See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/224504639

A Novel Approach to Indoor RSSI Localization by Automatic Calibration of the

Wireless Propagation Model

Conference Paper · May 2009

DOI: 10.1109/VETECS.2009.5073315 · Source: IEEE Xplore

CITATIONS READS

79 718

4 authors:

Paolo Barsocchi Stefano Lenzi

Italian National Research Council Italian National Research Council
97 PUBLICATIONS   833 CITATIONS    15 PUBLICATIONS   275 CITATIONS   

SEE PROFILE SEE PROFILE

Stefano Chessa Gaetano Giunta

Università di Pisa Università Degli Studi Roma Tre
179 PUBLICATIONS   2,921 CITATIONS    177 PUBLICATIONS   1,004 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

RUBICON - Robotics UBIquitous COgnitive Network View project

IPIN2016 View project

All content following this page was uploaded by Stefano Chessa on 01 June 2014.

The user has requested enhancement of the downloaded file.

 A Novel Approach to Indoor RSSI Localization
by Automatic Calibration of the Wireless
Propagation Model
Paolo Barsocchi, Stefano Lenzi, Stefano Chessa, Member, IEEE, Gaetano Giunta, Member, IEEE.
Abstract— We propose a novel localization algorithm of mobile
sensors based on wireless sensor networks providing RSSI
measurements between the mobile and the fixed sensors
(anchors) in the network. The algorithm selects and weights the
RSSI measurements according to their strength, and it uses a
propagation model to transform RSSI measurements into
distances, in order to estimate the position of the mobile. The
algorithm also uses a virtual calibration method of the
propagation model that does not require human intervention. By
an experimental setup we show that the localization algorithm
increases the performance with respect to the commonly used
least mean square algorithm showing also how to achieve a
wished accuracy increasing the anchor density.
I. INTRODUCTION & RELATED WORK
Localization of devices and people has been recognized as one
of the main building block of context aware systems [1], [2],
[3], [4], which have their main application field in Ambient
Assisted Living (AAL) applications. In these scenarios, the
general solution based on Global Positioning System (GPS) is
unfortunately not available since it works only in outdoor
environments. On the other hand a viable solution to indoor
localization exploits wireless sensor networks [16]. Sensor
network-based solutions can estimate the (unknown) location
of mobile sensors (placed on the users) with respect to a set of
fixed sensor (called anchors), whose position is known, by
using two different approaches, either range-based or range-
free. The former is defined by protocols that use absolute
point-to-point distance estimates for calculating location. The
latter makes no assumption about the availability or validity of
such information.
The effectiveness of these two localization approaches
depends on the behaviour and the requirements of the
applications using location information, and on the desired
error granularity. Acknowledging that the range-free solutions
have a coarse accuracy, these techniques are unsuitable in
applications where the location precision is one of the main
This work was supported in part by part by the European Commission in the
framework of the FP6 projects PERSONA (contract N. 045459) and
INTERMEDIA (contract N.38419).
Paolo Barsocchi is with the ISTI-CNR, Pisa Research Area, Via G.Moruzzi 1,
56124 Pisa, Italy (e-mail: paolo.barsocchi@isti.cnr.it).
Stefano Lenzi is with the ISTI-CNR, Pisa Research Area, Via G.Moruzzi 1,
56124 Pisa, Italy (e-mail: stefano.lenzi@isti.cnr.it)
Stefano Chessa is with the ISTI-CNR, Pisa Research Area, Via G.Moruzzi 1,
56124 Pisa, Italy (e-mail: stefano.chessa@isti.cnr.it) and with Computer
Science Department, University of Pisa, Largo B. Pontecorvo 3, 56127 Pisa,
Italy (e-mail: ste@di.unipi.it).
Gaetano Giunta is with the Department of Applied Electronics, University of
Roma Tre, via della Vasca Navale 84, Roma, Italy (e-mail: giunta@ieee.org).
978-1-4244-2517-4/09/$20.00 ©2009 IEEE
requirements (such as AAL). On the other hand, range-based
localization can provide the adequate precision because it
exploits measurements of physical quantities related to signals
travelling between the mobile sensors and anchors. Radio
signal measurements are typically the Received Signal
Strength Indicator (RSSI), the angle of arrival (AOA), the time
of arrival (TOA), and the time difference of arrival (TDOA).
Recently, localization based on a combination of AOA and
TDOA techniques have been proposed, that guarantee a high-
accuracy location but it requires a specific and complex
hardware. Since typical AAL applications are deeply involved
with users monitoring and may suffer from complex and too
invasive hardware, using specific localization hardware based
on AOA and TOA may result impractical. For this reason we
opt for RSSI-based localization that does not require any
special or a sophisticated hardware, and it is available in most
