An Autonomous Firefighting Robot

Ahmed Hassanein, Mohanad Elhawary, Nour Jaber, and Mohammed El-Abd

Department of Electrical and Computer Engineering
American University of Kuwait, Kuwait
[S00006531, S00011503, njaber, melabd - @auk.edu.kw]

Abstract: The field of firefighting has long been a features. At the end of the project the robot should be
dangerous one, and there have been numerous and capable of:
devastating losses because of a lack in technological
advancement. Additionally, the current methods  Traversing a map
applied in firefighting are inadequate and inefficient - Wall following.
relying heavily on humans who are prone to error, no - Obstacle detection/avoidance.
matter how extensively they have been trained. A recent  Interacting with fires
trend that has become popular is to use robots instead - Detect fire hazards.
of humans to handle fire hazards. This is mainly - Extinguish flames.
because they can be used in situations that are too - Fire an alarm.
dangerous for any individual to involve themselves in.  Providing external communications
In our project, we develop a robot that is able to locate
- Live Camera Feed.
and extinguish fire in a given environment. The robot
navigates the arena and avoids any obstacles it faces in
- Map representation via Bluetooth.
its excursion.
Through the remainder of this paper, a quick
Keywords: flame sensor; proximity sensor; infrared literature survey will be conducted leading onto the
sensor; ultrasonic sensor; digital compass; design specifics of hardware used, moving on to the
microcontroller. implementation to meet our goals and objectives
before evaluating the final results.
I. INTRODUCTION
II. LITERATURE SURVEY
Safety and speed are two of the most important
aspects of any rescue operation. When a building or A main inspiration that set us on the road to
an area is subject to a fire, it can be very dangerous for completing this project was the Trinity College
any of the rescuing parties to assess the situation Firefighting Robotics Competition [1]. For the
themselves without being subject to harm. A great competition an arena with 4 rooms was to be traversed
alternative to this would be to use robots. Thus the by a robot to locate one candle (representing a fire)
problem by definition is being able to locate and and extinguish it. Certain constraints such as map size,
extinguish a fire with minimal assistance of human number of rooms, room positions, time taken to
beings and as fast as possible using a robot. This is to traverse the arena and robot size were recreated for
reduce the likelihood of anyone being injured as well this project in order to measure the success of our
as increase the promptness of putting out the fire due robot against available standards. Also, the research
to the fact that a robot should not be affected by issues conducted led us to many other projects who had
such as smoke or lack of oxygen. The following are previously participated in this competition.
the goals of this project: From the different papers reviewed, certain
 To increase safety associated with firefighting. patterns started to appear that helped understand
 To create a way to check for fires without different sections of the task at hand. To being with,
involving human beings directly such that they the most popular fire detection technique is by using
are not exposed to any threats. ultraviolet sensors due to their high range and ability
 To be capable of traversing an arena without to detect fires through walls in some cases [2] [3] [4]
any concerns about inhaling smoke or burning [5]. Additionally, the most popular fire extinguishing
at any point during the exploration. system was an air fan [2] [6] [7]. This practice is not
 Decrease the time taken to traverse a map with realistic, but for a small scale robot and only a candle
potential fire. to turn off, it is very efficient in terms of how easy it is
to implement and how clean it is at extinguishing the
To accomplish the aforementioned goals, the fire. Navigation and maps were mostly done through
realized robot must have a number of important wall following [2] [4].

