design of a wearable cobot - flvc
TRANSCRIPT
Florida State University Libraries
Electronic Theses, Treatises and Dissertations The Graduate School
2006
Design of a Wearable CobotJason Yap Chua
Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected]
THE FLORIDA STATE UNIVERSITY
FAMU – FSU COLLEGE OF ENGINEERING
DESIGN OF A WEARABLE COBOT
By
JASON YAP CHUA
A Thesis submitted to the Department of Mechanical Engineering
in partial fulfillment of the requirements for the degree of
Master of Science
Degree Awarded: Spring Semester, 2006
ii
The members of the Committee approve the Thesis of Jason Yap Chua defended on February 23, 2006.
___________________________ Carl A. Moore
Professor Directing Thesis ___________________________ Rodney Roberts Outside Committee Member
___________________________ Patrick Hollis Committee Member
Approved: ____________________________________ Chiang Shih, Chair, Mechanical Engineering ____________________________________ Ching-Jen Chen, Dean, FAMU-FSU College of Engineering The Office of Graduate Studies has verified and approved the above named committee members.
iii
For all my ancestors who worked so hard to give me a better life and for all those that come after me to continue our ancestors’ legacy.
iv
ACKNOWLEDGEMENTS
I would like to express my gratitude to Dr. Carl Moore for his guidance and assistance in my
research. I would also like to express my appreciation to my committee members, Dr. Patrick
Hollis and Dr. Rodney Roberts for serving on my committee and providing additional guidance.
I must thank Dr. Hollis again for instilling an interest in mechanical design in me through classes
I took with him. Thanks go out to NASA for funding my research. I would also like to thank all
my colleagues in the Center for Intelligent Systems, Control and Robotics Lab for their
friendship and support. A big thanks goes out to Dan Baxter and the FSU Physics Lab for
machining the parts designed. I am indebted to William Kincannon for culturing the art of
engineering in me through machining. I must thank Dr. David Cartes for being a voice of
experience and wisdom. Thank you to Alicia Fontaine for support from an understanding peer.
Thanks also go out to Keith Larson for advice and consultation on engineering and problems
beyond. I cannot thank my family enough for all that they have done for me. Special thanks
must be given to my Grandfathers Chua Yap Chek and Cheng Phi as well as my Father Yap
Siong Chua for having the courage to cross oceans and the skill and determination to succeed in
foreign lands. I am forever indebted to them.
v
TABLE OF CONTENTS
List of Tables ................................................................................................ vii List of Figures ................................................................................................ viii Abstract ...................................................................................................... x 1. INTRODUCTION ......................................................................................... 1 1.1 Cobot History........................................................................................ 1 1.2 Telerobotics and Telepresence ............................................................. 7 1.2.1 Telerobotics ................................................................................ 7 1.2.2 Telepresence ............................................................................... 7 1.2.3 Virtual Constraints ...................................................................... 8 1.2.4 Powered Cobot ............................................................................ 9 1.3 Application ........................................................................................... 11 2. MECHANICAL DESIGN ............................................................................. 15 2.1 Arm Cobot ........................................................................................... 15 2.2 Transmission Ratio Design .................................................................. 16 2.2.1 Cable Transmissions ................................................................... 16 2.2.2 Design Parameters ...................................................................... 17 2.3 Capstan Design and Cable Selection ................................................... 19 2.3.1 Design Considerations ................................................................ 19 2.3.2 Loading Design Calculations ...................................................... 22 2.3.3 Loading Design Results .............................................................. 25 2.4 Cable Wrapping Design ....................................................................... 27 2.5 Cable Tensioning Design ..................................................................... 29 3. Results and Conclusions ................................................................................ 31 3.1 Cable Mounting and Tensioning .......................................................... 31 3.2 Results ................................................................................................ 34 3.3 Conclusions .......................................................................................... 35 3.4 Future Work ......................................................................................... 36 APPENDICES ................................................................................................ 38
vi
A Sava Cable Catalog .............................................................................. 38 B MathCAD Calculations ........................................................................ 51 C Engineering Drawings .......................................................................... 58 REFERENCES ................................................................................................ 105 BIOGRAPHICAL SKETCH .............................................................................. 107
vii
LIST OF TABLES Table 1: Joint Force (Col. 1-8, lb)/Torques (Col. 9-10, lb-in)............................. 18 Table 2: Cable Breaking Strength Adjustment ................................................... 21 Table 3: Sample Listing of Cables for Use in Transmissions ............................. 23 Table 4: Specification of Shoulder Reduction Stages ......................................... 25 Table 5: Specification of Elbow Reduction Stages ............................................. 25 Table 6: Specification of Forearm Reduction Stages .......................................... 26
viii
LIST OF FIGURES Figure 1: Virtual walls created by actuated steering .......................................... 1 Figure 2: Unicycle Cobot .................................................................................... 2 Figure 3: Unicycle Two-Link Cobot ................................................................... 3 Figure 4: Scooter Cobot ...................................................................................... 3 Figure 5: The scooter cobot guides the car door assembly ................................. 4 Figure 6: The Arm Cobot ................................................................................... 5 Figure 7: The Spherical CVT .............................................................................. 6 Figure 8: Telerobotic System With Visual and Position Feedback .................... 7 Figure 9: Telerobotic System with Virtual Constraints ...................................... 8 Figure 10: Traditional Robotic System ............................................................... 9 Figure 11: Multi-Degree of Freedom Cobotic System ....................................... 10 Figure 12: The DAWP Platform ......................................................................... 11 Figure 13: 6-DOF Cobot Master Controller ....................................................... 12 Figure 14: NASA’s Robonaut ............................................................................. 13 Figure 15: Force Reflecting Hand Controllers ................................................... 13 Figure 16: Wearable Arm Cobot ......................................................................... 15 Figure 17: Capstan Pair with Positive and Negative Torque Cables .................. 16 Figure 18: Experiment Configurations ............................................................... 17 Figure 19: Cobot Arm CG Location and Endpoint Loading for Shoulder Joint 19
ix
Figure 20: Multi-Stage Transmission and Equation for Adding Stiffnesses ...... 20 Figure 21: Pretensioning in Capstan Pairs .......................................................... 21 Figure 22: Wearable Arm Cobot ......................................................................... 26 Figure 23: Cable Travel in a Spooling Capstan .................................................. 27 Figure 24: Uneven Cable Travel in Capstan Pairs .............................................. 28 Figure 25: Sandwiched Capstan Tensioning Scheme ......................................... 30 Figure 26: Capstan Tensioning Scheme ............................................................. 31 Figure 27: Wearable Cobot Shoulder Transmission ........................................... 32 Figure 28: Cable Mounting Points ...................................................................... 33 Figure 29: Cables Spooling into Capstan Grooves ............................................. 34
x
ABSTRACT
Cobots are specially designed robots that use continuously variable transmissions (CVTs)
instead of traditional motor driven joints. Cobots are an attractive choice for telerobotic master
controllers because they are safe in contact with humans and are able to produce stable, high
quality virtual surfaces that can constrain the motion of the master to directions suitable for
telerobotic task completion. This thesis describes the design of a 3-DOF, cable driven, wearable
arm master controller and provides the details on the construction and assembly of the shoulder
joint. Solidworks CAD was used to design the wearable cobot. After shoulder capstans and
shafts were machined the device was assembled and tested. The rotational stiffness for the
shoulder joint was found to be 1.58*105 lb-in/rad and the start torque was found to be 16.5 lb-in.
