Access to this full-text is provided by Wiley.
Content available from International Journal of Distributed Sensor Networks
This content is subject to copyright. Terms and conditions apply.
Research Article
Application of Wireless Sensor Networks for
Indoor Temperature Regulation
Biljana Risteska Stojkoska,1Andrijana Popovska Avramova,2and Periklis Chatzimisios3
1Faculty of Computer Science and Engineering, Saints Cyril and Methodius University, 1000 Skopje,
e Former Yugoslav Republic of Macedonia
2Department of Photonics Engineering, Technical University of Denmark, 2800 Kongens Lyngby, Denmark
3Department of Informatics, Alexander TEI of essaloniki, 57400 essaloniki, Greece
Correspondence should be addressed to Biljana Risteska Stojkoska; biljanastojkoska@yahoo.com
Received December ; Accepted February ; Published May
Academic Editor: Hongke Zhang
Copyright © Biljana Risteska Stojkoska et al. is is an open access article distributed under the Creative Commons
Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
Wireless sensor networks take a major part in our everyday lives by enhancing systems for home automation, healthcare,
temperature control, energy consumption monitoring, and so forth. In this paper we focus on a system used for temperature
regulation for residential, educational, industrial, and commercial premises, and so forth. We propose a framework for indoor
temperature regulation and optimization using wireless sensor networks based on ZigBee platform. is paper considers
architectural design of the system, as well as implementation guidelines. e proposed system favors methods that provide energy
savings by reducing the amount of data transmissions through the network. Furthermore, the framework explores techniques for
localization, such that the location of the nodes can be used by algorithms that regulate temperature settings.
1. Introduction
Wireless sensor networks (WSNs) are able to eciently sense
various parameters with high accuracy and low power con-
sumption. e development of sensors and networks based
on sensor nodes has impacted and changed our everyday life.
Engaging WSNs in home and industrial monitoring systems,
medicine and healthcare systems, entertainment, education,
andsoforth,hasenlightenedandimprovedtheconceptof
modern living. A wireless sensor network consists of three
major elements []: sensor unit (used to take measurements),
computing unit (used to process data), and communication
unit (used to enable communication among the wireless
nodes). Dierent radio technologies can be used for com-
munication, such as ZigBee, Wi-Fi, Bluetooth, and Global
Systems for Mobile Communications (GSM). ZigBee as an
emerging technology has been proven to make WSN self-
congurable and self-healing while operating at low power
consumption [], a feature that is very important for wireless
sensors.
Intelligent smart home frameworks have been proposed
recently by the research community in [,]. e proposed
systems are used to monitor and report dierent parameters
in a home environment such as temperature, humidity,
andlight,aswellascontrollingdierentelectricaldevices
for lightning, air conditioning, or heating. In [], energy
optimization is based on a dynamic programming algorithm
that controls the usage of energy and sells it back to the
smart grid. In [], the authors have proposed a prototype
system for temperature monitoring in a university campus.
e purpose of this system is to provide optimal manage-
ment of the cooling system in order to reduce the power
consumption. e system consists of the client part with
web-based interface and MySQL database and two types of
nodes: coordinator node that is responsible for data gathering
and terminal nodes that measure temperature, humidity, and
light intensity. is prototype application is implemented
on a centralized network (utilizing a star topology). e
maximum distance between the coordinator and the terminal
nodes is meters in open space or meters when obstacles
Hindawi Publishing Corporation
International Journal of Distributed Sensor Networks
Volume 2014, Article ID 502419, 10 pages
http://dx.doi.org/10.1155/2014/502419
International Journal of Distributed Sensor Networks
are present. Both terminal and coordinator nodes use the
same microcontroller, which is based on the Arduino []
boards and XBee [] communication modules that support
the ZigBee standard. is system has several drawbacks.
