10-07-2012, 04:42 PM
The Effects of Variable Sound Speed on Localization in Underwater Sensor Networks
[attachment=27396]
Abstract
The field of Underwater Sensor Networks attempts to
set up a Wireless Sensor Network (WSN) underwater for the
purpose of exploration and monitoring by addressing challenges
unique to the oceanic environment. In an UnderWater Sensor
Network (UWSN), the knowledge of the location of a node can
both improve the operation of the network (e.g. geographical
routing) and add significance to the data that is collected. Many
localization techniques used in UWSN are based on
multilateration. Multilateration assumes the underwater speed of
sound to be constant. However, several studies have shown that
this speed varies with salinity, temperature and pressure. This
variation incorporates an error into the results obtained from
localisation. This paper presents an algorithm to find an estimate
of the underwater speed of sound at a particular location and
time. This estimate can be provided to multilateration to improve
the localization accuracy. The existing multilateration technique
need not be modified. Simulation studies show that our algorithm
improves the localization accuracy over a pure multilateration
technique that assumes the underwater speed of sound to be 1500
m/s.
Keywords- underwater sensor network; localisation; variable
speed of sound.
I. INTRODUCTION
Oceans cover vast areas and hold significant explored and
unexplored resources. Hence, there is great scope for the
exploration and monitoring of oceans. As Wireless Sensor
Networks (WSNs) excel at monitoring terrestrial
environments, there is a natural tendency to try and apply
them to oceanic environments. The resulting challenges spawn
the field of UnderWater Sensor Networks (UWSN).
Localisation is the process of determining the location of a
node in the network. The knowledge of location of a node in
the network can, not only improve the operation of the
network (e.g. geographical routing), but also add significance
to the data collected by the nodes (e.g. tsunami monitoring).
The accurate, range based localisation techniques used in
UWSN are based on multilateration. Multilateration
essentially constructs and solves a set of simultaneous
equations to localise the node. However, multilateration
assumes the speed of sound to be constant (usually 1500 m/s).
Several studies have shown that the underwater speed of
sound varies with salinity, temperature and pressure (depth)
[1, 2, 3]. The constant speed assumption incorporates an error
into the results of multilateration. We need to take into
consideration the variability of the underwater speed of sound
while performing localisation in order to yield an accurate
estimate of the position [4].
We propose an algorithm to estimate the underwater speed
of sound by measuring the signal characteristics between the
communicating nodes. This estimate of underwater sound
speed can be used by multilateration or multilateration based
localisation techniques so as to yield a better estimate of the
nodes location.
A. Motivation
The underwater speed of sound varies with salinity,
temperature and pressure (depth). However, the technique of
multilateration assumes the speed to be constant. On account
of this, an error is incorporated into the location of the node as
determined by multilateration. Till date, to the best of our
knowledge, there has been no significant research that has
focused on solving this problem. Reference [5] has proposed a
localisation algorithm for UWSN that does not require the
network to be time synchronised. It handles the effect of
variable underwater speed of sound by applying a standard
model of the variation of the sound speed and focuses on the
problem of localisation in a network that is not times
synchronised. Our work differs from [5] in that, we propose an
algorithm to estimate the underwater speed of sound from the
signal characteristics instead of proposing a new localisation
algorithm altogether. The underwater speed of sound
estimated by our algorithm can be used by any localisation
algorithm that used multilateration so as to improve the
localisation accuracy.
II. BACKGROUND
A. Underwater Sensor Networks
An UWSN is set of autonomous nodes that are deployed
over a particular region and that can communicate with each
other so as to form a network, measure certain physical
parameters of the region in which they are deployed and
forward these measurements to a control station without any
human intervention.
The challenges that UWSNs attempt to address are as
follows:
1) Acoustic Communication: RF waves are not suitable
for underwater communication. They are highly attenuated in
shallow, semi-conductive ocean water. All electromagnetic
waves, even those emitted by powerful lasers, are absorbed
almost completely within 1 km of propagation in ocean water
[6]. The Mica2 motes with Chipcon radio operating at 433
kHz when used underwater have a typical range of 1 m [6].
