows: 1 1 Kmax 1. (17) Whe is lare ad supposed to be ifiite, the expected value of the maximum umber of idetified tas i a sile slot will be 1 1 lim Kmax lim 1 e. 1 (18) Fiure 7 shows that the expected value of the umber of idetified tas i a sile slot is related to the umber of PN codes. Whe the umber of slots is fixed, the loer the PN code is, ad the more the umber of idetified tas there are i a sile slot. The maximum umber of idetified tas i a sile slot is /e; this occurs whe = m. We set the umber of PN codes as 1, 7, 15, 31, 63, 127, ad 255, ad the umber of tas as 1, 5, ad 1,. The aalysis results are show i Fi. 8. The loer a PN code is, the fewer the umber of slots there are for recoizi the ive tas. 2. Collisio Aalysis I our approach, if two or more tas use the same PN code to sed their respective ID i the same slot, the coflicts will occur. The umber of coflicti tas i a frame is ive by 1 1 NC,F me( Nctas ) 1. (19) m Whe the maximum capacity is achieved (that is, = m), the maximum umber of coflicti tas is 1 1 NC,max 1. (2) To compare our ati-collisio approach with Frame Slotted ALOHA, we set the umber of tas as 2,. Fiure 9 shows compariso results o the umber of coflicti tas uder differet combiatios of m ad. If = 1 (equivalet to the EPC C1 G2) whe the frame size is small but the umber of tas is lare, the too may collisios will occur ad the Number of coflicti tas 2, 1,8 1,6 1,4 1,2 1, =1 =7 =15 =31 5 1, 1,5 2, 2,5 3, 3,5 4, Number of slots, m Fi. 9. Compariso results o umber of coflicti tas betwee two ati-collisio approaches. ETRI Joural, Volume 37, Number 5, October 215 Xiaohua Cao ad Tiffay Li 935

8 umber of idetified tas will derade. Whe m is fixed ad is larer tha 1, the the umber of coflicti tas quickly declies with the icrease of. Ulike the Frame Slotted ALOHA, eve if the frame size is small ad the umber of tas is lare, there are t too may collisios if is eouh lare; ad the larer the umber of slots, the smaller the umber of coflicti tas. I additio, whe is small but m is lare eouh, the umber of coflicti tas is the same as the case whe is lare but m is small. So, we ca icrease the umber of slots to reduce the ta collisios due to the hih data rate of the UWB chael. I this way, the limitatio of PN resource ca be solved well. 3. Recoitio Efficiecy The recoitio efficiecy deotes the umber of successfully-idetified tas per uit of time; that is, S S, (21) TF where s is the umber of successfully idetified tas ad T F is the duratio of a sile frame. We set the umber of slots as m = /. The umber of idetified tas i oe frame is 1 1 S 1. The duratio of a sile frame T F ca be calculated by F Q ta ACK (22) T T ( mm ) T m T, (23) where m is the umber of idle slots. If m is lare, the we ca iore the T Q term ad simplify T F as TF m m Tta mt C ( ). (24) The umber of idle slots i a frame is expected by 1 m mp m1. (25) m I our ati-collisio approach, m = /. So, T F is TF 11 Tta 1 TACK. (26) To simplify (26), we set the coefficiet of duratio, α, as T ta /T ACK. The, T F ca be computed by 1 TF 11 1 Tta. (27) If the CDMA Slotted ALOHA i [32] is adopted, the the duratio of oe frame should be calculated by T F ( Tta TACK) STPN. (28) We assume the leth of a PN code to be β times that of ACK; that is, T PN = βt ACK. The duratio of oe frame ca the be calculated by 1 1 T F 1 1 Tta. (29) If the Frame Slotted ALOHA i the EPC C1 G2 stadard is adopted, the the trasmitti speed of dowlik is the same as i our approach; however, the uplik speeds are differet. We assume the uplik speed of our approach to be z times that of EPC C1 G2. The duratio of oe frame should be calculated