Improvements and future work are also discussed.
1
CHAPTER 1
INTRODUCTION
1.1 Cobot History
Cobots, or ‘COllaborative roBOTs’ are robots designed to present high quality
constraints safely to human users [1]. The initial driving force behind the invention of
cobots was to meet the need for a robot that could interact safely with automobile
assemblers. There are presently several different cobots, but the defining feature
common to all cobots is the continuously variable transmission (CVT) used in place of
traditional robot joints. The CVTs kinematically separates the cobot’s speed and
direction. For example, the unicycle cobot [1] has one steered wheel rolling on a flat
plane. The steered wheel is a CVT that controls the x and y-axis translational velocity
ratio as a continuous function of the wheel’s heading angle. The actuator, instead of
controlling the rotational velocity of the wheel, is used to steer the wheel.
Fig. 1. Virtual Walls Created by Actuated Steering
2
Fig. 1 shows this virtual wall created by actuated steering where upon reaching the
virtual wall the wheel is steered to roll along the line of the virtual wall. The downward
force and coefficient of friction determines the strength of the virtual wall.
The unicycle cobot, shown below in fig. 2, was created for research purposes and is a
purely passive device.
Fig. 2. Unicycle Cobot [1]
The Unicycle Two-Link Arm (UTLA) in fig. 3 was created for rehabilitation
research. The UTLA is spring loaded at the second rotational joint to provide a constant
down-force at the unicycle. The UTLA operates in 3 different modes. A virtual caster
mode senses the user’s intended motion direction and steers the wheel to accommodate.
The virtual path mode constrains the user to a specified path. The third mode, virtual
wall mode, is a combination of the first 2 modes in which the user may move freely in
certain areas but will be met with a virtual wall upon moving to the edge of the free
movement area.
3
Fig. 3. Unicycle Two-Link Cobot [2]
The scooter cobot in fig. 4 was designed for industrial material handling. It uses three
steered wheels to guide the platform.
Fig. 4. Scooter Cobot [3]
In free mode the scooter operates similarly to a chair on casters. In constrained mode
it follows a specified path but retains the ability to turn on its central axis. The Scooter
cobot was originally designed to assist in automobile assembly by allowing fewer
workers to place and remove parts, like a car door, accurately and easily as shown in fig.
5.
4
Fig. 5. The Scooter Cobot Guiding Car Door Assembly [3]
The arm cobot in fig. 6 was created for industrial design research and is able to create
a 3 dimensional virtual surface in the traditional x, y, z space. The user can qualitatively
determine car ergonomics such as door opening, from interaction with the arm cobot’s
end-effector. For example, the arm cobot can constrain the motion of the end effector to
simulate a sliding path of a door handle when opened by an occupant. The simulation
allows the user to experience the height and path of the door handle without creating a
physical prototype.
5
Fig. 6. The Arm Cobot [4]
The arm cobot’s revolute joints are connected to spherical CVT’s that allow for these
virtual constraints to be formed. The spherical CVT is shown below in fig. 7. In the arm
cobot, the input shafts of the three spherical CVT’s are connected to each revolute joint.
Each output shaft of the spherical CVT’s are connected to each other linking the CVT’s
mechanically. The spherical CVT creates a speed ratio between the input and output
shafts using a transmission sphere. Actuated steering wheels control the spinning axis of
the sphere allowing for different rolling speeds of the input and output shafts. The
actuated steered wheels in the spherical CVT are analogous to the steered wheel in the
unicycle cobot.
6
Fig. 7. The Spherical CVT [4]
As shown, previous cobots were created mainly for research, rehabilitative and
industrial uses. However, the unique qualities of intrinsic safety in contact with humans
and ability to create stable, high quality constraints make cobots ideal for telerobotics
implementation. The following sections will discuss telerobotics and the application of
cobots in telerobotics.
7
1.2 Telerobotics and Telepresence
1.2.1 Telerobotics
A typical telerobotic system consists of two robots. The follower, or slave robot,
performs a task at the direction of a human controlled master robot at a remote location.
This system is typically used when human presence at the task site is either impossible or
poses too great a danger to the human.
The first telerobotic systems typically consisted of a robotic manipulator at a remote
site controlled by a control panel or joystick. A user would manipulate the control panel
or joystick while observing the robot on a monitor. Since this type of system has only
visual and position feedback it is difficult to complete tasks.
