Mapping, Localization and Navigation for an Assistive Mobile Robot

in a Robot-Inclusive Space

Prabhu R. Naraharisetti

, Michael A. Saliba

and Simon G. Fabri

Faculty of Engineering, University of Malta, Msida, Malta

Keywords: Observability, Accessibility, Robot-Inclusive Space, Vision-Based SLAM, LiDAR-Based SLAM, GMapping,

Hector SLAM, Mapping, Path Planning, ROS Navigation Stack.

Abstract: Over the years, the major advancements in the field of robotics have been enjoyed more by the mainstream

population, e.g. in industrial and office settings, than by special groups of people such as the elderly or persons

with impairments. Despite the advancement in various technological aspects such as artificial intelligence,

robot mechanics, and sensors, domestic service robots are still far away from achieving autonomous

functioning. One of the main reasons for this is the complex nature of the environment and the dynamic nature

of the people living inside it. In our laboratory, we have started to address this issue with our minimal degrees

of freedom MARIS robot, by upgrading it from a teleoperated robot to an autonomous robot that can operate

in a robot-inclusive space that is purposely designed to adopt algorithms that are not very computationally

intensive, and hardware architecture that is relatively simple. This paper discusses the implementation of

suitable SLAM algorithms, to select the best method for mapping and localization of the MARIS robot in this

robot-inclusive environment. The emphasis is on the development of low-complexity algorithms that can map

the environment with lesser errors. The paper also discusses the 3D mapping, and the ROS based navigation

stack implemented on the MARIS robot, using just a LiDAR, a Raspberry Pi processor, and DC motors with

encoders as main hardware architecture, so as to keep low costs.

1 INTRODUCTION

In recent years, service robots have found their way

into public environments, such as in hospitals to

deliver food and medicines to patients in quarantine

(e.g. UBTECH robot) (Seidita et al., 2021), or in high

end restaurants to deliver food to the customers while

attaining a competitive advantage over their rivals

(e.g. SERVI robot (SERVI robot, 2023)). But perhaps

a more pressing requirement for these service robots

is in home settings, to serve the elderly and impaired

as their population is increasing globally. Some

companies have put their efforts to bring service

robots into home spaces (Polaris-Market-Research,

2023), but these robots are not very effective in

performing the daily household tasks to serve the

elderly and impaired satisfactorily. A primary reason

for this is the complex nature of the environment

including the dynamic nature of the people living in

https://orcid.org/0009-0007-0117-4888

https://orcid.org/0000-0002-1985-3076

https://orcid.org/0000-0002-6546-4362

it, that typically demand the use of highly complex

robot technologies (Sosa et al., 2018). High end

complex robots such as ASIMO (ASIMO robot,

2023) and SPOT (Boston Dynamics - Spot robot,

2023) may in principle be capable of performing

certain demanding household tasks, but these robots

would not be widely affordable. Thus, a minimally

complex robot that still has the functionalities to

perform daily household tasks, in particular to

address the specific needs of the elderly, could be

applied to operate in a robot friendly home

environment called a robot-inclusive space (RIS)

(Sosa et al., 2018). Prior to the present work, in the

first part of our Mobile Assistive Robot in an Inclusive

Space (MARIS) project, a survey was conducted to

understand the common needs of the elderly and

impaired (Aquilina et al., 2019). In subsequent work,

ten representative tasks were extracted to encapsulate

these needs, namely: 1) preparing or bringing

172

Naraharisetti, P., Saliba, M. and Fabri, S.

Mapping, Localization and Navigation for an Assistive Mobile Robot in a Robot-Inclusive Space.

DOI: 10.5220/0012172900003543

In Proceedings of the 20th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2023) - Volume 1, pages 172-179

ISBN: 978-989-758-670-5; ISSN: 2184-2809

medication; 2) heating a meal in a microwave oven;

3) operating a telephone; 4) preparing a small snack;

5) getting items from a refrigerator or cupboard;

6) taking out the garbage; 7) preparing tea or coffee;

8) arranging vegetables for chopping and/or cooking;

9) drying and putting away dishes; and 10) setting and

clearing a table (Naraharisetti et al., 2022). The RIS

principles such as observability and accessibility will

be considered in designing home environments that

are suitable for mapping and navigation by the

MARIS robot to perform the representative set of

tasks. The objective of this research is to conduct a set

of experiments to obtain quantitative evaluations that

determine the robot’s complexity and to illustrate the

fact that by increasing the inclusiveness of the home

environment, the robot complexity can be reduced.

