978-1-4244-5743-4/10/$26.00 ©2010 IEEE
`
`Abstract— In this paper, we give an overview of spatial
`identification (determining position and velocity) of modulated
`backscatter UHF RFID tags using RF phase information. We
`describe three main techniques based on PDOA (Phase
`Difference of Arrival): TD (Time Domain), FD (Frequency
`Domain), and SD (Spatial Domain). The techniques are
`illustrated with modeling and simulation example in free space
`and in presence of multipath using a multi-ray channel model for
`amplitude and phase of the received tag signal in deterministic
`environment. We also present and discuss the experiments
`performed in a real RFID warehouse portal environment.
`
`I.
`INTRODUCTION
`Real time locating systems (RTLS, also known as
`tracking,
`location sensing,
`location positioning,
`location
`localization, etc.) have always been a subject of strong interest
`in wireless
`industry. Classical wireless approaches
`to
`positioning employ triangulation using received signal strength
`indicator (RSSI) as in cell phones and time difference of arrival
`(TDOA) as in GPS. Good overviews of wireless positioning
`techniques can be found in [1-5]. Source localization has also
`been a subject of interest and research in acoustic [6-7] and
`underwater [8-9] communications, where mechanical instead of
`electromagnetic waves are used [10].
`Some of the classical wireless positioning approaches are in
`general applicable to RFID as discussed in [11-12]. However,
`modulated backscatter UHF RFID is a very short range
`narrowband technology, with typical tag read range on the
`order of 10-20 ft (for passive tags) and maximum available
`bandwidth of 26 MHz (unlicensed ISM band 902-928 MHz in
`US). Because the roundtrip signal delay is on the order of a few
`tens of nanoseconds and the bandwidth is narrow, RFID
`readers and tags can not operate in short pulse mode required
`for TDOA distance determination. Thus common
`tag
`localization techniques use either read/no read criteria (with
`tags as markers [13-14], readers as markers [15-16], or with
`adaptive power control on the reader side [17-18]) or use RSSI
`(Received Signal Strength Indicator) information from the tag
`signal [19-21]. However, RSSI is usually severely affected by
`the propagation environment, the tagged object properties, etc.
`and cannot be universally approximated with free space or
`similar distance-dependent path loss model. Besides, absolute
`calibration of RSSI is rather difficult. Finally, the tag
`backscatter loss varies with the power incident on the tag
`(which itself varies with the tag location) because input
`impedance of RFID tag IC is power dependent.
`
`
`Phase Based Spatial Identification of UHF RFID Tags
`
`Pavel V. Nikitin, Rene Martinez, Shashi Ramamurthy, Hunter Leland, Gary Spiess, and K. V. S. Rao
`Intermec Technologies Corporation
`6001 36th Ave W, Everett, WA, 98203, USA
`{pavel.nikitin, rene.martinez, shashi.ramamurthy, hunter.leland, gary.spiess, kvs.rao}@intermec.com
`
`102
`IEEE RFID 2010
`
`Fortunately, RFID readers perform fully coherent
`detection and recover the baseband phase of the coherently
`demodulated tag signal even though this information may not
`be visible or made available to the users. Thus, tag phase
`information is becoming increasingly important for such
`spatial
`identification
`advanced RFID applications as
`(determining tag position and velocity). We prefer to use this
`term because it suggests that one can not only obtain tag data
`information but also obtain additional information about tag
`spatial characteristics.
`the
`tag signal depends both on
`Phase of
`the
`propagation channel and the modulating properties of the tag,
`which can be both frequency- and power-dependent. However,
`all these factors are additive (rather than multiplicative) and in
`many cases can be calibrated out if phase difference of arrival
`(PDOA) is used. In this paper, we present an overview of phase
`based spatial identification of RFID tags. Some of the
`equations and terminology previously appeared in the cited
`references, but never were summarized together in application
`to modulated backscatter UHF RFID. To illustrate all three
`main techniques based on phase difference of arrival (TD-
`PDOA, FD-PDOA, and SD-PDOA), we show simulation
`results for a simple modeling example of a tag moving by a
`pair of reader antennas in the presence of multipath. We also
`experimentally demonstrate TD-PDOA technique using data
`collected on real Gen2 tags using our custom reader capable of
`accurately measuring RSSI and phase of the tag signal at
`different frequency channels. The data was collected in one of
`the most practical RFID environments – a warehouse portal.
