System and method for estimating range to an RFID tag

Abstract

The a system for measuring distance between an RFID reader and tag, including an adaptive linear combiner, which is a tapped delay line with controllable weights on each tap, and outputs that are summed and subtracted from a reference to produce an error signal. After a sufficient number of cycles, the weight distribution indicates the delay of the received signal with respect to the reference, and by extension determines the distance between the tag and receiver.

Claims

1. A system for determining distance from a plurality of antennas to a backscatter RFID tag comprising: a transmitter for transmitting a transmitted modulated signal to the tag via first and second antennas, a first receiver for receiving a first modulated backscattered signal having a first delay from the tag via said first antenna; a second receiver for receiving a second modulated backscattered signal having a second delay from the tag via said second antenna; a first linear adaptive combiner which is a delay line having first delay line taps with first controllable weights on each first delay line tap, and outputs that are summed and subtracted from a reference to produce a first error signal; and a second linear adaptive combiner which is a delay line having second delay line taps with second controllable weights on each second delay line tap, and outputs that are summed and subtracted from said reference to produce a second error signal; wherein modulation in the backscattered signal received by said first antenna is input to said first adaptive linear combiner, and said transmitted modulated signal is said reference to said first adaptive linear combiner; wherein modulation in the backscattered signal received by said second antenna is input to said second adaptive linear combiner, and said transmitted modulated signal is said reference to said second adaptive linear combiner; wherein the first and second controllable weights for said taps of said first and second linear adaptive combiners are approximately zero for all of said taps except for respective first and second combiner taps corresponding to the first and second delays of the backscattered signal; and wherein the distance from the RFID tag to each of said antennas is determined from said first and second delays.

2. The system of claim 1, wherein said first and second antennas are mounted in a toll plaza and the tag in a vehicle.

3. The system of claim 1, wherein said first and second linear adaptive combiners operate at a plurality of sample rates and said distance is determined from a plurality of results obtained at said plurality of sample rates.

4. The system of claim 3, wherein said distance is determined by selecting a maximum distance from said plurality of results.

5. The system of claim 1, wherein said first controllable weights are adjusted based on a least means square fit of said first error signal and said second controllable weights are adjusted based on a least means square fit of said second error signal.

6. A method for determining distance from a plurality of antennas to a backscatter RFID tag comprising: transmitting a modulated signal to the tag via first and second antennas, receiving a first modulated backscattered signal having a first delay from the tag via said first antenna; receiving a second modulated backscattered signal having a second delay from the tag via said second antenna; inputting modulation in the backscattered signal received by said first antenna to a first linear adaptive combiner which is a delay line having first taps with first controllable weights on each tap; inputting modulation in the backscattered signal received by said second antenna to a second linear adaptive combiner which is a delay line having second taps with second controllable weights on each tap; inputting said transmitted modulated signal as a reference to said first linear adaptive combiner and second linear adaptive combiner; summing and subtracting outputs of said first adaptive linear combiner from said reference to produce a first error signal; summing and subtracting outputs of said second linear adaptive combiner from said reference to produce a second error signal; wherein weights for said taps of said first and second linear adaptive combiners are approximately zero for all of said taps except for respective first and second combiner taps corresponding to the first and second delays of the backscattered signal and determining the distance from the RFID tag to each of said antennas based on said first and second delays.

7. The method of claim 6, wherein said first and second antennas are mounted in a toll plaza and the tag in a vehicle.

8. The method of claim 6, further comprising operating said first and second linear adaptive combiners at a plurality of sample rates and said distance is determined from a plurality of results obtained at said plurality of sample rates.

9. The method of claim 8, wherein said distance is determined by selecting a maximum distance from said plurality of results.

10. The method of claim 6, further comprising adjusting said first controllable weights are based on a least means square fit of said first error signal and adjusting said second controllable weights adjusted based on a least means square fit of said second error signal.

Description

DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 is block diagram of an embodiment of a system for distance measurement;

(2) FIG. 2 shows a reference signal (upper graph) and backscattered signal containing a small amount of noise and lagging by 15 samples or 15 nsec. (lower graph)

(3) FIG. 3 shoes a calculated final weight distribution after 1000 cycles (upper graph) and an error signal for the input in FIG. 2. The weights are initially zero.

(4) FIG. 4 is a diagram of a toll plaza antenna and a vehicle having an RFID tag.

(5) FIG. 5 is a diagram showing dimensions for a multiple antenna configuration.

