Wireless Sensor Networks for
Firefighting and Fire Investigation
Sarah A. Summers
Spring 2006
Abstract
Firefighting is one of the most dangerous professions in which people are employed. The
dangers associated with this profession are the result of a number of factors such as lack of
information regarding the location, size and spread of the fire. The use of wireless sensor
networks may be one way of decreasing the risks faced by firefighters and assisting in the
process of rapid extinguishment of the fire. They may also be able to provide fire
investigators with additional knowledge that could assist in determining whether a fire was
maliciously ignited.
Introduction
On a daily basis, firefighters enter potentially life threatening situations armed with little
more than a vague idea of the location, size and spread of the fire they are trying to
extinguish. One of the key factors that results in this lack of knowledge is that current fire
alarms often do little more than notify us of the presence of a fire. They provide little or no
information as to the location of the fire(s) or its spread. Where systems exist which can
provide some information regarding the location of the fire, they are powered by electrical
cables embedded in the buildings infrastructure. As a result, they can be attacked by fire
resulting in a total loss of information.
Although the main hazard to life is reduced significantly once the fire is extinguished,
problems remain for a fire investigator. Frequently, the destruction caused prior to the fire
being extinguished makes the determination of its origin and the manner of its spread
difficult to determine. In many cases, a fire investigator must rely on the eyewitness accounts
1
of the firefighters who fought the fire and/or individuals present when the fire was first
discovered. Any system that can provide knowledge to the Fire Department prior to the entry
into the fire ground and additional information to a fire investigator must be beneficial. This
project focuses on the various challenges that must be addressed for the deployment and use
of wireless sensor networks for initial detection of fire location and subsequent spread of fire
within a building.
Wireless Sensor Networks
Before looking at how wireless sensor networks can be used to assist firefighters in the
performance of their duties, it is first necessary to know something about wireless sensor
networks in terms of how they work; their capabilities and limitations.
A Wireless Sensor Network (WSN) is a network comprised of numerous small independent
sensor nodes or motes. They merge a broad range of information technology; hardware,
software, networking, and programming methodologies [1].
Wireless Sensor Networks can be applied to a range of applications [1] monitoring of space
which includes environmental and habitat monitoring, indoor climate control, surveillance
etc.; monitoring things for example structural monitoring, condition-based equipment
maintenance etc.; and monitoring the interactions of things with each other and the
surrounding space e.g., emergency response, disaster management, healthcare etc. The
majority of these applications may be split into two classifications: data collection and event
detection.
Each mote in a wireless sensor network is a self-contained unit comprised of a power supply
(generally batteries), a communication device (radio transceivers), a sensor or sensors,
analog-to-digital converters (ADCs), a microprocessor, and data storage [1, 2]. The motes
self organize themselves, into wireless networks (Figure 1) and data from the motes is
relayed to neighboring motes until it reaches the desired destination for processing [2].
2
Stargate
Mote
Figure 1:
Example of a Flat Network (adapted from [3])
Each mote has very limited resources in terms of processing speed, storage capacity and
communication bandwidth. In addition, their lifetime is determined by their ability to
conserve power [4]. These limitations are a significant factor and must be addressed when
designing and implementing a wireless sensor network for a specific application.
Data Collection versus Event Detection
As stated above, in general wireless sensor networks can be categorized into one of two
types, data collection or event detection networks. In many applications where data collection
is the goal, the sensors may be required to collect data for short periods at set times of the
day. In this case, most of the time the sensor node will be asleep thus conserving power.
However, where a wireless sensor network is to be employed for event detection, such as
detecting the ignition of a fire, it would be anticipated that the sensor nodes must remain
awake thus consuming their precious limited power [16].
3
Hardware and Tools for Developing a Wireless Sensor Network
A variety of hardware and tools are available for deploying and testing wireless sensor
networks. Brief descriptions are provided hereafter.
Sensor Mote Hardware Architecture
Currently one of the most popular research platforms is the Mica2 sensor mote shown in
Figure 2. It uses the TinyOS (TOS) Distributed Software Operating System, has a 325, 433
or 868/916 MHz multi-channel radio transceiver and an expansion connector that can be used
for light, temperature, RH, barometric pressure, acceleration/seismic, acoustic and magnetic
sensor boards [5]. Figure 3 shows the standard indoor injection molded housings available
for the Mica2.
