Content uploaded by D.M.G. Preethichandra
Author content
All content in this area was uploaded by D.M.G. Preethichandra on Oct 20, 2015
Content may be subject to copyright.
Content uploaded by D.M.G. Preethichandra
Author content
All content in this area was uploaded by D.M.G. Preethichandra on Oct 20, 2015
Content may be subject to copyright.
Design of a Smart Indoor Air Quality Monitoring
Wireless Sensor Network for Assisted Living
D.M.G.Preethichandra
Central Queensland University
School of Engineering and Built Environment
Rockhampton, Australia
e-mail preethi@ieee.org
Abstract— Wireless indoor air quality monitoring is the main
objective of this research in order to provide real time
information for assisted living. The indoor air quality measured
in the built environment provides a continuous stream of
information for seamless controlling of building automation
systems, and provides a platform for informed decision making.
Further, this low power sensor network design provides vital air
quality information under emergency and hazardous conditions
even without grid power for a reasonable time. The proposed
system has carbon dioxide, carbon monoxide, propane and
methane sensors. This prototype network was first built using a
hardware platform available in the market with industrial grade
gas sensors. The concept was verified with actual parameter
measurements under different real life situations. The results
reveal that the domestic indoor air quality may be extremely
different compared to what is expected for a quality living
environment.
Keywords— Air quality, sensor networks, wireless sensor
networks, gas sensor, online monitoring
I. INTRODUCTION
Indoor air quality monitoring is an essential part of built
environment control systems to ensure the indoor environment
is in an acceptable condition for living. The process of indoor
air quality monitoring involves different technologies namely
actual gas sensing, sensor networking, data mining and
decision making. There are some prior work on this topic from
different points of approach[1]-[3], but this paper proposes an
indoor air quality monitoring network for assisted living.
Accurate measurement of gas concentrations is the key of air
quality monitoring. However, this has to be within the budget
of the monitoring system as well. Therefore it is common to
select only a number of target gases and sacrifice a little from
the accuracy to bring down the cost while getting data within
an acceptable resolution.
Online air quality monitoring can be used in many different
applications ranging from quality of life improvement to
military operations. Occupancy monitoring is a typical
application where the number of occupants is estimated from
the amount of carbon dioxide detected in a given environment.
This information may be used to effectively control the
heating, ventilation and air conditioning(HVAC) system in the
building or may be used to remotely spy on the population in
the room[2,4]. However the military usage is out of scope of
this research and this paper mainly focuses on use of online gas
concentration monitoring only for civil purposes. This paper
mainly focuses on monitoring air quality to improve the living
condition in built environments and gather information in
disaster management situations.
Carbon dioxide, carbon monoxide, methane and propane
are the most common gases of interest in built environments.
Carbon dioxide is added to the environment in moderate
quantities from respiration under normal situations and in
excessive quantities from fires under hazardous conditions.
Carbon monoxide is mainly added to the built environment
from automobiles. A common source of methane is the
common garbage collected at designated areas adjacent to
buildings. Propane is mainly from cooking and heating gas
lines. Any of these gases may be life threatening if gone above
the accepted range for a considerable period of time. Therefore
continuous monitoring of each of them is vital in indoor air
quality monitoring system.
The HVAC system in a building can be programmed to
respond to the indoor air condition monitored from the
proposed network. Further, this kind of a distributed wireless
sensor network will help the staff of aged and disabled care
centers to make decisions on when to move them around for
people living indoor for extended periods due to physical
disabilities. This involves with different methodologies for data
analysis and fit for most suitable model in accordance with the
requirement of facility concerned[5,6].
Not only the data from within the building, this kind of
network can be easily used to monitor environmenental gas
concentrations as well[7,9]. Industrial air quality measurement
is another aspect of these networks[10].
II. T
HICK FILMMETAL OXIDE GAS SENSORS
Semiconductor gas sensors are very common in industrial
gas monitoring systems where the gas concentration is
represented by an electrical parameter. A common type of gas
sensors is thick film semiconductor gas sensor where the
semiconductor substrate resistance is varied dependent on the
amount of target gas presence. In this research a series of
semiconductor thick film type gas sensors with good selectivity
has been used for the target gasses. The manufacturer,
Figaro(Japan) prescribes precise control signals to the sensor in
order to obtain accurate measurements. The sensor consists of
two major parts, namely the heater and the sensor substrate.
The substrate has two terminals and its resistance is measured
as a representation of the amount of gas in the environment
while the heater provides the stabilized temperature needed for
the measurement. Fig. 1 shows the recommended setting for
the measurement where the heater control signal and sensor
measurement signal are coming from the sensor node
microcontroller while the sensor resistance is read by the same
microcontroller.
