Project Number: RBE.TP1.

CPS2

Robolink: Modular Assistive Robot Arm

A Major Qualifying Project Report
Submitted to the faculty of the
WORCESTER POLYTECHNIC INSTITUTE
In partial fulfillment of the requirements for the
Degree of Bachelor of Science

Advisor:
Taskin Padir

Submitted By:
Cassiopia Hudson
Gabriel Morell-Pacheco

1 Acknowledgments
The team would like to extend our thanks to several people who helped make this project
possible:

Taskin Padir, for his help and guidance as our advisor,

Ty Tremblay, for configuring the Robolink and getting us started,

Lillian Walker, for contributing her time and CAD experience,

Joe St. Germain, for his constant and helpful advice,

James Fleming, for his expertise on the Cyber Physical Systems project, and

Thomas Angelotti and Patrick Morrison, for opening the ECE shop to us.

2 Abstract
The goal of this project is to utilize the igus Robolink arm five degree of freedom
modular robot arm, to complete useful tasks for persons with no or limited mobility. These tasks
include driving the joystick of a wheelchair, flipping a light switch, and turning the pages of a
book. This is done through designing and building a modular interface for mounting the
Robolink arm onto an existing wheelchair project and implementing a universal control interface
in the software for future expansion of tasks and control methods.

3 Table of Contents
1

Acknowledgments ................................................................................................................... 2

Abstract .................................................................................................................................... 3

Table of Contents .................................................................................................................... 4

Table of Figures....................................................................................................................... 5

Table of Tables ........................................................................................................................ 6

Introduction ............................................................................................................................. 7

6.1

Explore existing technology .......................................................................................................... 7

6.2

Task Specifications ........................................................................................................................ 8

6.3

Design Specifications ..................................................................................................................... 8

Background ........................................................................................................................... 10
7.1

igus Robolink Applications ...................................................................................................... 10

7.2

Similar Projects ........................................................................................................................... 12

Methodology .......................................................................................................................... 15
8.1

Review of Previous Work ........................................................................................................... 15

8.2

Proof of Concept Experiments ................................................................................................... 15

8.3

Design Constraints ...................................................................................................................... 16

8.3.1

Mounting the Robolink .......................................................................................................... 16

8.3.2

Passive End Effector .............................................................................................................. 17

8.4

8.4.1

Mounting Arm ....................................................................................................................... 17

8.4.2

Power Supply ......................................................................................................................... 19

8.4.3

Controller ............................................................................................................................... 19

8.4.4

Wheelchair Joystick ............................................................................................................... 20

8.4.5

End Effector ........................................................................................................................... 22

8.5

Design Solutions .......................................................................................................................... 17

Completion of Tasks.................................................................................................................... 23

8.5.1

Joystick Code ......................................................................................................................... 23

8.5.2

Robolink Controller ............................................................................................................... 25

Results & Discussion ............................................................................................................. 27

9.1

Mounting Arm ............................................................................................................................. 27

9.2

Power Supply ............................................................................................................................... 30

9.3

Controller ..................................................................................................................................... 30

9.4

Wheelchair Joystick .................................................................................................................... 30

9.5

End Effector ................................................................................................................................. 30

9.6

Completion of Tasks & Experimental Results.......................................................................... 30

9.6.1

Drive Joystick ........................................................................................................................ 30

9.6.2

Flip Light Switch ................................................................................................................... 31

9.6.3

Turn Page of Book ................................................................................................................. 38

10 Conclusion ............................................................................................................................. 39
10.1

Goal Completion........................................................................................................................ 39

10.1.1

Task Specifications .............................................................................................................. 39

10.1.2

Design Specifications........................................................................................................... 40

10.2

Societal Considerations ............................................................................................................. 40

10.3

Future Recommendations......................................................................................................... 41

11 Bibliography .......................................................................................................................... 43
12 Appendix ................................................................................................................................ 48
12.1.1

Instructions for Use .............................................................................................................. 48

12.1.2

Repair History / Incident Report .......................................................................................... 49

12.1.3

Presentation Slides ............................................................................................................... 51

4 Table of Figures
Figure 1: Wheelchair with Joystick and Desk ................................................................................ 9
Figure 2: Robolink Cable Box ...................................................................................................... 10
Figure 3: igus Robolink arm ......................................................................................................... 11
Figure 4: FLASH Humanoid Robot.............................................................................................. 11
Figure 5: Robolink Fishtail Robot ................................................................................................ 12
Figure 6: Gripper with "Fingers" Tied .......................................................................................... 16
Figure 7: Mounting Fixture ........................................................................................................... 17
Figure 8: Mounting Structure........................................................................................................ 18
Figure 9: Raspberry Pi .................................................................................................................. 19
5

Figure 10: Simple Arcade Joystick ............................................................................................... 20

Figure 11: CTI Electronics Industrial Joystick ............................................................................. 21
Figure 12: Emergency Stop Button............................................................................................... 22
Figure 13: End Effector ................................................................................................................ 23
Figure 14: End Effector Over Joystick ......................................................................................... 23
Figure 15: Workspace of Arm Vertical ..................................................................................... 28
Figure 16: Workspace of Arm - Joystick Driving ........................................................................ 28
Figure 17: Workspace of Arm Horizontal ................................................................................. 29
Figure 18: Workspace of Arm - Over Desk .................................................................................. 29
Figure 19: Light Switch With Chair Facing Wall ......................................................................... 32
Figure 20: Light Switch With Chair at 45-Degree Angle ............................................................. 33
Figure 21: Forces on a Light Switch ............................................................................................. 35
Figure 22: Light Switch With Chair at 90-Degree Angle ............................................................. 37
Figure 23: Rocker Light Switch .................................................................................................... 38
Figure 24: Robolink Typing on Laptop ........................................................................................ 39
Figure 25: 80/20 Extruded Aluminum .......................................................................................... 40
Figure 26: Malleable End Effector ............................................................................................... 42
Figure 27: Motor Cable Pulled from Fastener .............................................................................. 49
Figure 28: Current Motor Axle Bend............................................................................................ 50

5 Table of Tables
Table 1: Similar Project Comparison ............................................................................................ 14
Table 2: Returns and Updates Rolling Average ........................................................................... 24
Table 3: Logic for Determining Wheelchair Speed ...................................................................... 24
Table 4: Example of Use............................................................................................................... 26
Table 5: Configuration 1 Necessary Force ................................................................................... 36
Table 6: Configuration 2 Necessary Force ................................................................................... 37

6 Introduction
Tasks such as drinking from a glass of water, preparing a meal, or doing housework are
daily activities most people can perform without much difficulty. For a person with limited
mobility, these Instrumental Activities of Daily Living (IADL) can be difficult, or even
impossible, without help or supervision. A disability statistics report written in 2000 that focused
on the use of mobility devices in the United States outlines that 1.7 of the 6.8 million Americans
who need assistive devices, need wheelchairs or scooters to help them with mobility [19]. More
than two thirds of these people (68.1 percent) also need assistance with at least one IADL [19].
Wheelchair mounted robotic arms have been developed since the early 1990s to help conquer
this problem [19].
The purpose of this project is to establish control of a wheelchair mountable modular
robotic arm to complete tasks useful for a person with no or limited mobility. The igus
Robolink is a five degree-of-freedom (DOF) modular robotic arm that is mounted to augment the
person's ability to complete tasks. The goals of this project are to utilize the Robolink to drive the
joystick of a wheelchair, flip a light switch, and turn the page of a book.