`
`
`
`II. TAG PHASE
`Most RFID readers perform fully coherent detection and can
`measure both the power and the phase of the tag signal. Tag
`phase can be best explained in phasor space (I-Q plane of the
`received baseband voltage) shown in Fig. 1. The complex
`demodulated voltage at the reader receiver at any given
`moment of time can be written as sum of three components:
`r
`r
`r
`v
`VV
`V
`V
`i
`=
`tag
`clutter
`leakage
`
` , (1)
`
`+
`
`+
`
`Vr
` is the voltage due to the reader transmit-receive
`where
`leakage
`leakage (including reflection from the mismatched reader
`Vr
`is the voltage due to the scatter from the static
`antenna),
`clutter
`tagVr
`i
`environment clutter; and
` is the voltage due to the
`backscatter from the tag when the tag IC (chip) is in state i.
`
`RFC - Exhibit 1017
`
`

`
`103
`
`These components can be assumed stationary during
`
`single tag interrogation. For example, if the reader reads 1000
`tags/s and the tag moves at 36 km/h (10 m/s), then during 1 ms
`that it takes to read the tag, the tag moves only by 1 cm which
`does not significantly affect propagation channel properties. All
`three components and their sources are shown in Fig. 1 and 2.
`
`
`Fig. 1. Complex demodulated voltage received by the reader.
`
`
`
`. (2)
`
`Fig. 2. Sources contributing to the complex demodulated voltage
`received by the monostatic reader: leakage, clutter, and tag.
`At the reader, the in-phase (I) and quadrature (Q) components
`of the received and demodulated tag signal are composed of
`DC and AC parts, as shown in Fig. 1:
`I
`I
`I
`QQ,
`Q
`+
`=
`+
`=
`ac
`dc
`ac
`dc
`The DC parts are due (in the diminishing order of importance)
`to the reader transmit–receive leakage, static environment
`clutter, and backscatter from the tag (which contains both static
`and modulated components). The reference point commonly
`used by readers is mid-point between tag constellation points
`(states 1 and 2 in Fig. 1).
`After the DC part is filtered out, the tag constellation is
`centered at zero and one can measure both RSSI and phase of
`the received tag signal as:
`r
`V
`2
`−
`tag
`Z
`
`+
`Z
`
`o
`
`2a
`
`c
`
`I
`
`=
`
`2
`
`r
`V
`1
`tag
`
`o
`
`21
`
`
`
`RSSI
`
`=
`
` (3)
`
`,
`
`2a
`
`c
`
`Q
`
`IQ
`
`=
`ϕ
`
`ang
`
`(
`r
`V
`2
`tag
`
`−
`
`r
`V
`1
`tag
`
`)
`
`=
`
`arctan
`
`ac
`where Z0 is the input impedance of the receiver (50 Ω).
`
`ac
`
` (4)
`
`For illustration, a typical tag signal seen at the reader (after
`the DC part removal) is shown in Fig. 3, both in IQ plane (tag
`signal constellation) and in time domain.
`
`Tag signal constellation
`
`30
`
`20
`
`10
`
`0
`
`-10
`
`-20
`
`-30
`
`Q (mV)
`
`10
`
`20
`
`30
`
`0
`I (mV)
`
`I
`
`Q
`
`-30
`
`-20
`
`-10
`
`30
`20
`10
`0
`-10
`-20
`-30
`
`Voltage (mV)
`
`0
`
`100
`
`200
`Samples
`
`300
`
`400
`
`
`
`Fig. 3. Typical tag signal received at the reader.
`In any propagation environment, the phase of the
`received tag signal can be written as:
`
`
`
` ϕϕϕϕ= ++
`, (5)
`prop
`o
`BS
`where
`the phase accumulated due
`to
`the
`propϕ is
`oϕ is the phase offset
`electromagnetic wave propagation,
`which includes phases of the cables and other reader and
`BSϕ is the backscatter phase of the
`antenna components, and
`tag modulation.
`In free space, using classical phasor formula for
`electromagnetic field propagating in free space (whose phase is
`proportional to the distance travelled), one can write:
`kd2
` , (6)
`−=ϕ
`prop
`
` c/f2k π=
`
`where
` is the wavevector (proportional to the
`frequency) and d is the distance to the tag, One can see that the
`phase given by equation 6 linearly varies with the distance to
`the tag. When the tag is moved away or towards the reader (in
`tagVr
`tagVr
`1
`2
`free space), both vectors
` and
` rotate simultaneously,
`making full rotation and causing 360 degrees tag signal phase
`2/λ of radial tag movement.