However, it is possible to use long range RF radios for a short
communication range of 6-10 m at a data rate of 1-8 kbps at
122 kHz frequency [7]. Large antenna and higher transmit
power are required to transmit RF waves underwater.
Similarly, the use of light is also restricted. Hence, acoustic
waves prove to be the best option. However, the use of sound
creates several constraints on UWSNs.
2) Low Bandwidth: The bandwidth available to acoustic
(sound) waves is of the order of 1 kHz [3]. This is much lower
than that available to RF waves in terrestrial networks (about 1
MHz).
3) High Latency: The underwater speed of sound (1500
m/s) being five orders of magnitude smaller than the speed of
RF in air or vacuum causes very high propagation time which
can causes several protocols to break down. For example,
traditional protocols for time synchronisation fail in case of
such large propagation delays.
4) Variable Latency: Not only is the propagation delay
large, it is also variable. The underwater speed of sound is a
function of salinity, temperature and pressure (depth) [1, 2, 3].
Therefore, the propagation delay varies from region to region
and from time to time. This variable latency causes an error in
localisation.
5) Adverse Communication Medium: On account of
various impurities dissolved in it, ocean water is a very
heterogeneous communication medium. Acoustic waves
propagating in ocean water suffer from absorption, scattering,
diffraction. The presence of air bubbles accentuates this effect.
Further, in shallow water, the acoustic waves are refracted and
reflected from the ocean surface and ocean bottom. This leads
to multipath propagation and, therefore, to interference at the
receiver. Further, the oceans are abundant with sources of
noise such as sailing ships, aquatic life and wind blowing
across the ocean surface.
6) Energy Constraints: As in WSNs, UWSNs are
battery powered and, therefore, energy constraint devices. The
two main strategies for recharging batteries in WSNs are to
use solar energy and to harvest energy from RF waves in the
environment. Both solar energy and RF energy propagate
poorly through the oceanic environment. Recharging of
underwater nodes remains a fairly unexplored area in UWSNs.
This accentuates the need energy conservation in UWSNs.
7) Failure Prone: The highly corrosive nature and high
pressures present in typical oceanic environment make nodes
highly susceptible to failure due to fouling and corrosion.
B. Localisation:
The process of localisation involves calculating the
position of one node given the position of several other nodes
whose position is known. The nodes whose positions are
known are known as anchors. In this paper, we are concerned
with a localisation technique called multilateration. It involves
solving the following simultaneous equations to obtain the
unknown location (x, y, z).
+ + +
= 0
=
2
−
2
=
2
−
2
=
2
−
2
= − −
+ +
+
+
+
‘’ is the Time Diffrence of Arrival (TDoA) and ( , , )
is the location for the th node and ‘’ is the sound speed.
These equations can be solved by using several robust
techniques from linear algebra such as Singular Value
Decomposition (SVD) and Gaussian Elimination.
Reference [8] classifies localisation techniques into range
based and range free techniques. The range based techniques
are used when localisation accuracy is important and most of
them are based on localisation.
III. PROBLEM DEFINITION
From the equations derived for multilateration in the
section II (B), we can see that in order to calculate the
constants , , and
, we need to know the underwater
speed of sound ‘’. Usually, the underwater speed of sound is
taken to be a constant of 1500 m/s. However, as the speed of
sound varies, this incorporates an error into the position
obtained from localisation. In this paper, we attempt to reduce
this localisation error.
IV. SOUND SPEED ESTIMATION
In this work, we find an estimate of the underwater speed
of sound at a particular place and time using signal
characteristics. This estimate is them provided to the
multilateration technique so as to obtain a better estimate of
the node’s location.
Ideally, the underwater speed of sound would be different
between each pair of nodes. The technique of multilateration,
however, assumes the underwater speed of sound to be the
same between each pair of nodes. Correcting this assumption
would require the technique of multilateration to be modified.
In this work, we attempt to find a suitable estimate of the
underwater speed of sound without necessitating any
modifications to the multilateration technique. Therefore, our
solution is compatible with current localisation techniques
already in use.
