An RF-based System For Tracking Transceiver-Free Objects: Dian Zhang, Jian Ma, Quanbin Chen, and Lionel M. Ni
An RF-based System For Tracking Transceiver-Free Objects: Dian Zhang, Jian Ma, Quanbin Chen, and Lionel M. Ni
An RF-based System For Tracking Transceiver-Free Objects: Dian Zhang, Jian Ma, Quanbin Chen, and Lionel M. Ni
1. Introduction
Real-time tracking of moving objects is highly in
demand and important to many applications, such as
vehicle tracking [2], battlefield surveillance [18], animal habitat monitoring [25], and patient tracking in
hospitals [5]. It has attracted a great deal of attention in
various research communities. GPS [21] is a technology well known for its accuracy. However, GPS only
works in outdoor environments without satellite signals
being blocked. Moving object tracking in indoor environments is more complicated and several technologies, such as video, pressure, infrared, and ultrasound,
have been proposed. These technologies are usually
costly, including infrastructure, deployment, and maintenance, and may have some restrictions placed on the
environments in which they are applied. For example,
1
In fact, we only take advantage of the communication behavior of sensor nodes, not the associated sensing device.
2. Related Work
Other than the RF technology, some other technologies might be used to track moving objects.
Video. In the computer vision area, by applying
many video-based algorithms on sequences of images
from a camera or synchronized images captured from
multiple cameras, moving people can be tracked [4] or
the number of people in a certain area can be counted
[22]. However, this technology is expensive and fails
in darkness. Moreover, it may violate the privacy of
people.
Infrared. In the context of infrared wireless networks, the limited range of an infrared network, which
facilitates the user location, is a handicap in providing
ubiquitous coverage. For example, some companies,
such as Acorel [1], already have some products based
on infrared technology. However, they can only record
how many people enter or exist in a restricted area by
monitoring that areas access, such as the door of a
room. Moreover, this technology requires careful and
dense deployment, and does not work in a more complicated environment.
Pressure. An optional localization technology is to
use acceleration and air pressure sensors to detect peoples footsteps [17]. The obvious drawbacks of this
technology are the high cost and need for careful deployment.
Ultrasound. The technologies using ultrasonic sensor networks usually require the target object to carry a
transmitter or receiver. Most of them adopt an ultrasonic Time-Of-Flight (TOF) method to obtain location
information. For example, in the Bat Ultrasonic Location System [19], the target object must carry a Bat
(transmitter), while this Bat periodically emits a short
pulse of ultrasound. The Cricket Location-Support
System [15] uses a combination of ultrasound and RF
to provide a location-support service. It allows applications running on mobile and static nodes to learn their
physical locations by using listeners (receivers) that
hear and analyze information from beacons spread
throughout the building.
As mentioned earlier, three indoor localization
models have been studied using the RF technology.
They are 802.11, active RFID, and wireless sensor
networks.
802.11. 802.11 technologies use a standard network
adapter to measure signal strengths. This approach
utilizes signal strength information gathered from
known multiple access points to locate objects. Some
approaches build a radio map of signal strength value
for each location [3] [24]. All these methods need the
target object to carry a radio detector (receiver).
RFID. LANDMARC can localize the target object
Pt G t G r 2
(1)
(4 )2 d 2
Pt Gt G r 2
(2)
(4 )2 (d 2 + 4h 2 )
Suppose the intensity magnitudes of the line-ofsight path and the ground reflection path are E1 and E2,
respectively. Eother is the intensity magnitude of other
radio propagation paths, such as the reflections of the
surroundings. The total received power by the receiver
P0 in the indoor static environment is expressed below.
2
(3)
P0 E1 + E 2 + E other
This value is almost stable in a static environment although noise does exist.
When a finite-sized object comes into this static environment, the target object will scatter the incident
power in various directions as illustrated in Figure 2.
According to radar equation [14], the received power
influenced by the target is
P G G 2
Pobj = t t 3 r2 2
(4)
(4 ) r1 r2
where r1 is the distance from the transmitter to the target object, r2 is the distance from the target object to
the receiver, and is the radar cross section of the target object. The radar cross section is defined as the
ratio of scattered power to incident power density.