To perform these representative tasks, a robot

platform should have the capability to move around

the environment while avoiding obstacles and

choosing the shortest routes to reach the target.

A first prototype of a mobile assistive robot

(MARIS-I), that was intended to be used in a RIS, and

that incorporates only teleoperated control, was

introduced in (Aquilina et al., 2019). The present

work discusses an autonomous system based upon the

Simultaneous Localization and Mapping (SLAM)

(Alsadik and Karam, 2021) approach implemented on

the MARIS Omni directional three-wheeled robot

base using a Light Detection and Ranging (LiDAR)

sensor. The work emphasizes the implementation of

autonomous robot functionalities to map the

environment, to self-locate in that environment, and

to navigate to a destination with minimal instructions

using Robot Operating System (ROS)-based SLAM

algorithms, and a ROS navigation stack that

integrates with a camera or LiDAR.

The rest of this paper is organized as follows:

Section 2 explains the general methodology adopted

in this paper. Section 3 describes the LiDAR-based

SLAM approaches implemented on the MARIS

robot. Section 4 discusses the vision-based SLAM

implementations on the MARIS robot. Section 5

describes the selection of the most suitable SLAM

method. Section 6 describes the ROS-based

navigation stack implemented on MARIS for

autonomous navigation. Finally, section 7

summarizes the work and briefly discusses the

ongoing work to improve the RIS and to implement

pick and place functionalities to the MARIS robot.

2 GENERAL METHODOLOGY

Based on the RIS principles of observability and

accessibility (Elara et al., 2018), some general features

in a one-floor home environment were mandated a

priori. These were: no very bright objects, no dark

colour wall paints, no sharp edges on the furniture, no

rough floor surfaces, a less clustered environment,

non-slippery floors, no uneven surfaces, non-

reflective floor surfaces, no rooms with heavy doors,

and no glass or transparent environments.

(a)

(b)

(c)

Figure 1: RIS environments (a) RIS 1 with more objects in

the environment, (b) RIS 2 with a hallway, (c) RIS 3 with

hallways on both sides of the bedroom.

For the LiDAR investigation, three different RIS

environments were designed using Gazebo (Gazebo

Building Editor, 2023) considering the general

features with minor improvements from one another

in terms of furniture placements, room location and

hallway designs as shown in figure 1.

RIS 1 environment consists of an enclosed kitchen

separated from the hall by a wall. So, eight of the ten

representative sets of tasks require the robot to move

from the kitchen to the hall. RIS 2 environment has a

hallway with a kitchen in the hall that allows the robot

to move with ease. The blue shaded regions in figure

1, represent the LiDAR rays that will be converted to

2D maps. Additionally, artificial landmarks such as

familiar shapes in the environment help to map the

environment precisely and help for better robot

localization. RIS 3 environment has two hallways with

a room at the center that allows the robot to move from

Mapping, Localization and Navigation for an Assistive Mobile Robot in a Robot-Inclusive Space

173

any direction to find the optimal path for

navigation. The LiDAR investigation was carried out

firstly in the ROS-RVIZ simulation environment

(Kam et al., 2015).

Both LiDAR and machine vision investigations

were then carried out experimentally in the Robotic

Systems Laboratory (RSL) of the Department of

Mechanical Engineering of the University of Malta. In

order to determine the least complex mapping suitable

for our MARIS robot to perform the representative set

of tasks, the approach shown in figure 2 was adopted,

which will be further elaborated in the rest of the

paper.

Figure 2: Approach to design a less complex SLAM to

navigate the robot.

3 LiDAR-BASED SLAM

APPROACH

3.1 Overview

In order to conform to our objectives to introduce a

less complex and more economically accessible

autonomous robot in the RIS home environments, we

implemented LiDAR based SLAM using a relatively

low cost 2-D RPLiDAR from SLAMTEC

(SLAMTEC RPLiDAR, 2023). The SLAMTEC 2-D

RPLiDAR emits and receives the reflected laser beam

and measures the time the beam takes to return. This

process is repeated for more than 8000 times per

second, producing a map of the surroundings with a

desirable point density. This method was investigated

only after considering its quick and precise solution to

create maps. The GMapping (Revanth et al., 2020)

and Hector LiDAR-based SLAM (Sat t et al., 2020)

algorithms were investigated and implemented on the

MARIS Omni directional three-wheeled holonomic

base, which is equipped with DC motors with

encoders and control hardware that runs on the

Arduino platform. To implement a portable high-level

control system, a widely used ROS software

framework was installed on a Raspberry Pi 4 single-

board computer running the Ubuntu operating system.