Centralized network could not cover campus taking into
account the limitations of the maximum distance between the
terminal nodes and the coordinator. e soware does not
support a coordinator to store data in a MySQL database, so
the idea that they should be stored at the client’s side is an
inappropriate solution. Arduino microcontrollers are expen-
sivesolutionforWSNbecauseoftheircomplexity,whichis
absolutely unnecessary for nodes that have a primitive task to
measure three physical parameters. In [], a WSN system is
proposed to control temperature. Unlike previous approaches
thatfocusedonthedesignofthenetwork,in[]theemphasis
is placed on data processing. e authors propose an analysis
of temperature readings using the variogram in order to make
a prediction of the temperature at each possible location in
the room. In [],aWSNbasedonstar-topologyisused
for temperature monitoring in a greenhouse. Since radio
range of the nodes is m, nodes can directly send their
data to the base station. e authors in []presentaweb-
based WSN interface that uses state-of-the-art technologies
for ecient habitat monitoring. e system is based on Mica
motes that are in interaction with remote users. e authors
in [] have installed a modular and extensible WSN in
a test and reference household called VILLASMART. e
energetic behavior of the building is modeled using indoor
and outdoor WSN readings (air and water temperature, solar
radiation sensor, weather conditions and power consumption
information). Grey-box estimation method is used for model
parameter determination, thus more precise predictions of
theindoortemperatureareachieved.
Our work proposes a wireless sensor network framework
for indoor temperature regulation (WSN-FITR). Homes,
classrooms, and halls are oen heated up by a number of tem-
perature controlled heaters. Users are usually not interested
in controlling the temperature at each separate heater. e
radiators, for example, are normally located just above the
oor or below windows and at the room’s walls. Furthermore,
the measurements do not show the real room temperature
as the temperature sensors are located just next to the
heaters. Neighbouring rooms with own heating elements also
inuence the temperature in the controlled room. is paper
presents detailed guidelines for ecient integration of WSN
into intelligent system for indoor temperature regulation. We
have discussed dierent WSN topologies regardless of the
deployment area. is paper shows how the node localization
methods can be used for room temperature optimization in
order to provide the most optimal tradeo between the time
it takes to reach the wanted temperature at a specic part
of the room and energy consumption. Even more, reduction
of the required data transmission through prediction meth-
ods is considered, which is important in order to increase
the battery life of the nodes and to extend the network
lifetime.
e rest of this paper is organized as follows. In the
Section , the framework architecture of WSN-FITR is pre-
sented, while Section provides an overview of ZigBee and
network topology of the proposed framework. Section gives
a detailed explanation of techniques for indoor localization
and clustering with respect to characteristics of the environ-
ment where WSN is deployed. Section compares techniques
for data reduction in WSN. Finally, we conclude this paper in
the last section.
2. System Architecture
e proposed WSN-FITR system for indoor temperature
regulation has a simple system architecture (Figure ). e
wireless nodes deployed in each room are programmed to
monitor the temperature. ere are two types of nodes:
sensor regulators and temperature controllers which are
interconnected in a ZigBee network. Sensor-regulator nodes
take samples at predened intervals and send data to the con-
troller node. e controller node is responsible for collecting
and processing sensor readings from its regulator nodes. It
actually serves as a gateway that forwards data to the central
(base) station. If the distance between the controller and the
base station is long, then the data will be forwarded using
multihop communication protocol. is data is stored in a
database and can be analyzed and processed on demand.
Users can access the data using smart phones or desktops via
the Internet.
Sensor regulators perform measurement and report local
temperature readings to the controller. ey are attached to a
device and regulate its action (e.g., increase/decrease heating)
in order to reach a certain temperature. Typical devices and
their corresponding actions are the following:
(i) heating bodies, such as central heating radiators,
electric radiators, or fans. e regulator can increase/
decrease the heating by using the valve controller,
(ii) air conditioners, which can be based on a fan. e
regulator can increase/decrease the cooling volume in
order to reach a certain temperature,
(iii) air ow, such as central air ow ventilation. ese
devices can regulate the degree of air circulation,
(iv) window shutters, such as outside curtains. By ris-
ing/lowering the curtains the inuence of the sun
energy can be regulated.
e temperature controller nodes are considered more
powerful than the sensor regulators. ey are expected to
have more advanced capabilities: memory, processing unit,
and steady energy supply. ey should be located higher up
in the room and away from all the heaters and windows in
a location that better represents the room’s temperature in
order to measure the most relevant temperature. Additionally
they are used to control the temperature at the premises
where the device is placed by generating instructions for
regulation. Temperature controllers represent the sink ele-
ment where the information is gathered and locally analysed.
Two dierent types of information are considered: static
and dynamic information. Static data is related to one-time
information, such as location of the nodes or type of the
nodes. Dynamic data is related to time variable parameters
such as temperature and energy cost.