6.1 Explore existing technology

Existing Robolink applications vary in scope from humanoid robots to industrial
processes. Research of Robolink applications, including a review of their Youtube playlist
igus robolink robot modules applications and references yielded no other cases of the
Robolinks use as a wheelchair mounted assistive robot arm (igus GmbH). This led to research
on the history of assistive robot arms. The team learned that work on assistive robotic arms
started in the early 1990's, an example of which is the Manus ARM made by Exact Dynamics
[39]. The ongoing trend is that robot arms are now smaller and lighter than their earlier
counterparts while increasing in functionality and usability. Advantages of the igus for human
assistive projects include its low cost, approximately $2,000 as opposed to the approximate
$48,000 JACO arm, and compact size, compared to arms like DORA [20], [30]. The igus
Robolink has a maximum of six DOF, however this configuration utilizes only five. Most
wheelchair mounted assistive robotic manipulators utilized 5-7 DOF ( Table 1: Similar Project
Comparison). As with most existing projects, the Robolink arm used has the extension capability
of just above a meter ( Table 1: Similar Project Comparison).

6.2 Task Specifications

A set of task specifications served as guidelines for the project. The most important task
specification involved driving the wheelchair using the Robolink arm. Specifications for driving
revolved around creating smooth motion of the chair for the passenger, so abrupt acceleration
and deceleration were to be avoided. The arm needed to move the joystick of the wheelchair
front, back, left, and right within a few seconds of receiving the command. The second task was
flipping a standard height light switch. The arm needed to reach out to a 48in. high light switch
on a wall, taking less than 15 seconds to flip the switch in either direction. The third task
involved flipping the pages of a hardcover book, taking less than 30 seconds to open the front
cover, and less than 30 seconds to turn a page. The time constraints were chosen arbitrarily based
on the teams opinions of reasonable time intervals.

6.3 Design Specifications

Several design specifications were developed to guide the design process. The Robolink
arm and its attached control box (12 x 12 x 4) containing the cables, motors, and controllers
(referred to as the cable box) shown in Figure 2: Robolink Cable Box needed to be stably
mounted on the wheelchair. This mounting needed to allow the end effector of the Robolink to
reach the joystick of the wheelchair at the end of the right armrest and the desk along the left
armrest of the wheelchair seen in Figure 1: Wheelchair with Joystick and Desk.

Figure 1: Wheelchair with Joystick and Desk

The arm also needed to be mobile, so an interface needed to be developed to draw power
from the wheelchairs power supply, and the Robolink needed to be controlled by a portable
small board or laptop.
One goal of this project involved experimenting with passive end effectors to avoid using
power and pneumatics to control a gripper, so a passive end effector needed to be developed. The
passive end effector needed to be modular and easily replaceable, while versatile enough to
complete multiple task specifications.

Figure 2: Robolink Cable Box1

7 Background
7.1 igus Robolink Applications
The Robolink system was made by igus and was released in 2011. It is composed of a
set of lightweight plastic joints, metal tubes, drive units, cables and end attachments that are
meant to work together to produce an infinite variety of low cost and low maintenance robot
arms.
The Robolink system (Figure 3: igus Robolink arm) has a large number of example
applications ranging from swimming robots, to pick and place industrial robots, to humanoid
robots. The team found a few good examples of the Robolink system in use in humanoid robotics,
all of which extend the maximum number of DOF afforded by the Robolink via complex
grippers and hands.

http://robot.wpi.edu/wiki/index.php/Robolink

10

Figure 3: igus Robolink arm2

The most impressive of these is the Flexible LIREC Autonomous Social Helper (FLASH) from
the Wrocaw University of Technology, which has seven DOF per arm and incorporates two
arms [21]. The Robolink arms on FLASH are purposed to express emotions by means of
gesticulation [21]. It is mounted on a self-balancing platform similar to a Segway, shown in
Figure 4: FLASH Humanoid Robot.

Figure 4: FLASH Humanoid Robot3

2
3

http://www.igus.com.sg/wpck/default.aspx?Pagename=robolink_articulatedarm&C=SG&L=en
http://www.flash.lirec.ict.pwr.wroc.pl/

11

Another humanoid, Humanoid Robot A1 (ADAM), has six DOF and two arms as well.
Both are some of the more complex examples of Robolink applications. On the other end of the
spectrum is a swimming robot from the German Research Center for Artificial Intelligence
(DFKI GmbH) that uses a Robolink joint to actuate a mechanical fish tail to propel a robot
through water seen below in Figure 5: Robolink Fishtail Robot. The research conducted suggests
that the Robolink arm used in our project is the only instance of Robolink currently being used
for assistive robotics.

Figure 5: Robolink Fishtail Robot4

7.2 Similar Projects

To learn more about assistive robotic arms, the team researched previous endeavors in the
field over the past twenty years. Table 1: Similar Project Comparison, organized by year, was
developed to compare eight similar projects to our work with the Robolink. The team members
based on photos, videos, and other research determined ratings in categories like Size and
Number of Tasks. The table also compares categories such as Control, Degrees of Freedom,
Maximum Extension, and Weight of Arm.
4

http://www.youtube.com/watch?v=4ajf4Ob-BGk&list=PL07B939DF7512B3B4

12

Many previous assistive wheelchair mounted robotic arms have used joysticks to control
the arm, but alternative control methods have been explored. One of the first arms by Exact
Dynamics used a combination of a camera and a touch screen interface that allows a human to
control the arm [40]. Touch screens have been used in robotic arms but more recently the use of
brain signal processing has been a focus of research. This enabled the use of assistive arms that
do not require physical input from the user, such as the movement of a hand to utilize a joystick,
so that the arm could be used by persons with no mobility. As the field has progressed, arms
have been getting smaller and lighter, with the weight of current generation arms being less than
half of the first iterations. The Robolink arm being used by the team is recent and weighs 10.3 kg
(arm, cable box, and mounting hardware). As well as being light, the Robolink arm the team is
using is highly compact even compared to other Robolink implementations. The standard stepper
motors used to drive the cables have been replaced with a custom drive module, which is far
smaller.
As seen in Table 1: Similar Project Comparison, many existing assistive robotic arms use
a gripper, one of the projects goals involved implementing a passive end effector to keep
complexity down. This goal is to prove that a well designed and thought out passive end effector
is as useful in performing a range of tasks as one that is specialized.
In order for the end effector to be effective, the system needs a workspace that allowed it
to perform its task. As seen in Table 1: Similar Project Comparison, the baseline minimum
number of degrees of freedom for an assistive robotic arm is five. Any fewer degrees and the
reach of the arm is limited and starts to have difficulty performing tasks due to not being able to
reach points in space. The Robolink configuration being used for this project has five, which
allows for versatile reach and task performance while keeping complexity down.
The reach of existing robotic arms has historically been approximately a meter with
variances of 20%. In the wheelchair configuration being used the reach of the Robolink arm is
1.187 meters, which is on the higher end of historical reaches.

