2015 IEEE International Conference on Advanced Intelligent Mechatronics (AIM)

July 7-11, 2015. Busan, Korea

RSSI Informed Phase Method for Distance Calculations

Ann Whitney, Johné Parker, Member, ASME, Mallory Engelhard, Zach Kratzer, Member, IEEE, and
John Fessler, Member, IEEE

environmental robustness of the phase method, a new

,

Abstract— In an attempt to find an accurate,

environmentally robust, and fast process for RFID distance method was created called RSSI Informed Phase.
estimations, a method was developed called RSSI Informed The remainder of this paper is organized as follows:
Phase. This method uses both phase and RSSI measurements first, the standard RSSI and phase methods will be discussed
to estimate the distance, and potentially location, of a moving along with how to calculate each, secondly the innovative
RFID tag. RSSI is initially used to find an approximate reader
RSSI Informed Phase method will be explained, next the
to tag separation distance. This distance value is applied to
find an approximate slope of the phase angle vs. frequency experimental design to test these methods will be described,
curve. The estimated slope can inform the phase distance then the results of the experiment will be analyzed, and
calculation, meaning fewer reads are required to find the actual finally the derived conclusion will be discussed.
phase angle vs. frequency slope. The reduction in the number
II. BACKGROUND
of necessary reads accelerates the localization process and
makes this method more robust for dynamic environments. A. RSSI Distance Estimation
While there are many different methods for
I. INTRODUCTION wireless localization [1][4][5][7][9][10][14], most begin
with a distance calculation [9][11][12][14]. RSSI is
When using RFID (radio frequency identification)
commonly used for distance estimation in RFID systems.
technology to estimate the distance between the reader and
Distance estimations can be found by using a calibrated
the tag, there are two main pieces of information which can
version of the Friis equation [6] as shown in Figure 1.
be extracted from the return signal of the tag. The most
commonly used metadata for predicting the reader to tag
separation distance is the return signal strength indication
(RSSI) [11]. However RSSI is extremely environmentally
dependent and it is not uncommon to have large amounts of
error in physically cluttered or RF (radio frequency) noisy
environments [1][2][11][13]. The other data which can be
gleaned from the tag’s return signal is the phase angle [3].
Since the phase angle of the return signal is dependent on the
frequency and read distance, this distance can be calculated
by using the measured phase angle at several different
frequencies. An issue with this method is that reading the tag
over a wide range of frequencies is time consuming,
especially if the tag is using pseudo-random frequency Figure 1. RSSI vs. Distance Best Fit Curve
hopping. This long sampling time means that, if the tagged
object is moving, the resulting distance estimation can be When an RFID tag is read, the resulting RSSI can
unreliable. However, distance calculations based on phase be compared to a known RSSI vs. distance curve to estimate
angle measurements are shown to be more environmentally the distance between the reader and the tag. However, RSSI
robust [8][11]. In an attempt to use the best attributes of is heavily dependent on the surrounding environment, and
both methods, the speed of the RSSI method and the can easily give inaccurate distance estimations in practical
situations.
B. Phase Based Distance Estimation
*Research supported by Lexmark International and the University of
Kentucky College of Engineering. Phase angle can also be used to calculate the reader
Ann Whitney is a Ph.D. student in the Department of Mechanical to tag separation distance. Initially the tag is read at one hop
Engineering at the University of Kentucky, Lexington, KY 40506-0503
(email: [email protected])
frequency, and the phase angle is measured by the reader.
Johné Parker is an Associate Professor of Mechanical Engineering at the Then the process is repeated for a number of hop frequencies
University of Kentucky, Lexington, KY 40506-0503 (phone: 859-257-6336 until a phase angle curve as shown in Figure 2 can be
x80647; email: [email protected]) constructed. Assuming the RFID reader and tag both remain
Mallory Engelhard is a MSTC student at Paul Laurence Dunbar High
School.
stationary, the relationship between the change in phase
John Fessler is an EMC engineer at Lexmark International, Lexington, angle and the change in frequency is directly proportional to
KY (email: [email protected]) the reader to tag separation distance.
Zach Kratzer is an EMC engineer at Lexmark International, Lexington,
KY (email: [email protected])

978-1-4673-9107-8/15/$31.00 ©2015 IEEE 1138

 Where:
d = predicted reader to tag distance (m)
c = speed of light (m/s)
φ = phase angle (radians)
f = frequency (Hz)
β = experimental offset (m)