International Journal of Distributed Sensor Networks
Room
Room Room
Controller
Regulator
Database Base station
Internet
F : System architecture.
Deployment diagram of the FIRT framework is given in
Figure . As illustrated, the temperature controller consists of
the following elements: local database, rules (which can be
basedonontologies[]), and sensors. Client should be able
to remotely congure the rules so the controller can meet
the temperature goals. e rules can include, for example,
temperature levels for several time intervals during dierent
days. e local database at the controller can be used in order
to store information regarding the temperature readings, as
well as the static information such as node type and node
location. e database can further contain information on the
size of the room being monitored, the number of expected
visitors, and other factors that can inuence the temperature
changes. All information that cannot be obtained from the
nodes, such as electricity cost, can be obtained from the
server. Furthermore, the server contains a database that can
be used by the controller to store historical data.
2.1. Temperature Optimization Framework. Finally a temper-
ature controller needs to make decisions by generating a set
of instructions for the dierent nodes. Hence, a regulation
method is required by the controllers in order to achieve
International Journal of Distributed Sensor Networks
Device
Server
Temperature controller
Sensor-regulator
Temperature
sensor
Applications:
localization, prediction
reporting, and instruction set
Device:
regulator data
regulator action
ZigBee radio
stack
ZigBee radio
stack
Applications:
localization,
aggregation, and
temperature
Local
database
Rules
ontology
Server
database
Web
service
Temperature
sensor
Data
Instructions
Client
Web
browser
Conguration
Data
prediction,
optimization
F : Deployment diagram of the system.
distributed decision regarding the temperature dierence
thatneedstobeachievedinacertainpointinordertosatisfy
the required overall temperature.
In order to provide a self-organized and cost-eective
solution, the WSN must provide the following function-
alities: self-localization, nodes clustering, data prediction,
distributed decision making, quality of service (QoS), and
so forth [,]. Sensor readings are useless if the location
where they are measured is not known or if the location
is wrong. us, a suitable localization algorithm should be
implemented in order to discover the location of the nodes.
is is important because manual recording of the nodes
positions is a very time-consuming solution and prone to
errors. e location information will be used by the controller
device to deduct the temperature set point that each of
the heater shall be commanded to, so that the temperature
dissipation from all heaters gives the wanted temperature at
the controller’s location and the controlled premises overall
(elaborated in Section ). Aer nodes discovering phase,
nodes can be divided into clusters. ere are many algorithms
purposed in the literature for optimal nodes clustering. In our
scenario, we assume that nodes deployed in close proximity
to each other belong to the same cluster.
Aer these two phases (discovering and forming the
clusters), we consider that a WSN is established. Nodes can
start measuring the temperature and forward the measured
readings to the nal destination (sink node). In order
to save energy, algorithms for data prediction should be
implemented on both sides: node and sink (elaborated in
Section ).
e temperature measurements are analysed by gener-
ation of temperature gradients. e temperature gradient
indicates the direction and the rate at which the temperature
changes within a particular location. e dynamic informa-
tion such as temperature gradients, time of the day, expected
visitors, and electricity cost and the static information such
as node type and node location represent an input to the
temperature optimization method.
e location of the wireless nodes together with the
temperature prediction methods allows the calculation of the
temperature gradient. Several methods can be used in order
to rank or assign weights to the dierent types of input such
as multiple attribute decision-making algorithms, genetic
algorithms, analytic hierarchy process, and fuzzy logic. Fuzzy
logic is suitable as fuzzy judgement matrices used for the
comparative analysis are close to the way the humans reect
and are very easy to implement. e fuzzy inference system
can be based on the standard Mamdani or Sugeno methods.
For each parameter that needs to be taken into consideration
membership function needs to be dened. Moreover, the
rules that need to be applied need to be carefully chosen. e
actual denition of the membership functions and the rules
will be considered in our future work.
Aer the collected data is analysed, the controller sends
instructions to the nodes (regulators). In order to determine
the appropriate actions the system should perform (open
window, increase radiator temperature, etc.), a mathematical
model is needed to calculate the heat transfer rate in a
certain object. Choosing the most suitable model depends
on the building where the WSN is deployed; that is, old
buildings and buildings constructed following the latest
recommendations and directives would have dierent models
due to dierent type of isolation. e following parameters
should be considered: heat resistance which depends on the
International Journal of Distributed Sensor Networks
thickness of the material, heat transmission coecient which
depends on the type of the material, and the temperature
dierence on the opposite sides of the walls, as well as the
area of the walls.
