Indoor Localisation using Wireless Sensor Networks

eggplantcinnabarMobile - Wireless

Nov 21, 2013 (3 years and 6 months ago)

130 views

ASK
-
IT International Conference


October 2006


1

Indoor Localisation using Wireless Sensor Networks

Chris Fowler, Jurgen Wager, Richard Fairchild

Transport Operations Research Group (TORG)

School of Civil Engineering and Geosciences

Newcastle University

NE1 7RU

Tel: +44 191 2226426, Fax: +44 191 2226505

Email: {c.p.fowler, r.g.fairchild, jurgen.wager}@ncl.ac.uk



Abstract


ASK
-
IT aims to harness the combined power of potable devices, wireless communication
protocols, bespoke service development, and adaptive interface design, to deliver on
-
demand
service
s for mobility impaired users. A key service to be provided by ASK
-
IT is localisation.
The context for the research presented in this paper is a user wishing to navigate within a
building, or building complex. This may be on a room
-
by
-
room basis, or locali
sing to a
specific coordinate relative to a building plan


an x,y position. Wireless sensor networks use
smart dust, or MOTEs to form ad
-
hoc networks. Using trilateration, a position can be
determined for an unknown MOTE located within a fixed network of
MOTEs. A reasonable
accuracy of between 0.5
-
1.2 meters has been achieved using the techniques presented here.
ASK
-
IT International Conference


October 2006


2

1. Introduction


The growth of mobile computing has been unprecedented. Mobile phones and personal
computing assistants (PDAs) are now widespread;

their use being distributed across most
population demographics, with this trend set to continue. The combination of portability and
the increasing coverage of wireless communications mean that Internet and phone services are
being delivered to a wider a
nd more technically
savvy

population. Services associated more
closely with static desktop computers are now available on mobile devices and wherever there
is network access. Consequently, terms such as
Pervasive Computing
,
Ambient Intelligence

and
Locatio
n based Services

are being used in connection with the latest technological
advances and the buying public are quick to test these concepts in the real world. Combining
portability, network access, the hardware capabilities of devices and network accessibl
y
information, provides a dynamic, flexible and extensible platform for research development
and evaluation. The ASK
-
IT project
[7]

is harnessing this platform to aid a user base with a
specific set of needs.


The mobility
-
impa
ired (MI) user covers a broad range of impairments. While avoiding
singling out any one set of users as in need of special attention, the reality is that certain user
profiles are not as yet included in the accessible information age in which we now live;
yet
this population. One of the key aims of the ASK
-
IT project is to bridge the gap
-

bringing the
clear benefits of on demand information service to all users, regardless of their personal
circumstances. It’s a challenge, but a laudable one.


ASK
-
IT harne
sses the combined power of device, communication, bespoke service
development, and adaptive interface design, to deliver on
-
demand services. These services are
filtered and delivered in a way considered most accessible to the user concerned. For example,
t
he visually impaired user might desire an audible interface, with voice
-
activated commands.
Cross
-
modal communications where sound, vision and perhaps haptics may be used in
combination, to both capture users’ commands and convey system events. To this end
, ASK
-
IT is concerned with helping users to access and interpret information across a wide range of
everyday activities: tourism and travel, navigation, domotics, and access to information


bridging the gap between the virtual space and the real world of
everyday peoples’ lives. As
such, the project has identified key, high level services, which when combined can satisfy a
set of clearly defined use
-
cases. By combining these use cases, usage scenarios have been
developed before being evaluated. Users will
access services through commonly available
devices such as phones and PDAs, as well as a specifically deigned device being developed as
part of the project.


A key service to be provided by ASK
-
IT is localisation. The context for the research presented
in
this paper is a user wishing to navigate within a building, or building complex. This may be
on a room
-
by
-
room basis, or localising to a specific coordinate relative to a building plan


an
x,y position. The goal of assisted route guidance in predicated o
n knowing where one is
before being able to navigate to where one wants to be. For the ASK
-
IT project a localisation
service is being developed which will calculate and deliver a position (encoded as a JSR179
[8]

address) to the

requester. This position can then be displayed on a map, or plan and
presented to the user in a form suited to their impairment.


