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

net/publication/258217560

Feasibility of ultrasound positioning based on signal strength

Conference Paper · November 2012

DOI: 10.1109/IPIN.2012.6418917

CITATIONS READS

29 456

3 authors, including:

C. Medina Sverre Holm

University of Granada University of Oslo
22 PUBLICATIONS 284 CITATIONS 262 PUBLICATIONS 4,859 CITATIONS

SEE PROFILE SEE PROFILE

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

HIWIRE View project

Mathematical Modelling and Numerical Simulation of an Ultrasound Technology for Breast Cancer Detection View project

All content following this page was uploaded by José C. Segura on 18 June 2014.

The user has requested enhancement of the downloaded file.

 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

Feasibility of ultrasound positioning

based on signal strength

C.Medina, J.C.Segura S.Holm

Dept. of Signal Theory, Networking and Communications Dept. of Informatics
CITIC-UGR, University of Granada University of Oslo
C/Rafael Gomez Montero n2, E-18071, Granada, Spain P.O.Box L080, NO-0316, Oslo, Norway
[email protected], [email protected] [email protected]

Abstract-The objective of this paper is to analyze the the communication between the nodes of the network. This
performance of an indoor positioning system based on the infrastructure can also be used to provide localization using the
ultrasound level, usually called RSSI (received signal strength RSSI measurements of the radio packages.
indicator). This paper builds on past experiences where RSSI
values from beams sent in different directions were compared [I]. Radio protocols such as ZigBee allow one to measure easy
There relative values of RSSI were measured but herein we go the RSSI value of the radio packages. From this value, the
one step further and perform localization based on absolute distance between a transmitter node and a receiver node can be
values of RSSI. The use of RSSI could potentially mean a estimated using propagation models. However, the
simplification of the traditional positioning systems design in confmement problem of the RF signals and the difficulty to
relation to other methods such as TOF (time-of-flight), being an model their propagation, leads to important errors in
alternative and interesting method for positioning. The distance positioning. These errors are typically of several meters, which
between nodes (transmitter and receiver) is estimated from RSSI are unacceptable for indoor location applications. In a building,
values using a signal propagation model in which the power for example, the mobile node could be located in another room
losses due to the spherical divergence and atmospheric or even another floor.
absorption are considered. However, in real conditions, other
factors related to the characteristics of the nodes have their Other technologies such as infrared or ultrasound are
impact on RSSI measurements. One of them has to do with the typically used to achieve greater precision. Using such
beamwidth of the transducers. The signal strength will be technologies, a target can be located with guarantees in an
conditioned by the angle of incidence of the transducers when enclosed space due to the confmement of these signals. The
these have a narrow beamwidth. On the other hand, the common reflections of the infrared signal are unlikely, so that the RSSI
use of wireless nodes in the deployment of these systems leads to measurement requires a direct line of sight between the
the RSSI measurement to be affected by the battery level in the transmitter and receiver devices. On the other hand, this signal
nodes. These effects introduce significant errors in the distance may be disturbed by the sun light and fluorescent light. By
estimation and therefore on the localization precision. Herein we contrast, the ultrasound signal is immune to these interferences
propose a mechanism for modeling the power loss due to the
and allows multiple reflections, so it does not require a direct
orientation of the ultrasonic transducers, as well as an algorithm
line of sight between the nodes to detect signal. This allows the
to compensate for the effect of the variations in the battery level
use a smaller number of sensors to detect the presence of a
on RSSI measurements. Some experimental results have been
obtained with 5 nodes (4 transmitters and 1 receiver) to show the
target in a closed environment. By ignoring the signal path,
quality of our compensation using a real positioning system. some problems such as partial obstructions and other factors
Location errors smaller than 10 cm for each coordinate are that affect the RSSI measurements are solved. Thus, these
obtained in contrast to errors of several meters which are features favor the use of ultrasound signals rather than infrared
normally attained in this kind of systems based on RF (radio signals.
frequency) RSSI-values.
If there is a direct line of sight between emitter and receiver
or the path of signal propagation is known, then it is possible to
Keywords- Signal strength; Ultrasound; Positioning; Battery
level; Angular dependence
estimate the distance between them from RSSI measurements
using an ultrasound propagation model. Although the location
precision is significantly better than that obtained with
measurements based on RSSI-RF, such precision is still
I. INTRODUCTION
insufficient for location applications where errors of few
Currently most local positioning systems (LPSs) are based centimeters are required. In this sense, it is common take
on wireless sensor networks (WSNs). The WSNs have a low advantage of the slow speed of propagation of ultrasonic signal
cost, offer scalability and also an easy deployment of location to estimate its time-of-flight (TOF) between a transmitter node
systems. These systems use typically a radio interface to allow and a receiver node. The TOF measurement is then used to

