IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS

Low Cost Autonomous Robot Cleaner using Mapping Algorithm based

on Internet of Things (IoT)
To cite this article: R J Ong and K N F Ku Azir 2020 IOP Conf. Ser.: Mater. Sci. Eng. 767 012071

View the article online for updates and enhancements.

This content was downloaded from IP address 172.85.104.214 on 21/03/2020 at 17:19

1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

Low Cost Autonomous Robot Cleaner using Mapping

Algorithm based on Internet of Things (IoT)

R J Ong1 and K N F Ku Azir1

1
School of Computer and Communication Engineering, Universiti Malaysia Perlis,
Pauh Putra Campus,02600 Arau, Perlis, Malaysia.

[email protected]

Abstract. The objective of this study is to design and develop a low-cost Arduino autonomous
robot cleaner using mapping algorithm to clean floor area of houses or offices. The idea is
basically to detect any obstacles with the help of sensor and sent its output to microcontroller
that will control the autonomous vacuum cleaner movement. A low-cost solution is proposed in
this study by using HC-SR04 ultrasonic sensor for obstacle avoidance and control by Arduino
UNO. By using an autonomous vacuum cleaner, user can turn ON the autonomous vacuum robot
to clean without any help of a human operator. Wall mapping and random mapping is being
applied in this study to find the effective mapping algorithm for autonomous robot cleaner.
Additionally, instead of using the traditional button or switch to activate the robot, voice
recognition through Google Assistant implemented in this project. Hence, this provides a more
user friendly platform not only for normal user but also help visually impaired people to activate
the robot.

1. Introduction
A robotic vacuum cleaner often referred to as a RoboVac, is an autonomous robotic vacuum cleaner
with intelligent programming and an electromechanical control system with high stability and rapid
functionality. In 2002, iRobot, an American advanced technology company, launched a robotic vacuum
cleaner called Roomba. The Roomba was able to change direction when it encountered an obstacle, spot
dirty spots on the floor, and detect steep debris to prevent the stairs from falling down. Additionally, an
autonomous robot can also learn new knowledge, e.g. adapting new methods to perform its tasks or
adapting to changing environments. The vacuum cleaner robot should have an artificial intelligence
mechanism to solve the problem of cleaning the entire environment area, taking into account certain
factors such as the number of turns and the length of the trajectory [1]. There have been new variations
in robotic vacuum cleaners on the market since 2002. For example, the Canadian bob sweep robot
vacuums cleaner, which use both mops and vacuum, or the Neato RoboticsXV-11 robot vacuum cleaner,
which uses laser vision instead of conventional ultrasonic models. In 2014, Dyson announced the release
of its new robotic vacuum cleaner, Dyson 360 Eye, equipped with a 360-degree camera mounted on the
top of the robotic vacuum cleaner or better navigation than other brands. The company is researching
robotics and AI to create better robotic vacuum cleaners in the future.
Four requirements should be met by a fully autonomous robot. First, the robot should be able to gain
environmental information. For a robot to associate behaviour with a location, it must know where it is
and be able to navigate from point to point. Current commercial robots navigate autonomously based on
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

the recognition of natural features. Second, the robot can move wholly or partially through its operating
environment without human support. The advantage of a robot vacuum cleaner compared to
conventional vacuum cleaners is more convenient due to its small size and can also be vacuumed by
itself. Thirdly, an autonomous cleaner should be able to avoid situations which are harmful to persons,
property or itself unless this is part of its design specifications.
In a nutshell, this study aims to develop an autonomous IoT-based vacuum cleaner that uses a low-
cost sensor to detect any obstacle and send its output to a microcontroller that controls the robot's
movement. With the standalone vacuum cleaner, the user can turn on the robot for cleaning using the
remote control or voice recognition through Google Assistant on any Android phones. Once this
machine is ON mode, it moves throughout the floor and efficiently covers an entire floor area. In
addition, the suction power of this robot must be sufficient to collect dust, sand, human or pet hairs and
some tiny rocks that are possible find out indoors. Hence, a small high-speed DC motor with 3000rpm
– 10000rpm is used to rotate a fan in order to produce strong air flow from one point to another and
archive vacuum function.