ROS-based SLAM techniques compatible with the

SLAMTEC LiDAR software development kit

provided a way to map the surroundings and localize

the robot.

3.2 GMapping

GMapping is based on a Rao-Blackwellized Particle

filer (RBPF) (Revanth et al., 2020; Sarkka et al., 2007)

SLAM approach that is widely used for robot

navigation, and uses a particle filter in which each

particle carries an individual map of the environment.

GMapping considers the movement of the robot and

compares it with the recent environment data to

decrease the inaccuracies of the robot pose. The

parameters considered in the research are kernel size,

linear update, resample threshold and the number of

particles that combinedly determine the accuracy of

the map. The inputs to the GMapping SLAM

algorithm are the robot transform, laser data and

odometry data that give the information of the robot

pose, and generate the 2D occupancy grid or map that

displays the obstacles and free spaces. The obtained

map can be saved using the map server package of

ROS to make the robot localize and navigate in the

map. Initially, the GMapping SLAM algorithm was

implemented. The laser inputs for GMapping SLAM

were obtained from the SLAMTEC 2D LiDAR

connected to Raspberry Pi. The odometry data that are

used to estimate the robot position and orientation

were obtained from the encoders of the three DC

motors on the MARIS base, by determining the speed

and distance travelled by MARIS.

3.3 Hector SLAM

Hector SLAM on the other hand uses a scan matching

algorithm based on the Gauss-Newton approach

(Cheng et al., 2021) that can be used without odometry

data. The high update rate and accuracy of modern

LiDAR hardware make the scan matching algorithm

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

174

sufficient for a robot to achieve accurate poses. The

scan matching algorithm matches the current scan to

the previous scan to determine the robot movement.

The parameters of the Hector SLAM including map

size, map update distance threshold, map update angle

threshold, laser minimum and maximum distance, can

be modified to obtain a better map of the environment.

The choice of these parameters makes the SLAM

adaptable to the specific environment, however

changing the parameters adds to the computational

cost of the algorithm. The speed of the algorithm and

frequency of data logging have an impact on map

accuracies. The parameters under each SLAM

algorithm mentioned before were modified uniformly

until a better map was reached. This kind of

adaptations helped us to determine the robot

complexity and quantify the RIS environments in

terms of inclusivity. Figure 3 shows the results of the

GMapping and Hector SLAM algorithms

implemented on the three RIS environments. For

instance, under predefined SLAM algorithm

parameters, higher complex environments will

generate distorted and less accurate maps than lesser

complex environments.

Enviro-

nment

GMapping Hector SLAM

Enviro-

nment

Enviro-

nment

Enviro-

nment

Figure 3: GMapping and Hector SLAM implemented on

RIS environments 1,2,3.

The second and third columns of Table 1 compare

the GMapping and Hector SLAM algorithms for the

three RIS environments combinedly, based on their

average values. For instance, the map error of

GMapping for the three RIS environments 1, 2, and 3

are 2cm, 2.4cm, and 2.1cm respectively. Their average

((2+2.4+2.1)/(3) = 2.17cm) is taken as a measure to

evaluate the accuracy of the SLAM algorithms. Other

factors such as noise and features detected in the maps

were assigned a value based on a widely followed 5-

point Likert scale (Allen and Seaman, 2007).

Table 1: Comparison of factors of GMapping, Hector,

Vision and LiDAR fusion SLAM.

Factors GMapping Hector Vision and

LiDAR fusion

mapping

Map accuracy

(average across

RIS 1, 2, 3)

2.17 cm

error

2.3 cm error 0.5 cm error

Time to build a

map

seconds

10 seconds 16 seconds

Noise of the

environment

3 (Likert

scale)

2 (Likert

scale)

3 (Likert

scale)

Number of

mapping

parameter

changes

4 5 No changes,

but under

uncluttered

and amiable

lighting

conditions

Computational

load

a) Memory

in %

18 6 42

b) CPU load

in %

89 16 98

Features detected

in the map

3 (Likert

scale)

4 (Likert

scale)

5(Likert

scale)