978-1-4673-1954-6/12/$31.00 ©2012 IEEE

 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

calculate the distance value between nodes considering the II. LOCATION BASED ON ULTRASOUND-RSSI
propagation speed of sound. The TOF of ultrasound signal MEASUREMENTS
enables us to estimate distances with a precision of centimeters
In this paper we use the TELIAMADE system [3] for
or even millimeters [3]. This precision may be useful for
carrying out the evaluation of our approaches. The
certain applications (e.g., industrial applications). However, for
everyday location applications a precision of tens of TELIAMADE system is based on a wireless network of smart
centimeters is often sufficient. The TOF measurement requires nodes. Network architecture is based on the ZigBee protocol
a precise synchronization of the nodes to achieve good which fits the low power requirements for the smart sensor
accuracy. This synchronization is particularly complex in network design. A star network topology is used with a
WSNs where it is necessary to apply compensation algorithms network coordinator and a set of end nodes. End nodes are
to solve problems such as the clock drift at the nodes [2]. In equipped with a low power microcontroller (PICI8F4620) [4],
addition, the TOF measurement requires advanced digital and a radio chip (CC2420) [5] implementing the 802.15.4
signal processing techniques which imply more hardware physical layer. The rest of the ZigBee stack is implemented by
resources in the design of the nodes and a higher cost of the software in the microcontroller. Each TELIAMADE node is
system. also equipped with a couple of low cost ceramic ultrasonic
In this paper we propose to use the ultrasonic RSSI transducers (400ST/RI20) [6] with a center frequency of 40
measurement to determine the position of a mobile node with kHz and a 6 dB bandwidth of 2 kHz. Therefore, a given node
an acceptable precision for most daily location applications. can be configured to transmit or receive an ultrasonic signal at
We think that RSSI is an alternative method for positioning a given time. Fig. 1 summarizes the operation of the
which includes simplicity of development and lower cost. This TELIAMADE system in its typical configuration. The devices
paper builds on past experience where RSSI values from beams located on the ceiling (denoted by Tx) represent the end nodes
sent in different directions were compared [1]. There relative configured as ultrasonic transmitters. The mobile node
values of RSSI were measured but in this paper we go one step (denoted by Rx) is configured as ultrasonic receiver. The
further and perform localization based on absolute values of configuration and control of the network nodes is done by
RSSI. sending of radio packets issued from the coordinator node.
The RSSI measurements between a transmitter node and a
The received signal might have followed a non-direct path
due to partial obstructions in the signal path or an inadequate receiver node are returned to the coordinator for determination
orientation of the transducers, causing a poor estimate of of the receiver node position. However, the current
distance. However, the RSSI measurement allows knowing the configuration of the system can be modified so that the
approximate position of the mobile node when this happens receiver node can perform this estimation using the hardware
(e.g. the room in which it is located). The use of RSSI resources available.
measurement instead of TOF measurement could potentially
mean a simplification of the system design, a lower cost of RSSI estimations are obtained by measuring the amplitude
signal processing and therefore a lower consumption. of an ultrasonic pulse of Tb duration, between a pair of nodes;
one acting as an ultrasonic transmitter and the other as an
However, to achieve an acceptable precision using RSSI
ultrasonic receiver. The ultrasonic burst generation is
measurement, two fundamental problems that typically occur
in this type of systems must be solved. One of them has to do performed by the EUSART (enhanced universal synchronous
with the orientation of ultrasonic transducers, where their asynchronous receiver transmitter module) of the
directional characteristics determine the emitted and received microcontroller. A sequence of alternating 0 and 1 bits is
signal strength. The other problem is the voltage loss of generated at a baud rate of twice the nominal transducer
batteries due to the consumption of the node, thus affecting the frequency (i.e. 80 kbps) with an appropriate length (e.g. a
RSSI measurement. pulse of 1 ms requires the transmission of a sequence of 80
alternating bits). The EUSART output is buffered though a
The paper is organized as follows. In Section II we present
digital inverter gate (74HC04) [7] to provide the needed
a propagation model of ultrasound signal to estimate the
current gain to drive the ultrasonic transducer. In reception,
distance from RSSI measurements, where losses by spherical
the signal is amplified and band-pass filtered before being
divergence and atmospheric absorption are considered. In
Section III, the effect of the ultrasonic transducer orientation is sampled and digitized, using the AID converter of the
analyzed, and the function that allows one to model the losses microcontroller with lO-bit resolution for a suitable dynamic
by orientation to correct the RSSI measurement is reported. range. This signal conditioning stage is implemented using a
Section IV shows the effect of battery voltage on the RSSI dual operational amplifier (LMC6482) [8] and some passive
measurement and the algorithm to compensate for this effect. components (resistors and capacitors). TELIAMADE nodes
In Section V, the improvement in the distance estimation by are battery powered, with a voltage between 2.7 and 3 volts.
considering our approaches is shown. There some location Fig. 2 shows the block diagram of the designed module for
results when these compensations are applied to the RSSI transmission and reception of ultrasonic signals. Details signal
measurements are also shown. These results are obtained using conditioning in transmission and reception are shown. In
the TELIAMADE system [3] in controlled scenarios where reception, the signal processing for the RSSI measurement is
orientation and battery effects appear. Finally, in section VI, briefly indicated. A detailed description of such processing is
we summarize the main conclusions of this work. presented in subsection II-A.
 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