2. Literature Review
As robotic technologies have evolved and become imperative, people have tried to alternate human work
with new mechanical advances in autonomy, especially where people can take risks at work [1]. Robots
have recently emerged as household appliances as human demands increase [10]. Robotics research for
household appliances is becoming more active than ever. The market for cleaning robots is expected to
grow from USD 1.83 billion in 2017 to USD 4.34 billion in 2023, with CAGR (Compound Annual
Growth Rate) at 16.21% between 2018 and 2023, according to semiconductor and electronics market
research. Besides, The Medium & Long Term Science and Technology Planning of China (2006 to
2020) have been clearly established that the autonomous will be a primacy enhancement of technology
in the future.
It is well known that cleaning a large area is an endlessly boring task that requires a lot of time and
effort [3]. Currently, the cleaning robot can be divided into the vacuum cleaner, the wiping robot, etc.
according to various functions [4]. The robotic vacuum cleaner was introduced earlier among all these
robots. Electrolux introduced the first autonomous vacuum cleaner in 1996. The early robotic vacuum
cleaners failed to avoid obstacles and low cleanliness. As a result, this cleaner cannot be successfully
brought to market. The UK technology company Dyson developed a robotic vacuum called DC06 in
2001. Due to its high price, however, it was never released on the market.
Since 2002, most of the cleaning robot’s research has focused on reducing development costs and
developing a highly efficient cleaning robot. There are many improvements made using various sensor
and path planning algorithms to archive the goal. In [4], suggested a multifunctional floor cleaner in
which can mop and also vacuum the floor. In [5], has developed an independent cleaning robot to remove
dust from the solar panel to maximize the solar panel’s life and efficiency. In [6], proposed an
autonomous robotic vacuum cleaner with 3D vector coordinates for the planning of robot routes. Then,
came the idea of Smart Dustbin is to place sensors in the container to detect the level of garbage. When
the garbage reaches the threshold, the status of the bin in the cloud is updated and a notification is sent
to the user. The garbage that accumulates in garbage bins is collected at the time of day, even if the
garbage bins are filled at any time. Therefore, the existing technology is not fruitful as the emerging
technology IoT is considered. IoT components such as sensors, detectors and actuators are specifically
integrated with the Intelligent System (IS) and inspection system to ensure efficient waste disposal [7].
There are many smart waste disposal projects that are described in the literature today. In [8], Waitkus
has implemented a waste bin and garbage collection monitoring system to monitor the capacity of the
containers, to notify the garbage trailer, and to schedule the collection of garbage based on customer
preferences. In [9], the ultrasonic sensor measures the levels of waste in the dustbin and sends the data
through a wireless ad hoc network to the server. In [10] used RFID tags and load cell sensor technology
for waste treatment.