3. ZigBee Overview and WSN-FIRT
Network Topology
ZigBee is a low-cost, low power, wireless mesh network
standard []. e low cost allows the technology to be widely
deployed in wireless control and monitoring applications (in
elds such as home automation, health care, and temperature
control). Low power usage allows longer life with smaller
batteries. Mesh networking provides high reliability and more
extensive range. ZigBee operates in the industrial, scientic,
and medical (ISM) radio bands (e.g. MHz in Europe,
MHz in the USA and Australia, . GHz in most jurisdiction
sworldwide). Data transmission rates vary from 20kbps in
the MHz frequency band to 250kbps in the . GHz
frequency band.
ZigBee builds upon the physical layer and medium access
control dened in the IEEE standard .. ( ver-
sion) for low-rate wireless personal area networks (WPANs).
e specication completes the standard by adding four
main components: network layer, application layer, ZigBee
Device Objects (ZDOs), and manufacturer-dened applica-
tion objects which allow for customization and favor total
integration.
Besides adding two high-level network layers to the
underlying structure, the most signicant improvement is the
introduction of ZDOs. ese are responsible for a number of
tasks, which include keeping of device roles, management of
requests to join a network, and device discovery and security.
Due to the fact that ZigBee nodes can go from sleep to
active mode in ms or less, the latency is low and devices
canberesponsive,particularlycomparedtoBluetoothwake-
up delays, which are typically around three seconds. Since
ZigBee nodes can sleep most of the time, average power
consumption can be low, resulting in long battery life.
e ZigBee network layer natively supports both star
and tree network topologies and generic mesh networks.
Every network must have one coordinator device, tasked
with its creation of the control of its parameters and basic
maintenance. Within star networks, the coordinator must be
the central node. Both trees and meshes allow the use of
ZigBee routers to extend communication at the network level.
e routers need to be constantly active, listening for network
trac. erefore, it is normally assumed that the routers
aremainspoweredsensors/devices.ebatterydevicesare
assumedtobesleepingandonlywakingupandpollingfor
data periodically or on demand (upon user interaction).
e network topology should adapt to the characteristics
of the controlled premises. For the proposed WSN-FITR
we consider two dierent topologies: star- and cluster-based
topologies. An example of a star topology is illustrated in
Figure whereoneclassroomisillustrated.Inthiscase
there is a single temperature controller that is responsible
for controlling the temperature at the dierent nodes. e
star topology is most appropriate for small areas where
there are no major obstacles so that the signal from the
nodes does not fade in high extent. For large premises,
cluster-based topology is preferred over mesh as in the latter
higher energy is required at the nodes due to the fact that
each node transmits its own readings and the readings of
other nodes. Figure illustrates the proposed WSN-FITR for
one oor in a commercial center where there are several
temperature clusters that are controlling a set of nodes. e
decision regarding the temperature regulation is distributed
among all controllers. ere is one main controller that has
wired connection to the server in order to retrieve/store
information towards the database at the server.
4. Localization and Clustering in Indoor WSN
Many algorithms have been proposed for ZigBee-based
WSN localization. Most of them consider a WSN deployed
in outdoor environment where GPS signals are available
and a global map of the network can be easily achieved
using well-known techniques for localization [,]. On
the other side, indoor localization methods should consider
dierent characteristics of the indoor surroundings where
WSN is installed. Finding position of indoor WSN is more
challenging since GPS signal is heavily attenuated by building
structures such as walls and roofs and there is absence of line
of sight to some satellites []. With only few exceptions, the
distances between the nodes of the network are necessary
to be known for accurate location prediction. Dierent
techniquesareusedtoobtainthedistances:
(i) RSSI (received signal strength indicator),
(ii) ToA (time of arrival),
(iii) AoA (angle of arrival),
(iv) TDoA (time dierence of arrival).
e techniques based on RSSI are easier to implement and
do not require additional hardware, as all standard wireless
devices possess features for measuring this value. But nding
the relationship between the signal strength attenuation and
the transmission distance in indoor environments is not a
trivial task [,]. In the indoor environment, the moving
objects inside the building can cause reection, diraction, or
absorption of the radio signals. us the algorithms are more
prone to errors due to multipath phenomenon. Additionally,
many other characteristics of indoor environments have to
be considered like temperature and humidity variations,
orientation of antenna, furniture rearrangements, presence of
humanbeings,andsoforth.