� 51
GNAL PULSE MODULATED ULTRASOUND
. SIGNAL TRANSDUCER (TX)
i .mi.
� JT.l =J&U1rA- M---: ( �- :F'-' ERC:- ---'r- rn
:
B
SQUARE CARRIER
(F!) 40 kHz)
v;

� =

C;
"
MICROCONTROLlER
(a)
ULTRASOUND

hi [0 - 4 0 8 0 - 4 0]
hq [1 0 7 0 7 0 1]
=

= - -

Figure I. Typical configuration of the TELlAMADE system. A coordinator MICROCONTROLlER

node is connected to a PC through which all other nodes of the network are
managed. A set of fixed nodes work as ultrasonic transmitters (Tx) and are
Figure 2. Block diagram of the designed module for transmission (a) and
placed in known positions (typically on the ceiling). The mobile node to be
reception (b) of ultrasonic signal in a TELIAMADE node. In transmission, the
located operates as ultrasonic receiver. Its position is determined by applying
signal is amplified before being driven to the transducer, in order to increase
multilateration from the estimated distances to the transmitter nodes. A the signal power and improve the range of the system. In reception, the signal is
propagation model is used for this purpose considering RSSI values. amplified and band-pass filtered before being sampled and digitized using the
AID converter of the node's microcontroller. The computation of the RSSI is
performed using a quadrature digital correlator (see subsection II-A).
The picture of a TELIAMADE node is shown in Fig. 3, where
it identifies the hardware components of the radio module and
the transmission and reception modules of ultrasonic signal. Due to interleaved sampling, the samples len) are only
known at even values ofn, while samples Q(n) are known at
odd values of n. However, since the signals are properly
A. Estimation RSSI-value sampled, len) and Q(n) can be recovered to a sampling
Computation of the RSSI is performed by the receiver frequency of Fs (i.e. at the same instant) using an interpolation
node using a quadrature digital correlator. A band-pass process. To this end, two band-pass quadrature filters are used
sampling scheme [9] is used to reduce the required memory to determine the values of the in-phase and quadrature
and processing resources. The signal is sampled at the receiver components at the instants where they are unknown. Table I
shows the way in which the in-phase and quadrature
using a sampling frequency (Fs) [ lO] given by
components are obtained by interpolation for an odd value of
M. The parameter m of the table represents the sample index n
4Fo
Fs= (1) expressed as module 4.
(2M -1)'
TABLE!. INTERPOLATED IN-PHASE AND QUADRATURE COMPONENTS
where Fo is the carrier signal frequency (40 kHz) and M is an FOR M ODD.

integer. The sampling frequency must satisfy the constraint of Compo m=O m=l m=2 m=3
Fs � 4B, where 2B is the signal bandwidth (double sideband). fen) +xi(n) +xq(n) -xi(n) -xq(n)
Then the quadrature sampling allows one to use relatively low
Q(n) +xq(n) -xi(n) -xq(n) +xi(n)
sampling frequencies of the order of 17.78 kHz (M=3) or even
12.31 kHz (M=5), so that memory resources and
computational cost in signal processing are reduced. where xi(n) and xq(n) are given by the following expressions:

Equation (2) shows the expression for a band-pass signal Xi(n)= 8x(n) -4(x(n -2) + x(n + 2)) (3)
using digital quadrature sampling which allows one to extract
samples of the in-phase component (I(n)) and quadrature
Xq(n)= 7(x(n + 1) -x(n -1)) + x(n - 3) -x(n + 3) .
component (Q(n)) to a frequency of Fs/2. From these in-phase and quadrature components, the outputs of
the digital signal correlator are given by
K-l K-l
x(n) = (
len) cos rrn (M -�)) - Q(n) sin (rrn (M -�)). (2) C;Cn)= IO f(n - k) Cq(n)= IO Q(n - k) , (4)
k= k=
 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

ULTRASOUND TRANSMISSION MOOULE

(8)
ZIGBEE MODULE

where R is the real distance between a transmitter node and the

receiver node and a is the atmospheric absorption factor [12].
The following expression shows the dependence of a (in
Np/m) with the temperature T (in K), the relative humidity h
(in %), the signal frequency [(in Hz) and the atmospheric

[
pressure Ps (in atm).

a =_
F 2
Ps
. 1.84 . 10-11 -
1
T 2
To +
()
( )]
Pso
ULTRASOUND REaPTION MODULE

2239.1 33 2
Figure 3. Picture of a TELIAMADE node. The regions identified by dashed
line show the hardware components of the radio module and the signal 5 5

=
2
conditioning modules for transmission and reception of ultrasonic signal. ( ;)
o
- ---r; 0.01278 e- 2 + 0.1068 e- --r;, 2 ' (9)
Fr.o + p Fr.N + p

= = =
r.O r.N
where K TbF's is the number of samples of ultrasonic pulse
and Tb is the pulse duration. The signal envelope is calculated

