RFID Assisted Vehicle Positioning in VANETs
Eun-Kyu Leea, Soon Y. Ohb,∗, Mario Gerlaa
aDepartment of Computer Science, University of California, Los AngelesLos Angeles, CA 90095, USA
bUtopiaCompression11150 Olympic Blvd. # 820, Los Angeles, CA 90064, USA
Abstract
With technological advancement, recent VANET applications such as safe
driving and emergency rescue often demand high position accuracy. Unfor-
tunately, however, conventional localization systems, e.g., GPS, hardly meet
new accuracy requirements. To overcome this limitation, this paper pro-
poses an RFID-assisted localization system. The proposed system employs
the DGPS concept to improve GPS accuracy. A vehicle obtains two different
position data: GPS coordinate from own GPS receiver and accurate phys-
ical position via RFID communication. Then, it computes GPS error and
shares it with neighbors to help them correct inaccurate GPS coordinates .
To evaluate the proposed system, we conduct extensive experiments both on
a simulator and on a real world test-bed. The simulation shows that, with
the RFID-assisted localization system, vehicles can acquire accurate position
both on a freeway and in an urban area. The results from the test-bed exper-
iments demonstrate that the proposed system is feasible in the real VANET
∗Corresponding authorEmail addresses: [email protected] (Eun-Kyu Lee),
[email protected] (Soon Y. Oh), [email protected] (Mario Gerla)
Preprint submitted to Pervasive and Mobile Computing May 16, 2011
environment.
Keywords: VANET, Localization, RFID
1. Introduction
In Vehicular Ad hoc Networks (VANETs), vehicles exchange data with
each other or with a roadside infrastructure in order to use various applica-
tions from Internet access to safety applications. The VANET applications
often assume that vehicles’ real-time position is provided. This is reasonable
because GPS receivers have become popular in vehicles today. For instance,
navigation applications make use of GPS data for finding their location.
However, critical applications that support safe driving require more accu-
rate position information so that advanced localization techniques have been
proposed to support them.
In wireless networks, signal propagation properties, e.g., Received Signal
Strength Indicator (RSSI) or Time of Arrival (ToA), have been exploited for
localization. Cellular localization [1] uses propagation delay of the signals
from transmission towers to calculate the “absolute” position. Kukshya et
al. [2] use the propagation properties to estimate distance between neighbor
vehicles and “relative” position. However, due to distortion and interference,
a wireless channel is too unstable to guarantee consistent and accurate po-
sition. Trilateration, a well-known positioning technique in VANETs to find
relative location, requires at least three neighbors within the radio range to
determine position so that it cannot easily provide location in sparse traffic
environments. Another VANET localization approach, data fusion, calcu-
lates position by integrating several types of data obtained from different
2
devices, e.g., a GPS, a digital camera, a smart-phone, etc [3, 4]. Position
accuracy in data fusion relies on the number and quality of engaging sensors
and pre-training may be necessary for better performance. Among existing
schemes, Differential GPS (DGPS) improves position accuracy to the level
of tens of centimeters in the best case [5]. In road environment, however, its
accuracy degrades steeply as a vehicle goes far away from a reference point.
In order to provide accurate position in VANET, this paper proposes a
novel localization system assisted by Radio-Frequency IDentification (RFID).
Main inspiration of our design is the notion of DGPS to improve GPS accu-
racy. A GPS vehicle, a vehicle equipped with a GPS system, obtains exact
position data from an RFID tag on a roadside unit using an RFID reader
while driving. Then, it broadcasts the calculated GPS error value to neigh-
bor vehicles via IEEE 802.11 radio. A non-GPS vehicle, which does not
have a GPS system, computes its position using our single peer localization
scheme. When the non-GPS vehicle encounters a vehicle having accurate
position data, they exchange position and travel information via RFID and
802.11 radio, respectively. At the end of this process, the non-GPS vehicle
can estimate its accurate position from the received data.
Our primary contribution is the design of a novel accurate localization
system that does not use signal strength or propagation properties. More-
over, a non-GPS vehicle is able to estimate its own accurate position from a
neighbor vehicle. To prove feasibility, we investigate the various parameters
of the RFID technology and analyze their impact on accurate positioning
in VANET. We also build a test-bed on a road and conduct experiments
driving a car equipped with the RFID system. Simulation and experiment
3
results show that the RFID-assisted localization is feasible on the road and
can provide sufficient accuracy for critical VANET applications.
The rest of the paper is organized as follows. Section 2 reviews related
works: two systems to improve GPS accuracy; and RFID system and its
VANET applications. Section 3 presents the proposed RFID-assisted local-
ization system, which is followed by its evaluation via simulation in Section
4. In Section 5 and 6, we describe our test bed set up on a real road envi-
ronment and demonstrate the experimental results, respectively. Finally, we
conclude the paper in Section 7.