4 VISION-BASED SLAM

IMPLEMENTATIONS

To move a robot autonomously to some desired

location, the spatial representation of the environment

should be known to the robot. The robot needs to have

a sensor or sensors that save the data of the

environment to enable robot localization. For the

autonomous robot (MARIS-II), this mapping refers to

the construction of the spatial environment to help the

robot perceive its surroundings and localize itself, and

to navigate accordingly. This SLAM process would,

in our case, involve continuously fusing onboard

sensor data from the LiDAR and/or the cameras and

from wheel encoders. With the recent advancements

in vision-based technologies, camera-based vision-

SLAM is gaining importance because it provides 3D

information of the environment. However, these

systems are mostly used in an indoor environment as

Mapping, Localization and Navigation for an Assistive Mobile Robot in a Robot-Inclusive Space

175

the camera range is limited and the machine vision is

sensitive to variations in light (Debeunne and Vivet,

2020).

As the objective is to adopt a suitable algorithm

with no compromise in performing the representative

set of tasks or subtasks, we have also implemented

and evaluated the vision-based SLAM. The vision-

based SLAM uses an Intel RealSense 435i depth

camera (Intel RealSense 435i depth camera, 2023)

that runs on Ubuntu, an open source Linux

distribution installed on a Raspberry Pi single board

computer. The resulting pure 3D map of the

environment of the laboratory (Figure 4 (a)), which is

9.15x5.73m in size is shown in figure 4 (b). The pure

3D map algorithm utilized only the camera to

perceive the environment, but we will need a map that

contains floor as well as static and dynamic obstacles

for autonomous navigation. So, we have implemented

vision and LiDAR fusion-based mapping on MARIS.

(a)

(b)

Figure 4: Image of the RSL (a) actual RSL (b) pure 3D map

of RSL.

The pure 3D map is fused with the LiDAR data

using the Real-Time Appearance-Based Mapping

algorithm (Labbé and Michaud, 2019) to obtain the

floor map thus giving a perception of the objects

around as well as the obstacles in its path (Figure 5).

When the RSL has a cluttered environment with

objects spreading unevenly around, the vision system

has a lot of noise as shown in figure 5(a). But after re-

arranging the chairs and objects around properly the

map shows minimal noise as in figure 5(b). The inbuilt

Inertial Measurement Unit (IMU) (Ahmad et al.,

2013) sensor in the RealSense camera also achieves

the tasks of localizing the robot in an environment, but

with the progress in the mapping, and the continuous

localization of the robot in a complex environment, the

usage of the central processing unit (CPU) of the robot

will also increase, since feature complexity and

processing time are correlated. The robustness of the

CPU is questioned in this situation as real-time

applications demand graphical processing units

(GPU) rather than traditional processors. Since the

Raspberry Pi 4, used in our research to connect to the

RealSense camera, has a CPU that may not be

sufficient to perform Realtime SLAM, external GPUs

would have to be connected. Other ways to

compensate for the load on the CPU is by using fewer,

and similar, objects in the environment to facilitate

their detection by the systems (Kamarudin et al.,

2014). Furthermore, the challenge that limits the

performance of Visual-based SLAM is due to the

disruptions in the lighting conditions that introduce

inaccuracies. The fourth column of Table 1 displays

the factors of vision-based mapping.

(a)

(b)

Figure 5: Vision and LiDAR fusion mapping of the RSL:

(a) in a cluttered RSL environment, (b) in a properly

arranged RSL environment.

5 SELECTION OF THE SLAM

METHOD

After studying the performance of the three SLAM

algorithms as implemented separately on the MARIS

robot, a selection needed to be made as indicated in

Figure 2. Table 2 summarizes the main features of the

three methods as extracted from Table 1. Feature 1,

less intensive computations, refers to performing

SLAM with minimally sophisticated processing and

memory storage devices. Feature 2, economical in

terms of price, refers to performing SLAM with

comparatively cheaper equipment. Feature 3, less

prone to distortions, refers to SLAM map with

minimal unevenness. Feature 4, modification to the

environment to facilitate SLAM (both LiDAR and

Vision based) refers to efforts such as ensuring smooth

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

176

Table 2: Comparison of SLAM methods for the MARIS

robot.

Desired Features

SLAM Algorithms

Vision

based

GMapp

ing

Hector

1. Less intensive

computations

No No Yes

2. Economical in

terms of price

No Yes Yes

3. Less prone to

distortions

No Yes No

4. Require less

changes in the

environment to

facilitate SLAM

No Yes Yes

surfaces and maintaining favourable illumination

(Naraharisetti et al., 2022).