Indoor localization methods can be divided into two
main categories []:
(i) deductive (mathematical) methods, that take into
account only the physical properties of signal prop-
agation. ey require the positions of the access
points, radio propagation model and map of the
environment,
(ii) inductive (ngerprinting) methods, that require a
previoustrainingphasewherethesystemlearnsthe
International Journal of Distributed Sensor Networks
Up
Up
Air ow Air ow
Cooling fan
Heating device
Shutter
Controller
Heating device
Shutter
Heating device
Shutter
Heating device
Shutter
F : Star architecture for a typical classroom.
Car parking
and wash
Food court
Up
Oce
OceOceOceOce
Controller
Controller
Main
Controller
Controller
controller
Controller
Controller
Controller
Controller
Controller
Controller
Oce
Up
Oce Oce Oce
Oce Oce Oce
Oce Oce Oce
Oce
Oce
Oce
Oce Oce Oce Oce Oce Oce
Oce
Oce
Oce
Oce
Oce
Oce
Oce
Oce
Oce
Oce
Oce
Oce
6m2
2m22m22m22m2
8m2
7m29m2
9m2
1m21m21m21m21m21m21m2
1m2
1m2
1m2
1m2
1m2
1m2
1m2
1m21m21m2
1m21m21m2
5m2
1m21m22m2
2m2
4m2
4m2
F : Clustered architecture for one oor in a shopping center.
RSS in each location. is phase can be very time-
consuming. In the next (positioning) phase, dierent
matching algorithm can be used in order to nd the
unknown location.
In [] the authors present an algorithm that com-
bines the advantages of both deductive and inductive meth-
ods. is hybrid method reduces the training phase with-
out a loss of precision. In []severalmatchingnger-
printing algorithms are investigated: the nearest neighbor
(NN) algorithm, the K-weighted nearest neighbor (KWNN)
algorithm, and the probabilistic approach based on the
kernel method. rough simulations it has been shown
that KWNN algorithm has the best indoor positioning
result.
Since localization is very crucial in our WSN-FITR, the
algorithm should be selected very carefully in accordance
with the characteristic of the environment where the network
should be deployed. If there are many walls and obstacles in
the environment, the deductive methods should be avoided
because they estimate the position mathematically. When
there are multiple access points and few walls in the envi-
ronment, inductive methods are not necessary as the training
phase can be very expensive.
International Journal of Distributed Sensor Networks
With the expansion of pervasive computing, most of
the service-based applications are expected to be contextual
awareness and location dependent. Still, indoor localization
is pioneering and nding appropriate methods will be a
challenge in the next years.
Aer determining the location, sensor nodes in WSN can
be geographically grouped into clusters. In each cluster one
representative node (cluster head) is chosen to coordinate
member nodes. e main advantages of WSN clustering are
not only to prolong the WSN lifetime but also to establish
collaboration between cluster members in order to provide
data aggregation and more accurate reports about the region
they sense. Many algorithms have been proposed in the
literature for WSN clustering [,].
5. Reductions of Data Transmissions
By reporting data measurement at each interval, WSN nodes
consume a great deal of energy, which reduces network life-
time and creates sucient communication overhead. Several
techniques have been developed to overcome these problems,
that is, to lower the communication overhead and to increase
energy saving. Most of them consider reducing the number
of radio transmissions.
ree main paradigms can achieve reduction of radio
transmissions:
(i) data compression, where well-known compression
techniques are used to compress consecutive mea-
surements.isapproachisusefulonlyiftheWSN
application does not require the data in real time and
canbeachievedregardlessofthenetworktopology,
(ii) in-network processing, where data are processed
on their way to the sink. is method is usu-
ally performed when summarization functions or
other queries are needed. It is appropriate only for
mesh-based, cluster-based, or hybrid-based network
topologies but cannot be implemented for star-based
network topology,
(iii) data prediction, where dierent prediction methods
areusedforpredictingthenextsensorreadings.Here,
each node runs a lter (or a model) that estimates the
next sensor reading. e sink runs exactly the same
models for each sensor in the network and makes
the same predictions. is approach is known as dual
prediction scheme (DPS).