of the standard wireless devices. Moreover RSSI-based
localization does not have a significant impact on local power
consumption, sensor size and cost, and for this reason it has
received considerable interest in the recent literature [5], [6],
[7].
The authors in [5] suggest that algorithms estimating distances
between two wireless devices based on their reciprocal RSSI
are unable to capture the myriad of effects on signal
propagation in an indoor environment. Nevertheless, the
authors in [6] have shown that despite the reputation of RSSI
as a coarse method to estimate range, it can achieve an
accuracy of about 2-3 meters root mean square in a testbed
experiment. Fading outliers can still impair the RSSI relative
location system, implying the need for a robust estimator. A
method to improve the quality of localization exploiting a
number of RSSI measurements averaged in a time window to
counteract interference and fading has been proposed in [7].
The main RSSI-based localization approaches are based on
fingerprint and on signal propagation modelling. In both cases
a mobile sensor is localized by means of fixed sensors (called
anchors) whose position is known in advance and that
exchange with the mobile sensors beacon packets in order to
collect sequences of RSSIs. The fingerprint schemes also
referred to as pattern matching, exploits the RSSI at the
mobile sensors as a function of the mobile position during an
off-line phase. Each mobile position is then identified by a set
of RSSIs. During the on-line phase the mobile location
estimation is performed by matching the actual signature of
the RSSI with the entries stored in a database available at the
anchors. The database entries are usually collected on a grid of
possible mobile positions within the area of interest, wherein
the grid spacing must be always chosen as a trade-off between
performance and time required to fix the mobile location. The
main drawback of this method is the extensive and accurate
measurements, during the off-line phase, required to create the
database. The creation of the database however is not
automatic: it is a human-based, time-consuming operation and
this is a practical barrier to its wider adoption.
The signal propagation modelling technique consists in
exploiting a large-scale path loss model that accounts both
free-space and loss due to obstructions; this model is then used
to estimate the RSSI for each mobile position. In indoor
environments the path loss model can also take into account
parameters such as the wall attenuation factor (WAF) and
floor attenuation factor (FAF) to model the effect of walls and
floors on the radio waves. Unfortunately, in particular in
indoor environments, RSSI is environment dependent since
the radio frequency signal suffers from reflection, diffraction
and multipath effects that make the signal strength rather
noisy. To overcome these problems wireless location systems
perform a preliminary calibration of the propagation model.
The calibration works in two phases: the training phase and
the estimation phase. In the training phase RSSI is measured at
a grid of points in the area of interest, and in the estimation
phase this information is used to estimate the propagation
model parameters. Clearly, the accuracy of the calibration
procedure depends on the number of points in the grid and on
the number of measures taken per point. It is also clear that
also this calibration (as it happens for the fingerprinting
technique) requires human intervention and is time expensive.
Unlike [12] where the localization algorithm was based on
standard Least Mean Square (LMS) algorithm, in this paper
we propose a novel localization algorithm that selects and
weights the RSSI measurements according to their strength,
and it uses an automatic training that only exploits information
from the anchors, without requiring human operators.
II. LOCALIZATION ALGORITHM
We assume a localization model comprising a set of anchors A
= {a1, a2, …, an}, a mobile M and a localization server. The
anchors have well known position on the map, identified by
the pair (xi, yi). Our localization system consists of two phases:
the training phase and the localization phase. In the training
phase each anchor broadcasts a beacon containing its
identifier. These beacons are used to measure the reciprocal
RSSIs among the anchors that, in turn, are used by the
localization server to calibrate the propagation model
parameters. In the localization phase each anchor periodically
emits a beacon packet containing its identifier. The mobile
node, which needs to be localized by the system, receives the
beacons from the anchors, computes the corresponding RSSI,
and sends to the localization server the pair <RSSI, anchor
id>. The localization server accumulates all the pairs and
using the proposed localization algorithm estimates the mobile
position.
III. TRAINING PHASE
In this section we describe the training phase based on virtual