DETAILED SPECIFICATION

(6) FIG. 1 is a block diagram of the system under consideration. The index k indicates the dimension of time. FIG. 1 shows an adaptive linear combiner, transversal filter format. The weights are updated at each time step as follows:

(7) ${\stackrel{\_}{W}}_{0,k+1}={\stackrel{\_}{W}}_{0,k}+{\mathrm{\mu .Math.}}_k{\stackrel{\_}{X}}_k,{\stackrel{\_}{W}}_{1,k+1}={\stackrel{\_}{W}}_{1,k}+{\mathrm{\mu .Math.}}_k{\stackrel{\_}{X}}_k,\mathrm{.Math.}\mathrm{.Math.}$${\stackrel{\_}{W}}_{N,k+1}={\stackrel{\_}{W}}_{N,k}+{\mathrm{\mu .Math.}}_k{\stackrel{\_}{X}}_k.$

(8) This is the LMS algorithm. The parameter μ is a scalar gain coefficient. Each weight is adjusted at every time step according to the previous weight and previous input data, system error, and gain coefficient.

(9) In an embodiment, the reader's transmitted CW energy is modulated during the uplink with a known pattern that will be recovered in the backscattered signal. The modulation in the backscattered signal becomes the input X.sub.k to the adaptive linear combiner, while the transmitted signal is the reference. The proper pattern of weights is a value of approximately zero for all but the one corresponding to the delay of the backscattered signal.

(10) The sample rate of the A/D converter is an important quantity affecting the resolution of the proposed technique. The filter operates on samples that are one time step apart, so the final distribution of weights has the same granularity. This implies that the base resolution of the system is no better than the inverse of the sample rate. For a sample rate of 700 MHz, for example, the sample time is 1.42 nsec. Using a free space signal propagation speed of 1 foot/nsec, this corresponds to a physical delay of 1.42 feet. This delay corresponds to the (round trip) total range between the tag and receiver. The height of the receiver antenna requires a translation of this range to the horizontal range, which is the parameter of interest. Deviations of a foot in the total range can imply several feet in the horizontal range.

(11) Basic Adaptive Filter

(12) Reference and backscattered signals are shown in FIG. 2. The sampling rate is 1 GHz. FIG. 2 shows a reference signal in the upper graph and in the lower graph a backscattered signal, containing a small amount of noise, and lagging by 15 samples or 15 nsec.

(13) FIG. 3 shows calculated final weight distribution, after 1000 cycles, and error signal for the input in FIG. 2. The weights are initially zero.

(14) All weights are approximately zero except for the one representing a delay of 15 samples. Note that when the true delay is not equal to an integral number of clock cycles, the resulting weight distribution is still primarily concentrated on a single integer, depending on interference levels. The algorithm produces a delay that effectively rounds the true delay to the nearest N seconds, where N is the time between samples.

(15) Complications from Reader Antenna Height

(16) The height of a typical reader antenna in a toll plaza (typically 18 feet) adds a complication to the problem, summarized in FIG. 4 and Table 1. FIG. 4 shows typical scenario dimensions as tabulated in Table 1, with range to tags for Δh=13 feet.

(17) TABLE-US-00001 TABLE 1 Horizontal Distance Distance to distance to tag tag in adjacent to tag (x, feet) (R, feet) lane (feet) 0 13 17.69 1 13.04 17.72 2 13.15 17.80 3 13.34 17.94 4 13.60 18.14 5 13.92 18.38 6 14.32 18.68 7 14.76 19.03 8 15.26 19.42 9 15.81 19.85 10 16.40 20.32 11 17.03 20.83 12 17.69 21.38

(18) Therefore, a resolving the range to approximately a tenth of a foot is required to determine horizontal positions of a tag to within a foot. This is an order of magnitude less than the 1.42 feet limit on the range imposed by the sampling rate. The tags that are one lane away (12 feet) have ranges that are only about 4 feet greater for the same horizontal distance.

(19) The adaptive filter can only resolve distances that are one ‘sample’ apart. However, utilizing different sampling rates will produce different range estimates. In general, changing the sampling rate in real time, producing one range estimate per sample rate, and choosing the maximum range estimate can theoretically achieve the required resolution.

(20) For example, if Δh is 13 feet, and actual true horizontal distance is 10 feet. The table below summarizes the estimated horizontal distances for 10 different clock rates. The noise and interference are negligible. Table 2 shows estimated distances for Δh=13 feet and x=10 feet, as a function of sample time. Note that the total range is 2*R.sub.est, and is equal to an integer multiple of the sample time.

(21) TABLE-US-00002 TABLE 2 Sample time Clock rate R.sub.est X.sub.est (nsec) (MHz) (feet) (feet) 1.428 700 15.7 8.8 1.449 690 15.94 9.2 1.47 680 16.17 9.6 1.492 670 15.66 8.7 1.513 660 15.89 9.1 1.535 651 16.11 9.5 1.556 643 16.34 9.9 1.577 634 15.77 8.9 1.599 625 15.99 9.3 1.62 617 16.2 9.7

(22) The best and worst results in the forth column are highlighted. The spread between best and worst results, highlighted in Table 2, is about 1 foot. However, as the horizontal distance is reduced, the spread increases, as seen in the next example with the same parameters as the previous example, except the horizontal distance is 4.1 feet.

