RFID Tags: Positioning Principles and Localization
Techniques
Mathieu Bouet
Laboratoire d’Informatique de Paris 6
Universit
´
e Pierre et Marie Curie
Paris, France 75016
mathieu.bouet@lip6.fr
Aldri L. dos Santos
Department of Informatics
Federal University of Paran
´
a
Curitiba, Paran
´
a, Brazil 81531-990
aldri@inf.ufpr.br
Abstract—RFID is an automatic identification technology that
enables tracking of people and objects. Both identity and location
are generally key information for indoor services. An obvious and
interesting method to obtain these two types of data is to localize
RFID tags attached to devices or objects or carried by people.
However, signals in indoor environments are generally harshly
impaired and tags have very limited capabilities which pose
many challenges for positioning them. In this work, we propose
a classification and survey the current state-of-art of RFID
localization by first presenting this technology and positioning
principles. Then, we explain and classify RFID localization
techniques. Finally, we discuss future trends in this domain.
Index Terms—RFID, localization, positioning algorithm.
I. INTRODUCTION
Radio Frequency IDentification (RFID) is widely used for
electronic identification and tracking. RFID offers substantial
advantages for businesses allowing automatic inventory and
tracking on the supply chain. This technology plays a key
role in pervasive networks and services [1]. Indeed, data can
be stored and remotely retrieved on RFID tags enabling real-
time identification of devices and users. However, the usage of
RFID could be hugely optimized if identification information
was linked to location.
This new dimension of context-awareness would support
the development of new strategies for autonomic and home
networking, mobility control, resource allocation, security, and
service discovery algorithms. Such functionalities would also
find application in indoor navigation, precise real-time inven-
tory, and in library management to retrieve persons or objects,
control access, and monitor events, for example. Classic RFID
systems provide coarse-grained location information. Their
readers are generally placed in strategic positions, like gates,
and their purpose is to detect tags that pass in their read range.
Thus, the localization accuracy of such systems corresponds
to the dimension of a cell formed by a reader.
With the popularity of indoor location sensing systems and
more globally of research on positioning in wireless networks,
RFID positioning issue has begun to emerge [2]. Localization
methods for RFID tags lie on the same principles than the
one for wireless networks. However, they are adapted to
the specific capabilities and constraints of this technology.
Indeed, tags have very limited capabilities in term of energy
and memory providing asymmetric short-range communica-
tion and centralized systems. Furthermore, future localization
methods will have to consider issues such as reader diversity,
mobility, security failures, among others.
This work presents a survey of the state-of-art related
to RFID localization. It introduces the current positioning
principles for indoor wireless networks in reason of their
numerous applications, and a classification of the main RFID
localization schemes found in the literature. The article is
organized as follows. Section II introduces the RFID technol-
ogy. Positioning principles for indoor wireless networks are
presented in Section III. The RFID localization techniques are
presented in Section IV. Finally, Sections V and VI provide
perspectives for RFID localization and conclude this survey.
II. THE RFID TECHNOLOGY
RFID networks are composed of three different entities,
RFID tags, readers, and servers, as shown in Fig. 1. All
RFID tags use radio frequency energy to communicate with
the readers. However, the method of powering the tags varies.
An active tag embeds an internal battery which continuously
powers it and its RF communication circuitry. Readers can
thus transmit very low-level signals, and the tag can reply
with high-level signals. An active tag can also have addi-
tional functionalities such as memory, and a sensor, or a
cryptography module. On the other hand, a passive tag has
no internal power supply. Generally, it backscatters the carrier
signal received from a reader. Passive tags have a smaller
size and are cheaper than active tags, but have very limited
functionalities. The last type of RFID tags is semipassive tags.
These tags communicate with the readers like passive tags but
they embed an internal battery that constantly powers their
internal circuitry.
RFID readers have two interfaces. The first one is a RF
interface that communicates with the tags in their read range
in order to retrieve tags’ identities. The second one is a
communication interface, generally IEEE 802.11 or 802.3, for
communicating with the servers.
