AUTONOMOUS ROBOTIC WHEELCHAIR

Authors
ALI AMMAR NAQVI

10-NUST-BE-ME-19

KOMAL ZULFIQAR

10-NUST-BE-ME-101

MUHAMMAD AQEEL ASHRAF

10-NUST-BE-ME-50

Supervisor
Dr.YASAR AYAZ
HoD, Robotics & Artificial Intelligence

DEPARTMENT OF MECHANICAL ENGINEERING

SCHOOL OF MECHANICAL& MANUFACTURING ENGINEERING
NATIONAL UNIVERSITY OF SCIENCES AND TECHNOLOGY
ISLAMABAD
JUNE, 2014

Autonomous Robotic Wheelchair

Authors

ALI AMMAR NAQVI

10-NUST-BE-ME-19

KOMAL ZULFIQAR

10-NUST-BE-ME-101

MUHAMMAD AQEEL ASHRAF

10-NUST-BE-ME-50

A thesis submitted in partial fulfillment of the requirements for the degree of

BS Mechanical Engineering

Thesis Supervisor:

Dr. YASAR AYAZ

Thesis Supervisors Signature:_____________________________________

DEPARTMENT OF MECHANICAL ENGINEERING

SCHOOL OF MECHANICAL & MANUFACTURING
ENGINEERING,
NATIONAL UNIVERSITY OF SCIENCES AND TECHNOLOGY,
ISLAMABAD
JUNE, 2014

Declaration
We certify that this research work titled Autonomous Robotic Wheelchair is our own work. The
work has not been presented elsewhere for assessment. The material that has been used from other
sources has been properly acknowledged / referred.

Signature of Students

ALI AMMAR NAQVI

10-NUST-BE-ME-19

KOMAL ZULFIQAR
10-NUST-BE-ME-101

MUHAMMAD AQEEL ASHRAF

10-NUST-BE-ME-50

Language Correctness Certificate

This thesis has been read by an English expert and is free of typing, syntax, semantic, grammatical
and spelling mistakes. Thesis is also according to the format given by the university.

Signature of Students

ALI AMMAR NAQVI

10-NUST-BE-ME-19

KOMAL ZULFIQAR

10-NUST-BE-ME-101

MUHAMMAD AQEEL ASHRAF

10-NUST-BE-ME-50

Signature of Supervisor
Dr.YASAR AYAZ
HoD, Robotics & Artificial Intelligence

ii

Copyright Statement

Copyright in the text of this thesis rests with the student authors. Copies (by any process)
either in full, or of extracts, may be made only in accordance with instructions given by the
authors and lodged in the Library of NUST, SMME. Details may be obtained by the
Librarian. This page must form part of any such copies made. Further copies (by any
process) may not be made without the permission (in writing) of the authors.

The ownership of any intellectual property rights which may be described in this thesis are
vested in NUST, SMME, subject to any prior agreement to the contrary, and may not be
made available for use by third parties without the written permission of the School of
Mechanical & Manufacturing Engineering (SMME), which will prescribe the terms and
conditions of any such agreement.

Further information on the conditions under which disclosures and exploitation may take
place is available from the Library of NUST SMME, H-12, Islamabad.

iii

Acknowledgements
We are thankful to our Creator Allah Subhana-Watala for having guided us throughout this project
at every step and for every new thought that He setup in our minds to improve it. Indeed we could
have done nothing without His priceless help and guidance. Whosoever helped us through the
course of our thesis, whether our parents or any other individuals was His will, so indeed none be
worthy of praise but Him.
We are profusely thankful to our beloved parents who raised us when we were not capable of
walking and continued to support us throughout in every department of our lives.
We would also like to express special thanks to our supervisor Dr. YASAR AYAZ for his help
throughout our thesis and also for Intro to Mechatronics & Robotics course which he has taught
us. We can safely say that we haven't learned any other engineering subject in such depth than the
one which he has taught.
We would also like to pay special thanks to Lab Engineer Khawaja Fahad Iqbal for his tremendous
support and cooperation. Each time we got stuck in something, he came up with the solution.
Without his help we wouldnt have been able to complete our thesis. We appreciate his patience
and guidance throughout the whole thesis.
We would also like to thank Dr. Omer Gillani and Dr. Mohsin Jamil, for being on our thesis
guidance and evaluation committee and express our special Thanks to Lab Engineer Ahmed
Hussain Qureshi for his help. We are also thankful to Ali Bin Wahid, Usama Siraj and Muhammad
Affan for their support and cooperation.
Finally, we would like to express our gratitude to all the individuals who have rendered valuable
assistance to our study.

iv

Dedicated to our exceptional parents and adored siblings whose

tremendous support and cooperation led us to this wonderful
accomplishment

Abstract
The Autonomous Robotic Wheelchair is being designed to enhance the capabilities of the
joystick-controlled SAKURA Wheelchair made available through SAKURAs collaboration with
RISE Lab, SMME. While an electric wheelchair can be successfully navigated by most persons,
people with Parkinsons, Multiple Sclerosis or other severe mental difficulties remain unable to
navigate a standard automated wheelchair in the same way. We have endeavored to design the
Autonomous Robotic Wheelchair so that these people can rely on our smart wheelchair to do the
navigation for them. Our Final Year Project is a small step towards this ambitious and important
goal being pursued by a variety of student teams working under the umbrella of the Mobile
Robotics Group at RISE Lab, SMME. Our project goals required that the wheelchair should be
able to perceive the layout of its surroundings, localize itself within that layout, and given
particular start and endpoints, it should be able to steer itself from one to the other following a
reasonable collision-free path. To this end, we analyzed the hardware configurations of various
other autonomous wheelchairs and adopted the hardware configurations which appealed to us the
most. We modified the existing wheelchair hardware to incorporate an LRF sensor and a tableunit for an on-board laptop and electronics housing. Following that, we developed our obstacle
avoidance and navigation codes based on C++ and Player\Stage. Our obstacle avoidance codes are
making use of real-time laser sensing by the LRF and our navigation codes are being controlled
either by the potential field navigation algorithm or by the initial position and headings provided
by the Player\Stage cfg files. Our obstacle avoidance algorithms are being monitored by
Player\Stage running on the Ubuntu open-source platform which conveys LRF inputs to our C++
code and our C++ code, in turn, sends the forward and turning velocities to our windows-based
LABVIEW program over a secure TP-Link Wireless Network. Once our Wheelchair becomes
fully functional, probably after a series of coding modifications, integration of further types of
sensors, rigorous lab-testing and optimization for environments bigger than the RISE Lab test
space, we will then endeavor to prototype and market this wheelchair as a healthcare product meant
to provide mobility and enhanced independence to the disabled and mentally challenged people
we

hope

to

Key Words: Autonomous Wheelchair, Obstacle Avoidance, Navigation, Player\Stage

vi

assist.

Table of Contents

Declaration ....................................................................................................................................................................i
Language Correctness Certificate ............................................................................................................................. ii
Copyright Statement ................................................................................................................................................. iii
Acknowledgements .....................................................................................................................................................iv
Abstract .......................................................................................................................................................................vi
Table of Contents ................................................................................................................................................ vii,viii
List of Figures ............................................................................................................................................................ xx
List of Tables ...............................................................................................................................................................xi
CHAPTER 1: INTRODUCTION...............................................................................................................................2
1.1

Background, Scope and Motivation .............................................................................................................2

1.2

Hardware .....................................................................................................................................................2

1.2.1

Wheelchair MC-2000 ..........................................................................................................................4

1.2.2

Laser Range Finder .............................................................................................................................5

1.2.3

NI-DAQ ..............................................................................................................................................6

1.3

Software .......................................................................................................................................................6

1.3.1

Ubuntu ................................................................................................................................................6

1.3.2

Player/Stage ........................................................................................................................................6

1.3.3

NetBeans .............................................................................................................................................6

1.3.4

LabView .............................................................................................................................................6

CHAPTER 2: DESIGN & FABRICATION................................................................................................................