by F ta ACK S ID T m( zt T ) T, (3) where T ID is the duratio that the reader dowloads ID iformatio. I the Frame Slotted ALOHA, = m. We assume the leth of a ID code to be γ times that of ACK; that is, T ID = γt ACK. The duratio of oe frame, T F, ca the be calculated by 1 1 T F z 1 Tta. (31) From (21), (27), (29), ad (31), we ca derive the recoitio efficiecies of our approach ad the two previous approaches S , Tta S 1 1 1, Tta S z 1 1. Tta (32) (33) (34) Accordi to the systemic hardware ad data size i practice, we let γ = 6 ad α = 1. If we use 7-bit or 15-bit or 31-bit Gold codes, the β would be 1, 1.3, ad 1.7, respectively; correspodily, z would be 1.95,.98, ad.49, respectively. Fiures 1 ad 11 show the ratios of maximum recoitio efficiecies of our approach to that of the other two approaches. From Fi. 1, whe the umber of tas exceeds a certai value, our approach with 7-bit Gold codes or 15-bit Gold codes or 31-bit Gold codes operates 1.4 times, 1.7 times, ad 3 times faster, respectively, tha the CDMA Slotted ALOHA; correspodily, it operates 14.5 times, 16.2 times, ad 18.5 times faster tha the EPC C1 G2 system, respectively. We assume ta data is 512 bits, the uplik speed is 1 Mp/s, ad the dowlik speed is 16 kbps. If is 7, the T ta is.4 ms, T ACK is.1 ms, ad α is 4. If is 31, the T ta is 1.64 ms, T ACK is.1 ms, ad α is 16. It ca be see from Fis. 12 ad 13 that more tha 2, tas ca be processed withi 3 ms i 936 Xiaohua Cao ad Tiffay Li ETRI Joural, Volume 37, Number 5, October 215

9 Ratio of recoitio efficiecies, R =7.5 =15 = , 1,2 1,4 1,6 1,8 2, Number of tas i read rae, Fi. 1. Compariso of maximum recoitio efficiecy with Frame Slotted ALOHA. Ratio of recoitio efficiecies, R =7 2 =15 = , 1,2 1,4 1,6 1,8 2, Number of tas i read rae, Fi. 11. Compariso of maximum recoitio efficiecy with CDMA Slotted ALOHA. Recoitio time,t (ms) =7, our approach =7, CDMA Slotted ALOHA =31,our approach =31,CDMA Slotted ALOHA , 1,2 1,4 1,6 1,8 2, Number of tas i read rae, Fi. 12. Compariso of recoitio time with CDMA Slotted ALOHA. our approach whe is 7; comparatively, CDMA Slotted ALOHA requires 6 ms to achieve the same result. If takes the same value amo the differet approaches, the it is evidet that our approach will eed much less recoitio time Recoitio time, T (ms) 2, 1,8 1,6 1,4 1,2 1, =7, our approach =7, CDMA Slotted ALOHA Frame Slotted ALOHA , 1,2 1,4 1,6 1,8 2, Number of tas i read rae, Fi. 13. Compariso of recoitio time betwee three approaches. tha that of the CDMA Slotted ALOHA approach. I additio, the Frame Slotted ALOHA approach requires 1,95 ms to process 2, tas ad ca recoize about 1, tas i oe secod, which is very much lower tha our approach ad that of CDMA Slotted ALOHA. 4. Complexity Discussio I our scheme, PN codes are stored i the ta memory (for example, 7-bit Gold codes oly occupy 63-bit ta memory) ad the spreadi process ca be achieved by XOR operatio o the biary represetatios of bits ad chips. The UWB trasmitter i a ta does ot eed the modules for backscatteri modulatio ad FM/Miller ecodi, which are ecessary to UHF RFID. Therefore, addi the CDMA UWB uplik will ot reatly icrease a ta s complexity. Sice the proposed ati-collisio approach is implemeted by the reader, the complexity is o the reader side. The reader eeds to have hih computi capability ad hih processi speed to obtai the sial amplitudes ad calculate the iverse of the correlatio matrix. All the detectio alorithms are accomplished by a diital sial processi block, which ca be implemeted i a FPGA device. Moreover, thaks to parallel computi capability, the receiver implemeted i FPGA will work much faster tha ay software implemetatio. Take toether, the proposed ati-collisio scheme has very little impact o the complexity of a overall system, sice a typical RFID system uses a reat umber of tas ad oly a few reader devices. V. Coclusio A dual-bad RFID scheme was preseted ad a feasible structure cosidered. Based o this structure, this paper proposed a efficiet CDMA-based full-duplex ati-collisio ETRI Joural, Volume 37, Number 5, October 215 Xiaohua Cao ad Tiffay Li 937

