Wireless Sensor Network, 2010, 2, 703-709
doi:10.4236/wsn.2010.29085 Published Online September 2010 (http://www.SciRP.org/journal/wsn)
Copyright © 2010 SciRes. WSN
Design of Building Monitoring Systems Based on Wireless
Sensor Networks*
Qifen Dong, Li Yu, Huanjia Lu, Zhen Hong, Yourong Chen
College of Information Engineerin g, Zhejiang University of Technology, Hangzhou, Zhejiang, Ch ina
E-mail: dongqifen0419@163.com, lyu@zjut.edu.cn, {hongzhen614, luhuanjia1020}@126.com, jack_chenry@163.com
Received June 9, 2010; revised July 11, 2010; accepted August 19, 2010
Wireless Sensor Network provides a potential technique for monitoring the indoor environment. This paper
presents a Building Monitoring system based on Wireless Sensor Networks. A clustering-based network
specified for building monitoring is proposed, which is inspired by LEACH (Low Energy Adaptive Cluster
Hierarchy) method. Further, two key ideas are used to implement the clustering-based network. First, the
configuration module of building management software is used to conduct all nodes in a room forming a
local cluster. This cluster formation method does not consume node energy. Second, because cluster-heads
cannot directly transmit packets to th e sin k node du e to limited wireless co mmunication range, the clu st er - h e a d
communications are represented by a multi-hop tree rooted at the sink node. The experiment has been made
to demonstrate the feasibility of the proposed results.
Keywords: Wireless Sensor Networks, Building Monitoring, LEACH, Cluster
1. Introduction
According to the 2007 UN Environment Program [1],
about 40 percent of total energy is used for heating,
cooling, lightin g and ventilation of buildings. Subs tantial
savings can be made by applying these functions only
when and where they are needed. Such control is only
possible when indoor conditions such as temperature,
relative humidity or light are measured. Wired sensors
could certainly be installed at sensing locations, but such
a step requires significant effort and an additional set of
wires throughout a building. Estimates of the cost to de-
ploy wires ranged from $2.2 per meter for new buildings
to $7.19 per meter for existing constructions in 2002 [2].
In addition, running wires in existing constructions in-
duces other problems, such as destroying appearance of
buildings. Transmitting sensing data wirelessly provides
a significant benefit for monitoring indoor environment.
However, traditional wireless systems suffer from their
own disadvantages, such as high running cost, and high
energy consumption of monitoring devices.
Wireless Sensor Network (WSN) [3] which co nsists of
dense sensor nodes that continuously observe physical
phenomenon provides an opportunity for building moni-
toring [4,5]. Thomas Schmid [6] reported their experienc e
with the implementation, deployment and operation of
SensorScope, an indoor environmental monitoring net-
work based on WSNs. Literature [7] demonstrated an
industrial-strength wireless sensor network application
for indoor environment monitoring. This application is
integrated WSNs with a Building Management System.
Won-Suk Jang showed how advanced WSN technologies
can be used to monitor conditions in and around build-
ings [8]. A WSN system was deployed in a number of
residential and commercial buildings in [9]. W.S. Jang
and W. M. Healy investigated WSN performance metric
for building monitoring applications [10]. It appears that
WSN provides huge potential for building monitoring.
However, it still needs getting more attention continu-
The main contributions of this paper are as follows: (1)
a Building Monitoring system based on WSNs (BMW
SNs) is presented; (2) inspired by LEACH (Low Energy
Adaptive Cluster Hierarchy) method [11,12], a cluster-
ing-based network specified for building monitoring is
proposed; (3) the configuration module of building
management software is used to conduct all nodes in a
room forming a local cluster. This cluster formation
method does not consume node energy; (4) because
cluster-heads cannot d irectly transmit packets to the sink
node due to limited wireless communication range, the
*Supported by the Key Project of Zhejiang Provincial Ministry of
Education under Grant NO.ZD2007003, and the Zhejiang Provincial
ural Science Foundation of China under Grant NO.Y1080163.
Copyright © 2010 SciRes. WSN
cluster-head communications are represented by a multi-
hop tree rooted at the sink node.