2
1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

Most of the works above require inconvenient operations and/or costly solutions. Therefore, there is
essentially to reduce the system design complexity, cost-effective and universal availability of the
infrastructure [11]. The growth of information technology and the use of pervasive computing
technology have occurred in many applications, including waste disposal [12]. Therefore, in [2] propose
fully automated indoor waste disposal to replace conventional disposal facilities by using a mobile waste
collection robot. This method is definitely a time effective process than currently used methods.
Mapping and localisation are two closely related activities in the autonomous robot cleaner. The map
created can be used in a variety of applications, such as route planning, safety, cleaning, monitoring and
just to name a few. Since the autonomous robot does not require any human input for its work. Therefore,
the localization and navigation of the robot will be the main concern in the design. High-priced, highly
functional sensors such as cameras and laser scanners provide a wealth of information about surrounding
environments and are therefore widely used. Both sensor types were successfully used for the creation
of raster maps. In particular,' Kinect ' sensors have recently come to the fore, as they are cheaper but
provide rich 3D information. In [2] et.al presented a robot navigation system based on the Kinect device
and encoder to create a 2D map. Ronny Mardiyanto’s system is tested in a laboratory setting. The
effective distance is between 60 and 600 cm. Their rotary encoder sensor has an average error of 5.05%
on the created 2D map and an error distance of 2.23%. This 2D map creator took 1.46 seconds to create
each map in one place. They are disadvantageous, however, because they do not recognize transparent
objects like windows. Additionally, high information quality is difficult to develop an algorithm to
manage errors and noise. It is therefore not currently possible to promote such sensors to the market. In
contrast, ultrasonic sensors are practicable for obstacles avoidance and robot mapping. The HC-SR04
ultrasonic sensor with a range of up to 4 meters is a cost-effective solution for mapping robots, rather
than using an expensive laser sensor or camera. In [20] suggest a new approach to creating a raster map
for a mobile robot with sonar data. Experimental results and evaluations in the home environment
demonstrate the validity of the methods proposed. Complete raster maps, which were created in the way
suggested by the research, probably reduce the development cost. In the meantime, Atanas Dimitroy
and others presented a prototype of a mobile robot suitable for mapping [16]. It uses a HC-SR04
ultrasonic sensor that is a cost-effective alternative to sensor mapping instead of a costly laser sensor.
The proposed robot was tested in the university laboratory and can be improved by visualizing real-time
data. In fact, sonar data may not always be accurate because of specular reflective indoor operation. In
[17] explains that ultrasonic sensors and compass sensors are used in the mapping algorithm. The sensors
used in this document are limited to the environment. Ultrasonic sensors can only detect the distance
between 3 cm and 400 cm. Ultrasonic sensors, however, is cheaper compared to another same
functionality sensor, and the raster maps created with ultrasonic sensors can most often be reduced by
the raster map making methodology proposed in the development of smart robots for low cost.
The current autonomous robotic vacuum cleaner production methods can be classified into three
types, namely random walk, video camera location and laser location. In [18] present a new effective
global localization approach for residential environments that adapts the notion of free space density to
a ceiling - mounted camera. This vacuum cleaner has a camera which captures the ceiling image and
uses the lens during transportation. The use of image processing technology creates a map that allows
for more efficient cleaning. On the other hand, iRobot Roomba, one of the most popular robotic vacuum
cleaners, used an algorithm called iAdapt and used a laser rangefinder to scan around and under
furniture, avoid stairs, and clean up very efficiently. In contrast, the random walk algorithm only avoids
obstacles with the help of some sensors, and the robot itself does not recognize the work environment.
The first two methods, however, are not cost efficient due to the camera module and the cost of the laser
rangefinder is relatively higher. In [19] assess a cost-efficient, lightweight, low noise and low
maintenance robotic system. It is said that mapping technology is absent in its design because of the
cost effective problem. Data shows that the robot covers the whole area more quickly with combined
motion algorithms (random walk, s shape and spiral) based on numerous tests. The test result shows that
their design is fully adapted to the planning of an effective mode for performing the cleaning task in an
unknown environment [20] They also suggest a dynamic return to recharge, including the relative

3
1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

coordinates of the starting point and the current position. Based on the above research results, path
planning is one of the keys to better cleaning robot performance. In this cyber-physical age, the
autonomous cleaner can be improved by implementing Internet of Things (IoT) with a number of
sensors, with some calculations, to develop a new algorithm to solve the problem in path planning or
robotic navigation. For that reason, the motto of this project is to propose a highly efficient and cost-
effective cleaning robot using an odometer and with the help of several sensors.

