The Use of Virtual Fixtures as Perceptual Overlays to Enhance Operator Performance in Remote Environments

Abstract

This report introduces the notion of virtual fixtures for use in telepresence systems. Tools and fixtures in the real world (e.g., a ruler guiding a pencil) enhance human performance by guiding manual operations, providing localizing references, reducing mental workload, and increasing precision. Virtual fixtures are computer-generated percepts overlaid on top of the reflection of a remote workspace which can provide similar benefits. Because such perceptual overlays are virtual constructions they can be diverse in modality, abstract in form, and custom tailored to individual situations. This study investigates the potential of virtual fixtures by implementing simple combinations of haptic surfaces and auditory sensations as virtual perceptual aids in a standardized telemanipulation task. Subjects viewed the remote environment through a vision system while wearing an upper-body exoskeleton. Eight subjects controlled a slave robot arm to perform standard Fitts' law peg-insertion tasks with and without the aid of a variety of virtual fixtures. Fixtures composed of haptic and auditory perceptual overlays increased operator performance up to 70%. Simple fixtures devised from basic elements can be powerful perceptual aids; a workstation environment might be developed to allow a teleoperator to design and implement virtual fixtures interactively. (Keywords: Telepresence, Telerobotics, Virtual Reality, Augmented Reality, Mixed Reality, AR, VR, MR, HCI )

Full text

9
&*-*
A
R
M
S
T
R
O
N
G
L
A
B
O
R
A
T
O
R
Y
AL/CF-TR-1994-0089
THE
USE
OF
VIRTUAL
FIXTURES
AS
PERCEPTUAL
OVERLAYS
TO
ENHANCE
OPERATOR
PERFORMANCE
IN
REMOTE
ENVIRONMENTS
Louis
B.
Rosenberg
CREW
SYSTEMS
DIRECTORATE
BIODYNAMICS
AND
BIOCOMMUNICATIONS
DIVISION
WRIGHT-PATTERSON
AFB
OH
45433-7901
■
™
—EMBER
1992
19950322
149
INTERIM
REPORT
FOR
THE
PERIOD
JUNE
1992
TO
JULY
1992
Approved
for
public
release;
distribution
is
unlimited.
|
AIR
FORCE
MATERIAL
COMMAND
WRIGHT-PATTERSON
AIR
FORCE
BASE,
OHIO
45433-7901

NOTICE
When
US
Government
drawings,
specifications,
or
other
data
are
used
for
any
purpose
other
than
a
definitely
related
Government
procurement
operation,
the
Government
thereby
incurs
no
responsibility
nor
any
obligation
whatsoever,
and
the
fact
that
the
Government
may
have
formulated,
furnished,
or
in
any
way
supplied
the
said
drawings,
specifications,
or
other
data,
is
not
to
be
regarded
by
implication
or
otherwise,
as
in
any
manner,
licensing
the
holder
or
any
other
person
or
corporation,
or
conveying
any
rights
or
permission
to
manufacture,
use
or
sell
any
patented
invention
that
may
in
any
way
be
related
thereto.
Please
do
not
request
copies
of
this
report
from
the
Armstrong
Laboratory.
Additional
copies
may
be
purchased
from:
National
Technical
Information
Service
5285
Port
Royal
Road
Springfield
VA
22161
Federal
Government
agencies
and
their
contractors
registered
with
Defense
Technical
Information
Center
should
direct
requests
for
copies
of
this
report
to:
Defense
Technical
Information
Center
Cameron
Station
Alexandria
VA
22314
TECHNICAL
REVIEW AND
APPROVAL
AL/CF-TR-1994-0089
The
voluntary
informed
consent
of
the
subjects
used
in
this
research
was
obtained
as
required
by
Air
Force
Regulation
169-3.
This
report
has
been
reviewed
by
the
Office
of
Public
Affairs
(PA)
and
is
releasable
to
the
National
Technical
Information
Service
(NTIS).
At
NTIS,
it
will
be
available
to
the
general
public,
including
foreign
nations.
This
technical
report
has
been
reviewed
and
is
approved
for
publication.
FOR
THE
COMMANDER
THOMAS
J.
MOORE,
Chief
Biodynamics
and
Biocommunications
Division
Crew
Systems
Directorate
Armstrong
Laboratory

REPORT
DOCUMENTATION
PAGE
Form
Approved
OMB
No.
0704-0188
o
w
„
mr)l
.„
hlir
j.„
f
nr
this
collection
of
information
is
estimated
to
average
l
hour
per
response,
including
the
time
for
reviewing
instructions,
searching
existing
data
sources,
n^her
nSLintaminq
the^
data
n^edeS
and
completing
and
reviewmg
the
collection
of
information.
Send
comments
regarding
this
burden
estimate
or
any
other
aspe«of
this
?ol
Mton
oHnformat
oincluding
jugged
ons
for
reducing
this
burden.
?o
Washington
Headquarters
Services
Directorate
for
Information
iOperations
and
Report
,1215
Jefferson
Davlf^hway
Su
™
204
ArMngton,
VA
22202-4302,
and
to
the
Office
of
Management
and
Budget,
Paperwork
Reduction
Project
(0704-0188),
Washington,
DC
20503.
1.
AGENCY
USE
ONLY
(Leave
blank)
2.
REPORT
DATE
September
1992
3.
REPORT
TYPE
AND
DATES
COVERED
Interim
-
June
1992
-
July
1992
4.
TITLE
AND
SUBTITLE
The
Use
of
Virtual
Fixtures
as
Perceptual
Overlays
to
Enhance
Operator
Performance
in
Remote
Environments
6.
AUTHOR(S)
Louis
B.
Rosenberg
7.
PERFORMING
ORGANIZATION
NAME(S)
AND
ADDRESS(ES)
Center
for
Design
Research
Stan
ford
University
Stanford
CA
94305
9.
SPONSORING/MONITORING
AGENCY
NAME(S)
AND
ADDRESS(ES)
Armstrong
Laboratory,
Crew
Systems
Directorate
Biodynamics
and
Biocommunications
Division
Human
Systems
Center
Air
Force
Materiel
Command
Wright-Patterson
AFB
OH
45433-7901
5.
FUNDING
NUMBERS
PE
-
61101F
PR
-
ILIR
TA
-
CB
WU
-
31
8.
PERFORMING
ORGANIZATION
REPORT
NUMBER
10.
SPONSORING/MONITORING
AGENCY
REPORT
NUMBER
AL/CF-TR-1994-0089
11.
SUPPLEMENTARY
NOTES
12a.
DISTRIBUTION/AVAILABILITY
STATEMENT
Approved
for
public
release;
distribution
is
unlimited.
12b.
DISTRIBUTION
CODE
13.
ABSTRACT
(Maximum200words)
This
report
introduces
the
notion
of
virtual
fixtures
for
use
in
telepresence
systems.
Tools
and
fixtures
in
the
real
world
(e.g.,
a
ruler
guiding
a
pencil)
enhance
human
performance
by
guiding
manual
operations,
providing
localizing
references,
reducing
mental
workload,
and
increasing
precision.
Virtual
fixtures
are
computer-generated
percepts
overlaid
on
top
of
the
reflection
of
a
remote
workspace
which
can
provide
similar
benefits.
Because
such
perceptual
over-
lays
are
virtual
constructions
they
can
be
diverse
in
modality,
abstract
in
form,
and
custom
tailored
to
individual
situations.
This
study
investigates
the
potential
of
virtual
fixtures
by
implementing
simple
combinations
of
haptic
surfaces
and
auditory
sensations
as
virtual
perceptual
aids
in
a
standardized
telemanipulation
task.
Subjects
viewed
the
remote
environment
through
a
vision
system
while
wearing
an
upper-body
exoskeleton.
Eight
subjects
controlled
a
slave
robot
arm
to
perform
standard
Fitts'
law
peg-insertion
tasks
with
and
without
the
aid
of
a
variety
of
virtual
fixtures.
Fixtures
composed
of
haptic
and
auditory
perceptual
overlays
increased
operator
performance
up
to
70%.
Simple
fixtures
devised
from
basic
ele-
ments
can
be
powerful
perceptual
aids;
a
workstation
environment
might
be
developed
to
allow
a
teleoperator
to
design
and
implement
virtual
fixtures
interactively
14.
SUBJECT
TERMS
telepresence
force
feedback
force-reflection
virtual
reality
exoskeleton
17.
SECURITY
CLASSIFICATION
OF
REPORT
UNCLASSIFIED
18.
SECURITY
CLASSIFICATION
OF
THIS
PAGE
UNCLASSIFIED
19.
SECURITY
CLASSIFICATION
OF
ABSTRACT
UNCLASSIFIED
15.
NUMBER
OF
PAGES
50
16.
PRICE
CODE
20.
LIMITATION
OF
ABSTRACT
UL
NSN
7540-01-280-5500
Standard
Form
298
(Rev.
2-89)
Prescribed
by
ANSI
Std.
Z39-18
298-102

