Passive UWB RFID For Tag Localization Architectures And Design

4IEEE SENSORS JOURNAL, VOL. 16, NO. 1, MARCH 1, 2016with Π(t) denoting the rectangular function of unitary durationfor t [0, 1]. In this way, the polarity of the backscatteredsignal changes at each chip time Tc according to the kth tag’s(k)code value cn , with k T , while the polarity of the wholesymbol composed of Ns pulses is modulated according to the(k)data bit bm { 1} transmitted from the tag to the reader[32].Each interrogation signal is backscattered by the tags aswell as by the surrounding scatterers of the environment.Specifically, the signal at the ith receiver due to the signaltransmitted by the jth transmitter, can be written asTSr̆i,j (t) ri,j(t) ri,j(t) n(t)Tri,j(t)4(3)Sri,j(t)wheredenotes the tags contribution andaccounts for the static contribution (comprehensive of multipath)between transmitter and receiver.5 Finally, n(t) models theadditive white Gaussian noise (AWGN) with one-sided powerspectral density N0 .The tags’ contribution is given by6&'! %(j,k)Tri,j(t) sTj (t) hdown (t) · m(k) (t) h(k,i)up (t) (4)k T(j,k)where hdown (t) is the downlink channel impulse response(CIR), de-embedded of the propagation time, related to the(k,i)link jth transmitter - kth tag, hup (t) is the uplink CIR relatedto the link kth tag - ith receiver.It is possible to rearrange (4) asTri,j(t) p 1Nc 1! N!!(k) (k) Td(j)n cn bm wi,j,k (t mTs nTc τi,j,k )k T m 0 n 0(5)where τi,j,k is the signal TOA, considered with respect to the(j,k)(k,i)Ttransmitter’s clock, and wi,j,k(t) g(t) hdown (t) hup (t)is the channel response to g(t) for the kth tag. Specifically, itis τi,j,k (dTj dRi )/c, where dTj is the distance betweenthe jth transmitter and the tag, dRi is the distance betweenthe tag and the ith receiver,7 and c denotes the speed of light.Note that the round trip backscattering channel is stronglyunfavorable from the energetic point of view, since the received backscattered signal experiences twice the path lossbetween the reader and the tag8 [35]–[37]. In fact, the distancedependence of the received signal power scales, in free space,with the fourth power of the reader-tag distance, resultingin a maximum reading distance much smaller than for anordinary one-hop communication link [38]. It is important tounderline that the tag backscattering behavior is impacted bythe presence of the object on which it is attached. However,differently from standard UHF-RFID, the typical materialswhich constitute the objects where tags are attached to do4 Hereonly the tag’s antenna mode component is included [12].neglect the presence of fast moving objects and tags in the environment, as well as the effects of diffuse clutter [6], [33].6 Operator denotes the convolution.7 Both dTj and dRi are function of the tag index k, however thisdependence has been omitted for notation convenience.8 The effect is more accentuated due to the classical carrier frequency around4 GHz usually adopted for UWB signals [34].5 Wenot detune significantly the tag thanks to the large bandwidthadopted [39].Together with the tag response, the receivers collect thesignal reflected by the surrounding environment, which composes the static contribution (clutter component). Such a staticScontribution ri,j(t) in (3) isSri,j(t) sTj (t) h(j,i)(t)c(6)(j,i)having indicated with hc (t) the CIR of the link betweenthe jth transmitter and the ith receiver, independent of tags’backscattering. As before, (6) can be reformulated asNp 1 Nc 1Sri,j(t) ! !Sd(j)n wi,j (t mTs nTc τi,j )(7)m 0 n 0where τi,j is the signal TOA, considered with respect to the(i,j)Stransmitter’s clock, and wi,j(t) g(t) hc (t) is the channelresponse to g(t) for the static contribution. Specifically it isτi,j di,j /c, where di,j is the distance between the jthtransmitter and the ith receiver.Discrimination between the useful contribution backscattered from the kth tag and those reflected by other tags andthe direct transmitter-to-receiver signal is ensured by a despreading procedure at the receiver (see Fig. 3). Specifically,the separation is enabled by the adoption of different spreadingcodes, since the tag signals are modulated according to the(j)(k)composed code {dn · cn }, while the static contributionindependent of tags’ backscattering is modulated according to(j)the transmitter code {dn } only [6]. In the following, startingfrom the signaling scheme here described, the management ofthe tags in the scenario is described.III. TAG M ANAGEMENTWhen multiple UWB tags based on backscatter modulationare deployed in the environment, several aspects have to beaccounted. Here the solutions developed for tag managementare reported in a unitary form, in order to drive the subsequent discussion about the receiver design and the networkdeployment