Wireless Sensor Networks use Smart Dust, or MOTES to form ad
-
hoc networks. An ad
-
hoc
network is one that is formed from a set

of communication nodes, which come together in an
ad
-
hoc way. They allow for very flexible and dynamic network topologies to be formed in
ASK
-
IT International Conference


October 2006


3

real
-
time. Nodes are allowed to join and leave the network at any time. The network is broken
if any two nodes, or gr
oups of nodes, do not have a connecting node. A connection is made
when the signal pattern (the area of active communication) of one MOTE intersects with that
of a neighbouring MOTE. Network topology is therefore an important factor as is the distance
betw
een MOTES. Given this infrastructure, groups of MOTES within a building for example
can form an ad
-
hoc network, which can then be exploited for various applications. An active
MOTES network provides a communications platform through which data can be propa
gated
and received between on
-
board pluggable sensors, with applications in many areas. For
example, environmental monitoring and control is possible through such an infrastructure. In
the case of localisation, it is the communication characteristics of th
e MOTES that provide a
unique opportunity for localisation. Using well
-
understood received signal strength
characteristics it is possible to determine the distance between two points on a network given
the signal strength. If this calculation is repeated a
cross a number of MOTES for a given point
then a position can be determined. By using a fixed network within the building and have a
roaming,
mobile

MOTE, which is able to move through the network while still remaining
connected, it is possible to determin
e the position of the mobile MOTE relative to the fixed
network. If the fixed MOTES are mapped onto a building plan, then a position relative to the
building can be determined.


This paper presents work on the effectiveness of localisation techniques for i
ndoor
localisation using the MicaZ MOTE and sensor board. The purpose is to determine accuracy
and highlight any issues that may need to be overcome in order to use a MOTE based
localisation service for ASK
-
IT. The technique used is trilateration. This tec
hnique is for
MOTES in a two
-
dimensional plane
[6]
. ASK
-
IT requires localisation within a three
-
dimensional network, so by using a correcting algorithm to map from 3D to 2D a position can
be found at any point and height. This p
aper is structured as follows: Section 2 discusses
MOTES and their use in localisation and also reviews existing work in the area; Section 3
describes scenarios from which a set of requirements for internal localisation using MOTES
can be determined; Secti
on 4 describes localisation techniques used in this work; Section 5
presents the evaluation of the results relative to the requirements; Section 6 concludes the
paper.


2.

Wireless Sensor Networks

Wireless sensor networks is a name given to two or more device
s that communicate with each
other and also incorporate an ability to access, or sense, their immediate environment.


The devices in a wireless sensor network communicate using data transmission protocols such
as Transmission Control Protocol (TCP) over a
wireless data transmission at the physical
layer. Standard physical layer solutions include wi
-
fi (IEEE 802.11b/g), Bluetooth (IEEE
802.15.1), Zigbee (IEEE 802.15.4) and other non
-
standard proprietary protocols.


Because of the computational power needed
to run such protocols, most devices used in a
wireless sensor network contain on
-
board processors. The generic term for one of these
devices is a mote. It is widely accepted that motes are the modern day precursors to nanometre
(10x10
-
6
m) sized smartdust
devices.


Many manufacturers of motes incorporate a modular design consisting of a main board with
the processor, memory, wireless device, aerial and batteries. A connector on this mainboard
ASK
-
IT International Conference


October 2006


4

allow the addition of daughterboards that can contain sensors su
ch as temperature, light,
sound, magnetism etc. There are also application specific sensors such as GPS.

This project will be using the Crossbow Technologies Mica2 motes, transmitting at 906MHz.
Before designing a wireless sensor network, the character
istics of the motes needs be taken
into consideration. These include inter
-
mote distance and mote height from the ground. In
[1]
[2]

it is suggested that a mote height of around 100 cm is optimal and

a viable
communication distance of around 50
-
55 meters.


There are several methods of determining the position of a single mote in a network: range
-
based or range
-
free. Broadly speaking, a range
-
based approach determines the range of one
mote to another
by measuring the distance between the two motes using the received signal;
this could be by measuring the signal propagation time or comparing the received signal
strength against a theoretical or empirical model to determine distance. Although using the
received signal strength indicator is deemed a bad estimator of link quality, work has been
carried out to prove that new systems based on IEEE 802.15.4 standard can provide a reliable
indicator of link quality and hence provide justifiable distances
[3]
.