According to scattering theory [6], in the dynamic
environment, E1, E2 and Eother will not change their values. The total received power by the receiver is the
sum of incident and scatted waves as below
P E1 + E 2 + E other + E obj
(5)
2
MPL
PL(1m to MPL)
PL(2m to MPL)
PL(3m to MPL)
1.8
1.6
1.4
Dynamic of RSSI(dB)
1.6
1.2
1
0.8
0.6
0.4
(6)
That is, if the condition Pobj << P0 is satisfied, the difference of signal strength between static environment
and dynamic environment is approximately proportional to Pobj. Here we omit the proof part because of
the page limitation.
m
i =1
b ai
m
r12 = x + h p2
2
r22 = x + + hp2
2
1.2
1
0.8
0.6
0.2
-2
-1
0
1
Distance to midpoint (m)
P Pobj
1.4
0.4
0.2
0
-3
MVL
VL(0.75m to MVL)
VL(1.5m to MVL)
VL(2.25m to MVL)
1.8
0
-3
-2
-1
0
1
Distance to midpoint (m)
2
2
d
d
r12 r22 = x + h 2p x + + h p2
2
2
= x 4 + 2 ( h p2
d2 2 d2
) x +
+ h p2
4
4
d2
+ h p2 d 2
= x 2 + h p2
4
midpoint of MPL or MVL has the largest RSSI dynamics. For each PL or each VL, although there are some
fluctuations at some positions due to noise behavior, in
general, the RSSI dynamics are larger as the person
moves closer to its midpoint. We also observe that
when the object position is on the MPL or MVL, the
RSSI dynamics are much larger than it is when the
person is at the same place on the other PLs or VLs.
From our experiments, we further verified that the
relationship between the person-position and the associated RSSI dynamics does obey the signal dynamic
property.
4. Methodology
Extending our signal dynamic model, we deploy
sensors in a regular 2D grid array, in which the positions of all the nodes are predetermined. Each node
broadcasts beacon messages periodically and listens to
the beacons from its neighbors as well.
The basic idea of our methodology is to detect the
RSSI dynamics due to the movement of objects. Because noise behavior sometimes can also generate
some RSSI dynamics, in order to reduce the influence
of noise, we set a dynamic threshold for each wireless
link. Only when the measured RSSI dynamics on a link
is larger than its threshold, we call the link as an influential link. Based on the signal dynamic property, the
influential links tend to cluster around the object position. Thus, we can employ these influential links to
locate the target objects.
In this section, first we present the midpoint and intersection algorithms, which can be applied to track a
single object without calibration. Then, in order to improve the accuracy and track multiple objects, we propose the best-cover algorithm, which requires calibration on the wireless links with the same distance.
midpoint algorithm utilizes the midpoints of the influential links to estimate the object location. We assign a
weight value to the midpoint of each influential link
according to its RSSI dynamic value. The object position can be calculated as the weighted average of these
midpoints positions.
An example is illustrated in Figure 7, where a
thicker influential link implies that it has a larger RSSI
dynamic value. For example, link ab is one of the influential links. m is the midpoint of link ab. From the
signal dynamic property, link ab having a larger RSSI
dynamic value means it is closer to the object position,
so is the midpoint m of link ab. In this example, there
are four midpoints on four influential links (some of
them are overlapped).
Since the coordinates of all the sensor nodes are
known in advance, for each influential link, the midpoint coordinate (xi, yi) can be calculated. The weight
of each influential link is pi, which is the RSSI dynamic value of the link. Assuming we have in total n
influential links, we compute the weighted average of
all the midpoints and get the object position coordinate
value (xobj, yobj) as below.
n
p x
i
xobj =
i =1
p y
i
i =1
yobj =
i =1
i =1
Since we only need to compute the midpoint coordinate of each influential link, the complexity of the
midpoint algorithm is O (n), where n is the number of
influential links.