2. Related Works
2.1. GPS Error Correction
Differential GPS was used first for a maritime navigation system [6].
DGPS leverages the scientific observation that the distance to the satellites
is so far that GPS receivers within a nearby area on the Earth experience
the same signal propagation delay, which leads to the same GPS error. In
the DGPS system, a reference point is installed near the coast line, e.g., a
lighthouse, and calculates GPS error that is the difference between a GPS
value received from own GPS receiver and exact location measured on its
installation. GPS error is then delivered to ships passing nearby so that they
correct their GPS coordinates. Because there are no obstacles distorting GPS
signals in the sea, the reference point and the ships passing near the point
have the same GPS error. In this way, each ship is able to maintain accurate
position. In an urban area, however, tall buildings obstruct and distort GPS
signals so that each block shows different GPS error values even within a
4
small area. Thus, the DGPS system cannot guarantee consistent position
accuracy. In addition, it is too expensive to be installed in a vehicle.
Dead Reckoning (DR) is another GPS error correction technique [7],
which is applied when a GPS signal is temporarily unavailable. Once a
mobile node fails to receive GPS signals, it estimates current position based
on last recorded location from the GPS system and its mobility information,
e.g., speed, orientation, and time. Vehicular applications employ DR in such
places where the GPS signal cannot be reached such as tunnels and indoor
parking lots. DR guarantees accurate position only for a short time because
estimation error quickly accumulates. The estimation error depends on ac-
curacy of on-board speed and orientation sensors. As a term of comparison,
a speedometer has an error range of ±3 ∼ ±10%, and a digital compass or a
gyroscope has 10◦/s of orientation error at maximum.
2.2. RFID System
An RFID system is composed of an RFID tag storing an object identity
and an RFID reader that accesses tag’s data. The difference of the RFID
system from traditional bar code is its use of electromagnetic waves for data
transmission. The RFID system operates in various frequency bands and
provides corresponding radio ranges. Among them, many applications prefer
a passive RFID system in Ultra High Frequency (UHF, 860MHz∼2.45GHz)
because of low tag price (less than 10 cents) and relatively long radio range
(up to approximately 10m). In the passive RFID system, the RFID reader
emits RF radio waves eliciting a signal back from a tag. The RFID tag storing
its ID in the memory works without an external power source. More precisely,
upon receiving the radio waves, the tag absorbs energy and pumps back the
5
waves modulated with own ID signature. This battery-free operation is the
most remarkable advantage of the system.
2.3. RFID Applications in VANET
The RFID system has been applied to VANET applications in two differ-
ent ways. First, an RFID tag is attached to a vehicle, and an RFID reader
is installed on a roadside. In the Automatic Toll Collection (ATC) system,
the RFID reader installed on a toll booth reads a RFID tag on a vehicle
when it passes by the booth. The ATC system identifies the passing vehicle
and charges toll road fee automatically. In the Automatic Vehicle Location
(AVL) in Vejle, Denmark, RFID tags are attached to the front bumper of
buses, and RFID readers are installed on the road along the bus route in
order to identify passing buses. In Edinburgh, UK, RFID tags are attached
to buses and emergency vehicles so that RFID readers on the roadside mon-
itor the flow of public transportation and detect emergence situation. Upon
receiving this information, a traffic control center controls traffic signals to
resolve traffic congestion and to prevent car accidents.
In a second type of applications, the RFID reader is on the vehicle,
whereas RFID tags are on the roadside unit. In Road Beacon System (RBS)
[8], RFID readers on the vehicles gather road information from RFID tags
buried under the pavement. In [9], an RFID tag is assumed to have accu-
rate position data, and vehicles with an RFID reader update their locations
when passing over the tags on the road. Previous research has addressed
the issue of installing the RFID reader on a vehicle without any verification,
but this paper proposes a novel localization system with protocol design and
verification with real implementation and experiments. Moreover, we tackle
6
a new possibility that a vehicle is equipped with both the RFID tag and
the RFID reader, which, we believe, is the future direction of VANET RFID
applications.
3. RFID-Assisted Localization System
3.1. Preliminary
We assume that all the vehicles are equipped with the RFID system and
IEEE 802.11 radio (e.g., 802.11p-Dedicated Short Range Communications
[DSRC] [10]), but a GPS receiver is installed only in a fraction of vehicles.