Figure 2:
Mica2 Sensor Mote
4
Figure 3:
Injection molded housing assembly for the Mica 2
TinyOS (TOS), TOSSIM and TinyViz
TinyOS
The limited processing capabilities of the motes, precludes the use of standard operating
systems. As a result, the motes use TinyOS. TinyOS is a programming framework,
developed at the University of California, Berkeley, for embedded systems and a set of
components that enable building an application specific operating system into each
application [6]. The programming language capabilities for TinyOS are provided through a
stylized version of C using a customized compiler called nesC [7].
TOSSIM
TOSSIM is a simulator for TinyOS networks, users can compile a TinyOS application into
the TOSSIM framework where they can then debug, test, and analyze algorithms in a
controlled and repeatable environment [8].
5
TinyViz
TinyViz is a graphical interface that can be used with TOSSIM. It can be attached to a
running simulation and with the use of plug-ins can be used to visualize such things as
network traffic. Users can write their own plug-ins for events they wish to visualize.
In view of the potential expense of real-world trials, the development of a simulator to assess
the feasibility of a wireless sensor network for this application would need to be carried out.
It is anticipated that this could be undertaken as additional research.
Current Research into Wireless Sensor Networks and Firefighting
In the past, research into using wireless sensor networks for fire detection has predominantly
focused on the detection and tracking of wildfires. It is only in the last few years, since the
attack on the World Trade Center, there has been increasing focus on the safety of and
communication between firefighters on the fire ground. This has resulted in researchers
looking at the possibilities of using wireless sensor networks when fighting structural fires.
One group at UC Berkeley has been working on a project called FIRE (Fire Information and
Rescue Equipment) [9]. The objective of this project is to develop both hardware and
software tools to improve the safety, efficiency and effectiveness of firefighting.
The research is comprised of three sections, a wireless sensor network called SmokeNet,
which forms the basis of the whole project, a head mounted display unit for individual
firefighters called FireEye, and an incident command system called eICS which will be a
visual display showing information such as resource allocation, location of personnel on the
floor plans of the building, and biometric data of firefighters including air supply and heart
rate.
The SmokeNet implementation is based on TinyOS and utilizes Crossbow wireless smoke
and temperature detecting sensor nodes. In a non-alert state, these nodes check the status of
6
the environment every 10 seconds and send data obtained along with their battery condition
via a multi-hop network to a central building node logger computer every 5 minutes. On
detecting a fire, a node sends an alert message which places the whole network in an alert
state. In this state, each node checks for fire every five seconds and reports to the data logger
every two minutes if no fire is detected to confirm that it is still alive [10].
The system also allows for the Incident Commander to connect into the wireless sensor
network, via eICS, to determine the location of the fire and also track firefighters and their
health status. It also allows for data to be sent to firefighters wearing the FireEye.
Another group has been working on a project called “Siren – Context-aware Computing for
Firefighting” [11]. The system that they are developing supports tacit communication among
firefighters using a context-aware messaging application. Each firefighter carries a WiFi
enabled PDA with a built in Berkeley motes sensor board. The mote in the PDA collects data
from motes which are pre-deployed in the building to inform the firefighter of hazards and
immediate danger. In addition, the pre-deployed motes also serve as location beacons, thus
enabling a firefighter to navigate his/her way through the building. Each PDA connects to the
PDA’s of other firefighters in a peering mode.
The Challenges Faced by Firefighters
As stated above, over the last few years, two groups [10, 12] have been carrying out research
into development of information providing tools for use by firefighters. In order to better
understand the needs of firefighters, both groups interviewed personnel from the Fire
Department. Their findings are directly relevant to this project and are reviewed here.
The interviews of Steingart et al [10] revealed that the Fire Chiefs considered the following
as the most important pieces of information during an emergency:
1.
Proximity of the firefighters to danger.
2.
The health status of the firefighters.
3.
Better radio communication.
7
4.
Location and ambient temperature of the firefighters.
5.
The validity of building floor plans.
The firefighters interviewed indicated that although they believed that technology was a good
thing, they were concerned about unreliability and increasing dependence on technology.
Another significant factor that was identified was that Fire Departments have little money for
technology or equipment that is not mandated by a standards body.
In their research, Jiang et al. [12] identified the incident commander (IC) as being the most
information intensive position, since in addition to coordinating the overall response strategy,
they must also manage the available personnel and resources. In terms of the dangers faced
by firefighters, the IC’s were concerned with being aware of the following dangers:
1.
“Flashover, sudden ignition of the contents in a room.”