THIS
PAGE
INTENTIONALLY
LEFT
BLANK
11

PREFACE
This
work
was
performed
with
facilities
and
support
from
the
Armstrong
Laboratory
with
headquarters
at
Brooks
Air
Force
Base,
Texas,
and
its
Crew
Systems
Directorate,
Human
Sensory
Feedback
Group,
at
Wright-
Patterson
AFB,
Ohio.
Support
for
this
work
was
provided
in
part
through
the
Air
Force
Office
of
Scientific
Research
Laboratory
Graduate
Fellowship
Program
(LGFP).
The
work
is
part
of
a
graduate
thesis
done
in
conjunction
with
the
Center
for
Design
Research,
Stanford
University,
Stanford,
California.
Accesion
For
NTIS
CRA&l
DTIC
TAB
Unannounced
Justification
_
By
Distribution
/
D
D
Availability
Codes
Avail
and/or
Special
m

THIS
PAGE
INTENTIONALLY
LEFT
BLANK
"IV

TABLE
OF
CONTENTS
Page
INTRODUCTION
1
Telepresence
and
the
Transparent
Interface
1
The
Concept
of
Perceptual
Overlays
2
The
Virtual
Fixture
Metaphor
3
Perceptual
Work-Station
Environment
5
Fitts'
Law
Performance
Test
6
Possible
Performance
Advantages
of
Virtual
Fixtures
9
EXPERIMENTAL
HARDWARE
1:L
Virtual
Fixture
Performance
Testbed
11
Task
board
1:L
MBA
exoskeleton
master/Merlin
robot
arm
slave
12
Monocular
vision
system
14
Virtual
fixture
board
14
Overall
Test
Setup
16
EXPERIMENT
DESIGN
&
PROCEDURE
18
Experiment
Design
18
Experimental
Protocol
22
RESULTS
25
DISCUSSION
31
Analysis
of
Each
Test
Fixture
31
Virtual
Fixture
1
31
Virtual
Fixture
2
33
Virtual
Fixture
3
33
Virtual
Fixture
4
34
Virtual
Fixture
5
35

Virtual
Fixture
6
36
Virtual
Fixture
7
36
Virtual
Fixture
8
37
CONCLUSIONS
39
ACKNOWLEDGEMENT
40
REFERENCES
41
VI

LIST
OF
FIGURES
Figure
Pa
9
e
1
Measurements
Used
to
Define
Task
Difficulty
7
2
Photograph
of
AF/NAVY
Teleoperator
Task
Board
11
3
Photograph
of
Subject
Wearing
MBA
Exoskeleton
12
4
Schematic
of
Task
Board
13
5
Complete
Telepresence
System
Test
Setup
15
6
Virtual
Fixtures
Tested
19
7
Sample
Trajectory
of
Peg
20
8
Performance
Curves
for
Each
Virtual
Fixture
Tested
...
28
Vll

Table
9
LIST
OF
TABLES
Page
1
Index
of
Difficulties
(I.D.)
of
Tasks
Tested
22
2
Subject
Testing
Schedule
24
3
Mean
Movement
Times
for
All
Subjects
for
Fixtures
1,
2,
3,
4,
and
6
25
4
Mean
Movement
Times
for
All
Subjects
for
Fixture
5
...
26
5
Mean
Movement
Times
for
All
Subjects
for
Fixtures
7
and
8
26
6
Index
of
Performance
(Ip)
for
All
Subjects
for
Fixtures
1,
2,
3,
4,
and
6
26
7
Index
of
Performance
(Ip)
for
All
Subjects
for
Fixture
5
27
8
Index
of
Performance
(Ip)
for
All
Subjects
for
Fixtures
1,
6,
7,
and
8
27
Mean
Coefficient
of
Variation
for
All
Subjects
27
10
Percent
Decrease
in
Movement
Time
for
Each
Fixture
...
29
11
Percentage
Decrease
in
Movement
Times
for
Fixtures
7
and
8
29
12
Percent
Increase
in
Index
of
Performance
(Ip)
for
Each
Fixture
29
13
Percentage
Increases
in
Index
of
Performance
(Ip)
for
Fixtures
7
and
8
30
vm