Finally, one or several servers constitute the third part of
an RFID system. They collect tags’ identities sent by the
reader and perform calculation such as applying a localization
Fig. 1. Architecture of a classic RFID system.
method. They also embed the major part of the middleware
system and can be interconnected between each others.
RFID systems can be classified in two main categories in
accordance with their usage: monitoring and authorizing [3].
The first class includes RFID systems where tags are attached
in an inseparable way to the items they identify. Such networks
provide the capability to check, monitor and authenticate
which tags are present in the interrogation zone. Classic
utilizations are livestock or people embedded with RFID tags.
The second class includes RFID systems where RFID tags are
not permanently attached to entities. Due to this property, the
identity of the entity in possession of the RFID tag cannot
be verified. Typical usages of authorizing RFID systems are
access control in a building where tags are embedded inside
cards or keys.
III. POSITIONING PRINCIPLES FOR INDOOR WIRELESS
NETWORKS
Radio propagation in indoor environment is subject to
numerous problems such as severe multipath, rare line-of-
sight (LOS) path, absorption, diffraction, and reflection [4].
Since signal cannot be measured very precisely, several indoor
localization algorithms have been proposed in the literature.
They can be classified in three families: distance estimation,
scene analysis, and proximity.
A. Distance Estimation
This family of algorithms uses properties of triangles to
estimate the target’s location. The triangulation approach, il-
lustrated in Fig. 2, consists in measuring the angle of incidence
(or Angle Of Arrival - AOA) of at least two reference points.
The estimated position corresponds to the intersection of the
lines defined by the angles. On the contrary, the lateration
approach, illustrated in Fig. 3, estimates the position of the
target by evaluating its distances from at least three reference
points. The range measurements techniques use Received Sig-
nal Strength (RSS), Time Of Arrival (TOA), Time Difference
Of Arrival (TDOA), or Received Signal Phase (RSP).
1) RSS: The attenuation of emitted signal strength is func-
tion of the distance between the emitter and the receiver. The
target can thus be localized with at least three reference points
and the corresponding signal path losses due to propagation.
Several empirical and theoretical models have been proposed
to translate the difference between the transmitted and the
received signal strength into distance estimation. The RSS-
based systems usually need on-site adaptation in order to
reduce the severe effects of multipath fading and shadowing
in indoor environments.
2) TOA: The distance between a reference point and the
target is also proportional to the propagation time of signal.
TOA-based systems need at least three different measuring
units to perform a lateration for 2-D positioning. However,
they also require that all transmitters and receivers are pre-
cisely synchronized and that the transmitting signals include
timestamps in order to accurately evaluate the traveled dis-
tances. If more than three reference points are available, the
least-squares algorithm or one of its variants can be used in
order to minimize the localization error.
3) TDOA: The principle of TDOA lies on the idea of
determining the relative location of a targeted transmitter by
using the difference in time at which the signal emitted by a
target arrives at multiple measuring units. Three fixed receivers
give two TDOAs and thus provide an intersection point that
is the estimated location of the target. This method requires
a precise time reference between the measuring units. Like
TOA, TDOA has other drawbacks. In indoor environments, a
LOS channel is rarely available. Moreover, radio propagation
often suffers from multipath effects thus affecting the time of
flight of the signals.
4) RSP: The RSP method, also called Phase Of Arrival
(POA), uses the delay, expressed as a fraction of the sig-
nal’s wavelength, to estimate distance. It requires transmitters
placed at particular locations and assumes that they emit pure
sinusoidal signals. The localization can be performed using
phase measurements and the same algorithm than TOA or
phase difference measurements and the same algorithm than
TDOA. The disadvantage of the RSP method when applied
in indoor environments is that it strongly needs a LOS signal
path to limit localization errors.
Fig. 2. Triangulation: the estimated location is calculated with the angles
formed by two reference points and the target node.