2.
“Backdrafts, explosions that occur when an oxygen starved fire suddenly receives
oxygen.”
3.
“Hidden fires in walls, attics, and other unseen areas.”
4.
“Structural hazards, including structural collapse and toxic gases from burning
hazardous materials.”
Jiang et al. noted that at the present time, there are no special technologies available to
firefighters to assist in avoiding the problems described above.
Another important factor identified in both studies was that information regarding the status
of the fire is predominantly communicated face to face or via radio. The firefighters
interviewed stated that the noise associated with the fire and firefighting activities made
communication difficult. This was exacerbated by the presence of radio dead zones on the
fire ground.
8
Fire Data
In order to develop a wireless sensor network for use in fire detection and tracking, it is
important, not only to identify the challenges faced by firefighters, but also to gain some
understanding about the potential temperatures involved, the factors may affect them and the
development of smoke. An appreciation of this sort of information enables assumptions to be
made when designing the wireless sensor network in terms of deployment, protection of the
sensors, and determining the increases in temperature that will cause the sensors to trigger an
alarm.
The information presented here is from a full scale house fire experiment [13]. The structure
used for the experiment was a two story, single family dwelling of wood construction. The
simulation involved the accidental ignition of a flaming fire by a space heater. Two arrays of
thermocouples were used, one array located in the room where the fire was ignited, at a
distance of approximately 3 meters from the point of ignition and the other array in an
adjacent room. The thermocouples were distributed at various distances from the ceiling
downwards.
Figure 4 shows the temperatures recorded by the thermocouple in the room where the fire
was ignited. As can be seen, there was a significant temperature rise, 121°C in the first five
seconds, and 187°C in the first twenty seconds after the fire was ignited at the thermocouple
located at a height of 2.26m (7.42 ft).
9
Figure 4:
Thermocouple temperatures from the room in which the fire ignited.
Figure 5 shows the comparative data from thermocouples located in the room adjacent to
that which the fire was ignited. It can be seen that the initial temperature rise is not as rapid.
In the first five seconds, the temperature at the thermocouple located at a height of 2.26m
(7.42 ft) rose only 7°C in the first five seconds, and 121°C in the first 20 seconds after the
fire was ignited.
Figure 5:
Thermocouple temperatures from the room adjacent to the one in which the
fire ignited.
10
From both figures 4 and 5, it can be seen that the greatest temperature rise is recorded by the
thermocouples located closest to the ceiling, the temperature rise being considerably less the
closer the thermocouple to the floor.
It should be noted that this is not a comprehensive investigation into fires and their associated
temperatures, but rather a small snapshot. The initial ignition and subsequent spread of a fire
will depend upon many factors, such as whether the fire was accidentally or maliciously
ignited (with the use of accelerants), the type of structure in which the fire occurs,
ventilation, fire loading (amount of combustible materials present) and whether the fire
ignites as a flaming fire or a smoldering fire etc. It should be noted the initial temperature
rise in a smoldering fire will be negligible in the early stages of the fire.
Requirements for Wireless Sensor Networks for Firefighting
It is apparent that many of the issues and information needs identified by the Fire Department
could be addressed by the use of pre-deployed wireless sensor networks. Table 1 shows the
problems to be addressed and the potential wireless sensor network solution.
Problem
Proximity of firefighters to danger
WSN Solution
Pre-deployed temperature, smoke, oxygen,
accelerometer and olfactory sensors
Flashover
Pre-deployed temperature sensors
Backdrafts
Pre-deployed oxygen sensors
Hidden Fires
Pre-deployed temperature and smoke
sensors
Structural collapse
Pre-deployed Accelerometers
Toxic gases
Pre-deployed Olfactory sensors
Table 1:
Potential Wireless Sensor Usage (adapted from [15])
11
In designing and implementing a pre-deployed wireless sensor network for firefighters, there
are number of requirements that must be satisfied, first and foremost whatever system is
implemented, it must be reliable and cost effective, otherwise whatever its potential benefits
it is unlikely to be generally adopted. Assuming the reliability and cost issues can be
overcome then the following are the challenges that must be addressed:
1.
The possibility of false alarms must be minimized. It should be obvious that the
occurrence of false alarms should be minimized as they waste the time and resources
of the Fire Department and can potentially result in fire fighting resources being
unavailable to immediately attend genuine incidents
2.
The wireless sensor network must be secure in order to prevent malicious triggering
of false alarms and malicious sending of false data.