3. Method
The overall of this study is to design an autonomous vacuum cleaner that can perform well without
human assistance. In the early stages of design, network connections must be established to ensure that
all hardware is connected with each other. This can be done by either check the serial monitor in Arduino
IDE or check the list of devices connected to the same local network. Once the system is online, the
robotic vacuum cleaner can be either turn on or off by using voice command via Google assistant as
shown in Figure 1. Meanwhile figure 2 shows the block diagram of the study. First and foremost, voice
command has been implemented with Google Assistant to help the visually impaired. However, there is
no way for Google Assistant to connect directly to NodeMCU. Therefore, Blynk and IFTTT are
implemented to act as an intermediate bridge to allow the Google Assistant to control the microcontroller
to perform as the ON / OFF switch of the robot. To achieve this must have some pre-setting in both
IFTTT and also Blynk.
There are two algorithms can be chosen by the user which are wall follow and random walk. By
selecting different algorithm, the decision to take will be different when the robotic cleaner duel with
obstacle avoidance. After the device is activated, the vacuum pump will start to work. Moreover, all the
sensors will begin to collect environment data and process within Arduino UNO. The robot will keep
moving around the workplace until it runs out of battery or receives an OFF command from the user.
Figure 3 shows that the subroutine will be called when the random walk algorithm is chosen by the
user. First and foremost, the robot will check is the obstacle right in front of the robot. If yes, then the
robot will stop for a while in order for the DC motor to take action. The robot is programmed to reverse
a bit and following by turn left (reverse the left wheel and forward the right wheel). If no, then the robot
will check either the obstacle is at left or at right and apply the same algorithm but rotate in the opposite
direction. While there is no obstacle, the robot will keep moving forward.
Figure 4 shows that the subroutine will be called when the wall follow algorithm is chosen by the
user. First and foremost, the robot will check is the obstacle right in front of the robot or the robot
reaching a corner. If yes, then the robot will stop for a while in order for the DC motor to take action.
The robot is programmed to reverse a bit and following by turn right (reverse the right wheel and forward
the left wheel). If no, then the robot will check is the robot reaching an edge or not. If yes, the robot will
stop for a while and follow by slightly forward and turn left. Furthermore, the ultrasonic sensor at left
will always make sure the robot moves along the wall with constant speed.
The backbone of this project is to connect the sensor to the microcontroller to control the actuator to
perform certain tasks under certain conditions. Ultrasonic sensor used to avoid obstacles in this project.
For instances, the robot is programmed to take action when there is an obstacle 10cm in front of the
ultrasonic sensor. The red colour board at the top left in figure 5 is L298N H- bridge motor driver. There
are two DC motor connected to this board and power supply by two 3.7V 18650 3000mAh rechargeable
li-ion battery. There are two EN pins which are ENA and ENB have to connect to PWM pins from
Arduino UNO in order to control the speed of rotation.

4
1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

Figure 1. Flow chart of the study.

Figure 2. Block diagram of the study.

5
1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

Figure 3. Flow chart of random walk.

Figure 4. Flow chart of follow algorithm.

6
1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

Figure 5. Circuit diagram of the study.

4. Results and Analysis

The testing of the automous robot cleaner was done by implementing both of the mapping algorithm to
the automous robot cleaner. The movement of a robot can be divided into several simple movements as
in Table 1. The truth table is generated based on L298N H-bridge motor driver after a few times testing
programmed by Arduino UNO. The combinations of these movements are successfully achieved both
wall follow and random walk algorithm.

Table 1. Truth table of motor driver

Action Left Motor Right Motor ENA ENB IN1 IN2 IN3 IN4

Forward On On Const Const 1 0 1 0

Reverse On On Const Const 0 1 0 1
Turn Left Off On 0 Const 0 1 1 0
Turn Right On Off Const 0 1 0 0 1
Stop Off Off 0 0 0 0 0 0

There are four situations considered during design. Figure 6 shows that the four situations that happened
during testing and the robot effectively overcome the problem by itself without human assist. First, the
robot is doing well and moving along the wall with a gap of 4cm with the wall. The robot is capable to
adjust itself automatically when either going too far from the wall or too close to the wall. Second, the
robot escapes from a corner and continue its path to the right intelligently. Third, the robot is doing well
when come to the wall edge so that it can close back to the path and continue its task.
For a random walk algorithm, the robot proved that it is smart enough to avoid obstacle before hitting
on it. In addition, there are not only static obstacles but also dynamic obstacles such as people walk
through the path of the robot as shown in figure 7. Thus, the robot will stop for a while and take action
before hitting on such dynamic obstacles.