For the WSN-FITR system we need the sensor mea-
surements up-to-date; hence data compression is not an
appropriate solution. Regarding network topology, we can
choose among data prediction and innetwork processing. If
the network topology of WSN-FITR is star based, we should
apply data prediction methods. For dierent topologies we
should consider innetwork processing or combination of
both.
In order to compare these techniques for data reduction,
dierent algorithms were implemented in MATLAB. For
the evaluation, a set of experimental data from Intel Lab
[] was used. e MicaDot sensors deployed in the
laboratory were equipped with weather boards and measured
temperature once every seconds. e measurements were
collected between February and April , . We run the
simulations for dierent error margins 𝐸max (ranging from
.∘Cto
∘C).
5.1. Data Prediction. e most appropriate models (lters)
for DPS are based on time-series forecasting: moving average
(MA), autoregressive (AR) model [], autoregressive mov-
ing average (ARMA) [], least mean square (LMS) [], and
LMS with variable step size (LMS-VSS) []. We implemented
and evaluated LMS, LMS-VSS, second-, fourth-, and tenth-
order MA and ARMA. Figure shows the reduction gain
foreachofthesealgorithmssimulatedontwonodesform
the Intel Berkeley Research Lab network []. e metric
is the reduction of transmissions in percentage (Figure ,
upper) and the dierence between the predicted and the true
value (Figure , lower), that is, mean square error (MSE).
e horizontal axis represents the value of the threshold of
𝐸max.
In Figures and , the results of the algorithms are repre-
sented by dierent colors: blue for MA(), yellow for MA(),
redforMA(),andgreenforARMA.eresultsshow
that, concerning the reduction of transmissions, the ARMA
algorithm is constantly better than the other two (Figure ).
MA() is second in this regard, and MA() is slightly behind.
In most of the results from the simulations we have done, the
dierence is greater for threshold values in the [, .] range.
However, there is very little dierence for values greater than
..
Another conclusion, concerning the performance of the
MA algorithm that can be drawn from the results, is that both
MSE and percentage of sent data are positively correlated
with the order of the MA algorithm. is is more clearly
visible on Figure ,whereonlythethreeMAalgorithms
(MA(),MA(),andMA())arecompared.Although
ARMA always performs better than the other techniques, this
algorithm uses consecutive readings in order to generate
the prediction model, while the others (MA, LMS, and LMS-
VSS) operate on real time.
5.2. In-Network Processing in WSN-FITR. If the WSN is
deployed on vast region, the radio signals would be far from
the sink node and multihop routing is needed for data to
reach its destination.
In order to calculate the reduction in this case, we divided
Intel network []intoclusters(Figure ). Four algorithms
were used for evaluation, MA(), MA(), LMS, []and
LMS-VSS []. e clustering parameter was geographic
position, that is, the Euclidian distance. We assume that
each sensor sends its reading to the cluster head, which is
responsible for resending the reading to the sink. As a result,
each reading is sent twice, except for the readings taken at the
cluster head.
Figure shows the reduction for cluster containing
nodes:,,,,,and.Forerrormarginof.
∘C,
LMS-VSS shows an average gain of .% compared to LMS
International Journal of Distributed Sensor Networks
0 1 2 3 45
0
20
40
60
80
100
0 1 2 3 45
0
5
10
15
MSE
LMS
LMS-VSS
MA(2)
ARMA
Emax
Emax
Data sent (%)
(a)
0 1 2 3 45
0
20
40
60
80
100
0 1 2 3 45
0
2
4
6
8
10
MSE
Emax
Emax
LMS
LMS-VSS
MA(2)
ARMA
Data sent (%)
(b)
F : Data reduction for dierent algorithms for node (a) and node (b).
0 1 2 3 4 5
0
20
40
60
80
100
0 1 2 3 4 5
0
5
10
15
MSE
ARMA
MA(2)
MA(4)
Emax
Emax
Data sent (%)
F : Results from simulations performed on dierentnodes.
algorithm, MA() is better than LMS-VSS by %, while
MA()outperformsMA()byapproximately%.Additional
reduction can be achieved if cluster head performs summa-
rization function (Average, Minimum, Maximum, etc.) and
forwards only the calculated aggregate to the sink. In this
case, the data reduction is far greater (% reduction of the
total messages sent for the given error margin of .∘C).