1.2.2 Telepresence
Telepresence is a property of telerobotic systems that describes the amount and
quality of feedback from the slave robot and its surroundings [5]. The telerobotic system
shown in fig. 6 is indicative of many early telerobotic systems and consists of a master
robot controlling a manipulator with a pencil at a remote site. These systems exhibited a
minimal amount of telepresence using only position control and visual feedback control.
In the case of fig. 8 is would be relatively easy for the user to accidentally commence a
motion causing the slave to break the pencil. Without increased telepresence it would be
nearly impossible to handle unstable materials or perform surgery via a telerobotic
system.
Fig. 8. Telerobotic System With Visual and Position Feedback
8
1.2.3 Virtual Constraints
Greater bandwidth in electronic communication and advances in force transducer
technology have allowed the integration of effective force feedback in telerobotic
systems. With force feedback the user can prevent breaking the pencil tip in the fig. 8 by
detecting the amount of downward force applied by the slave on the pencil.
Pictured below in fig. 9 is an example of a telerobotic system equipped with virtual
constraints.
Fig. 9. Telerobotic System with Virtual Constraints
The objective is to draw a straight line at the remote site. The user controlling the
master robot cannot easily draw a straight line without the aid of a straight edge. If the
master controller could be programmed with a straight edge virtual constraint the user
could very easily complete the task. In this situation virtual guides help constrain the
user motion to a straight path and the user determines the speed at which the line is
drawn. This can be easily applied to creating constraints in a multi-dimensional
workspace [6].
For a virtual constraint to be useful it must be firm enough to resist forces against it
and smooth enough that is does not increase frictional drag on the master controller. It is
difficult to implement this control concept safely on traditionally actuator driven robot
systems. For example, when the human user applies a force perpendicular to the virtual
9
ruler constraint in fig. 9, the master robot should respond with an equal and opposite
constraint force to counter that of the human user. The actuators used in a traditional
robot setup are capable of creating large and unpredictable joint velocities that could be
life threatening to the human user [4]. Furthermore, the control structure in a traditional
robot system employs a zero order hold. The zero order hold and sampling of the robots
real-time sensor data creates a situation in which the user can extract energy from a
virtual constraint that is supposed to feel passive [7, 8]. As the user repeatedly uses the
constraint, the repeated energy addition and subtraction could cause the manipulator to
become unstable. Creating high quality virtual constraints for traditional robot systems is
still difficult despite maximized sensor resolution and sampling rates. The use of
powered CVT’s in place of traditional actuators can yield firm, high quality constraints.
1.2.4 Powered Cobot
In a traditional robotic system the actuator is placed directly on each joint shown in
fig. 10. These actuators act as power level and signal level actuators, which requires
actuators of relatively high torque production to maintain rigid virtual constraints. With
large actuators maintaining safety is problematic. Safety and smoothness of the system
can be increased by separating the power level and signal level duties among an increased
number of actuators [1]. A powered cobot can fulfill this requirement.
Fig. 10. Traditional Robotic System
10
A powered cobot with ‘n’ DOF requires ‘n + 1’ actuators, one more than a
conventionally actuated robot of the same dimensionality [1]. In a cobot, ‘n’ number of
actuators are used to adjust the transmission ratios or ‘steer’ each CVT. This steering
angle is denoted by “γ” in fig. 11. The CVT steering actuators are signal level actuators
that cannot affect the speed of the cobot’s endpoint. This is tantamount to a driver in a
parked car turning the steering wheel; while the proposed heading is changed there is no
motive force to realize that change in direction.
The remaining actuator is the power level actuator, denoted by “τ0”, used to amplify
the user’s motions. In the arm cobot each joint is coupled to a CVT. Each CVT is joined
in parallel by a common shaft driven by the power assist actuator, shown in fig. 11. This
powered cobot architecture relies on only one power level actuator which allows for a
more manageable system in terms of control as compared to a system with three power
level actuators.
Fig. 11. Multi-Degree of Freedom Cobotic System
11
1.3 Application
Agencies such as NASA and Argonne National Laboratories are calling for more
sophisticated master robot systems that can provide increased telepresence through the
use of cobot created virtual constraints. The Dual Arm Work Platform (DAWP), shown
with a control station in fig. 12, at the Argonne National Laboratory is used for
disassembly in nuclear reactor cores [9]. The platform has strategically placed cameras
that allow remote control via a series of joysticks and monitors. This particular setup in
fig. 8, provides a limited amount of telepresence, as discussed previously.
Fig. 12. The DAWP Platform [10]
On the DAWP platform, tools are often broken in the middle of performing an
operation. After replacing the tool, 90% of the operation time is spent repositioning the
tool at the original worksite [9].
The time spent repositioning the tool could be greatly reduced if a greater
telepresence at the remote site could be established. This can be achieved by using force
feedback coupled with virtual fixtures [6]. Force feedback would allow the user to sense
the amount of force imposed on the tool, thereby aiding the user in the prevention of tool
breakage. Should a tool break, virtual constraints could be created based on the position
12
of the previous tool worksite to guide the user back to the exact site of the work quickly
and accurately.
A 6-DOF cobot master controller, shown below in fig. 13, has been created to replace
the outdated position feedback joystick controllers that were formerly in use [9]. This
controller employs force feedback and virtual constraints in a joystick interface to aid the
user in task completion.
Fig. 13. 6-DOF Cobot Master Controller [9]
NASA’s Robonaut project intends to aid and even replace the need for manned space
walks by teleoperation of a robot possessing the same configuration and dexterity as a
human, shown in fig. 14 [11]. It is possible that Robonaut could exhibit the same
problems in operation as the DAWP.
13
Fig. 14. NASA’s Robonaut [12]
Currently, Robonaut can function as an anthropometric tool in simple task completion
such as picking up and carrying items, inserting bolts into holes and carrying out other
simple instructions via hand signals. Remote control of Robonaut includes the use of two
Force Reflection Hand Controllers (FRHC). These controllers, shown in fig. 15, can
provide force feedback for the user at the hands.