=
as with F [/Ps' Fr.o [r,o /Ps and Fr.N [r,N /Ps. The terms
[r,o and [r,N are respectively the relaxation frequencies of
Ce(n) Cl(n) + q(n). (5) oxygen and nitrogen whose value depends on the relative
humidity and temperature. All the frequency terms in (9) are
Finally the ultrasonic RSSI-value is estimated by detecting the expressed in Hz.
maximum of the envelope
The distance measurement R can be estimated by equating to
zero the passive sonar equation (7) and substituting (8)
(6)

B. Using RSSI-value for location

b.= (RSS1o -RSSI) -2010g10 (:J + a(R -Ro)· (10)
A simple way to determine the position of the mobile node

=
using RSSI is to construct a map of RSSI values of the The derivative of (10) with respect to R is
localization environment. This involves taking a large number
of measurements at different reference points. The position of ob. 20
the mobile node is estimated by comparing the current RSSI­ - - -a . (11)
oR Rlog(10)
value with other RSSI values measured at different reference

=
points. However, this approach is not viable for location in
large spaces or when a high precision is required due to the From (10) and (11) the value of R is estimated using the
considerable workload that this implies. Newton-Raphson iterative method for the initial approxima­

=
tion R Ro
A more efficient solution is based on predicting the
distance between nodes (transmitter and receiver) by applying
a propagation model for the ultrasonic signal. The mobile node
� b.
R R- . (12)
position can then be calculated from the estimated distances to ob./oR
transmitter nodes using multilateration. The distance between
nodes is obtained using the passive sonar equation in a way For a typical scenario with one mobile node and a set of fixed
similar to that described in [11]: transmitter nodes, the mobile position can be calculated by
multilateration using the estimated distances to each
RSSI === RSSlo -PL, (7) transmitter node. These distances are denoted by Rb where the
subscript i refers to each transmitter node and takes values
where RSSI is the estimated power level to any distance (R), within the range i=I ...N, where N is the total number of
RSSlo is the measured power level to a known distance (Ro) transmitter nodes. The position of these nodes is known and it
is denoted by its coordinates (x, y, z) in the form (tix ' tiy , tiz)'
(under reference conditions of temperature and relative
humidity). In (7) we assume RSSI> (NL + DT) , where NL is
The position vector of the mobile node (m) is calculated by
the Noise Level and DT is the minimum Detection Threshold
of signal. The term PL is the Power Loss due to the signal solving an overdetermined system of linear equations given by

=
propagation. In this paper we use propagation model which the following expression
includes both losses due to spherical divergence and
atmospheric absorption. Its expression is given by A'm ' b (13)
 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

where approximated considering the resonance membrane radius (a)

of the transducer. In this case a=3.6 mm, which is measured
tlX - tzx tly - tzy tlz - tzz using a digital caliper. The parameter e is the angle of
tlx - t3x tly - t3y tlz - t3z incidence of the transducer, and}l(x) is the Bessel function of
A= tlx - t4x tly - t4y tlz - t4z the first kind and order l. The function fee) is normalized to
unity at e =0° (easily seem as}l(x) behaves like}l(x) == xj2
for small x). The obliquity factor is a mix of the factor for a
hard baffle and a soft baffle. This mixed obliquity factor is
5( tlx z - t z + RZ - RZ)
z - t zx + tlz - t z + tlz often found to describe practical ultrasound transducers quite
O· z zy zz z 1
y well. From (16), the loss factor by orientation is found from the
5 z z z z z z Z
O ( tlx - t3x + tl - t3 + tlz - t3z + R3 - R1 )
Z angle of incidence for the transmitter and for the receiver as
·
y y
b= O 5( tlx
z - t4x Z - t4z
z + tlz - t4z + tlz Z + R4Z - R1Z)
·
y y (17)
5 Z Z Z Z Z Z Z Z
O . ( tlx - tNx + tl - tN + tlz - tNz + RN - R1 )
y y In practice it is rather cumbersome to account for the loss
factor as the two angles need to be found. In a future system we
A closed solution for m can be obtained using the least would therefore like to get rid of this factor, i.e. to use
squares method. Solving for m, we have omnidirectional transducers. This is feasible for the receiver as
there exist MEMS-based sensors which are intrinsically
m = ((A" A)-I. A'. b)'. (14) omnidirectional. However, care must be taken when they are
However, factors such as the orientation of the transducers or mounted in a system as there must be no shadowing from the
variations in battery level of the nodes will influence the RSSI mounting or the printed circuit board.
measurement. To compensate for these losses and reduce the The way to get an omnidirectional transmit transducer is to
localization error due to an incorrect estimation of distance, make the active radius a smaller. As we are not aware of any
the RSSI expression is rewritten as such transducer which is readily available now, one could
instead use mUltiple transmitter transducers. For instance four
transmitters spaced 90° apart could be used combined with a
measurement of the response for each one in turn. This could
where !:::"e is the loss factor due to the orientation of the give enough information to resolve the effect of the
transducers and !:::"V is the loss factor due to changes in the transmitter's angle of incidence. The disadvantage is the
battery voltage. The following sections III and IV describe the reduced update rate as four times, as many transmit pulses need
form in which!:::"e and!:::"V are estimated. to be sent.
The orientation problem is well known in WLAN-based
fmgerprinting positioning systems also where it is presently not
III. RSSI LOSS DUE TO ORIENTATION OF TRANSDUCERS corrected for [14].
Most of ultrasound-based localization systems make use of
narrowband transducers. These transducers are characterized
by a high emission power and high sensitivity. These features IV. RSS[ LOSS DUE TO BATTERY CONSUMPTION