4m between 2 nodes
1.6
MPL
MVL
1.4
1.2
1
0.8
0.6
sensor
0.4
0.2
0
-4
-3
-2
-1
xobj =
p
i =1
xi
, yobj =
i =1
p
i =1
yi
i =1
Since we need to calculate the intersection coordinate of each two influential links, the total algorithm
complexity is O (n2), where n is the number of influential links.
1m between 2 nodes
6m between 2 nodes
4m between 2 nodes
1.6
1.6
1.6
MPL
MVL
1
0.8
0.6
0.4
0.2
1.4
1.2
0
-2
MVL
1.4
1.4
1.2
1
0.8
0.6
0.4
-1
-0.5
0.5
1.5
0
-4
1.2
1
0.8
0.6
0.4
0.2
0.2
-1.5
MPL
MVL
MPL
-3
-2
-1
0
-5
-4
-3
-2
-1
5. Performance Evaluation
In this section, we first describe our experimental
setup and phases of the tracking system. Second, the
investigation of sensor distance and dynamic threshold
are given. Third, we compare the performance of the
three tracking algorithms proposed in the last section.
Lastly, the tests of a moving object and multiple objects are provided.
0.9
0.9
0.8
0.8
Midpoint
s + 0.1 dB
0.7
Midpoint
0.7
Intersection
Intersection
Best-cover
Best-cover
s= Variance
in static
s - 0.1 dB
s - 0.2 dB
0.6
CDF
0.6
CDF
s + 0.2 dB
2
1.8
1.6
1.4
1.2
1
0 .8
0 .6
0 .4
0 .2
0
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
Midpoint
Intersection Best-cover
0.5
0.1
0.5
1.5
2
2.5
3
Tracking error (m)
3.5
4.5
0.5
1.5
2
2.5
3
Tracking error (m)
3.5
4.5
Our methods utilize the RSSI dynamic value between the static environment and the dynamic environment to calculate the target object position. Among
the three algorithms, the midpoint and intersection algorithms can be used for applications without requiring
high accuracy and with only one moving object. If
some calibrations are performed in advance, which test
the dynamic behavior and measure the possible object
area by different sensor distances, the best-cover algorithm significantly beats the previous two. Furthermore, our system can also support moving object tracking. Even if there are multiple objects in our system, as
long as the objects are not tightly close to each other,
the best-cover algorithm is able to locate them. If the
objects are tightly close to each other, we would recognize them as one object.
The algorithm complexity relies on the number of
influential links. Even in a very large area covered by
the sensor grid, as long as the moving objects are not
crowded, the number of the influential links stays very
limited. For the midpoint and intersection algorithms,
the complexity is nearly constant. For the best-cover
algorithm, the complexity still depends on the side
length of the grid array and the scanning step.
As future work, we would like to try a larger area
covered by the sensor grid. Furthermore, we may try
other settings or irregular topologies instead of the grid
setting to deploy sensors. For example, we may consider hexagon structure for each cell in the deployment.
At each cell, the number of sensor pairs with the same
distance will be larger than that in the grid setting. This
may help to improve the accuracy of location estimation. Our solution to multiple moving objects is restricted. Some probabilistic models may be used to
analyze the dynamic behavior of the multiple moving
traces in the future.
Acknowledgements
China NSFC Grant 60533110, the National Basic Research Program of China (973 Program) under Grant
No. 2006CB303000, and the HKUST Digital Life Research Center Grant.
References
[1] ACOREL Corporation, http://www.acorel.com.
[2] Advanced Tracking Technologies, Inc., http://www.a
dvantrack.com.
[3] P. Bahl and V. N. Padmanabhan, "RADAR: an inbuilding RF-based user location and tracking system,"
in Proceedings of the Nineteenth Annual Joint Conference of the IEEE Computer and Communications Societies, 2000.
[4] Q. Cai and J. K. Aggarwal, "Automatic tracking of human motion in indoor scenes across multiple synchronized video streams," in Proceedings of the Sixth International Conference on Computer Vision, 1998.
[5] T. Gao, D. Greenspan, M. Welsh, R. R. Juang, and A.