978-1-4673-7509-2/15/$31.00 ©2015 IEEE

 With all of these considerations in mind, the team E. Live Feed
made up their own decision on which sensors and Since in a real life situation a robot will enter into
systems to use which is discussed in the following an environment with fire hazard unaccompanied by
section. humans, there needs to be a way for them to see what
III. DESIGN & ANALYSIS is happening inside. Therefore, the robot had a
mounted JMC 2.4Ghz wireless surveillance camera to
In this section, the selected hardware components show this. The camera had a separate system with its
to be used for this project are reviewed. Keep in mind own transceiver that will receive the signal sent by the
that the competition constraints were thought of as the camera and display it on an LCD screen.
design process was taking place.
F. Map Representation System
A. Robot Kit
In order to have a map representation on a
The robot kit used for this project was the K- computer, several components were obtained to form
Junior V2 peripherals in the kit made it a suitable the map representation system. These include a digital
choice [8]. The brain of the computer was a PIC compass, an Arduino Mega microcontroller and an
microcontroller. Additionally, there were 10 IR additional Bluetooth since more pins were required in
sensors (5 spread out across the front of the robot at additional to a new brain to increase processing
different angles, 4 on the bottom and one sensor on power. The job of the Arduino microcontroller was to
the rear), a few LEDs, a Bluetooth module and a built- collect data from the digital compass and send it to
in buzzer. The robot kit also included an expansion MATLAB, via Bluetooth, for processing and map
general I/O board that would allow interfacing of representation. It would also interface with the K-
other electronics, in addition to an ultrasonic sensor Junior to receive data on the movement of the robot
that could be installed on the robot. The IDE provided also to be forwarded to MATLAB to assist in map
with the kit, was programmable in C. representation with the data from the digital compass.
B. Flame Sensor 1) Digital Compass
From the research conducted at the early stages of Digital compasses are a type of sensor that tells
the project, there were two promising flame sensors the robot how much they have turned from their
that could be used, the DFRobot sensor [9] and the original position. The digital compass used in this
YWR sensor [10]. They were easy to integrate with project was the GY-26 [11]. The ports used on the
the K-Junior V2 Robot via the I/O shield, both were digital compass were RX/TX which required the use
simple to program and on paper they had the same of another microcontroller as the K-Junior V2 did
detection range. After testing, it was found that the not have these ports available, which is another
YWR had a superior practical range, therefore it was reason for using the Arduino Mega microcontroller.
used and the DFRobot sensor was discarded.
2) Bluetooth Module
C. Navigation Sensors Since a new microcontroller was required for
The 10 IR sensors on the K-Junior V2 were processing and interfacing with MATLAB, a
sufficient and suitable for obstacle avoidance. They separate Bluetooth module had to be used. The
have an appropriate detection range, of about 5cm module used is the Kootek BT2s Bluetooth Module
away from the robot. In addition, the team decided to [12] due to its ease of programmability to interface
use the Ultrasonic sensor available with the robot kit the microcontroller with MATLAB.
for traversing purposes and wall following [8]. The
entrances to the rooms were marked with black lines, IV. IMPLEMENTATION
and these lines would be detected through IR sensors In this section, the implementation of the robot is
on the bottom of the robot, so that it recognizes when discussed, including the block diagram of the system,
it enters a room and can start scanning for a fire. the map chosen and the hardware components. Seen
in Fig. 1 and Fig. 2 is the final robot implementation.
D. Fire Extinguishing System
As mentioned earlier in the Literature Survey A. Block Diagram of the System
section, the most common fire extinguishing system The block diagram of the system is shown in Fig.
for a small scale robot is a fan. This is because the 3. The direction of the arrow specifies the flow of
fire is being simulated by a candle and a fan can data; the component with the arrow head receives
extinguish that easily, also water can be avoided data from the other component. In addition, the
which can leak and damage the robot or any of the dashed lines that surround the Arduino Mega, K-
sensors attached. The fan used was a 12V DC but it Junior Robot and camera indicate that the
was run on a 9V battery. However, its strength was aforementioned components are separate sub-systems
more than sufficient to blow out any flames it mounted on top of the robot.
encountered.
 Digital
Compass
Interface
Circuit for
Fan