13

Table 1: Similar Project Comparison

Robot Arm Name:

Robolink

RAPUDA (Robotic Arm

for Persons with Upper
Limb DisAbilities)
Human in Loop Integration
of An Arm Mounted
Wheelchair Robot Based
on RT Middleware
JACO | Rehab Edition

PerMMA (Personal
Mobility and Manipulation
Appliance)
DORA (Door Opening
Robotic Arm)
iARM (Assistive Robotic
Manipulator)

Developed By:

igus

Systems Research
Institute (AIST)

2013

Control:

Brain Signals,

Arm:

Size:
(Compact (1)
to Bulky (5))

End
Effector:

Number of

Weight can

Tasks: (Few (1)

lift

to Many (5))

(Payload):

Degrees
Extension:

of
Freedom:

10.3 kg

Passive

N/A

1.187 m

5.897 kg

Gripper

0.454 kg

0.914 m

5.000 kg

Gripper

1.000 kg

1.070 m

5.700 kg

Gripper

1.500 kg

0.900 m

Signals, Touch 28.600 kg

2.500

0.830

kg/arm

m/arm

Joystick

Ref:

[16]

2010

Joystick

[13]
[48]

Japan
Sugano Lab
Waseda

Joystick,
2010

University

Kinova

University,
University of

2010

Joystick

Joystick, Brain
2009

Screen

Pittsburgh
UMass Lowell

Touch Screen,
Brain Interface

Carnegie Mellon

2009

Joystick

2x
Gripper

Exact Dynamics

2009

Southern Florida

2008

[44]

[18]
[20]

[26]
[34]

[30]

Gripper

N/A

1.245 m

9.000 kg

Gripper

1.500 kg

0.900 m

13.750 kg

Gripper

4.500 kg

1.082 m

[1]

User Interface, 14.300 kg

Gripper

1.500 kg

0.800 m

[40]

Joystick, or
Single Button

Mounted Robotic Arm)

[46]

13.063 kg

Keypad,

University of

Robotic Manipulator)

Developed:

Weight of

Intelligent

WMRA (Wheelchair

Manus ARM (Assistive

Year

Brain Signals

[31]
[14]
[15]

Touch Screen
Exact Dynamics

~1992

Camera

8 Methodology
The goal of this project is to utilize the igus Robolink arm to drive a wheelchair, flip a
light switch, and turn the page of a book. The teams approach to the project involved
understanding the igus arm, its capabilities, and constraints. Design solutions for making the
arm mobile included mounting the arm on the wheelchair, interfacing with the wheelchairs
power supply, investing in a joystick to control the chair, and designing a custom passive end
effector for the Robolink.

8.1 Review of Previous Work

The igus Robolink arm utilized for this project was configured prior to the teams work
[39]. The teams first steps in this project involved reviewing and understanding the initial
configuration of the cable box, and the documentation of the arm from igus. The team was also
provided with the forward and inverse kinematics of the arm, including the Jacobian matrix. The
team reviewed the configuration of the 5 DOF arm and kinematics calculations. Code was
provided to run the arm with an Xbox controller. The Xbox controller code was very useful
throughout the project.

8.2 Proof of Concept Experiments

Several proof of concept experiments were conducted as the team brainstormed
approaches to the design and task specifications of the project. The first task the team
experimented with was utilizing the Robolink to move the joystick of a controller. The
wheelchair did not have a joystick for the team to work with, so to conduct the test, the team
utilized an Xbox controller and a GameCube controller. Since a passive end effector was
required but not yet developed the team tied the fingers of the igus gripper together seen
below in Figure 6: Gripper with Fingers Tied.

Figure 6: Gripper with "Fingers" Tied

The team utilized the Xbox controller code to move the fingers over the different joysticks of
the controllers to determine if the control was precise enough to move a joystick. The results
showed that the arm could be used to control a joystick.
The second proof of concept experiment revolved around finding possible mounting
positions for the Robolink on the wheelchair based on the workspace of the arm. Taking into
account the design specification that the end effector needed to reach both the wheelchairs
joystick on the end of the right armrest and the desk attached to the left armrest, the team
measured the workspace of the arm from different mounting configurations. The conclusion was
that the most appropriate place to mount the arm was on the back of the wheelchair, with the arm
coming over the right shoulder of the chair.

8.3 Design Constraints

8.3.1 Mounting the Robolink
The team faced many constraints in achieving the design goals. Mounting the cable box
and arm on the wheelchair posed the issue of finding suitable space to mount the arm. Since
assistive wheelchairs are already loaded with technology, available mounting space is limited. As
well as being limited the mounting point needs to support the 10.3 kg weight of the robot arm.
The arm also needs to be interfaced with the chairs mobile power supply and laptop. As

16

explained in the proof of concept experiment above, the workspace of the arm with respect to
both the joystick and desk needed to be accommodated.
8.3.2 Passive End Effector
Another important design constraint involved not utilizing the igus gripper as the end
effector. This decision was made in an attempt to save space by not mounting pneumatics to
control it. A simpler solution was necessary, in the form of a passive end effector. This end
effector needed to be modular, easy to remove and replace the existing gripper. This end effector
also needed to be versatile enough to work for multiple tasks including joystick driving.

8.4 Design Solutions

8.4.1 Mounting Arm
The Robolink arm needed to be mounted such that the end effector could reach both the
joystick at the end of the right armrest and the desk of the wheelchair on the left armrest. The
preliminary experiment showed that the best place to mount the arm was the back of the chair,
with the arm coming over the right shoulder of the chair. Possible mounting solutions involved
attaching the cable box to the back of the chair using the headrest, back pocket of the chair, or
the unused mounting fixture at the base of the chair seen below in Figure 7: Mounting Fixture.

Figure 7: Mounting Fixture

17

The first design version involved utilizing the mounting fixture and 1 steel square pipes.
The pipes would be bent to an L-shape to attach to the mounting fixture. An L-shaped aluminum
plate would be machined to attach to and support the cable box. For extra support, nylon
mounting straps would attach the box to the headrest.
This first design was evaluated to be too complex. A second version of the design also
utilized the empty mounting fixture at the base of the chair, but consisted of only 80/20
Aluminum parts. Several pieces of 1 in. 80/20 pipe and several brackets were utilized to screw
directly into the back plate of the box. The box was mounted several inches off center such that
the arm came directly over the right shoulder. The mounting fixture with the back plate of the
box attached can be seen below in Figure 8: Mounting Structure.

Figure 8: Mounting Structure

18

8.4.2 Power Supply

The Robolink drive module takes 24V DC and at least 10A. The arm needed to be mobile,
so it needed to interface with the wheelchairs power supply. Once an empty 24V Anderson
Power Pole connector was located on the chair, about 5 ft. of wiring was run from the base of the
chair to the power connector on top of the box to power the arm.
8.4.3 Controller
To achieve a mobile system, the igus arm needed to run from a mobile computer. The
first mobile computer the team attempted to utilize was a Raspberry Pi, shown below in Figure 9:
Raspberry Pi. The Raspberry Pi was small and powerful, which would have been ideal for this
application. Issues occurred with building the existing software against the EposManager library
on the Raspberry Pi. The issue was that the only provided EPOS driver was available for x86 and
not ARMel. EposManager is a necessary Robot Operating System (ROS) package for controlling
the arm, making the Raspberry Pi unusable for this application.