Fig. 15. Force Reflection Hand Controllers [12]
14
While this control setup provides more telepresence than a simple joystick, creating
an even more intuitive device that yields a 1-to-1 correlation in motion between the user
and master robot can increase telepresence even further. A wearable master cobot would
allow for expansion of force feedback to include individual joints of the user’s arm
instead of only imposing forces at the user’s hand. Furthermore, the presence of virtual
constraints via-powered CVT, could help guide Robonaut’s tools.
The author’s design proposed in this thesis will be used in the development of a
master controller for Robonaut. While Robonaut is fully anthropometric in its arm
design, a 3-DOF wearable cobot design will be explored in lieu of a more complex 7-
DOF cobot design.
15
CHAPTER 2
Mechanical Design
2.1 Arm Cobot
The arm cobot design created by Dr. Carl Moore, not to be confused with the
wearable arm cobot design discussed in this paper, integrated 3 spherical CVTs with a 3-
DOF manipulator, but suffered problems with backlash and compliance. The wearable
arm cobot design aims to eliminate backlash through the use of cable transmissions while
minimizing device weight and maintaining joint stiffness.
Fig. 16. Wearable Arm Cobot
16
The wearable arm cobot design, shown above in fig. 16, will be mated to a new
spherical CVT design currently in development. The previous spherical CVT design
employed polyurethane in-line skate wheels. These softer rubber-like wheels allowed
compliance and creep during CVT operation. This compliance at the CVT is propagated
out to the endpoint of the manipulator and is contraindicative to forming hard, high
quality virtual constraints. The new spherical CVT design will employ metal-to-metal
contact yielding a more controllable system with a compact design and a maximum
sustainable input torque of 5 lb-in.
In the opinion of the writer, the literature concerning cable transmission design is
limited and shallow in exposition. In the following sections the design of a cable
transmission will be discussed in depth including figures and calculations.
2.2 Transmission Ratio Design
2.2.1 Cable Transmissions
Cable and belt transmissions have been widely used in power transmission in which
the belt or cable is run in only one direction. In robotic applications it becomes necessary
to accommodate for both positive and negative torques created at the joint of a
manipulator. In this paper, the cable transmission design will consist of multiple pairs of
capstans and complimentary pairs of cables for positive and negative torque as shown in
the figure below.
Fig. 17. Capstan Pair with Positive and Negative Torque Cables
17
Cable transmission design in this paper is approached from 2 directions using an
iterative method. At one end of the transmission is the CVT, which has a maximum slip
torque of 5 lb-in. At the other end of the transmission, the end-effector of the wearable
cobot, is a maximum loading requirement reflecting that of the maximum torque a human
can sustain. After initial calculations from one end of the device the calculations are
iterated back and forth to find a set of variables that satisfy the requirements of a
telerobotic master controller.
2.2.2 Design Parameters
In order to maintain safety in the wearable cobot design the maximum torque a
human can sustain must be reduced to accommodate the 5 lb-in maximum input torque of
the CVT. Only a very general range of maximum sustainable torques is needed as the
CVT can be adjusted to accommodate the user. The transmission ratio for each DOF is
determined by dividing the maximum sustainable user torque by the maximum
sustainable CVT input torque (5 lb-in).
Fig. 18. Experiment Configurations
18
In-lab tests were conducted to determine the maximum sustainable joint torques of a
human-male for the shoulder and elbow joints including forearm rotation. Fig. 18 depicts
the experiments performed. The grey dots represent the joint axis that was tested and the
accompanying arrows depict the direction the arm was exerting force or torque.
Experiments 9 and 10 tested forearm torque and are depicted in fig. 18 by a head-on view
of the test subject’s right fist. All test subjects were male graduate students 22-29 years
of age. The experimental data in fig. 18 coincide with the numbers in table 1.
Table 1. Joint Force (Col. 1-8, lb)/Torques (Col. 9-10, lb-in)
Subject
Forearm
Length
Upper
Arm
Length
1
2
3
4
5
6
7
8
9
10
1 13.5” 12” 10 15 18 22 12 15 18 11 21.25 20
2 11.5” 11” 11.5 15 25 25 15 9.5 23 20 16.25 17.5
3 11.75” 11.5” 21 25 25 30 14 18 24 13 15 22.5
4 11.5” 10.75” 12 16 15 25 14.5 11 15 12 17.5 20
To maintain safety for a more generalized range of users, an average was taken for
the maximum sustainable joint torques at the weakest arm orientation. These values are
14 lbs (column 8), 13.62 lbs (Column 1) and 17.5 lb-in (column 9) for the shoulder,
elbow and forearm rotation joints, respectively. When these values are added to the
design target weight of 13 lbs at 12.6 inches from the shoulder, when arm is fully
extended, and divided by the maximum sustainable input torque of the CVT, the
transmission ratio can be found.
The center of gravity for the lower arm is 3.5 lbs at 5 inches from the elbow. The
forces exerted on the end-effector are at 24 inches from the shoulder and 12 inches from
the elbow. The target weight and maximum loadings can be seen in fig. 19 below. The
diameter of the wrist sleeve is 5 in. The transmission ratios for the shoulder, elbow and
forearm rotation are 100:1, 36:1 and 5:1 respectively. Note that the transmission ratio for
the forearm rotation is necessary to maintain joint stiffness rather than to reduce the
forearm torque, this will be further explained in later sections.
19
Fig. 19. Cobot Arm CG Location and Endpoint Loading for Shoulder Joint
2.3 Capstan Design and Cable Selection
2.3.1 Design Considerations
Joint stiffness and cobot arm weight are the main considerations during capstan
design and cable selection to create a highly controllable master manipulator. It is
kinematically advantageous to have a very light but very rigid mechanism to maintain the
integrity of virtual constraints.