3.
Due to the rapidity with which fire can spread, it is essential that the sensor node(s)
detecting the event begin transmitting data as soon as and event is detected. Failure to
do so could result in alarms not being triggered and data of value to the Fire
Department being lost. In addition, the sensor node(s) detecting the event should also
wake other sensor nodes in the vicinity of the event.
4.
The Fire Department must be able to connect to the wireless sensor network in the
building of interest. A system for monitoring a fire is only of value if the Fire
Department can connect into the network to retrieve the data.
5.
The network must be self modifying to ensure that data can still be transmitted
if/when individual sensors fail. The very nature of a fire will undoubtedly result in the
destruction or failure of sensors. Consequently, in order to ensure that data regarding
the condition of the fire continues to be sent and/or received, the network must be self
modifying.
12
6.
There must be rapid transfer of data. To be of any value to firefighters, data from the
various sensors comprising the network must be rapidly and correctly received.
7.
The positions of the sensors must be known. In order to provide an accurate picture of
the location and spread of the fire, the locations of the sensors within the building
must be known.
8.
There should be a visual display show the location and spread of the fire and
temperatures within the building.
9.
The sensors and their housings must be protected as much as possible from the heat
generated by the fire in order to keep them functional as long as possible, without
degrading the ability of the sensors to detect changes in temperature or any other
parameters that they are sensing.
In order to address all of the needs highlighted in Table 1, a multitude of different sensors
would need to be pre-deployed in a building. Although motes are rapidly decreasing in price,
this option does not seem feasible at the current time. In addition, until the true value of a
wireless sensor network system can be proved, it would seem sensible to concentrate on a
minimal set of sensor motes that could provide the basic information required by the Fire
Department.
On the basis of the above and the assumption that in general fires within buildings are
detected by smoke or rapid temperature rises, this project considers the deployment of
temperature and smoke sensors as a bare minimum.
FAWSNet (Firefighters And Wireless Sensor Networks)
FAWSNet (Firefighting And Wireless Sensor Network) is the name that is being given to the
suggested implementation. It is the intention that FAWSNet be comprised of a graphical user
interface for use by the Incident Commander, and a pre-deployed sensor network, similar in
13
some ways to the FIRE project currently in progress at UC Berkeley. It is intended that
FAWSNet be used by Incident Commanders to track the initial location and subsequent
spread of the fire. If this proves to be effective then further additions to the system could be
considered.
In an ideal situation, a variety of wireless sensors would be deployed, first and foremost
would be temperature sensors as it is these, in most cases, that will detect the presence of fire.
Other sensors would include smoke detectors, infra-red sensors and motion detectors.
The wireless sensor network would be connected via the internet to the Fire Department
Control. Thus providing an instant emergency call if a fire were detected. By being
connected over the internet, as soon as a fire is detected, the controllers will be able to
provide the deployed firefighters with an initial location of the fire within the building and
also an update as to the spread of the fire during their journey to the fire ground. On arrival at
the fire ground, the Incident Commander would then link into the wireless sensor network
and be provided with a constant stream of information about the location and spread of the
fire. It is the intention that the owner/occupier of each building with a wireless sensor
network would provide the Fire Department with detailed floor plans of the premises. These
would be automatically displayed along with the locations of the sensors and their current
status.
Once at the fire ground, the Incident Commander’s laptop would take over from the
building’s central monitoring system as the main data depository for the wireless sensor
network. In this way, if the building’s central monitoring/control system was attacked by fire,
the wireless sensor network would remain functional as long as sensors were functional
within the building.
Minimizing False Alarms
On the basis that fire will be detected by either temperature sensors or smoke detectors, care
must be taken when setting triggering levels. For example, it would be undesirable if the
14
temperature sensor were to indicate that a fire had ignited when in fact all that had happened
was that the temperature had risen due to the central heating installation starting up. From the
data obtained from the full scale house burn [13], it is known that the initial temperature rise
when a fire ignites is very rapid, in the case of a flaming fire.
If the assumption is made that a central heating system will not cause the temperature to rise
more than 10°C above the preset value, then it can be said that any temperature rise greater
than 10°C above the room temperature setting can be considered to be caused by the ignition
of a fire. Obviously, even if the temperature does rise more than 10°C above the preset value,
it is unlikely that it would rise more than 10°C in 5 seconds. As a result, the temperature
sensor should be programmed to create an alarm only if the temperature rises more than 10°C
in 5 seconds.