Wireless
Camera

IR
Sensors Fig. 3. Block diagram of the system

US B. Chosen Map & Environment

Sensor
The environment that the robot needs to traverse has
several characteristics that need to be taken into
Fig. 1. Photograph of final robot design (first angle)
consideration; the size, the material of the arena, and
the location of the candles within the environment.
The specifications that were followed in this project
Flame
Sensor
matched those used in the Trinity College
Firefighting competition [1]. The dimensions of the
Arduino Mega
with Bluetooth
arena are shown in Figure 4. The color and material
Module of the arena was wooden and the entrances of rooms
Blower had black tape to mark them. In the competition there
Fan was only one candle placed on the arena, but for this
project several candles were placed in different
rooms. Occasionally on different trials, random
obstacles would be places on the environment to
challenge the robot.
C. Movement & Navigation
The IDE for the K-Junior V2 robot already came with
some sample codes some of which were utilized for
this project after some tweaking, such as the
weightings of the sensors that were scaled up for this
project. For avoiding obstacles, the values of the
sensors on the robot are first obtained and a speed is
Fig. 2. Photograph of the final robot design (second angle)
then set.
As can be seen in Figure 3, the robot interacts with
the flame and infrared proximity sensors, as well as
the ultrasonic sensor. The flame sensor sends a
continuous analog signal to the robot that varies with
the detection of a flame and how far it is. The
ultrasonic sensor is used for wall following. The
robot decides if it is close enough to the flame and
sends a signal out to the fan, turning it on. The robot
also interacts with the Arduino Mega via an IO pin
telling it if it is in navigation mode or fire detection
mode. The Arduino Mega also receives data from the
digital compass and sends it, via the Bluetooth
module, to the Personal Computer for displaying the
current position of the robot on MATLAB. Another
system that does not interact with the robot but is
mounted within the system is the Camera which
provides a live feed displayed on an LCD Screen so Fig. 4. Dimensions of the arena as specified by the Trinity College
that fire personnel can observe the robot’s path. Firefighting Competition
This is achieved based on the following equation
provided with the IDE sample codes:
𝑊× 𝑅
𝑆𝑝𝑒𝑒𝑑 = + 𝑀𝑖𝑛 𝑆𝑝𝑒𝑒𝑑 (1)
𝐼𝑅 𝐺𝑎𝑖𝑛

Where W is the weighting of the sensors and R is the

reading of the sensors. The left and right speeds are
set by multiplying the sensor weightings for each of
the motors by the sensor reading.

TABLE I.
OBSTACLE AVOIDANCE SENSOR WEIGHTINGS
Array Sensor Left Right
Element Position Weighting Weighting
0 Left 8 -8 Fig. 5. The robot moving towards a flame
1 Front left 12 -12
2 Front -16 -20 E. Map Representation System
3 Front Right -12 12 The purpose of Bluetooth communication is
4 Right -8 8 purely for map representation on a PC, such that the
5 Rear 8 8
position of the robot on the map can be seen from a
bird’s eye view. To do this the Arduino Mega
Furthermore, to better traverse the arena, it was Microcontroller was used, in addition to a Bluetooth
decided to use a wall-following method within the Module and a digital compass.
code. The ultrasonic sensor provided with the robot kit
was installed on the right of the robot. The algorithm The digital compass reading is taken once the
used is as follows: every time the robot moves 15 robot starts its excursion and this is used as a
steps (a step is around 1 cm and 15 steps would be the reference angle. Any other angle taken afterwards is
full length of the robot), it checks the distance to the used to calculate how much the robot has diverted
right of the robot. If this value is greater than 55cm, from the starting angle. The robot then turns
then it indicates an empty space that the robot could
accordingly on the map shown on the MATLAB
go into and thus it turns right. The value of 55cm was
screen. Fig. 6 shows the control panel, where the user
used because that was the width of the corridor and
the robot should not turn if it is going down the establishes the connection, sets the starting position
corridor. of the robot on the map and then follows its
movement as shown in Fig. 7. The robot is
D. Fire Interactions represented by a blue circle and the yellow dot on the
The fire interactions are fire detection, sounding robot determines the direction it is facing. Since a live
an alarm and fire extinguishing. Before detecting the feed is provided, the user might see an object they
fire the robot has to move into a room. If the robot want to mark on the map. Therefor we have allowed
detects a black line, using the infrared sensors on the the user to drop pins to mark the map. Fig. 7 shows 2
bottom, it knows that it is about to enter a room. pins dropped, one for fire (in red) and the other for a
Therefore it moves into the room by taking 15 steps survivor (in green).
forward and then starts a 360-degrees-turn while
scanning for a fire using the flame sensor. This 360
degrees rotation is done at a low speed so that the
sensor has time to successfully detect a flame. Once a
flame is detected the robot will start moving closer to
it in an attempt to extinguish it with the fan as seen in
Fig. 5. In addition, a buzzer will turn on and four
LEDs will turn on and off.
As the flame sensor is detecting a fire the Blower
fan is turned on and produces an air stream while the
robot simultaneously moves slowly towards the fire.
Once the flame sensor doesn’t detect anything, the
robot does one last check by turning 360 degrees one
more time while checking for a flame. This is because
as the fan is producing an air stream to extinguish the
flame, it could be strong enough to make the flame
very weak but not fully extinguished. In this case, the
second 360 degree turn allows the robot to re-detect
the flame, moving forwards once more and attempt to
extinguish the fire. Fig. 6. The Map Control Panel for the robot
 Bluetooth system was going to be used, but as the
code got more complex and processing overhead
became a problem, the team resorted to a separate
Bluetooth module to interface the Arduino Mega with
MATLAB.