INTRODUCTION
Telepresence
and
the
Transparent
Interface
The
fundamental
purpose
of
a
telepresence
system
is
to
extend
an
operator's
sensory-motor
facilities
and
problem
solving
abilities
to
a
remote
environment
[15].
Telepresence
is
achieved
by
projecting
the
operator's
manipulatory
dexterity
to
a
remote
environment
while
reflecting
sensory
feedback
so
realistically
that
the
operator
feels
present
in
the
remote
site
[1].
To
enhance
operator
performance
and
understanding
within
remote
environments,
most
research
and
development
of
telepresence
systems
has
been
directed
towards
increasing
the
transparency
of
the
link
between
operator
and
environment.
Much
of
this
work
has
focused
upon
improving
the
fidelity
and
presentation
of
the
reflected
sensory
information.
Other
work
has
focused
upon
making
the
projection
of
the
operator's
kinematics
and
dynamics
more
naturally
mapped
into
the
remote
environment
[4,
5,
6,
9,
10].
The
culmination
of
such
research
efforts
could
be
a
transparent
link
between
human
and
machine;
a
user
interface
through
which
information
is
passed
so
naturally
between
operator
and
environment
that
the
user
achieves
a
complete
sense
of
presence
within
the
remote
site.
Although
unattainable
by
current
technology,
the
linking
of
an
operator
to
a
remote
environment
through
a
perfectly
transparent
interface
should
yield
a
human-machine
system
that
allows
an
operator
to
perform
just
as
well
in
the
remote
environment
as
in
a
real
environment.
Although
the
achievement
of
such
an
interface
would
be
an
important
milestone
in
human-
machine
systems,
the
motivation
for
the
work
described
in
this
paper
is
somewhat
different.
Rather
than
looking
for
methods
of
enhancing
the
transparency
of
the
interface,
this
work
is
actually
focused
upon
corrupting
the
transparency
of
the
interface
by
introducing
abstract
perceptual
information
between
human
and
machine.
It
is
believed
that
abstract
percepts
overlaid
on
top
of
the
reflection
of
the
workspace
can
be
implemented
as
perceptual
tools
that
enhance
an
operator's
performance
and
understanding.
It
is
not
suggested
that
such
an
approach
is
a
replacement
for
a
high
fidelity
link
between
operator
and
environment,
but
rather
a
supplement.
The
culmination
of
work
along
these
lines
could
be
a
telepresence
system
which
does
more
than
simply
project
an
operator's

abilities
into
a
workspace,
but
actually
enhances
and
expands
the
operator's
abilities
within
that
workspace
beyond
natural
capabilities.
The
Concept
of
Perceptual
Overlays
When
asked
to
draw
a
straight
line
in
the
real
world,
human
performance
can
be
greatly
enhanced
by
using
a
simple
tool
such
as
a
ruler.
The
use
of
a
ruler
reduces
the
amount
of
mental
processing
required
to
perform
the
task,
speeds
an
operator's
line
drawing
ability,
and
most
of
all
allows
an
operator
to
draw
a
significantly
better
line
than
if
no
ruler
had
been
used
at
all.
Without
a
ruler,
line
drawing
is
a
manual
task
that
requires
constant
visual
supervision
and
hand/eye
coordination.
With
a
ruler,
line
drawing
is
not
only
faster
and
straighter,
but
the
dependance
on
visual
feedback
is
reduced,
freeing
up
that
modality
for
other
uses.
What
is
more,
a
ruler
is
often
used
as
a
barrier
to
protect
against
dangerous
or
destructive
failures,
protecting
the
work-piece
from
the
slip
of
a
pencil
or
knife.
Such
guidance
and
protection
allows
the
operator
to
ease
mental
criteria
for
task
success
and
failure,
reducing
the
level
of
concentration
devoted
to
the
task.
Although
a
simple
tool
by
any
standard,
a
common
ruler
is
clearly
a
powerful
performance
aid
in
manual
line
drawing
tasks.
Although
the
use
of
a
ruler
to
assist
in
straight-line
drawing
is
an
effective
means
of
enhancing
human
performance,
can
such
a
process
be
generalized
beyond
line
drawing?
Ruler-use
can
be
thought
of
as
nothing
more
than
a
process
of
overlaying
abstract
sensory
information
on
top
of
a
workspace.
Thus,
a
ruler
can
be
generalized
as
a
particular
"perceptual
overlay"
designed
to
enhance
line
drawing
performance.
In
the
case
of
a
ruler,
the
overlaid
sensory
information
represents
a
rigid
surface
that
is
perceived
haptically
and
visually
by
the
user.
By
overlaying
this
generic
piece
of
sensory
information
on
top
of
the
workspace,
the
user
has
reduced
the
mental
and
physical
demands
of
the
straight
line
drawing
task
and
performance
is
greatly
enhanced.
If
a
simple
ruler-like
perceptual
overlay
can
so
greatly
enhance
the
performance
of
real
world
manipulatory
tasks
like
straight
line
drawing,
it
seems
that
computer
generated
perceptual
overlays
could
be
developed
within

virtual
environments
to
enhance
the
performance
of
tele-manipulation
tasks
within
remote
worksites.
Just
as
a
ruler
can
be
overlaid
on
top
of
a
real
workspace,
such
virtual
perceptual
overlays
could
be
overlaid
on
top
of
the
sensory
feedback
from
a
remote
workspace.
The
Virtual
Fixture
Metaphor
Because
the
abstract
notion
of
overlaid
sensory
information
is
as
difficult
to
conceptualize
as
it
is
to
talk
about,
I
will
introduce
a
virtual
fixture
metaphor
as
a
means
of
describing
such
computer
generated
sensations
as
concrete
physical
structures.
It
must
be
stressed
that
the
point
of
this
metaphor
is
intended
to
facilitate
the
understanding
of,
and
interaction
with
perceptual
overlays
and
should
not
be
taken
so
literally
as
to
limit
the
scope
of
the
perceptual
overlay
concept.
Virtual
fixtures
are
thus
defined
as
abstract
sensory
information
overlaid
on
top
of
reflected
sensory
feedback
from
a
remote
environment.
Although
overlaid
on
top
of
the
user's
perception
of
the
remote
environment,
virtual
fixtures
are
completely
independent
of
all
information
from
the
remote
site
and
are
thus
immune
from
communication
delays
and
bandwidth
limitations.
Like
the
ruler
guiding
the
pencil,
virtual
fixtures
overlaid
on
top
of
a
remote
workspace
could
act
to
reduce
mental
processing
required
to
perform
the
task,
reduce
the
work
load
of
certain
sensory
modalities,
and
most
of
all
allow
precision
and
performance
to
exceed
natural
human
abilities.
Although
virtual
fixtures
could
be
functionally
equivalent
to
fixtures
in
the
real
world,
there
are
many
advantages
inherent
to
virtual
fixtures
because
they
are
computer
simulations
rather
than
real
physical
hardware.
When
overlaid
on
a
workspace,
the
fixtures
only
interact
with
the
user
and
not
with
the
workspace
itself.
Thus
fixtures
can
occupy
the
same
physical
space
as
objects
in
the
workspace.
This
means
that
the
workspace
geometry
imposes
no
constraints
upon
the
placement
or
configuration
of
virtual
fixtures.
What
is
more,
such
fixturing
has
no
mass,
has
no
physical
or
mechanical
constraints,
requires
no
machining
time
or
maintenance,
can
be
easily
prototyped
and
modified,
and
can
essentially
be
transported
to
remote
locations
using
nothing
more
than
standard
communication
links.