of Sec. IV and Sec. V.A. Tag SynchronizationTag synchronization is a crucial operation in the UWBRFID systems. In fact, if code generators of tags are completely free running, the reader must perform an exhaustivecode acquisition search in order to synchronize its local codegenerator, used for the de-spreading, with that of the intendeduseful tag to be detected. In addition, the level of readertag synchronization imposes constraints on the tag codesassignment to counteract MUI, as detailed in Sec. III-C.To accomplish this task, an additional narrowband link, forexample in the UHF band or in the 2.4 GHz band, can beused to derive the synchronization signal necessary to wake-uptags, initially in an idle state, and to reset the tags’ spreadingcode generators. The same wake-up signal can also be used toenergize the tag by exploiting energy harvesting techniques,thus making the tag fully passive, that is, energy autonomous.In fact, due to the stringent emission limits on the UWB

N. DECARLI et al.: PASSIVE UWB RFID FOR TAG LOCALIZATION: ARCHITECTURES AND DESIGNCW signalUWB interrogationReaderVoltage - Tag 1SynchThresh.m(1) (t)Thresh.m(2) (t)Tag 1Voltage - Tag 2Tag 2phase is sufficient to accomplish the identification purpose.De-spreading is operated, as described in [32], by accumulating the responses of the Ns pulses that compose a symbol. Inparticular, each response of the channel to a transmitted pulse(j)is multiplied at receiver side by the code element d n of the(k)intended transmitter and the code element c̃n of the intendedtag, and then summed up to the others composing a symbol.(j)(k)Codes {d n } and {c̃n } denote the periodically repeated(k)(k)sequences of period Ns Nc Npc as c̃n ! c⌊n/Npc ⌋ and(j)(k)(k)d n ! dfor n 0, 1, . . . , Ns 1, with c̃ c̃n ,⌊n/Npc ⌋Fig. 5. Wake-up synchronization scheme and signaling.mask, no significant energy can be extracted by the UWBinterrogation signals of readers to power up the tags.Among the various possibilities, an interesting solution isto perform the reset of the tag code generator on the fallingedge of a wake-up continuous wave (CW) carrier received bytags, as shown in Fig. 5. Note that the CW powers up tags,by charging a capacitor via the antenna and a rectifier circuit.Once the CW signal has ended, the falling of the resultingvoltage is used for initiating the backscatter modulation ofthe UWB signals. In this way the propagation-dependentcapacitor charge time9 does not play a significant role inthe synchronization jitter, since the discharge starting eventis not affected by the path loss and depends only on the tags’circuitry. Such solutions have been proposed and experimentedin [19], [40]. The maximum offset after this procedure isexpected to be in the order of a few microseconds. Moreover,if the narrowband link exploited for wake-up is modulatedwith a proper signal (e.g., exploiting amplitude shift keying(ASK)/on-off keying (OOK) modulations), the demodulatedsignal at tag side can serve as reference to lock the local tagoscillator, and helps to prevent the clock drift effects [25].Such clock drifts could be present at tag side due to theexpected poor characteristics of the low-cost oscillators drivingthe backscatter modulator.10 The clock drift determines a timeslide between the modulation signal in the tag and the despreading at the reader that might compromise the correctdetection of the signal. The longer is the duration of the UWBpacket, that is, Np Ts , the more the drift effect is detrimental.B. Tag DetectionOnce tags are synchronized, the UWB reader-tag communication begins, since tags start modulating the transmittedpulses that are successively received by the readers (seeFig. 5).As described in Sec. II-B, tag multiple access is realizedthanks to the assignment of different spreading codes to tags.In this case, since we are in presence of a code divisionmultiple access (CDMA) and each spreading code is uniquelyassigned to a tag, a decision at the output of the de-spreading9 This is due to the narrowband CW signals that may experience selectivelychannels, and to different reader-tag distances.10 An example of ad-hoc oscillator for tags based on UWB backscatteringhas been presented in [41].5n Ns(j)(j)d n Ns d n , where ⌊·⌋ indicates the floor operation. Such ade-spreading operation allows isolating the signals related to aspecific transmitter-tag pair from clutter and interference, andproduces a processing gain counteracting the receiver noise[32].The optimal processing technique requires a matched filter[32] but its implementation is often too complex due tothe large bandwidth so energy-based detection techniques areusually considered [21]. Generically, tag detection is realizedat each reader by checking if the de-spreading output levelrelated to