7
1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

Figure 6. Four situations to overcome by wall follow mapping algorithm.

Figure 7. Random walk mapping algorithm and obstacles avoidance.

Figure 8 shows the screenshot of Google Assistant and Blynk when using voice command. As mentioned
earlier in the previous chapter, there are three optional commands in order to activate the robot. When
Google Assistant receives a valid command such as “turn off” in the figure 8 then it will reply “Yes sir
turning off vacuum cleaner” as an acknowledgement receive command successfully.

Figure 8. Google Assistant and Blynk for Voice command.

5. Conclusion
The objective of this project was to plan and deployment a cheap and smart Internet of Things (IoT)
based autonomous vacuum cleaner that could work in an unknown environment. Although the robot has
a minimum number of sensors, this goal has been met by deployed algorithms. Moreover, voice
command fulfilled successfully and the output is same as expected. Google Assistant are available on
every android smart phone, hence, this can be an ease for development and also provide a more user

8
 1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

friendly platform not only for normal user but also visually impaired people. Walk follow and random
walk algorithm implemented successfully which that all the situation considered can be overcome by
the robot automatically. Even so, robot development provides endless possibilities to improve the
performance and to overcome more situations for unknown environment. Future robot development
therefore includes mapping technology, automatic charging algorithm and automatic dirt disposal.

Acknowledgments
Authors wishing to acknowledge assistance or encouragement from colleagues, special work by
technical staff and financial support from Ministry of Higher Education Malaysia and Universiti
Malaysia Perlis.

References
[1] Maddukuri S V P K, Renduchintala U K, Visvakumar A, Pang C and Mittapally S K (2017). A
low cost sensor based autonomous and semi-autonomous fire-fighting squad robot.
Proceedings - 6th Int. Symp. on Embedded Computing and System Design (ISED), p 279–
283. (doi: 10.1109/ISED.2016.7977097).
[2] Mardiyanto R, Anggoro J and Budiman F (2015). 2D map creator for robot navigation by
utilizing Kinect and rotary encoder. In Int. Seminar on Intelligent Technology and Its
Applications (ISITIA) p 81–84. (doi:10.1109/ISITIA.2015.7219957).
[3] Ravindhiran P, Gopal P, Gladwin S J and Rajavel R (2017). Automated indoor waste
management system employing wavefront algorithm and received signal strength indicator
values-based mobile robot. In IEEE Region 10 Humanitarian Technology Conf. (R10-HTC)
p 284–289. (doi: 10.1109/R10-HTC.2017.8288958).
[4] Wickramaarachchi W H C, Chamikara M A P and Ratnayake R A C H (2018). Towards
implementing efficient autonomous vacuum cleaning systems. Proc. IEEE Int.l Conf. on
Industrial and Information Systems (ICIIS), p 1–6. (doi: 10.1109/ICIINFS.2017.8300385).
[5] Hong Y, Sun R, Lin R, Yu S and Sun L. (2014). Mopping module design and experiments of a
multifunction floor cleaning robot. In Proc. of the 11th World Congress on Intelligent Control
and Automation. p 5097–5102. (doi: 10.1109/WCICA.2014.7053581).
[6] Aravind G, Vasan G, Kumar T S B G, Balaji R N and Ilango G S (2014). A Control Strategy for
an Autonomous Robotic Vacuum Cleaner for Solar Panels, Texas Instruments India
Educators' Conf. (TIIEC), Bangalore, p 53-61. (doi: 10.1109/TIIEC.2014.18\)
[7] Bai Y W and Hsueh M F (2012). Using an adaptive iterative learning algorithm for planning of
the path of an autonomous robotic vacuum cleaner. In 1st IEEE Global Conf. on Consumer
Electronics (GCCE). p 401–405. (doi: 10.1109/GCCE.2012.6379640).
[8] Murugaanandam S, Ganapathy V and Balaji R (2018). Efficient IOT Based Smart Bin for Clean
Environment. Int. Conf.on Communication and Signal Processing (ICCSP), p 715–720.
[9] Folianto F, Low Y S and Yeow W L (2015). Smartbin: Smart waste management system. IEEE
10th Int. Conf.on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP)
p 1–2. (doi: 10.1109/ISSNIP.2015.7106974).
[10] Chowdhury B and Chowdhury M U (2008). RFID-based real-time smart waste management
system. Australasian Telecommunication Networks and Applications Conf. (ATNAC). p 175–
180. (doi: 10.1109/ATNAC.2007.4665232).
[11] Talha M, Upadhyay A, Shamim R and Beg M S (2018). A cloud integrated wireless garbage
management system for smart cities. Int. Conf. on Multimedia, Signal Processing and
Communication Technologies (IMPACT). p175–179. (doi:10.1109/MSPCT.2017.8363999)
[12] Wahab M H A, Kadir A A, Tomari M R and Jabbar M H (2014). Smart recycle bin: A conceptual
approach of smart waste management with integrated web-based system. Int. Conf. on IT
Convergence and Security (ICITCS). p1–4. (doi: 10.1109/ICITCS.2014.7021812).