allow good accuracy in the RSSI measurement and a long Nodes in wireless sensor networks (WSNs) are often
distance range. However, they have usually a highly directional powered by batteries. Depending of their configuration and
radiation pattern with a beamwidth less than e = ± 40 ° as workload, the consumption of batteries can be greater or less,
defmed by 6 dB power loss. Therefore, variations in the causing voltage differences between network nodes. A
orientation of the transducers will affect the RSSI measurement reduction of the voltage level in the nodes affects the RSSI
and thus the distance estimation. The power loss by orientation measurement. A lower voltage in the transmitter nodes results
will be different depending of the transducer's features. In this in a smaller signal gain at the output of the amplifier stage. On
paper we propose a theoretical model to estimate the power the other hand, a voltage drop in receiver node leads to a
loss due to the orientation angle of the transducers. This model reduction of the dynamic range of the AID converter, then the
considers the transducer surface as an uniform circular aperture received signal is detected with an amplitude greater than its
[13] actual value and therefore with greater power.
Using the propagation model described in Section II, the
}1(2nu)
fee) = (1 + cose) · ....:
- :....- distance between nodes can be estimated considering a
2nu reference power value RSSlo measured to a fixed distance Ro
a under known atmospheric conditions of temperature and
wit h u= �sin e . (16) relative humidity. Since the RSSI measurement is conditioned
by battery level of the nodes, the battery level must be
Equation (16) is the well-known Airy pattern for a circular considered in the RSSlo estimation. The conditions under
aperture multiplied by an obliquity factor (1+cose). The term a which RSSlo is estimated are known as reference conditions.
is the radius of the circular aperture and its value is Therefore, since voltage conditions in the nodes may differ
 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

from the reference conditions, the RSSI measurements will be

affected by such variations.

A voltage drop in the transmitter batteries causes a lower

RSSI value, and thus the estimated distance is higher than
expected. This is due to the fact that the signal power loss is
attributed only to the signal propagation from the transmitter
node to the receiver node. By contrast, a voltage drop in the
receiver batteries causes the RSSI measurement to be detected
with higher value. As a result, the estimated distance is now
lower than expected.
To compensate for battery effects, we use the reference
voltage levels and the real voltage levels to adjust the RSSI
measurement to the voltage reference conditions. This loss
factor is expressed as

(18)
Figure 4. Power-loss due to the orientation of the transducers. The filled dots
correspond to power losses experimentally obtained from RSSI measurements
using the transducers. By contrast the unfilled dots correspond to the estimated
where VTO and VRO are respectively the voltage levels of the power losses using the loss factor in (17).
transmitter and receiver in the reference conditions, while VT
and VR correspond to actual voltages at the nodes when a RSSI
measurement is performed.

V. EXPERIMENTAL RESULTS

In this section we show the effects of the transducer's angle

of incidence and the battery level on RSSI measurement. An E
�5
incorrect estimate of the RSSI level introduces errors in the
e
distance measurement when the signal propagation model is iii
(]) 4
o
c:
applied. In subsection V-A, some results about it are shown.
os
This leads to significant localization errors using ti 3
multilateration. Results for location accuracy achieved with the Ci
TELIAMADE system are shown in subsection V-B.
A. Accuracy in distance measurement
We begin by analyzing the effect of the angle of incidence
o
of the transducers in the RSSI measurement. In Section III, we -80
propose to use the Airy function for modeling the power loss of
the transducers with the angle. For this, a pair of nodes placed
on a swivel tripod are used. The nodes are placed facing each Figure 5. Distance errors obtained using the signal propagation model from
RSSI measurements taken at different angles. In grey, distance errors without
other at a fixed distance (about 2.0 m) and a certain height by-angle loss compensation are shown. In black, the distance errors when loss
above the floor (about 1.6 m). One of them (receiver) is fixed factor due to angle is used for compensate of the RSSI measurements.
while the other node (transmitter) is rotated at different angles
in the range [-80°, 80°]. The angles are measured using a
graduated compass located on the rotation platform of the The gray bars show the errors obtained when the compensation
tripod. of loss power due to angle is not applied. By contrast the black
bars show the errors when RSSI measurements are
To verify the quality of our approach, we use as reference compensated considering the loss factor /:::"0' An increase of the
the losses pattern obtained experimentally with our ultrasonic error when the angular misalignment between transducers
transducers. Fig. 4 shows the power loss curve for the Airy grows can be seen. For example for angles exceeding ± 50°,
model (unfilled dots) and the power loss curve from the errors of several meters are obtained. However, these errors are
experimental RSSI measurements using the transducers (filled significantly reduced when the RSSI measurement is
dots). The similar behavior of both curves indicates that our compensated with the loss factor /:::"0' In this case, the errors are
model is consistent with experimental measurements. less than 0.6 m for angles less than ±65°.
Fig. 5 shows the distance error obtained from RSSI A new experiment was performed to assess the effect of the
measurements taken at different angles using the ultrasonic battery levels on the RSSI measurements. As in the previous
signal propagation model. Temperature and relative humidity experiment, we consider a pair of nodes separated by a fixed
values are taken for each measurement, being their average distance (about 2.0 m). In this case both transmitter and
values of 21.4 o( and 39 % respectively. receiver are aligned so that then the orientation angle between
 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