The smoke detector is another of the sensors which could be the first to detect the ignition of
a fire and cause a false alarm to be raised and they can also cause false alarms. There are two
types of smoke detector, the ionization smoke detector and the photoelectric smoke detector.
In general, the ionization type of detector is more suitable for detection of fast flaming fires,
while photoelectric types are better for detecting slow smoldering fires [14]. Both types can
produce false alarms, for example dust and dirt accumulations in both types of detector can
cause them to become more sensitive. They can also be triggered by certain aerosol products.
In view of this fact, data needs to be obtained on how rapidly an aerosol product would
disperse so that a triggering condition can be set.
For the purposes of FAWSNet, it would be better to utilize the photoelectric type of sensor
since it would be anticipated that the temperature sensors would detect fast flaming fires due
to the significant temperature rise, however, they may be slower at detecting smoldering fires
where the initial temperature increase is negligible – thus allowing the fire to go undetected
for longer.
15
Wireless Sensor Network Security
It is not unknown for fires to be maliciously ignited, therefore it must be assumed that
arsonists or other individuals may attempt to gain access to the wireless sensor network in
order to trigger a false alarm, or worse modify the data collected by the sensors resulting in
false data being transmitted. In order to keep the current design and implementation as
simple as possible, no further consideration will be given to this aspect in the current report.
However, it is an area that needs further consideration and should be considered in future
work.
Scheduled Testing and Event Detection
The failure of sensor nodes is not uncommon. As a result, if a wireless sensor network is to
be of value it is essential that scheduled testing is carried out to confirm that the nodes
comprising the network are still functional, and have adequate power resources. For the
proposed fire detection and tracking application, it is anticipated that daily testing of the
nodes would be adequate. However, this will depend upon the environment in which the
sensor nodes are deployed. For example, it could be expected that the node failure rate would
be higher in a factory environment than in an office environment.
As stated earlier in this paper, fires can spread rapidly, so any pre-deployed sensor nodes
must rapidly detect the event and immediately start to transmit data. This poses a potentially
serious problem, since in order to detect an event a sensor node must be awake, and whilst in
the wakeful state the node will be consuming its limited power resources. It is not sufficient
for the node to detect the event it must also have the power resources to transmit the data
regarding the event rapidly and accurately. Obviously, if all the sensor nodes were to remain
in a continuous wakeful state the life expectancy of the network, due to the limited power
resources, would be seriously curtailed. Consequently, it is necessary to develop a schedule
to maximize the life of the nodes whilst still providing adequate detection capabilities.
Although, this particular application does not preclude the replacement of batteries since the
16
sensor nodes are accessible, relying on people to regularly replace batteries is hazardous
since it is often a job that will be postponed with potentially disastrous effects.
He et al. [17] proposed a “sentry service”, for an energy wireless sensor networks which
could be used in FAWSNet. Energy of the nodes is conserved by selection of a subset of
nodes which are defined as sentries, to detect events. The remaining nodes in the network are
allowed to remain in a low power sleep state. If an event occurs, the sentries wake other
nodes in the vicinity of the event. Obviously the energy reserves of the sentries will be
depleted more rapidly, so it is necessary to rotate the sentry responsibility among the nodes in
the network to achieve uniform energy dissipation across the network. If this scheme is to be
used in FAWSNet it will be essential to ensure, that at all times, there is a minimum of one
sentry node in a room within the building. Obviously, if the rooms within a building are
particularly large it would be necessary to have several nodes defined as sentry nodes.
Self-Modifying Network
It has already been stated that the temperatures encountered in a fire can be extremely high
and are likely to quite rapidly destroy sensors in the vicinity of the event. Therefore, it is
essential that if a mote is destroyed, the network self modifies in order that data can continue
to be transmitted.
Sensor Node Location
It is anticipated that when the sensor nodes are deployed throughout the building, each will
have a unique identification number. This data will be incorporated with the building floor
plans supplied to the Fire Department. As a result on attendance at an incident, the location
of the sensor nodes with respect the various rooms within the building will be known.
17
Sensor Protection
Sensors by their very nature are sensitive instruments that can potentially be easily damaged.
Since during a fire event, the temperatures in the building near the fire location will be high,
it is necessary to protect the sensors and their housings as much as possible from the heat
generated by the fire in order to keep them functional as long as possible, without degrading
the ability of the sensors to detect changes in temperature. Most sensors currently available
are supplied in an injection molded plastic housing. Obviously, this type of housing is
particularly susceptible to heat damage. Therefore, it would be necessary to utilize a more
heat resistant casing in which house the sensor.
