Radio Interferometric Tracking of Mobile Wireless Nodes
Branislav Kusy
Janos Sallai
Gyorgy Balogh
Akos Ledeczi
Vanderbilt University, USA
akos@isis.vanderbilt.edu
Vladimir Protopopescu
Johnny Tolliver
Frank DeNap
Oak Ridge National Lab, USA
protopopesva@ornl.gov
Morey Parang
University of Tennessee, USA
parangm@ornl.gov
ABSTRACT
Location-awareness is an important requirement for many
mobile wireless applications today. When GPS is not appli-
cable because of the required precision and/or the resource
constraints on the hardware platform, radio interferometric
ranging may offer an alternative. In this paper, we present a
technique that enables the precise tracking of multiple wire-
less nodes simultaneously. It relies on multiple infrastruc-
ture nodes deployed at known locations measuring the posi-
tion of tracked mobile nodes using radio interferometry. In
addition to location information, the approach also provides
node velocity estimates by measuring the Doppler shift of
the interference signal. The performance of the technique is
evaluated using a prototype implementation on mote-class
wireless sensor nodes. Finally, a possible application sce-
nario of dirty bomb detection in a football stadium is briefly
described.
Categories and Subject Descriptors: C.2.4[Computer-
Communications Networks]:Distributed Systems
General Terms: Algorithms, Experimentation, Theory
Keywords: Wireless Sensor Networks, Radio Interferome-
try, Tracking, Localization, Location-Awareness, Mobility
Acknowledgments: This work was supported in part by
TRUST (The Team for Research in Ubiquitous Secure Tech-
nology, NSF award number CCF-0424422), ARO MURI
grant SA5212-11087 and a Vanderbilt Discovery Grant. We
would like to express our gratitude to Andras Nadas, Miklos
Maroti and Peter Volgyesi for their valuable contributions
to this work. We are indebted to Jay Lowe for providing
access to the Vanderbilt Football Stadium, the anonymous
reviewers and our shepherd, Venkat Padmanabhan, for their
constructive comments.
1. INTRODUCTION
The Global Positioning System (GPS) has made the tran-
sition from a few early adopters to a sizeable percentage of
the population in many technologically advanced countries
around the world. Most new cars are offered with built-in
GPS systems. Top-of-the-line cell phones have built-in GPS
Copy right 2007 As s o ciation f or Com puting M achinery. ACM acknowl-
edges that this contribution was authored or co-authored by an employee,
contractor or affiliate of the U.S. Government. As such, the Government re-
tains a nonexclusive, royalty-free right to publish or reproduce this article,
or to allow others to do so, for Government purposes only.
MobiSys ’ 07, June 11-14, 2007, San Juan, Puerto Rico, USA.
Copyright 2007 ACM 978-1-59593-614-1/07/0006 ...$5.00.
receivers. On the commercial side, trucking companies have
adopted GPS for asset tracking, for example. And of course,
the military is using it everywhere from soldier navigation
to smart munition guidance. In fact, the European Union
announced that they will build a competing system called
Galileo to reduce their dependence on American technology.
The advantages of GPS are numerous. It provides accu-
rate positioning anywhere on Earth. Receivers are relatively
small, cheap, and power conscious. The service itself is free.
On the other hand, GPS has shortcomings too, making it
less applicable to certain application domains. The accuracy
of traditional GPS is a few meters. Differential GPS can get
two orders of magnitude more precise, but it relies on ex-
pensive hardware. Typical handheld GPS receivers last only
a few hours on a single charge. Recently, single chip GPS
receivers became available, but they still cost tens of dollars
and typically consume 50-100mW. Moreover, GPS does not
work well indoors, in cluttered urban environments or un-
der dense foliage. Finally, GPS can be jammed making it
vulnerable in military applications.
Hence, when a wireless mobile application requires geolo-
cation information with one or more of the following proper-
ties: (1) high precision (a few centimeters), (2) low per node
cost (a few dollars), (3) low power consumption (10-50mW),
and (4) robustness to environmental conditions, then alter-
native solutions need to be applied. We have recently intro-
duced a novel localization method based on radio interfer-
ometry [19, 13] primarily targeted at wireless sensor network
node localization for static deployments. Later we applied
the same underlying technique to tracking a single mobile
node [12]. Here we extend this approach to multiple tracked
objects and to estimate the velocity, along with the loca-
tions of the tracked objects. We also describe a dirty bomb
detection and localization system that utilizes the technique.
The underlying idea of radio interferometric ranging is to
have two nodes transmit unmodulated radio waves at close
frequencies, thereby creating an interference field. Measur-
ing the phase offsets at different points provides information
on the location of those points and that of the transmitters.
If we have a set of static nodes at unknown positions and
we make enough measurements, the relative coordinates of
the nodes can be reconstructed. For tracking, we can place
several nodes at known positions to create reference points
similar in functionality to GPS satellites. The system then
uses interferometric ranging to determine the unknown po-
sitions of mobile nodes.
Radio interferometric localization and tracking satisfies
three of the requirements above: high precision, low cost,
139

and low power. It can achieve centimeter scale accuracy.
Currently, it is implemented on COTS Crossbow Mica2
mote devices with no additional hardware. We utilize the
mote’s Texas Instruments CC1000 radio which costs about
$3. This radio could be added to other hardware platforms
with an estimated incremental cost of less than $10. Its
receive power requirement is about 20 mW.
