,(((,QWHUQDWLRQDO&RQIHUHQFHRQ5RERWLFV$XWRPDWLRQ$UWLILFLDOLQWHOOLJHQFHDQG,QWHUQHWRI7KLQJV5$$,&21

1RYHPEHU'HFHPEHU'KDND%DQJODGHVK

Designing and Optimization of An Autonomous

Vacuum Floor Cleaning Robot
H.A.Shakhawat Hossen Prayash1 , Md. Ragib Shaharear2 , Md. Farhanul Islam3 , Saiful Islam4 , Noushad Hossain5 ,
and Shamik Datta 6
School of Electrical and Electronic Engineering,
Shahjalal University of Science & Technology, Sylhet-3100, Bangladesh.
habib.prayash@gmail.com1 , shaharearshuvo@gmail.com2 , md.farhanulislam@gmail.com3 , m.saiful.i.d@gmail.com4 ,
sourav1023@gmail.com5 , eeeshamik@yahoo.com6 .

Abstract—With the advent of Automation, Robots are seeing Since the late 1990s, better developed cleaners equipped with
widespread use in Industrial manufacturing, Automated retail, limited suction power have been developed. Then the footstep
Surveillance and Security. In the same breath, home automation of iRobots Roomba vacuum cleaner in 2002 has been followed
has also evoked interest among different user groups. Currently,
high-priced floor cleaning robots produced industrially are beyond by many other companies launching similar products. In early
the purchasing capability of normal households in third world 2008, Samsung made the first robotic vacuum cleaner which
countries. In this paper, we have introduced an initiative to build a maps its environment and systematically navigates in a home[5].
cost-effective floor cleaner robot. Using a vacuum cleaning system, Several other products have followed since then including LGs
we designed and implemented a cleaning robot. The paper has RoboKing or Neato Robotics laser-equipped vacuum robot.
described how an operational floor cleaning robot prototype has
been developed along with the techniques of the indoor mapping Currently some prominent brands cleaning robots are irobot,
and positioning system. Raspberry Pi and Arduino Mega were used Neato and bobsweep. Tech-giants like Hitachi, Siemens and
in unison as the processor of the robot. This robot can map and Cyberworks are producing robotic cleaning systems which are
clean a certain area of a building floor autonomously. One novel fairly costly[6]. While less renowned start-ups like iRobot,
contribution of this work is the introduction of a GPS module to dyson, eufy have got success by producing cheaper cleaning
the Raspberry Pi so that the robot can be aware of its location
and move in the right direction. Human assistance is only required robots[7].
for waste disposal and maintenance. Our future goal is to add a So, an initiative is taken to develop a low cost cleaning
reliable autonomous waste disposal feature in the prototype. robot which is affordable to common households. Total cost
Index Terms—Cleaning Robot, Floor Cleaner, LIDAR, Wall Fol- of building our cleaning bot is $655. Our prototype robot can
low, IR Proximity Sensor, Raspberry Pi 3, GPS, Vacuum Cleaning, clean dust and small size waste objects efficiently owing to
Obstacle Avoidance, Magnetic direction Sensor.
its smart dust cleaning algorithm which is something new in
I. I NTRODUCTION the cleaning bot technology. When the battery is down, it can
automatically go to the nearest charging hub and charge itself
The last few decades have experienced significant influx of
up. A mapping and tracking system is available in this prototype
robotic appliances used in household activities. Floor cleaning
robot to increase its efficiency. It creates a map of the designated
is one of the most taxing and time consuming chores. And this
area and generates the cleaning path according to the map.
is where automation comes into play, taking over jobs where
Finally, the area is cleaned with the help of indoor positioning
manual effort is necessary. Rising level of pollution,increased
system. Also it has a system to detect and avoid real-time objects.
level of consumption goods followed by humongous amounts
of garbage have necessitated a meticulous cleaning of our The paper is arranged as follows: A brief description of the
surroundings in a quotidian manner. Floor cleaning Robot is the prototype is given in section II, System assembling process
best possible technical solution to do this and thus makes life and Software integration are described in section III and IV
more easier[1]. respectively. In section V details of the genetic algorithm are
A vacuum cleaner is an electro-mechanical appliance com- discussed, while the conclusion comes in section VI.
monly used for cleaning floors, furniture, rugs and carpets by
suction .Nowadays, consumers are paying appreciable attention
to Robotic Vacuum Cleaners (RVC). The market size had al-
ready grown to 350 million dollars in 2005[2]. While in 2014,
II. D ESCRIPTION OF S YSTEM
The global family vacuum cleaners market size was valued
at USD 11200 Million[3]. A new market report published by
Transparency Market Research claims that the global house-
hold vacuum cleaners market is expected to reach a value of The primary purpose of the cleaner robot is to clean the floor
US$18,936.9 Mn by 2026[4]. A high demand for fitting yet dirt from the different places of a building. To clean the floor,
affordable dust cleaning appliance is driving this market growth. some mechanism or process is required. Here in the prototype,
A fully automated cleaner bot can save both time and money. a vacuum cleaning process is introduced for cleaning.