If
we
explore
the
concept
of
virtual
fixtures
using
the
simple
ruler
example
as
the
starting
point,
the
first
elements
to
consider
might
be
rigid
planar
surfaces.
Such
fixtures
would
be
composed
of
haptic
sensations
generated
by
reflecting
simulated
forces
to
the
operator
through
a
force-
reflecting
master.
As
the
operator
interacts
with
the
modeled
surfaces,
the
reaction
forces
would
be
computed
and
reflected
appropriately.
Of
course,
such
fixtures
are
by
no
means
limited
to
rigid
surfaces.
Abstracting
the
fixturing
concept,
we
might
consider
modeling
compliant
surfaces,
damped
surfaces,
even
viscous
or
coulomb
frictional
contacts.
In
fact,
the
simulation
environment
offers
such
freedom
that
fixtures
could
even
be
developed
as
attractive
or
repulsive
fields.
Although
fixtures
composed
of
haptic
sensations
offer
endless
possibilities,
the
fixturing
concept
is
not
limited
to
that
modality.
Abstract
fixtures
could
be
composed
of
visual,
auditory,
even
tactile
sensations
used
alone
or
in
cross-modal
combinations.
For
example,
if
a
haptic
fixture
composed
of
rigid
planar
surfaces
or
attractive
force
fields
was
developed
to
aid
a
particular
task,
an
audio,
visual,
or
vibratory
signal
could
be
mapped
to
various
locations
along
the
fixture
to
enhance
interaction.
Such
additional
modalities
could
be
used
to
indicate
deviations
from
a
trajectory,
proximity
to
a
danger
zone,
even
provide
feedback
of
velocities
or
accelerations.
Abstracting
the
fixturing
concept
further,
we
could
imagine
virtual
fixtures
imbued
with
particular
visual
qualities
to
enhance
interaction
with
environments.
For
example,
virtual
fixtures
composed
of
haptic
surfaces
could
be
modeled
with
optical
properties
to
suit
the
task
at
hand.
The
fixture
might
be
invisible
to
the
user
if
the
operator
gains
no
benefit
from
visual
cues,
it
could
be
made
to
look
like
a
solid
object
if
rich
visual
cues
are
useful
for
the
task,
it
could
even
be
made
to
look
like
a
transparent
glassy
solid
if
visual
cues
are
important
but
the
user
wants
to
avoid
obscuring
the
workspace.
Fixtures
might
even
be
designed
as
visual
filters
to
block
particular
distractions,
enhance
contrast,
provide
depth
cues,
even
magnify
a
part
of
the
workspace.
One
can
even
imagine
the
benefit
of
a
fixture
composed
of
compliant
surfaces
which
changes
color
or
brightness
with
compression.
Previous
work
with
targeting
cues
[3],
predictive
displays
[2,
14,
20,
21,
22],
and
perspective
overlays
[11,
25]
has
demonstrated
that
overlaid
visual
cues
can
enhance
performance
and
understanding
within
a
teleoperation
environment.

If
the
description
of
virtual
fixtures
thus
far
seems
too
abstract,
a
simple
example
may
drive
the
concept
home.
Imagine
a
situation
where
a
teleoperating
surgeon
performs
a
delicate
procedure
on
a
patient.
Although
such
uses
of
telerobotics
are
still
in
the
research
phase,
it
is
an
application
that
demands
a
high
degree
of
human
performance
within
a
remote
workspace.
Now
imagine
that
a
virtual
fixture
is
being
used
by
the
doctor
to
enhance
his
abilities
in
this
procedure.
The
fixture
might
appear
to
the
doctor
like
a
flat
plane
of
glass
with
a
grooved
guide
for
the
scalpel.
The
glass-like
virtual
fixture
might
actually
pass
directly
through
a
patient's
body,
preventing
the
scalpel
from
penetrating
below
a
particular
depth
but
not
obscuring
vision
of
the
tissue
below.
By
sliding
the
scalpel
along
the
edge
of
a
groove
in
the
fixture,
the
surgeon
could
make
a
perfect
incision.
The
slightest
deviation
from
the
target
trajectory
might
be
reported
by
aa
audio
or
tactile
signal.
The
power
and
flexibility
of
such
a
system
would
be
unmatched
by
actual
physical
tooling.
Besides
the
fact
that
such
a
fixture
in
the
real
world
could
not
pass
directly
through
a
patient's
body,
it
could
not
be
put
in
place
at
the
touch
of
a
button,
removed
at
the
touch
of
another
button,
or
easily
altered
as
conditions
change.
What
is
more,
virtual
fixturing
does
not
have
to
be
fabricated,
sterilized,
monitored,
or
maintained.
Perceptual
Workstation
Environment
Although
the
notion
of
virtual
fixturing
does
seem
promising,
if
the
development
of
fixtures
requires
complex
computation
or
intimate
knowledge
of
the
workspace
to
be
effective,
robot
autonomy
might
be
a
preferred
solution
to
the
task
at
hand.
If,
on
the
other
hand,
effective
fixtures
could
be
developed
out
of
basic
building
blocks
and
quickly
implemented
by
a
teleoperator
in
an
interactive
environment,
fixtures
could
be
used
in
unstructured
or
changing
environments
unsuitable
for
autonomous
systems.
Thus,
it
is
proposed
that
a
workstation
type
environment
be
developed
to
allow
teleoperators
to
design
and
implement
assistive
fixtures
when
confronted
with
an
unknown
task
in
an
unstructured
environment
upon
first
encounter.
With
such
a
workstation
in
mind,
the
study
described
in
this
paper
investigates
the
use
of
simple
haptic
surfaces
and
auditory
fields
as
perceptual
aids
in
a
simple
peg
insertion
task.
Although
a
workstation
environment
could
implement
more
sophisticated
surfaces
or
fields
including
a
more
diverse
array
of
sensory
modalities,
it
was

thought
that
if
simple
combinations
of
forces
and
sounds
could
be
made
into
effective
fixtures,
the
potential
of
virtual
fixtures
would
be
adequately
displayed.
Fitts'
Law
Performance
Test
To
quantify
teleoperator
performance
in
a
remote
manual
task,
a
Fitts'
Law
paradigm
was
chosen
because
of
its
general
acceptance
as
a
robust
measure
of
human
performance
[7].
Although
extensive
use
of
Fitts'
Law
has
been
documented
in
human
performance
literature,
little
work
has
been
done
to
extend
the
paradigm
to
a
telepresence
environment.
McGovern
(1975)
used
a
Fitts'
Law
task
to
demonstrate
the
merit
of
a
closed
loop
master-slave
system
as
compared
to
an
open
loop
system
[13].
Hill
(1979)
used
a
Fitts'
task
to
demonstrate
performance
differences
as
a
function
of
force
feedback
from
the
manipulative
system
[8].
Pepper
(1988)
was
the
first
to
use
the
Fitts*
task
in
a
true
telepresence
scenario.
While
previous
work
had
employed
Fitts'
Law
with
the
teleoperator
in
direct
view
of
the
workspace,
this
work
used
Fitts'
Law
to
compare
a
variety
of
viewing
conditions
which
included
remote
visual
links
between
operator
and
workspace
[16].
These
studies
have
shown
that
a
Fitts'
Law
paradigm
is
appropriate
for
analysis
of
perceptual-motor
performance
within
teleoperated
and
true
telepresence
systems.
Fitts
(1954)
established
a
means
of
quantifying
human
performance
in
terms
of
information
processing
capacity
of
the
neuromotor
system.
He
developed
a
relationship
between
the
speed
and
the
accuracy
of
human
motor
performance
and
demonstrated
that
the
speed
of
a
task
of
requiring
a
particular
accuracy
is
bounded
by
the
capacity
of
the
neuromuscular
system
control
movements.
Fitts
argued
that
if
manual
control
was
limited
by
the
information
processing
rate
of
the
peripheral
and
central
nervous
system,
movement
times
would
be
limited
by
the
information
processing
demands
of
the
task.
This
concept
is
apparent
if
we
think
of
a
task
requiring
great
accuracy
such
as
threading
a
needle.
Why
do
we
perform
a
threading
task
with
infuriating
sluggishness?
Such
a
task
is
limited
by
human
information
processing
capacity
(i.e.,
how
fast
we
can
perceive
the
environment,
actuate
our
limbs,
and
adjust
for
error).
A
task
such
as
threading
a
needle
requires