In the development of the Whole Arm Manipulator (WAM) the equation for
mechanical torsional stiffness was derived [13]. Mechanical torsional stiffness k for a
cable transmission is given by:
2 21*
2 (1 )
N R AEk
NL x N=
+ − (2.1)
The variables in eqn. 2.1 are defined as follows. The cable manufacturer, Carl Stahl
Sava Cable (See App. A), sells cable with a Young’s modulus (E) of 6.3 Mpsi. ‘A’ is the
cross sectional area of the particular cable chosen. The reduction site location ‘x’ and the
transmission length ‘L’ are set equal to each other since speed reduction is done at the
joint. The capstan size ratio (N) is the larger capstan diameter divided by the smaller
20
capstan diameter and the driven capstan radius (R) is usually the radius of the larger
capstan in the pair.
For multi-stage transmissions, successive values of torsional stiffness k can be added
similarly to resistors in parallel to find the effective stiffness keff as shown in fig. 20.
Fig. 20. Multi-Stage Transmission and Equation for Adding Stiffnesses
To preserve the integrity of the cables, a few manufacturer suggested guidelines must
be followed. The capstan diameter should generally be at least 25 times the cable
diameter and the maximum dynamic operating load should be no more than 10% of the
cable breaking strength. The design guidelines give a strength percentage according to
variation of capstan diameter to cable diameter ratio as seen in table 2 below. More in-
depth design guidelines regarding mounting procedure and cable selection can be found
in appendix A. Furthermore, to preserve the stability and uniformity of cable
performance there should be a relatively consistent cable line stress for each
transmission.
21
Table 2. Cable Breaking Strength Adjustment
To eliminate backlash the cable transmissions must be pretensioned. The WAM
manipulator uses a pretension of the added static and dynamic cable loadings to prevent
the complimenting cable from going slack under high loading. For example, if a total
maximum torque is 100 lb-in and the capstan it is working on has a 5 in radius, the line
load in the cable will be 20 lbs. This requires that the pretension on each cable (positive
and negative torque) be at least 20 lbs to prevent cable slack as one cable line load will go
to 40 lbs and the other cable’s load will go to 0 lbs as illustrated below in fig. 20. It is
advantageous to set the cable pretension at higher than that of the loading.
Fig. 21. Pretensioning in Capstan Pairs
22
Keeping these guidelines in consideration, only the variables of cable size, capstan
size and transmission ratio remain. Capstan size is chosen based on readily available
material and number of reduction stages as well as aesthetics. The cable transmissions
and their housing were created to be hidden somewhat by the user’s body when viewed
from the front. Cable size is chosen to fulfill the above guidelines and to give an
appropriate stiffness. These calculations are then further iterated to give a reasonable
agreement among the different guidelines.
2.3.2 Loading Design Calculations
The calculations discussed in this section include the calculations for one pair of
capstans and are taken from the final iteration of the shoulder transmission design for the
wearable cobot. The complete calculations for the shoulder, elbow and forearm rotation
joints can be found in appendix B.
In section 2.2.2 the maximum user joint torque was found to be 14 lbs at 24 inches, or
336 lb-in. The target arm weight was 13 lbs at 12.6 inches creating the design target
torque of 163.8 lb-in. When the maximum user torque and the design target torque are
added the total maximum torque T at the shoulder joint is just under 500 lb-in. A 100:1
reduction is necessary to accommodate the CVT input torque of 5 lb-in. This reduction
will be divided among three stages of 2:1, 5:1 and 10:1. Only the first stage reduction of
2:1 will be discussed in detail.
The shoulder capstan diameter was chosen to be 6.25 in. and the complimenting
smaller capstan would have a diameter of 3.125 in. In eqn. 2.1, the capstan size ratio ‘N’
is given by the larger capstan diameter divided by the smaller capstan diameter.
6.25
23.125
inN
in= = (2.2)
23
Now that a shoulder capstan size has been chosen a cable can be chosen. Shoulder
torque ‘T’ is divided by the shoulder capstan radius to determine the cable line loading
‘CL’.
500160
3.125
lb inCL lb
in
−= = (2.3)
The cable chosen must have a breaking strength of 10 times the working load,
therefore a cable of 1600 lb breaking strength or more should suffice. An abridged table
of available cables is shown below in table 3.
Table 3. Sample Listing of Cables for Use in Transmissions
Cables 2124, 2125 and 2126 are all suitable for use with the shoulder capstan but the
best choice is cable 2126 for its 7x19 construction allowing for greater cable flexibility.
An explanation of cable construction can be found in appendix A. This cable also has the
properties of a 1760 lb breaking strength and a diameter of 1/8 in. At this point the cable
breaking strength must be adjusted according to table 2 in section 2.3.1. The minimum
recommended capstan diameter for a cable of 7x19 construction is 25 times the cable
diameter or in this case 3.125 in., the size of the complimenting smaller capstan. The
effective breaking strength must now be reduced to 92% of that stated in table 2.
0.92*1760 1619lb lb= (2.4)
It is now confirmed that the adjusted breaking strength is still 10 times the working
load. The line stress must now be determined for comparison to the line stresses in the
24
other two stages of reduction. Having similar line stresses in a transmission allows for a
more uniform behavior in cable stretch.
Again, the cable loading ‘CL’ is 160 lb and the cross sectional area ‘A’ of a 1/8 in.
diameter cable is 0.0123 in2. The stress ‘S’ is given below in eqn. 2.5.
2
16013040
0.0123
lbS psi
in= = (2.5)
The stresses for the other two stages were 10430 psi and 12430 psi. While the three
stresses are not equal, they are within 80% of each other. Further investigation of this
phenomenon is warranted in future research with particular focus on cable cycle life and
the affect of the different stresses on overall device performance
After having verified the cable and capstan diameter selection, the stiffness for this
stage can be found using eqn. 2.1. The transmission length ‘L’ is 9.66 in., which is
determined by allowing for clearances between capstans and CVT shafts. The driven
capstan radius ‘R’ is 3.125 in. The reduction position ‘x’ is set equal to the transmission
length ‘L’ to maximize stiffness and the young’s modulus ‘E’ for the cables is 6.3 Mpsi.