them is 0°. In this way the RSSI measurements are not affected Test 1
by losses due to the orientation of the transducers, thus the
effect of the battery can be analyzed separately. The nodes are
powered by external power supplies instead of their internal
batteries (AA batteries) in order to have accurate and adjustable
control of the voltage level in the nodes. The nodes are capable
of estimating its voltage level ( Vin) in an autonomous way
using an adjustable voltage threshold (Vth) whose resolution is
of ± 46.3 mY. Such inaccuracy can lead to a maximum 31 315 32 325 33
distance error of 32 mm. Battery level M
Test 2
To see the effect of the batteries in the processes of
transmission and reception of ultrasonic signal, two tests are --+- Without compensation
carried out. In Test 1, the receiver node is supplied with a fixed
voltage of 3.30 V while the transmitter node changes its
voltage level in the range [2.65, 3.30] V. Temperature and
relative humidity values are taken for each measurement. In
this case, their average values are 21.7 °C and 34 %
respectively. On the other hand, Test 2 is based on providing a 44.31 '----'----'_...l..---'-_.l.--'--''--1._-'--'-_'-
2.65 2.7 2.75 2.8 2.85 2.9 2.95 3 305 31 315 32 325 33
fixed voltage of 3.30 V to the transmitter node, by changing
Battery level M
now the voltage of the receiver node in that same range. The
average value of temperature for this experiment is of 22.5 °C , Figure 6. RSSI measurements obtained at different voltage levels. In Test I, the
with equal relative humidity value. Fig. 6 shows the RSSI receiver node is powered with a fixed voltage of 3.30 V by changing the
voltage of the transmitter node in the range [2.65, 3.30] V. In Test 2, the
measurements obtained for both tests. The filled dots curve transmitter node is powered with a fixed voltage of 3. 30 V by changing the
shows the RSSI values measured at different voltages when the voltage of the receiver node in that same range. The curve of filled dots shows
battery loss compensation is not used (case A). By contrast, the the RSSI values measured at different voltages without to use loss
unfilled dots curve shows the compensated RSSI compensation due to battery (case A). The curve of unfilled dots shows the
RSSI measurements when the battery-loss factor is used (case B).
measurements using the loss factor as indicated in (15) (case
B). The reference voltage levels CVTO ' VRO ) for the calculation
of t.v are set to 3.30 V. The reference power value RSSlo is The significant increase of the error for voltages values below
measured considering the reference voltages. In Fig. 6 the 2.75 V is due to the malfunctioning of some hardware
curve of filled dots shows the effect expected in the RSSI components of the node's design when these are working
measurements when voltage conditions in the nodes is different below this voltage.
from the reference voltage conditions. Test 1 lets one see that a
voltage drop in the transmitter node reduces the signal power TABLE II. ESTIMATED DISTANCE ERRORS BETWEEN TWO NODES
due to a lower gain in transmission. On the other hand, Test 2 SEPARATED BY A FIXED DISTANCE OF 2M CONSIDERING DIFFERENT VOLTAGE
LEVELS. THE DISTANCE IS CALCULATED FROM THE RSSII MEASUREMENT
shows that a voltage drop in the receiver node increases the USING AN ULTRASONIC SIGNAL PROPAGATION MODEL. THE RESULTS SHOWN
received signal power by reducing the dynamic-range of the IN TABLE CORRESPOND TO THE EVALUATION OF THE TESTS I AND 2
AID converter. The compensation algorithm proposed in this DESCRIBED ABOVE. THE CASE A SHOWS ERRORS WHEN THE COMPENSATION OF
BATTERY IS NOT APPLIED WHEREAS THAT CASE B SHOWS ERRORS USING SUCH
work reduces these effects. The curves of unfilled dots COMPENSATlON.
represented in Fig. 6 show the effect of this compensation. The
nearly flat slope of the compensated curves demonstrates the
(Test 1) (Test 2)
effectiveness of our adjustment. Battery distance error (mm) distance error (mm)
(V)
To see the effect of such compensation in the final accuracy Case A CaseB Case A CaseB
of the system, the RSSI measurements are converted to 3.30 0.1 -9.4 0.0 -21.5
distance values. The distance is estimated from the RSSI 3.25 -32.5 14.8 22.6 -21.3
measurement using an ultrasonic signal propagation model. In 3.20 -70.9 12.0 43.9 -22.7
Table II are shown the distance errors obtained for the Test 1 3.15 -103.5 15.6 82.5 -6.0
and Test 2 considering the cases A and B described above. Data 3.10 -137.0 18.9 102.3 -9.0
report that a significant improvement in the accuracy of the -174.4 19.0 137.0
3.05 3.6
system is achieved when the RSSI measurements are
3.00 -218.3 14.2 157.6 1.3
compensated considering the current batteries level of the
2.95 -247.8 23.5 188.3 9.9
nodes with regard to reference voltage conditions. Such
2.90 -294.1 18.1 217.4 16.6
compensation allows one to limit the maximum error in the
2.85 -335.1 18.3 247.7 24.3
estimation of distance to a value less than 40 mm. It should be
mentioned that these error measurements are affected by 2.80 -381.3 15.1 273.5 28.2