many
fine
adjustments
and
is
thus
limited
by
how
quickly
perceptions
and
adjustments
can
be
performed
(i.e.,
limited
by
human
bandwidth).
To
measure
human
information
processing
capacity
as
an
indication
of
performance
within
a
telepresence
environment,
subjects
were
tested
on
a
peg
insertion
task.
Like
threading
a
needle,
the
difficulty
of
peg
insertion
is
a
function
of
tolerance,
the
difference
between
the
peg
diameter
and
the
hole
diameter.
The
tighter
the
fit
of
the
peg
in
the
hole,
the
more
accuracy
required
of
the
operator,
and
thus
the
slower
the
maximum
performance
speed.
By
requiring
subjects
to
perform
a
standardized
peg
insertion
task
as
fast
as
they
can,
completion
times
could
be
measured
and
compared
to
task
difficulty
to
yield
information
processing
capacity
for
that
task.
By
doing
such
an
analysis
upon
subjects
with
and
without
the
implementation
of
various
virtual
fixtures,
variations
in
task
completion
times
indicate
performance
changes
resulting
from
fixture
use.
Thus,
by
using
Fitts'
Law
analysis
of
a
standardized
peg
insertion
task,
the
effectiveness
of
various
virtual
fixtures
as
perceptual
aids
could
be
quantified.
W
=
D-d
W
ID
=
-log
2
"2Ä
Figure
1:
Measurements
used
to
define
task
difficulty

Through
an
extensive
review
of
human
psychomotor,
perceptual,
and
cognitive
test
batteries,
the
Naval
Oceans
Systems
Center
developed
a
peg
insertion
performance
task
specifically
representative
of
teleoperator
manipulative
activities
[16,
24].
The
test
battery
requires
subjects
to
move
pegs
of
various
diameters
between
holes
of
varied
spacing.
Movement
times
for
peg
motions
are
recorded
and
correlated
with
task
difficulty.
As
defined
by
Fitts
[7],
the
binary
Index
of
Difficulty,
ID,
for
the
one
dimensional
peg
transfer
task
can
be
computed
as:
W
ID
=
-log2
ry
»
(bits/response)
where
A
is
the
amplitude
of
the
motion
and
W
is
the
peg
tolerance
defined
as
the
difference
between
the
hole
diameter
and
the
peg
diameter.
These
quantities
are
shown
in
Figure
1.
Fitts'
Law
[7]
relates
task
completion
time
to
task
ID
by
defining
the
movement
time,
mt,
as
follows:
mt
=
ki
ID
+
k
2
(sec)
where
kj
and
^
are
characteristic
constants
of
the
individual
operator
and
represent
the
slope
and
intercept
of
the
Fitts'
Law
curve.
The
reciprocal
of
the
slope
of
the
Fitts'
Law
curve
(1/kj)
has
units
(bits/sec)
(which
is
identical
to
units
of
capacity
for
an
information
channel)
and
is
thus
an
accepted
measure
of
human
information
processing
capacity
[17].
Because
the
slope
is
a
coarse
measure
and
is
susceptible
to
distortions
resulting
from
changes
in
strategy
between
tasks
of
different
difficulty,
a
more
robust
measure
of
processing
capacity
was
also
computed,
called
the
Binary
Index
of
Performance.
This
value,
abbreviated
as
Ip,
describes
the
information
processing
capacity
required
of
the
operator
to
perform
a
task
of
a
particular
difficulty
[7].
Ip
also
has
units
(bits/sec)
and
is
defined
as
follows:
-I-
i
W
P
=
"mt
g2
2A
(bits/sec)

(or)
ID
Ip
=
^
(bits/sec)
r
mt
where
mt
is
the
movement
time
required
to
complete
the
task.
Index
of
Performance
Ip
is
an
accepted
measure
of
the
information
processing
capacity
of
the
teleoperator
[19].
Possible
Performance
Advantages
of
Virtual
Fixtures
Fitts
[7]
demonstrated
that
performance
in
manual
tasks
is
limited
by
the
information
processing
capacity
of
the
central
and
peripheral
nervous
systems.
Virtual
fixtures
could
enhance
performance
by
either
reducing
the
information
processing
demands
of
a
given
task
or
by
increasing
the
information
processing
capacity
of
the
operator.
Since
the
fixtures
are
perceptual
overlays
and
do
not
exist
in
the
workspace,
it
would
be
incorrect
to
suggest
that
fixtures
could
in
any
way
modify
the
given
task.
If
fixtures
cannot
alter
the
task,
how
can
they
act
to
reduce
the
processing
demands
of
the
task?
Although
the
task
itself
remains
the
same,
virtual
fixtures
can
modify
how
the
workspace
is
perceived,
alter
how
the
task
is
conceptualized,
and
can
thus
be
designed
to
reduce
the
processing
demands
of
the
task.
For
example,
by
implementing
haptic
fixtures
as
physical
guides
or
barriers,
the
operator
might
have
fewer
alternative
movements
to
consider
in
the
perception
of
the
workspace
than
really
exist
in
the
actual
task
workspace.
Virtual
guides
or
barriers
simplify
the
perception
of
the
task
workspace,
reduce
the
information
processing
demands
on
the
nervous
system,
and
thus
increase
operator
performance.
Thompson
(1977)
demonstrated
a
similar
effect
by
showing
that
completion
times
in
part
mating
tasks
could
be
reduced
by
eliminating
degrees
of
constraint
[26].
Virtual
fixtures
do
not
eliminate
constraints
required
by
a
task,
but
they
can
assist
the
operator
in
achieving
some
of
the
required
constraints.
It
is
also
hypothesized
that
the
additional
percepts
offered
by
virtual
fixtures
provide
localizing
references
to
the
remote
workspace
(i.e.,
interaction
with
the
fixtures
gives
the
operator
a
better
sense
of
the
physical

relationship
between
kinesthetic
and
proprioceptive
output
and
workspace
geometry).
Such
localization
enhances
the
illusion
of
presence,
provides
the
operator
with
a
better
understanding
of
the
workspace,
and
reduces
the
operator's
reliance
on
sensory
feedback.
This
localizing
effect
should
act
to
reduce
the
overall
information
processing
demands
of
the
task
and
ultimately
enhance
operator
performance.
Finally,
virtual
fixtures
could
also
enhance
performance
by
providing
alternative
information
pathways
by
introducing
feedback
from
additional
sensory
modalities.
Since
the
nervous
system
is
parallel
in
architecture,
the
introduction
of
alternative
pathways
could
act
to
increase
the
information
processing
capacity
of
the
operator.
To
test
these
hypotheses,
the
following
experiment
measured
operator
performance
with
and
without
the
use
of
simple
virtual
fixtures.
The
fixtures
tested
were
chosen
to
offer
insight
into
all
three
modes
of
operation
presented
above:
Simplification
of
the
perception
of
the
task,
enhanced
localization
to
remote
workspace,
and
increased
capacity
due
to
alternative
modes
of
sensory
feedback.
10