These features, including the computational load

and other factors are shown in Table 2, based on which

the Hector SLAM as investigated and tested on our

MARIS robot will be used for autonomous navigation.

Even though GMapping localizes the robot better than

Hector SLAM, the latter is sufficient and proved to be

reliable under restricted speed and angular velocities

of the robot in RIS environments 1,2, and 3.

Hector SLAM based experiments were also

performed in our RSL to evaluate the robustness of the

algorithm (Figure 6). The environment of the lab with

8 office chairs, tables with almost 24 supports and

wires under the tables resemble a cluttered

environment. The experiments helped us to

understand and tune our robot to navigate better in the

RIS environments. The parameters that were tuned,

shown in Table 3, were considered to adjust the robot

complexity to serve in any RIS environments

considering proper implementation of RIS principles.

Figure 6: Hector SLAM map of our RSL under fixed SLAM

parameters.

6 ROS-BASED ROBOT

NAVIGATION

Robot navigation involves planning routes and

moving

the robot safely and conveniently through an

Table 3: Tuned mapping parameters of the robot after

experiments.

Parameters Reason Values

Speed of the

robot

Robot moving at high

speed distorts the map

0.3m/s

Angular

velocity

Higher velocity will

distort map

1 rad/s

Map update

distance

threshold

The robot has to travel

to have an angular

change

0.2 m

Map update

angle threshold

The robot has to have

an angular change

0.2 rad

economically beneficial route from one point to

another. In order to implement autonomous navigation

on the MARIS robot by integrating the LiDAR and

other hardware architecture, an open source ROS

navigation stack (Setup and configuration of

Navigation stack, 2023) shown in figure 7 was chosen.

Figure 7: ROS Navigation stack architecture implemented

on MARIS.

The ROS navigation stack requires the Adaptive

Monte Carlo Localization (AMCL) ROS nodes to

localize the robot moving in a 2D space, and AMCL

takes the data from the laser scans to determine the

pose of the robot (Matias et al., 2015). Sensor

transforms refers to publishing the stream of LiDAR

data over ROS, and the odometry information

(Publishing Odometry information over ROS, 2023)

can be published using the transform (“tf”)

(transform-tf, 2023), which tracks the coordinate

frames and transforms points and vectors between the

coordinate frames. The map server node (Map_server,

2023) of ROS gives the map data to the ROS move

base package (Move_base package, 2023) that links

global and local planners to perform the navigation

task. The global cost map and the local cost map (setup

and configuration of Navigation stack, 2023) store the

information of the obstacles to create a long-term plan

to navigate the robot without colliding with the

obstacles. The base controller accepts the command

Mapping, Localization and Navigation for an Assistive Mobile Robot in a Robot-Inclusive Space

177

velocity topic from the moving base, which gives the

robot linear and angular velocities at that instant and

converts them to individual wheel velocities.

The ROS navigation stack for a differential (two-

wheeled) robot is supported with the required

information on its website. However, our MARIS

robot uses an omni directional three-wheeled

architecture, which is not readily supported in the

stack. So, changes in parameters such as the Odom

model type from “differential” to “omni-corrected” in

the AMCL launch file in ROS; changes in

odom_alpha1,2,3,4 that determine the expected noise

in the odometry rotation and translation estimate from

rotational and translational components of robot

motion; and the inclusion of a new odom_alpha5 that

determines the translation-related noise parameters

that are used to identify only the tendency of the robot

to translate in a perpendicular direction of travel

etc.,(AMCL parameters, 2023) were made so as to

implement a new navigation stack for the omni

directional three wheeled MARIS robot shown in

figure 8 (a).

Finally, as Hector SLAM is selected instead of

GMapping and vision-based SLAM, we have used the

pose obtained from the Hector mapping and supplied

the odometry message to the AMCL of ROS

navigation stack implemented on the MARIS robot.

This navigation is computationally low and accurate

as the 2D-RP LiDAR is used to obtain laser scan

matching. The navigation stack determines the

presence of obstacles in the environment and avoids

them. For instance, the four dark circles on the map

labelled with a red circle shown in figure 8 (b) are the

four supports of a table which, together with the

navigation stack, supply the local and global map to

the robot to move to the required destination.