2.1

Wheelchair Design .........................................................................................................................................

2.1.1

Motivation...........................................................................................................................................4

2.1.2

Proposed Model ..................................................................................................................................5

2.2

Fabrication ...................................................................................................................................................5
2.2.1

Materials and Processes ......................................................................................................................5

2.2.2

LRF Mount .........................................................................................................................................6

2.2.3

Laptop Table .......................................................................................................................................5

CHAPTER 3: WHEELCHAIR CONTROL ...............................................................................................................

3.1

Joystick Control .............................................................................................................................................

3.2

Shifting Control ...........................................................................................................................................5

3.3

LabView Control .........................................................................................................................................6

3.4

Control Graph Analysis ...............................................................................................................................5

CHAPTER 4: OBSTACLE AVOIDANCE .................................................................................................................

4.1

Literature Review ..........................................................................................................................................

4.2

Pseudo Code ...............................................................................................................................................4

4.3

Integrating LRF and Player/Stage ...............................................................................................................5

vii

4.4

Simulations .................................................................................................................................................5

CHAPTER 5: NAVIGATION ......................................................................................................................................

5.1

Literature Review ..........................................................................................................................................

5.2

Potential Field Algorithm ...........................................................................................................................4

5.3

BrushFire Pseudo code ...............................................................................................................................5

5.4

BrushFire Simulations ................................................................................................................................5

5.5

Runtime Navigation ....................................................................................................................................5

CHAPTER 6: NETWORKING....................................................................................................................................
6.1

Introduction ...................................................................................................................................................

6.1.1

Wireless Networking ............................................................................................................................4

6.1.2

Advantages of wireless Networking .....................................................................................................5

6.2

Scope ..........................................................................................................................................................4

6.3

Literature Review .......................................................................................................................................5

6.4

Server/Client ...............................................................................................................................................5

6.4.1

Sockets .................................................................................................................................................5

6.4.2

IP Adresses ..........................................................................................................................................5

6.4.3

Byte Order ...........................................................................................................................................7

6.5

Applications ..................................................................................................................................................4

CHAPTER 7: TESTING & OPTIMIZATION ...........................................................................................................

7.1

Testing Limitations .......................................................................................................................................

7.2

Obstacle Avoidance Testing .......................................................................................................................5

7.2.1

Single Obstacle Avoidance .................................................................................................................6

7.2.2

90 Avoidance .....................................................................................................................................4

7.2.2

Variable Angle Avoidance ..................................................................................................................5

7.3

Dynamic Obstacle Avoidance.....................................................................................................................5

7.4

Navigation Testing ......................................................................................................................................5

CHAPTER 8: CONCLUSIONS & RECOMMENDATIONS ...................................................................................

8.1

Conclusion ....................................................................................................................................................
8.2

8.2

Learning Outcomes ...............................................................................................................................4

Future Recommendations ...........................................................................................................................5

APPENDIX A............................................................................................................................................................. 45
REFERENCES .......................................................................................................................................................... 49

viii

ix

List of Figures
Fig 1.1: SAKURA Electric Wheelchair MC 20001
Fig 1.2: LRF Sensor...3
Fig 1.3: NI 6009 USB3
Fig 1.4: Player\Stage User Interface..5
Fig 1.5: LabVIEW Graphical Interface ....6
Fig 2.1: Wheelchair Mechanical Design Features.7
Fig 2.2: PRO\E Model for the Wheelchair8
Fig 2.3: Final Hardware Representation of the Wheelchair..9
Fig 2.4: Wheelchair Fabrication Materials 10
Fig 2.5: Wheelchair Fabrication Processes 11
Fig 2.6: LRF Sensor Mount..13

List of Tables

Table 2-1: Geometrical Data of IST Turbine [90] ........................................................................................................6

xi

CHAPTER 1: INTRODUCTION
1.1

Background, Scope and Motivation

The Autonomous Robotic Wheelchair is one of the flagship research projects of the Modular
Robotics Group at RISE Lab, SMME. We are working on the SAKURA automated wheelchair
MC-2000 and incorporating sensors onto the wheelchair in order to enable intelligent motion.

For the purpose of our Final Year Project, we have worked with an LRF sensor and used it to
enable the wheelchair to perceive the obstacles in its environment and to navigate its path around
those obstacles. Other undergraduate and graduate groups working on the Autonomous Robotic
Wheelchair are working with a large variety of sensors such as the webcams and headsets etc.

For us, the motivation for working on the Autonomous Wheelchair project was the opportunity
to work on something that is at the cutting edge of robotics research and simultaneously a
comprehensive engineering challenge in itself. Not only that the autonomous wheelchair is also a
deeply humanitarian project as well for it seeks to empower and mobilize disabled people
suffering from Parkinsons, Multiple Sclerosis and Tetraplegia etc.

The scope of our project is to enhance the capabilities of the SAKURA wheelchair in order to
stimulate high maneuverability and navigational intelligence for the wheelchair. It is desired that
the wheelchair should be able to perceive the layout of its surrounds, localize itself within that
layout and given particular start and end points, the wheelchair should be able to steer a
reasonable collision-free pathway.

Fig 1.1: Sakura Electric Wheelchair MC 2000

1.2

Hardware

1.2.1 SAKURA WHEELCHAIR MC-2000

We have used the SAKURA Wheelchair MS-2000. The wheelchair is currently joystick
controlled with two on-board batteries connected in series providing net 24V to the wheelchair.
The front two wheels of the wheelchair are passive wheels and the back two wheels are powered
with a differential drive. The joystick provides velocity values to the wheelchair in the form of
forward or angular velocities. The wheelchair also has an efficient clutch system and a safety
breaking mechanism which comes into action if the wheelchair is not started properly.

1.2.2 LRF SENSOR

We have used the SICK LMS-200 LIDAR Laser scanner and interfaced it with UBUNTU and
Player\Stage. The cone angle for LMS-200 is 180 and it has a measurement range of 80m with
10mm resolution and +/-15mm systematic error. The LRF weighs approximately 4.5kg and is
210mm 156mm 155mm in size. In order to work, the LRF sensor requires 24V supply
voltage and around 0.67Amps current. The LRF we have used was provided to us by the RISE
Lab. We have tested and installed onto the wheelchair the LRF sensor which showed the most
consistent LRF scan results.

Fig 1.2: LRF

Sensor

1.2.3

NI-DAQ

The NI USB-6009 provides basic DAQ functionality for applications

such as simple data acquisition, portable measurements and academic
experiments. The NI USB 6009 provides connections to eight singleended analog input (AI) channels, two analog output (AO) channels,
12 digital input/output (IO) channels, and a 32 bit counter with a full
speed USB interface. In order to use the NI USB-6009 we can use the
DAQ Assistant using LabVIEW in order to configure the virtual and
measurement channels which will drive our wheelchair.

Fig 1.3: NI 6009 USB

1.3

Software

1.3.1

UBUNTU

Ubuntu is an Open-Source Linux operating system. We have opted to

use this for the Autonomous Robotic Wheelchair project since it is
anticipated that the project will lead to ongoing high level research
and development in the field of autonomous motion planning and
navigation at the RISE Lab. The Version of Ubuntu we have worked
on for the purpose of our project is: Ubuntu 12.04. It offered regular
upgrades, software patches and downloadable applications which we
availed at various times during our project.

1.3.2 PLAYER\STAGE
The Player Project creates Open-Source software under the GNU Public License that enables
advanced research in robotics. It is the most widely used robot control interface in the world.

