2016 5th International Conference on Informatics, Electronics and Vision (ICIEV) 909

Design and Fabrication of an Autonomous Fire

Fighting Robot with Multisensor Fire Detection
Using PID Controller

Tawfiqur Rakib M. A. Rashid Sarkar

Department of Mechanical Engineering Department of Mechanical Engineering
Bangladesh University of Engineering and Technology Bangladesh University of Engineering and Technology
Dhaka, Bangladesh Dhaka, Bangladesh
[email protected] [email protected]

Abstract— Recently, Multisensor Fire Detection System general interface for supervised computer was designed in his study.
(MSFDS) is one of the important research issues. Here, a fire Viguria et al. [12] built an aerial/ground robot team applied to fire
fighter robot is fabricated providing extinguishment platform. detection. They used a disturbed market based algorithm, called S+T
The base of the robot is made of the wood of ‘Rashed tree’, and coordinated in between aerial and ground vehicles. Their
locally known as ‘Kerosene wood’. There is about 1 liter water simulations showed that if the number of services increased,
reserving capacity. An arduino based simple algorithm is used communication and energy requirement would also increase. Khoon
for detection of fire and measurement of distance from fire et al. [13] described the development of an Autonomous Fire
source while the robot is on its way to extinguish fire. When the Fighting Mobile Platform (AFFMP) that is equipped with the basic
fire is detected and the robot is at a distance near to fire, a fighting equipment that can patrol through the hazardous site via a
centrifugal pump is used to throw water for extinguishment guiding track with the aim of early detection for fire. They used
purpose. A water spreader is used for effective extinguishing. It is flame sensor, motor driver circuitry and LDR sensor and kept the
seen that velocity of water is greatly reduced due to the use of extinguishing as simple as possible. For early warning, a vision
water spreader. Two sensors: LM35 and Arduino Flame Sensors sensor based fire detection method is proposed by Ko et al. [14].
are used to detect the fire and distances on its way towards fire. They developed an AVM classifier for fire pixel verification.
Sensitivity of these sensors at different day times and distances is
Kim et al. [15] made a portable fire evacuation guide robot system
tested through analog reading of the serial monitor.
and demonstrated that robot system can be thrown into a fire to
gather information, locate displaced people and evacuate them. They
designed the robot with aluminum compound metal for thermal
Keywords-Multisemsor; Arduino; Rashed tree; Kerosene wood;
resistance with waterproofing and an impact distribution frame for
Algorithm; Detection; Extinguishment.
impact resistance. White et al. [16] developed a vehicle mounted fire
I. INTRODUCTION fighting system and included a series of flame and heat retardant
coverings placed on all exposed parts of the system to prevent
Industrial development in Bangladesh has become a key figure to damage from exposure to extreme heat.
obtain progress and to cope up with developed countries. But
recently, the development of industries has faced serious difficulties Our effort is to develop an autonomous fire fighter robot which is
due to accidental situations like fire havoc. Prevention of and constructed by locally available fire resistant and water-proof
intervention in such scenarios are the most talked topics in the field materials and performs on an arduino based fire detection and
of technology and science. Sometimes it is almost impossible to extinguishment algorithm. The robot is also fabricated so that it can
intervene in a these situations without the help of robot. Because save itself from fire by keeping safe distance from the source. At
robot has some size and sensing advantages that human lacks of. It is different distances from the fire and at different day time, the
not feasible to appoint a person to patrol for accidental fire where a performance of the robot is evaluated by performing sensitivity tests
robot can do the patrolling [1]. So a robot is made to patrol for fire on the sensors taking serial monitor readings in Arduino.
detection and early warning in domestic, industrial and
environmental cases [2-5]. II. FABRICATION OF MECHANICAL COMPONENTS
All the materials used in the fabrication of the robot were found
This robot may be used in industrial environment and even in
locally. It is the purpose of the design to demonstrate a fabrication
household areas where there is more probability of occurring
process of Fire Extinguisher Robot that can perform safely.
accidental fire [6-8]. Different sensors are used and fusion of their
performances is ensured by an intelligent algorithm in Arduino A. Making of the platform:
computing platform or by soft computing techniques [9-10]. Su et al.
For the purpose of making the base light, “Kerosene wood” is
[11] made an automatic fire detection system using Adaptive Fusion
used. This is famously known as th the wood of “Rashed tree”. The
Algorithm (AFA). Multisensor Fire Detection System (MSFDS)
density of this wood is about 370-400 kg/m3. So it can provide more
along with Visual Basic to receive information was used and a