The manufacturer’s specification provides the heater turn
on timing and sensor resistance read timing. For example, the
carbon monoxide sensor (TGS2442) has a 14ms heating pulse
in every second and a 5ms read-control signal immediately
after the heater pulse is off. The actual read should occur at the
middle of the read-control pulse, in other words 2.5ms after the
read-control pulse is on.
Fig. 1. Schematic diagram of thick film gas sensors and
control circuit.
If the current through R
S
is i
S
, heater resistance is R
H
, and
Sensor Resistance is R
S
, when the Q
2
is turned on by read
control pulse,
=
+
(
)
(1)
=
(
)
(2)
=
(
)
(
)
(3)
The value of sensor resistance can be calculated from the
equation 3 using V
out
value read from the sensor circuit. The
microcontroller looking after this sensor sends the heater and
read control signals at required intervals and the analog to
digital converter(ADC) of the microcontroller converts the
V
out
voltage at the appropriate time. Then the R
S
/R
0
is used in
conjunction with manufacturer’s datasheet to obtain the target
gas concentration where R
0
is the resistance of the sensor
substrate at a known gas concentration. Sensor resistance R
S
is
dependant on the ambient temperature and humidity. To have
an accurate measurement of gas concentration, the two
parameters have also to be measured. The chosen Libelium
Waspmote hardware platform has a built in temperature
monitor and the humidity sensor has to be integrated on to the
sensor module. Further, the sensor control module has a
software programmable real time clock which will be useful if
a particular node of the sensor is to be programmed to activate
at desired time intervals of the day. Moreover, when the
network become highly populated with large number of nodes
and the links become busier, then the time stamp from each
node will be very useful. Since this real time clock is software
programmable, the base station can synchronize all the nodes
to its clock easily.
III. S
ENSOR NODE DESIGN
A. Sensor node hardware
The intended sensor node design consists of several
modules controlled by a microcontroller. The main modules
are RF module, control module, sensor module and power
module. The RF communication was handled through ZigBee
protocol using a XBEE module from Digi International. RF
communications module with 2dB antenna which can
communicate over 200m line of site distance, which is more
than enough for indoor operations.
One of the main objectives of this research is to develop a
microcontroller based ultra low power sensor node to gather
information on indoor air quality. However, in the first phase, it
was decided to use sensor nodes available in the market as
proof of concept. The sensor control module prototype has
been designed with a Waspmote hardware module available
from Libelium Communications. He network operation is
explained in the next section.
The selected sensor node has a built in power control circuit
which recharges a battery either from USB or from a solar
panel of 12V. The control system also provides provisions for
eight analog ports which can be configured as inputs or eight
digital I/O pins. These were used to implement the control and
data busses to communicate with the sensor module. The
sensor module consists of carbon dioxide, carbon monoxide,
methane and propane gas sensors integrated into it.
B. Power Consumption
Power consumption is not a major issue as long as the built
environment is connected to the power grid where these
sensors can directly be fed from it. However, indoor gas
condition monitoring is vital in emergency evacuation
situations where the mains power may be turned off under high
danger situations or by the emergency itself. Each sensor node
has its own rechargeable battery which getting charged from
the mains power when available, and automatically switches to
battery power whenever the mains power is cut off. The chosen
Wasp mote sensor hardware can seamlessly transfer from
mains to battery, therefore no rebooting will occur during the
changeover. An ultra low-power sensor network node is vital in
such situations where it can sense and send information on
indoor air quality for a longer time without mains power for
disaster management team to make correct decisions.
However, the manufacturer recommends that the gas sensor
in the sensors module needs to be heated-up for more than two
hours before taking reliable readings. Therefore this module
has to be powered up continuously, but the average power
consumption by the sensor is very small as the heater pulse is a
50mA, 14ms pulse in every 1000ms. There are different
V
cc
Sensor
Read
Control
V
ou
Sensor
Heater Control
R
H
R
S
R
L
1k
Sensor
1k
Q
1
Q
2
options available for heater stabilizing where Jelicic et.al. [11]
propose a similar results for a short term turned off sensor
compared to a continuous turned on sensor.
Fig. 2. Sensor node schematic diagram
Fig. 2 illustrates the block diagram of the proposed sensor
node. A separate node without the sensor module act as the
base station or the gateway to the host computer. Each RF
module has a 16digit IP address and they can be configured to
have a common network ID or different network IDs.
Depending on the requirement these nodes can be
reprogrammed to change the network topology. Only the last
six digits change from RF module to RF module and any
particular sensor node using a RF module with known IP
address can directly be addressed without disturbing the other
nodes in the neighborhood.