EXPERIMENTAL
HARDWARE
Virtual
Fixture
Performance
Testbed
To
implement
a
standardized
Fitts'
Law
performance
test
in
a
telepresence
environment,
the
following
hardware
subsystems
were
used:
Peg
Insertion
Task
Board,
MBA
Exoskeletal
Master/Merlin
Robot
Arm
Slave,
Monocular
Vision
System,
and
Virtual
Fixture
Board.
Figure
2:
Photograph
of
the
AF/Navy
Teleoperator
Performance
Evaluation
Battery
Task
Board
used
for
implementing
peg
insertion
Fitts'
Law
task.
Also
shown
is
the
Merlin
Robot
Arm
used
as
the
slave
in
the
teleoperated
system.
Task
Board:
(AF/Navy
Teleoperator
Performance
Evaluation
Battery)
As
shown
in
Figure
2,
a
specialized
task
board
was
used
to
implement
the
performance
task.
This
task
board
was
originally
developed
by
the
Naval
Oceans
Systems
Center
to
implement
a
controlled
peg-in-hole
test
battery
for
telemanipulation
systems
[16,
24].
Holes
on
the
board
are
arranged
to
accommodate
a
wide
range
of
peg
sizes
and
movement
amplitudes.
This
study
made
use
of
four
holes
(diameter
2.00
cm)
spaced
to
allow
horizontal
peg
movements
of
4.00
cm
and
16.00
cm.
Three
pegs
(diameters
0.75
cm,
0.98
cm,
1
1

1.50
cm)
were
used
in
the
task
to
vary
the
peg
insertion
tolerance.
The
bottom
of
each
hole
contains
a
microswitch
connected
to
a
PC
via
a
Digital
I/O
controller
card.
This
PC
controls
and
monitors
the
peg
insertion
portion
of
the
experiment.
The
computer
is
equipped
with
a
real-time
clock
for
measuring
peg
movement
times
to
the
nearest
millisecond.
The
timer
is
started
when
the
peg
is
extracted
from
the
start
hole
and
stopped
when
inserted
into
the
target
hole.
MBA
Exoskeletal
Master/Merlin
Robot
Arm
Slave
As
shown
in
Figure
3,
the
MBA
Exoskeleton
Master
is
a
dual-arm,
full
upper
body
exoskeleton
which
can
transduce
motion
in
seven
degrees
of
freedom
for
each
arm
[12].
In
its
current
configuration,
the
device
is
used
only
as
a
sensor;
no
force
information
is
reflected
back
to
the
user
through
the
exoskeleton.
Thus,
peg
insertions
were
performed
based
only
on
visual
and
auditory
feedback
from
the
workspace.
This
reduced
sensory
feedback
environment
offered
a
good
testbed
for
the
prowess
of
virtual
fixtures.
Figure
3:
Photograph
of
subject
wearing
MBA
kinesthetic
exoskeleton
used
as
the
master
in
the
teleoperator
system.
12

In
this
experiment,
the
position
of
the
subject's
right
hand
was
tracked
and
fed
to
the
slave
robot
arm.
An
end-effector
was
constructed
for
the
slave
robot
which
housed
the
peg
and
allowed
for
easy
interchange
of
the
various
diameter
pegs.
Because
of
the
difficulty
of
the
teleoperated
peg
insertion
task
without
force
feedback,
the
slave
robot
was
programmed
to
maintain
its
end-
effector
orientation
perpendicular
to
the
board
at
all
times.
This
arrangement
simplified
the
task
and
eliminated
the
possibility
of
subjects
torquing
the
peg
in
the
hole.
To
further
simplify
the
task,
the
robot
was
constrained
to
move
only
in
a
horizontal
plane
level
with
the
target
holes.
Thus,
the
task
only
required
two-dimensional
positioning
of
the
peg
in
the
xy
plane
with
x
defined
into
the
board,
and
y
defined
along
the
horizontal
row
of
target
holes
as
shown
in
Figure
4.
It
should
be
noted
that
although
the
task
was
constrained
to
planar
positioningit
was
sufficiently
difficult
to
require
subjects
to
practice
for
two
45-minute
sessions
before
training
was
complete.
The
task
was
designed
to
be
simple
in
concept
but
difficult
enough
that
the
subjects
would
not
approach
the
upper
bound
of
the
slave
robot's
ability
to
perform
the
task.
Figure
4:
Schematic
of
Task
Board
with
xy
plane
of
allowable
peg
motion.
To
protect
the
task
board
and
the
robot,
the
peg
was
rigidly
constrained
in
all
degrees
of
freedom
except
along
the
x
axis
pointing
into
the
task
board.
Along
that
axis,
the
peg
floated
on
a
stiff
spring
so
that
strong
forward
impacts
of
the
peg
with
the
board
would
not
jar
the
board
or
damage
the
robot.
13

Monocular
Vision
System
A
previous
study
using
an
identical
task
board
performed
by
the
Naval
Oceans
Systems
Center
has
shown
that
stereo
vision
feedback
offers
no
performance
advantage
over
monocular
projections
used
in
this
peg
insertion
task
[16,
24].
Thus,
a
simple
monocular
vision
system
was
implemented
to
provide
visual
feedback
for
this
experiment.
The
system
was
designed
as
an
inexpensive
means
of
creating
the
illusion
of
operator
presence
while
providing
visual
feedback
of
the
workspace.
The
system
used
7X
power
binocular
optics
focused
upon
a
distant
color
video
monitor
which
displayed
the
output
of
a
single
camera
in
the
workspace
focused
on
the
task
board.
The
monitor
was
placed
at
a
distance
from
the
optics
in
such
a
way
that
the
magnification
of
the
video
image
created
the
illusion
that
the
task
board
was
within
reach
of
the
operator's
hands.
Robot
end-effector
motion
was
scaled
to
match
operator
hand
motion
so
that
the
apparent
end-effector
position
corresponded
to
the
user's
kinesthetic
sense
of
arm
position.
Such
a
correlation
between
kinesthetic
feedback
and
visual
feedback
of
robot
end-
effector
position
greatly
enhanced
the
user's
sense
of
presence
within
the
workspace.
Testing
of
the
system
revealed
best
results
when
the
angle
of
the
camera
incident
on
the
workspace
was
closely
matched
by
the
angle
that
the
magnifying
optics
were
incident
upon
the
distant
video
monitor.
When
these
angles
were
not
similar,
conflicting
perspective
cues
hindered
the
illusion
of
presence.
Figure
5
includes
a
rough
schematic
of
the
vision
system
as
part
of
the
overall
system
hardware.
Virtual
Fixture
Board
Rather
than
using
a
force
reflecting
exoskeleton
to
model
the
rigid
impedance
surfaces
which
compose
the
virtual
fixtures,
it
was
thought
that
the
preliminary
tests
of
virtual
fixtures
should
not
be
influenced
by
hardware
limitations
of
force
reflecting
devices.
Thus,
a
Fixture
Board
was
designed
which
allowed
real
rigid
surfaces
to
be
fabricated
from
acrylic
sheets.
The
MBA
exoskeleton
would
interact
with
the
real
acrylic
surfaces
and
reflect
that
information
to
the
user.
As
far
as
the
user
is
concerned,
the
haptic
perception
of
these
surfaces
was
coming
from
the
exoskeleton
and
was
just
as
much
"virtual"
information
as
if
it
was
truly
computer
generated.
The
benefit
of
this
14