calibration proposed in [12], whose objective is to adapt the
theoretical propagation model to the environment where it is
actually used. The virtual calibration procedure achieves this
goal without human intervention, by exploiting only
information obtained from the anchors. In particular, the
anchors preliminary exchange beacons to compute reciprocal
RSSI and the localization server uses this information to
configure the parameters of the theoretical propagation model.
The large-scale path loss model considered in this paper is a
further generalization of the one-slope model that consists in
adding an attenuation term due to losses introduced by walls
and floors penetrated by the direct path:
L(d ) = l 0 + 10α log10 (d) + WAF + FAF (1)
db
where FAF is the floor attenuation factor and WAF is the wall
attenuation factor expressed as:
N
WAF = ∑ k i l i (2)
i =1
where ki is the number of penetrated walls of type i, and li is
the attenuation due to the wall of type i. For the sake of
simplicity we assume that the localization system works in one
floor, hence the attenuation term in Equation (1) due to the
propagation among different floors was disregarded.
The parameters of the propagation model (1) are: l0 (the path
loss at distance of 1 meter), α (the air attenuation factor), and li
for each i∈[1,N]. Since l0 should be estimated in a free space it
is not affected by the environment, but it depends on physical
properties of the devices hardware and it can be estimated a
priori, thus it is not object of virtual calibration.
To calibrate parameters α and li the virtual calibration exploits
two heuristics: the global virtual calibration (G-procedure)
and the per-wall virtual calibration (W-procedure). The G-
procedure assigns the same parameters for every wall, based
on all RSSI measurements obtained from any pair of anchors.
Instead, the W-procedure provides an attenuation factor for the
walls that directly affect the communication between specific
pairs of anchors, and it uses G-procedure for all the other
walls. To the purpose of comparison we also have used
another heuristic (called H-procedure) that estimate the
propagation model parameters by using the RSSI
measurements on a grid of points in the environment, as it is
done with fingerprinting. Hereafter we use the following
notation:
ƒ C = {(ai, aj)∈A : ai and aj can communicate directly}
ƒ Ri = {aj1, aj2,…, ajn} is the set containing the ids of all
anchors in the same room, hence ∀(a, b) ∈ Ri : a≠b the
communication between a and b is “wall-free”;
ƒ Wn is the set of all pairs (ai,aj) such that the communication
channel between ai and aj crosses exactly n wall(s);
A. Global virtual calibration
The G-procedure considers a single virtual type of wall. This
leads (1) to:
L(d) db = l 0 + 10α log10 (d ) + kl w (3)
where k identifies the number of wall crossed by the signal
and lw is the attenuation introduced by the wall on the signal.
During the virtual calibration phase we can estimate all the
required parameters (l0, α, lw).
Substituting d(i,j) as actual distance between anchors ai and aj,
and k(i,j) as the number of wall crossed by the direct path
between anchors ai and aj in (3), we obtain an estimation mi are given by:
RSSI′(i,j) of the actual RSSI:
⎛ n n ⎞
( )
RSSI (′i, j ) = l 0 − 10α log10 d(i, j ) + k(i, j )l w
(4)
⎜ ∑ mi X i ∑ miYi
⎜
⎟
⎟
( )
∀i, j : a i ,a j ∈ C ( X G , YG ) = ⎜ i =1n , i =n1 ⎟ (6)
⎜ ⎟
⎜ ∑ i ∑ mi
The estimated RSSI′(i,j) differs from the measured RSSI(i,j) by m
⎟
an error component ε(i,j). We assume that all ε(i,j) are identically ⎝ i =1 i =1 ⎠
distributed and uncorrelated among themselves. Recalling that which, for equal masses mi simplifies to:
l0 is estimated a priori, the approximation of the remaining
⎛ n n ⎞
parameters (α,lw) that minimizes the least mean square error ⎜ ∑ X i ∑ Yi ⎟
║RSSI − RSSI′║2 can be achieved by direct method. The ⎜ i =1 ⎟
( X G , YG ) = ⎜ , i =1 ⎟ (7)
computation cost for direct method solving a linear least mean
square estimator problem is polynomial [14]. ⎜ n n ⎟
⎜ ⎟
⎝ ⎠
B. Per-wall virtual calibration
An example of how the localization algorithm works is shown
With this technique we estimate an individual attenuation in Figure 1, where a sensor node M is within the
factor for each wall between any pair of anchors belonging to communication range of four anchors. The three anchors a1,
C. Let us assume there are q different types of wall in the map a2, and a3 with a greater RSSI are selected and the six
of the building and let F = {f1, f2… fq} the set of attenuation intersection points I1,…, I6 (n = 6) are evaluated. The mobile
factors for each type of wall. Thus the Equation (4) building position M is then estimated by using the Equation (7).
the linear system becomes:
q V. EXPERIMENTAL RESULTS
( )
RSSI (′i , j ) = l0 − 10α log10 d(i, j ) + ∑ kh (i, j ) f h
(5) In this section we present the performance obtained using our
h =1
∀i, j : (ai , a j ) ∈ C
localization algorithm combined with the automatic virtual