10 approach. The mai cotributios of this paper ca be summarized as follows: 1) A full-duplex ati-collisio approach has bee proposed. 2) Mathematic models for the performace aalysis of the proposed approach have bee formed. 3) Throuh simulatios, the followi importat coclusios have bee draw: The umber of idetified tas i a sile slot is related to the umber of PN codes. The maximum umber of idetified tas i a slot is /e ad occurs whe = m. Icreasi the umber of slots ca reduce the ta collisios due to the hih data rate of the UWB chael i our approach. More tha 2, tas ca be processed withi 3 ms i our approach but 6 ms i CDMA Slotted ALOHA whe the umber of PN codes is 7. I the future, we pla to evaluate the power cosumptio of a CDMA-based RFID system with full-duplex ati-collisio approach ad compare it aaist both the EPC C1 G2 system ad the CDMA UWB RFID system. Refereces [1] S. Sarma, D. Brock, ad D. Eels, Radio Frequecy Idetificatio ad Electroic Product Code, IEEE Micro., vol. 21, o. 6, Dec. 21, pp [2] P.V. Nikiti ad K.V.S. Rao, Theory ad Measuremet of Backscatteri from RFID Tas, IEEE Ateas Propa. Ma., vol. 48, o. 6, Dec. 26. pp [3] K.C. Shi, S.B. Park, ad G.S. Jo, Ehaced TDMA-Based Ati-collisio Alorithm with a Dyamic Frame Size Adjustmet Stratey for Mobile RFID Readers, Sesors, vol. 9, o. 2, Sept. 29, pp [4] EPClobal, Class 1 Geeratio 2 UHF Air iterface Protocol Stadard V1..9, Ja. 25. [5] B. Zhe, M. Kobayashi, ad M. Shimizu, Framed ALOHA for Multiple RFID Objects Idetificatio, IEICE Tras. Commu., vol. E88-B, o. 3, Mar. 25, pp [6] O. Ba, S. Kim, ad H. Lee, Idetificatio of RFID Tas i Dyamic Framed Slotted ALOHA, It. Cof. Adv. Commu. Techol., Pyeocha, Rep. of Korea, Feb , 29, pp [7] G. Mazurek, RFID System with DS-CDMA Trasmissio, Proc. KKRRiT-26 Cof., Poza, Polad, Jue 7 9, 26, pp [8] A. Loeffler, F. Schuh, ad H. Gerhaeuser, Realizatio of a CDMA-Based RFID System Usi a Semi-active UHF Traspoder, It. Cof. Wireless Mobile Commu., Valecia, Spai, Sept. 2 25, 21, pp [9] E.H. Dia ad B. Jabbari, Spreadi Codes for Direct Sequece CDMA ad Widebad CDMA Cellular Networks, IEEE Commu. Ma., vol. 36, o. 9, Sept. 1998, pp [1] T. Nakaishi ad T. Ikeami, Throuhput Performace of CDMA-ALOHA i S-Bad Lad Mobile Satellite ad Stratospheric Platform Chaels, IEEE It. Symp. Pers., Idoor Mobile Radio Commu., Lodo, UK, Sept , 2, pp [11] I. Opperma et al., UWB Wireless Sesor Networks: UWEN A Practical Example, IEEE Commu. Ma., vol. 42, o. 12, Dec. 24, pp [12] N. Helleputte ad G. Giele, A 7 pj-pulse Aalo Frot-Ed i 13 m CMOS for UWB Impulse Radio Receivers, IEEE J. Solid-State Circuits, vol. 44, o. 7, July 29, pp [13] Z. Zou et al., A Efficiet Passive RFID System for Ubiquitous Idetificatio ad Sesi Usi Impulse UWB Radio, Elektrotechik If. Techik J., vol. 124, o. 11, 27, pp [14] Z. Zou et al., A Low-Power ad Flexible Eery Detectio IR- UWB Receiver for RFID ad Wireless Sesor Networks, IEEE Tras. Circuits Syst., vol. 58, o. 7, July 211, pp [15] K. Mariam et al., Aalytical Study of the Outae Probability of ALOHA ad CSMA i Bouded Ad Hoc Networks, Europea Wireless Cof., Lucca, Italy, Apr , 21, pp [16] L.G. Roberts, ALOHA Packet System with ad without Slots ad Capture, ACM SIGCOMM Comput. Commu. Rev., vol. 5, o. 2, Apr. 1975, pp [17] Y. Mauire ad R. Pappu, A Optimal Q-Alorithm for the ISO 18 6C RFID Protocol, IEEE Tras. Autom. Sci. E., vol. 6, o. 1, Ja. 29, pp [18] P. Pupuwiwat ad B. Static, A RFID Explicit Ta Estimatio Scheme for Dyamic Framed-Slot ALOHA Ati-collisio, It. Cof. Wireless Commu. Netw. Mobile Comput., Chedu, Chia, Sept , 21, pp [19] L.