a specific tag code is above a certain threshold. Whenmore than one reader is tuned to the same tag, the decisionon the tag presence can be taken by properly combining thedifferent observations in the central unit. The detection anddemodulation of multiple tags requires the replication of thesame receiver structure, with multiple de-spreaders each tunedto a specific tag code. Alternatively, the same de-spreader canbe re-used for different tags in different interrogation cycles,changing properly the tag code, at the expense of the refreshrate, that is the capability of the system of offering a newreading of a specific tag.It is important to underline that even if tag detection hasbeen properly performed, the receiver has to continue tracka certain tag code to follow the tag clock drift. Note thatin UWB backscattering, the clock drift at tag side does notaffect the TOA of the backscattered pulse (which is determinedonly by the transmitter, supposed to be equipped with a highaccuracy clock), but it affects the pulses modulation operatedby spreading codes. In addition to detection, demodulation of(k)the bits {bm } allows data communication between tags andreaders. In this manner the tag ID can be transferred to thenetwork or, if the tag has embedded sensors, measurementscan be transmitted. Data modulation can be accomplished viabinary phase shift keying (BPSK) modulation of the overallsequence of Ns pulses composing a symbol, as described in[32].In order to have an idea of potential operating ranges,Table I shows the number of pulses Ns required to ensurea theoretical reader-tag distance, in monostatic configuration,when an energy detector receiver is considered, for a signalto-noise ratio (SNR) which let to guarantee a probability offalse alarm 10 3 and a probability of detection 0.9, underthe European mask constraint [21]. In the same table themaximum refresh rate when a packet of 128 bits is transmitted

6IEEE SENSORS JOURNAL, VOL. 16, NO. 1, MARCH 1, 2016TABLE IM INIMUM NUMBER OF PULSES FOR A TARGET READER - TAG READING RANGE AND REFRESH RATE WITH A 128 BIT PACKET TRANSMISSION .Reader-Tag distanceNsRefresh Rate1m610173 Hz2m88694 Hz3m441138 Hz4m139344 HzTABLE IIS IMULATION PARAMETERS .ParameterEIRPBandwidthCenter frequencyPulse intervalReader antenna gainTag antenna gainTag lossesPulses per symbolReceiver noise figureSignal-to-Noise RatioSymbolPTWfcTpGrGtLtNsFSNRValue 11 dBm1.5 GHz4 GHz128 ns5 dBi1 dBi2 dB327684 dB13 dBfrom the tag to the reader is also shown.11 Notice that therefresh rate shown in Table I can be significantly increased if(k)a code cn is uniquely assigned to a tag. In fact, in this case,there is direct correspondence between the UWB spreadingcode and the tag ID, so tag detection is sufficient for providingto the reader the tag identity. Here and in the following ofthe paper, results have been obtained considering the systemparameters reported in Table II. It is important to underlinethat the larger is Ns (i.e., the symbol time), the longer willbe the maximum operating reader-tag distance, thanks to anincreasing processing gain at receiver side. As reported inSec. II-B, the number of pulses per symbol Ns is determinedby both the code length Nc and by the number of pulsesper chip Npc . In general, increasing the number of pulsesper chip, with a fixed PRP, decreases the energy consumptionsince the tag average switching rate between open and shortcircuit is decremented. Furthermore, as described in [21], thishelps on avoiding detrimental effects due to synchronizationmismatches between tags and receivers.C. Multi-Tag InterferenceWith the proposed CDMA scheme, all tags present in themonitored area simultaneously respond to the reader interrogation, without the need of any anti-collision protocol asrequired in the EPC Gen.2 RFID standard [42], [43]. Suchan access technique poses particular constraints in terms ofcode assignment to tags. The behavior of the adopted codes,and the degree of synchronization between readers and tags,determine the amount of interference at the receiver. Suchinterference may produce false alarms during the detectionof a desired tag [25] and worsen the demodulation performance [32]. Moreover, due to the two-hop propagation linkcharacteristic of backscatter propagation, near-far interferenceeffects can drastically affect the performance; in fact, a tag11 The number of bits has been chosen larger than the standard 96 bits of theelectronic product code (EPC) provided by the Gen.2 UHF-RFID to accountfor possible sensor data and parity.5m340018 Hz6m70499 Hz7m130595 Hz8m222773 Hz9m356832 Hz10 m543871 Hzclose to a reader can produce an irreducible interference onthe detector output of a farer tag, preventing the possibility of acorrect detection. An extensive analysis of