978-1-5090-1269-5/16/$31.00 ©2016 IEEE

 910
`
surface area with light weight. Its thickness is 1 inch. It also has good III. CONTROL AND SENSOR UNIT
strength and shock resistance capacity. It is also worth mentioning
that this wood is also used to make the base of boat for its water A. Control Unit:
resistance property. The dimension of the platform is 12×7×1. The logic control unit of the robot is solely an Arduino Uno.
Arduino Uno provides an Integrated Development Environment
B. Mobility of the robot: (IDE) which provides a base of programming for the C and C++
This is a high functioning mobile robot. For the mobility of the languages. The Uno is a microcontroller board based on
robot, two wheels made of Nylon and a caster ball is used. This is the ATmega328P. It has 14 digital input/output pins (of which 6 can
mainly a rear wheel drive type of vehicle. Two motors are mounted be used as PWM outputs), 6 analog inputs, a 16 MHz quartz crystal, a
with the nylon wheel by two separate shafts. The wheels are of 3 inch USB connection, a power jack, an ICSP header and a reset button.
diameter and 0.8 inch thickness. The motors’ movement is controlled The input/output pins are connected in different ways with the motor,
by a motor driver (L298) which responses to the signal from arduino. pump and sensors through jumper wires and breadboard. In this way,
The motors have high torque and speed. These are locally made it establishes control over each component of the robot and the robot
motor which do not give the same speed even when unloaded. The does the job as per the programme induced in the Arduino Uno.
robot also seems not to have uniformly distributed weight. So, an The power supply control unit is solely a LiPo battery. To power
algorithm is established which is programmed to control both motors the motor, pump and the whole circuit a Lithium polymer battery is
separately. An extreme calibration is then done to match the speed of used. It is a 2200 mAH and 12V battery. It is a rechargeable
both motors so that they can go straight in a specific direction. The battery of lithium-ion technology in a pouch format. This type of
use of caster ball is made for simplicity and its low cost. The robots’ batteries is lighter but also less rigid. To power the arduino, the
movement is mainly controlled by rear wheels. Caster ball is used for voltage of the battery is stepped down to 8V by 7808IC. Because
front support providing a flexible mobility. arduino operates best at this voltage. Motors and pump selected for
C. Water spreading Mechanism: the robot can be best powered at 12V. So choosing a 12V LiPo
battery was efficient.
The water container has the capacity to contain at least 1L water.
It is made of strong cardboard which has water resistant property. B. Sensor unit:
The pump is set with the container by screw and glue mechanism. Multisensor Fire Detection System (MSFDS) is used for the
The glue used here is water proof. A long aluminum pipe is fitted
detection and extinguishment purpose. This system includes
with the pump. The pipe is about 2 feet long because water should be
thrown by keeping a safe distance from fire. A locally made water the use of more than one sensor at a time and also makes the
spreader is used at the front of the pipe to spread the water. The use collaboration of the sensors. There are two highly sensitive
of the spreader reduces the velocity of water from the pump greatly sensors connected to the Arduino Uno: Arduino Flame Sensor and
but spreads the water effectively for extinguishment purpose. So a LM35.
safe distance from the source is maintained by making the pipe long.

Figure 1: Labelled Design of Mechanical Body (Designed in SolidWorks 2013)

 911
`

Figure 2: Original Prototype of Autonomous Fire Fighting Robot.