Alm, "Vital signs monitoring and patient tracking over a
wireless network," in Proceedings of the Twentyseventh IEEE EMBS Annual International Conference,
2005.
[6] J. Jackson, "Classical electrodynamics," in 3rd ed., John
Wiley & Sons Inc, 1998.
[7] J. Ma, Q. Chen, D. Zhang, and L. M. Ni, "An empirical
study of radio signal strength in sensor networks using
MICA2 nodes," in Technical Report, Department of
Computer Science and Engineering, the Hong Kong
University of Science and Technology, 2006; http://
www.cse.ust.hk/~majian/papers/mica2rss_TR.pdf.
[8] M. Marti, P. Vlgyesi, S. Dra, B. Kus, A. Ndas, .
Ldeczi, G. Balogh, and K. Molnr, "Radio interferometric geolocation," in Proceedings of the Third International Conference on Embedded Networked Sensor
Systems, 2005.
[9] S. Meguerdichian, S. Slijepcevic, V. Karayan, and M.
Potkonjak. "Localized algorithms in wireless ad-hoc
networks: Location discovery and sensor exposure," in
Proceedings of the Second ACM International
Sympsium on Mobile Ad Hoc Networking and Computing, 2001.
[10] D. Moore, J. Leonard, D. Rus, and S. Teller, "Robust
distributed network localization with noisy range measurements," in Proceedings of the Second International
Conference on Embedded Networked Sensor Systems,
2004.
[11] L. M. Ni, Y. Liu, Y. C. Lau, and A. P. Patil, "LANDMARC: indoor location sensing using active RFID," in
Proceedings of the First IEEE International Conference
on Pervasive Computing and Communications, 2003.
[12] D. Niculescu and B. Nath. "Ad hoc positioning system
(APS)," in Proceedings of IEEE Global Telecommunications Conference, 2001.
[13] D. Niculescu and B. Nath. "Ad hoc positioning system
(APS) using AOA," in Proceedings of IEEE the Conference on Computer Communications, 2003.
[14] D. Pozar, "Microwave engineering," in 2nd ed., John
Wiley and Sons, Inc., 1998.
[15] N. B. Priyantha, A. Chakraborty, and H. Balakrishnan,
"The Cricket location-support system," in Proceedings
of the Sixth Annual International Conference on Mobile
Computing and Networking, 2000.
[16] T. S. Rappaport, "Wireless communications: principles
and practice," in 2nd ed., Upper Saddle River, NJ: Prentice Hall PTR, 2002.
[17] J. O. Robert and D. A. Gregory, "The smart floor: a
mechanism for natural user identification and tracking,"
in Proceedings of the CHI 2000 Conference on Human
Factors in Computing Systems, 2000.
[18] Rockwell Scientific Company, http://wins.rockwells
cientific. com/.
[19] R. Want, A. Hopper, V. Falcao, and J. Gibbons, "The
active badge location system," in ACM Transactions on
Information Systems, vol. 10, pp. 91-102, 1992.
[20] A. Woo, T. Tong, and D. Culler, "Taming the underlying challenges of reliable multihop routing in sensor
networks," in Proceeding of the First International Conference on Embedded Networked Sensor Systems, 2003.
[21] G. Xu, "GPS: Theory, algorithms and applications," in
Berlin: Springer-Verlag, 2003.
[22] D. B. Yang, H. H. Gonzalez-Banos, and L. J. Guibas,
"Counting people in crowds with a real-time network of
simple image sensors," in Proceedings of the Ninth
IEEE International Conference on Computer Vision,
2003.
[23] XBOW Corporation, "XBOW MICA2 mote specifications," http://www.xbow.com.
[24] J. Yin, Q. Yang, and L. M. Ni, "Adaptive temporal radio
maps for indoor location estimation," in Proceedings of
the Third IEEE International Conference on Pervasive
Computing and Communications, 2005.
[25] P. Zhang, C. M. Sadler, S. A. Lyon, and M. Martonosi,
"Hardware design experiences in ZebraNet," in the Second ACM Conference on Embedded Networked Sensor
Systems, 2004.