The remainder of the paper is organized as follows.
Section 2 describes the overview of BMWSNs. The im-
plementation of the clustering-based network is given in
Section 3. Section 4 shows experimental results. Conclu-
sion is given in Section 5.
2. System Overview
Figure 1 shows the architecture of BMWSNs. Several
sensor nodes are placed in each room. Because the data,
such as temperature, humidity and light intensity, sensed
by nodes in the same room are highly correlated, a node
termed cluster-head is installed in each room. The clus-
ter-head receives data from other nodes in the same room,
performs data aggregation, and forwards data to a sink
node which is connected to a computer throu gh a RS- 232
connection. The building supervisor can get the indoor
environment with the help of building management
software installed on the computer. So it is possible to
control the electro-devices.
Figure 2 shows hardware images. The sensor node
powered by 2 AA batteries consists of a main board and
a sensor board. The main board contains a microproces-
sor (ATMEGAL128L) and a radio (CC2420). The sensor
board contains temperature/ humidity sensor (SHT11),
light sensor (STL2550), and human detection sensor
(BISS0001), and is plugged into the main board via a
41-pin connector. The sink node is composed of the main
board and a bottom board which integrated UART con-
nector, power outlet, and program interface. At the
computer, the building management software is written
in Qt, using Sqlite as the database.
The BMWSNs has the following features from the
view of end-user.
1) Displaying cu rve: it displays curves of temperature,
humidity, light intensity, and human detection of the
selected room.
2) Configuring node’s attributes: the node’s attributes
can be configured by plugging the node into the bottom
board. These attributes include long address, room num-
ber, normal/abnormal measurement period for each sen-
sor, as well as alarm threshold for each sensor. This con-
figuration module is useful in cluster formation.
3) Tipping alarm information: the alarm information is
flipped automatically when the measurement exceeds the
corresponding alarm threshold..
Figure 1. The architecture of WBMSNs.
CC2420 Atmel 128
41-Pin Connector
2-AA Batteries
STL 2550
SHT 11
Human Detection
41-Pin Connector
UART Connector
Power Outlet
Program Interface
41-Pin Connector
(a) main board (b) sensor board (c) bottom board
Figure 2. The sensor node consists of (a) and (b); the sink node is compose d of (a) and (c ).
Copyright © 2010 SciRes. WSN
4) Storing data: the historical data is stored in the data-
base. It can provide a reference for investigating power
saving scheme.
3. Implementation of the Clustering-Based
The periodic listen and sleep mechanism in S-MAC [13]
protocol is adopted to save node energy. This section
focuses on how sensor nodes process their original sens-
ing signals and transmit data to the sink node. This is a
hot spot and difficult problem in the study of WSNs. It
relates to data aggregatio n, routing, topology control and
so on. Moreover, it should be oriented to the specified
As mentioned previously, there is strong correlation
among the data from nodes in the same room, so we con-
structed a clustering-based network inspired by LEACH
method. All nodes in a room form a local cluster, with
one node acting as the cluster-head. All non-cluster-head
nodes must transmit their data to the clu s ter-head.
In LEACH, it is assumed that all cluster-heads could
communicate with the sink node directly. However, the
wireless communication distance is limited, so this as-
sumption is impracticable. Thus it is necessary to design
a simple and feasible protocol for transmitting data from
a cluster-head to the sink node.
3.1. Cluster Formation
In LEACH, nodes elect themselves to be cluster-heads
with a certain probability. Due to space constraints, the
details of the cluster formation in LEACH are omitted,
interested readers should refer to [11,12]. However, this
algorithm is not suitable for building monitoring applica-
tion. There are two reasons for this. First, performing this
process is difficult and power-wasting in actual applica-
tion. Second, it is assumed that the data from all nodes is
highly correlated. In fact, the correlation only exists
among nodes in the same room. Therefore, it would be
better to cause nodes in the same room to form a local
cluster using a simple method. To do so, the configura-
tion module of the building management software is used.