Range free methods generally include anchor motes that have fixed locations, the locations of
mobile motes in the network can be determined by the calculation of the relative distances to
several anchor motes


a process genera
lly referred to as triangulation
[5]
. Other techniques
include ring overlapping which is reported to achieve a more accurate location estimate than
the triangulation method
[5]
.

3.

Localisation System S
cenarios

To test and evaluate the use of wireless sensor networks (WSN) for effective indoor
localisation for MI users, a series of usage scenarios have been devised. Each will evaluate the
effectiveness and suitability of WSN for a given set of requiremen
ts. The requirements at this
stage are only concerned with measurable parameters sufficient to fulfil the scenario aims.

Zones

Localisation is often associated with a particular point (x,y,z coordinate), however, this level
of detail is not always a requi
rement. The zones scenario described here is one such case in
point and has a particular relevance to ASK
-
IT use
-
cases.


The ASK
-
IT use
-
case 4.2 illustrates a system interaction where a user requires help via human
involvement. For example, this might be
the case where a MI user may have fallen, or they’re
lost in unfamiliar surroundings. In
-
order to provide assistance the helper must be able to
locate the subject. An additional scenario, might be a building that must be evacuated and as
part of that evacu
ation able helpers are dispatched to locate and assist any mobility
-
impaired
persons still in the building. A possible system is illustrated in
Figure
1
. The setup of the
network involves having a numbered MOTE placed in each room,

and any connecting
corridors
-

each MOTE can be queried in turn from the server using its unique identifier. The
placement of the MOTES must be such that they form an ad
-
hoc network. The mobile MOTE
is associated with a particular user. An additional adva
ntage of this system is that both
attendance and location can be determined through the one system; if the user is present and
where the user is located.
In this case the location is a room.

ASK
-
IT International Conference


October 2006


5


Figure
1
. Zones Localisation Scenario

Following the call to evacuate, most likely through sounding of the fire alarm, the scenario
runs as follows:
Figure
1
.1 the system requests the location of all mobile MOTEs;
Figure
1
.2
Each fixed MOTE is a
sked to report the signal strengths of all MOTEs visable within its
signal footprint;
Figure
1
.3 the server determines form the signal strengths received where the
mobile MOTES is located
-

as the fixed MOTE positions are known, th
en the mobile position
will be the room where the fixed MOTE reported the highest signal strength when connecting
to the mobile MOTE.
Figure
1
.4 the system informs the able bodies helper where the MI user
can be found by matching t
he known location of the fixed MOTE to the localted mobile
MOTE.
Figure
1
.5 is an optional part to this scenarios involves the server sending
information/instructions to the MI user via a WiFi network to their PDA display.


In this

scenario the server requests the location of the mobile MOTE. Given the need to locate
MI users then this is appropriate. The same setup is able to test the case where the MI user
requests their location. They must send a localisation request to the serve
r via the MOTES
network. On receiving this request the localisation is done and the result sent.

Open Access

Room, or zone level granularity is not always sufficient. Consider a visually impaired user
wanting to navigate through an unfamiliar space, for e
xample an office or canteen area, or a
large shopping center. For this they will need localisation based on an accurate position. This
scenario illustrates a plausible use
-
case for determining a coordinate position within an open
space.


Figure
2
. Open Area Localisation Scenario

ASK
-
IT International Conference


October 2006


6

In this scenario the MI user initiates the localisation request.

Figure

2
.1 the MI user’s device requests a localisation coordinate from the system passing the
ident
ifier for the mobile MOTE they have; this is sent via the WiFi network to the server.

Figure

2
.2 the server requests the signal strength data from all MOTES in the network.

Figure

2
.3 each MOTE reports the signal strength between itself and any other visible
MOTE.
Figure
2
.4 the server filters the results using only those signals between the
static MOTEs and the mobile MOTE in question. Using these dat
a and a 3D to 2D
ASK
-
IT International Conference


October 2006


7

mapping algorithm determines the x,y position of the mobile MOTE.

Figure
2
.5 the position is mapped relative to the room layout and sent to the MI user’s PDA
via the WiFi network.

Requirements for scenarios

Examini
ng the scenarios above a set of high
-
level requirements for the system to support them
can be derived.
Table
1

shows a list of requirements with a description of each. The
evaluation discussion will refer back to this list to demon
strate the validity of the approach
take.