3) Digital Compass Operation

Once the digital compass was tested successfully
with Bluetooth and the Arduino Mega, problems
arose when the system was mounted on the K-Junior.
After some research was carried out, it seemed that
the digital compass’ readings were being affected by
the electromagnetic waves from other sensors and
devices near it on the robot. To solve this issue, a rod
was used to elevate the digital compass away from
any interference and get clear readings from it.
Fig. 7. An example of the pins that could be dropped on the map

F. Overlall operation
In Figure 8, the means of which the robot is
programmed and designed to behave is evident. Once
the robot is on, the obstacle avoidance algorithm is
also in play. The user then continuously checks if two
laps of the arena are completed when the robot
returns to its starting point, otherwise the robot
checks if a black line is detected (signifying the
entrance of a room). If it is not detected, it loops back
to the obstacle avoidance algorithm, otherwise it
moves beyond the black line by 15 steps and into the
room and starts scanning for a flame by moving
around 360 degrees. If a flame is detected, the alarm
and fan are turned on until the fire is distinguished. If
it is not distinguished, the robot moves closer to the
flame until it is. In the case there is no flame detected,
the 360 degree scan is completed while continuously
checking for a flame. Once it is completed, the robot
turns 180 degrees and exits the room.

G. Major Issues Faced

1) Accurate Robot Movement
The K-Junior V2 robot has an unsolvable problem
with accurate movement. This is because the motors
are not synchronized and have different speeds. The
best solution to make the robot move as straight as
possible was through obstacle avoidance and sensor
weightings. In TABLE I. it can be seen that the
weightings for the left and right are almost always the
exact opposite with the exception to the rear sensor
and the front sensor. For the rear sensor they are both
equal which makes sense since the sensor is neither
on the right nor the left. The same should be true for
the front sensor but it actually is not. The weightings
are both negative when it comes to the front sensor,
but one side has a higher value to try to compensate
for the skewered movement of the robot.