maximum allowed resolution to estimate the current level of 2.75 -426.6 13.2 305.6 1l.l
batteries in the nodes (errors of 32 mm may occur because of 2.70 -485.6 39.6 338.4 21.6
this imprecision in the estimation of voltage). 2.65 -532.4 38.4 372.0 33.3
 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

B. Indoor positioning accuracy in a real environment Absolute positioning error

50 ,-----.-----,
In the previous subsection V-A, the distance measure
accuracy is discussed considering the compensation for
orientation and consumption of batteries in the RSSI
measurement. In this subsection we show the localization
accuracy of the TELIAMADE's system by applying
multilateration using the estimated distances. Location
measurements are performed in an empty room of dimensions
(7x5x2.5) m, using four nodes transmitter and a receiver node.
The transmitter nodes are placed in the central area of the
ceiling, forming a rectangle of dimensions (2.5x 2) m, with
their ultrasound transducers facing down. This reduces the
multipath effects from signals from the surrounding walls. The
receiver node is placed in different known coordinate points
inside the projected area by transmitter nodes. This node is
placed with its transducer ultrasonic facing the ceiling. Fig. 1 50 100 150 �o 250 300
helps us understand the deployment of the nodes in this new error (mm)
experiment. Figure 8. Histogram of the absolute error positioning from the measurements
carried out at all positions.
The location accuracy of the system is evaluated at 10
different positions for the receiver node considering several The unfilled points correspond to the positions where the
heights. At least 300 RSSI measurements are taken for each receiver node is placed considering only the coordinates x and
transmitter node for each position. The RSSI measurements are y. The filled points correspond to the set of localization
compensated taking into account the angle of incidence measurements obtained by applying multilateration.
between the transducers. This angle is estimated by
trigonometry using the known coordinates of the transmitter The proximity of the experimental-points clouds to the
nodes and the receiver node for each position. The effect of the theoretical values demonstrates the effectiveness of the
batteries in the RSSI measurement is also compensated using compensation approaches proposed in this work. Fig. 8 shows
the voltage values estimated by the nodes in each the histogram of the absolute positioning error considering all
measurement. Finally the RSSI value is converted to distance measurements carried out. The shape of the error distribution is
using signal propagation model described in section II-B. approximately log-normal with mean 10 cm and a standard
Temperature and relative humidity values are considered at deviation of 4.5 cm.
each measurement to achieve a correct estimate of a. In Table Table III shows the estimated mean error for each component
III are shown the average values of the temperature and relative (x, y, z) at the considered test points. Note that errors of less
humidity obtained from the set of measurements taken for each than 10 cm for each component are obtained. The last column
position. Fig. 7 shows the 2D-representation of the evaluated of the table shows the root-mean-square error (rmse) obtained
test points. The z component is omitted for better visualization from of the measurements taken at each position. A global
of the distribution of the measurements. analysis of the data (considering all error measurements) shows
that 95 percent of these errors are less than 19.5 cm. This
2 D-Position of the receiver node at different test points
1� ,---.---,---�--� demonstrates the feasibility of using the RSSI measurement in
daily location applications since these errors are perfectly
bearable.

TABLE Ill. AVERAGE VALUE OF THE LOCALIZATION ERROR OBTAINED

FOR EACH COMPONENT (x, Y, z) AND THE ROOT MEAN SQUARE ERROR (RMS)
OF THE MEASUREMENTS TAKEN IN THE TEST POINTS. AVERAGE TEMPERATURE
AND RELATIVE HUMJDlTY VALUES MEASURED FOR EACH POSITION ARE GIVEN ..

T RH x-error y-error z-error rmse

Pos.
(OC) (%) (mm) (mm) (mm) (mm)
PI 21.6 41 -108.5 3.84 -65.3 149.2
P2 21.5 41 -87.9 -18.1 4.0 112.4
P3 23.3 32 49.7 -33.0 -0.8 84.3
P4 21.6 41 -44.6 -18.1 42.7 88.6
'���0--�300=-�4=OO--�5=O-�6=00��7=OO�-=800�-=��-=100=0--�1100 P5 23.3 32 49.5 36.2 -44.7 92.8
x (mm)
P6 21.6 41 -29.3 32.1 19.9 75.8

Figure 7. 2D-representation of the evaluated location positions. The unfilled P7 23.2 32 117.9 -4.6 108.5 169.8
points correspond to the positions where the receiver node is placed (using only P8 23.2 32 -50.3 -21.18 61.7 100.2
coordinates (x,y)). The filled points correspond to the set of localization P9 21.3 42 17.3 -18.7 142.6 160.9
measurements obtained by applying multilateration from the RSSI
measurements. PIO 21.3 42 2.7 114.2 49.6 143.8
 2012 International Conference on Indoor Positioning and Indoor Navigation, 13_15th November 2012