From the results, it can be concluded that, for these
datasets, the MA() always performs better compared with
other prediction techniques.
01 2 3 4 5
0
20
40
60
80
100
0 1 2 3 4 5
0
5
10
15
MSE
ARMA
MA(2)
MA(4)
MA(10)
Emax
Emax
Data sent (%)
F : Results from simulations performed on dierent nodes.
6. Conclusion
InthispaperwehaveproposedaWSNframeworkforindoor
temperature regulation. We provided an overview of systems
architecture and in detail elaborated designing guidelines
for ZigBee-based network. In order to reduce the energy
consumption,weproposedadatareductionstrategybased
on dual prediction scheme that uses dierent methods for
time-series forecasting. rough simulations on real world
dataset we showed that these lters can achieve reduction of
data transmissions of more than %.
International Journal of Distributed Sensor Networks
50
51
52 53
54
8
49
48
47
46
45
44
43
42 41
40
7
9
56
4
10
2
3
1
33
34
35
36
37
31
32 30
29
28
23
27
26
25 24
22
21 20
19
18
17
16
15
14
13
11
12
39
38
F : A clustered view of the Intel Lab [].
00.5 1 1.5 2
0
10
20
30
40
50
60
70
80
90
100
Data sent (%)
LMS
LMS-VSS
MA(2)
MA(4)
Emax
F : Data reduction in the cluster containing the nodes: , ,
, , , and .
For future work, we intend to investigate in the
above-mentioned data prediction techniques using dier-
ent datasets with dierent measured parameters, like light
intensity or humidity, since they can aect the temperature
parameter. We also plan to explore machine learning tech-
niques for choosing the best technique. Furthermore, for
distributed decision making, fuzzy logic is very suitable to
be implemented in constrained WSNs, since, for temperature
regulation, there are only a few important parameters that
holdmostoftheinformationandmakerulepruningpossible.
Conflict of Interests
e authors declare that there is no conict of interests
regarding the publication of this paper.
References
[] R. Mittal and M. P. S. Bhatia, “Wireless sensor networks for
monitoring the environmental activities,” in Proceedings of the
IEEE International Conference on Computational Intelligence
and Computing Research (ICCIC ’10), pp. –, December .
[] “ZigBee Alliance,” http://w ww.zigbee.org/.
[] C.Zhang,M.Zhang,Y.Su,andW.Wang,“Smarthomedesign
based on ZigBee wireless sensor network,” in Proceedings of
the 7th International ICST Conference on Communications and
Networking in China (CHINACOM ’12), pp. –, August
.
[] A. A. Khan and H. T. Mouah, “Energy optimization and
energy management of home via web services in smart grid,” in
Proceedings of the IEEE Electrical Power and Energy Conference
(EPEC ’12), pp. –, IEEE, Ontario, Canada, October .
[] V. Boonsawat, J. Ekchamanonta, K. Bumrungkhet, and S.
Kittipiyakul, “Xbee wireless sensor networks for temperature
monitoring,” in Proceedings of the 2nd ECTI-Conference on
Application Research and Development (ECTI-CARD ’10),Pat-
taya, Chonburi, ailand, May .
[] “Arduino Platforms,” http://w ww.arduino.cc/.
[] “Xbee Wireless RF Modules,” http://www.digi.com/xbee/.
[]C.C.Castello,R.X.Chen,J.Fan,andA.Davari,“Context
aware wireless sensor networks for smart home monitoring ,”
e International Journal of Autonomous and Adaptive Commu-
nications Systems,vol.,pp.–,.
[] M. Demirbas, K. Y. Chow, and C. S. Wan, “INSIGHT: internet-
sensor integration for habitat monitoring,” in Proceedings of
the Advanced Experimental Activities on Wireless Networks and
Systems (EXPONWIRELESS) Workshop (as part of WOWMOM
2006), pp. –, Bualo, NY, USA, June .
[] B. Stojkoska and D. Davcev, “Web interface for habitat moni-
toring using wireless sensor network,” in Proceedings of the 5th
International Conference on Wireless and Mobile Communica-
tions (ICWMC ’09), pp. –, IEEE, August .
[] A. Caracas, F. Mller, O. Sundstrm, C. Binding, and B.