[attachment=27396]
Abstract
The field of Underwater Sensor Networks attempts to
set up a Wireless Sensor Network (WSN) underwater for the
purpose of exploration and monitoring by addressing challenges
unique to the oceanic environment. In an UnderWater Sensor
Network (UWSN), the knowledge of the location of a node can
both improve the operation of the network (e.g. geographical
routing) and add significance to the data that is collected. Many
localization techniques used in UWSN are based on
multilateration. Multilateration assumes the underwater speed of
sound to be constant. However, several studies have shown that
this speed varies with salinity, temperature and pressure. This
variation incorporates an error into the results obtained from
localisation. This paper presents an algorithm to find an estimate
of the underwater speed of sound at a particular location and
time. This estimate can be provided to multilateration to improve
the localization accuracy. The existing multilateration technique
need not be modified. Simulation studies show that our algorithm
improves the localization accuracy over a pure multilateration
technique that assumes the underwater speed of sound to be 1500
m/s.
Keywords- underwater sensor network; localisation; variable
speed of sound.
I. INTRODUCTION
Oceans cover vast areas and hold significant explored and
unexplored resources. Hence, there is great scope for the
exploration and monitoring of oceans. As Wireless Sensor
Networks (WSNs) excel at monitoring terrestrial
environments, there is a natural tendency to try and apply
them to oceanic environments. The resulting challenges spawn
the field of UnderWater Sensor Networks (UWSN).
Localisation is the process of determining the location of a
node in the network. The knowledge of location of a node in
the network can, not only improve the operation of the
network (e.g. geographical routing), but also add significance
to the data collected by the nodes (e.g. tsunami monitoring).
The accurate, range based localisation techniques used in
UWSN are based on multilateration. Multilateration
essentially constructs and solves a set of simultaneous
equations to localise the node. However, multilateration
assumes the speed of sound to be constant (usually 1500 m/s).
Several studies have shown that the underwater speed of
sound varies with salinity, temperature and pressure (depth)
[1, 2, 3]. The constant speed assumption incorporates an error
into the results of multilateration. We need to take into
consideration the variability of the underwater speed of sound
while performing localisation in order to yield an accurate
estimate of the position [4].
We propose an algorithm to estimate the underwater speed
of sound by measuring the signal characteristics between the
communicating nodes. This estimate of underwater sound
speed can be used by multilateration or multilateration based
localisation techniques so as to yield a better estimate of the
nodes location.
A. Motivation
The underwater speed of sound varies with salinity,
temperature and pressure (depth). However, the technique of
multilateration assumes the speed to be constant. On account
of this, an error is incorporated into the location of the node as
determined by multilateration. Till date, to the best of our
knowledge, there has been no significant research that has
focused on solving this problem. Reference [5] has proposed a
localisation algorithm for UWSN that does not require the
network to be time synchronised. It handles the effect of
variable underwater speed of sound by applying a standard
model of the variation of the sound speed and focuses on the
problem of localisation in a network that is not times
synchronised. Our work differs from [5] in that, we propose an
algorithm to estimate the underwater speed of sound from the
signal characteristics instead of proposing a new localisation
algorithm altogether. The underwater speed of sound
estimated by our algorithm can be used by any localisation
algorithm that used multilateration so as to improve the
localisation accuracy.
II. BACKGROUND
A. Underwater Sensor Networks
An UWSN is set of autonomous nodes that are deployed
over a particular region and that can communicate with each
other so as to form a network, measure certain physical
parameters of the region in which they are deployed and
forward these measurements to a control station without any
human intervention.
The challenges that UWSNs attempt to address are as
follows:
1) Acoustic Communication: RF waves are not suitable
for underwater communication. They are highly attenuated in
shallow, semi-conductive ocean water. All electromagnetic
waves, even those emitted by powerful lasers, are absorbed
almost completely within 1 km of propagation in ocean water
[6]. The Mica2 motes with Chipcon radio operating at 433
kHz when used underwater have a typical range of 1 m [6].