The vehicles broadcast packets to one-hop neighbors via 802.11 radio while
the RFID tag/reader set is used for short-ranged RFID communication. A
lane on the road is 3m wide, and RFID tags are installed either in roadside
units or on the road surface. Figure 1 illustrates the proposed localization
system in a freeway, consisting of RFID assisted GPS and single peer local-
ization. Terminology used in this paper is following.
Figure 1: The proposed localization system: RF-GPS and single peer localization.
• Stationary RFID tag is a short radio range (e.g., 3∼4m) passive tag
affixed to a roadside unit, e.g., a speed sign, or attached to road surface.
7
The tag stores own absolute position data and transmits it to passing
vehicles.
• Mobile RFID tag is a passive tag attached to a vehicle. It stores the
vehicle’s ID.
• Mobile RFID reader is an interrogator on a vehicle that reads data
from either a stationary or a mobile RFID tag.
• Reference vehicle is a GPS vehicle that has obtained absolute position
data from a stationary RFID tag. It is the mobile version of a reference
point in the DGPS system. After calculating GPS error, it broadcasts
the error to neighbors via the 802.11 radio.
• GPS coords is coordinate data obtained from a GPS receiver on a ve-
hicle. Abs(olute) coords is coordinate data stored in a stationary RFID
tag reporting the exact position. Accurate coords (Accurate position
data) is vehicle’s position data within 3m error range. Diff(erential)
coords is difference between GPS coords and Accurate coords in a ve-
hicle. It represents GPS error at the point.
• Travel data is information of a vehicle’s movement including vehicle
ID, Accurate coords, speed, and orientation.
3.2. RF-GPS
RFID assisted GPS (RF-GPS) improves GPS position accuracy by ex-
ploiting reference vehicles on a road. Unlike the traditional DGPS system
that uses a fixed reference point, a GPS vehicle becomes a moving reference
8
point temporarily. Upon passing by a stationary RFID tag, the vehicle ac-
quires Abs coords and calculates GPS error value using own GPS coords.
Then, it broadcasts GPS error (Diff coords) to neighbors to support them to
correct their GPS coords.
Stationary RFID contact : Stationary RFID tags that store Abs coords
in the memory are installed on roadside units. When a (non-) GPS vehicle
travels into the radio range of the stationary tag, TS in Figure 1, the mobile
RFID reader on the vehicle obtains Abs coords from TS via RFID commu-
nication. Due to short radio range of RFID communication (3∼4m), only
those vehicles traveling on the lane closest to the road divider (or to the
curb) can read the Abs coords from TS. If the vehicle has a GPS system,
then it calculates Diff coords by subtracting Accurate coords from own GPS
coords. The vehicle now becomes a reference vehicle (VR in Figure 1). Note
that a non-GPS vehicle cannot be a reference vehicle since it does not have
GPS coords.
Broadcasting : The reference vehicle VR broadcasts Diff coords to one-hop
neighbors via the 802.11 radio. Since vehicles within the radio range, say,
on the freeway, are highly likely to have the same GPS error, a nearby GPS
vehicle (VG in Figure 1) can calculate their Accurate coords using Diff coords
received. VR does not repeat Diff coords broadcasting, and VG also does not
forward it to the neighbors. Furthermore, if they hear broadcast, vehicles will
not broadcast Diff coord within short interval even though they pass/get Abs
coords from the stationary RFID tags. We call this mechanism as broadcast
restriction. Therefore, the Diff coords broadcasting overhead is minimum.
However, it is still possible that receives multiple Diff coords from different
9
neighbors. Since two consecutive stationary RFID tags are apart longer than
the radio range (5km in the evaluation), multiple Diff coords usually come
from vehicles encountering the same stationary tag. In this case, VG ignores
newer Diff coords if they are the same value; otherwise, it uses the latest
information to calculate own Accurate coords. We also note that non-GPS
vehicles (VN) within the range cannot compute Accurate coords since they
do not have own GPS coords.
3.3. Single Peer Localization
In the basic RF-GPS scheme, a non-GPS vehicle obtains accurate position
only via the stationary RFID contact. To enhance positioning of non-GPS
vehicles, we propose a single peer localization scheme. When a non-GPS
vehicle encounters a GPS vehicle having Accurate coords, they establish two
wireless connections; mobile RFID contact and IEEE 802.11 peer-to-peer
link. Upon receiving data via the connections, the non-GPS vehicle computes
its accurate position.
Mobile RFID Contact : When a non-GPS vehicle (VN) passes a GPS ve-
hicle (VG), VN reads VG’s ID, IDG, via the mobile RFID contact: The mobile
RFID reader on VN accesses the mobile RFID tag on VG. For future calcula-
tion, VN records mobile RFID contact information, e.g., vehicle’s ID (IDG)
and contact time (TM), and then they establish peer-to-peer connection via
802.11.