The Player robot device interface provides a network server for robot control. Player has a clientserver architecture which allows robot control programs to be written in any programming
language and to run on any computer that has a network connection to the robot. Your Player
clients can talk to Player over a TCP socket, reading data from sensors, writing commands to
actuators, and configuring devices on the fly.
The Stage is Players backend simulator. It simulates multiple
robots moving around concurrently and sensing a two dimensional
bitmap environment. The sensor models provided include sonar, LRF,
webcam and odometery. But Players modular architecture makes it
easy to add support for new hardware and commonly the control
programs written for Stages simulated robots work just as well on
real hardware as well.

Before implementing our algorithms on the wheelchair for trail-runs,

we tested them in Player\Stage simulations on a P3AT robot model.
Player has a variety of robot models available for simulations, a
decent amount of online learning materials and programming
examples, and an active user/developer community which continually
contributes new drivers. The following is a look at the Player User
Interface.

Fig 1.4: Player\Stage User Interface

1.3.3

NETBEANS

NetBeans is an Open-Source integrated development environment for

developing programs in Java, PHP, C/C++, and HTML5. The
NetBeans platform allows applications to be developed from a set of
modular software components called modules. NetBeans also offers
cross-platform support for both Windows and Linux-based OS and
has numerous plugins and APIs available.
1.3.4

LABVIEW

LabVIEW is a visual design and development environment from National

Instruments. Its programming language G is a dataflow programming
language using which we can develop programs or subroutines called
virtual instruments or VIs. LabVIEW has three main components: the
front panel, the block diagram, and the icon connector pane.
In LabVIEW, we can build the front panel user-interface using controls
and indicators. Controls are knobs or dials etc. while indicators are graphs,
LEDs or other displays. After we have built the user interface, we can
insert our code into the block diagram using VIs and structures to control
our front panel objects. Furthermore, LabVIEW communicates with
hardware for data acquisition, vision and motion control using devices
8

such as the NI 6009 USB which we have used, and LabVIEW has inbuilt
features for connecting our programs to the web using the LabVIEW web
server and software standards such as TCP/IP networking.
For our project, we have used LabVIEW in order to receive our
velocity values from across the TP-Link wireless network. We feed in our
laptops IP address to enable a Windows-based computer and Ubuntubased computer to network. We use NetBeans and TCP broadcasting to
repeatedly update our forward and angular velocity values for the
wheelchair. LabVIEW translates these into power values for the
wheelchairs differential drive and these variable power signals are then
sent to the wheelchair control using the NI DAQ USB. The following is a
snapshot of LabVIEWs
interface.

Fig 1.5: LabVIEW Graphical

Interface

CHAPTER 2: DESIGN & FABRICATION

The wheelchair given to us was provided by SKURA (Japan) and it was joystick controlled. In
order to make this wheelchair autonomous/intelligent, certain alteration in its design was needed.
So our final year project work distribution includes a major task which we named Design &
Fabrication.

2.1 Design
We were required to make a design which was totally detachable and could accommodate LRF
(Laser Range Finder) on it. Furthermore we were also required to design a laptop table in front of
the wheelchair user which could simultaneously serve as a mounting for sonar sensors or
webcams as well. This table also was also meant to have a shelf where different electronic
circuitry could be placed.
So we designed the following new features in our wheelchair:
i.

LRF sensor mount;

ii.

Laptop table (including sonar mount & electronic housing)

Our design has following three characteristics:

Light
Weight

Mechanically
Stable

Detachable

Fig 2.1: Wheelchair Mechanical Design Features

10

2.1.1 Motivation
We studied various designs of the autonomous wheelchairs that are being developed around the
world in different universities like:

The Intelligent Wheelchair Project @ MIT CSAIL

The SmartWheeler Robotic Wheelchair Project @ McGill

The Wheelchair Project @ Carnegie Mellon

Rolland Bremen Autonomous Wheelchair - Universitt Bremen

Fig 2.2: PRO\E Model for the Wheelchair Hardware

11

2.1.2 Proposed Model

The above diagram is the assembly of the wheelchair consisting of different separately made
components. We designed our proposed wheelchair model in Pro-E software. In the picture above,
the dotted region shows the newly designed portion of the wheelchair. The above drawn model
has exactly the same dimensions as of the original wheelchair. During modeling different complex
joints were simplified. This modeling gave us a rough idea about how it will look after fabrication
and this also helped us in reviewing our design again and again to make it more accurate. After
this all our detailed diagrams, we used for the next stage i.e. Fabrication.

2.2 Fabrication
For fabrication purposes, we

used the

workspace of

Manufacturing

Resource Center- SMME.

Three things were fabricated
i.

LRF Mount

ii.

Laptop Table

iii.

Extended arm rests

namely:
LRF Mount

(to support table)

Fig 2.3: Final Hardware Representation of the Wheelchair

2.2.1 Materials & Processes

Following materials were used during this fabrication:

Stainless
Steel

Aluminium

Cast Iron

Fig 2.4: Wheelchair Fabrication Materials

12

Wood

Following processes were used in fabrication:

Welding

Riveting

Drilling

Threading

Cutting

Fig 2.5: Wheelchair Fabrication Processes

2.2.2 LRF Mount

LRF Sensor
LRF Sensor Mount
(i)Aluminium Sheet
(ii) Cast Iron hollow pipe
(iii) Bush

Fig 2.6: LRF Sensor Mount

2.2.3 Laptop Table

Laptop Table
Laptop

Extended Arm Rests

Sonar Sensor Mount

Fig 2.7: Wheelchair Laptop Table

13

CHAPTER 3: WHEELCHAIR CONTROL

3.1 Joystick Control
Initially the wheelchair provided by SAKURA (Japan) is joystick controlled. The person sitting
on the wheelchair has a joy stick on his right side. The person has to
move joy stick in the direction in which he/she is intended to go. For
example if person pushes joy stick forward then wheelchair will move
in the forward direction. Similarly if person pushes joy stick backward
then wheelchair will move in the backward direction.
Joy stick can be move at any angle. The amount by which the joy stick
is moved is proportional to the amount of voltage signal send to the
internal circuitry and then that specified voltage signal is translated
into a specific speed. This wheelchair is rare wheel driven and it has
differential drive.
Fig 13: Joystick

3.2 Shifting Control : Joystick

LabVIEW

In order to make wheelchair autonomous, we shiftedits control from joystick to LabView. For
this purpose the wires were detached from the joystick and they were inserted into the NI-DAQ.
NI-DAQ is the real time USB developed by National Instruments and it connects hardware with
software. In this case hardware is our wheelchair and software is LabView.

Fig 14: Wheelchair Control Representation

Wheelchair

NI-DAQ

14

LabView

Fig 15: Shifting control form joystick to LabView

3.3 LabVIEW Control

Fig 16: Graphical LabVIEW Control

Description:
This LabView program was made to control wheelchair motion. Previously this was controlled
by joystick. A continuous while loop is implemented so that a required value is being sent to
wheelchair at all times. There is a stop button in the program by which code can be terminated at
any time. Actually an angular velocity and a forward velocity values are given to the program
15

then these values are converted to a specific voltage value. This conversion is done by two
functions; one made for forward speed and another for angular speed. This function is made by
using straight line equation i.e. y= mx + c on the velocity-voltage graphs (graph is explained in
the next section). These voltage values for forward and angular speed the then send to wheelchair
via NI-DAQ. Constant voltages of 1.85 volts are being provided to wheelchair when wheelchair
is stationary.

3.4 Control Graph Analysis

Following table shows our acquired observations during control shifting:

Voltage vs. Rpm vs. Velocity

Table 1: Voltage Vs. RPM Vs. Velocity for the Wheelchair

Voltages were provided by LabView software via NI-DAQ. RPM of the rare wheels were
measured by the tachometer. Then velocity was calculated on the basis of rpm.