The Arduino flame sensor is very sensitive to IR wavelength at The LM35 is a precision integrated-circuit temperature devices
760 nm ~ 1100 nm light. These types of sensors are used for short with an output voltage linearly-proportional to the Centigrade
range fire detection. It has four pins. A) VCC: Positive voltage input: temperature. This sensor gives a proportional value of the
5V for analog 3.3V for Digital, B) A0: Analog output, C) D0: Digital temperature as an output. When the robot moves in the direction of
output, D) GND: Ground. It takes the input from the environmental the flame, LM35 identifies significant temperature rise near the
data and gives an analog output. If there is a significant change of IR source of the flame and changes the output. The robot is programmed
wavelength, it detects as a flame and increases the analog output. to stop when this output changes significantly. Since LM35 has to
Then the robot is programmed in such a way that it stars to move in come very near to the source for heat detection, two measures of
the direction where the flame is detected. safety has been taken. One, LM35 sensor is connected with the
control unit through a long jumper wire inside a pipe of about 10 inch
long. So, the sensor maintains a considerable distance from the main
body. Two, when it detects temperature rising, the motor stops and
within a fraction of second, it gets back to a safe and accommodated
distance to throw water.

IV. WORKING PRINCIPLE

The working principle is very simple. At the beginning of
operation, the robot rotates on its own axis. Arduino will control the
motor through the motor driver. To rotate, one motor will remain off
and other one will move, thus resulting in rotation of the body. The
robot does this rotation to search fire at different corners of the room.
Eventually Arduino flame sensor will detect the fire if there is any.
Then the robot will move forward to the fire slowly. Here, algorithm
is needed to make a slow approach of the robot towards fire. Because
the robot needs to stop at a certain limit and it definitely does not
want to mess up with the fire. The LM35 sensor senses the heat. If it
finds a remarkable increase in the temperature, it stops and moves
back slightly to accommodate extinguishing. Then a centrifugal
pump throws water at fire to put it out. A unique mechanism is used
to enable spreading of water.
Figure 3: Connections of sensors with control unit.
 912
`
The flowchart of the whole operation is given below: direction of fire. At 3 inch distance intervals, the readings are noted.
When the robot comes near the source, temperature rises and the
output of LM35 changes which is also recorded. The experiment was
Arduino Sending signal to done on 22nd December, 2015. The average temperature of the day
uno. motor through motor was 26 degree Celsius.
driver to rotate.
VI. RESULTS AND DISCUSSION
A. Sensitivity of flame sensor at day and night:
Sensitivity of Arduino Flame Sensor is tested at day and night
without any fire and from 3 feet distance away from the fire source
Moving Arduino flame and the results are demonstrated in Figure 4 and 5 respectively. It is
towards fire. sensor getting signal seen from Figure 4 that at 7 o’clock, the sensor gives higher output
of fire flame. with or without fire source. But as the time goes on, the reading
keeps falling. This is because of the presence of infrared rays in the
sunlight. Nearly at 12-13 o’clock, the intensity of sunlight increases
to the highest of the day and thus increases the availability of infrared
rays to which the flame sensor responds. So, the output reading is the
lowest at this time of the day with or without fire source. But after 15
o’clock, the sun’s intensity starts to fade away and as a result the
LM35 getting Stops and relay switch output reading starts to increase. It reaches the maximum value at the
signal of makes the pump on end of the day when the Sun sets off. It is also observed that
temperature throughout the day, the serial reading profiles almost maintained
rising. constant offset and almost parallel patterns were found.