‹,(((


Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 06,2020 at 06:26:43 UTC from IEEE Xplore. Restrictions apply.
 Arduino position of the bot. A suction pipe from the vacuum cleaner is
Mega & placed at the center so that it can easily suck up the dirt. The
Sensors vacuum cleaner is mounted in the middle of the main body. The
sensors are mounted in front of the main body so that any kind
of obstacle can easily detect. The robot looks like a cylinder
GPS Raspberry having 7 inches in radius and 1.5 feet in height.
LIDER There are two packs of ten cells of 3.7V and 3600 mAh Li-
Module Pi
ion batteries where 10 cells are connected in series making 37V
and two packs are connected in parallel to increase the current
Motor rating. By connecting these two pack batteries in parallel, we
Driver & got 7200mAh=7.2Ah. The input power rating is (7.2Ah*37V
Vacuum = 266.4Wh). The power consumption of vacuum cleaner is
Cleaner 30W. We used four DC gear motors. The power consumption
of these motors is 72W. So, the total power consumption of the
Fig. 1: Block Diagram of the System cleaning Robot is 102W. The input power rating is (7.2Ah*37V)
266.4Wh. When the battery is fully charged, this robot can clean
The custom made vacuum cleaner was constructed with a DC continuously for 3.7hours on battery.
motor. This motor is mounted inside the body of the vacuum There is an autonomous process for auto recharge of the
cleaner to make it easy to suck the dirt up. The mounted DC battery namely ’auto docking and charging process’ which is
motor rotates an omni-directional fan that creates a vacuum and necessary to make the robot auto rechargeable. It can be line
intakes the dirt with air inside. back for the charger if it runs low on battery power, even if
The mechanical structure is constructed by using an acrylic its in a distant room. And most importantly, while cleaning
glass which is actually Polymethyl methacrylate also known as floor if the battery runs out, it stores the current location in
Plexi-glass. The plexi-glass is cut according to the requirement the memory, goes to the docking position and recharge and then
of the structure. The structure is made by making a shaped like remembers the last location and start from where it left off.
a circle. There are three decks one over another so that the This solves a significant limitation of the traditional cleaning
components can be mounted on each of the decks. In the lower robot refers that the traditional robot can go back to the charger,
part of the body, the suction cap is mounted below. but it doesnt know how to resume from where it stopped for
charging as it doesnt have a map. The lithium-ion battery pack
of the robot lasts at least for two hours, which is sufficient, but
full autonomous operation including recharging means that the
battery life doesnt matter.
III. S YSTEM A SSEMBLING P ROCESS
The cleaner robot is a combination of different electrical
and mechanical components such as- Raspberry Pi (model
3B), Arduino Mega, servo motor, GPS module, LIDAR, motor
driver, relay module, vacuum cleaner, proximity sensors, wires,
buck module, diodes, mechanical structure etc. At first, these
components are assembled in the mechanical structure than the
electrical connections were set up.

Fig. 2: Schematic Diagram of Electrical Unit.

The cleaning system includes another two DC geared motors,

two rotating brush, sprockets for moving chain from geared
motor to rotating brush and aluminum sheet for supporting
vacuum cleaner. This DC geared motor has been fitted on the
acrylic sheet with metallic element holder and sprockets put in Fig. 3: An image of fully functional cleaning robot
with it which have been fitted into the shaft of the motor. All
components are installed on the lower part of the acrylic sheet. In the first stage, Power is given to the buck modules from
Operation of the rotating brush is to move the dirt to the center battery through diode and a kill switch. At the starting LiPo



Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 06,2020 at 06:26:43 UTC from IEEE Xplore. Restrictions apply.
(Lithium polymer-ion) battery is used for testing purpose, then Start
when the robot is ready, rechargeable Li-ion pencil batteries
are used for power supply. Diodes are used to avoid reverse
connection problems. Three buck modules were used. One is System Check
for empower the vacuum which has 30V output and another
two has 5V and 24V output respectively for sensors, switching
relays and motors. We use 30V terminal voltage to the vacuum Start
because lower than 30V can not create enough suction to clean Everything No Blinking
the small particles of dust. From the 5V output all the sensors, Connected? LED for
Arduino and Raspberry Pi get the power. As mentioned above Maintenance
various types of sensors are used in the cleaning Robot, so there yes
are buck modules to provide huge amount of currents. For this
Move Sensors to
reason high power buck modules are used to serve this purpose.
the Default
Four proximity sensors are used for detecting near obstacles,
Position
although LIDAR is used to have good mapping but for instant
Manual
appearance of any obstacles we use proximity sensors. Proximity
modification
sensors are connected to the Arduino input pins. As mentioned
and
above LIDAR is used for mapping the surroundings[8], it is
Approval
also connected to the Arduino input pin. Four servo motors are
used in the bot. One is for charging dock and another three are
for moving the LIDAR in 360 degree space and all these servo Not Found
Map Checking Create Map
motors are connected to the output of the Arduino. Arduino
takes the inputs from the Proximity sensors and LIDAR, it
Found
also sends command to the servo motors for moving LIDARs
according to the Raspberry Pi’s command at the same time when Generate
taking data from LIDAR. An output of the Arduino is also the Path for Check Check
connected to the relay module to turn on and off the vacuum Cleaning current current
cleaner. Motor driver’s input is connected to the Arduino output position position
pins to control the speed and direction of the motors. Arduino
is connected to the Raspberry Pi for communication between Ring the
sensors and Raspberry Pi. A GPS module is also connected to Buzzer
the Raspberry Pi so that the robot can be aware of its location. for 30s
All the connection of the system were made by connecting wires.
Check for Non- Present
IV. S YSTEM I NTEGRATION (P ROGRAMMING ) Stop
Static Object
To acquire data from the devices and sensors e.g. GPS, LI-
DAR, IR proximity sensors and Ultrasonic Sensors and process Not Present
them with necessary instructions, a software is designed. The
Start Auto
Complete
Check Battery Low docking and
charging
Condition charging
process
process
OK
Start Cleaning
and Move to the
Next Position

Check the
No End Position

yes
Fig. 4: Robot’s Cleaning Path (Generated). Stop Cleaning

sensors are used to perform different tasks. They all require

different type of instructions. End
Fig. 5: Flowchart of th System’s working process (Simplified)
 The software was implemented on Arduino mega and Raspberry

Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 06,2020 at 06:26:43 UTC from IEEE Xplore. Restrictions apply.
Pi to make the robot autonomous and fully operational. V. A LGORITHM D ETAILS