where kh(i,j) is the wall number of type h crossed by the signal

considering the direct path between anchors ai and aj in (3).
The path loss exponent α used in this equation is previously
evaluated with the G-procedure. The evaluation of parameters
fi∈F is achieved with the same methodology used in
Subsection A, by means of the least mean square estimator.
IV. THE LOCALIZATION ALGORITHM
The training phase is followed by the localization phase,
where the localization server accumulates all the pairs of
<RSSI, anchor_id> received from the mobile and it computes
the position of the mobile by estimating the distance of the
mobile from each anchor based on the received RSSIs and
exploiting the propagation model calibrated as described in the
previous section.
Among all the received RSSI measurements the localization
server selects only those measures coming from the three
anchors with the greatest RSSI (without loss of generality let
a1, a2, and a3 be such anchors). This because the anchors that
have a smaller RSSI are farther from the mobile, thus the
RSSI values they produce are noisier and their use increase the
localization error. This effect was also confirmed during the
experimental setup.
Using the RSSI obtained from a1, a2, and a3 the localization
server estimates the distance of the mobile from these three
anchors. Using these distances as a radius of a circle, the
localization server estimates the intersection points between
them (Figure 1). Summarizing we use a trilateration method
by using the three anchors with higher RSSI in order to
decrease the bias introduced by the other anchors. The main
idea of the localization algorithm is to treat the intersection
points, estimated at (Xi, Yi), as point masses mi and to find the
center of gravity (centroid) of all these masses. In the most
general form, the coordinates of the centroid of n point masses
I1
a1 I5
I4 a2
M
a3
I3 I6
a4
I2
Fig. 1. Intersection points measured from the mobile node.
calibration procedure of the propagation model. In the next
subsections we present the setup of our experiment, the
performance comparison among different type of virtual
calibration and the performance of the proposed localization
algorithm.
A. Experimental setup
We performed the experiments in our laboratory. It is a typical
office environment with an area of approximately 7m by 11m.
It has desks, chairs, cabinets, computers, monitors, etc. This
environment is harsh for wireless communication due to
multi-path reflections from walls and the possibility of
interference from electronic devices. Figure 2 shows the
layout of the laboratory and the deployment of the anchors in
the rooms.
For the experiments we used a wireless sensor network
composed of 7 MicaZ [15] equipped with the Chipcom
CC2420 radio subsystem implementing the IEEE 802.15.4
 φW = stdclb (RSSI )− slfclbW (RSSI ) (9)
2
Where stdclb is the function computing the distance based on
the propagation model calibrated with the H-procedure, slfclbG
provides the distance by means of the propagation model
calibrated with G-procedure and slfclbW provides the distance
by means of the propagation model calibrated with W-
procedure.
With Global virtual calibration (G-procedure) the estimated
parameters are α = 1.45 and l1 = l2 = lw = −8.96dB.
With W-procedure the estimated parameters are α= 1.45,
l1=−8.33dB, and l2=−7.3dB. The main difficulty for this
calibration method is due to the different number of samples
used to estimate the single wall attenuation. In our case we