Under the current configuration all network nodes in the
network have the same network ID making the RF
communications local to the network. The base station is
connected to a PC via USB port and the measurement data are
fed into a database in the PC. However the base station can
send data via RF link to another node connected to a remote PC
in a different geographical locations for emergency conditions.
The network topology selected for the initial testing
network is the star topology with auxiliary multiple hop links
to find a lost node in case of repetitive direct communications
problems. This situation is a common scenario in built
environment as objects may come in between two RF modules
blocking their line of sight. If the obstructing object acts as a
RF signal absorber then the direct link will be lost. Therefore
the network is programmed in a way that the established
network links are tested for a predefined number of times for
transmission and if not successful the base station will send a
command to adjacent nodes to relay the message to the missing
node.
C. Data Fusion
The data obtained from each node has to be analyzed
individually to check whether it has gone beyond the maximum
allowable level for that particular gas for that particular node
location. Then the data obtained from each node for multiple
gasses need to be fused to make a decision to decide whether it
is a good living environment or not under the current
concentrations of them. Next level is to fuse data obtained from
multiple nodes to make a decision on the suitability of the
complete sensor network area for human health.
IV. I
MPLEMENTATION
The first stage of proposed sensor network was
implemented using Libelium Waspmote sensor nodes and gas
sensors from Figaro. Each sensor node is programmed to
measure the target gas concentration in real time and transmit
measured value upon receiving a request from the base station.
Under this scheme the sensor node is actively listening to the
network base station and as soon as an interrogation signal is
received, it will check the MAC address of the packet to decide
whether it is for it or not. Since the Base station signals are on
broadcast mode, every node in the network will receive all the
requests, but respond only if it was addressed to them.
Start
Base station send
interrogation signal
to node (i)
Wait or skip the
node if failed for 10
consecutive rounds
Does node (i)
respond?
N
Sensor node (i)
starts
measurement.
Send result to
Base station
Y
Change node no
(i) to next
Transfer data to
Base station
Want to stop?
Stop
Y
N
Fig. 3 Flowchart of sensor network version 1 implemented
Fig. 3. Shows the flow diagram of the sensor network
version 1 operation. The main disadvantage of this version is
that each sensor node is awaked full time and consumes full
amount of power. However, this version is better suits for gas
concentrations monitored at very short time intervals because
Control
Module
Power Module
Microcontrolle
r
RF
Module
Data
Control
Sensor
Module
there is no time to put the nodes into sleeping mode to save
power.
As mentioned earlier, the base station will activate direct
links to individual sensor node in a round robin operation,
however it may use multicast transmission commands to
recover a missing node in the network. If a particular sensor
node has difficulties for direct RF communications with the
base station for a prolonged time, the base station will send an
alarm to system administrator to look in to the matter.
In the version 2 of network design, individual nodes will
autonomously measure the gas concentrations at set time
intervals and store them against real time in a local database.
They will be transferred to the base station when interrogated
by the base station and the successfully transmitted data will be
deleted from the local database. Fig.4. shows the operational
flow chart for this.
Start
Base station send
interrogation signal
to node (i)
Wait or skip the
node if failed for 10
consecutive rounds
Does node (i)
respond?
N
Y
Change node no
(i) to next
Transfer data with
time stamp to
Base station.
Clear data buffer
Want to stop?
Stop
Y
N
Synchronize node
clocks to base
station clock
Command every
node to start
measurement at
provided time
interval
Node (i) wake up
Put modules not
necessary for
measurement to
sleep mode
Fig. 4. Flowchart of sensor network version 2 implemented
This version of the sensor network is of particular interest
for disaster management situations when the mains power is
cutoff. The sensor nodes can provide gas concentrations for a
longer period with the saved battery power with partial
sleeping. However, the sensor module has to be powered on for
at least a couple of minutes before making any measurement.
In the first stage of the project a carbon dioxide sensor
module was used with a sensor node to measure indoor CO
2
concentrations in a bedroom during the night time. The sensor
node measures CO
2
concentration at every 20 seconds and
sends measurement data to the base station upon an
interrogation over the RF link using ZigBee protocol. The
selected bedroom is 4m x 4m in size and all windows were
closed for two hours without any occupants before starting the
measurements. The sensor node was placed 1.5m above the
ground closed to the wall opposite to the bed head. Room fan
or air conditioner was not operated during the night.
V. R
ESULTS AND DISCUSSION
Figure 5 shows the measured CO
2
concentration in the
selected bedroom during the night from 7:00pm to 6:00am. The
graph clearly shows that the CO
2
concentration is in the range
of 400ppm, which is the standard concentration in clean air,
before the occupants were accommodated. It starts to rise at a
lower rate when the occupants were in but still the door was
open while the windows were closed. The next section of the
graph shows a linear increment in CO
2
concentration when all
the openings to the room were closed while the occupants were
sleeping. Last section is opened doors and windows with no
occupants inside the room in the morning. It can be clearly
seen that the fresh air from outside has brought down the CO
2
concentration rapidly when the windows and door were open.