The rotational stiffness ‘k’ is given below in eqn. 2.6.
2 2 252 *(3.125 ) *0.0123 *6.31
* 1.563*102 9.66
in in Mpsik lb in
in= = − (2.6)
The transmissions for the elbow and wrist rotation are calculated in a similar matter.
Complete calculations can be seen in appendix B.
The overall torsional stiffness keff for the shoulder transmission was found to be
1.049*105 lb-in. The linearized deflection at the end-effector of the cobot arm can be
calculated assuming a rigid arm. The angular deflection ‘D’ can be found by dividing the
total maximum shoulder torque ‘T’ by the effective torsional stiffness ‘keff’. This can be
seen below in eqn. 2.7.
3
5
5004.765*10
1.049*10
lb inD rad
lb in
−−= =
− (2.7)
25
The sine of D is then multiplied by the arm length of 24 in. to yield the linearized arm
deflection (0.114 in.) due to transmission stiffness in eqn. 2.8.
sin( )*24 0.114D in in= (2.8)
An acceptable deflection in the cobot arm was determined by approximating the
deflection in the flesh of a human fist against a tabletop. The writer observed that the
flesh at the bottom of his own fist deforms when pressure is applied allowing for up to
0.25 inches of deflection. While no formal testing was conducted in this area it does
warrant further research to determine a quantification of a ‘good’ virtual constraint metric
for robot-human interaction.
2.3.3 Loading Design Results
The results of the calculations and iterations in finding the transmission specifications
are given below in tables 1, 2 and 3 for each transmission stage.
Table 4. Specification of Shoulder Reduction Stages
Stage A1-A2 A2-A3 A3-A4
Capstan Size Ratio
(inches)
6.25 : 3.125
10.0 : 2.0
10.0 : 1.0
Rotational Stiffness
(lb-in/rad)
1.563 * 105
1.344 * 106
4.186 * 105
Line Stress
(psi)
13040
10430
12430
Table 5. Specifications of Elbow Reduction Stages
Stage B1-B2 B2-B3
Capstan Size Ratio
(inches)
6.25 : 2.0 11.25 : 1.0
Rotational Stiffness
(lb-in/rad)
1.2 * 105 7.804 * 10
5
Line Stress
(psi)
12020 12730
26
Table 6. Specification of Forearm Reduction Stages
Stage C1-C2 C2-C3 C3-C4
Capstan Size Ratio
(inches)
5.0 : 2.0
2.0 : 1.0
3.0 : 3.0
Rotational Stiffness
(lb-in/rad)
6.034 * 103
1.319 * 103
8.768 * 103
Line Stress
(psi)
8704
8704
190.11
The nomenclature used in tables 4, 5 and 6 coincides with fig. 21 of the entire
wearable cobot design below.
Fig. 22. Wearable Arm Cobot
27
2.4 Cable Wrapping Design
Cable wrapping design requires that a cable spool consistently onto its capstan to
ensure consistent performance. Cable wrapping design includes cable spacing to prevent
scrubbing, supporting the cable on 1/3 of its circumference and consistency in axial cable
travel. These three criteria can be fulfilled through the use of a helical groove for the
cable to spool into.
Scrubbing in cable wrapping is the rubbing of cables as they spool and unspool on a
capstan. This leads to cable wear and loss of cable strength due to cable fatigue
decreased cross-sectional area. Scrubbing can be prevented by cutting a helical groove
for the cable to spool into with a helix height slightly more (~0.005 in) than that of the
cable diameter. The capstans in the wearable cobot design have helical groove cuts that
will support the cable on 1/2 of its circumference.
Axial cable travel is inherent to cable transmissions but seldom addressed in design.
It must be noted that spooling a cable onto a capstan requires that it takes up space in the
axial direction of the capstan. Below, in fig. 22, a capstan spooling a cable onto itself is
depicted. The beginning position is denoted by ‘X0’ and the distance the cable travels
axially is denoted by ‘d’.
Fig. 23. Cable travel in a spooling capstan
28
Consistency in axial cable travel between capstan pairs is crucial in maintaining
consistent cable line tension. Uneven cable travel between 2 capstans forces stretch in
the cable. This stretch, in turn, creates a tension in the cable that will adversely affect
device performance, as the cable tension will become a function of loading and capstan
position. The added cable stress may also shorten cable life by driving the working load
beyond design specifications. In a 1/8 in. cable (sava #2126) stretch on the order of
0.001 in. yields approximately 10 lbs of cable tension. Large amounts of cable stretch are
created by incorrectly mounting capstans or through faulty capstan design.
Take, for example, a capstan pair of 1 inch (capstan 2) and 5 inch (capstan 1)
diameters. They both share a helix of 0.1 inch per revolution. The starting position of
the cable is denoted by the boxed ‘1’ and the final cable position, after one revolution of
capstan 1, by the boxed ‘2’ in figure 23. One revolution of capstan 1 will cause capstan 2
to turn 5 times. Since the helixes of the capstans are not matched according to the axial
travel and capstan diameter the axial cable travel at capstan 2 is 5 times that of capstan 1.
This would cause approximately 0.013 inch of cable stretch over a transmission length of
6 inches. Cable stretch of 0.013 inch will cause very noticeable, if not problematic,
changes in cable tension in all but the smallest of cables
Fig. 24. Uneven Cable Travel in Capstan Pairs
Matching axial cable travel prevents stretching and keeps cable tension consistent
throughout the range of operation; this is achieved by taking the helix of the smaller
capstan and multiplying it by the ratio between the 2 capstans. For example, if a size
29
ratio between capstans is 10 to 1 then every 1 turn on the large capstan will produce 10
turns on the smaller capstan. Since the cable is wrapping and unwrapping from a helix,
the axial travel distance over 1 turn on the large capstan should equal the same axial
travel distance as 10 turns on the smaller capstan. This will yield a matched axial cable
travel preventing cable stretch.