Fig. 2. Map of the building used for the experiments. have 5 anchors to estimate the first wall and other 3 anchors
for the second wall. Therefore, in order to resolve the
Equation (5) with the least mean square estimator, we weigh
standard. The experiments consist in a set of measures
the WAF parameters with a number directly proportional to
between a pair of anchors or between an anchor and 12 points
the number of established links between pairs of anchors and
of the grid (in the case of H-procedure). Each measure collects
inversely proportional to the number of anchors.
300 RSSIs, where every RSSI is averaged over a set of 100
Table II summarize the results obtained showing the
samples. Each sample is obtained exchanging a beacon packet
Cumulative Distribution Function (CDF) of φG and φW. This
between two sensors every 1/32 second, using the highest
table highlights that the W-procedure has a better performance
transmission power of the MicaZ.
than the G-procedure. Not surprisingly, the W-procedure
As mentioned in Section III the l0 parameter can be estimated
outperforms the G-procedure, due to the better accuracy in the
a priori as the path loss at a reference distance. To this
walls modelling. Figure 3 shows the CDF obtained by using
purpose we preliminary evaluated l0 measuring the RSSI
our localization algorithm (solid line) for each calibration
between two anchors deployed at 1 meters distance, obtaining
procedure. As depicted in this figure the G-procedure
l0 = −10.06.
performs like the commonly used H-procedure, in terms of
We also performed the H-procedure to obtain the reference
localization error and that the W-procedure is practically
parameters of the propagation model to be used for the
identical to H-procedure. This means that virtual calibration
comparison with G-procedure and W-procedure techniques.
procedure results in the same localization error as the
The parameters obtained with all the calibration methods are
expensive H-procedure.
shown in Table I.
C. Localization performance
B. Calibration performance
In this subsection we analyse the accuracy of our localization
In this section we present the results of the measurement
algorithm in the environment showed in Figure 2.
campaign aimed at comparing the performance of the three
Figure 3 shows the CDF obtained by using our localization
heuristics proposed in the previous section.
algorithm with the propagation model calibrated using the
We performed two phases of measurements. The first phase
virtual calibration method (solid line) compared with a
was aimed at performing the calibration of the propagation
commonly used LMS algorithm [12] [13] (dotted line). As
model using the three different procedures, namely: H-
depicted in this figure our localization algorithm outperforms
procedure, G-procedure, and W-procedure. The second phase
the LMS algorithm, despite the fact that the LMS algorithm
was aimed at measuring the RSSI on a grid of points in the
requires a time-consuming measurement campaigns for its
environment, and to compare the localization error of the
calibration.
procedures.
It worth to note the relationship between the anchor density
We use the H-procedure as the reference technique, and we
and the accuracy of the localizatio algorithm. In fact, Figure 4
evaluate the performance of the two virtual calibration
shows how the location error performs by changing the
techniques in terms of the following formula:
number of anchors.
φG = stdclb(RSSI )− slfclbG (RSSI ) (8)
2 As expected the error distance decreases with the increasing of