Sensor Deployment
From the data shown in figures 4 and 5, it can be seen that the temperatures recorded by the
thermocouples differs significantly the closer the sensor is to the floor. In terms of the
deployment of temperature sensors, it makes sense to deploy an array of sensors at varying
heights. The thermocouples located close to the ceiling will detect the ignition of the fire and
raise the alarm. However, due to the temperatures to which they would be exposed they
would be expected to fail rapidly. By deploying some sensors at a lower level, it would still
be possible to monitor the status of the fire.
The motes may be multi-sensing (temperature and smoke) or single-sensing with temperature
and smoke sensors situated at different locations within the building. The configuration
chosen for this project is single-sensing sensors. Figure 6 shows a potential layout of sensor
motes within a single storey building. Located at various points around the building are
temperature sensors, shown as blue circles, rather than a single sensor, they denote an array
of sensors located at different levels below the ceiling of the room. Also shown in the figure
are the smoke detectors, shown as brown circles.
18
21
22
Room 3
Room 1
28
6
7
Room 2
29
30
20
23
27
32
31
5
19
26
35
33
18
4
Corridor
24
34
25
8
9
3
10
15
36
37
38
Room 4
11
Room 5
16
39
41
Figure 6:
2
14
1
40
13
17
12
Potential Layout of sensor motes within a building
Figure 7 shows a cross section of the room and the potential layout of sensors.
Array of
Temperature
Sensors
Figure 7:
Smoke
Detector
Cross section of room showing the sensor layout.
19
Wireless Sensor Network Operation during a Fire Event
If we consider that a fire is ignited in Room 4, as shown by the red circle in Figure 8, and
that mote 40 is a sentry mote and is the first mote to detect a change in temperature. The
motes will be programmed to send an event alarm message if a temperature rise of greater
than 10 degrees in 5 seconds is detected. Mote 40 is shown in red to indicate that it has
detected a fire; it will wake motes in its vicinity, place them on alert (denoted by the change
in color to orange) and start to send data. In addition to waking motes in the vicinity of the
event, it would also be advisable to place a wake up call to the remaining motes in the
building. Obviously, the number of motes that initially detect a rise in temperature above the
trigger level will depend on the size of the room and their location and whether or not they
are assigned as sentry motes at the time of the event.
Figure 8:
A fire has ignited in Room 4
20
As the temperature rises and the fire spreads, more motes will be detecting the event and will
want to pass data. In order to avoid conflict, the functional motes closest to the event and
experiencing the greatest rise in temperature should be given priority.
It is anticipated that the graphical user interface available to the incident commander would
be similar to the floor plan shown in figure 8. As temperatures in a particular area rise, the
room on the floor plan could be designated a specific color dependent on the temperatures
being reported thus providing instant visual feedback as to conditions. Actual temperatures
could be displayed alongside the floor plan in the GUI or in another panel.
Sensor Event Triggering Algorithms
Each mote installed in the building will have to be preprogrammed such that it knows when
an event has occurred. The following basic algorithms are suggested, although these are
likely to be modified with further research.
Temperature Sensors
1.
If temperature sensing mote detects temperature increase > 10°C in 5 seconds then
fire ignited, trigger alarm.
2.
If alarm triggered, determine closest neighboring sensors (temperature and smoke)
and send them a wake up call.
3.
Send location of detecting mote and detected condition to controller.
4.
Draw mote detecting fire on GUI as red circle.
5.
If a neighboring mote detects temperature rise > 10°C but less than 50°C and alarm
has already triggered, draw motes as an amber square to indicate heat associated with
fire.
6.
Send location of mote and event to controller.
7.
If a neighboring mote detects temperature rise > 70°C and alarm has already
triggered, draw motes as an amber circle to indicate fire spreading.
8.
Send location of mote and event to controller.
21
With the exception of the 10°C trigger temperature, the temperatures given in the above
algorithm are rough estimates. Further research into the temperature increases associated
with fire spread needs to be undertaken to establish best settings.
It should be noted that the temperature rise for motes located at great distances from the
ceiling will, in general, experience a low rise in temperature during the initial stages of a fire.
As a result, it may be necessary to modify the above temperature algorithm to take account of
this especially if the sentry service is to be successfully utilized.
Smoke Detectors
1.