This equation differs slightly from the one given by

Nikitin [11], with the addition of parameter β. This constant
β is an offset based on the particular experimental setup
being used, and can be found by calibration. Most likely it
will be related to the length of cable between the antenna
and radio within the reader.
Figure 2. Example Of A Sawtooth Phase Curve
While phase-based distance estimations are more
The phase angle vs. frequency sawtooth curve in environmentally robust than RSSI, there are drawbacks to
Figure 2 ranges from 0 to π due to the way in which phase each method. While RSSI distance estimations are
angle is calculated by the reader, using Equation (1) [3]. environmentally dependent, the phase method is quite time
consuming, especially if pseudo-random frequency hopping
𝑄𝑄𝑛𝑛𝑛𝑛 is used. Therefore the phase angle method would only be
𝜑𝜑 = tan−1 � � (1) reliable on stationary tags.
𝐼𝐼𝑛𝑛𝑛𝑛
Where:
III. RSSI INFORMED PHASE METHOD
φ = phase angle
Qnk = quadrature, and For accurate and reliable distance estimations in
Ink = in-phase practical applications, a method is needed which is
environmentally robust, and rapid enough to be used for
In order to use the sawtooth curve to estimate mobile objects. Therefore, a method has been devised,
distance it is necessary to find the slope of the curve, which which uses the speed of RSSI distance estimation and the
represents the amount the phase angle varies with a change environmental robustness of phase distance estimation. This
in frequency. The slope can be more easily calculated after new method is called RSSI Informed Phase.
linearizing the sawtooth curve. This linearization is In the standard phase method, a large number of
accomplished by adding or subtracting increments of π for phase angles are needed at a wide range of frequencies to
each of the jumps in the sawtooth curve; therefore creating determine where the peaks and valleys of the sawtooth curve
one relatively straight line as shown in Figure 3. are located. Without knowing the location of these peaks
and valleys, the correct number of increments of π cannot be
added to linearize the phase angle curve. In addition, if the
frequency location of the peaks and valleys of the sawtooth
curve are inaccurate, this could result in a drastically
different slope and therefore produce a large error in the
distance calculation.
The idea of the RSSI Informed Phase method is
that, in order to calculate the slope of the phase curve, it is
not necessary to find the phase angles at every possible
frequency. If the approximate slope, frequency location of
the peaks and valleys, or number of increments of π are
known, a relatively accurate slope of the phase vs. frequency
line can be calculated with as few as two phase data points.
Ideally these points would have a significant frequency
Figure 3. An Example of a Linearized Phase Graph
separation to increase the accuracy of the slope. However,
when solely relying on two phase data points, there is no
Note that it is not important if the sawtooth curve is way of knowing how many increments of π, or how many
shifted up or down during this process as only the slope of jumps in the sawtooth curve are between the two points.
the line is needed. Once a relatively straight line is created, The number of increments of π depends on the reader to tag
the slope can be calculated with linear regression. The distance, and greatly impacts the slope of the phase vs.
resulting slope can be used to calculate reader to tag frequency line.
separation distance with Equation (2). To solve this problem, the RSSI and phase are
𝑐𝑐 𝜕𝜕𝜕𝜕
measured at two or three different frequencies spread out
𝑑𝑑 = − + 𝛽𝛽 (2) within the range of possible frequencies. For instance, one
4𝜋𝜋 𝜕𝜕𝜕𝜕
measurement can be taken at the lower end of the frequency
range, one at the upper end, and one in the middle. The

1139
average RSSI value can then be calculated using these data
points, and this value can be used to initially approximate
the reader to tag distance with the experimentally found
RSSI vs. distance best fit curve. This initial distance
estimation is applied to Equation (2) to find an estimate of
the phase vs. frequency slope. With this approximate slope,
the two or three collected phase data points are enough to
calculate a more environmentally robust distance estimation
by use of the phase angle data. The RSSI-estimated slope is
made into a line which passes through one of the phase angle
points, as shown in Figure 4.
Figure 6. Calculation of Final Phase Slope