`change for every
`
`

`
`104
`
`III. TECHNIQUES
`
`A. TD-PDOA (Time Domain Phase Difference of Arrival)
`TD-PDOA allows one to estimate the projection of the tag
`velocity vector on to the line of sight between the tag and the
`reader by measuring tag phases at different time moments. It
`can be viewed as a form of measuring Doppler shift to
`determine the speed of the mobile node [22-24]. The technique
`is illustrated in Fig. 4, where the tag moves with a constant
`speed some distance away from the reader antenna.
`
`
`
`
`
`∂∂
`
`Fig. 4. TD-PDOA illustration.
`
`
`By measuring the phase of the tag signal at two different
`time moments (at the fixed frequency), assuming that other
`two components of the tag phase (phase offset and tag
`backscatter phase) do not change, and taking the derivative of
`the phase with respect to time, we can calculate tag radial
`velocity projection as:
`c
`ϕ
`Vr
` . (7)
`−=
`f4
`t
`π
`A theoretical application of this technique to RFID recently
`appeared in [25], where network analyzer was used for
`channel measurements. Note that the velocity calculated from
`equation 7 is instantaneous tag velocity, which may change
`with time (if the tag accelerates/decelerates).
`B. FD-PDOA (Frequency. Domain Phase Difference of Arrival)
`FD-PDOA allows one to estimate the distance to the tag by
`measuring tag phase at different frequencies. It can be viewed
`as a form of ranging using well known frequency modulated
`continuous wave (FM CW) radar [26-27], or similar harmonic
`[28-31] radar. The technique is illustrated in Figure 5.
`
`∂∂
`
`Fig. 5. FD-PDOA illustration.
`
`
`By measuring the phase of the tag signal at several (two or
`more) frequencies, taking the derivative of the phase with
`respect to frequency, and assuming that other two components
`of the tag phase (phase offset and tag backscatter phase) do
`not change with frequency or can be calibrated out, and the tag
`has not moved much (much less than wavelength) during the
`measurements, we can calculate the range to the tag as:
`c
`ϕ
`d
` . (8)
`−=
`f
`4
`π
`
`Note that by using several different reader antennas and
`applying FD-PDOA to each of them, one can localize the tag
`in three dimensions, like it is done in FM CW radars
`[27].Various applications of this technique to RFID in free-
`space environment recently appeared in [32-38]. Since the FD-
`PDOA technique is similar to FM CW radar, it can work for
`both moving and stationary tags.
`C. SD-PDOA (Spatial Domain Phase Difference of Arrival)
`SD-PDOA allows one to estimate the bearing (direction to
`the tag), or the angle of arrival, by measuring phases of the tag
`signal at several receiving antennas. It can be viewed as a form
`of direction-of-arrival estimation using phased array antenna
`[39-40]. Many signal processing
`techniques have been
`developed in this field to improve angle estimation accuracy
`[41-42]. The technique is illustrated in Fig. 6 for the bistatic
`reader configuration (separate transmit and receive antennas).
`
`
`
`Fig. 6. SD-PDOA illustration.
`2 ϕϕ − 1
` of the received
`By measuring the phase difference
`
`tag signal at two different receiving antennas (at the fixed
`frequency channel) and attributing
`it
`to
`the path
`d
`d −
`difference
`, we can approximately calculate two-
`2
`1
`dimensional tag bearing as:
`
`⎥⎦⎤
`
`)
`
` , (9)
`
`(
`c
`2 ϕϕ 1
`−
`f2
`a
`π
`where a is the spacing between the two receiving antennas.
`Transmitted antenna can be located anywhere (the phase offset
`can be calibrated out). Phase measurements on antennas 1 and
`2 can either be done simultaneously (in this case, RFID reader
`needs to have 2-channel receiver) or sequentially (the tag can
`be interrogated several times, while the receiving antennas are
`switched between tag queries).
`Equation 9 assumes that the tag is far so that simple
`trigonometry can be used. In general, a set of hyperbolic
`equations can be used for localization, like it is done in GPS
`[43]. Note that many variations of SD-PDOA are possible in
`RFID. For example, an array of several receiving antennas can
`be used to locate tag in three dimensions [44-45]. Other
`variations can be found in several issued or pending patents
`[46-48]. Reader antennas can also be used in monostatic mode
`(simultaneous transmit/receive). If monostatic configuration is
`used, equation 9 needs to be modified (the denominator in this
`case becomes π4 instead of π2 ). The bistatic configuration is
`preferable in SD-PDOA because in all received tag responses,
`the tag is powered in the same way by separate high gain
`transmit antenna. Also, receive antennas do not need to have
`high gain and can be made small and omnidirectional.