If smoke detected for time > 30 seconds then fire ignited, trigger alarm.
2.
If alarm triggered, determine closest neighboring sensors (smoke and temperature)
and send them a wake up call.
3.
Send location of detecting mote and detected condition to controller.
4.
Draw mote detecting fire on GUI as red square.
5.
If a neighboring mote detects smoke and alarm has already triggered, and nearest
temperature sensors have not been triggered draw smoke detector as an amber square
to indicate smoke in that area.
6.
Send location of mote and event to controller.
In order to carry out a simulation to test the effectiveness of the above algorithms, it would
also be necessary to produce an algorithm that could effectively simulate a fire occurring
within the building. It would be necessary to inject the temperature rise associated with a fire
into a mote and then spread the conditions to other motes in order to see how the simulation
runs. After a predetermined period of time, it would be necessary to disable one or motes to
simulate the destruction of the mote by the fire and see if the network was successful in self
modifying and continued to pass data.
22
Future Work
This paper has focused on the important aspects with regard to the types of sensors deployed,
how they should operate in the event of a fire etc. Obviously, there are many other aspects
that need to be considered, for example, how the wireless sensor network connects,
communicates and transfers data to the graphical user interface, the self modification of the
network in the event of a mote failure. These are all issues that need to be addressed in future
work are briefly described below.
Viability of Temperature Sensors in a Fire Environment
It should be apparent from the preceding discussion that sensors deployed in a fire
environment are exposed to extremely harsh conditions in terms of the temperatures to which
they are exposed. Although it has been seen that temperature sensors located at low level in
a room in which a fire is ignited will detect a relatively rapid rise in temperature, those
deployed close to the ceiling will detect changes more rapidly thus potentially providing a
more rapid notification of the presence of a fire. The concern is that even when located at low
levels, sensors may be destroyed before they are able to provide meaningful data.
TOSSIM Simulation
The next step in the research would be to create a simulation using TOSSIM to determine if
the suggested implementation would be feasible. It is intended that the simulation would
adapt the location tracking simulation developed by Sinha [6].
Development of Graphical User Interface
If the TOSSIM simulation proved to be successful then a graphical user interface would need
to be developed. In the real world simulation this is what would be used by an Incident
Commander to track the spread of the fire. It should clearly show the conditions existing
within the building in terms of location of the fire, spread of the fire, smoke and hotspots. It
23
will be essential that the data shown represent as closely as possible the real time events
taking place within the fire ground.
Inclusion of Additional Sensors
Mention was made of the possibility of including other types of sensors, such as infra-red and
motion sensors. The addition of these to the model may provide additional valuable
information for tacking the fire over and above that provided by the temperature sensors and
smoke detectors. Their inclusion may in fact provide a more accurate location for the
position of the fire at any given time, since the temperature sensors will only provide the
location in which the temperature is rising.
Wireless Sensor Network Security
One of the concerns already mentioned is the possibility that unauthorized users could gain
access to the wireless sensor network and trigger false alarms or introduce malicious data.
Consequently, it will be necessary to investigate ways in which the network can be protected
from such intervention.
Physical Testing of Implementation
Once all of the above considerations have been addressed, it would be necessary to carry out
a physical test of the implementation. Testing of this nature is both time consuming and
expensive. It would require the co-operation of the Fire Department and the setting up of a
suitable test environment. In addition to pre-deploying wireless sensors in the environment, it
would also be necessary to utilize additional instrumentation such as thermocouples and
cameras to ensure that the data being collected and transmitted to the graphical user interface
was indeed accurate and being received in a timely manner.
24
Conclusions
There are many risks associated with fire fighting, not least of which are the unknown factors
such as location and spread of the fire. On a theoretical basis, it would appear that the use of
a wireless sensor network, such as FAWSNet, could assist in diminishing these unknowns by
potentially providing Incident Commanders with real time information regarding the location
and spread of the fire.
At the current time, FAWSNet remains theoretical. There are many issues that must still be
addressed not least of which is can the sensors which first detect the fire withstand the rapid
temperature rises expected for long enough to transmit meaningful data. Before further work
is undertaken on this project, it would be advisable to establish if currently available sensors
can withstand the temperatures to which they are likely to be exposed, and what if any
protection can be provided that will allow to remain operational for as long as possible in the
environmental conditions experienced in a fire.