The slope of this final line is the input for Equation

(2), which yields the final RSSI Informed Phase distance
estimation.
IV. EXPERIMENTAL DESIGN
To test the efficacy of the RSSI Informed Phase
method, it was compared to the standard RSSI and phase
distance estimation methods previously discussed. In
addition, measurements were taken while the RFID tag was
moving at various speeds in order to observe the accuracy of
these methods in dynamic situations. It was hypothesized
that the standard phase would perform well when the tag
Figure 4. The Initial Location of the Phase Angle Points and the Relative was stationary or moving slowly, but would have high error
RSSI Calculated Slope. when the tag moved more quickly. In contrast, it was
predicted that the RSSI Informed Phase method would have
Next, increments of π are added or subtracted from somewhat greater error than the standard phase method with
the remaining phase data points in order to align them as a stationary or low speed tag, but maintain relatively low
closely as possible with the RSSI predicted line, as shown in error even when the tag speed was increased. The standard
Figure 5. RSSI method was hypothesized to perform well with low
distance errors considering that a semi-anechoic chamber
was used for testing.
The measurements were taken in a 5 meter semi-
anechoic chamber with an M6e ThingMagic radio, a circular
polarized antenna, and a vertically polarized Alien
Squiglette RFID tag. Due to the fact that the testing was
conducted in a semi-anechoic chamber, the frequencies
could be stepped through incrementally (rather than pseudo-
randomly) to make the standard phase method as time
efficient as possible. For the same reason, the three phase
data points for the RSSI Informed Phase method could also
be specifically selected.
The necessary data was collected for all three
Figure 5. Phase Data Points Shifted Closer to RSSI Calculated Line.
methods with the RFID tag stationary, as well as traveling at
speeds ranging from 0.01 m/s through 0.1 m/s, towards the
The final step improves upon the initial RSSI
reader antenna. The actual distance was then considered to
estimation by using a best-fit line from the resulting
be half way between the beginning and ending location. For
relocated phase data points with linear regression (Figure 6).
consistency, all measurements began at 3 meters separation
distance.
V. RESULTS
The average error of the predicted distance was
calculated for each tag speed, using each method of distance
estimation. When the tag was completely stationary, all
three methods had minimal error. However, once the tag
began to move, the standard phase method had large and

1140
continually increasing distance errors, which were often
greater than the separation distance itself. The RSSI
Informed Phase method and the standard RSSI method were
very comparable, with the standard RSSI method having
slightly less error overall (See Figure 6). The standard error
for each predicted distance was calculated, but found to be
negligible.

Figure 8. Linearized Phase Angle vs. Frequency for Mobile Tag

(0.06 m/s)

Figure 8 is an excellent demonstration of how the

typically negative slope of the phase angle vs. frequency line
can begin to shift while the measurements are being
collected due to the movement of the tag. For this reason,
only collecting two or three samples and reducing the
sample time makes the RSSI Informed Phase method
Figure 6. Distance Error of the Three Methods of Distance Estimation superior for dynamic tag localization.
In this experiment, the RSSI method performed
VI. DISCUSSION similarly to the RSSI Informed Phase method; however, this
data was collected in a semi-anechoic chamber where
Phase methods for distance estimations are conditions are nearly ideal for using RSSI for distance
understood to be more environmentally robust than RSSI estimations. It is not unlikely that in a practical setting the
methods; however this is assuming that the RFID tag is standard RSSI method of distance estimation could have
immobile. If, instead, the tag is moving, it could easily take much larger amounts of error.
too long to measure the phase angle at every (or nearly
every) frequency. When this happens, instead of creating a VII. CONCLUSION
straight phase vs. frequency line, the line begins to shift as Three different methods for calculating RFID
the separation distance is changing. Figure 7 is a linearized reader to tag separation distance were analyzed: standard
phase vs. frequency graph when the tag is stationary, and it RSSI, standard phase angle, and RSSI Informed Phase. The
can be seen that the phase data points are clustered along a RSSI Informed Phase method is a new method created by
negatively sloped straight line. the authors with the intention of combining the
environmental robustness of the phase angle method with
the speed of a simple RSSI distance estimation. This
method used the average of two or three RSSI values to
calculate an approximate phase vs. frequency slope used in
phase distance estimations. Increments of π were added or
subtracted from the phase angle of the two or three measured
data points, until they aligned as closely as possible to the
initial line estimate. Finally, a new line was calculated with
the altered phase angle data points and used to find the
improved distance estimate.
When the methods were tested in a semi-anechoic
chamber, it was found that the RSSI Informed Phase method
and standard RSSI method both had similarly low errors at
all of the tested tag speeds. However, this is to be expected
Figure 7. Phase Angle vs. Frequency Line Measured from Stationary Tag.
of the standard RSSI distance estimation method, given the
In contrast, if the tag is mobile, the approximately relatively clean environment of the semi-anechoic chamber.
10 seconds of data collection become problematic for While the standard phase method had slightly
measuring the phase angle vs. frequency line. Figure 8 lower distance errors when the tag was stationary, the RSSI
shows how the phase angle vs. frequency curve can shift and Informed Phase method had significantly less error when
result in a positive slope. This extreme change in the slope measuring mobile tags. Therefore, based upon this
of the line leads to large errors in the predicted distance. experiment it would appear that the RSSI Informed Phase
method may be the best of the three methods in a practical
environment with a mobile RFID tag.