Figure 9: Raspberry Pi5

The remaining options for controlling the Robolink were laptops. The team investigated
purchasing a small laptop to control the system. The small laptop would have been ideal because
it would have been lightweight and compact, while supporting the EposManager. This small
laptop was ordered used and was found to have issues with the boot drive. The team determined
that it was not worth it to devote any more time to it, and moved to the Lenovo ThinkPad laptop
5

http://images.bit-tech.net/content_images/2013/03/raspberry-pi-case-competition-update/pi1l.jpg

19

provided by WPI. A ThinkPad labeled Amazon was the computer utilized for controlling the
wheelchair. The team was assigned a ThinkPad labeled Nile. The Nile laptop's software
installation was 32-bit, which is required to run the EposManager, while the Amazon laptop's
existing software installation was 64-bit, and could not be used for the Robolink.
8.4.4 Wheelchair Joystick
The highest priority task involved driving the joystick of the wheelchair with the igus
arm. The biggest challenge to this accomplishment was the lack of a joystick to work with, since
the original joystick had been disassembled and determined inadequate for the application. The
team needed to purchase a joystick with both directional control and a button to be used as an
emergency stop. Options included game system controllers, such as Xbox, GameCube, or Wii.
Proof of concept experiments showed promise with the GameCube controllers grey joystick.
These game controllers were all potentially clumsy to mount on the chair, however, so
arcade joysticks were explored next. A simple arcade joystick held much potential; it was cost
efficient and the proper size to mount on the wheelchair. It can be seen in Figure 10: Simple
Arcade Joystick. Three concerns were raised with this option, though. The ball on the end of the
joystick would be difficult to push around, and the joystick was digital not analog, which was
necessary to control the speed of the wheelchair, and there were not buttons available.

Figure 10: Simple Arcade Joystick6

https://www.sparkfun.com/products/9182

20

The team then researched industrial joysticks, which were deemed to provide high
compatibility with the team's vision for an end effector. They were an ideal size for being
manipulated. They provided analog control, which was necessary for fine tuned wheelchair
control, and some were USB compatible, which would make working with Robot Operating
System (ROS) straightforward. The main downfall of the industrial joysticks was cost. A CTI
Electronics industrial joystick was ordered for the wheelchair. It can be seen in Figure 11: CTI
Electronics Industrial Joystick.

Figure 11: CTI Electronics Industrial Joystick7

The team soon discovered that an emergency stop button would be necessary. A McMaster-Carr
turn-to-reset Emergency Stop button was ordered and mounted on the back of the cable box to
cut power to the Robolink, shown in Figure 12: Emergency Stop Button.

http://www.ctielectronics.com/OEM-Analog-Industrial-Joysticks/Inductive-Industrial-Joystick.php#N24

21

Figure 12: Emergency Stop Button

8.4.5 End Effector

One goal of the project was to utilize a passive end effector, as mentioned in the
Constraints section above. A Robotics Engineering Undergraduate student volunteered to assist
the team in designing a passive end effector for the arm. While Walkers project was under
development, the team tied the fingers of the igus gripper together to test and prove concepts.
Walkers end effector was tailored to work with the CTI Electronics joystick and made to
modularly fit the end effector attachment of the Robolink. It consisted of a 1 in. deep cup that fit
over both the joystick and a light switch. The custom end effector was 3D printed. Photos of the
custom end effector can be seen below in Figure 12: End Effector and Figure 14: End Effector
Over Joystick.

22

Figure 13: End Effector

Figure 14: End Effector Over Joystick

8.5 Completion of Tasks

Two distinct pieces of code have been written for this project. The first one interfaces the
joystick with the wheelchair's control systems. The other controls the arm mainly to drive the
joystick and perform other tasks.
8.5.1 Joystick Code
The joystick control code converts the raw joystick input data from the axes into
something that ROS can comprehend. It subclasses some of the methods and overrides others
that were written for Xbox controller. Some issues with the joystick itself were overcome in the
code. When the joystick nears the edges of its range (in the extremes), all the values go to zero.
To counteract this, a weighted rolling average function was written to check against the last few
values that the joystick published and output the maximum needed value if the present joystick

23

values are determined to be in the extremes. The function that returns and updates the rolling
average can be seen below in Table 2: Returns and Updates Rolling Average.

Table 2: Returns and Updates Rolling Average

def _ij_x_avg(self):
# this function returns the rolling weighted average for
# the X values of the joystick value timeline
return average(self.x_timeline,weights=WEIGHTS)

def _ij_y_avg(self):
# this function returns the rolling weighted average for
#the Y values of the joystick value timeline
return average(self.y_timeline,weights=WEIGHTS)

def _ij_x_update(self,val):
# This function updates the Value Timeline of the X values
#by removing the last one and adding the latest as the first
self.x_timeline.pop()
self.x_timeline.insert(0,val)

def _ij_y_update(self,val):
# This function updates the Value Timeline of the Y values
# by removing the last one and adding the latest as the first
self.y_timeline.pop()
self.y_timeline.insert(0,val)

The logic for determining the wheelchairs speed based on the joystick values can be seen below
in Table 3: Logic for Determining Wheelchair Speed.

Table 3: Logic for Determining Wheelchair Speed

if ij_x > POS_MIN_VAR and ij_x < POS_MAX_VAR:
# if the value of x isn't in the joystick extremes
if ij_x not in extremes:
# convert the joystick value to an integer
x_map = convert_joy_value_to_int(1.0-ij_x)
# update the rolling average
ijO._ij_x_update(x_map)

24

elif ij_x < NEG_MIN_VAR and ij_x > NEG_MAX_VAR:

# if the value of x isn't in the joystick extremes part 2
if ij_x not in extremes:
# convert the joystick value to something usable
x_map = convert_joy_value_to_int(-1.0-ij_x)
# update the rolling average
ijO._ij_x_update(x_map)

else:
# we are in the extremes and it's terrible here
# so we get the current rolling average
a = ijO._ij_x_avg()
print a
# if the rolling average is above the rolling floor we set it to our max mapped
value
# if the rolling average is under the rolling floor we set it to our negative max
mapped value
if a > ROLLING_FLOOR:
x_map = ROLLING_MAP
elif a < -ROLLING_FLOOR:
x_map = -ROLLING_MAP/2

8.5.2 Robolink Controller

8.5.2.1 Command Message

The Robolink control code is composed of two classes that speak to existing Robolink
driver functionality. The first class is the CommandMessage class, which is initialized with the
following arguments:
'positions' : a list of position values
'timeinterval': a list of time intervals
'controller': a controller to modify
defaulting to the standard 'RobolinkControl' class
'reset': whether or not to expose a reset parameter
defaults to True
'steps': a list of arbitrary steps to execute
'backsteps': a list of steps to execute when being run in reverse
25

 'startingstep': a starting step

defaults to the first given
'method': which base method to modify
defaults to pose
can be set to None
'method_sub_map': what properties of the method to modify the values of 'positions'
defaults to the ['position.x','position.y','position.z'] objects
'control_mode': which control mode to use with EPOS
defaulting to pose control
'iterative_control': whether or not to use iterative control
defaults to False
'DRO': a list of command names not to reset on
The only required argument is the list of positions. It exposes five methods, most importantly
those to run the steps, run them forward, and to run through them backward. It also supports
executing any arbitrary position that was given to the Command Message either regularly or
iteratively
8.5.2.2 Command Acceptor

The second class is the Robolink Command Acceptor, which opens a ROS subscriber to
wait for and execute commands. It is initialized with a dictionary of CommandMessage instances.
Upon initialization it returns a list of available commands that it is listening for on the subscriber.
Upon receiving a command via a publisher it proceeds to run it until completion.
8.5.2.3 Example of Use

A command can be simple or override default behavior as needed. Two examples are
shown in Table 4: Example of Use with the command acceptor being initialized after defining
the CommandMessages. The first defines a set of points in space to iterate to. The second defines
a set of angles that the arm should replay.