5) AOA: AOA consists in calculating the intersection of
several direction lines, each originating from a beacon station
or from the target. At least two angles, measured with direc-
tional antennae or with an array of antennae and converted in
direction lines, are needed to find the 2-D location of a target.
Nevertheless, this technique requires complex and expensive
equipments and notably suffers from shadowing and multipath
reflections.
B. Scene analysis
Scenes analysis approaches are composed of two distinctive
steps. First, information concerning the environment (finger-
Fig. 3. Trilateration: the estimated location corresponds to the intersection
point of three circles.
prints) is collected. Then, the target’s location is estimated
by matching online measurements with the appropriate set
of fingerprints. Generally, RSS-based fingerprinting is used.
The two main fingerprinting-based techniques are: k-nearest-
neighbor (kNN) also known as radio map, and probabilistic
methods.
kNN - The kNN method consists in a first time in measuring
RSS at known locations in order to build a database of RSS
that is called a radio map. Then, during the online phase, RSS
measurements linked to the target are performed to search for
the k closest matches in the signal space previously-built. Root
mean square errors principle is finally applied on the selected
neighbors to find out an estimated location for the target.
Probabilistic Approach - The problem stated in probabilis-
tic approaches is to find the location of a target assuming
that there are n possible locations and one observed signal
strength vector during the online phase according to posteriori
probability and Bayes formula. Thus, the location with the
highest probability is chosen. Generally, probabilistic methods
involve different stages such as calibration, active learning,
error estimation, and tracking with history.
C. Proximity
The last type of localization techniques in indoor environ-
ments is based on proximity. This approach relies on dense
deployment of antennae. When the target enters in the radio
range of a single antenna, its location is assumed to be the
same that this receiver. When more than one antenna detect
the target, the target is assumed to be collocated with the one
that receives the strongest signal. This approach is very basic
and easy to implement. However, the accuracy is on the order
of the size of the cells.
IV. RFID LOCALIZATION SCHEMES
Several RFID localization methods have been proposed.
They utilize the indoor localization principles and are adapted
to the characteristics of the RFID technology. Due to the very
limited capabilities of tags and contrary to ad-hoc and sensor
networks, the localization is always centralized. With passive
tags or sparse reader deployment, the proximity approach is
privileged. On the contrary, when tags have more energy and
thus larger read range or when readers are densely deployed,
more elaborated techniques can be applied to localize tags.
RFID localization schemes can be classified into three fam-
ilies: lateration with distance estimation, scene analysis with
the deployment of extra reference tags, and constraint-based
approach.
A. Distance Estimation
1) SpotON [5]: SoptON is based on RSS measurements
from adjustable long range active RFID tags. The approach is
simple: multiple readers collect signal strength measurements
in order to approximate distance through a function defined
with empirical data. Classic laterations are then performed to
localize tags.
2) SAW ID-tags [6]: Surface Acoustic Wave Identification
(SAW ID) tags are completely passive. They utilize pulse
compression techniques and a large number of coding pos-
sibilities. Each tag is interrogated with the time inverse of its
impulse response. Then, it retransmits the correlated signal.
This retransmitted signal shows an autocorrelation peak. The
response with the highest amplitude identifies the searched tag.
The distance between each reader i and the tag is measured
based on TOA as follows:
d
i
=
T
total,i
− T
SAW
− T
sys
− T
cable,i
c
0
. (1)
The time delay T
sys
caused by the system and the time delay
T
cable,i
due to the cables between each receiving antenna
and the demodulator are calculated during a pre-calibration
phase. The time delay T
SAW
is equal for all tags. When
three estimated distances are available, the system performs
a trilateration to localize the tag.
3) LPM [7]: The Local Position Measurement (LPM)
system uses active tags. Since it is based on the TDOA
technique, readers are synchronized with the help of reference
tags (RT) at well-known and fixed positions that operate
continuously. After having received an activation command,
the selected measurement tag (MT) responds at time t
MT
.