`
`
`
`−
`
`⎢⎣⎡
`
`θ
`
`≈
`
`arcsin
`
`

`
`105
`
`IV.
`
`MODELING AND SIMULATION EXAMPLE
`
`A. Geometry
`Consider the following scenario with the geometry shown in
`Figure 7. Two reader antennas are mounted at the same height
`above the ground (shown in yellow color), and the tag is
`traversing with a constant speed V along the x-axis parallel to
`the infinite wall (shown in green color). The reader antenna is
`directional patch and the tag is a dipole oriented along the z-
`axis (radiation patterns are shown in blue and pink color
`respectively). The following parameters are used: V=5 mph,
`h
`tagh =1.2 m,
`b=1.4 m, d=1.3 m,
`=1.1 m, a=0.5 m.
`
`reader
`
`
`
`
`
`Ground
`
`WallWall reflection
`
`Top view
`
`Tag
`
`Wall
`
`y
`
`x
`
`z
`
`2
`
`V
`
`1
`
`V
`
`2
`
`d 1
`
`V
`
`d 2
`
`1
`
`d
`
`b
`
`Reader antennas
`
`Ant 1
`
`a
`
`Ant 2
`
`Fig. 7. Geometry of the modeling and simulation example.
`
`B. Channel model
`There exist various radio propagation environments. The real
`propagation environment (the most interesting case) is never
`free-space. Modeling methods and models
`for many
`propagation environments are well known in literature [49-55].
`The unique feature of UHF RFID is its short range and the fact
`that it is based on modulated backscatter. To accurately model
`RFID channel, both forward and reverse links need to be
`considered at the same time. Essentially, RFID channel is a
`double fading channel: each fade is experienced twice, in
`forward and reverse link [56]. A typical RFID use case
`scenario involves indoor multipath environment (for example,
`a warehouse) with the line-of-sight and very few major
`reflections. Such environment
`is used
`in our example.
`Normally, signal received in wireless environment where the
`distance to the transmitter is large and there are many
`independent scatterers can be modeled statistically. However,
`
`as mentioned before, passive UHF RFID is a short range
`technology (even though current passive tags may have range
`of 50 ft, most practical use case scenarios involve tag reading at
`5 to 20 ft distances). While in some cases well known
`theoretical distributions (Rician, Rayleigh, Nakagami, log-
`normal) can be used to fit the data [57-59], in general the signal
`distribution is primarily shaped by the specific geometry and
`tag trajectories. The multipath in such short range environment
`is highly deterministic and strongly depends on particular
`arrangement of nearby reflectors. While numerical 3D EM
`modeling tools can be applied to such environments and can be
`especially useful for analyzing complicated cases where
`various tagged objects are present, such tools normally require
`significant computation time due to the large problem size
`relative to the wavelength. Below we describe a deterministic
`multi-ray model that allows one to calculate the RSSI and
`phase of the tag signal received at the reader. Because the
`described model is essentially given by closed-form equations,
`it can be easily implemented (in Excel or Matlab) and used for
`system analysis and in RFID network simulators [60].
`The path gain of the deterministic multipath channel
`between the reader and the tag can be written as:
`2
`
`2
`
`, (10)
`
`H
`
`⎠⎞
`
`⎟⎟
`
`o
`
`πλ
`
`d4
`
`⎝⎛
`
`⎜⎜
`
`G
`
`path
`
`=
`
`−
`
`d(jk
`
`−
`
`)d
`o
`
`i
`
`e
`
` , (11)
`
`io
`dd
`
`i
`
`Γ
`i
`
`gg
`
`it
`
`N
`
`∑=
`
`1i
`
`1H
`+=
`
`ig are the
`g and
`od is the length of the direct ray path,
`where
`angle-dependent normalized signal gains of the reader and the
`iΓ is the angle-dependent reflection coefficient
`tag antennas,
`id is the length of the i-th
`of the i-th reflecting object,
`reflected ray path, and N is the total number of reflections. This
`multi-ray model is similar to the models described in [61-62]
`for short range scenarios where other wireless technologies are
`used for highway applications and where the size of reflectors
`is large compared to wavelength. The complex factor H can be
`interpreted as the channel response due solely to multipath. In
`free space environment, there is no multipath (H=1), and the
`channel path gain is log-linear, 20 dB/decade. In our example,
`there are two reflections: from the wall and from the ground.