Table 4: Example of Use

cmddict ={
'simple': CommandMessage(
positions=[(0,0.05,0),(0,.1,0)],

26

reset=False
),
'morecomplex': CommandMessage(
positions = [(-8,5,0,89,0),(-8,4,0,90,0),(-8,2,0,88,0),(-8,0,0,87,0),(-8,-1,0,85,0),],
method=None,
reset=True,
method_sub_map=METHOD_SUB_MAP_JOINTS,
control_mode=JOINT_VELOCITY,
DRO=['drive_stop','drive_fwd','drive_back'],
backsteps=[0,1,2]
),
}

cmdacceptor = CommandAcceptor(cmddict)

There are a large number of default settings that can be overridden in order to perform tasks
beyond the default scope.

9 Results & Discussion

9.1 Mounting Arm
The arm mounting structure consisting of 80/20 Aluminum pipes and brackets worked
well for this application. The structure was stable and strong, both in the sense that it did not
wiggle or move and that it attached to a very stable fixture on the wheelchair. The design was
also simple; it took very few pieces, did not require any custom work, and was easy to put
together. This design also achieved modularity, as its attachment to the fixture was
straightforward, and easy to put on or remove. The workspace of the Robolink was also very
efficient from this mounting position the desk and joystick could both be reached, as well as a
range of vertical and horizontal locations, shown in Figure 15:Workspace of Arm Vertical,
Figure 16: Workspace of Arm Joystick Driving, Figure 17: Workspace of Arm Horizontal,
and Figure 18: Workspace of Arm Over Desk.

27

Figure 15: Workspace of Arm Vertical

Figure 16: Workspace of Arm - Joystick Driving

28

Figure 17: Workspace of Arm Horizontal

Figure 18: Workspace of Arm - Over Desk

29

9.2 Power Supply

Interfacing with the wheelchairs power supply worked well. While it was difficult to
locate an open 24V DC rail with no knowledge of the wheelchairs electronics, once it was found
the interfacing was simple. This interfacing also kept within the modular spirit of the project, and
an Anderson Power Connector was the only necessary components, and was easy to connect and
disconnect from the system.

9.3 Controller
Overall, the Lenovo ThinkPad laptops worked well for this application. Working with
both Amazon and Nile on the mobile wheelchair was clumsy, but given that the EposManager
could not work on Amazon, it was necessary to use both laptops.

9.4 Wheelchair Joystick

The CTI Electronics joystick was definitely the right tool for the job, despite the cost. It
was USB compatible, which worked extremely well for this application. The code developed
worked well for wheelchair driving, and can be run simply on the Amazon laptop.

9.5 End Effector

The custom 3D printed end effector was optimal for use with the CTI joystick. It kept
within the modular theme since it was easy to attach and remove, and therefore worked well with
the project. The end effector was passive, not adding bulk to the system and not drawing any
power. It also served as an experiment to illustrate the usefulness of passive end effectors.
Though an immobile person would not be able to change out end effectors on their own, they can
still be useful.

9.6 Completion of Tasks & Experimental Results

9.6.1 Drive Joystick
The highest priority goal for this project involved driving a joystick mounted on the
wheelchair. While driving using the arm is achievable by executing the control movements
manually, there are a number of issues that the team came across that stand in the way of
automating it. To automate the wheelchair driving functions, the team recorded a series of angles
to play back and fed them into a CommandMessage. Testing yielded that telling the arm to go

30

back and forth between two points in space would never yield the exact same result. The arm
would move to approximately the same point in space but up to a half inch away from its
previous location. In an attempt to improve on the repeatability the team ensured that the drive
cables were did not have slack and were appropriately tightened. This in turn lowered the margin
of error but not enough to be satisfactorily repeatable due to another issue the team noticed;
when the power for the arm was cycled, the repeatability of tasks suffered due to the fact that the
arm drooped slightly then took the new position on power up as its default position. The team
attempted to work around this by accounting for the droop before power down so that the arm
would be in its default position even with the droop occurring. However, this wasn't ideal for a
deployed setup with an actual person occupying the chair.
The team also had to work around the fact that moving to a position outside of the driving
range while the end effector was still over the joystick would force the end effector to be caught
on the joystick and put massive strain on the arm as it attempted to move. This was mitigated by
adding a flag to the command message and having the arm move to a safe point above the
joystick before continuing on to the next task.
Part of the issue with repeatability was the feedback mechanism used by the command
acceptor. The Robolink arm feedback topic returned integers for angles instead of other number
types. When the end of arm was over forty-five centimeters away from the rotating joint, the
resolution for a degree is just under a centimeter, which was not accurate enough.
Circumference of end effector tube:

Margin of error:

9.6.2 Flip Light Switch

The second priority goal of this project involved flipping a standard height light switch.
This task proved more complex than initial thoughts portrayed. An incident occurred during an
experiment when the Robolink was mounted on the chair and run utilizing the Xbox control code
rather than the Command Message/Command Acceptor code. The Xbox control code crashed
but the motors continued to run. A team member was sitting in the wheelchair at the time, and
the arm drove into the team members leg. Though not enough force was put on the team
31

member to cause pain or injury, a motor axle for one of the degrees of freedom bent and needed
replacement. This incident provided valuable information to the team; the arm was potentially
good for an immobile person, as it would break before causing injury, and the force at the end
effector was very low.
This made the team wonder if there would be enough force to flip a light switch. To
measure the force to flip a light switch, the team utilized a force gauge, hooking the end over the
light switch. The team had access to two switches in Atwater Kent; old switches which required
a force of 3 lbf to flip, and new switches which required a force of 2 lbf to flip. Flipping a switch
on or off took the same amount of force.
The team conducted an initial experiment with the Robolink arm on the wheelchair
utilizing the Xbox controller code. The team first placed the wheelchair directly facing a
standard height light switch. The first problem presented itself; the arm was not long enough to
reach the wall in front of the chair with the footrest of the chair in its down position, seen below
in Figure 19: Light Switch With Chair Facing Wall.

Figure 19: Light Switch With Chair Facing Wall

32

The team corrected this by moving the chair at an approximate 45-degree angle with
respect to the wall, and utilizing the rotational DOF at Joint 0 of the arm to place the end effector
near the light switch seen in Figure 20: Light Switch with Chair at 45-Degree Angle. The arm
was then moved to a position such that Joint 2 was at an approximate 150-degree angle with the
end effector on the light switch. This was done in hopes that the weight of the arm would assist
in forcing the switch down. Both joints were driven to push downward on the switch
simultaneously.