(a) (b)

Figure 8: ROS based Navigation stack of MARIS, (a)

hardware, (b) global map.

7 CONCLUSIONS

In order to perform the representative set of tasks in a

one-floor home environment, the MARIS robot

requires mobility features such as obstacle avoidance

and collision free navigation. One of the objectives of

the MARIS research project is to design a minimally

complex robot that is able to perform the tasks in a

robot-inclusive environment, and as such this research

has involved conducting an experimental study to

determine which SLAM algorithm to best adopt in the

MARIS robot.

Initially, three RIS environments were considered

to implement and evaluate two LiDAR-based SLAM

algorithms, GMapping and Hector SLAM. The

computational intensity and map quality were

compared to select one SLAM that is most suitable for

these RIS environments. Hector SLAM was found to

be the more suitable.

Later a vision-based SLAM was also investigated

and implemented, as this can generate a 3D perception

or map of the environment. But with the increase in

the quality of the map, the computational intensity also

increases, which violates our objective of designing a

minimal complex robot. So, Hector SLAM was the

chosen algorithm. We thus tested the MARIS robot in

our RSL environment to tune and establish the optimal

Hector SLAM mapping parameters for our RIS

environments. In this work, it has been shown that

when the environment has been set up to

accommodate a robot, i.e. as a RIS, then LiDAR-based

Hector SLAM can match the functionality of the much

more computationally demanding LiDAR-based

GMapping or Visual-based SLAM methods. Based on

the selected Hector SLAM method with tuned

parameters, a new ROS based navigation stack was

implemented for the MARIS robot, for further

experimentation and development.

The mapping of the environment was found to

have inaccuracies if the environment is featureless or

has limited features, and there was also some

divergence from the real map because of the

accumulated inaccuracies. Thus, in a RIS environment

that facilitates the use of the robot, there should be

features that are recognizable. However, such features

may create a cluttered environment that may further

cause problems for the robot to navigate, and thus a

compromise must be found between these two

conflicting aspects. Ongoing work involves the

development of a system to optimize the number of

features that an environment should have to obtain a

better mapping that enables the robot to navigate with

fewer inaccuracies. The environment should obey

basic RIS principles, such as observability,

accessibility and manipulability.

For future developments, the MARIS robot is

envisioned to also have autonomous pick and place

capabilities using existing computer vision

technologies that can detect and track the object.

Since the MARIS robot is intended to serve the

MARIS

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

178

elderly and the impaired in performing household

tasks, it is also intended to implement voice-based

manipulation capabilities to provide the robot with

wider functionality.

REFERENCES

Ahmad, N., Ghazilla, R. A. R., Khairi, N. M., & Kasi, V.

(2013). Reviews on various inertial measurement unit

(IMU) sensor applications. International Journal of

Signal Processing Systems 1, no.2, pp. 256-262.

Allen, I. E., & Seaman, C. A. (2007). Likert scales and data

analyses. Quality progress 40, no.7, pp. 64-65.

Alsadik, B., & Karam, S. (2021). The simultaneous

localization and mapping (SLAM)-An overview. Surv.

Geospat. Eng. J, 2, pp. 34-45.

AMCL parameters, (2023). [Online]. Available:

http://wiki.ros.org/amcl.

Aquilina, Y., Saliba, M. A., & Fabri, S. G. (2019). Mobile

Assistive Robot in an Inclusive Space: An Introduction

to the MARIS Project. In Social Robotics: 11th

International Conference, ICSR 2019, Proceedings 11

pp. 538-547.

ASIMO robot-The world’s most advanced robot, (2023).

[Online]. Available: https://asimo.honda.com/.

Boston Dynamics - Spot, (2023). [Online]. Available:

https://www.bostondynamics.com/products/spot.

Debeunne, C., & Vivet, D. (2020). A review of visual-

LiDAR fusion based simultaneous localization and

mapping. Sensors 20, no.7, p. 2068.

Elara, M. R., Rojas, N., & Chua, A. (2014). Design

principles for robot inclusive spaces: A case study with

Roomba. In 2014 IEEE International Conference on

Robotics and Automation (ICRA), pp. 5593-5599.

Gazebo Building Editor, (2023). [Online]. Available:

https://www.classic.gazebosim.org/tutorials?cat=build

_world&tut=building_editor.

Intel RealSense 435i depth camera, (2023). [Online].