The equation for matching axial cable travel through the respective capstan helixes is
given below. The ratio of the large capstan radius (D1) to the small capstan radius (D2)
multiplied by the small capstan helix height per turn (H2) should equal the larger capstan
helix height per turn (H1).
12 1
2
DH H
D= (2.9)
Please note that the limiting factor in this design is the helix on the smaller capstan as
it has more wrappings than the larger capstan. Therefore it is advantageous to design
capstan thickness starting from the smaller capstan.
2.5 Cable Tensioning Design
Cable transmission design depends on pre-tensioning the transmission cables to
increase stiffness and eliminate backlash [13]. A pre-tensioning method that is simple,
compact and non-destructive to the capstan shafts is desirable. A common method for
smaller operational torques is to sandwich the capstans between a retaining ring and nut
creating friction between the sandwiched capstans to maintain cable tension. The pre-
tensioning of the sandwich capstan design may require 2 people to achieve. The
wearable cobot is designed to accommodate torques of up to 500 lb-in at the shoulder
joint and while the static calculations confirm that the sandwiched capstan design would
support this torque, it is difficult to investigate creep between the 2 capstans.
In the sandwiched capstan method, shown in fig. 23, a ball termination is swaged
onto both ends of the cable. These fittings are placed into the capstan and can be
tensioned using opposing torques created by conventional wrenches via the capstan
30
extensions. At this point nuts are used to create frictional forces between the capstans
that will hold the cable tension. This method has been used in cable tensioning for the
WAM manipulator [14].
Fig. 25. Sandwiched Capstan Tensioning Scheme
In the wearable cobot the cable tensioning design consists of 2 cables, one for each
the positive and negative torques. Each cable is hard connected to the capstans with
fittings. Below is a diagram of the tensioning scheme in fig. 24. In the wrist
transmission the sandwiched capstan tensioning method is used because of the relatively
small capstan sizes and forces.
31
Fig. 26. Capstan Tensioning Scheme
The cable is outfitted with a threaded or ball termination fitting on either end. The
ball termination fits into the small capstan and the threaded termination fitting threads
into a block that fits into the large capstan. A bolt is threaded into the other side of the
threaded block and is used to tighten the cabling. In the next chapter we will verify this
tensioning scheme as well as cable wrapping methods and also verify shoulder joint
stiffness.
32
CHAPTER 3
Results and Conclusions
3.1 Cable Mounting and Tensioning
The guidelines for mounting and tensioning the shoulder transmission are discussed
in this section. These guidelines are applicable to all transmissions designed in this
paper. An isometric view of the shoulder transmission is shown in fig. 27.
Fig. 27. Wearable Cobot Shoulder Transmission
33
The cable used in this transmission must be carefully mounted to avoid damage to
itself. As mentioned in the previous section the cable is held into the capstans by either a
swaged ball fitting or a threaded lug fitting as shown in fig. 28.
Fig. 28. Cable Mounting Points
The ball fitting is held by a hole drilled into the side of the smaller capstan. The
threaded lug fitting is threaded into a block, which is placed into the larger capstan and
tensioned by a bolt.
34
Fig. 29. Cables Spooling into Capstan Grooves
The cable must rest easily within the machined groove on each capstan as shown in
fig. 29. Rhythmic creaking of the cable is an indicator of a cable resting into a groove
that is too small or other mechanical problem. Oiling the cable and groove with a
multipurpose lubricant will help prevent some premature cable wear and also oxidation of
the steel cable.
To create the correct amount of tension, locking pliers and a torque wrench must be
used. Suppose a cable should be tensioned to 10 lbs and it is on a capstan of 10 inches.
The complementing smaller capstan shaft must be held steady with the locking pliers and
the torque wrench must impose a torque of 100 lb-in on the 10-inch capstan shaft.
Plucking the cable and recording the tone produced will indicate the 10 lbs of tension.
35
The transmission stage must then be tensioned via tensioning bolts until the cable
produces that same tone. The tensioning method is explained below.
When tensioning the cable each stage must be tensioned individually and completely
independent of the other stages. This means that if certain stages share common shafts,
such as stage 1 and stage 2, stage 1 must be mounted and tensioned first before stage 2
capstans can even be placed on its respective shafts. This process will allow for the
greatest facility in cable tensioning, as manipulation of key components may be
necessary to achieve the correct cable tension.
To achieve the desired tension a specific process must be followed. Tensioning, i.e.
the turning of the bolt to tension the cable, must only occur at the extreme of capstan
travel. This will allow for greater ease of cable movement as the contact area between
the cable and the groove is minimized. This is achieved by turning the capstan to the
extreme of its travel and then turning the appropriate tensioning bolt. The process is then
repeated for the complimenting bolt by turning the capstan set to its other extreme of
travel.
The tensioning of the cable at the extreme of travel will create a high, local tension
from the large capstan to the initial spooling site of the smaller capstan. When the
capstans are turned back and forth the tension will distribute evenly across the entire
cable. The tensioning process must be repeated several times (4-5 times) to ensure the
appropriate and uniform tension throughout the cable.
3.2 Results
The starting torque for the shoulder transmission was found to be approximately 16.5
lb-in. The linearized deflection of the arm under maximum loading is 0.566 inches.
Empirical testing shows 0.49 inches of this deflection can be attributed to the bending of
the arm. Therefore the linearized deflection due to cable compliance at 24 inches from
the shoulder joint is 0.076 inches.
Shaft A3 suffers from a noticeable amount of deflection when under full loading.
Shaft A3 and shaft A4 bow towards each other due to the loading. This bowing creates a
variable cable tension due to inconsistent transmission length. The cable groove width on
36
the large capstan on shaft A3 was machined narrower by 0.1 inches. This also creates a
variable cable tension due to inconsistent axial cable travel between shafts A3 and A4.
The bearing in shaft A4 contains a foreign object, most likely a small aluminum chip that
produces a rhythmic noise while in use and is likely adversely affecting performance.