802.11 Peer-to-Peer Connection: The mobile RFID contact triggers an
802.11 peer-to-peer connection: VG broadcasts a message containing its ID,
accurate position, and travel data. Because VG knows that VN is in the
vicinity (within 3 ∼ 4m), it reduces transmission power to minimize signal
10
interference. Upon receiving the message, VN records current time TG and
acquires data tuple, {ID′G, (xG, yG), SG, OG}, which are vehicle’s ID, Accu-
rate coords, speed, and orientation of VG. VN verifies the data comparing
two IDs, IDG and ID′G. VN also takes own travel data {SN , ON} to compute
its Accurate coords (xN , yN). Define ∆T = TG − TM , ∆O = OG − ON , and
∆L = width of the lane. Then, xN and yN can be evaluated by Equation
(1).
xN = xG +∆T ∗ (SG − SN ∗ cos (∆O))
yN = yG −∆T ∗ SN ∗ sin(∆O)−∆L(1)
Dead-Reckoning : Dead-Reckoning is used by a GPS vehicle when GPS
signal is not available. It estimates current position based on previously
recorded position and vehicles speeds over elapsed time and course. In the
RF-GPS system, non-GPS vehicles employ DR to compensate for unavailable
real-time GPS coordinate data. After obtaining Accurate coords from the
stationary RFID contact or the single peer localization, the non-GPS vehicle
keeps updating current position using DR. Yet, its error could accumulate
quickly and go beyond some satisfactory level. In the following subsection,
position accuracy of RF-GPS and DR is analyzed in detail.
3.4. Position Accuracy
In the maritime navigation, DGPS provides accurate position to a ship
up to 300km away from the reference point. The U.S. Department of Trans-
portation estimated error that is 0.67m per 100km from the coast-line ref-
erence point [11]. But, the same level of accuracy is not possible in the
VANET scenario. Obstacles on the road affect GPS signal delays, and thus
11
GPS receivers even in a small area could experience different GPS error.
The proposed mobile DGPS approach provides accurate position using dis-
tributed RFID tags and moving reference vehicles. The GPS error measured
by the reference vehicle is broadcasted only to one-hop neighbor vehicles via
802.11 radio. Thus, propagation of obsolete and erroneous position is pre-
vented. The accuracy level of the mobile DGPS approach is strongly related
to the frequency of GPS error measurement, and thus DGPS error can be
reduced by pervasive deployment of stationary RFID tags that provide Abs
coords. In addition, traffic pattern is also critical to DGPS error.
Many researchers have investigated and published GPS and DR errors
in the literature. Shengbo et al. [12] show that the position estimation
error in a conventional DR scheme is around 1%∼2% of travel distance.
According to such error rate, around 20m of position error occurs in 30s at
100Km/h vehicle speed. The estimated error rate in a VANET scenario,
however, could decrease, because vehicles travel only along the roads and
rarely turn their orientation sharply. Specifically, on a freeway, the estimated
error rate goes down to 0.3% of travel distance [7]. Say, there would be 2.5m
estimation errors in 30s at 100Km/h vehicle speed. We expect 3m error
range in DGPS, which is the width of a lane. At 100Km/h speed, 3m error
occurs after driving 150m with 2% error rate and 1,000m with 0.3% error
rate. Once calculating elapsed time, DR can keep within 3m position error
for 5.39s and 35.97s with 2% and 0.3% of error rate, respectively. Similarly,
to support less than 3m position error at 60km/h vehicle speed, an accurate
position data should be renewed every 8.98s and 59.88s for 2% and 0.3%
of estimation error rate, respectively. On a freeway, vehicles drive fast, but
12
they do not turn their orientation abruptly, which cuts down the estimation
error. In an urban street, on the other hand, frequent orientation changes
happen. However, it is expected that the average vehicle speed would be
less than 60km/h, which cuts down the estimation error, too. We take into
account these conditions in our experimental evaluation. For example, our
experiment monitors how often vehicles can update their position, and we
compare the measured update intervals with our threshold values, 36s and
60s (these values are driven from 0.3% error rate with 100km/h and 60km/h
speed, respectively).
4. Simulation Experiment
We implement and validate RF-GPS using QualNet [13], a packet level
network simulator. We design a 4-lane freeway in which all vehicles drive in
one direction at various speeds. Vehicles are installed with 4m range RFID
system and 250m range 802.11 radio. Table 1 describes the default values
of scenario parameters, and simulation experiments use these values unless
explicitly stated. We measure the latency until a vehicle acquires the first
accurate position data. In addition, average interval of position acquisition
is also measured to assess the ability of non-GPS vehicles to maintain their
accurate position data.