Fire extinguishing by
water

Flowchart 1: Working Principle of Fire Fighter Robot

V. STUDY OF PERFORMANCE
A. Problem Specification:
Arduino Flame Sensor and LM35 are voltage sensitive sensors.
They can take the input of 0-5V. These voltages are changed into an
analog reading (0-1023) in Arduino serial monitor reading. This
conversion gives us a better range and a good variation in the output.
So, the robot can be programmed at different circumstances. When
the flame sensor identifies the flame, a change in the analog output is Figure 4: Effect of sensitivity of flame sensor at day
obtained. This sensor has different sensitivity at different daytimes.
So, the performance of the robot is varies with time. LM35 also
responds to the temperature change and changes the analog output
when the robot is near the source. Here, Sensitivity is the output
reading obtained from the serial monitor. The main aim of our study
is to check the sensitivity of flame sensor at different daytimes and at
different distances with or without fire source and to observe the
response of LM35 sensor. In this way, our study shows the best and
worst possible situation of the day, the fabricated model of fire
fighter robot has to handle.
B. Test Procedure:
The test is done to check the sensitivity of arduino flame sensors
and LM35 at different day times and at different distances from the
fire source. Different test is done placing a candle at different
distances from the arduino flame sensor. The best stable reading
obtained at the serial monitor analog output is at 3feet away from the
fire source. From that distance, readings are taken with or without the
fire source at different daytime with half hour interval. When the
flame sensor detects the fire, the robot starts to move towards the Figure 5: Effect of sensitivity of flame sensor at night
 913
`
In Figure 5, it is evident that the readings are almost similar robot is programmed to stop and on the process, it makes the
throughout the night with and without the fire source. As the effect of necessary extinguishment.
Sun ceases to exist around 23-03 o’clock, the output reading without
the fire source is the highest and it becomes the lowest with the fire Table 1. Change of output readings of LM35
source. So the gap between these two readings is highest in this
portion of the night. This range of time gives the most adequate Temperature (°C) Serial Reading(AO)
difference for the proper distinguishment between fire source and 26 55
normal condition. So it is fact that flame sensor detects fire
irrefutably without any inconvenience at night. 40 88