`The power of the tag signal received by the monostatic reader
`can be written as:
`
`it
`
`RSSI
`GGP
`K
`2
`t=
`t
`ath
`tG is the gain of the
`tP is the output power of the reader,
`where
`monostatic reader antenna and K is the tag backscatter gain
`(correspondingly, -K is the tag backscatter loss) which defines
`how much of the incident RF power is converted to the
`backscattered modulated power. The phase of the tag signal
`received at the reader is given by equation 5 where the phase
`due to the roundtrip propagation channel can be written as:
`kd2
`)Harg(
`2
`. (13)
`−=ϕ
`+
`prop
`
`. (12)
`
`2p
`
`o
`
`

`
`106
`
`. (15)
`
`C. Results
`Figures 8 and 9 show the RSSI and the phase of the tag
`signal which would be measured at the receiving antenna 2 at
`915 MHz as functions of tag position along the x-axis. On
`antenna 1, one would see the same curves but shifted by 0.5 m
`(antenna 1 is to the left of antenna 2). One can see that in this
`example where directional antennas are used, the reflection
`from the ground is small, and the reflection that influences the
`RSSI of the tag the most is coming from the wall. Figures 10,
`11, and 12 give the tag radial speed projection V2, the range to
`the tag d2 (using two frequencies, 914 MHz and 915 MHz), and the
`bearing to the tag α, all obtained from the tag phase using
`three PDOA techniques described above (TD, FD, and SD).
`One can see that the multipath has very strong effect on the
`tag speed, range, and bearing calculated with PDOA. The most
`robust technique is TD-PDOA, which even in the presence of
`multipath allows one to find the direction of the tag movement
`(ingress/egress) and the center crossing (when the tag passes by
`the reader).
`
`To make the problem tractable, we make several simplifying
`assumptions listed below. Both ground and wall are modeled as
`flat sheets of perfect conductors (the general expressions of
`angle-dependent
`reflection
`coefficients
`for
`non-ideal
`conductors are well known and can be found for example in
`[55]). Only single reflections from either reflector (the ground
`or the wall) are considered. The vertically oriented tag antenna
`is assumed to have the radiation pattern of an ideal half-
`wavelength dipole with 2 dBi gain, approximated based on
`well known angle-dependent expression [63] as:
`, (14)
`(g
`sin
`)
`,
`3
`φθ
`=
`θ
`whereθis the angle in the vertical plane (see Fig. 7, side view).
`The normalized angle-dependent reader antenna pattern in both
`vertical and horizontal planes iss approximated with the
`analytical function based on antenna half-power beamwidths:
`
`⎥⎦⎤
`⎠⎞
`
`⎟⎟
`
`⎝⎛
`
`⎢⎣⎡
`
`cos
`
`⎥⎦⎤
`⎠⎞
`
`⎟⎟
`
`⎝⎛
`
`⎢⎣⎡
`
`(g
`r
`
`,
`)
`φθ
`
`=
`
`cos
`
`θππ
`φππ
`sin
`sin
`⎜⎜
`⎜⎜
`6
`2
`6
`2
`θ
`φ
`dB3
`dB3
`whereϕis the angle in the horizontal plane (see Fig. 7, top
`view). We assume for simplicity that the phase patterns of both
`tag and reader antennas are isotropic (independent of angle).
`Both horizontal and vertical beamwidths of the circularly
`polarized reader antenna are 60 degrees, and the horizontal and
`vertical linear gains are 6 dBil. The tag sensitivity is -12 dBm,
`and the tag backscatter loss (at threshold) is 10 dB, increasing
`linearly (1 dB/dB) with increasing incident power. The tag
`backscatter phase is assumed constant (independent of power).
`These assumptions are based on measurements described in the
`next section. The reference point is chosen so that
`.
`0
`+ϕϕ
`=
`o
`BS
`The reader output power is 30 dBm and the frequency is 915
`MHz. For every tag position, the paths and angles of three rays:
`direct, ground reflection, and wall reflection are computed and
`used in equations 10 and 11 to calculate the RSSI and the phase
`of the tag at the reader receiver using equations 12 and 13. The
`model was experimentally verified and agreed well with the
`data (tag RSSI and phase) collected in the portal environment.