The node’s attributes need to be configured before the
node is placed. Using the room number, the node’s short
address is calculated according to equation (1).
nodem m
FF RRn  (1)
andRdenote the floor and the room of this floor,
respectively. For example, if the configured room num-
ber is 324, then3, 24FR
and m
Rare the maxi-
mum number of nodes that are allowed in each floor and
in each room, respectively. The two values are deter-
mined by the designer. n denotes the sequence that the
node is deployed in a room.
The node which is first placed in a room is treated as
the cluster-head automatically. So the cluster-head’s
short address is known to the other nodes in the same
room. Let all nodes in a room form a local cluster. Ob-
viously, this cluster formation mechanism does not re-
quire node energy expenditure, and the relationship be-
tween the node and the room is also established. Besides,
it is easy to add or remove a node as long as the number
of deployed nodes in a room is less thanm
Each node starts to measure temperature, humidity,
light intensity, and human detection periodically after
deployment. The measurements are packed and sent to
the cluster-head every twelve hours. Then the data ag-
gregation is performed in the cluster-head by equation
min max
||||||||if for hum an detection sensor
other wise
vv vv
result vv v
where i
vis the sensor measurement value of node i for
a certain period,
is the total number of nodes in the
local cluster, andmin 1,
min i
max 1,
max i
. Excep-
tionally, for human detection sensor, i
vis equal to one if
node ihas detected the human otherwise i
vis equal to
zero. Performing data aggregation at the cluster-head
decreases much amount of data which needs to be trans-
mitted to the sink node. Because the energy used for
computation is much less than the energy for communi-
cation, this aggregation process reduces the overall sys-
tem energy expenditure.
A node speeds up sensor acquisition frequency once
one of the sensor measurements exceeds the correspond-
ing alarm threshold. A warning message is sent to the
cluster-head as soon as the measurement exceeds the
alarm threshold continuously for ten periods. The clus-
ter-head forwards this warning message directly and dis-
cards the data from other cluster members. So the elec-
tro-devices can be regulated as quickly as possible.
Remarkably, the data aggregation makes sense only
when all nodes in a room are synchronized in time. To
do so, the cluster-head broadcasts a synchronization
message periodically. The newly added node don not run
the sensors until receiving the synchronization message.
3.2. Cluster-Heads Transmission Protocol
As mentioned above, the wireless communication range
is limited, so cluster-heads need transmitting packets to
the sink node via multi-hop manner. Denote the sink
node as0
S, and the cluster-head communications are re-
presented through a multi-hop tree rooted at 0
S (see
Figure 3). There are two key sub-processes in imple-
menting the multi-hop tree. The two sub-processes are
Copyright © 2010 SciRes. WSN
Figure 3. An example of multi-hop tree. The red circle de-
notes the sink node, while the yellow ones denote clus-
tree initialization and tree maintenance. The goal of the
tree initialization is to initialize a tree with cluster-heads
and the sink node. The tree maintenance is aimed at up-
dating the tree structure to balance energy consumption
among cluster-heads.
3.2.1. Tree Initialization
All cluster-heads are separated originally. The sink node
forms a “network” and its depth in the network is zero.
The aim of the tree initializatio n is to exp and th e n etwork
to construct a tree rooted at 0
S. The initialization pro-
cedure is described as follows.
Step 1 A cluster-head that plans to join the network
broadcasts a join-network request message. This clus-
ter-head is termed CH. Meanwhile, a timer with fixed
time period is scheduled. Then CH waits for response
message during this time. The waiting time period should
be long enough so that it allows multiple neighbors to
response to the request.
Step 2 Neighbors of CH would receive the join-net-
work request message. The neighbor sends out a join-
network response message containing its depth in the
network if it has joined the network, otherwise it discards
the request.
Step 3 CH chooses the neighbor whose response mes-
sage has the strongest received signal strength (RSS) as
its candidate parent. The RSS is indicated by the re-
ceived signal streng th indicator in cc2420 ch ip. Then CH
unicasts an association request message to the selected
candidate parent.
Step 4 Upon reception of the association request mes-
sage, the candidate parent responses it and adds CH to its
children table.