B. Change of output of flame sensor when the robot moves 55 123

towards fire:
Figure 6 shows us the effect in output of the flame sensor at
10AM and 10PM when the robot detects a fire and starts moving VII. CONCLUSION
towards the source. When it starts moving towards the source, the The Fire Fighting Robot is fabricated with locally available
analog output will decrease because of the increase in the intensity of materials and some tests are done to observe its effectiveness at
the source. It is also observed that the readings at 10PM are higher different situation. As the Fire Fighter Robot has to endure different
than that at 10AM at the highest distance away from the source. The situation, this effectiveness test will help us to make a better model.
fall of the reading is steeper at 10PM than at 10AM. Because at Addressing to the pertinent design of the robot and the tests,
10AM, the infrared rays from the Sun disturbs the analog reading of following closing remarks can be drawn-
the sensor. But it is also found that near the safe possible distance to
fire source, the readings at both day and night are almost same.  The platform of the robot is made of locally available wood
Because the effect of the fire source becomes dominant. It is also of ‘Rashed Tree’. It has enough strength and the capacity of
observed that at near distance to the fire source, the curve becomes resistance to fire and water. Its low density also helps to
equally steeper for both the cases. Because there are increased make a less weighted structure with a firm support.
infrared rays with the increase of the intensity of the source. So, there  The water container is made of white cardboard which is
is a rapid attenuation for the output reading in case of 10PM and in waterproof. This provides resistance to water leakage.
the case of 10AM, the attenuation is slight.  The pipes are made of aluminum alloy which is fire
resistant. These carry water at a safe and calibrated distance
to accommodate extinguishing and another pipe is also
used to carry LM35 sensor to detect heat at a measured
distance.
 The water spreader at the front of the pipe reduces the
velocity of the water thrown by the pump greatly but
spreads the water for effective extinguishment.
 In presence of daylight, the analog output readings are the
least. But it provides us a nearly constant difference of 300-
350 between the readings of the output with or without the
fire source. So, an efficient algorithm can be made at this
circumstance even with the worst possible output reading.
 In day, sunlight is absolutely a dominant factor to control
the values of the output reading.
 At night, the readings have quite large value without fire
source and the least value is obtained without the presence
of fire. This provides a good range of detection.
 The best possible range for detection of fire source is from
23-03 o’clock. This provides the best platform to program
the device during any time of the day.
Figure 5: Change of output of flame sensor when the robot starts moving  After detection of fire, the robot starts to move towards the
towards fire Source. source. With the change of temperature, the output of
LM35 is changed.
C. Change of output of LM35:
 This change of output of LM35 is non-linear. So, at
When the robot starts to move towards the fire source, different distance, the readings are taken and the robot is
temperature starts to rise. LM35 is a sensor which can detect heat at a calibrated to stop at safe distance from the source.
very near distance. Even sometimes for small source of fire, it has to
The Fire Fighting Robot is effective enough to fight against fire on
come in contact of the fire source to detect the existence of fire. The
a small scale. It can sense fire flame better at darker places. It is made
tricky design of the robot allows the LM35 sensor to detect heat at
as a preventer robot. Because it can detect fire instantly and can
least at a distance 6 inch away from the source. Because the sensor
extinguish it before spreading. This multisensory based robot may be
remains even at more close to the heat source as it is placed at the
a solution to all fire hazards. With enough funding and scope, this
front side of the robot through a pipe. It also allows the robot to keep
design of robot can also fight against large fire with larger reserving
itself at a safe distance. So, LM35 detects temperature rise and
capacity and an improved sensing unit can provide even an earlier
changes the output. When the change crosses an irrefutable range, the
detection of fire at all circumstances.
 914
`
REFERENCES [9] J. H. Kim, B. Keller and B.Y Lattimer, ”Sensor fusion based seek-and-
find fire algorithm for intelligent firefighting robot,” Advanced
Intelligent Mechatronics (AIM), 2013 IEEE/ASME International
[1] I. Doshay, "Robotic fire protection system," U.S. Patent No. 6,364,026, Conference, pp. 1482-1486, July 2013.
April 2002. [10] E. Mahdipour and C. Dadkhah, "Automatic fire detection based on soft
[2] E. P. Krider, R. C. Noggle, A. E. Pifer and D. L. Vance, "Lightning computing techniques: review from 2000 to 2010," Artificial
direction-finding systems for forest fire detection," Bulletin of the Intelligence Review, vol.42, no. 4, pp. 895-934, Dec 2014.
American Meteorological Society vol. 61, pp. 980-986, September [11] K. L. Su, "Automatic fire detection system using adaptive fusion
1980. a algorithm for fire fighting robot," Systems, Man and Cybernetics, 2006.
[3] M. Hefeeda, and M. Bagheri, "Wireless sensor networks for early SMC'06. IEEE International Conference on, Vol. 2, pp. 966-971,
detection of forest fires." Mobile Adhoc and Sensor Systems, 2007. October 2006.
MASS 2007. IEEE International Conference on, pp. 1-6, October 2007. [12] A. Viguria, I. Maza, and A. Ollero, "Distributed service-based
[4] S. L. Rose-Pehrsson, S. J. Hart, T. T. Street, F. W. Williams, M. H. cooperation in aerial/ground robot teams applied to fire detection and
Hammond,D. T. Gottuk, M. T. Wright and J. T. Wong, ”Early warning extinguishing missions," Advanced Robotics, vol. 24, no. 1-2, pp. 1-23,
fire detection system using a probabilistic neural network”, Fire January 2010.
technology, vol. 39, pp. 147-171, April 2003. [13] T. N. Khoon, P. Sebastian, and A. B. S. Saman, "Autonomous fire
[5] R. C. Luo, T. Y. Hsu, T. Y. Lin and K. L. Su, "The development of fighting mobile platform," Procedia Engineering vol. 41, pp. 1145-
intelligent home security robot." IEEE International Conference on 1153,December 2012.
Mechatronics, 2005. ICM '05, pp. 422-427, July 2005. [14] B. C. Ko, K. H. Cheong and J. Y. Nam, "Fire detection based on vision
[6] T. L. Chien, H. Guo, K. L. Su and S. V. Shiau, “Develop a multiple sensor and support vector machines," Fire Safety Journal vol. 44, pp.
interface based fire fighting robot,” Mechatronics, ICM2007 4th IEEE 322-329, April 2009.
International Conference on, pp. 1-6, May 2007. [15] Y. D. Kim, Y. G. Kim, S. H. Lee and J, H. Kang, "Portable fire
[7] K. L. Su, L.T. Chien, and Jr H. Guo, "Design a low cost security robot evacuation guide robot system."Intelligent Robots and Systems, 2009.
applying in family," International conference on autonomous robots and IROS 2009. IEEE/RSJ International Conference on, pp. 2789-2794,
agents. Palmerstn North, New Zealand. pp. 367-372, December 2004. October 2009.
[8] R. C. Luo, T. Y. Lin, and K. L. Su, "Multisensor based security robot [16] W. R. White, "Vehicle mounted fire fighting system." U.S. Patent No.
system for intelligent building," Robotics and autonomous systems, 5,836,398, November 1998.
vol. 57, no. 3, pp. 330-338, March 2009.