approach
is
that
the
reflected
perception
of
these
surfaces
was
perfectly
modeled.
The
surfaces
felt
crisp
and
real,
free
from
the
bandwidth
limitations
of
most
force
reflecting
devices.
The
drawback
of
this
approach
was
that
more
abstract
fixtures
were
impossible
to
generate
using
this
method.
The
virtual
fixtures
were
constructed
out
of
acrylic
sheets
and
positioned
upon
a
wooden
platform
in
front
of
the
user.
To
make
the
fixtures
quickly
interchangeable,
they
were
constrained
and
positioned
on
the
wooden
platform
by
locating
pins.
The
exoskeleton
was
fitted
with
a
teflon
cap
at
the
end
of
the
hand
grip
which
was
used
as
the
contact
surface
between
the
exoskeleton
and
the
fixture.
The
fixtures
were
treated
regularly
with
an
oil-
based
lubricant
so
that
little
friction
was
perceivable
between
the
teflon
and
the
acrylic
surfaces.
Although
frictional
surfaces
could
make
for
effective
virtual
fixtures,
the
intent
of
this
study
was
to
look
at
simple
surfaces
modeled
only
as
rigid
impedances
free
from
any
viscous
or
coulombic
damping.
Remote
Environment:
TASK
BOARD
"CAMERA
ROBOT
ARM
Operator
Space
VISION
SYSTEM
Figure
5:
Complete
Telepresence
System
developed
to
implement
**"?
testing
of
teleoperator
performance
in
a
standardized
peg
insertion
task
with
and
without
the
aid
of
virtual
fixtures.
15

Although
the
haptic
fixtures
were
modeled
physically
rather
than
computationally,
auditory
fixtures
were
tested
that
were
pure
computer
simulations.
Simple
compliant
surfaces
were
modeled
in
which
the
compression
of
the
surface
was
proportional
to
a
linear
change
in
pitch.
Tones
were
generated
on
a
PC
and
fed
to
the
user
via
stereo
headphones.
These
trials
served
as
a
testbed
for
alternate
modality
fixtures
as
well
as
a
proof
of
concept
of
purely
computer-generated
fixtures.
Overall
Test
Setup
Having
described
each
of
the
major
components
of
this
test
setup,
it
is
important
to
clearly
describe
the
system
as
a
whole
before
getting
into
the
details
of
the
subject
testing.
As
shown
in
Figure
5,
the
system
is
divided
into
two
physically
separate
parts:
the
remote
environment
and
the
operator
space.
The
remote
environment
contains
the
task
board,
the
merlin
robot
arm,
and
a
single
video
camera
pointed
at
task
board.
The
camera
is
positioned
so
that
the
incident
perspective
is
similar
to
what
a
human
operator
would
see
if
standing
directly
in
front
of
the
board
and
performing
the
peg
insertions
in
person.
The
operator
space
contains
the
exoskeleton,
the
vision
system,
and
the
virtual
fixture
board.
Once
inside
the
exoskeleton
and
vision
system,
the
subject
is
presented
with
a
projection
of
the
image
from
the
camera
in
the
remote
environment.
The
subject
is
given
the
illusion
that
the
task
board
is
situated
directly
before
him,
within
reaching
distance
of
the
exoskeleton.
In
reality,
the
task
board
is
on
the
opposite
side
of
the
laboratory,
behind
the
subject
and
completely
out
of
view.
The
fixture
table
is
placed
directly
in
front
of
the
subject
in
such
a
way
that
it
cannot
be
seen
when
looking
through
the
vision
system,
but
feels
as
though
it
occupies
the
same
space
as
the
apparent
image
of
the
task
board.
Thus,
virtual
fixtures
implemented
on
the
fixture
board
feel
as
though
they
are
overlaid
on
top
of
their
perception
of
the
remote
environment.
The
remote
robot
arm
is
slaved
to
the
right
hand
position
of
the
exoskeleton.
Thus,
when
the
subject
moves
his
right
hand
so
as
to
interact
with
the
image
of
the
task
board,
the
end-effector
on
the
slave
arm
follows.
Because
the
subject
cannot
see
his
own
hand
when
looking
into
the
vision
16

system
but
does
see
the
remote
robot
end-effector
in
the
position
where
he
feels
his
hand
to
be,
a
sense
of
presence
within
the
remote
environment
is
created.
The
subjects
also
wear
a
set
of
stereo
headphones
for
use
in
implementing
auditory
perceptual
overlays.
Auditory
surfaces
and
fields
can
be
modeled
on
the
control
computer
and
interacted
with
by
the
user.
The
complete
system
provides
a
powerful
testbed
for
projecting
subjects
into
a
remote
environment
and
overlaying
haptic
and
auditory
information
on
top
of
the
reflected
percepts.
17

EXPERIMENT
DESIGN
&
PROCEDURE
Experiment
Design
The
goal
of
this
study
was
to
investigate
the
effect
that
the
presentation
of
virtual
fixtures
has
upon
performance
in
a
standard
telemanipulation
task.
Eight
simple
fixtures
were
developed
for
comparison
to
a
no-fixture
control
case.
Six
of
the
test
fixtures
were
purely
haptic
sensations
while
two
fixtures
introduced
both
haptic
and
auditory
sensory
information.
It
should
be
noted
that
the
design
of
these
test
fixtures
was
not
motivated
by
finding
the
most
effective
perceptual
aid
for
this
particular
peg
insertion
task.
This
particular
set
of
test
fixtures
was
chosen
to
provide
general
insight
into
the
design
and
use
of
virtual
fixtures
and
to
allow
the
evaluation
of
some
of
the
basic
building
blocks
from
which
fixtures
can
be
made.
When
designing
the
test
fixtures,
the
peg
insertion
task
was
not
thought
of
as
a
single
motion,
but
rather
as
a
combination
of
two
phases:
ballistic
motion
and
fine
positioning.
The
ballistic
phase
was
defined
as
the
time
from
which
the
peg
is
removed
from
the
start
hole
until
it
makes
first
contact
with
the
target
hole.
The
fine
positioning
phase
was
defined
to
begin
at
the
end
of
the
ballistic
motion
and
continue
until
the
peg
is
properly
positioned
in
the
hole
[17].
This
decomposition
of
the
peg
insertion
task
was
found
to
be
a
useful
conceptual
guide
for
fixture
design.
Some
of
the
test
fixtures
were
targeted
at
aiding
ballistic
motion,
some
were
targeted
at
enhancing
fine
motion,
and
some
were
intended
to
help
both.
Figure
6
shows
a
schematic
representation
of
each
of
the
eight
fixtures
tested.
These
fixtures
are
shown
overlaid
on
top
of
the
task
board
as
they
are
perceived
by
the
subjects.
Although
these
fixtures
are
represented
graphically
in
this
figure,
they
are
perceived
only
as
forces
or
sounds
by
the
subjects.
Although
instilling
visual
qualities
to
fixtures
is
a
viable
application
of
perceptual
overlays
and
an
important
topic
for
future
investigation
into
virtual
fixtures,
this
study
was
restricted
only
to
haptic
and
auditory
sensory
modalities.
18