Requirement

Discussion

Network

Network connectivity is a key component of the system. The nature of ah
-
hoc
networks are such that provided nodes are within range of their neighbours,
then the network is connected
. The network required consists of a fixed set of
MOTEs and mobile MOTEs, which roam while remaining connected. The
topology of the fixed network needs to be such that all MOTEs are connected,
with a link to the server from one MOTE. Although we can locate

within a 3D
network, the fixed motes must be in the same plane.

Connectivity

The topology of the fixed network needs to be understood and mapped to the
physical space.

Signal
Strength

The signal strength is required in order to determine the position o
f the mobile
MOTE. To understand how this parameter performs a calibration exercise is
carried out.

Mapping

A position of the mobile MOTE relative to the fixed network must be
determined. In
-
order to find this position relative to the building or room, t
he
position must be mapped to building plans and room layouts.

Interface

At this stage a simple layered interface showing a map layout and the
requested position is sufficient to demonstrate results.

Accuracy

A calibration and determination of the accura
cy of the MOTEs for localisation
under a set of conditions typical of those found in the scenarios.

Table
1

Requirements

ASK
-
IT International Conference


October 2006


8

The above requirements are suitable for a system that enacts all stages of the scenarios
presented. This pape

ris focusses in evaluating only the use of signal strength for finding a
position of a given set of MOTEs.

4. Localisation using MOTES

For this work trilateration was used to localise a mobile MOTE within a static fixed network.
Trilateration uses simple
geometry to determine the locale of a variable point using its
distance from a set of known fixed points. Using this method a minimum of 3 fixed points is
required. MOTEs must be in a single plane.



“Standing at B, you want to know your location relative

to the reference points P1,
P2, and P3. Measuring r1 narrows your position down to a circle. Next, measuring
r2 narrows it down to two points, A and B. A third measurement, r3, gives your
coordinates at B. A fourth measurement could also be made to reduce

error.”
[6]

Figure
3

Trilateration technique for localisation on a 2D plane

Calibration Testing

Many factors affect the outcome of localisation using motes
-

interference caused by furniture
and fitting
, reflections from surfaces and walls can cause fluctuations or non
-
linear
characteristics. In the test area an attempt to minimise these errors by mote placement, and
clearing of furniture was done, however, these conditions are normal in a real world sit
uation.
ASK
-
IT is aiming to localise users in many different contexts, so although a
clean

test area
would give the best calibration and testing result, a more realistic setup would give realistic
results and help to answer the question as to whether motes

can be used for the ASK
-
IT
indoor localisation service. The test area is Space 2 in Culture Lab on Newcastle University’s
city campus, see
Figure
4
.

ASK
-
IT International Conference


October 2006


9


Figure
4

Space 2 Culture Lab, Newcastle University

Th
e first stage, and in order to determine effective localisation, calibration of the equipment is
required. The method to be used requires an accurate distance between the known fixed motes
and the mobile mote.
Figure
5

below shows
a plot of signal strength over distance. This was
determined by first marking a line with ½ meter gradations on the floor of the test area to a
distance of 5 meters. A receiving mote was placed at the 0m mark. Then, one by one with
each mote in the test se
t, a mote was place at ½ meter intervals starting from 1m up to 5m. At
each point the received signal strength was measured. This measurement was repeated 30
times and an average taken for each mote at each gradation. As can be seen in
Figure
5

a very
consistent signal to distance ration is seen for all the test motes. From 0
-
3.5 meters a
reasonably linear relationship is shown. Beyond 3.5 meters the signal becomes less linear. To
achieve a good result it was therefore decided to li
mit the range of the motes to 3.5 meters so
as to give the most accurate locale for the mobile mote.


-85
-80
-75
-70
-65
-60
-55
-50
-45
-40
0.00
1.00
2.00
3.00
4.00
5.00
6.00
Distance (m)
Signal Strength (dB)

Figure
5

Signal strength over distance ratio

2D to 3D localisation

Trilateration is performed using the signal to distance rati
o to determine the distance between
a transmitting and receiving mote on a single 2D plane. This will determine the position of an
unknown mote within a network of fixed motes, where their positions are known. This will
not, however, be the case in a real
-
world scenario such as an ASK
-
IT user wishing to navigate
inside a shopping centre of hospital complex. In this case the fixed network is likely to be
located on the ceiling or high enough on the wall infrastructure to be out of harms way. The
user will ca
rry the mobile mote. With the fixed motes at ceiling height and the mobile mote at
approximately 1 meter height, a single plane is not possible
-

the motes are now arranged in a
3
-
dimentional network. To compensate for this a projection from the 3D to 2D i
s needed.