Jansen, “VillaSmart: wireless sensors for system identication
in domestic buildings,” in Proceedings of the IEEE Innovative
Smart Grid Technologies Europe (ISGT Europe’13),Copenhagen,
Denmark, October .
International Journal of Distributed Sensor Networks
[] A. Rossello-Busquet, J. Soler, and L. Dittmann, “A novel home
energy management system architecture,” in Proceedings of
the UKSim 13th International Conference on Modelling and
Simulation (UKSim ’11), pp. –, April .
[] B. Yahya and J. Ben-Othman, “Energy ecient and QoS
aware medium access control for wireless sensor networks,”
Concurrency Computation Practice and Experience,vol.,no.
, pp. –, .
[] J. Ben-Othman, K. Bessaoud, A. Bui, and L. Pilard, “Self-
stabilizing algorithm for ecient topology control in Wireless
Sensor Networks,” Journal of Computational Science,vol.,no.
, pp. –, .
[] B. Stojkoska, I. Ivanoska, and D. Davcev, “Wireless sensor net-
works localization methods: multidimensional scaling versus
semidenite programming approach,” in ICT Innovations 2009,
pp. –, Springer Berlin, Heidelberg, Germany, .
[] B. R. Stojkoska and D. Davcev, “MDS-based Algorithm for
Nodes Localization in D Surface Sensor Networks,” in Proceed-
ings of the 7th International Conference on Sensor Technologies
and Applications (SENSORCOMM ’13),pp.–,.
[] L.Cheng,C.Wu,Y.Zhang,H.Wu,M.Li,andC.Maple,“A
survey of localization in wireless sensor network,” International
Journal of Distributed Sensor Networks,vol.,ArticleID
, pages, .
[] A. Awad, T. Frunzke, and F. Dressler, “Adaptive distance
estimation and localization in WSN using RSSI measures,” in
Proceedings of the 10th Euromicro Conference on Digital System
Design Architectures, Methods and Tools (DSD ’07), pp. –,
August .
[] A. Faheem, R. Virrankoski, and M. Elmusrati, “Improving
RSSI based distance estimation for .. wireless sensor
networks,” in Proceedings of the IEEE International Conference
on Wireless Information Technology and Systems (ICWITS ’10),
August .
[]J.Lloret,J.Tomas,M.Garcia,andA.Canovas,“Ahybrid
stochastic approach for self-location of wireless sensors in
indoor environments,” Sensors,vol.,no.,pp.–,
.
[] S.-S. Jan, L.-T. Hsu, and W.-M. Tsai, “Development of an indoor
location based service test bed and geographic information
system with a wireless sensor network,” Sensors,vol.,no.,
pp. –, .
[] S. Zhang, J. Fan, J. Jia, and J. Wang, “An ecient cluster-
ing algorithm in wireless sensor networks using cooperative
communication,” International Journal of Distributed Sensor
Networks,vol.,ArticleID,pages,.
[] A. A. Abbasi and M. Younis, “A survey on clustering algorithms
for wireless sensor networks,” Computer Communications,vol.
, no. -, pp. –, .
[] “Intel lab data.Web Page,” http://db.lcs.mit.edu/labdata/labdata
.html.
[] Y.-A.leBorgne,S.Santini,andG.Bontempi,“Adaptivemodel
selection for time series prediction in wireless sensor networks,”
Signal Processing,vol.,no.,pp.–,.
[] J. Lu, F. Valois, M. Dohler, and M.-Y. Wu, “Optimized data
aggregation in WSNs using adaptive ARMA,” in Proceedings
of the 4th International Conference on Sensor Technologies and
Applications (SENSORCOMM ’10),pp.–,Venice,Italy,
July .
[] S. Santini and K. Romer, “An adaptive strategy for quality-based
data reduction in wireless sensor networks,” in Proceedings of
the 3rd International Conference on Networked Sensing Systems
(INSS ’06), pp. –, Chicago, IL, USA, June .
[] B. Stojkoska, D. Solev, and D. Davcev, “Data prediction in
WSN using variable step size LMS algorithm,” in Proceedings
of the 5th International Conference on Sensor Technologies and
Applications,Nice,France,.
Available via license: CC BY
Content may be subject to copyright.
Content uploaded by Biljana Risteska Stojkoska
Author content
All content in this area was uploaded by Biljana Risteska Stojkoska on Feb 04, 2016
Content may be subject to copyright.