Figure 20: Light Switch With Chair at 45-Degree Angle

This resulted in the end effector simply staying on the light switch. When the wheelchair
was pulled back from the wall, the end effector was approximately 7 in. lower than the light
switch on the wall. This shows that the arm was attempting to force the switch down but was not
exerting enough force to flip the switch. The result was the same for both the old and new
switches. The team was concerned, as a repeat of the above incident was not desirable and could
potentially result in another broken motor axle.
33

This prompted the team to perform a calculation of force exerted by the end effector
assuming the arm was pointed directly upward. The nominal torque for the motor was multiplied
by the gear ratio to get the output torque from the motors. This was then divided by the radius of
the motor wheels to find the force exerted after the motor wheels, F1. This force was then
multiplied by the radius of the rope in the joint divided by the link length to get the force at the
end of the joint, F2. This process was repeated until the force at the end effector was found.

From motor:

End of Link 1:

End of Link 2:

End of Link 3 (End Effector):

34

Though this calculation only accounted for one position and was therefore not entirely
accurate, it showed the minimal force the end effector could exert. This made sense, as it was not
enough force to hurt a person in the chair or flip a light switch, but it was enough to move the
high quality joystick. The Robolink utilized for this project used small Maxon Motors to make
the system more mobile, but were not as strong as the stepper motors igus generally uses for
the Robolink.
The team performed calculations with the Jacobian matrix for the Robolink system to
prove that the motors could not provide enough torque in multiple configurations. The
force/torque relation,

, where is a vector of joint torques and F is a vector of the forces

and moments on the end effector, was utilized for these calculations. The team assumed no force
in the y-plane and that the moments were considered negligible. The forces in the x and z planes
were calculated using trigonometry and can be seen in Figure 21: Forces on a Light Switch.

Figure 21: Forces on a Light Switch

35

Where:

[ ]

For sets of joint angles, the output torque was compared to the normal torque and stall
torque for the motors. For the configuration described above, with the arm at a 45-degree angle
to the wall Table 5: Configuration 1 Necessary Force was generated, where i, the angle between
Link 1 and Link 2, was in range of 40-70 degrees, and j, the angle between Link 2 and Link 3,
was in range of 120-150 degrees.

Table 5: Configuration 1 Necessary Force

Configuration: [45, i, 0, j, 0]

2 lbf Force

3 lbf Force

Nominal Torque (0.0827 Nm)

None

None

Stall Torque (0.782 Nm)

None

None

This data illustrated that the teams suspicions were true; no combination of joint angles
would reach and flip the light switch in this configuration. The team decided to work with
another arm configuration, seen in Table 6: Configuration 2 Necessary Force, where k, the

36

rotation from the shoulder, was in range of 60-120 degrees, seen below in Figure 22: Light
Switch With Chair at 90-Degree Angle.

Figure 22: Light Switch With Chair at 90-Degree Angle

Table 6: Configuration 2 Necessary Force

Configuration: [0, 90, k, 90, 0]

2 lbf Force

3 lbf Force

Nominal Torque (0.0827 Nm)

for 0 < k < 61 deg

for 0 < k < 60 deg

Stall Torque (0.782 Nm)

for 0 < k < 71 deg

for 0 < k < 67 deg

This configuration was much more promising than the first, as it utilized the motion of a
stronger joint than the end effector. In the case of a standard height light switch, this

37

configuration would still not work, since k would need to be much closer to 90 degrees. The arm
could potentially have enough force to flip a lower light switch.
The team suspects that the Robolink would be capable of flipping other kinds of light
switches, such as rocker switches like the one shown below in Figure 23: Rocker Light Switch.

Figure 23: Rocker Light Switch8

9.6.3 Turn Page of Book

Due to unforeseen delays in the project, the team was unable to thoroughly experiment
with turning the pages of a book. As a brief experiment, a laptop was placed on the wheelchairs
desk, seen in Figure 24: Robolink Typing on Laptop. The laptop was well within the workspace
of the Robolink, and the end effector of the Robolink could exert enough force to type on the
keys of the laptop. The team speculates that the Robolink would have been capable of turning the
pages of a book, given a rubberized end effector. This would have been a useful task to perform,
as it would majorly improve the quality of life of a person with no mobility.

http://www.homecontrols.com/Leviton-LevNet-Self-Powered-Wireless-Light-Switch-Single-Rocker-LVWSS0SD0x

38

Figure 24: Robolink Typing on Laptop

10 Conclusion
10.1 Goal Completion
10.1.1

Task Specifications

Some of the tasks specifications for this project were met. The arm was able to safely
drive the wheelchair using the industrial joystick with its custom end effector even with caveats.
However, the arm was unable to flip a standard height light switch due to the lack of necessary
force at the tip. The team was unable to test the opening of a book on the desk, however the team
tested the viability of the workspace that would ultimately be used to open the book and the
reach of the arm in that workspace.
39

10.1.2

Design Specifications

The design specifications for this project were all met. The arm needed to be mounted on
the wheelchair. The 80/20 Aluminum structure fulfilled this requirement while keeping in the
modular spirit of the project. The whole Robolink system needed to be mobile on the wheelchair.
This was achieved by interfacing the arm with the wheelchairs power supply and utilizing
laptops to control the system. A passive end effector was 3D printed to interact with the joystick.

10.2 Societal Considerations

Through the course of this project, the team needed to address the societal concerns
revolving around wheelchair mounted robotic arms. These concerns involve ethics, safety, and
aesthetics of the project.
The ethical concerns of this project involve the safety of the occupant of this wheelchair.
The team designed the mounting structure for the arm such that the arms workspace remains out
of the way of the occupant. In addition, it is possible to add software limitations to prevent the
arm from entering specific areas of the workspace, adding an additional level of safety. As well
as avoiding the occupant, the end effector of the arm can exert very little force (about lbf).
This means that the arm will break before it is able to injure an occupant of the wheelchair in
case of software failure. An emergency stop button has also been added to the cable box of the
arm that cuts power to all motors controlling the arm. Though the team could not find regulations
on wheelchair velocity or acceleration, the code that controls the wheelchairs joystick was
designed to avoid any sudden acceleration and deceleration in order to provide a smooth ride.

Figure 25: 80/20 Extruded Aluminum

40

The simple mounting design of the arm conforms to the design qualities of the existing
hardware and is pleasing to the eye. The design consists of non-abrasive materials, such as the
80/20 Aluminum shown in Figure 25: 80/20 Extruded Aluminum, with smoothed edges to
improve the quality and safety of the arm. All wiring is away from the occupant and moving
parts of the chair, keeping the look clean and safe. The passive end effector designed for this
project is also pleasing to the eye, made of 3D printed red plastic (which can be recycled),
matching the red headrest of the wheelchair.

10.3 Future Recommendations

The team would like to leave suggestions for future work on the Robolink project. The
most common problem the team encountered was a lack of sufficient motor bracing. The design
of the cable box did not support the cable wheels from both sides, leading to issues with bent
motor axles. The cables were very strong, so in a situation where a degree of freedom was forced
beyond its limit, the axle of the motor for that DOF would bend. The team recommends finding
some design solution to this weakness for future work on the project.
Another recommendation involved adding accuracy to the control system for the arm.
Repeatability and precision were difficult with the control code given by to the team. Joint angles
were integers of degrees, and were not repeatedly reached between power cycles. The team
recommends looking deeper into the control software, and potentially adding a feedback loop to
add more precision to the system.
An alternative to editing the control software could involve different passive end
effectors. Since the code cannot provide the precision required for the 3D printed end effector, an
end effector of a more malleable material could work well, as seen below in Figure 26:
Malleable End Effector.