`
`Free space
`With ground
`With ground and wall
`
`-40
`
`-45
`
`-50
`
`-55
`
`-60
`
`-65
`
`RSSI (dBm) a
`
`
`
`
`
`-3
`
`-2
`
`1
`0
`-1
`Tag position (m)
`
`2
`
`3
`
`Fig. 8. RSSI of the tag signal received on antenna 2 at 915 MHz
`
`Free space
`With ground
`With ground and wall
`
`180
`135
`90
`45
`0
`-45
`-90
`-135
`-180
`
`Phase (deg) a
`
`-3
`
`-2
`
`1
`0
`-1
`Tag position (m)
`
`2
`
`3
`
`Fig. 9. Phase of the tag signal received on antenna 2 at 915 MHz.
`
`Free space
`With ground
`With ground and wall
`
`-3
`
`-2
`
`2.5
`1.5
`
`0.5
`-0.5
`
`-1.5
`-2.5
`
`Speed (m/s) a
`
`
`Fig. 10. Tag radial speed relative to ant. 2 obtained with TD-PDOA.
`
`1
`0
`-1
`Tag position (m)
`
`2
`
`3
`
`Free space
`With ground
`With ground and wall
`
`01234567
`
`Range (m)
`
`-3
`
`-2
`
`
`Fig. 11. Range to the tag from ant. 2, obtained with FD-PDOA.
`
`0
`-1
`Tag position (m)
`
`1
`
`2
`
`3
`
`Free space
`With ground
`With ground and wall
`
`90
`
`45
`
`0
`
`-45
`
`-90
`
`Bearing (deg)
`
`-3
`
`-2
`
`1
`0
`-1
`Tag position (m)
`
`2
`
`3
`
`
`
`Fig. 12. Bearing to the tag α obtained with SD-PDOA.
`
`

`
`107
`
`V.
`EXPERIMENTS
`Experiments on using phase based techniques to
`
`determine tag range, etc. in free space environment have been
`described by various researchers (see e.g. [36]). In this paper,
`we would
`like
`to discuss
`the real challenging RFID
`environment with strong multipath. To illustrate one of the
`aforementioned approaches (TD-PDOA) in such practical
`scenario, we performed tag RSSI and phase measurements in a
`standard warehouse portal (a gate through which tagged goods
`travel and which is equipped with antennas connected to RFID
`reader). The tag RSSI and phase characteristics in the portal
`strongly depend on the geometry of the portal and its antennas,
`the characteristics of the tags, their relative arrangement and
`trajectories, the type of tagged goods, etc. This is a good
`example of challenging RFID environment. Tag RSSI in portal
`has recently been analyzed in [64] using battery-assisted tag
`device which recorded signal strength. In our measurements,
`we used our own custom reader hardware which was capable
`of accurately measuring both RSSI and phase of the received
`tag signal. Below, we concentrate on time domain phase
`measurements to show how TD-PDOA can be utilized in
`practice.
`
`
`Fig. 13. Experimental RFID portal.
`We performed our measurements
`in
`the experimental
`warehouse portal photographed
`in Figure 13, with
`the
`dimensions and the coordinate system as shown. Four antennas
`were connected to one monostatic RFID reader with four
`sequentially switched antenna ports. A remote controlled robot
`carried the styrofoam column with a cardboard piece and RFID
`tags attached and moved along one of the lines 0-6 (along x-
`axis), traversing the portal and periodically stopping to allow
`the reader to take measurements of the RSSI and phase of the
`received tag signal on all frequency channels in FCC ISM band
`(902-928 MHz) from all four antennas. The reader transmit
`power was constant (30 dBm). A large number of tag RSSI and
`phase data was collected (6 lanes, 4 reader antennas, >20 tags,
`>200 robot steps, >50 frequency channels).
`
`The portal was made of metal, and the warehouse floor was
`re-barred concrete. The tags were Avery Dennison dipole-like
`AD-222 Gen2 tags [65], with Impinj Monza 2 chip. The tags
`were mounted both horizontally and vertically. The reader
`antennas were circularly polarized Huber & Suhner RFID
`antennas (model number SPA 915/60/10/0/RCP [66]), with the
`maximum gain 10 dBi, axial ratio 2.5 dB, horizontal and
`vertical beamwidths 45 and 60 degrees (the antennas were
`rotated 90 degrees to minimize coverage outside the portal).