16

Voltage-Velocity Graph
1.5

Velocity (m/s)

1
0.5
0
0

0.5

1.5

2.5

3.5

-0.5
-1
Voltage (V)
Fig 17: Voltage-Velocity Graph

The graph above shows the trend of voltage vs. velocity. We gave wheelchair different sets of
voltages through LabView and then measured the resultant speed of wheelchair by measuring
rpm of its rare wheels.
Following conclusions can be drawn from the above graph:

Maximum forward speed of wheelchair is 1.22 m/s @ 0 volts

Maximum backward speed of wheelchair is 0.75m/s @ 3.7 volts

Wheelchair is stationary i.e. zero velocity @ 1.85 volts

There is a voltage window where wheelchair does not move due to inertia and that is 1.25
volts 2.3 volts

Wheelchair maximum forward speed is greater than maximum backward speed

Graph between voltage and velocity is discontinued and approximately linear

The main purpose of drawing this graph was to know exactly about the speed output when a
specified voltage is provided. This knowledge helped during the developing of obstacle
avoidance and navigation codes and it also helped us during real-time wheelchair testing phase.

17

CHAPTER 4: OBSTACLE AVOIDANCE

4.1

Literature Review

Obstacle Avoidance is a means of a robot being able to move

around in an unknown environment without unknown obstacles
coming in its path. It is one of the most important aspects of mobile
robotics.

According to Seigwart et al, local obstacle avoidance focuses on

changing the robots trajectory as informed by its sensors during realtime robot motion. The resulting function is both a function of robots
current or recent sensor readings and its goal position and relative
location to the goal position.

In developing our obstacle avoidance algorithms for the Autonomous

Robotic Wheelchair, we were able to benefit significantly from the
work completed in lieu of our MRDS semester project for our Intro to
Mechatronics and Robotics class taught last semester by Dr. Yasar
Ayaz and Dr. Umar Gillani.

We studied and practiced coding for primarily the obstacle avoidance

algorithms presented in the Introduction to Autonomous Mobile
Robotics by Seigwart et al. For example:

1) The Bug Algorithm:

The Bug Algorithm is the simplest most intuitive obstacle avoidance
algorithm. The basic idea is to follow the contour of the obstacle in
order to circumnavigate around it. It has two different forms Bug1
and Bug2.

18

In Bug1 algorithm, the robot fully circles around the obstacle

and then moves along the line of shortest path towards the endpoint.
This is the obstacle avoidance technique we have been able to
implement in our code. With Bug2 algorithm, the robot starts to
follow the objects contour but departs immediately when it able to
move directly towards the object. This would have been more efficient
and future coding on the Autonomous Wheelchair could attempt to
develop this.

Fig 18: Bug Algorithms

19

2) The VFH Algorithm:

Developed by Bronstein & Koren, the VFH algorithm creates a local
map of the environment around the robot which translates onto an
occupancy grid populated by sensor readings. The VFH algorithm
generates a polar histogram, in which the X-axis represents the angle
alpha at which the obstacle was found, and the Y-axis represents the
probability grid that there is an obstacle in that direction based on the
occupancy grids cell values.
We were able to obtain and study a basic implementation of
the VFH algorithm in the example files for Player\Stage. However, we
did not pursue it further due to its redundancy with the potential field
path-planning algorithm which we were working on implementing at
the time. Also the there are now more efficient VFH+ and the VFH*
algorithms which have also been developed. There are a handful of
other promising obstacle avoidance algorithms presented in Seigwart
et al, however we were able to investigate and understand only the
ones discussed.

20

Fig 19: VFH

polar histogram

4.2

Pseudo-Code
From Jennifer Owens How to Use Player\Stage manual we
were able to obtain a basic obstacle avoidance function which we were
able to expand upon in different ways. Our obstacle avoidance
functions attempt to update motor speeds if there is an obstacle to be
avoided. We update the function just before sending data to the
motors. Thus, obstacle avoidance overwrites onto our navigation
behavior for the robot.

Once the obstacle is no longer in its way, the robot moves along as it
was. This allows us to transition from any navigation behavior into
obstacle avoidance and then back again, as per the requirements of
our control structure.

We wrote three distinct obstacle avoidance codes for the Autonomous

Robotic Wheelchair. These are discussed as follows:

1) Left-Right Obstacle Avoidance:

This was the first obstacle avoidance code that we developed. It was
based on the obstacle avoidance example code obtained from the
Player\Stage tutorial files. It created and initialized two variables
newspeed and newturnrate within a continuous for loop. Next, it
used the inbuilt Player\Stage LRF commands lp.GetMinRight() and
21

lp.GetMinLeft() in order to see to the left and right of the robot for
obstacles. The readings obtained are processed and we check to see if
either the left or the right distances to the obstacle are greater than
100. Based on that, we calculate robots newspeed and newturnrate
which is limited to values of -40 to +40. And use the SetSpeed
command to set these motor speeds.

Pseudocode for Left-Right Obstacle Avoidance:

Enable Motors
Initialize Newspeed & NewTurnRate
Obtain minR & minL
Check to see if obstacle is close enough to avoid?
If either distance is smaller; Turn the Other Way
Calculate NewSpeed using minR & minL
Calculate NewTurnRate using minR & minL
SetSpeeds
This code was implemented with the robot wandering in the
Cave cfg environment. It was not tested in real-time on the
wheelchair because we were providing the wheelchair with the path
beforehand from our navigation algorithm and we wanted primarily
that the wheelchair should be able to detect and avoid the obstacles in
its path rather than the obstacles coming to the left and right close to
its path. A snapshot of the Left-Right Obstacle avoidance simulation
has been provided in the Section 4.4.
2) GoTo-based Obstacle Avoidance:
This was the second Obstacle Avoidance code we developed. Our
GoTo-based Obstacle Avoidance Code countered the problems of the
Left-Right Obstacle Avoidance. However, we could not implement
this one in the actual lab trials as well.
22

In GoTo-based Obstacle Avoidance, we declared our Data variables

robotSpeed.forwardSpeed and robotSpeed.turningSpeed, we
initialized them to zero and set the motor speeds to their values. Then
inside a continuous for loop, we gave forward speed an initial value of
0.25 and zero turning speed to get the robot moving along its provided
path. The Robot.Read() command was used to update LRF values
obtained from the LRF simulation proxy and we read in the LRF
readings of minimum distances at angles 88 to 91 using the
lp.GetRange() command. Doing so checks to see if there is an obstacle
directly in front of the robot which if the robot was to continue
moving along its given path it would collide with. If the obstacle is in
front of the robot and the lp.GetRange() value of distance in front of
the obstacle comes out as less than 1.5 we make the robot turn around
an obstacle whose dimensions we assume to be known. So, if there is
an obstacle closer than 1.5m, the robot would immediately stop. It
would detect its X and Y coordinates at that point using the
robot.GetXPos() and robot.GetYPos() commands. Thereafter, it
would use those x and y position values to navigate around the
obstacle using a sequence of GoTo commands that take the robot out
of its path, enable it to cross the obstacle from its left and makes it
come back directly in the position it would have been moved along to
had there been no obstacle in its path. Lastly it checks to see if the
endpoint has been reached, at which point it would stop immediately.

Pseudocode for GoTo-based Obstacle Avoidance:

Enable Motors
Initialize robotSpeed.ForwardSpeed to some initial value
Initialize robotSpeed.TurningSpeed to zero
Use Robot.Read()

23

Use GetRange() to look in front of the robot

Check to see if obstacle in front close than set distance (1.5)?
If Yes, read robot coordinates at current position
Use GoTo commands with current position coordinates to move
around obstacle
Else Continue moving along original path
Check to see if EndPoint has been reached?
If Yes, Stop.