Raspberry Pi works as central processing unit. Arduino mega

is used as a sub-system to read the analog data from IR proximity
sensor and Ultrasonic Sensor and send it to the main processing The algorithm is based on Genetic Algorithms (GA) which
unit. Then the Raspberry Pi processes the data and performs is done is several steps to find the correct path[9]. Each gene
necessary tasks. represents the robot position and some of the chromosomes
represent also the mini-path. In addition, this algorithm helps
Mapping is an important task for the cleaning robot. It is very the robot to pass through every part of the room by avoiding
fundamental to the cleaner robot that it possesses the entire map obstacles using different sensors. The algorithm can be divided
of the area that it has to clean, a more specific map of rooms into several parts.
and the possible location of obstacles that it has to avoid and its 1) Modeling of the room: In this step, a map of the room is
current location. We used a GPS module and LIDAR unit to get created using different sensors and camera data. The positions
the map of the building floor and guide the robot accordingly. of the obstacles are also determined and stored in the robots
Specifically, the LIDAR unit is used for room mapping and memory, which used later for planning the path. The model of
GPS module to guide the robot to the right direction. LIDAR the room always updates during the run time based on different
measures the distance of the object from the robot with very sensor and camera data.
good accuracy. By spinning the LIDAR unit, systems can get
the detailed distances required for mapping. When mapping was 2) Path planning: The path which is covered by the robot
completed, the location of the docking and charging position from a start position to the position with passing over all the
was known by the robot and store it in the memory. accessible positions and avoiding obstacles that are named as
This map can be edited manually by operator if there is anything the global path. The global path is divided into several mini
wrong. Then the Raspberry Pi generates the cleaning path paths. The mini path is always shorter than the range of sensors.
according to the map using genetic algorithms. The Genetic Algorithms helps the vacuum cleaner robot to
After completing the mapping and path generation process find associate degree economical path and clean all accessible
the cleaner robot start its main mission of sweeping every areas from a begin position to an exit position. It consists of
accessible area in the entire environment avoiding obstacles. several steps which are shown in the flowchart below.
During cleaning process the cleaning robot must face obstacles
like walls, tables, chairs, doors and moving people. Infrared 3) Fitness function and selection: A fitness function is nec-
proximity sensors and ultrasonic sensors are used for obstacle essary to know the detailed description and solution of the
detection. A proximity sensor can detect the presence of nearby problem. It is directly related to the constraints of navigation
objects without any physical contact. But the range of object coverage. To evaluate the fitness of individual mini-paths three
detection is small, thats why we used another sensor for distant parameters are taken into consideration: the total distance of
object detection called ultrasonic sensor. The ultrasonic sensor the mini-path, the number of consecutive unclean cells and
is mainly used for measuring the distance to the object. But the total distance of every position cell relative to the present
if the distance is small, it shows garbage value which can robot position. An appropriate fitness function of mini-path is
be misleading. So infrared proximity sensor also used for low constructed as, where is the constant number and are the distance
distance measuring purposes. After that, everything is combined of mini-path, the number of free cells and the distance of the
with an algorithm to form a complete mapping. In a word, whole free cell relative to the present golem position severally.
coordination process consists of GPS, LIDAR, IR proximity
sensor and Ultrasonic Sensor.
When battery parameters indicate that charging is necessary, the
cleaning robot can call the memory for the location to start
executing the docking and charging sequence automatically to
take the robot to the charging position. There is a complete map
of 3rd floor of IICT building of our university where the robot is
placed. Using the map, the robot start cleaning. First, the robot
gets its own position by using GPS and other proximity sensors.
Then the bot started to clean by following a side wall in a circular
pattern. After finishing one complete circle in a certain zone, the
bot added that area ti the clean list and again start cleaning by Fig. 6: Mini Path Planning Picture
increasing the distance from the wall. Repeating this process
continuously when the bot reached the center of the zone, the
zone is marked as a clean zone and the bot move to the next The mini-path contains some points which are occupied by the
uncleaned zone and repeat the whole process. During the scan, robot during its movement. These positions become cleaned but
if any obstacle is found in its path, it will follow the edges of the others remain unclean until they are taken into consideration
that obstacle until it riches to its desired path again. by the next mini-paths planning.



Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 06,2020 at 06:26:43 UTC from IEEE Xplore. Restrictions apply.
 Begin represents the Euclidean distance as shown in the following
figure.
The number of unclean cells: It is the number of consecutive
Initialization unclean cells that belong to the mini-path from the current robot
position. The total distance of each position cell relative to
the current robot position: It represents the distance between
the current robot position and the unclean cell position which
belongs to the mini-path. The ideal fitness is the solution that
contains no redundant visited and obstacles cells. Stochastic
yes tournament selection with elitism is applied based on fitness
Quit?
values.
no

Selection

Cross-over

Mutation

End

Fig. 7: Genetic Algorithm Diagram to Find Efficient Path[10]

Fig. 9: Crossover Technique Picture

4) Crossover: After two mini-paths were selected by the

selection operator, they will be considered as parents. The
crossover operator allows the generation of children to inherit
the genetic code of the mixture between the father and the
mother. We opted for a one-point crossover techniques and the
two children are then added to the population. An example of
a crossover operator is illustrated in Fig. 9 in which we notice
that from the point crossover (the third gene), the genes value of
chromosome1 is replaced by the genes value of chromosome2
and the same case for the chromosome2. Therefore, we obtained
a new two mini-paths.