To evaluate the proposed localization system, we design four different
scenarios. First, we run simulations with different traffic volumes on the road.
Next, the percentage of non-GPS vehicles varies. Third, we run simulations
with different speed ranges. Last, fraction of RFID-enabled vehicles varies.
13
Scenario parameter Default value
Number of vehicles 200
Percentage of Non-GPS vehicles 50%
Speed range of vehicles 20∼30m/s
Interval of stationary RFID tag 5km
Table 1: Scenario Parameters
(a) CDF of getting accurate position (b) CDF of non-GPS vehicles
Figure 2: Traffic volume.
4.1. Traffic Volume
In this experiment, 50∼400 vehicles are deployed on 5-km-long 4-lane
freeway; thus average inter-vehicle distances are 50m ∼ 400m. Figure 2(a)
illustrates that the more vehicles on the road, the more opportunities to
obtain accurate position data. If the number of vehicles increases, there
are more GPS vehicles (50% of total vehicles) that have accurate position,
and thus non-GPS vehicles are able to get more chances to acquire Accurate
coords via single peer localization. Figure 2(b) demonstrates how often a non-
GPS vehicle can renew accurate position data. 60.2% of non-GPS vehicles
successfully renew it within 36s in the case of 200 vehicles; this value jumps
to 85.2% (with 300 vehicles) and to 96.5% (with 400 vehicles). Within 60s
14
interval, more than 83% of non-GPS vehicles renew accurate position when
there are more than 200 vehicles. This means that most non-GPS vehicles
can obtain their accurate position by updating every 60s via single point
localization and Dead-Reckoning.
4.2. Fraction of Non-GPS Vehicle
In this experiment, we change the percentage of non-GPS vehicles from
0% to 80%. Increasing number of non-GPS vehicles affects the latency per-
formance, as shown in Figure 3(a), since the number of reference vehicles
decreases. Figure 3(b) represents update intervals of accurate position in
terms of the population of non-GPS vehicles. The interval value increases
slowly until 65% of non-GPS vehicles, but it rapidly increases thereafter so
that the interval is too long to keep accurate position.
(a) Probability of getting accurate posi-
tion in GPS vehicles and non-GPS vehicles
(b) Mean and median of renewing accu-
rate position in a non-GPS vehicle
Figure 3: Fraction of non-GPS vehicles.
4.3. Impact of Speed Variables
This experiment investigates the effect of vehicle speed on localization
performance. We divide speed ranges into three groups; 15∼30m/s, 20∼30m/s,
15
(a) Speed configuration (b) Cumulative percentage of non-GPS ve-
hicles
Figure 4: Speed variables.
and 25∼30m/s. In addition, each group has two different settings: first, all
vehicles randomly select speed from the given speed, and the other case is
that lanes are divided faster lane and slower lane (vehicle speed is still se-
lected from the given range). Figure 4(a) depicts the speed configuration.
The lines show the speed ranges, and the small rectangles in the middle of
the lines are average speed of vehicles. The bars represent standard deviation
indicating speed variations. Figure 4(b) exhibits the average update inter-
val of accurate position in non-GPS vehicles. High speed and large speed
variation shorten the update interval since this configuration provides more
chance to encounter GPS vehicles.
4.4. Fraction of RFID-Enabled Vehicle
In this experiment, we assume all vehicles have a GPS system, but an
RFID system is installed on the portions of vehicles. Other parameters are
same to those in Table 1. In Figure 5(a), the performance of RF-GPS de-
grades as the number of RFID-enabled vehicles decreases. With 20% of
RFID-enabled vehicles, performance degrades 60 times compared to the 100%
16
(a) Probability of getting accurate posi-
tion in GPS vehicles
(b) Messages on RF-GPS
Figure 5: Fraction of RFID-enabled vehicles.
case, mainly because the number of broadcast messages containing Diff co-
ords decreases. This reasoning is enumerated with the results in Figure 5(b),
showing that the average number of broadcast messages proportionally in-
creases to the number of the RFID-enabled vehicles. The update intervals of
Diff coords are also shown in Figure 5(b) (see the third bars). If the RFID
system is installed on 80% of vehicles, each GPS vehicle receives Diff coords
and thus corrects GPS error every 45s. The results imply that RF-GPS can
maintain position accuracy even in the GPS deprived zones provided if there
are sufficient numbers of RF-GPS vehicles and stationary RFID tags. For
example, properly deployed stationary RFID tags in a tunnel can help GPS
vehicles maintain accurate position.