The time difference t
diff
of the corresponding signals at each
reader R
i
can thus be calculated as follows:
c
0
t
diff
(R
i
) = c
0
(t
MT
−t
RT
)+kMT −R
i
k−kRT −R
i
k. (2)
The weighted mean squares method is then utilized to estimate
the locations of the tags with at least three different readers.
4) RSP [8]: The authors propose to apply the RSP tech-
nique, which they called Direction Of Arrival (DOA), to the
localization of passive RFID tags. Their approach consists in
placing two readers at specific locations in order to calculate
the phase difference and thus the direction of a moving tag.
When several observations are available, the estimation can be
improved by using the least-square fitting technique. With two
pairs of readers, two oblique angles are obtained and utilized
for a triangulation calculation as following:
(x
e
, y
e
) =
µ
H.
tan(θ
1
) tan(θ
2
)
tan(θ
1
) + tan(θ
2
)
,
H
2
.
tan(θ
1
) − tan(θ
2
)
tan(θ
1
) + tan(θ
2
)
¶
(3)
Where H is the distance between the centers of the two arrays
formed by the pairs of readers and θ
1
and θ
2
the estimated
DOAs of the tag by the two arrays.
B. Scene analysis
1) Landmarc [9]: This system is based on the kNN
technique. Reference tags which are fixed tags with known
positions are deployed regularly on the covered area. Readers
have eight different power levels. This approach consists in
selecting the k nearest reference tags from the unknown active
tag with the following indicator for each reference tag j:
E
j
=
v
u
u
t
n
X
i=1
(θ
j,i
− S
i
)
2
(4)
Where n is the number of readers, S
i
the RSS of the tag
measured by the reader i, and θ
j,i
the RSS of the reference
tag j measured by the reader i. E denotes the relationship
between each reference tag and the unknown tag. The k nearest
reference tags’ coordinates are then used to localize the tag:
(x
e
, y
e
) =
k
X
i=1
w
i
(x
i
, y
i
), with w
i
=
1
E
2
i
P
k
j=1
1
E
2
j
. (5)
The reference tag with the smallest E has the largest weight.
2) VIRE [10]: VIRE uses the principle of Landmarc [9],
that is 2D regular grid of reference tags. Nevertheless, this
method introduces the concept of proximity map. The whole
sensing area is divided into regions where the center of each
region corresponds to a reference tag. Every reader maintains
its own proximity map. If the difference between the RSS
measurement of the unknown tag and the RSS measurement
of a region is smaller than a threshold, the region is marked
as
0
1
0
. The fusion of all the n readers’ maps provides a
global proximity map for the tag. Two weighting factors are
defined. The first one demonstrates the discrepancy of the RSS
measurements between the selected reference tags and the tag:
w
1i
=
n
X
j=1
|θ
j,i
− S
i
|
n × θ
j,i
(6)
Where n is the number of readers, S
i
the RSS of the tag
measured by the reader i, and θ
j,i
the RSS of the reference
tag j measured by the reader i. The second weighting factor
is a function related to the density of selected reference tags.
The densest area has the largest weight:
w
2i
=
p
i
P
n
a
j=1
p
j
(7)
Where n
a
is the number of total regions and p
i
denotes the
ratio of conjunctive possible regions to the whole area. The
coordinates of the tag are finally calculated as following:
(x
e
, y
e
) =
n
a
X
i=1
w
1i
× w
2i
(x
i
, y
i
). (8)
3) Simplex [11]: This method is also based on the deploy-
ment of reference tags. It requires that the n readers have
K transmission power levels. For the localization of a tag, the
readers start with the lowest power level and gradually increase
the transmission power until they receive the response from the
tag. In the mean time, each reader also receives the responses
from reference tags. The distance L
i,j
between a reader i and
a tag j is then estimated by averaging the distances from the
reader to all reference tags detected in the same power level
but not in the previous power levels. The location of j is
calculated by minimizing the error function defined as:
²
j
=
n
X
i=1
Ã
L
i,j
− L
ˆ
i,j
L
i,j
!
2
. (9)
The simplex method is used to minimize ²
j
.