Fig. 8: Total Distance of Mini-path Representation by Euclidean

Distance

The total distance of mini-path: It is calculated by the sum of

each distance between two cells position in which this distance Fig. 10: Mutation to Incorporate the Exploration Impact



Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 06,2020 at 06:26:43 UTC from IEEE Xplore. Restrictions apply.
 5) Mutation: It is used to incorporate the exploration impact. [6] Zhang Hong and Song Zhen-Hua. 2009. Research on multi-sensor fusion
The genes value is modified by one of the neighborings prece- of underwater robot navigation system. in Proceedings of the 2009 interna-
tional conference on Robotics and biomimetics (ROBIO09). IEEE Press,
dent gene. The selected point of the chromosome is random as Piscataway, NJ, USA, 1327-1330.
shown in figure 10. [7] The best affordable robot vacuums of 2018,USA TODAY(2018).
[8] N. Kawasaki and U. Kiencke, ”Standard platform for sensor fusion on
VI. C ONCLUSIONS advanced driver assistance system using Bayesian Network,” IEEE Intel-
ligent Vehicles Symposium, 2004, Parma, Italy, 2004, pp. 250-255. doi:
The primary purpose of the Cleaner Robot is to clean the 10.1109/IVS.2004.1336390
[9] Rekha Raja, Ashish Dutta, and KS Venkatesh, New potential field
floors efficiently. As earmarked, it can clean the dirt with method for rough terrain path planning using genetic algorithm for a
efficacy. The robot can detect obstacles and can avoid those 6-wheel roverRobotics and Autonomous Systems,volume-72,pages-295–
to clean the area . The robot not being a large-sized one can 306,2015,Elsevier
[10] Mohammad Sazzadul Hoque, Md Mukit, Md Bikas and Abu Naser An
be mistakenly hit by unwary pedestrians. We aim to introduce implementation of intrusion detection system using genetic algorithm,arXiv
an automated fix to this problem in the later versions. The preprint arXiv:1204.1336,2012
mapping process involving a GPS module is quite complex.
While working in an indoor environment , its slightly hard to
track the robot using GPS. GPS signals are carried through
waves at a frequency that does not move easily through solid
objects. The signals from satellites are attenuated and scattered
by roofs, walls etc. Using more sensitive GPS chips can get a
fix. As a low power DC motor was used in the vacuum cleaner,
the suction process is not optimized for a huge bunch of dust
and dirt. So, these drawbacks should be resolved. We will use
a garbage scanner that will detect garbage and only then will
start the vacuum cleaner. This way, we can reduce the power
consumption of the robot. The robot can clean a plane floor.
Our future endeavor would be to modify it so that it can clean
the stairs too.

VII. ACKNOWLEDGMENT
This work is funded by University Research Centre and
supported by the Department of Electrical and Electronic En-
gineering of Shahjalal University of Science and Technology,
Sylhet 3114, Bangladesh. This research project is university
funded project and some brilliant researchers Sajid Hasan,
Muntasir Mahdi, Ahmed Udoy, Sihab Anik have contribution
to complete this research work. The whole research team was
guided by Mohammad Kamruzzaman Khan Prince, Assistant
Professor, Department of Electrical and Electronics, Dr. Refat
Kibria, Associate Professor, Department of Computer Science
and Engineering, and Md Saiful Islam, Assistant Professor,
Department of Computer Science and Engineering. We are
grateful to the University Research Centre and Department of
Electrical & Electronic Engineering of Shahjalal University of
Science and Technology for funding and lab supports.

R EFERENCES
[1] Sung JY., Guo L., Grinter R.E., Christensen H.I. (2007) My Roomba Is
Rambo: Intimate Home Appliances. In: Krumm J., Abowd G.D., Seneviratne
A., Strang T. (eds) UbiComp 2007: Ubiquitous Computing. UbiComp 2007.
Lecture Notes in Computer Science, vol 4717. Springer, Berlin, Heidelberg.
[2] H. Lee, et al., Research roadmap of intelligent robots, Foundation of Korea
Industrial Technology, 2005, pp. 224-225.
[3] Household Vacuum Cleaners Market Analysis By Product (Upright, Can-
ister, Central, Robotic, Drum, Wet/dry), And Segment Forecasts To
2022.Grand View Research. Report ID: 978-1-68038-553-3
[4] Household Vacuum Cleaners Market Global Industry Analysis, Size, Share,
Growth, Trends, and Forecast, 2018 2026.Transparency Market Research.
Rep Id : TMRGL442
[5] W.Y. Jeong and K.M. Lee. CV-SLAM: A new ceiling vision-based SLAM
technique. In Int. Conference on Intelligent Robots and Systems (IROS),
August.2005



Authorized licensed use limited to: Auckland University of Technology. Downloaded on June 06,2020 at 06:26:43 UTC from IEEE Xplore. Restrictions apply.