The disadvantage of our technique, though, is that it is
susceptible to RF multipath, so currently it does not work
well indoors. Even though we expect the interferometric
ranging to provide accurate localization in large indoors fa-
cilities, such as sports arenas where GPS would inevitably
fail, at this moment, we do not have an effective localization
algorithm for the environments with significant multipath.
Our initial research, however, shows promise in resolving
the multipath problem. If multipath is present, the phase
offset depends not only on the four distances between the
nodes, but also on the relative amplitudes of the reflected
signals compared to the direct signal, as well as the locations
of the points of reflection. We have mathematically modeled
the expected phase offset of the interferometric signal and
made the following observation: if we have a bound on the
number of reflected signal components and have enough car-
rier frequencies at our disposal, we will eventually be able to
collect more data than the number of unknowns in our the-
oretical model. This is because with each frequency, we can
accurately measure two values: the relative phase offset and
the relative signal strength between the two receivers. Note
that the signal strength does contain information about the
reflected paths. If we measure at slightly different frequen-
cies and hence, wavelengths, the composite signal, consisting
of the line of sight component and multiple reflected ones,
will have different amplitudes at different frequencies be-
cause of the different phases in the summation. As the num-
ber of unknowns in our model does not increase, the system
will become solvable if enough measurements are available.
Another advantage of our technique is that it is relatively
easy to experiment with multiple antenna designs, differ-
ent radio frequencies, and densities of the infrastructure
nodes. Furthermore, the potential of using radio interfer-
ometry with lower frequencies which are less susceptible to
reflections due to their longer wavelength may significantly
simplify the algorithms for indoor localization. To summa-
rize, we believe that radio interferometric ranging is an im-
portant technique showing great promise to overcome the
main GPS drawback – the lack of indoor localization. Note,
however, that we do not claim that our system is a replace-
ment of GPS, as we can only cover a geographically limited
area.
The rest of the paper is organized as follows. First we
review related work and briefly summarize the theory be-
hind interferometric ranging and highlight previous results.
Then we present the description, analysis and evaluation of
the enhanced technology for tracking multiple moving nodes
simultaneously. We also describe application of our track-
ing techniques to a dirty bomb detection system and its
demonstration in a football stadium. Finally, we conclude
the paper with a discussion and future research directions.
2. RELATED WORK
Object tracking systems are used to determine the tra-
jectory of one or more moving targets from partial location
information provided by sensors. In general, the location
of an object in the physical world is inferred from different
sensed signals, such as acoustic, infrared, or seismic signals.
Algorithms for object detection, classification and identifi-
cation are, therefore, important part of tracking systems.
For example, Acoustic ENSBox [7] was used to detect and
localize different animal species such as marmots or dusky
antbirds. However, the majority of the wireless sensor net-
work (WSN) based tracking systems eliminate the need of
object detection by assuming that the tracked nodes coop-
erate with the system [8, 23, 9, 1].
Cooperative tracking systems need to address multiple is-
sues. First, the systems need to decide which set of sensors
to activate to track a particular target and where to send
the collected location data. This problem can be solved in
an information centric [28] or location centric [3] way. Next,
the activated sensors collect ranging measurements to esti-
mate the location of the target, utilizing acoustic signals [7,
25], radio signal strength [9, 1, 27], radio time of flight [6,
16], or phase difference [19]. Finally, filtering and smoothing
techniques, such as extended Kalman filters [2, 22] or par-
ticle filters [26, 11], are applied to the calculated location
estimates to remove the effects of the measurement noise
and improve tracking accuracy.
In our work, we concentrate solely on the localization
and ranging problem of the tracking systems. In partic-
ular, we focus on obtaining highly accurate location esti-
mates that can be calculated fast, to allow for high loca-
tion update rates. In this regard, the best published results
were achieved with systems based on acoustic ranging tech-
niques. Acoustic ENSBox [7] achieved 5 cm average 2D
localization error in a large scale experiment, with 1.5 de-
gree average orientation error. Cricket nodes were used to
achieve a few cm 2D localization error with 1 Hz update rate
at short ranges [21]. SLAT achieved 7 cm 3D localization
error indoors [25]. While the achieved localization accuracy
is impressive, these systems require hardware components
which have considerable cost, form factor, and power con-
sumption, if they are required to work over large distances.
State-of-the-art GPS receivers are facing similar problems,
despite their relatively low price ($50) and power consump-
tion (50—100 mW). Due to their small hardware cost over-
head, the RF based systems that use the radio chips included
on the sensor nodes are an attractive option for tracking
systems. Unfortunately, radio signal strength systems do
not show a great promise in achieving high location accu-
racy [27]. A promising sensor network RF time of flight
solution described in [16] proposed to utilize pairwise round
trip time of flight measurements to relax the strict global
time-synchronization requirement and achieved an average
error of 0.9 m outdoors and 2 m indoors. Despite the sig-
nificant progress in many of these ranging approaches, we
believe that the radio interferometric ranging provides a su-
perior alternative due to its low-cost (requires only CC1000
radio chip), low power (20 mW receiving power), and high
accuracy (4 cm) over large ranges (100 m) [13].
The main problem of tracking moving objects is that the
target changes its location during the measurement as the
ranging measurement takes a finite amount of time. One
solution is to divide the global space-time region into space-
time cells [17]. The size of these cells should approximate
a region where the signature of the target remains nearly
constant. Sequential Monte Carlo localization was shown to
achieve improved accuracy of localization when utilizing the
140