25
References
[1]
Overview of Sensor Networks, D. Culler, D. Estrin and M. Srivastava, Computer,
Vol. 37, Issue No. 8, August 2004, pp. 41 – 49,
http://csdl2.computer.org/comp/mags/co/2004/08/r8041.pdf
[2]
Sensor Network Operation,
http://www.intel.com/research/exploratory/sensornetwork_operation.htm
[3]
Intel Motes and Wireless Sensor Networks,
http://www.intel.com/research/downloads/SNOverviewCD.pdf
[4]
Sensor Nets/ RFID, http://www.intel.com/research/exploratory/instrument_world.htm
[5]
Crossbow Technology Inc. Data Sheet,
http://www.xbow.com/Products/Product_pdf_files/Wireless_pdf/MICA2_Datasheet.p
df
[6]
Sensor Based Efficient Multi-Floor Location Tracking, D. Sinha, Masters Thesis,
University of Colorado at Colorado Springs.
http://cs.uccs.edu/~gsc/pub/master/dsinha/doc/DEVJANI_PROJECT_REPORT.doc
[7]
Wireless Sensor Networks, pt 1: Introduction, Eliana Stavrou
http://webhosting.devshed.com/c/a/Web-Hosting-Articles/Wireless-Sensor-Networkspt-1-Introduction/
[8]
TOSSIM: A Simulator for TinyOS Networks, P. Levis and N. Lee, September 17,
2003, http://www.cs.berkeley.edu/~pal/pubs/nido.pdf
[9]
Design of Monocular Head-Mounted Displays for Increased Indoor Firefighting
Safety and Efficiency, J. Wilson, D. Steingart, R. Romero, J. Reynolds, E. Mellers, A.
Redfern, L. Lim, W. Watts, C. Patton, J. Baker and P. Wright, Proceedings of SPIE -Volume 5800, Helmet- and Head-Mounted Displays X: Technologies and
Applications, May 2005, pp. 103-114,
http://bmi.berkeley.edu/fire/SPIEpaperFinal.pdf
26
[10]
Augmented Cognition for Fire Emergency Response: An Iterative User Study, D.
Steingart, J. Wilson, A. Redfern, P. Wright, R. Romero and L. Lim, Proceedings of
the 1st International Conference on Augmented Cognition, Las Vegas, NV, 22-27
July 2005, http://bmi.berkeley.edu/our_lab/images/AugCogFinal.pdf
[11]
Siren: Context-aware Computing for Firefighting, X. Jiang, N. Y. Chen, J. I. Hong,
K. Wang, L. Takayama and J. A. Landay, Proceedings of Second International
Conference on Pervasive Computing (Pervasive 2004). , Vienna, Austria, April 18 April 23, 2004, http://dub.washington.edu/projects/siren/pubs/pervasive2004-sirenfinal.pdf
[12]
Ubiquitous Computing for Firefighters: Field Studies and Prototypes of Large
Displays for Incident Command, X. Jiang, J. I. Hong, L. A. Takayama and J. A.
Landay, CHI Letters (Human Factors in Computing Systems: CHI 2004). , Vienna,
Austria, April 24 - 29, 2004, http://dub.washington.edu/projects/siren/pubs/CHI2004firefighters-final.pdf
[13]
Full-Scale House Fire Experiment for InterFIRE VR, Report of Test, A.D. Putorti Jr
and J. McElroy, November 2, 1999, Revised April 10, 2000,
http://www.interfire.org/features/fire_experiment.asp
[14]
System Smoke Detectors, http://www.systemsensor.com/html/guides/A05-1003.pdf
[15]
Application of Wireless Sensor Mote for Building Risk Monitoring, N. Kurata, B. F.
Spencer, Jr, and M. Ruiz-Sandoval,
http://www.unl.im.dendai.ac.jp/INSS2004/INSS2004_papers/PosterPresentations/P8.
pdf, (2003)
[16]
On Random Event Detection with Wireless Sensor Networks, P. K. Dutta, Masters
Thesis, Ohio State University, 2004,
http://www.cs.berkeley.edu/~prabal/pubs/masters/dutta04masters.pdf
27
[17]
Energy-Efficient Surveillance System Using Wireless Sensor Networks, T. He, S.
Krishnamurthy, J. A Stankovic, T. Abdelzaher, L. Luo, R. Stoleru, T. Yan, L. Gu, J.
Hui and B. Krogh, MobiSYS ’04, June 6 -9, 2004, Boston, MA,
http://www.cs.virginia.edu/papers/tracking-mobisys04.pdf
28