4) Kalman filtering [12]: This approach also utilizes ref-
erence tags. The first step consists in calculating with RSS
measurements from two readers the distance D
i
between each
reference tag and the target tag. The location of the tag is
obtained by solving with the minimum mean squared error
algorithm the system of non-linear equations:
(x
i
− x
e
)
2
+ (y
i
− y
e
)
2
= D
2
i
∀i = 1, ...n. (10)
The second step consists in building a probabilistic map of
the error measurement for the readers’ detection area. The
first step is applied for each reference tag in order to calculate
their corresponding error probability distribution function with
the help of their estimated location and their real location.
The Kalman filter is then used iteratively on this online map
to reduce the effect of RSS error measurement and thus to
improve the accuracy of the localization.
5) Scout [13]: Scout belongs to the family of probabilistic
localization techniques. This method also utilizes reference
tags and several readers. Active tags are localized following
three steps. First, the propagation parameters are calibrated us-
ing on-site reference tags. Secondly, the distance between the
targeted tag and the readers is estimated with a probabilistic
RSS model. Finally, the location of the tag is determined by
applying Bayesian inference. Iteratively, predicted beliefs are
calculated and then corrected with observations until obtaining
a good model resulting in an estimated area.
C. Constraint-based approach
1) 3-D Constraints [14]: This approach is only based on
connectivity information. They are used to define inclusive
constraints, that is if a reader can detect a tag that means
that the distance between them is inferior to the read range,
and exclusive constraints, the complementary with readers that
cannot detect the tag. The space is discretized into points in
order to delimit the detection area of the readers. The mean
of the set of points that respect the maximum of constraints
corresponds to the estimated location of the tag.
Table I briefly compares RFID localization schemes. They
are classified according to their approach of the problem.
Some of them require the deployment of reference tags which
provide finer data but also considerably increase the cost of the
system and the maintenance. Except two schemes that are built
on very specific properties of passive tags, they all concerns
active tags which have larger capacities. Accuracies cannot be
directly compared since the systems do not consider the same
hypothesis.
TABLE I
RFID LOCALIZATION SCHEMES
Localization Scheme Positioning Algorithm Reference Tags Target Space Dimension
SpotON [5] (2000) RSS lateration No Active 3-D
SAW ID-tags [6] (2003) TOA lateration No Passive 2-D
LPM [7] (2004) TDOA weighted mean squares No Active 2-D
RSP [8] (2007) RSP/AOA No Passive 2-D
Landmarc [9] (2003) kNN Yes Active 2-D
VIRE [10] (2007) kNN Yes Active 2-D
Simplex [11] (2007) kNN optimization Yes Active 3-D
Kalman filtering [12] (2007) RSS mean squares and Kalman filetering Yes Active 2-D
Scout [13] (2006) RSS Bayesian approach Yes Active 2-D
3-D Constraints [14] (2008) Range-free optimization No Active 3-D
V. PERSPECTIVE
Pervasive networks are potentially rich in term of infor-
mation. The quantity and the diversity of their components
could be used to increase the accuracy of RFID localization.
Therefore, future localization methods should take into ac-
count several aspects:
• RF model. Most of the current methods perform RSS
measurements. However, they generally use models de-
veloped for wireless networks. RFID propagation has
some particularities that should be considered in an
appropriate RF model.
• Reader redundancy. Reader redundancy should be more
exploited to obtain more data with the respect of problems
in dense deployments of readers [15].
• Reader diversity. Localization in a RFID network with
readers that have different read ranges, antennas, and
capacities could be an interesting and more realistic
approach [16].
• Intelligent constraints. Intelligent constraints could be
deduced from meta-information. For example, two tags
fixed on the same package define a bound on the physical
distance between them [16].
• Mobility. Hybrid systems with static and mobile readers
should be considered to increase the quantity and the
diversity of collected data.
• Scalability. The scalability of the RFID localization tech-
niques should be carefully studied in order to define the
amount of tags that can be read in a given period; the rate
of successful reading and its impact on the accuracy; and
how long takes the localization calculation.