5. RFID Test-bed Experiment
The goal of our test-bed experiments is to prove the feasibility of RFID
communication in a highly mobile VANET scenario. We install an RFID
reader at the front bumper of a vehicle and place RFID tags on a road. We
17
Table 2: Hardware specification of the used RFID system.
RFID reader
Frequency 910∼914MHz
RF power 4W EIRP
Read distance Less than 5m
Modulation ASK
Radio access FHSS
RFID reader
antenna
Angle 60◦(3dB)
Gain 6dBi
Size 215(W)×420(L)×55(H)
Forward FW Jamie McMaster
RFID tagData 64bit
Data rate 256kbps
deploy RFID tags on road surface instead of the road side unit because this
setting simulates the harshest environment for RFID communication. During
experiments, we measure RFID read rate while the vehicle is driving over the
RFID tags.
5.1. RFID System Specification
We select a UHF RFID system because of its long range and fast trans-
mission time. Table 2 summarizes the specification of the used RFID sys-
tem. The reader is KIS900RE operating in 900MHz-914MHz [14]. It pro-
vides anti-collision algorithm for multiple readers using Frequency-Hopping
Spread Spectrum (FHSS) in 200kHz bandwidth. An RFID reader antenna
KIS900AE has 60◦ of angle and 6dBi of gain. The EM4222 chip is used in
an RFID tag to transmit 64bit data at 256kbps rate. For anti-collision, each
18
Figure 6: RFID system: reader, reader antenna, and tag.
tag gives random jitter, pause time, before sending data out. The maximum
pause time is 62.5ms. Figure 6 shows the RFID system including a computer
for gathering and processing RFID data.
5.2. RFID Communication in VANET
In the VANET scenario, an RFID reader and an RFID tag encounter
during a very short time period since the reader at the vehicle moves at a
very high speed. We define RFID read latency as a required time period
during which the reader successfully obtains the tag data after it meets a
valid tag.
To compute the read latency, we use 68◦ of an RFID reader angle [9], and
Figure 7 presents the corresponding RFID read area where a signal from the
RFID reader could reach. The width (x1) and the length (x2) of the area are
19
Figure 7: RFID read area.
calculated by Equation 2.
x1 = 2× h× tan 34◦
x2 =h
tan(56◦ + θ◦)+
h
tan(56◦ − θ◦)
(2)
where h and θ are the height and the pitch angle of the reader antenna
(−56◦ < θ < 56◦), respectively. If h = 37.5cm and θ = 45◦, then we
can compute x1 = 58.58cm and x2 = 185.63cm. Using these values, the
maximum time, during which a tag can stay in the moving RFID read area,
is derived in terms of vehicle speed as shown in Table 3. The values in the
second column “Theoretical values” are computed by Equation 2, whereas
those in the next column “Experimental values” result from our experiments.
A big gap between two values is mainly attributed that the RFID signal range
in the real world (i.e., in the experiment) is shorter than theoretical values.
For instance, x2 value is only 1m instead of 1.85m in practice.
Data transmission rate at the RFID tag introduces another issue. Accord-
20
Table 3: Moving speed of RFID read area (time to pass over a fixed RFID tag with
h = 37.5cm and θ = 45◦).
Speed [km/h] Theoretical value[s] Experiment value[s]
10 0.665 0.360
20 0.332 0.180
30 0.222 0.120
40 0.166 0.090
50 0.133 0.072
60 0.111 0.060
70 0.095 0.051
80 0.083 0.045
90 0.074 0.040
100 0.067 0.036
21
Figure 8: Laboratory experiment setting. The reader antenna is fixed on a 30cm-height
iron frame and tag is placed on the floor.
ing to specification, the tag has 256kbps data rate, and it takes 0.22ms to
transmit 64bit data. However, in our experiment, the average read latency
records 38.89ms because of the pause time at the tag, i.e., the maximum
pause time is 62.5ms. To solve this problem, we propose RFID performance
enhancement technique and antenna setting in the next subsection.
5.3. Laboratory Experiment and RFID System Installation
To investigate the best setting that enhances RFID performance, we con-
duct laboratory experiments. We install a reader antenna with h = 30cm
and pitch angle, θ = 30◦ as shown in Figure 8.
5.3.1. RFID Reader Antenna
To improve read rate, we install dual RFID antennas as shown in Figure
9. If one reader antenna is mounted, the read area is 86cm in width while the
width is extended to 130cm with dual antennas. Thus, RFID read rate is also
expected to increase. As shown in Figure 10, a vehicle can contact tags even
22
Figure 9: Measured RFID read area (18ms read latency).
Figure 10: Influence of dual RFID reader antennas.
though it deviates from the center of a lane. In the test-bed experiments,
dual RFID reader antennas is used; otherwise explicitly stated.