3) SetSpeeds-Based Obstacle Avoidance:

This was the Obstacle Avoidance code which we eventually used for
trial runs in the lab. Logically it follows the same sequence of
instructions as out GoTo-based Obstacle Avoidance code except that
it uses a set of SetSpeed commands in conjunction with various sleep()
calls in order to navigate around the obstacle instead of the GoTo
commands. It was important to develop this code because ultimately
we needed that the forward speed and turning speed values for the
robot should change against different time calls and we should be able
to pass these changing forward and turning speed values to drive our
wheelchair.

For this code, we again declare our data variables

robotSpeed.forwardSpeed and robotSpeed.turningSpeed, we
initialized them to zero and set the motor speeds to their values. This
is an important first step because not setting these speed values to zero
initially causes the wheelchairs automatic safety system to activate
causing the wheelchair to lock itself and give a beep sound.
Once weve passed zero values to the motors for some time, we give
the wheelchair an initial forward speed of 0.8. This is the minimum

24

initial speed for which the wheelchair would move. For power values
below this velocity level, the wheelchair does not move. Thereafter,
inside a continuous for loop we use Robot.Read() command to update
LRF values obtained from the LRF attached to the wheelchair.
Details of how we can set up Player\Stage in order to read in these
LRF values will be covered in the next section. As before, we read in
LRF readings of minimum distances at angles 88 to 91 using the
lp.GetRange() command. And we check to see if at some point all of
these GetRange() values would be less than 1.3 or some other specified
distance directly in front of our robot? If the obstacle is present, the
robot halts and reads its X and Y coordinates using the
robot.GetXPos() and robot.GetYPos() commands. Thereafter, it uses
those x and y positions to navigate around the obstacle using a
sequence of SetSpeed commands that take the robot out of its path,
make it cross the obstacle and makes it come back to its path after
having crossed the obstacle. Lastly, it checks to see if the endpoint has
been reached at which point the robot stops.
Pseudocode for SetSpeeds-based Obstacle Avoidance:

Enable Motors
Initialize ForwardSpeed & TurningSpeed to zero
Initialize ForwardSpeed to 0.8
Use Robot.Read()
Use GetRange() to look in front of the robot
Check to see if obstacle in front close than set distance (1.3)?
If Yes, read robot coordinates at current position
Use SetSpeed commands with current position coordinates to move
around obstacle
Else Continue moving along original path

25

Check to see if EndPoint has been reached?

If Yes, Stop.
In this Obstacle Avoidance code, the wheelchair takes 90 turns
around the obstacle. We have also written and tested a modified form
of the SetSpeeds based obstacle avoidance code. In this version,
instead of using 90 turns, the wheelchair moves around the obstacle
at some angle. We are using different values for sleep() to give
different values for the angle at which the wheelchair moves out of its
path. The advantage of this variation is that the wheelchair has to
cover less overall distance while it is avoiding the obstacle. But the
problem with this technique is that the wheelchair is frequently
unable to properly swerve around the obstacle. Depending on where
the LRF will actually detect the obstacle, the angle required for
obstacle avoidance will change from one trail run to the next. Our
code does not have the capability for calculating a changing obstacle
avoidance angle in this way every single time and this is one area
which calls for further coding in order to perfect the performance of
the Autonomous Robotic Wheelchair.

4.3

Integrating LRF and Player\Stage

In order to integrate the SICK LMS 200 with the Player\Stage

software for real time obstacle avoidance, one has to first execute a
series of terminal window instructions so that the LRF can be hooked
up with our Player code. The commands are as follows:
cd\dev

// create a device folder

ls l tty*

// displays pointers for all the root

commands
sudo chmod +r ttyUSB0

// enables read at port zero

sudo chmod +w ttyUSB0

// enables write at port zero

ls l ttyUSB0

// displays the output at USB Port zero

26

ls l ttyU*

// shows the USB outputs

sudo chmod a+w ttyUSB0 //access and write for USB0

ls l ttyU*
Along with following this sequence of instructions, we also
modify our cfg file so that instead of simulating an LRF sensor
player recognizes the actual LRF SICK LMS 200 and tries to take the
reads from it. We point the laser proxy to dev/ttyUSB0 in the cfg file
and set baud rate for sampling and resolution parameters.

Thereafter, we open a second terminal window and use the following

command:
playerv --laser --position2d
This command causes a stage window to open in which we can
visually see the output of our LRF sensor. This stage window can be
kept open and we can use a third terminal window to open a second
Stage window in which we can run our robot while were doing
obstacle avoidance in real time.

27

4.4

Simulations

In this section, we will discuss the simulations we generated using our

Lef-Right Obstacle Avoidance and GoTo-based Obstacle Avoidance
codes.

In Left-Right Obstacle Avoidance, the robot wanders around in the

cave environment and turns right or left based on the position of
obstacles. The
following is a snapshot
of the Left-Right
Obstacle Avoidance
code.

Fig 20: Left-Right Obstacle Avoidance

28

In the GoTo-based Obstacle

Avoidance, the robot moves along its path
and if it finds obstacles it navigates
around it and returns to keep moving
along its original path until endpoint is
reached. The following snapshots show
our GoTo-based obstacle avoidance
simulations for a straight axis and for an
inclined axis.

Fig 21: GoTo Based Obstacle

Avoidance in X and Y Axis

29

CHAPTER 5: NAVIGATION

5.1

Literature Review

According to Introduction to Autonomous Mobile Robots by Seigwart

et al, for a mobile robot, the specific aspect of cognition directly linked
to robust mobility is navigation competence. Given partial knowledge
about its environment and a goal position or a series of positions,
navigation encompasses the ability of a robot to act based on its
knowledge and sensor values so as to reach its goal positions as
efficiently and as reliably as possible.
Therefore, navigation refers to the robots ability to determine its own
position in its frame of reference and then to plan a path towards its
goal position. Thus, for navigation, the robot requires representation
i.e. a map of the environment and the ability to interpret that
representation. Navigation can be defined as a combination of three
fundamental competencies: (i) Self-Localization (ii) Path-Planning (iii)
Map-building and map interpretation.

Also there are two different types of Navigation:

(i) Global Path Planning: which deals with finding a suitable path
from a starting point to a goal point using a given representation of
the environment. We attempt to find the path with the lowest cost for
the journey.
(ii) Local Path Planning: defines path points taking into account the
vehicle dimensions and kinematic constraints. Thus, in local path
planning we are looking for a planned path with small deviations
which bring the robot back to its planned path.

In developing our navigation code we studied the various algorithms

presented in the Introduction to Autonomous Mobile Robots by
30

Seigwart et al. The following is a brief discussion of the algorithms

that we reviewed:

1) Breadth First Search:

The Breadth First Search algorithm begins at the start node and
explores all of its neighboring nodes. For each of these nodes it
explores all of its unexplored neighbors marking its nodes as either
visited or open. In this way the algorithm iterates until it reaches it
goal where it terminates. The breadth first search always returns the
path with the fewest number of edges between the start and end
nodes. If we assume that the cost of all individual edges in the graph is
constant than the breadth first search returns the optimal-cost path.

2) Depth First Search:

In contrast to the Breadth First Search, the Depth First Search
Navigation algorithm expands every single node to its deepest level, its
branches are removed and the search backtracks by expanding the
next neighboring node of the start node until its deepest level. Depth
First Search offers the advantage of space complexity, in that it stores
only the single path from the start node to the goal node.

31

5.2

Potential Fields Algorithm

For Potential Field navigation we start out by creating a two dimensional discretized
configuration space of X and Y coordinates. We can then attach a potential value to each of these
grid configurations and denote as a potential function or gradient spread over the entire 2D field
within which the robot moves.