1141
 VIII. FUTURE WORK
A more in-depth analysis of the benefits and
drawbacks of the RSSI Informed Phase in comparison to
other standard methods of distance estimation for RFID tags
is currently in progress. In addition, the authors plan to
investigate how these methods perform in a range of
situations or environments.

ACKNOWLEDGMENT
The authors would like to thank Dr. Julie Whitney, Senior
Technical Staff Member at Lexmark International, for her
support.

REFERENCES
[1] Bouet, Mathieu, and Aldri L. Dos Santos. RFID Tags:
Positioning Principles and Localization Techniques. Proc. of
Wireless Days, Duabi. N.p.: IEEE, 2008. 1-5.
[2] Calis, Gulben, Burcin Becerik-Gerber, Ahmet B. Goktepe, Shuai
Li, and Nan Li. "Analysis of the Variability of RSSI Values for
Active RFID-based Indoor Applications." Turkish Journal of
Engineering & Environmental Sciences (2013): 186-210.
[3] Carrick, John C. Methods and Apparatuses for RFID Tag Range
Determination. ThingMagic Inc., assignee. Patent US
20100109903A1. 6 May 2010.
[4] Chen, Lun-Chi, Ruey-Kai Sheu, Hui-Chieh Lu, Win-Tsung Lo,
and Yen-Ping Chu. "Object Finding System Based on RFID
Technology." APWeb 2006 International Workshop (2005): 383-
96. Springer.
[5] Cicirelli, Grazia, Annalisa Milella, and Donato Di Paola. "RFID
Tag Localization by Using Adaptive Neuro-Fuzzy Inference for
Mobile Robot Applications." Industrial Robot: An International
Journal (2012): 340-48. Emerald Group Publishing Limited.
[6] Friis, Harald T. "A Note on a Simple Transmission Formula."
Proceedings of the I.R.E. and Waves and Electrons (1946): 254-
56.
[7] Gezici, Sinan. "A Survey on Wireless Position Estimation."
Wireless Personal Communications 44.3 (2008): 263-82.
Springer.
[8] Hekimian-Williams, Cory, Brandon Grant, Xiuwen Liu,
Zhenghao Zhang, and Piyush Kumar. Accurate Localization of
RFID Tags Using Phase Difference. Proc. of RFID 2010,
Orlando, FL. IEEE, n.d.
[9] Joho, Dominik, Christian Plagemann, and Wolfram Burgard.
Modeling RFID Signal Strength and Tag Detection for
Localization and Mapping. Proc. of Proceedings of the IEEE
International Conference on Robotics and Automation, Japan.
N.p.: n.p., n.d. 3160-165.
[10] Mo, J P T, and D R Pearson. "Localization of Position Using
Radio Frequency Identification Array." Journal of Engineering
Manufacture 225 (2010): 675-84.
[11] Nikitin, Pavel V., Rene Martinez, Shashi Ramamurthy, Hunter
Leland, Gary Spiess, and K.V.S. Rao. "Phase Based Spatial
Identification of UHF RFID Tags." IEEE RFID 2010 (2010):
102-09.
[12] Whitney, Ann, John Fessler, Johné Parker, and Nathan Jacobs.
Received Signal Strength Indication Signature for Passive UHF
Tags. Proc. of IEEE/ASME International Conference on
Advanced Intelligent Mechatronics, France, Besancon. N.p.:
n.p., 2014.
[13] Wu, Rong-Hou, Yang-Han Lee, Hsien-Wei Tseng, Yih-Guang
Jan, and Ming-Hsueh Chuang. Study of Characteristics of RSSI
Signal. Proc. of Industrial Technology, Chengdu. IEEE, n.d.
[14] Zhou, Junyi, and Jing Shi. "Localization of Stationary Objects
Using Passive RFID Technology." International Journal of
Computer Integrated Manufacturing 22.7 (2009): 717-26.
Taylor & Francis.

1142