Appropriate solutions for all the above mentioned problems are discussed in the section
3.4 Future Work.
3.3 Conclusions
A 3-DOF wearable arm cobot was designed for use as a master controller in
telerobotic operations. A detailed design of the wearable arm cobot is complete at this
time. Its wearable design allows for an increased telepresence in terms of a 1-to-1
correlation in human arm movement. The wearable cobot’s cable driven transmissions
allow joint torques to be suitably reduced for use with CVTs. This design is also the first
to incorporate a wearable mechanism with spherical CVTs.
The shoulder transmission has been built, assembled and tested. Difficulties in
assembly mainly dealt with convenience of disassembly to ease the wrapping of cable.
An alternate mounting method between the capstans and shafts would help to better
accommodate. Further research is warranted to determine an appropriate start torque for
this system. Since the start torque is 16.5 lb-in, the application of 0.688 lbs (at 24 inches
from the shoulder) to start movement will noticeably affect the sensitivity of the cobot. It
should be noted that at the time this paper was written the robot had been assembled for 2
weeks and a break in period of a few thousand cycles should be considered for the
bearings and cables. The deflection due to cable compliance at 24 inches from the
shoulder joint is 0.076 inches. This translates to a mechanical rotational stiffness of
1.58*105 lb-in/rad, approximately 1/3 more stiff than theorized. This can be improved
upon by amending the shaft bowing issues discussed in the previous section.
37
3.4 Future Work
Future work includes completing fabrication and testing of the remaining wearable
cobot. A minor redesign will be implemented to increase the rigidity of the system by
adding a bearing to further support shafts A3 and A4. The bearing on shaft A4 will be
replaced. The large capstan on shaft A3 will be remade to the precise specifications.
Also, lightening of components and the addition of bearings at shafts A3 and A4 may
reduce the start torque. Safety stops will also be designed and placed to protect the user
and to prevent the device from exceeding its limits.
The current design will employ the use of a 6-axis force transducer but alternative
intent sensing measures at the CVT level must be investigated. The development of
adjustable cobot arm length must also be investigated to accommodate a wide variety of
users.
A new mounting method using ShaftlocTM
sleeves should be considered to allow
greater flexibility in capstan mounting. The new all-metal-contact CVT will be
implemented with the new powered cobot control algorithm.
38
APPENDIX A
Sava Cable Catalog
Select Pages from Sava Cable Catalog Pertaining to Design and Cable Selection
39
40
41
42
43
44
45
46
47
48
49
50
51
APPENDIX B
MathCAD Calcations
Stiffness Calculations for Individual Transmissions
52
53
54
55
56
57
58
APPENDIX C
Engineering Drawings
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
REFERENCES
[1] Peshkin, M. A., Colgate, J. E., Wannasuphoprasit, W., Moore, C. A., Gillespie, B.
and Akella, P., Cobot Architecture, IEEE Trans. Robot. Automat., vol. 17, pp. 377–
390, Aug. 2001 (13)
[2] http://lims.mech.northwestern.edu/projects/utla/
[3] Wannasuphoprasit, W., Akella, P., Peshkin, M.A., Colgate, J.E. “Cobots: a novel
material handling technology,” ASME Vol. 98-WA/MH-2, 1998.
[4] Moore, Carl A. “Design, Construction, and Control of a 3-Revolute Arm Cobot,”
Ph.D. dissertation, Dept. Mech. Eng., Northwest Univ., IL, 1999
[5] Schloerb, David W. A Quantitative Measure of Telepresence. Presence:
Teleoperators and Virtual Environments. Vol. 4, no. 1, pp. 64-80. Winter 1995.
[6] Rosenberg, Louis B. Virtual Fixtures: Perceptual Tools for Telerobotic
Manipulation. IEEE Conference Proceedings, 1993, pp. 76-82.
[7] Colgate, J. E. and Brown, J. M., 1994, “Factors Affecting the Z-Width of a Haptic
Display,” Proc. IEEE International Conf. On Robotics and Automation, pp. 3205-
3210.
[8] Colgate, J. E. and Schenkel, G. C., 1997. “Passivity of a Class of Sampled-Data
Systems: Application to Haptic Interface,” Journal of Robotic Systems, 14(1), pp.
37-47.
[9] Faulring, E. L., Colgate, J. E. and Peshkin, M. A. A High Performance 6-DOF
Haptic Cobot. Proc. IEEE ICRA, pp. 1980-1985, 2004.
[10] Heckendorn, F. and Kress, R., “Outline for Large-Scale System Operations and
D&D Report,” U.S. Dept. of Energy, WSRC-TR-2000-00364.
[11] Peters, R. A., Campbell, C., Bluethmann, W. J. and Huber, E. Robonaut Task
Learning through Teleoperation. Proc. IEEE ICRA, pp. 2806-2811, 2003
[12] http://robonaut.jsc.nasa.gov/status/Jul_Robonaut_Status_03.htm
106
[13] Salisbury, K., Townsend, W., Eberman, B. and DiPietro, D. Preliminary design
of a whole-arm manipulation system (WAMS). Robotics and Automation, 1988.
Proceedings 1988 IEEE International Conference, April 1988 Page(s):254 - 260 vol.1
[14] Townsend, William T., “The Effect of Transmission Design on the Force-
Controlled Manipulator Performance,” Ph.D. dissertation, Artificial Intelligence
Laboratory, M.I.T., MA, 1988.
107
BIOGRAPHICAL SKETCH
Jason Yap Chua
Jason Yap Chua was born January 24, 1980 in Jacksonville, Florida. After completing
the International Baccalaureate program at the Stanton College Preparatory School he
attended Florida State University with a Bright Futures Academic Scholarship to obtain
his Bachelor’s Degree in Mechanical Engineering. During his undergraduate studies
Jason worked as a machinist in the FAMU-FSU College Department of Mechanical
Engineering Machine Shop. After his graduation in 2003, he pursued graduate studies at
Florida State University. His research interests are primarily in robotics, wearable
robotics, controls and mechanisms.