The height h and the angle θ of the reader are also critical for RFID
performance, since they determine beam shape and direction. Through ex-
tensive experiments in our laboratory setting, we find that RFID read latency
is within 18ms with 0 ∼ 30cm height and 20◦ ∼ 40◦ angle. For simple in-
stallation, the reader is placed at 30cm height and with 30◦ angle in the test
vehicle.
23
Figure 11: RFID tag cluster models.
Figure 12: Average read latency with four RFID tag cluster models.
5.3.2. RFID Tag Antenna
Like the dual reader antenna, we design an RFID tag cluster to enhance
communication reliability and to improve read latency. Because of the ran-
dom pause time, each tag in a cluster sends back signal at different time. In
24
Table 4: Parameters for deployment.
Reader antenna value RFID tag value
Numbers 1 or 2 Yaw angle 0◦
Height 30cm Pitch angle 0◦
Pitch angle 30◦ Number of tags in Cluster 3 1, 3, or 4
our study, the first response from tags in the cluster is considered as RFID
read latency. Note that the increasing number of tags does not always guar-
antee better read latency, because too many tags in the cluster could lead to
collision between tags inside the cluster.
We consider four types of RFID tag clusters as shown in Figure 11. A
tag consists of a tag chip (a black spot at the center) and a dipole antenna.
Cluster 1 and 2 represent horizontal and vertical integration, respectively. In
the next, a space is given between two neighboring tags, which is expected
to enhance the signal sensitivity. In Cluster 4, the tag chips share one dipole
antenna. We place each cluster within the target read distance (0.5m) and
measure read latency as we increase the number of tags in a cluster from 1
to 5. Figure 12 shows the average read latency with four RFID tag cluster
models. The results verify that as a cluster includes more tags, read latency
shortens. In Cluster 2, 3, and 4, the time values go down below 18ms when
there are 2, 3, and 4 member tags. With 5 member tags, however, perfor-
mance gets worse due to tag collision. For our experiments, we use Cluster 3
having 3 or 4 members of RFID tags since they show the best performance.
25
6. Test-bed Experiment Results
Figure 13(a) displays a test vehicle. The RFID reader antenna is mounted
on the front bumper, which is connected to the location determination server.
Figure 13: Testbed: RFID system, a vehicle, and a road.
26
Table 5: Test scenarios with variables.
Test1 Test2 Test3 Test4 Test5 Test6
Number of RFID reader antenna 1 2 2 2 2 2
Number of RFID tags in Cluster 3 1 1 3 4 3 4
Interval of neighboring RFID tags 2m 2m 2m 2m 5m 5m
Maximum speed [km/h] ∼100 ∼80 ∼80 ∼80 ∼80 ∼100
The server allows a user to monitor RFID data read in real-time while driving.
The tags are attached to the surface of the test road which is pictured in
Figure 13(c), and Figure 13(b) depicts the read area. Default parameters for
both the reader antenna(s) and the tags are denoted in Table 4. We conduct
6 tests by changing 4 variables as represented in Table 5.
In the experiments, we measure average duplication read and average
read rate. The average duplication read is the averaged number of receiving
duplicated data from the same tag, and the average read rate is fraction of
RFID tags successfully read over the total number of encountered tags.
6.1. Effect of Antenna Diversity
Test 1 and Test 2 share the same configuration except the number of
RFID reader antenna. In Figure 14(a), single and dual antenna systems
show similar duplication number and they decline as vehicle speed increases.
Furthermore, in Figure 14(b), average read rates drop rapidly as vehicles’
speed becoms faster. After 40km/h of speed, the read rates drop to 40 ∼
50%. Unfortunately, we could not find a big difference between them, since
the test vehicle travels along the center of the lane. But, as shown in Figure
10, the dual reader antenna widens the read area so we believe that tag miss
is decreased in the real world.
27
Figure 14: Results from Test 1 and Test 2.
28
Figure 15: Experiments with dual RFID reader antenna (Test 2 to Test 6).
6.2. Effect of Tag Multiplicity
With dual RFID reader antennas, we vary the number of member tags
in Cluster 3 and the interval between tag clusters, i.e., Test 2 to Test 6. The
results are shown in Figure 15.
In Figure 15(a), the duplication read is proportional to the number of tags.
As the speed becomes faster, the duplication read goes down. At 80km/h, all
the cases result in less than 5 times. The low number of duplication read is
29
highly connected to performance as shown in Figure 15(b); it shows around
40% of RFID read rate. At 70km/h, on the other hand, 80% and 94% are
shown when using multiple tags. Upon comparing Test 3 and Test 4 directly,
a cluster having 3 member tags performs slightly better than 4-tag-cluster.