The potential field method treats the robot as a point under the influence of an artificial potential
field U(q). The robot moves by following the field just as a ball were rolling downhill under the
action of gravity. The basic idea is that the endpoint is pulling an attractive potential onto the
robot and the obstacles within the robots environment are exerting a repulsive potential onto the
robot. Thereafter the potential field of the robot is computed as a sum of the attractive field of the
goal and the repulsive fields of the obstacles.
U(q) = Uattr(q) + Urep(q)

At every point in time, the robot will look at the potential field vector at that point and
start going in that direction. The attractive and repulsive potentials can be found using any
of the linear or quadratic formulae with various scaling factors for the endpoint attraction
and the repulsion by obstacles. Some of the problems associated with the potential field
algorithm are the Local Minima Problems in which robot is unable to navigate its way to the
endpoint of the repulsive potentials due to the obstacles chocking its pathway.

32

5.3

BrushFire PseudoCode

In order to implement the Potential Field Navigation method for our code, we implemented the
BrushFire Navigation scheme. The BrushFire algorithm is a clearance-based path finding
algorithm which overlays a grid onto the entire navigational space of the robot.
In the BrushFire algorithm weve represented the 2D navigational space of our robot as a grid of
pixels. For example, if weve used a 400 X 400 jpg image for our robot environment, we will
first break it down into a grid of blocks. Each of the block is then checked using nested iterative
for loop of pointers, if the pixels are occupied by obstacles we mark them with 1s and if theyre
free and the robot can travel to them we mark them as zero.

In order to compute the attractive potential of the endpoint we compute it as a function of the
distance of each pixel from the goal. On the other hand, the repulsive potential of the obstacles is
generated by traversing the map such that at each pixel visited the 8 neighbors connected to it are
updated to a value one greater than the value of the current pixel. Doing so we have a grid where
the occupied pixels are blocked out and the empty pixels close to the obstacles have higher grid
values than those adjacent to them and the pixels as they go towards the endpoint have a
decreasing grid value.

Fig 22: BrushFire Algorithm Grid

33

The architecture of our BrushFire Navigation program was made in the following way:

(i)

CreateMap:

the CreateMap program takes a jpg or a png image as its input and it overlays a grid onto the
image marking the occupied grid blocks as (1) and the unoccupied free space grid blocks as (0)
(ii)

BrushFireMap:

the BrushFireMap program generates the BrushFire grid. It uses nested for loop to access every
grid block and for each occupied grid block marked with (1) provided by the CreateMap grid it
updates the cells around those obstacles to show repulsion. Then it goes around to the cells
around those updated blocks and keeps updating values on the grid to show to the rippling effect
of the repulsion produced by the obstacles in the navigation space.
(iii)

DistanceMap:

the DistanceMap program is used to generate grid values to denote the attractive potential of the
endpoint for the robot. Once again we make use of nested for loops to iterate to all the grid
blocks and the blocks closer to the endpoint are marked with lower and lower values with the
endpoint having the value of zero and the starting point having the maximum value on the grid
(iv)

PotentialMap:

the PotentialMap program relies on the outputs of the BrushFireMap denoting the repulsion
potential and the DistanceMap denoting the attractive potential. The PotentialMap program uses
the grid output from both to compute a grid showing the net potential function at each point. The
potential function is computed as follows:
Potential Function = P * DistanceMap[x][y] + N*pow(1/BrushFireMap[x][y], 3)
The values of P and N are varied by looking at the results of different simulation trials. Usually P
is around 20 and N is on the order of 10^5
(v)

RunBrushFireMap

the RunBrushFireMap program is the ultimate client program of CreateMap, BrushFireMap,

DistanceMap, and PotentialMap. Based on the output of the PotentialMap program, the
runBrushFireMap causes the robot to move from its start point to its endpoint. For our simulation
we have used a for loop to get the robot moving along its PotentialMap path of shortest cost from
the start to end. At each grid point, it checks to see if endpoint has been reached, if not it keeps
moving along its PotentialMap shortest cost path.
34

5.4

BrushFire Simulations

We have tested the BrushFire in Player\Stage simulations only. We were able to obtain
successful test trails in some of the BrushFire simularions but not all. Some of the problems
arose to the programming complexity for the algorithm and due to the local minima problems of
the BrushFire algorithm.

The following are screen shorts of our BrushFire Simulations which enables the robot to move
from start point (provided by the cfg file) to endpoint (specified in the code). We started out by
using simpler jpg images for testing as shown:

Fig 23: BrushFire Navigation

Eventually we progressed to more complicated scale model representations of our RISE Lab
workspace and more realistic model-type obstacles placed within that workspace. In almost all of
these simulations the robot was able to determine the shortest cost path to its end point specified.
However, perhaps making the obstacles as models within the cfg file or for some other reason
the BrushFire algorithm almost always crashed during more challenging test runs.

35

5.5

Run-Time Navigation
We were not able to use our BrushFire programs for real runtime navigation. There were not only logical problems in
implementation of the algorithm but also the test constraints which
became a significant hurdle in the way. For example, we need to
provide forward and angular velocities in order to physically drive
our wheelchair using the NI 6009 USB. However, the BrushFire code
does not navigate based on velocity variables, it navigates based the
relative grid values from start point to end point. Logically we could
not translate the output of the latter into giving velocity values to the
power drive of our wheelchair. Another significant problem was the
small test space and the limited scope of obstacle avoidance and
navigation we could afford within that test space as well as the front
two wheels of the wheelchair being idle and causing significant
deviations betweens the expected and actual wheelchair motion.

In view of these and other problems, we simplified our navigation

scheme for the wheelchair based on a solution proposed by Lab
Engineer Ahmed Hussain Qureshi. We used our Player\Stage cfg file
as our means for simple navigation. Using our cfg file, the robot was
given an initial start position and an initial angular heading in order
to move the robot along that heading. The robot was thus able to
move along the line we wanted it move along, meanwhile our code was
used to provide an end point along the length of the line the robot
travels along. Once the robot has reached that specified point along
the line, the robot motion terminates.
Since most of our testing on the wheelchair required for the chair to
move along a straight line while avoiding obstacles in its path, we
were able to successfully use this navigation technique in our
simplified simulations and trail runs in the lab.

36

CHAPTER 6: NETWORKING

6.1

Introduction

networking between computers allows communication between different users and programs,
thus enabling sharing of information between computers located at different locations. The
information shared on networks is accessible to all the computers connected on the network.
There are two main types of Computer networks,
(I)

Wireless Network

(II)

Wired Network
We used Wireless Networking for our project as it enabled us to communicate with the

wheelchair from a distance without any wired connection, therefore enabling free motion of the
Wheelchair without any hindrance.
6.1.1 Wireless Networking
Wireless networking is a flexible data communications system, which uses wireless media
such as radio frequency technology to transmit and receive data over the air, minimizing the
need for wired connections. Such networks are used to replace wired networks and are most
commonly used to provide last few stages of connectivity between a mobile user and a wired
network.

Fig. 24: Wireless Networking

37

Wireless networks use electromagnetic waves to communicate information from one point to
another without relying on any physical connection. Radio waves are often referred to as radio
carriers because they simply perform the function of delivering energy to a remote receiver.
The data being transmitted is superimposed on the radio carrier so that it can be accurately
extracted at the receiving end. Once data is superimposed (modulated) onto the radio carrier,
the radio signal occupies more than a single frequency, since the frequency or bit rate of the
modulating information adds to the carrier. Multiple radio carriers can exist in the same space
at the same time without interfering with each other if the radio waves are transmitted on
different radio frequencies. To extract data, a radio receiver tunes in one radio frequency while
rejecting all other frequencies. The modulated signal thus received is then demodulated and
the data is extracted from the signal.
6.1.2 Advantages of Wireless Networking
Wireless networks offer the following advantages over traditional wired Networks:

Mobility: They provide mobile users with access to real-time information so that they can
roam around in the network without getting disconnected from the network. This mobility
supports productivity and service opportunities not possible with wired networks.