• Metric. Finally, RFID localization schemes cannot be
directly compared since they lie on different hypothesis.
An interesting metric to be precisely defined would be
the accuracy versus the cost of the whole system.
VI. CONCLUSION
This paper surveys the current state-of-art of RFID
localization. The presented techniques were classified
according to their approach: distance estimation, scene
analysis, or topological constraints. Designed for passive
or active tags, some techniques require the deployment of
reference tags when others necessitate specific equipment to
take into account changes in environment and to proceed to
calibration. On the contrary, some techniques are designed
so as to be more cost effective and more easily adaptable to
the utilization of different equipments. Globally, in terms of
scalability and availability, these RFID positioning techniques
have their own important characteristics when applied in
real environments. The choice of technique and technology
(passive or active tags) significantly affects the granularity
and accuracy of the location information but also the whole
cost and the efficiency of the RFID system.
REFERENCES
[1] R. Want. An introduction to RFID technology. IEEE Pervasive
Computing, 5(1):25–33, Jan.-March 2006.
[2] A. Cangialosi, J.E. Monaly, and S.C. Yang. Leveraging RFID in hospi-
tals: Patient life cycle and mobility perspectives. IEEE Communications
Magazine, 45(9):18–23, Sept. 2007.
[3] T. Hassan and S. Chatterjee. A taxonomy for RFID. In Proc. of HICSS,
2006.
[4] T. Rappaport. Wireless Communications: Principles and Practice.
Prentice Hall PTR, Upper Saddle River, NJ, USA, 2001.
[5] J. Hightower, R. Want, and G. Borriello. SpotON: An indoor 3D location
sensing technology based on RF signal strength. Technical report, Univ.
of Washington, Dep. of Comp. Science and Eng., Seattle, WA, Feb.
[6] T.F. Bechteler and H. Yenigun. 2-D localization and identification based
on SAW ID-tags at 2.5 GHz. IEEE Trans. on Microwave Theory and
Techniques, 51(5):1584–1590, 2003.
[7] A. Stelzer, K. Pourvoyeur, and A. Fischer. Concept and application of
LPM - a novel 3-D local position measurement system. IEEE Trans. on
Microwave Theory and Techniques, 52(12):2664–2669, Dec. 2004.
[8] Y. Zhang, M. G. Amin, and S. Kaushik. Localization and tracking of
passive RFID tags based on direction estimation. International Journal
of Antennas and Propagation, 2007.
[9] L.M. Ni, Y. Liu, Y.C. Lau, and A.P. Patil. LANDMARC: indoor location
sensing using active RFID. In Proc. of PerCom, pages 407–415, 2003.
[10] Y. Zhao, Y. Liu, and L.M. Ni. VIRE: Active RFID-based localization
using virtual reference elimination. In Proc. of ICPP, 2007.
[11] C. Wang, H. Wu, and N.-F. Tzeng. RFID-based 3-D positioning
schemes. In Proc. of INFOCOM, pages 1235–1243, 2007.
[12] A. Bekkali, H. Sanson, and M. Matsumoto. RFID indoor positioning
based on probabilistic RFID map and kalman filtering. In Proc. of
WiMOB, 2007.
[13] X. Huang, R. Janaswamy, and A. Ganz. Scout: Outdoor localization
using active RFID technology. In Proc. of BROADNETS, 2006.
[14] M. Bouet and G. Pujolle. A range-free 3-D localization method for
RFID tags based on virtual landmarks. In Proc. of PIMRC, 2008.
[15] D.Y. Kim, B.J. Jang, H.G. Yoon, J.S. Park, and J.G. Yook. Effects
of reader interference on the RFID interrogation range. In Proc. of
European Microwave Conference, pages 728–731, 2007.
[16] N. Vaidya and S.R. Das. Rfid-based networks: exploiting diversity and
redundancy. SIGMOBILE Mob. Comput. Commun. Rev., 12(1):2–14,
2008.