Until 70km/h, 3-tag-cluster’s RFID read rate maintains over 90%. In total,
we can conclude that an RFID tag cluster model contributes to enhancement
of the RFID read rate.
A comparison between Test 3 and Test 5, i.e., different tag intervals,
demonstrates that 5m of interval performs a bit better than 2m. At 80km/h,
the duplication reads in both cases go below 5 times, but 5m of interval out-
performs 2m by 37.1% in the RFID read rate. This implies that deployment
of tags at a short interval degrades performance at a high speed.
In order to clarify the effect of the tag distance interval, we compare
Test 6 and Test 4 results. The duplication read in the case of 5m interval
does not easily fall down below 5 till 100km/h. The RFID read rate also
shows more than 80% for all the speed parameters except for at 70km/h.
In conclusion, 4 member tags deployed every 5m, performs the best in our
test-bed experiments.
7. Conclusion
This paper has presented a new localization system in VANETs, RF-
GPS, which exploits a RFID system. It develops the mobile version of the
DGPS system. RF-GPS calibrates GPS error and thus allows a vehicle to
compute its accurate position. Moreover, a vehicle, which does not have
a GPS receiver or cannot use the receiver temporarily, is also able to es-
30
timate its accurate position with the single peer localization scheme. The
proposed RFID-assisted localization system has been evaluated extensively
via simulations and real world experiments. The results from QualNet-based
simulations showed the impact of traffic volume and speed variations on the
performance of the RF-GPS system. We also estimated the consequences
of penetration of a GPS system and an RFID system over the road. The
test-bed experiments focused more on the feasibility of the RFID system on
a vehicular environment. In particular, we evaluated the reliability of RFID
communication over various vehicle speed ranges. The results showed that
the off-the-shelf commodities could tolerate fast-moving vehicles: An RFID
reader on a vehicle can access data of an RFID tag on road surface while the
vehicle drives fast. The simulations and the real world experiments together
show feasibility and performance of the proposed RF-GPS system.
References
[1] A. Varshavsky, M. Chen, E. deLara, J. Froehlich, D. Haehnel, J. High-
tower, A. LaMarca, F. Potter, T. Sohn, K. Tang, I. Smith, Are gsm
phones the solution for localization?, in: 7th IEEE Workshop on Mobile
Computing Systems and Applications (WMCSA), 2006, pp. 20–28.
[2] V. Kukshya, H. Krishnan, C. Kellum, Design of a system solution for
relative positioning of vehicles using vehicle-to-vehicle radio communi-
cations during gps outages, in: IEEE Vehicular Technology Conference
- Fall, Vol. 2, 2005, pp. 1313–1317.
[3] R. Schubert, M. Schlingelhof, H. Cramer, G. Wanielik, Accurate po-
31
sitioning for vehicular safety applications - the safespot approach, in:
IEEE Vehicular Technology Conference - Spring, Dublin, Ireland, 2007.
[4] A. Boukerche, H. Oliveira, E. Nakamura, A. Loureiro, Vehicular ad hoc
networks: A new challenge for localization-based systems, in: IEEE
Computer Communications, Vol. 31, 2008, pp. 2838–2849.
[5] N. Drawil, Improving the vanet vehicles’ localization accuracy using
gps receiver in multipath environments, Master’s thesis, University of
Waterloo (2007).
[6] Navigation center differential gps, http://www.navcen.uscg.gov/?pageName=dgpsMain.
[7] T. King, H. Fubler, M. Transier, W. Effelsberg, Dead-reckoning for
position-based forwarding on highways, in: International Workshop on
Intelligent Transportation (WIT), 2006, pp. 199–204.
[8] R. B. S. (RBS), in: http://www.roadbeacon.com/.
[9] H. Chon, S. Jun, H. Jung, S. An, Using rfid for accurate positioning, in:
International Symposium on GNSS/GPS 2004, 2004.
[10] Standard Specification for Telecommunications and Information Ex-
change Between Roadside and Vehicle Systems - 5 GHz Band Dedicated
Short Range Communications (DSRC) Medium Access Control (MAC)
and Physical Layer (PHY) Specifications (Sept. 2003).
[11] 2001 federal radionavigation plan, Tech. rep., Department of Trans-
portation and Department of Defense (March 2002).
32
[12] Q. Shengbo, D. Keliang, L. Qingli, An effective gps/dr device and al-
gorithm used in vehicle positioning system, in: IEEE ITS Conference,
2003.
[13] Scalable Networs Inc., QualNet, http://www.scalble-networks.com.
[14] Kiscom, in: http://www.kiscom.co.kr/.
33