2) Establishing Bluetooth Commuincation Fig. 8. The flowchart of the basic algorithm of the robot
Another issue faced of great importance was with
the Bluetooth communication. Initially the K-Junior’s
4) Accurate Map Represenation VI. CONCLUSION
The final problem was the representation on the In this paper we have gone over the
map, since there was no pin point of the position to implementation of an autonomous firefighting robot.
send to MATLAB. The team decided to roughly The hardware components of the robot were a body,
estimate the speed of the robot through trial and error infrared and ultrasonic sensors for navigation, flame
and update it on the map to the best possible sensor for fire detection and a fan for fire
representation, which proved to be fairly accurate. extinguishing. There was also a live feed and map
representation provided to allow the journey of
V. EVALUATION extinguishing fire to be visible to humans and not just
for the robot, in case any abnormalities occur.
A. Results and Analysis
TABLE II represents the information collected There is plenty of room for expanding over this
after numerous trial runs of the robot. It is important project. For instance, allowing the robot to navigate
to note that the time taken for the robot to completely through different floors as houses or buildings on fire
traverse the arena means that the robot has completed will usually be more than one level. Additionally, a
2 sequences around the arena and returned to its space for survivors could be added along with human
starting position. 2 rounds were chosen as the robot detection through image processing to identify those
will start to repeat the same behavior again and this is survivors and carry them to safety.
redundant. The average time taken for the robot to
complete trials came to 7 minutes and 48 seconds. REFERENCES
[1] Trinity College. Trinity College Robot Contest. Retrieved
TABLE II. Novemeber 2014, from http://www.trincoll.edu/events/robot/
RESULTS FROM TRIAL RUNS
[2] Playne, Matthew. "Fire-Fighting Robot." Industrial Systems
Time Fires Room Room Room Room Design & Integration. (2006): n. page. Web. 10 Nov. 2013.
Trial
Taken Extinguished 1 2 3 4
[3] Davis, Joe. "FIRE FIGHTING ROBOT." (2010) : Web. 10
1 8.32.18 3/3 Yes Yes Yes Yes Nov. 2013.
2 8.20.77 1/1 Yes No Yes Yes
[4] Mobley, Clint, and Aaron O’Connor. "Trinity College Fire-
Fighting Robot Competition." Evansville. N.p., n.d. Web. 28
3 6.34.68 4/4 Yes Yes Yes No Oct 2013.
[5] Sharma, Aman. "A FULLY AUTOMATED FIRE
FIGHTING ROBOT." Irnetexplore. Dronacharya College of
Engineering. Web. 28 Oct 2013.
Furthermore, on average 3 out of the total 4 rooms
are regularly visited on each run and this is due to the [6] Xu, James. "Industrial Systems Design and Integration." Fire
Fighting Robot. (2009): n. page. Web. 10 Nov. 2013, 1989.
fact that robot is affected by various factors which
[7] Gupta, Rishi, and Sandeep Saini. "Fire Righting Robot."
include obstacles, time taken to extinguish fires and gobookee. N.p.. Web. 28 Oct 2013.
the number fires in the arena. The fires extinguishing [8] K-Team Mobile Robotics. n.d. K-Junior V2 Super
proved to be 100 percent accurate as this is the most Educational Pack. [online] Available at: http://www.k-
important factor in the entire project and thus it was team.com/mobile-robotics-products/k-junior/edu-pack
focused on. It is worth noting that the robot does not [Accessed: Oct 2013].
know how many flames there when it is started. [9] DFRobot. n.d. Flame Sensor. [online] Available at:
http://www.dfrobot.com/index.php?route=product/product&
B. Map Representation on MATLAB product_id=195#.UsRyqPQW2So [Accessed: Dec 2013].
The 2D map produced on MATLAB was [10] TaoBao. n.d. [online] Available at:
http://item.taobao.com/item.htm?spm=a1z10.5.w4002219218
relatively accurate in comparison to that of the actual 8891.12.yAgGQA&id=10352669854 [Accessed: Dec 2013].
positioning of the robot on the wooden arena thus [11] ITead Studio. (2011, May). Digital Compass GY-26 Module.
proving successful. Retrieved November 2014, from
http://imall.iteadstudio.com/im120712009.html
[12] Virtuabotix. (2011). Virtuabotix BT2S Bluetooth to Serial
Slave for Arduino & More! Retrieved November 2014, from
https://www.virtuabotix.com/product/bt2s-bluetooth-serial-
slave-arduino-versalino-microcontrollers/