41

Figure 26: Malleable End Effector9

9 http://www.bastiansolutions.com/blog/index.php/2013/06/12/innovative-robotic-end-of-arm-tool/#.UhzpIWRoQx8

42

11 Bibliography
1. Alqasemi, Redwan M., and Rajiv V. Dubey. "Chapter 3: A 9-DoF Wheelchair-Mounted
Robotic Arm System: Design, Control, Brain-Computer Interfacing, and
Testing." Advances in Robot Manipulators. Ed. Ernest Hall.,
2010. Http://www.Intechopen.com/books/advances-in-Robot-manipulators/a-9-DofWheelchair-Mounted-Robotic-Arm-System-Design-Control-Brain-Computer-Interfacingand-Testing Intech. Web. 2013.
2. Alqasemi, Redwan M., et al. "Analysis, Evaluation and Development of WheelchairMounted Robotic Arms." www.smpp.northwestern.edu. July 1 2005.Web.
<http://www.smpp.northwestern.edu/ICORR2005/proceedingsCD/pdffiles/papers/ThP01
_33.pdf>.
3. Chung, Cheng-Shiu MS, and Rory A. PhD Cooper. "Literature Review of WheelchairMounted Robotic Manipulation: User Interface and End-User Evaluation". RESNA
Annual Conference 2012. 2012.http://resna.org/conference/proceedings/2012/StudentScientific/Robotics/LIT
ERATUREREVIEWOFWHEELCHAIRMOUNTEDROBOTICMANIPULATION.html. Print.
4. Deyle, Travis. "Telescoping Robot Arm for Wheelchair - Mounted Object
Retrieval." www.hizook.com. Aug 16 2010.Web.
<http://www.hizook.com/blog/2010/08/16/telescoping-robot-arm-wheelchair-mountedobject-retrieval>.
5. dfkvideo. "Entwicklung eines Antribskonzepts dr das Robotergelenk
Robolink." www.youtube.com. Ed. Praxissemester-Arbeit am Robotics Innovation Center
des DFKI. Sep 23 2010.Web. <http://www.youtube.com/watch?v=4ajf4ObBGk&list=PL07B939DF7512B3B4>.
6. E. Chaves, B. S., et al. "Clinical Evaluation of a Wheelchair Mounted Robotic
Arm". Teachnology & Disability: Research, Design, Practice & Policy. June 19, Atlanta,
GA. Web.

43

7. Evenchick, Eric. "Toorcamp: MC Hawking Robotic Wheelchair." hackaday.com. Aug

14, 2012 2012.Web. <http://hackaday.com/2012/08/14/toorcamp-mc-hawking-roboticwheelchair/>.
8. Fox, Stuart. "Robotic Arm Opens Dorrs for the WheelchairBound." www.popsci.com. Nov 24 2009.Web.
<http://www.popsci.com/technology/article/2009-11/robotic-arm-opens-doorswheelchair-bound>.
9. Gray, Jeremy. "Trussville students build robotic arm to aid quadriplegics
(video)." blog.al.com. April 23, 2013 2013.Web.
<http://blog.al.com/spotnews/2013/04/trussville_students_build_robo.html>.
10. Gutierrz, Jon. "Wheelchair-Mounted Robotic Arm Is A Total
Snoozefest." www.uproxx.com. Sept 27 2010.Web.
<http://www.uproxx.com/gammasquad/2010/09/wheelchair-mounted-robotic-arm-is-atotal-snoozefest/>.
11. hairny8. "Robotic Arm for Wheelchair - 2010 SJSU Senior Project Presentation
(1/2)." www.youtube.com. Ed. Aimy Huynh, et al. May 21, 2010 2010.Web.
<http://www.youtube.com/watch?v=_ZqYBd9rTJA>.
12. Hillman, Michael, et al. "A Wheelchair Mounted Assistive Robot". International
Conference on Rehabilition Robotics. Stanford, CA. ICORR ,
1999. http://rehabrobotics.org/icorr1999/papers/papers/hillman.pdf. Print.
13. Hollister, Sean. "Lightweight robot arm connects to your wheelchair, stoops in your stead
(video_." www.engadget.com. Aug 1 2010.Web.
<http://www.engadget.com/2010/08/01/lightweight-robot-arm-connects-to-yourwheelchair-stoops-in-you/>.
14. "iArm." www.exactdynamics.com.Web. <http://www.exactdynamics.nl/site/?page=iarm>.
15. "iARM Product Overview." iarmrobot.com. 2012.Web.
<http://iarmrobot.com/overview.shtml>.
16. "igus Robolink." www.igus.com.Web.
<http://www.igus.com/wpck/default.aspx?pagename=robolink>.

44

17. igus GmbH. "igus robolink robot modules applications and references."
www.youtube.com.Web.
<http://www.youtube.com/playlist?list=PL07B939DF7512B3B4>.
18. "Jaco | Rehab Edition." www.kinovarobotics.com. 2012.Web.
<http://kinovarobotics.com/products/jaco-rehab-edition/>.
19. Kaye, H. S., Kang, T. and LaPlante, M.P. (2000). Mobility Device Use in the United
States. Disability Statistics Report, (14). Washington, D.C.: U.S. Department of
Education, National Institute on Disability and Rehabilitation Research.
20. "Kinova JACO Robot Manipulator (Research
Edition)." www.robotshop.com. 2013.Web. <http://www.robotshop.com/kinovatechnology-jaco-robot-manipulator-academic-edition-2.html>.
21. LIRECWRUT. "FLASH Mobile Dexterous Social Robot, LIREC,
WRUT." www.youtube.com. Mar 6 2012.Web. University of Technology, Poland.
<http://www.youtube.com/watch?v=o8kf14SZkaQ&list=PL07B939DF7512B3B4>.
22. Malte. "Humanoid robot A1: robolink arms." www.mtahlers.de. July 19 2010.Web.
<http://www.mtahlers.de/index.php/robotik/humanoider-torso/robolink-arme>.
23. Natasha. "Robotic Arm Developed By Columbia University, Can Be Controlled By
Facial Muscles." webmuch.com. May 2013.Web. <http://webmuch.com/robotic-armdeveloped-by-columbia-university-can-be-controlled-by-facial-muscles/>.
24. Neergaard, Lauran. "Paralyzed man uses brain-powered robot arm to
touch." www.nbcnews.com. Oct 10 2011.Web.
<http://www.nbcnews.com/id/44843896/ns/health-mens_health/t/paralyzed-man-usesbrain-powered-robot-arm-touch/#.UZz1gopDuKk>.
25. Ostrovsky, Gene. "DORA Robot Arm a New Accessory for
Wheelchairs." www.medgadget.com. Nov 24 2009.Web.
<http://www.medgadget.com/2009/11/dora_robot_arm_a_new_accessory_for_wheelchai
rs.html>.
26. "PerMMA - Personal Mobility & Manipulation Appliance." www.cmu.edu.Web.
<http://www.cmu.edu/qolt/AboutQoLTCenter/PressRoom/ces-2012/permma.html>.