However, it is possible to use long range RF radios for a short
communication range of 6-10 m at a data rate of 1-8 kbps at
122 kHz frequency [7]. Large antenna and higher transmit
power are required to transmit RF waves underwater.
Similarly, the use of light is also restricted. Hence, acoustic
waves prove to be the best option. However, the use of sound
creates several constraints on UWSNs.
2) Low Bandwidth: The bandwidth available to acoustic
(sound) waves is of the order of 1 kHz [3]. This is much lower
than that available to RF waves in terrestrial networks (about 1
MHz).
3) High Latency: The underwater speed of sound (1500
m/s) being five orders of magnitude smaller than the speed of
RF in air or vacuum causes very high propagation time which
can causes several protocols to break down. For example,
traditional protocols for time synchronisation fail in case of
such large propagation delays.
4) Variable Latency: Not only is the propagation delay
large, it is also variable. The underwater speed of sound is a
function of salinity, temperature and pressure (depth) [1, 2, 3].
Therefore, the propagation delay varies from region to region
and from time to time. This variable latency causes an error in
localisation.
5) Adverse Communication Medium: On account of
various impurities dissolved in it, ocean water is a very
heterogeneous communication medium. Acoustic waves
propagating in ocean water suffer from absorption, scattering,
diffraction. The presence of air bubbles accentuates this effect.
Further, in shallow water, the acoustic waves are refracted and
reflected from the ocean surface and ocean bottom. This leads
to multipath propagation and, therefore, to interference at the
receiver. Further, the oceans are abundant with sources of
noise such as sailing ships, aquatic life and wind blowing
across the ocean surface.
6) Energy Constraints: As in WSNs, UWSNs are
battery powered and, therefore, energy constraint devices. The
two main strategies for recharging batteries in WSNs are to
use solar energy and to harvest energy from RF waves in the
environment. Both solar energy and RF energy propagate
poorly through the oceanic environment. Recharging of
underwater nodes remains a fairly unexplored area in UWSNs.
This accentuates the need energy conservation in UWSNs.
7) Failure Prone: The highly corrosive nature and high
pressures present in typical oceanic environment make nodes
highly susceptible to failure due to fouling and corrosion.
B. Localisation:
The process of localisation involves calculating the
position of one node given the position of several other nodes
whose position is known. The nodes whose positions are
known are known as anchors. In this paper, we are concerned
with a localisation technique called multilateration. It involves
solving the following simultaneous equations to obtain the
unknown location (x, y, z).
+ + +
= 0
=
2
−
2
=
2
−
2
=
2
−
2
= − −
+ +
+
+
+
‘’ is the Time Diffrence of Arrival (TDoA) and ( , , )
is the location for the th node and ‘’ is the sound speed.
These equations can be solved by using several robust
techniques from linear algebra such as Singular Value
Decomposition (SVD) and Gaussian Elimination.
Reference [8] classifies localisation techniques into range
based and range free techniques. The range based techniques
are used when localisation accuracy is important and most of
them are based on localisation.
III. PROBLEM DEFINITION
From the equations derived for multilateration in the
section II (B), we can see that in order to calculate the
constants , , and
, we need to know the underwater
speed of sound ‘’. Usually, the underwater speed of sound is
taken to be a constant of 1500 m/s. However, as the speed of
sound varies, this incorporates an error into the position
obtained from localisation. In this paper, we attempt to reduce
this localisation error.
IV. SOUND SPEED ESTIMATION
In this work, we find an estimate of the underwater speed
of sound at a particular place and time using signal
characteristics. This estimate is them provided to the
multilateration technique so as to obtain a better estimate of
the node’s location.
Ideally, the underwater speed of sound would be different
between each pair of nodes. The technique of multilateration,
however, assumes the underwater speed of sound to be the
same between each pair of nodes. Correcting this assumption
would require the technique of multilateration to be modified.
In this work, we attempt to find a suitable estimate of the
underwater speed of sound without necessitating any
modifications to the multilateration technique. Therefore, our
solution is compatible with current localisation techniques
already in use.