Installation speed and simplicity: Installing a wireless system can be fast and easy and can
eliminate the need to pull cable through walls and ceilings.

Reach of the network: The network can be extended to places which cannot be wired

More Flexibility: Wireless networks offer more flexibility and adapt easily to changes in the
configuration of the network.

Reduced cost of ownership: While the initial investment required for wireless network
hardware can be higher than the cost of wired network hardware, overall installation
expenses and life-cycle costs can be significantly lower in dynamic environments.

Scalability: Wireless systems can be configured in a variety of topologies to meet the needs
of specific applications and installations. Configurations can be easily changed and range
from peer-to-peer networks suitable for a small number of users to large infrastructure
networks that enable roaming over a broad area.

38

6.2

Scope

In this project we used two computers operating in parallel. The first computer had Ubuntu as an
Operating System with Player/Stage and NetBeans. The working algorithm gave the velocity
output stored in variable form that was to be transmitted. We used wireless communication to
transmit these velocity values from the navigation algorithm to the LabView operated computer.
This LabView code further translated these values into voltage outputs, thereby enabling the
accurate movement of the wheelchair

Velocity
6.3

Server

Client

Literature Review

The Networking task required a lot of literature reading as the coding for creating a network and
data transfer was a fairly new for us. This involved the study of standard networking tutorials
and manuals. Following are some of the Tutorials that were used for Networking.
(I)

Beej Networking tutorial

(II)

Lantronix Networking tutorial

6.4

Server/Client

The Network communication is based on a server/client system. We used the server to transmit
float type data to the client by casting it into a binary value. The Client interpreted this data type
in binary and then re-calculated the values in float type data.
6.4.1 Sockets
The first step for creating a server was to create a parent socket. There are two types of sockets
Stream sockets and Datagram sockets. We used Stream Sockets for our server because Stream
sockets are reliable two-way connected communication streams. If we two output items into the
socket in the order 1, 2, they will arrive in the order 1, 2 at the opposite end and they will be
error free.

39

6.4.2 IP addresses
The IP address of the parent socket connection to the server was known. It was in the form of
IPv4. This IP address was used by the client program to connect with the server on the network
and pick up the data.

6.4.3 Byte order

The computer stores data bytes in two different orders.
(I)

Little-Endian: When representing a two byte digit like b34f, this type of byte order
will store the sequential bytes 4f followed by b3 in the memory. This storage method
is called Little-Endian.
Big-Endian: When we want to represent a two byte like b34f, then well need to

(II)

store it in two sequential bytes b3 followed by 4f. This number, stored with the big
end first, and the small end last is called the Big-Endian order

Thus we understand that for Intel microprocessors that use little-Endian, The computer may have
been storing bytes in reverse order. To counter this problem we cast the element in binary order
and use the function network byte order that makes sure that the byte order remains the same in
both computers. We make use of the long type variables that enable four byte storage.

6.5

Application

The whole networking process can be summed up in the following points:

(I)

Setting up a Server

(II)

Making a Responsive Client

(III)

Establishing a wireless Network for communication

(IV)

Sending and Receiving Data

40

CHAPTER 7: TESTING AND OPTIMIZATION

7.1

Testing Limitations

7.2

Obstacle Avoidance Testing

We performed the obstacle avoidance testing on the wheelchair
within the RISE Lab test space. The wheelchair was set up by
connecting the LRF to its power cords hooked up to the wheelchairs
battery. The USB Port from the LRF was connected to the Port0 of
Laptop 1 running Ubuntu and Player\Stage which was setup in order
to enable it to read LRF readings in real time. Meanwhile Laptop 2
running LabVIEW was set up onto the wheelchair table and this
laptop was connected wireless with the other one over the TP-Link.
Laptop 2 was also hooked up to the joystick control of the differential
drive of the wheelchair via the NI 6009 USB. We check the wheelchair
emergency brake system once and run our code on Player\Stage and
run the LabVIEW code once a connection has been established
between the two computers. The follow are the descriptions of the
main types of Obstacle Avoidance trails we performed:

7.2.1 Single Obstacle Avoidance

For our trail run for Single Obstacle Avoidance we used our SetSpeeds based Obstacle
Avoidance algorithm. The algorithm made the wheelchair halt for a few seconds initially in order
allow a network to setup and also so as not to activate the emergency brake mechanism of the
wheelchair. After a while, the wheelchair is provided forward velocity of 0.8 which is the
minimum velocity required for driving the wheelchair. Once the wheelchair detects obstacle at
1.3m or closer directly in front of it, the wheelchair notes its coordinates and uses them to move
around the obstacle in front whose dimensions we have assumed to be known.
7.2.2 90 Avoidance
Our Single Obstacle Avoidance is of two types both based on SetSpeeds Obstacle
avoidance. One is 90 avoidance which makes use of 90 angles in order to move around the
41

obstacle. After detecting the obstacle the wheelchair moves out of its path horizontally, it
crosses the obstacles at a specified distance from it and comes back to its original path line
horizontally after making yet another 90 turn. Problems arise during 90 Obstacle
Avoidance when the front wheels behave differently and trace a different angle (not
equivalent to 90) from one trial run to the other. A possible solution for this is to keep the
testing conditions as similar as we can so that test results can be repeated based on our time
calibrations. Video 1 provided shows Single Obstacle Avoidance with 90 turns.

7.2.3 Variable Angle Avoidance

Variable Angle Avoidance is essentially the same thing as 90
avoidance except that it makes the wheelchair avoid the obstacle in
front at an angle. Consequently, the wheelchair travels less distance
than it does in case of 90 avoidance. As before, we again use
SetSpeeds based avoidance but this time, instead of using time calls
for long enough to allow the wheelchair to make 90 angles we only
allow it to make smaller 45-60 angles and to go out of the path of the
obstacle and to come back to the original path in the same way as
before.

There is a major problem with implement angular obstacle avoidance

using our code, based on the point at from which the wheelchair
detects the obstacle in front of it. Its turning angle constantly varies
from one trial run to the next. This point keeps changing because of
the LRF scan variations and the inaccuracies caused by the front
wheels, thus one trail run may not lead to a second successful trial run
and we need to rely on hit-and-trial to get this to work for any specific
test condition. Variable Angle Avoidance is shown in Video 3.

42

Fig 25:
90 degrees and Angular Obstacle Avoidance

7.3

Multiple Obstacle Avoidance

Our Multiple Obstacle Avoidance code makes use of the Single Obstacle Avoidance with 90
turns consecutively. Like before, wheelchair is prepped for motion by giving it zero velocity
values. After a while, it is given a forward velocity of 0.8m to get it started. When the first
obstacle is detected, the wheelchair halts and makes a 90 turns and moves horizontally to avoid
the first obstacle. It continues on its new path this time and as it encounters a second obstacle in
front of it. It again makes a 90 turn and avoids the obstacle by moving horizontally.
Another slight variation we could have programmed for this code could have been multiple
obstacle avoidance such that the obstacle returned to its original path along which it started. It is
easily possible by making slight modifications on our current code.

7.4 Dynamic Obstacle Avoidance

This is another small task we could accomplish using our current code however, we have not yet
perfected it during trails runs. Simplistic Dynamic Obstacle Avoidance is also possible using our
current Single Obstacle Avoidance Code. However, we have not had the time to implement it in
lab as such. We need to use the same code as that we used before for Single Obstacle Avoidance.
However, this time once the obstacle has been detected we need to give the wheelchair a velocity
somewhat greater than 0.8 say 1.8 or something which would cause the wheelchair to move
faster than the other dynamic object moving directly in front of it and to maintain that speed
during and after obstacle avoidance. Further testing is encouraged in this direction since it would
quickly yield results.