Step 5 Receiving an association response message in-
dicates that CH has joined the network successfully. CH
stores information of the parent node and its depth in the
network. CH’s depth in the network is one more than the
depth of parent node.
If CH cannot receive any join-network response mes-
sage when the timer expires or it fails to request associa-
tion, it tries to join the network again after a random de-
Using this method, neighbors of the sink node join the
network firstly. Then do the two-hop cluster-heads of the
sink node. A tree rooted at0
Sis formed gradually.
3.2.2. Tree Maintenance
Each cluster-head can send packets to its parent after the
tree is formed. However, the energy consumption among
these cluster-heads is very different even they have the
same depth in the tree, because each of them has a dif-
ferent number of children. Therefore, it is necessary to
perform tree maintenance to balance the energy con-
To do so, each cluster-head updates its parent periodi-
cally, and the update period is computed randomly to
avoid radio interference. The updating process is similar
to the tree initialization sub-process, and the difference
between them is that the join-network response message
from nodej contains more parameters about nodejin
the updating process. These parameters include the chil-
dren number
c, the residual energy
E, and the depth
din the tree. The combinations of them are considered
in Step 3. Denote E as the initial energy of 2 AA bat-
teries. A cluster-head ichooses the neighbor n as its
parent with probabilityin
P, as calculated by equation (3):
(1 /)(1 /)(/),
(1 /)(1 /)(/)
in i
 
where i
pis the set of cluster-heads that response to the
join-network request from cluster-headi, and
12 3
are nonnegative weighting factors for each
4. Experiments
The BMWSNs is operated in buildings, and the material
quality of the buildings, environmental condition, and so
on can affect the distance and quality of the wireless
communication. Figure 4 shows the packet loss rate
versus the wireless communication distance between two
nodes which are separated by a wall. It indicates that the
packet loss rate is lower than 20%. The packet is re-
transmitted if it is lost, then the packet loss rate is less
than 4%. This demonstrates that deploying a cluster-head
in each room can guarantee the network connectivity if
all rooms are smaller than the size of 115.
The BMWSNs is tested with 21 nodes deployed ran-
domly in several rooms. The node’s attributes are con-
figured using the configuration module of the building
management software before the node is deployed. The
actual deployment is shown as Figure 5. The sink node
is deployed in room 323. Each room has a cluster-head
and several sensor nodes. The figure between brackets
Copyright © 2010 SciRes. WSN
01 2 3 4 5 6 7 8 910 11 12 13 14 15
packet loss rate
Figure 4. The packet loss rate VS. the distance.
denotes the node’s shor t address. In th e test, the maximal
number of nodes that are allowed in each floor and each
room are 100 and 10, respectively.
Figure 6 is a network topology graph as the BMWSNs
starts up. The dotted lines with arrow denote communi-
cation links. It demonstrates the feasibility of the pro-
posed clustering-based network. All nodes in a room
form a star network centered on the cluster-head and all
the cluster-heads build a multi-hop tree rooted at the sink
node. Along this tree, all packets can be delivered to the
sink node. For example, the route from nodes in room
322 to the sink node is as follows:
322 322 323NS CHCHSINK
where NSn and CHn denote the sensor nodes and the
cluster-head in room n, respectively.
Figure 7 shows the temperature curve of room 323. It
also can display curves of other parameters by clicking
on the icon at upper-right corner. This provides reference
for controlling electrical devices in real time.
We regulate the temperature of room 322 at 16C
(The upper and lower thresholds of temperature are
18 C
and 26C
, respectively), then the alarm informa-
tion is flipped automatically (see Figure 8).
Figure 9 shows historical temperature analysis chart.
It provides a basis for building energy-saving plan.
5. Conclusions
The system architecture of BMWSNs is presented and an
overview of system features is given. This paper focuses
on the two key methodologies which are used to imple-
ment the clustering-based network specified for building
monitoring. The experiment was made to demonstrate
the feasibility of the BMWSNs. However, there exist
shortages inevitably as a result of exploratory research
on applying WSNs to building monitoring. We will per-
form further tests to find problems, and then improve it
from both hardware and software. Future plans for the
BMWSNs include
Figure 5. The actual deployment.