ASK
-
IT International Conference


October 2006


10



Figure
6

3D to 2D projection

Figure
6

above illustrates how the MOTE localisation compensates for the 3
-
dimentional
aspect of the tests. The red and green dots represent mobile motes, each po
sitioned at a
different height.

5. Evaluation

The evaluation was split into two parts. The first looked at localising in a 2D plane. For this a
set of four fixed motes where positioned to form a network at approximately 1 meter above
the floor level. The p
ositions of the motes were recorded. A fifth mote was then place in the
network. It’s position was also recorded relative to the previous four so that a comparison
between the actual position and the calculated position could be made. A series of eight tes
ts
where then carried out and the results recorder. Each test was repeated multiple times to show
groupings of positions. For each test the fifth mote was rotated through 45 degrees. This was
to see what affect this would have on the resulting position cal
culated. In the real world a user
will be actively moving and turning so this was considered important to examine.



Figure
7

Results for 2D network with mote rotation

The results in
Figure
7

show a wide r
ange of positions. Those for 45, 90 and 135 degrees are
the closest to the expected with only ½ meter error from the actual position. The full range of
results shows errors of up to 6 meters from the expected. It is difficult to determine why this
ASK
-
IT International Conference


October 2006


11

is. Perh
aps through signal reflection, or signal pattern as the groupings for each angle suggest
that this characteristic does have an impact.


The tests were repeated for a 3D network. The fixed motes were positioned on the lighting
tracks with the locations proj
ected on to the floor plan so the mobile motes location could be
measured relative to the fixed motes.



Figure
8

Results for 3D network with mote rotation

The results in
Figure
8

show a better position,
with a good grouping average positions and
most results within 1 meter of the expected. The overall spread of results was better for the 3D
network. It is not clear as to why this is the case.

6. Conclusion

This paper has presented work on evaluating the s
uitability of MicaZ wireless motes for
indoor localisation. The technique used enabled a mobile mote to be localised within a static
network of motes of known positions. Both 2D and 3D networks where examined. The results
for both networks produced a posit
ion with an accuracy of ½ meter for certain mote rotation
angles relative to the fixed network. An encouraging result was the 3D network as this
showed greater consitency across the full range of angles with an accuracy to within 1.2m for
all rotation angl
es.


References

[1]


Anastasi, G., Borgia, E., Conti, M., Gregori, E., Passarella, A. (2005) Understanding the
real behavior of Mote and 802.11 ad hoc networks: an experimental approach.
Pervasive
and Mobile Computing
, no 1, pp 237
-
256.

[2]


Fairchild, R. (2003)

Wireless Sensor Networks in Transport
. MSc Thesis, University
of Newcastle upon Tyne, England.

[3]


Srinivasan
, K.,

and Levis
, P. (2006)
RSSI is Under Appreciated.

In
Proceedings of the
Third Workshop on Embedded Networked Sensors (EmNets 2006)
.

ASK
-
IT International Conference


October 2006


12

[4]


Liu, C., Wu,

K., and He, T. (2004)
Sensor localization with Ring Overlapping based
on Comparison of Received Signal Strength Indicator
. In
2004 IEEE International
Conference on Mobile Ad
-
hoc and Sensor Systems (MASS)
, October.

[5]


He, T., Huang, C., Blum, B., Stankovic,
J., Abdelzaher, T. (2005)
Range
-
Free
Localization and its Impact on Large Scale Sensor Networks
.
ACM Transactions on
Embedded Computing System (TECS)
, Vol. 4(Issue 4),
TECS.

[6]


Wikipedia
Trilateration

http://en.wikipedia.org/wiki/Trilateration


[7]


Ambient Intelligence System of Agents for Knowledge
-
based and Integrated Services for
Mobility Impaired Users.
ASK
-
IT ( IST
-
2003
-
511298),

http://www.ask
-
it.org/

[8]


JSR 1
79: Location API for J2ME, http://jcp.org/en/jsr/detail?id=179