TABLE I TABLE II
PARAMETERS AND PERFORMANCE COMPARISON AMONG PROCEDURES CDF OF THE ERRORS

G-procedure W-procedure H-procedure Error [m] CDF of φG CDF of φW

α 1.45 1.45 1.46 0.25 0.604 0.896

l1 -8.96 -8.33 -8.21 0.5 0.732 0.919
l2 -8.96 -7.30 -6.40 1 0.868 0.973
1.5 0.901 0.984
 CDF CDF
1 1
G−procedure
0.9 W−procedure 0.9
H−procedure
0.8 G−procudure(LMS) 0.8
W & H−procedure(LMS)
0.7 0.7

0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3

0.2
0.1
0 0
0 0.5 1 1.5 2 2.5
Localization error [m]
Fig. 3. CDF of the localization error.
the density of the anchors (anchor per square meters). For
example, if we want to achieve the 90th percentile location
error within 2 meters we must deploy an anchor density
greater than about 0.092, which in our experimental testbed
means 3 anchors per room.
In conclusion we obtained a better results with respect the
RADAR [9], DALS [10] and the Bulusu [8] algorithms that
obtains a 75th percentile location error under 5 m, the 87th
percentile location error of about 9 m, and the 90th percentile
location error within 3 m, respectively. Our localization
performance are similar to that of MoteTrack system [11] that
achieves a 50th percentile and 80th percentile location-tracking
accuracy of 0.9 and 1.6 m respectively, but our algorithm does
not require the expensive training phase of MoteTrack.
VI. CONCLUSIONS
We considered the problem of indoor localization based on
RSSI and standard IEEE 802.15.4 radio interfaces. In
particular we propose a novel localization algorithm based on
sensor networks that identify the position of the mobile node
as a function of the RSSI measurements obtained exchanging
beacon packets between the mobile and the anchors. The
algorithm selects and weights the RSSI measurements
according to their strength, and it uses a propagation model to
transform RSSI measurements into distances. The algorithm
also uses a virtual calibration of the propagation model that
does not require any human intervention.
We showed by an experimental setup that the performance of
the virtual calibration, in terms of accuracy of the estimated
distances, is close to that achievable with more expensive ad-
hoc methods, and that the localization algorithm increases the
performance with respect to the commonly used LMS
algorithm showing also how to achieve a wished accuracy
increasing the anchor density.
0.2 2
0.092 [anchor/m ]
2
0.1 0.091 [anchor/m ]
2
0.078 [anchor/m ]
3 3.5 4 4.5 0 1 2 3 4 5 6
Localization error [m]
Fig. 4. CDF of the location error by changing the anchor density.
[2] G.D. Abowd, C.G. Atkeson, J. Hong, S. Long, R. Kooper, and M.
Pinkerton, Cyberguide: A mobile context-aware tour guide. Wireless
Networks, 3(5): 421-433, 1997
[3] Y. Sumi, T. Etani, S. Fels, N. Simonet, K. Kobayashi, and K. Mase, C-
map: Building a context-aware mobile assis- tant for exhibition tours,
Community Computing and Support Systems, Social Interaction in
Networked Communities, 137-154, London, UK. Springer-Verlag, 1998.
[4] K. Cheverst, N. Davies, K. Mitchell, A. Friday, and C. Efstratiou,
Developing a context-aware electronic tourist guide: some issues and
experiences, SIGCHI conference on Human Factors in Computing
Systems, pp. 17-24, 2000.
[5] E. Elnahrawy, X. Li, and R. Martin, The limits of localization using
signal strength: a comparative study, in First Annual IEEE
Communications Society Conference Sensor and Ad Hoc
Communications and Networks, 2004, pp. 406-414.
[6] N. Patwari, I. Hero, A.O., M. Perkins, N. Correal, and R. ODea, Relative
location estimation in wireless sensor networks, IEEE Transactions on
Signal Processing, vol. 51, no. 8, pp. 2137-2148, 2003.
[7] P. Bergamo and G. Mazzini, Localization in sensor networks with fading
and mobility, in The 13th IEEE International Symposium on Personal,
Indoor and Mobile Radio Communications, vol. 2, 2002, pp. 750-754.
[8] Bulusu, N., Heidemann, J., and Estrin, D. GPS-less low cost outdoor
localization for very small devices. IEEE Personal Communications
Magazine 7, 5 (October 2000), 28–34.
[9] Bahl P, Padmanabhan VN (2000) RADAR: an in-building RF-based
user location and tracking system. In: INFOCOM, pp 775–784
[10] Christ T, Godwin P (1993) A prison guard duress alarm location system.
In: IEEE ICCSTI, Ottawa, Ont., Canada, 13-15 Oct 1993, 106-116
[11] K. Lorincz and M. Welsh, “MoteTrack: a robust, decentralized approach
to RF-based location tracking,” Personal Ubiquitous Comput., vol. 11,
no. 6, pp. 489–503, 2007.
[12] P. Barsoccchi, S. Lenzi, S. Chessa, G. Giunta, Virtual calibration for
RSSI-based indoor localization with IEEE 802.15.4. in IEEE ICC 2009,
Dresden, Germany, June 14-18 2009.
[13] Borrelli, A. et.al., “Channel models for IEEE 802.11b indoor system
design”, IEEE International Conference on Communications, pp 3701-
3705, 20-24 June 2004
[14] Ake Bjorck, “Solution of Equations in RN”, vol. 1, NorthHolland, 1990
[15] Crossbow Technology Inc., http://www.xbow.com.
[16] Paolo Baronti, Prashant Pillai, Vince Chook, Stefano Chessa, Alberto
Gotta, and Y. Fun Hu, “Wireless Sensor Networks: a Survey on the State
of the Art and the 802.15.4 and ZigBee Standards”, Computer
Communications, 30 (7): 1655–1695, 2007

REFERENCES
[1] R. Want et. Al., The Active Badge Location System. ACM Transactions
on Information Systems, 10(1):91-102, 1992

View publication stats