`Because the tag received different power at different points
`inside portal volume, we needed to know how tag backscatter
`phase behaved as a function of incident power. We measured it
`separately, with the variable attenuator connected to the
`monostatic reader output and the tag placed (first as is, then on
`cardboard) in anechoic chamber at the distance of d=3 ft from
`the transmitting antenna (which was linearly polarized, with 6
`dBi gain) as shown in Fig. 14. The reader power was constant
`(30 dBm), and the attenuator allowed to change the output
`power in 1 dB steps. The phase offset due to the attenuator was
`separately measured at each attenuation level with the network
`analyzer and taken into account in measurements.
`
`
`Fig. 14. Experimental setup for tag backscatter characterization.
`Figure 15 shows the measured backscatter gain and the
`backscatter phase as functions of power incident on the tag.
`Note that since only the relative phase change is important, any
`reference point can be chosen. We chose the reference point so
`that the tag phase is zero at threshold power when the tag is
`measured as is. The measured threshold backscatter loss for the
`AD-222 tag on cardboard (when incident power was equal to
`threshold tag sensitivity) was about -10 dB. Measurements in
`Figure 15 are given
`for one
`frequency, 915 MHz.
`Measurements at other frequencies showed similar results and
`confirmed that the tag phase variation due to the varying power
`of the incident signal (less than ten degrees) is small compared
`to phase changes due to tag motion (several hundred degrees).
`
`As is
`On cardboard
`
`0
`-4
`Power (dBm)
`
`4
`
`8
`
`-12
`
`-8
`
`90
`
`60
`
`30
`
`0
`
`-30
`
`-60
`
`-90
`
`Backscatter phase
`
`As is
`On cardboard
`
`-8
`
`-12
`
`-16
`
`-20
`
`-24
`
`-28
`
`-32
`
`Backscatter gain
`
`-12
`
`-8
`
`0
`-4
`Power (dBm)
`
`4
`
`8
`
`
`
`
`Fig. 15. Backscatter gain and phase as functions of power incident on
`the tag at 915 MHz in free space (as is and on cardboard).
`
`

`
`108
`
`We performed measurements for all tags traversing our portal
`at different heights at all six lanes. A typical measured tag
`phase when the tag traverses through our portal is shown in
`Fig. 16 for the case when the tag is at z=0.93 m above the
`ground, it travels on lane 3 (y=-0.1 m) and the signal is
`received on one of the lower reader antennas (y=-1.45 m) at
`902 MHz. For some tag positions (when tag enters/leaves
`portal), the tag was not powered up, and hence the phase value
`was not available as can be seen from Fig. 17.
`
`180
`
`90
`
`0
`
`-90
`
`-180
`
`Phase (deg)
`
`-2
`
`-1.5
`
`-1
`
`0.5
`0
`-0.5
`Tag position (m)
`
`1
`
`1.5
`
`2
`
`
`
`Fig. 16. Measured phase of the tag traversing through the portal.
`
`Center crossing
`
`Tag is not powered
`
`0.4
`0.3
`0.2
`0.1
`0
`-0.1
`-0.2
`-0.3
`-0.4
`
`Tag speed (m/s)
`
`0.5
`0
`-2 -1.5 -1 -0.5
`Tag position (m)
`
`1
`
`1.5
`
`2
`
`
`Fig. 17. Radial speed of the tag traversing the portal: calculated from
`data with TD-PDOA and the actual (zero speed means tag is not read).
`An application of TD-PDOA technique to the measured
`phase data is shown in Fig. 17 and allows one to calculate the
`radial speed projection of the tag. As one can see, the
`ingress/egress direction of the tag movement (radial speed
`positive or negative) and the point when the tag crosses the
`center (radial speed is zero) can easily be identified.
`The FD-PDOA experimental results with the portal were
`similar to the results presented in Fig. 11 for the modeling and
`simulation example: the multipath had very strong effect on
`ranging with FD-PDOA. As for SD-PDOA, our experimental
`data was not well suited to apply this technique because the
`four-antenna reader was monostatic and the pairs of antennas
`mounted on each side of standard portal were arranged
`vertically instead of horizontally.
`We have also performed a series of experiments in anechoic
`chamber where we observed that both FD-PDOA and SD-
`PDOA can work really well for UHF RFID in the absence of
`reflections. Similar results have been noticed by other
`researchers (see e.g. [36]).
`
`VI.