45

27. Polito, Joe. "This Week in Research: Wheelchair-mounted robotic

arm." www.usforacle.com. April 4 2011.Web. <http://www.usforacle.com/this-week-inresearch-wheelchair-mounted-robotic-arm-1.2530642#.Uhe_7WRoQx9>.
28. puzzlebox. "Brain-Controlled Wheelchair Robot." www.instructables.com. Sept 3, 2012
2012.Web. <http://www.instructables.com/id/Brain-Controlled-Wheelchair-Robot/>.
29. Raak, Martin. "robolink Joint kit." robot.wpi.edu. Jan 2013.Web.
<http://robot.wpi.edu/wiki/images/7/7d/Robolink_Documentation.pdf>.
30. Rapacki, Erin B., Christopher Niezrecki, and Holly A. Yanco. "An Underactuated
Gripper to Unlath Door Knobs and Handles." robotics.cs.uml.edu. 2009.Web.
<http://robotics.cs.uml.edu/fileadmin/content/publications/2009/tepraFINAL_id333_rapa
cki_v2.pdf>.
31. Rapacki, Erin. "The Design of an Underactuated Wheelchair-Mounted Robotic Arm to
Unlatch Door Knobs and Handles." 2009.Web.
<http://users.wpi.edu/~nems/Rapacki.pdf>.
32. "Researchers Develop "Brain-Controlled" Wheelchair Robotic Arm." news.usf.edu. Jan
28 2009.Web. <http://news.usf.edu/article/templates/?a=1130>.
33. "Researchers Develop "Brain-Controlled" Wheelchair Robotic
Arm." www.singularityhub.com. Feb 11 2009.Web.
<http://singularityhub.com/2009/02/11/researchers-develop-brain-controlled-wheelchairrobotic-arm/>.
34. "Robotic Arm's Big Flaw: Patients in Wheelchairs Say It's 'Too
Easy'." www.sciencedaily.com. Sept 24 2010.Web.
<http://www.sciencedaily.com/releases/2010/09/100923184453.htm>.
35. "Robotic system may put more tasks within wheelchair users'
reach." www.reasearch.va.gov. Feb 3 2010.Web.
<http://www.research.va.gov/news/research_highlights/robotics020310.cfm#.UhvnXmRoQx9>.
36. Santos, Alexis. "Columbia University's low-cost robotic arm is controlled by facial
muscles, we go face-on (video)." www.engadget.com. May 4, 2013 2013.Web.
<http://www.engadget.com/2013/05/04/columbia-university-robotic-arm-controlled-byfacial-muscles-hands-on-video/>.

46

37. Schwartz, Ariel. "The Jaco Robotic Arm Can Scratch Your Back, Hands Out
Drinks." www.fastcompany.com. Jan 23 2011.Web.
<http://www.fastcompany.com/1720283/jaco-robotic-arm-can-scratch-your-back-handout-drinks>.
38. Sherer, Kyle. "Wheelchair-mounted robotic arm controlled by thought
alone." www.gizmag.com. Feb 10 2009.Web. <http://www.gizmag.com/brain-controlwheelchair-robotic-arm/10954/>.
39. Tremblay, Ty and Taskin Padir. Modular Robot Arm Design for Physical Human-Robot
Interaction. Systems, Man and Cybernetics (SMC), 2013 International Conference on.
Manchester, UK.
40. Tsuia, Katherine M., et al. ""I Want That": Human-in-the-Loop Control of a WheelchairMounted Robotic Arm." robotics.cs.uml.edu. 2011.Web.
<http://robotics.cs.uml.edu/fileadmin/content/publications/2011/2011-JAAB-tsui-et-alfinal.pdf>.
41. TullyFoote. "Documentation - ROS Wiki." www.ros.org. July 18, 2013 2013.Web.
<http://www.ros.org/wiki/>.
42. Vieru, Tudor. "Robotic Arms Inprove Life for Wheelchair
Users." news.softpedia.com. Nov 26 2009.Web.
<http://news.softpedia.com/news/Robotic-Arms-Improve-Life-for-Wheelchair-Users127981.shtml>.
43. "Visual Control Interface for Manus ARM." robotics.cs.uml/edu.Web.
<http://robotics.cs.uml.edu/research/current-research/visual-control-interface-for-manusarm/>.
44. Wang, Wei, and Yuki Suga. "Human in Loop Integration of an Arm Mounted Wheelchair
Robot Based on RT Middleware." 2010.Web.
<http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5756822&tag=1>.
45. Wang, Hongwu, et al. "Performance Evaluation of the Personal Mobility and
Manipulation Appliance (PerMMA)." Medical engineering & physics.0 Web.
46. "Wheel chair robot project." www.sugano.mech.waseda.ac.jp.Web.
<http://www.sugano.mech.waseda.ac.jp/project/wheelchair/index.html>.

47

47. "Woman Users Her Thought to Control Robot to Feed Chocolate - Amazing New
Technology news." www.youtube.com. Dec 17 2012.Web.
<http://www.youtube.com/watch?v=C7H_M8-dBHc>.
48. . " RAPUDARobotic Arm for Persons with Upperlimb DisAbilities." www.youtube.com. July 27 2010.Web.
<http://www.youtube.com/watch?feature=player_embedded&v=yqOUexGKusg#!>.

12 Appendix
12.1.1

Instructions for Use

Mounting & Startup:

o

To mount the arm simply slide the 80/20 square pipes into the mounting fixture at the
back of the CPS chair. The masking tape should line up with the fixture

Plugging in (Note: BLUE is PWR; WHITE is GND):

Attach Anderson Pole Connector to open 24V rail on CPS chair

Attach the Emergency Stop Button to the open dual lock tape on cable box

Attach TE Circular Connector

Turn on CPS and Robolink

If Robolink joints are in their marked startup positions (should point straight
out), the software limitations will be correct

If not, use Tys Xbox code (http://robot.wpi.edu/wiki/index.php/Robolink) to

put it in startup position, then power cycle

Running code:

https://github.com/robolink

To run the Robolink

Run roscore

Run roslaunch robolink robolink.launch

Run roslaunch robolink CommandAcceptor.launch

To run the CTI Joystick

Run roscore

Run roslaunch rad rad2_indjoy.launch

48

12.1.2

Repair History / Incident Report

Broken Motors in White Box:

M121594 002 - Broken Hall Sensor

o

Was the spare (6th) motor, never used due to broken sensor

M121594 004 - Bent Axle

o

Was the motor for the yellow cable (Joint 3), was found bent on 5/28/13 with
cable pulled out of metal fastener

Figure 27: Motor Cable Pulled from Fastener

M121594 003 - Bent Axle

o

Was the motor for the black cable (Joint 1), axle bent when the code crashed
while mounted on the CPS chair - the motors continued to run and end effector
pressed on the leg of the teammate seated in CPS chair. While not enough force
was exerted to injure the team member, the motor axle bent. After this, the
emergency stop button was added

Yellow cable was repaired after the bent axle incident

Repair consisted of cutting the cable and knotting on a new piece such that the knot
remains on the wheel through the range of motion

49

The knot was a strong sailors knot, sewn through several times (green thread), and
super-glued

Current status:
o

One motor axle slightly bent, still useable:

Figure 28: Current Motor Axle Bend

50

12.1.3

Presentation Slides

51

52

53

54

55

56

57

58

59

60

61

62

63

64