9
 1st International Symposium on Engineering and Technology (ISETech) 2019 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 767 (2020) 012071 doi:10.1088/1757-899X/767/1/012071

[13] Yakoubi M A and Laskri M T (2016). The path planning of cleaner robot for coverage region
using Genetic Algorithms. Journal of Innovation in Digital Ecosystems. 3(1) p 37 – 43, (doi:
10.1016/j.jides.2016.05.004).
[14] Hasan K M, Al-Nahid A and Reza K J (2014). Path planning algorithm development for
autonomous vacuum cleaner robots. In Int. Conf. on Informatics, Electronics & Vision (ICIEV)
p 1–6. (doi: 10.1109/ICIEV.2014.6850799).
[15] Dimitrov A and Minchev D (2016). Ultrasonic sensor explorer. In 19th Int. Symp. on Electrical
Apparatus and Technologies (SIELA). p 1–5. (doi: 10.1109/SIELA.2016.7542987)
[16] Fathan A M, Jati A N and Saputra R E (2016). Mapping algorithm using ultrasonic and compass
sensor on autonomous mobile robot. In Int. Conf. on Control, Electronics, Renewable Energy
and Communications (ICCEREC). p 86–90. (doi:10.1109/ICCEREC.2016.7814986)
[17] Ribacki A, Jorge V A M, Mantelli M, Maffei R and Prestes E (2018). Vision-based Global
Localization using Ceiling Space Density. 2018 IEEE International Conference on Robotics
and Automation (ICRA), Brisbane, p 3502-3507. (doi: 10.1109/ICRA.2018.8460515)
[18] Hasan K M, Al-Nahid A and Reza K J (2014). Path planning algorithm development for
autonomous vacuum cleaner robots. In Int. Conf. on Informatics, Electronics & Vision
(ICIEV). p 1–6. (doi:10.1109/ICIEV.2014.6850799).
[19] Bai Y W and Hsueh M F (2012). Using an adaptive iterative learning algorithm for planning of
the path of an autonomous robotic vacuum cleaner. In 1st IEEE Global Conf. on Consumer
Electronics (GCCE). p 401–405. (doi:10.1109/GCCE.2012.6379640).
[20] Lee S-J, Lee K and Song J-B (2014). Development of advanced grid map building model based
on sonar geometric reliability for indoor mobile robot localization. In 11th Int. Conf. on
Ubiquitous Robots and Ambient Intelligence (URAI). p292–297.
(doi:10.1109/URAI.2014.7057437).

10