`CONCLUSIONS
`In this paper, we gave an overview of phase based
`
`spatial identification of modulated backscatter tags. We
`described three main approaches (TD-PDOA, FD-PDOA, and
`SD-PDOA), presented a modeling and simulation example and
`analyzed experimental data collected in real multipath RFID
`scenario: a warehouse portal. We also described
`the
`deterministic multipath channel model for UHF RFID systems
`which was used in our simulations.
`It was observed both in simulations and in experiments that
`TD-PDOA technique was fairly robust to multipath. The
`ingress/egress direction of the tag movement (radial speed
`positive or negative) and the point when the tag crosses the
`center (radial speed is zero) could easily be identified.
`The real challenge is to make both FD-PDOA and SD-
`PDOA techniques work reliably in an arbitrary environment,
`where location of short range multipath sources is unknown.
`The fact that the available bandwidth is small (<30 MHz in
`UHF ISM band) makes this problem especially difficult. This
`is a subject of current and future research.
`
`REFERENCES
`[1] H. Liu et al., “Survey of Wireless Indoor Positioning Techniques and
`Systems”, IEEE Transactions on Systems, Man, and Cybernetics, vol.
`37, no. 6, Nov. 2007, pp. 1067 - 1080
`[2] M. Vossiek et al., “Wireless local positioning”, IEEE Microwave
`Magazine, vol. 4, Issue 4, Dec. 2003, pp. 77 – 86
`[3] G. Yanying, A. Lo, L. Niemegeers, “A survey of indoor positioning
`systems for wireless personal networks”, IEEE Communications
`Surveys & Tutorials, vol. 11, no. 1, 2009, pp. 13 – 32
`[4] P. Gulden, S. Roehr, M. Christmann, “An overview of wireless local
`positioning system configurations”, IEEE Int. Microwave Workshop on
`Wireless Sensing, Local Positioning, and RFID, Sept. 2009, pp. 1 – 4
`[5] K. Pahlavan, L. Xinrong J. P. Makela, “Indoor geolocation science and
`technology”, IEEE Communications Magazine, vol. 40, no. 2, Feb.
`2002, pp. 112 - 118
`[6] H. Schau, A. Robinson, “Passive source localization employing
`intersecting spherical surfaces from time-of-arrival differences”, IEEE
`Transactions on Acoustics, Speech and Signal Processing, vol. 35, no.
`8, Aug 1987, pp. 1223 – 1225
`[7] K. Frampton, “Acoustic self-localization
`in a distributed sensor
`network”, IEEE Sensors Journal, vol. 6, no. 1, Feb. 2006, pp. 166 – 172
`[8] B. Friedlander, “A Passive Localization Algorithm and Its Accurancy
`Analysis”, IEEE Journal of Oceanic Engineering, vol. 12, no. 1, Jan
`1987, pp. 234 – 245
`J. Smith, J. Abel, “The spherical interpolation method of source
`localization”, IEEE Journal of Oceanic Engineering, vol. 12, no. 1, Jan
`1987, pp. 246 – 252
`[10] M. Fowler, H. Xi, “Signal models for TDOA/FDOA estimation”, IEEE
`Transactions on Aerospace and Electronic Systems, vol. 44, no. 4, Oct.
`2008, pp. 1543 – 1550
`[11] M. Bouet, A. L. dos Santos, “RFID tags: Positioning principles and
`localization techniques”, 1st IFIP Wireless Days, Nov. 2008, pp.1 – 5
`[12] J. Zhou, J. Shi, “RFID localization algorithms and applications—a
`review”, Journal of Intelligent Manufacturing, vol. 20, no. 6, December,
`2009, pp. 695-707
`[13] C. Yu, R. Liu, “Application of RF Tags in Highway Reference
`Markers”, International IEEE Conference on Intelligent Transportation
`Systems, Oct. 2008, pp. 464 – 469
`[14] Youngsu Park et al., “Improving position estimation on RFID tag floor
`localization using RFID reader transmission power control”, IEEE Int.
`Conference on Robotics and Biometrics, Feb. 2009, pp. 1716 - 1721
`
`[9]
`
`

`
`109
`
`[15] E. Bruns et al., “Enabling Mobile Phones To Support Large-Scale
`Museum Guidance”, IEEE Multimedia, vol. 14, no. 2, April-June 2007,
`pp. 16 - 25
`[16] A. Reza, T. Geok, “Investigation of Indoor Location Sensing via RFID
`Re