The couple of sharp spikes shown on the graph are due to the
operator go close to the sensor node to check the operating
conditions and breathing out at very close proximity to the
sensor.
Figure 5. CO
2
concentration measured in a bedroom
overnight
Even though the occupants didn’t notice it and got used to
this condition for a long time, the carbon dioxide
concentration is more than four tines the normal and well
above the permissible level for healthy living. This
measurement data can be used to activate ventilation system in
an automated HVAC system to provide fresh air. Therefore
this kind of continuous air condition monitoring will be an
essential integrated part in future built environment systems to
ensure a good quality of living. This can be achieved by
incorporating this data with suitable algorithms to control the
HVAC system.
Carbon monoxide, methane and propane sensors work
similarly in the network and development of dedicated and
optimized low power sensor nodes is underway.
R
EFERENCES
[1] J.Y. Kim, T.Z. Shin, and M.K. yang, “ Design of a Smart gas Sensor
System for Room Air-Cleaner of Automobile(Thick-Film metal Oxide
Semiconductor gas Sensor)”, Journal of Electrical Engineering and
Teechnology, Vol.2, Issue 3, pp 408-412, 2007
[2] Zhenyu Han, Gao, R.X., and Zhaoyan Fan, “Occupancy and indoor
environment quality sensing for smart buildings”, Instrumentation and
Measurement Technology Conference (I2MTC), pp 882-887, 2012
[3] Wongchoosuk, C, Khunarak, C., Lutz, M., and Kerdcharoen, T., “WiFi
electronic nose for indoor air monitoring”, 9th International Conference
on Electrical Engineering/Electronics, Computer, Telecommunications
and Information Technology (ECTI-CON), pp 1-4, 2012
[4] A.K. Melikov, J. Kaczmarczyk, “Air movement and perceived air
quality”, Building and Environment, Vol. 47, pp 400-409, 2012.
[5] J.C.M. Pires, S.I.V. Sousa, M.C. Pereira, M.C.M. Alvim-Ferraz, F.G.
Martins, “Management of air quality monitoring using principal
component and cluster analysis—Part I: SO
2
and PM
10
”, Atmospheric
Environment Vol.42, pp 1249–1260, 2008.
[6] J.C.M. Pires, S.I.V. Sousa, M.C. Pereira, M.C.M. Alvim-Ferraz, F.G.
Martins, “Management of air quality monitoring using principal
component and cluster analysis—Part I: SO
2
and PM
10
”, Atmospheric
Environment Vol.42, pp 1261–1274, 2008.
[7] J.C.M. Pires, S.I.V. Sousa, M.C. Pereira, M.C.M. Alvim-Ferraz, and
F.G. Martins, “Management of air quality monitoring using principal
component and cluster analysis—Part I: SO
2
and PM
10
, Atmospheric
Environment 42, 200,8 pp 1249–1260
[8] J.C.M. Pires, S.I.V. Sousa, M.C. Pereira, M.C.M. Alvim-Ferraz, and
F.G. Martins, “Management of air quality monitoring using principal
component and cluster analysis—Part II: CO, NO
2
and O
3
”,
Atmospheric Environment 42, 2008, pp 1261–1274
[9] D. Lazik and S. Ebert, “Improved Membrane-Based Sensor Network for
Reliable Gas Monitoring in the Subsurface”, Sensors 2012, Vol. 12, pp
17058-17073
[10] M. reza Akhondi, A. Talevski, S. Carlsen, and S. Petersen,
“Applications of Wireless Sensor Networks in the Oil, Gas and
Resources Industries”, 24th IEEE International Conference on Advanced
Information Networking and Applications (AINA), 2010 , pp941-948.
[11] V. Jelicic, M. Magno, D. Brunelli, G. Paci, and Luca Benini, “Context-
Adaptive Multimodal Wireless Sensor Network for Energy-Efficient
Gas Monitoring”, IEEE Sensors Journal, Vol. 13, no.1, 2013, pp 328-
338.
This is a copy of the final manuscript of:
D.M.G. Preethichandra, "Design of a Smart Indoor Air Qiuality Monitoring Wireless Sensor
network for Assisted Living", IEEE Instrumentation and Measurement Technology
Conference(I2MTC2013), Minneapolis, USA, may, 2013 pp 1306-1310
DOI: 10.1109/I2MTC.2013.6555624