43

7.5 Navigation Testing

Our Run-Time based Navigation techniques making use of the cfg file to do simple navigation
for the wheelchair using its start point and its angular heading at the start point worked perfectly
in every trail run. The wheelchair was able to end up at exactly where we wanted it to go in
almost every single trial run.

44

CHAPTER 8: CONCLUSION & RECOMMENDATIONS

8.1

Conclusions

The Autonomous Robotic Wheelchair is an important technical solution to the problems posed
by the traditional powered wheelchairs. It can enable the end-user to use the wheelchair by
himself without the need of any other person, irrespective of their disability or physical
condition. For people with Parkinsons, Multiple Sclerosis or Tetraplegia, an autonomous
wheelchair can be a means to an independent, fulfilling life.
In the course of our Final Year Project entitled Autonomous Robotic Wheelchair we have been
able to make decent progress towards our dream of developing such a wheelchair here in
Pakistan. We could not have gotten anywhere had it not been for the SAKURA automated
wheelchair provided to us at RISE Lab. The automated wheelchair had the electronic control and
the hardware which was a necessary prerequisite for our project.

We were able to modify the existing hardware of the wheelchair to mount our LRF and the
laptop table with electronics housing. Our hardware design is modular, simplistic and completely
detachable. In the course of future work on the wheelchair hardware, subsequent sensor mounts
can the designed either on top of our own mounts or in place of the holes and fasteners that we
have already used.

As for the project tasks of Obstacle Avoidance and Navigation, we were able to make
respectable progress on both the fronts. Not only could we successfully implement Obstacle
Avoidance and Navigation in our simulations, we were able to implement both during our trails
runs in the lab. However, only on a more restricted scale and approximating the behavior of our
original algorithm commands by using sleep calls with different time values.

Finally, we were able to successfully upgrade the wheelchair from its joystick control to laptop
control while making use of sensor readings all in real time. A robust LabVIEW program was
used to drive the wheelchair by giving it power/ velocity values and Player\Stage was used to
45

collect data from the LRF for successful obstacle avoidance in almost every single trial run.

8.1.1 Learning Outcomes

When we started out with the Autonomous Robotic Wheelchair at the start of the year we did not
have a lot of background knowledge to go on. In fact, the only thing we had was the idea of the
final shape of the Autonomous Robotic Wheelchair that we wanted to get to. And as we started
out, it was unclear how exactly we would have to proceed in order to get to that final stage.

We took the first steps under the guidance of Dr. Yasar Ayaz and Sir Khawaja Fahad Iqbal and
eventually the road ahead became clear step by step. We started by getting acquainted with
Ubuntu and the Player\Stage interface. We studied from the Player\Stage manual and built and
tested several smaller Player\Stage programs. Eventually we started integrating our C++ coding
into our Player\Stage Simulations and programming for more complicated problem scenarios.

We learning coding in LabVIEW with Ali Bin Wahid who had already developed a small
LabVIEW program to convert the wheelchairs joystick control to laptop control. We
complicated upon his original program as well incorporate the Wireless settings and TCP
broadcasting for LabVIEW so that LabVIEW could communicate in real time with our
Player\Stage code.

Another challenging task was setting up the Player\Stage codes to work on the hardware rather
than in simulations. The LRF had to be upgraded from the simulations proxy to the USB Port 0
proxy and next, we had to network the computers. The crux of our networking was based on the
wireless TP-Link and in our C++ code that used sendData commands repetitively each time a
SetSpeed call was made in our code.

Another learning outcome of the project was getting to understand the actual behavior of the
wheelchair as it differed from the faultless simulations we could create in Player\Stage. Several
test variables modified the behavior of our wheelchair in the real time e.g. the surface friction of
the test space, the uncontrollable variations in the front wheels motion and the amount of battery

46

power available to the wheelchair.

8.2

Future Recommendations

Future work on the Autonomous Robotic Wheelchair could involve improvements on the current
Obstacle Avoidance and Navigation theorems. We have only made use of the most simplistic
algorithms at this point which are good enough for room level movements but in order to make
the wheelchair more enhanced we need to move towards building level navigation and for that
complication navigation algorithms like the BrushFire or the DFS or BFS algorithms need to be
implemented for wheelchair motion.
The wheelchair hardware also needs slight improvements. In
its current state, the wheelchair table is not positioned correctly to
accommodate fully the wheelchair users legs. The table platform
needs to be raised and also contraptions need to be installed for other
sensors.

Lastly there is also a lot of scope for improvement in the design of our
current LabVIEW program for driving the wheelchair. As it is our
LabVIEW program is fairly simplistic and is merely translating the
velocity values being received from the code. We need to develop a
LabVIEW program that could interact with multiple Player\Stage
C++ codes reading and processing data from multiple sensors. Once
we have developed techniques to incorporate such processing within
LabVIEW itself the wheelchair will be become vastly more intelligent
than it is right now.

There is a long road ahead calling for extensive efforts in

programming especially and students with a strong programming or
technical background would be best suited to take on this project in
the future.

47

APPENDIX A

48

REFERENCES

http://playerstage.sourceforge.net/
http://playerstage.sourceforge.net/player/player.html

http://www.ni.com/pdf/manuals/371303m.pdf

http://beej.us/guide/bgnet/
http://beej.us/guide/bgnet/output/print/bgnet_USLetter.pdf
http://www.lantronix.com/resources/networking.html

[1] Ruffles, P.C.; 2001, Expanding the Horizons of Gas Turbine in Global Markets; ISABE
2001-1010
[2] Dring, R. P., Joslyn, H. D., Hardin, L. W., and Wagner, J. H., 1982, "Turbine Rotor-Stator
Interaction," ASME Journal of Engineering for Power, vol.104, pp.729-742
[3] Arndt, N., 1993, Blade Row Interaction in a Multistage Low Pressure Turbine, ASME
Journal of turbomachinery, Vol. 115, pp. 370-376
[4] Denton J.D., 1993, Loss Mechanisms in Turbomachines:, ASME Journal of
Turbomachinery, Vol.115
[5] Denton J.D., 1993, Loss Mechanisms in Turbomachines:, ASME Journal of
Turbomachinery, Vol.115
[6] R.E. Walraevens, A.E. Gnllus, Stator-Rotor-Stator Interaction inAn Axial Flow Turbine And
its Influence on Loss Mechanisms, AGARD CP 571, 1995, UK, pp 39(1-13).
[7] Dawes, W.N., 1994, A Numerical Study of the Interaction of Transonic Compressor Rotor
Over Tip Leakage Vortex with the Following Stator Blade Row, ASME Paper No. 94-GT156
[8] Sharma, O,P., Renaud, E., Butler, T.L., Milasps, K., Dring, R.P., and Joslyn, H.D., 1988,
Rotor- Stator Interaction in Multi- Stage- Axial Flow Turbines, AIAA Paper No. 88-3013

49

[9] Busby, J.A., Davis, R.L., Dorney, D.J., Dunn, M.G., Haldeman , C.W., Abhari, R.S., Venable,
B.L., and Delany, R.A., 1999, Influence of Vane-Blade Spacing on Transonic Turbine Stage
Aerodynamics: Part II- Time- Resolved Data and Analysis:, ASME Journal of
Turbomachinery , Vol. 121, pp. 673-682
[10]

Collar (1947), "The Expanding Domain of Aeroelasticity," Journal of the Royal

Aeronautical Society 51-1.

[11]

Bell Loyed, Three dimensional Unsteady flow analysis in vibrating Turbine Cascades,

Durham University, 1999.

[12]

Jeff Green, PHD Thesis, Controlling Forced Response of a High Pressure Turbine Blade,

Royal Institute of Technology, Stockholm, 2006.

50