Copyright © 2010 SciRes. WSN
1) Usually, electric devices are regulated only when
the data exceeds the alarm threshold. Therefore, nodes
could store normal data locally in order to save com-
munication energy consumption, and the user queries
data depending on the need. Thus nodes deployed in the
building construct a distributed storage and query data-
Figure 6. The network topology.
Figure 7. The temperature curve of room 323.
Figure 8. The alarm information.
Figure 9. The historical temperature analysis chart.
2) In current WBMSNs, a node is always the clus-
ter-head once it is configured as a cluster-head. This
leads to exhausting the energy of these cluster-heads
quickly. In order to solve this problem, it is necessary to
explore a mechanism for switching cluster-heads auto-
3) How to deploy nodes to optimize the network per-
formance is also the focus of further research.
5. References
[1] B. Robert, “Buildings Can Play a Key Role in Com- bat-
in g C l i m a te Change,” 2007.
[2] K. M. Michael and R. B. Michael, “Pros & Cons of Wire-
less,” American Society and Heating, Refrigerating, and
Air-Conditioning Engineers Journal, Vol. 44, No. 11,
November 2002, pp. 54 - 61.
[3] L. M. Sun, J. Z. Li, Y. Chen and H. S. Zhu, “Wireless
Sensors Networks,” Tsinghua University Press, Beijing,
[4] A. Willing, “Wireless Sensor Networks: Concept, Chal-
lenges and Approaches,” Elektrotechnik & Information-
stechnik, Vol. 123, No.6, June 2006, pp. 224-231.
[5] C. Adam and W. Jakub, “On Applications of Wireless
Sensor Networks,” Internet - Technical Development and
Applications, Springer, 2009, pp.91-99.
[6] S. Thomas, D. F. Henri and V. Martin, “Sensor Scope:
Experiences with a Wireless Building Monitoring Sensor
Network,” Proceedings of the Workshop on Real-World
Wireless Sensor Networks, Stockholm, 2005.
[7] J. Y. Chang, J. Y. Kim and O. Kwon, “Demo Abstract:
Control City – Integrating Wireless Sensor Networks and
Building Management Systems,” Proceedings of the 2009
International Conference on Information Processing in
Sensor Networks. San Francisco, February, 2009, pp.
Copyright © 2010 SciRes. WSN
[8] W. S. Jang, W. M. Healy and J. S. Miroslaw, “Wireless
Sensor Networks as Part of a Web-Based Building Envi-
ronmental Monitoring System,” Automation in Construc-
tion, Vol. 17, No. 6, August 2008, pp. 729-736.
[9] D. Tessa, G. Elena and B. James, “Wireless Sensor Net-
works to Enable the Passive House- Deployment Experi-
ences,” Smart Sensing and Context, Berlin Heidelberg:
Springer, 2009, pp. 177-192.
[10] W. S. Jang and W. M. Healy, “Wireless Sensor Network
Performance Metrics for Building Applications,” Energy
and Buildings, Vol. 42, No. 6, 2010, pp. 862-868.
[11] W. B. Heinzelman, A. P. Chandrakasan and H. Balak-
rishnan, “An Application-Specific Protocol Architecture
for Wireless Micro-sensor Networks,” IEEE Transactions
on Wireless Communications, Vol. 1, No. 4, 2002, pp.
[12] W. B. Heinzelman, A. P. Chandrakasan and H. Balak-
rishnan, “Energy-Efficient Communication Protocol for
Wireless Micro-sensor Networks,” Proceedings of the
33rd Hawaii International Conference on System Sci-
ences. Hawaii, 2000, pp. 1-10.
[13] W. Ye, J. Heidemann and D. Estrin, “An Ene rgy -Ef fici ent
MAC Protocol for Wireless Sensor Networks,” Proceed-
ings of the 21st International Joint Conference on IEEE
Computer and Communications Societies, New York,
June 2002, pp. 1567-1576.