-C. Wa ad H.-C. Liu, A Novel Ati-collisio Alorithm for EPC Ge2 RFID Systems, It. Symp. Wireless Commu. Syst., Valecia, Spai, Sept. 6 8, 26, pp [2] S. Gopala, G.N. Karystios, ad D. Pados, Capacity, Throuhput, ad Delay of Slotted ALOHA DS-CDMA Liks with Adaptive Space-Time Auxiliary-Vector Receivers, IEEE Tras. Wireless Commu., vol. 4, o. 1, Ja. 25, pp [21] A. Sakata et al., Throuhput Compariso of CSMA ad CDMA Slotted ALOHA i Iter-vehicle Commuicatio, It. Cof. ITS Telecommu., Sophia Atipolis, Frace, Jue 6 8, 27, pp [22] R. Gold, Optimal Biary Sequeces for Spread Spectrum Multiplexi (Corresp.), IEEE Tras. If. Theory, vol. 13, o. 4, Oct. 1967, pp [23] G. Mazurek, Active RFID System with Spread-Spectrum Trasmissio, IEEE Tras. Autom. Sci. E., vol. 6, o. 1, Ja. 29, pp [24] C. Mutti ad C. Floerkemeier, CDMA-Based RFID Systems i Dese Scearios: Cocepts ad Challees, IEEE It. Cof. 938 Xiaohua Cao ad Tiffay Li ETRI Joural, Volume 37, Number 5, October 215

11 RFID, Las Veas, NV, USA, Apr , 28, pp [25] Y.F. We et al., Desi of Chipless UWB RFID System Usi a CPW Multi-resoator, IEEE Ateas Propa. Ma., vol. 55, o. 1, Feb. 213, pp [26] R. Ael, L. Atoio, ad G. David, Semi-passive Time-Domai UWB RFID System, IEEE Tras. Microw. Theory Tech., vol. 61, o. 4, Apr. 213, pp [27] S.S. Choi ad S. Kim, A Dyamic Framed Slotted ALOHA Alorithm Usi Collisio Factor for RFID Idetificatio, IEICE Tras. Commu., vol. E92-B, o. 3, Mar. 29, pp [28] M. Grilo et al., Novel Dual-Bad RFID Atea Cofiuratio with Idepedet Tui Adjustmet, Microw. Opt. Techol. Lett., vol. 54, o. 9, Sept. 212, pp [29] J.-P. Curty et al., Remotely Powered Addressable UHF RFID Iterated System, IEEE J. Solid-State Circuits, vol. 4, o. 11, Nov. 25, pp [3] M.B. Nejad et al., A Novel Passive Ta with Asymmetric Wireless Lik for RFID ad WSN Applicatios, IEEE It. Symp. Circuits Syst., New Orleas, LA, USA, May 27 3, 27, pp [31] H. Vot, Efficiet Object Idetificatio with Passive RFID Tas, It. Cof. Pervasive Comput., vol. 2414, Zurich, Switzerlad, Au , 22, pp [32] S. Yu ad P. Yu, The RFID Mechaism Desi Based o CDMA ad Hash Fuctio, It. J. Radio Freq. Idetificatio Techol. Appl., vol. 3, o. 4, Nov. 211, pp Xiaohua Cao received his PhD deree i iformatio techoloy from Wuha Uiversity of Techoloy, Chia, i 28. He is curretly a professor at Wuha Uiversity of Techoloy. His research iterests iclude iformatio techoloy ad loistics plai. He has held or bee holdi more tha te major research projects, such as Natioal Natural Sciece Fud, Support Project of Hubei Sciece ad Techoloy, Retured Overseas Studets Research Fud of Natioal Educatio Miistry, ad Hubei Natural Sciece Fud. Tiffay Li received her PhD deree i electrical eieeri from the Electrical Eieeri Departmet, Texas A&M Uiversity, Collee Statio, USA, i 212. She is a associate professor with teure i electrical ad computer eieeri, Lehih Uiversity, PA, USA. She is a director with the Iformatio ad Commuicatio Research Laboratory, Lehih Uiversity. Her research iterests iclude wireless commuicatios/etworks ad data storae. She was a IEEE Distiuished Lecturer betwee 212 ad 213. ETRI Joural, Volume 37, Number 5, October 215 Xiaohua Cao ad Tiffay Li 939