VI. CONCLUSIONS [2] C.Medina, J.C.Segura and A. I. Torre, "Accurate Time Synchronization
of Ultrasonic TOF Measurements in IEEE 802. I5. 4 based Wireless
In this paper the performance of an indoor posltlOning Sensor Networks," AD-HOC NetlV orks, August,2012., "in press".
system based on ultrasound RSSI is analyzed. The use of RSSI DOl: 10.10 16/j.adhoc.2012.07.005.
could potentially mean a simplification of the traditional [3] C.Medina, J.C.Segura and A. I. Torre, "A wireless sensor network
positioning systems design in relation to other methods such as oriented to localization with sub-centimeter accuracy" - "Una red
TOF, being an alternative and interesting method for inalambrica de sensores orientada a localizacion con precision
subcentimetrica," in Proc. 26th Simp. Nacional URSI, Madrid, Spain,
positioning. The distance between nodes is estimated from
September, 2011. subcentimetrica," in Proc. 26th Simp. Nacional URSI,
RSSI values using a signal propagation model in which the Madrid, Spain, September, 20II.
power losses due to the spherical divergence and atmospheric [4] Microchip, "28/40/44-Pin Enhanced Flash Microcontrollers with 10-Bit
absorption are considered. However, other factors have their AID and nanoWatt Technology," [Online]. Available:
impact on RSSI measurements such as the beamwidth of the http://wwl.microchip. comldownloads/en/DeviceDoc/39626e. pdf.
transducers or the voltage drop of the batteries at the nodes. All [5] Texas Instruments, "CC241O 2.4 GHz IEEE 802.15. 4/ZigBee-ready RF
this introduces significant errors in the distance estimation and Transceiver," [Online]. Available:
therefore on the positioning accuracy when muItilateration is http://www.ti. comllit/ds/symlink/cc2420.pdf.
applied. [6] Prowave,"400STI20/SRI20 datasheet," [Online]. Available:

Herein we propose an approach for modeling the power http://www.prowave. com.tw/pdf/T400S12.pdf.

loss due to the orientation of the ultrasonic transducers, as well [7] Philips Semiconductors, "74HC04, 74HCT04 Hex inverter datasheet,"
[Online]. Available: http://www.nxp. comldocuments/data sheet174HC.
as an algorithm to compensate for the effect of the variations in
[8] National Semiconductor, "LMC6482IN," [Online]. Available:
the battery level on RSSI measurements.
http://www.ti. com/lit/ds/snos674c/snos674c. pdf.
To demonstrate the quality of our compensations some [9] W.Waters and B. R. Harret, "Bandpass signal sampling and coherent
experimental results are obtained using a real location system detection," IEEE Trans. on Aerospace and Electronic Systems, vol.AES-
in a controlled environment. The results show a good accuracy 18, no. 4, pp. 731-736, November, 1982.
of the system with rms location errors lower than 17 cm, in [10] J.L. Brown Jr. , "On quadrature sampling of bandpass signals," IEEE
Trans. on Aerospace and Electronic Systems, voI.AES-15, no.3, pp. 366-
contrast to errors of several meters that typically are obtained
371, May, 1979.
for these systems based on RF RSSI measurement [15].
[II] S.Holm, "Airborne Ultrasound Data Communications: The Core of an
Indoor Positioning System," in Proc. IEEE Ultrasonics Symposium,
Rotterdam, Netherland, September, 2005.
ACKNOWLEDGMENT
[12] H.E. Bass, L. C. Sutherland, AJ.Zuckerwar, D.T. Blackstock and D.
This work was partly supported by the Regional Government Hester, "Atmospheric absorption of sound: Further developments,"
of Andalusia under research project P08-ICT-03886. We are Journal of the Acoustical Society of America, vol. 97, no. 1, pp. 680-683,
1995.
pleased with the posItIve collaboration provided by
[l3] Sophocles J.Orfanidis, "Electromagnetic Waves and Antennas, Chap. 17-
researchers from the Department of Signal Theory, Radiation from Apertures, " Rutgers University, August, 2010.
Networking and Communications at University of Granada [14] D. Lymberopoulos, Q.Lindsey and A.Savvides, "An Empirical
and from the Department of Informatics at University of Oslo. Characterization of Radio Signal Strength Variability in 3-D IEEE
802. 15. 4 Networks Using Monopole Antennas," EWSN, Zurich,
Switzerland, February, 2006.
REFERENCES
[15] G. Cherntanomwong y D. Suroso, "Indoor Localization System using
[I) S.Holm and C.Nilsen, "Robust ultrasonic indoor positioning using Wireless Sensor Networks for Stationary and Moving Target,"
transmitter arrays," in Proc. Int. Con! on Indoor Positioning and Indoor Information, Communications and Signal Processing (ICICS), Beijing,
Navigation, Zurich, September, 2010. China, 2011.

View publication stats