the code assignmentstrategies for tags and an overview of their design guidelinesis reported in [21]. We want here to stress the importance ofadopting balanced, or quasi-balanced, codes (i.e., codes withzero mean value) that allow the clutter cancellation at receiverside [32], as will be discussed in the next section.A main limitation of the analyzed CDMA scheme is represented by the necessity of providing each tag with a uniquespreading code. Table III reports the number of differentavailable codewords for several code families of interest forthe UWB-RFID system, considering typical codes lengths.As example, a good choice allowing a substantial tradeoff between interference mitigation, complexity and detectionperformance is Nc 128, as shown in [21]. In this case,adopting balanced Gold codes, which represent good candidates, only 65 different tags can be managed. Note that, incase of good reader-tag synchronization, the same spreadingcode can be assigned to several tags with a different initialphase shift, providing that this is greater than the possiblelevel of asynchronism of the system. In this manner thenumber of manageable tags is increased. A new alternativesolution capable of overcoming the limitation on the numberof different tags will be introduced in Sec. VI.D. Clutter RemovalThe RFID-UWB system suffers from clutter that is determined by the environmental response not depending ontag’s backscattering [12]. In monostatic networks the clutterincludes the transmitting-receiving antenna coupling, whichcan be avoided with time-gating operations on the receivedsignal [22], resulting in blind zones around the reader wherethe tag cannot be detected. Another clutter contribution isgiven by the tag’s structural mode, which is the backscatteredcomponent independent of the tag’s antenna load. In thiscase, it has been proposed to separate in the time-domainthe structural and the antenna mode by adopting a delay lineat tag side between the antenna port and the backscatteringmodulator [22], [30]. The last clutter contribution is given bythe response of the environment. The ensemble of all theseSsignals determines the static contribution ri,j(t) at receiverside.The clutter component must be properly canceled in order toprovide robust tag detection. The proposed CDMA approachis intrinsically immune to clutter if balanced codes are adoptedat tag side, regardless of the reader’s code [6], [32]. In fact,with the de-spreading, each static response to a transmitted(k)pulse is multiplied for the tag code cn and accumulated;the balanced tag code makes then null the clutter output

N. DECARLI et al.: PASSIVE UWB RFID FOR TAG LOCALIZATION: ARCHITECTURES AND DESIGN7TABLE IIIN UMBER OF AVAILABLE CODEWORDS FOR DIFFERENT CODE FAMILIES .NcOrthogonal (balanced)M-sequencesGold 865255/25625516-RX components peak amplitude [V]10-1Clutter component(object reflection)10 -2Interference componentfrom the opposite 7possibility of ADC saturation and poor quantization. For thisreason, the signal characteristics expected at the receivingantenna port are now analyzed. Specifically, the received signalis composed of the following components:10 -310 -410 -5Backscattered component from tag(antenna mode)10 -610 -7012345678910Reader-Tag distance [m]Fig. 6. Typical dynamic range of a UWB-RFID system.contribution after the accumulation of the Ns pulses. Alternative solutions for clutter suppression deal with classical radarsignal processing, such as filtering techniques that removethe estimated background response when the tag is absent[22]. The main drawbacks are represented by the necessityof channel estimation and very high-speed signal processing,additionally with severe limitations in terms of tag movement.From the implementation point of view, one of the mainchallenges is the digital removal of the clutter component,whose level is usually several orders of magnitude higherthan signals from tags, as will be studied in Sec. IV-A, thusconditioning the required dynamic range at the receiver.1) The signals backscattered by the tags, related to theinterrogation of a specific transmitter, which the readerwants to demodulate;2) The signals backscattered by the tags related to theinterrogation signal coming from other transmitters (i.e.,an interference component);3) The interference of other transmitters, that is, the ensemble of signals emitted by other transmitters independentof tags’ backscattering;4)

by presenting in a unitary form a possible and effective design of passive UWB-RFID networks, with particular emphasis on system-related aspects.1 The considered UWB-RFID network is composed of readers monitoring an area where tags have to be localized. In particular, tags are semi-passive and based on UWB backscatter modulation, where the low .