Available: https://www.intelrealsense.com/depth-

camera-d435i/.

Kamarudin, K., Mamduh, S. M., Shakaff, A. Y. M., &

Zakaria, A. (2014). Performance analysis of the

microsoft kinect sensor for 2D simultaneous

localization and mapping (SLAM) techniques. Sensors

14, no.12, pp. 23365-23387.

Kam, H. R., Lee, S. H., Park, T., & Kim, C. H. (2015). Rviz:

a toolkit for real domain data visualization.

Telecommunication Systems 60, pp. 337-345.

Labbé, M., & Michaud, F. (2019). RTAB ‐ Map as an

open‐source lidar and visual simultaneous localization

and mapping library for large‐scale and long‐term

online operation. Journal of field robotics, 36, no.2, pp.

416-446.

Map_server, (2023). [Online]. Available:http://wiki.ros.

org/map_server.

Matias, L. P., Santos, T. C., Wolf, D. F., & Souza, J. R.

(2015). Path planning and autonomous navigation using

AMCL and AD. In 2015 12th Latin American Robotics

Symposium and 2015 3rd Brazilian Symposium on

Robotics (LARS-SBR), pp. 320-324.

Move_base package, (2023). [Online]. Available:

http://wiki.ros.org/move_base.

Naraharisetti, P. R., Saliba, M. A., & Fabri, S. G. (2022).

Towards the Quantification of Robot-inclusiveness of a

Space and the Implications on Robot Complexity. In

2022 8th International Conference on Automation,

Robotics and Applications (ICARA), pp. 39-43

Polaris-Markert-Research. Household Robots Market

Share, Size, Trends, Industry Analysis Report, (2023).

[Online].Available:https://www.polarismarketresearch

.com/industry-analysis/household-robots-market.

Publishing Odometry information over ROS, (2023).

[Online].Available:http://wiki.ros.org/navigation/Tutor

ials/RobotSetup/Odom.

Revanth, C. M., Saravanakumar, D., Jegadeeshwaran, R.,

& Sakthivel, G. (2020, December). Simultaneous

Localization and Mapping of Mobile Robot using

GMapping Algorithm. In 2020 IEEE International

Symposium on Smart Electronic Systems (iSES)

(Formerly iNiS), pp. 56-60.

Saat, S., Abd Rashid, W. N., Tumari, M. Z. M., & Saealal,

M. S. (2020). Hectorslam 2d mapping for simultaneous

localization and mapping (slam). In Journal of Physics:

Conference Series, Vol. 1529, No. 4, p. 042032.

Särkkä, S., Vehtari, A., & Lampinen, J. (2007). Rao-

Blackwellized particle filter for multiple target

tracking. Information Fusion 8, no.1, pp. 2-15.

Seidita, V., Lanza, F., Pipitone, A., & Chella, A. (2021).

Robots as intelligent assistants to face COVID-19

pandemic. Briefings in Bioinformatics 22, no.2,

pp.823-831.

Servi robot from bear robotics, (2023). [Online]. Available:

https://www.bearrobotics.ai/restaurants.

Setup and Configuration of the Navigation Stack on a

Robot, (2023). [Online]. Available:

http://wiki.ros.org/navigation/Tutorials/RobotSetup.

SLAMTEC RpLiDAR, (2023). [Online]. Available:

https://www.slamtec.com/en/LiDAR/A1.

Sosa, R., Montiel, M., Sandoval, E. B., & Mohan, R. E.

(2018). Robot ergonomics: Towards human-centred

and robot-inclusive design. In DS 92: Proceedings of

the DESIGN 2018 15th International Design

Conference, pp. 2323-2334.

“transform-“tf”, (2023). [Online]. Available: http://wiki.

ros.org/tf.

Wu, M., Cheng, C., & Shang, H. (2021). 2D LIDAR SLAM

Based On Gauss-Newton. In 2021 International

Conference on Networking Systems of AI (INSAI), pp.

90-94.

Wang, H., Yu, Y., & Yuan, Q. (2011, July). Application of

Dijkstra algorithm in robot path-planning. In 2011

second international conference on mechanic

automation and control engineering, pp. 1067-1069.

Yang, X. (2021). Slam and navigation of indoor robot

based on ROS and lidar. In Journal of physics:

conference series, Vol. 1748, No. 2, p. 022038.

Mapping, Localization and Navigation for an Assistive Mobile Robot in a Robot-Inclusive Space

179