(23) TABLE-US-00003 TABLE 3 Sample time Clock rate R.sub.est X.sub.est (nsec) (MHz) (feet) (feet) 1.428 700 13.56 3.9 1.449 690 13.04 1.0 1.47 680 13.23 2.5 1.492 670 13.43 3.4 1.513 660 13.62 4.1 1.535 651 13.04 1.0 1.556 643 13.22 2.4 1.577 634 13.41 3.3 1.599 625 13.59 4.0 1.62 617 12.96 (imaginary!)

(24) Not only is the spread between best and worst estimates much larger, but the estimate is actually imaginary for one clock rate. This is because the horizontal distance is computed by X.sub.est=√{square root over (R.sup.2.sub.est−Δh.sup.2,)} and the estimated value for R.sub.est is smaller than the value of Δh. The example underscores the fact that at short horizontal distances, a single estimate (i.e. one clock period) can be off by several feet or not even be a real number.

(25) In general, this technique would require several clock cycle changes during one uplink period, producing several estimates for the range. The maximum value among the estimates is the closest to the true range.

(26) Complications of Variable Tag Height

(27) Additional complication comes from variation in the tag height. In order to estimate the horizontal range (X.sub.est), the value of Δh must be assumed. If the assumed Δh is too high, the value for X.sub.est becomes imaginary at short ranges. If it is too low, the value for X.sub.est suffers from low accuracy at short ranges. This is demonstrated in Table 4, which shows. Estimated horizontal distances with an assumed Δh of 13′ where actual Δh values vary per column.

(28) TABLE-US-00004 TABLE 4 X X.sub.est (feet) for X.sub.est (feet) for X.sub.est (feet) for (feet) true Δh = 12′ true Δh = 13′ true Δh = 14′ 10 8.6 9.9 11.2 9 7.5 9.0 10.4 8 6.2 8.0 9.5 7 4.8 6.9 8.7 6 3.3 6.0 7.8 5 0 5.0 7.2 4 j3.0 4.0 6.5 3 j4.1 2.9 6.0 2 j4.6 2.0 5.5 1 j5 0.8 5.3

(29) If the assumed Δh is near the true Δh, the error is small. The error is as high as 5′ for the other cases, and imaginary ranges result if the assumed Δh is too high.

(30) There are at least 2 ways to deal with this problem. The simplest and least accurate is to adjust the tag height assumption in real time when an imaginary range is computed. Another method is to add a second antenna and receiver, using the known separation between the antennas to determine a better assumption for Δh and therefore a more accurate horizontal distance estimate.

(31) Method 1: Adjustment of Assumed Height

(32) The simplest but least accurate approach is to begin the estimate process with an assumption for Δh that represents an average value. Then it is reduced as needed, when the resulting horizontal range estimate becomes imaginary. Consider the second column in Table 4, where the assumed Δh is 13′ and the true Δh is 12′. The error at a 10′ range is 1.4′ and increases as the tag moves closer to the antenna, reaching 5′ when the tag is 5′ away. At closer distances, the estimated range is imaginary with an assumed Δh of 13′, but if the assumed Δh is reduced to 12′, the error will be less than 1′. There may be more sophisticated methods of adjusting the assumed Δh that result in increased accuracy.

(33) Altering the assumed Δh does not require re-running the adaptive filter through an additional cycle. The filter specifically computes the range R, not the horizontal range. The imaginary number results from the single calculation X.sub.est=√{square root over (R.sup.2.sub.est−Δh.sup.2)}, and altering the assumed Δh and re-calculating can be done without re-estimating X.

(34) Method 2: Multiple Antennas

(35) For a given assumed height Δh, the estimated horizontal ranges are:
X.sub.1est=√{square root over (R.sup.2.sub.est−Δh.sup.2)} and X.sub.2est=√{square root over (R.sup.2.sub.est−Δh.sup.2)}.

(36) and the difference between the two,
X.sub.1est−X.sub.2est=√{square root over (R.sub.1est−Δh)}−√{square root over (R.sub.2est−Δh)},

(37) can be compared to the known quantity X.sub.1−X.sub.2.

(38) If the estimated difference is greater/less than the true difference, the assumed Δh must be reduced/increased. This is repeated until estimated and true values of X.sub.1−X.sub.2 are within an acceptable tolerance. This method requires an additional antenna and processing bank. The potential for imaginary ranges is still present, and would require reducing the assumed value of Δh, as in the case of a single antenna.