Figure
6:
Virtual
fixtures
shown
projected
on
top
of
the
task
board
as
perceived
by
the
subjects.
Also
shown
are
top
views,
looking
down
from
above
the
task
board.
The
fixtures
are
shown
graphically
here,
but
they
are
perceived
haptically
or
auditorily
by
subjects.
19

As
shown
in
Figure
6,
the
eight
fixtures
tested
are
composed
of
simple
combinations
of
planar
surfaces.
Fixture
1
is
simply
a
rigid
horizontal
surface
oriented
like
a
table
top
in
the
workspace
and
positioned
so
that
contact
with
the
surface
will
result
in
vertical
alignment
of
the
operator's
hand
with
the
holes
in
the
peg
board.
Fixture
2
is
just
like
Fixture
1,
but
it
includes
a
second
surface
which
is
parallel
to
the
plane
of
the
task
board
and
located
three
inches
back
from
the
board.
When
using
this
fixture,
subjects
operate
in
the
space
between
the
vertical
plane
and
the
task
board.
This
arrangement
provides
some
limits
upon
the
operator's
ballistic
motion.
Fixture
3
is
similar
to
Fixture
2,
except
the
vertical
plane
is
angled
so
as
to
converge
with
the
task
board
as
the
target
hole
is
approached.
This
arrangement
acts
like
a
funnel,
confining
and
guiding
ballistic
motion
towards
the
target
hole.
Figure
7:
Sample
trajectory
of
peg
from
start
hole
to
target
hole
as
guided
by
a
virtual
fixture.
As
shown,
the
fixture
does
not
define
the
motion
of
the
operator's
hand,
but
rather
defines
the
boundaries
of
the
motion.
Fixture
4
is
identical
to
Fixture
3,
except
a
third
plane
is
added
directly
in
front
of
the
target
hole.
This
fixture
influences
the
ballistic
phase
even
further,
stopping
hand
motion
when
close
to
the
target
hole.
Fixture
6
is
similar
to
20

Fixture
3,
except
a
second
funnel-like
fixture
is
added
in
front
of
the
target
hole.
Thus,
ballistic
motion
is
guided
as
was
done
by
Fixture
3,
and
fine
motion
is
guided
by
the
additional
surfaces.
Figure
7
shows
a
sample
trajectory
superimposed
upon
Fixture
4,
making
the
fixture
implementation
clear.
This
drawing
demonstrates
how
the
use
of
a
fixture
might
influence
the
trajectory
of
the
peg.
As
shown
in
the
figure,
the
rigid
surfaces
do
not
define
the
motion
of
the
peg,
but
rather
influence
the
operator's
trajectory
by
confining
the
boundaries
of
peg
motion.
Fixture
5
is
very
different
from
those
presented
thus
far
in
both
its
geometry
and
implementation.
The
key
is
that
his
fixture
was
not
interacted
with
by
the
subject's
right
hand
(the
hand
that
performed
the
peg
insertion
task),
but
rather
was
designed
for
interaction
only
with
the
unused
left
hand.
Fixture
5
is
a
rigid
impedance
plane
parallel
to
the
task
board
and
located
approximately
0.5"
in
front
of
the
board.
The
subject
would
place
his
left
hand
upon
the
planar
surface
while
performing
the
task
with
the
right
hand.
The
intent
of
this
fixture
was
to
isolate
the
effect
of
localization
on
performance.
It
was
hypothesized
that
all
fixtures
provided
some
localizing
information
to
the
users
which
enhanced
their
understanding
of
the
geometry
of
the
workspace
and
allowed
them
to
better
correlate
their
kinesthetic
sense
of
hand
position
to
the
remote
site.
Because
interaction
with
other
fixtures
would
guide
or
limit
the
operator's
motion,
it
was
impossible
to
isolate
performance
increases
due
to
the
localizing
effect
alone.
Thus,
Fixture
5
was
designed
to
interact
only
with
the
unused
hand
and
therefore
only
affects
performance
by
providing
localizing
cues.
Fixture
7
and
Fixture
8
are
the
only
fixtures
which
implement
both
haptic
and
auditory
information.
Fixture
7
is
identical
to
Fixture
1,
but
also
introduces
a
texture-like
field
of
auditory
information.
The
information
is
represented
as
a
series
of
surfaces
perpendicular
to
the
task
board
as
shown
in
black
in
Figure
6.
Operator
interaction
with
these
surfaces
results
in
the
production
of
an
audible
tone.
The
pitch
of
the
tone
increases
from
left
to
right
across
the
task
board.
Fixture
8
is
identical
to
Fixture
6,
but
it
introduces
a
single
compliant
surface
in
front
of
the
target
hole
as
shown
in
black
in
Figure
6.
This
surface
is
modeled
as
a
compliant
surface
with
a
proportional
21

stiffness
such
that
interaction
with
this
surface
produces
a
tone
proportional
to
compression.
Experimental
Protocol
A
series
of
tests
was
run
to
evaluate
subject
performance
using
each
fixture
configuration.
Operator
performance
was
recorded
during
test
periods
which
included
12
practice
and
36
timed
peg
insertion
trials
for
each
fixture
studied.
A
single
trial
consisted
of
moving
a
peg
from
a
designated
start
hole
to
a
designated
target
hole.
The
holes
were
referenced
by
numerals
(3,
4,
5
or
6)
located
above
each
hole
as
shown
in
Figure
4.
Two
different
peg
motions
were
studied
in
these
tests:
a
16
cm
motion
from
hole
3
to
hole
6
and
a
4
cm
motion
from
hole
5
to
hole
6.
The
36
trial
period
was
divided
into
three
groups
of
12
trials.
Each
of
these
groups
required
the
subject
to
perform
the
insertion
task
using
a
different
peg
size.
After
the
completion
of
each
group
of
12
trials,
subjects
were
instructed
to
rest
for
approximately
two
minutes
while
a
new
peg
size
was
installed
in
the
robot
end-effector.
The
use
of
two
motion
amplitudes
(4
cm
and
16
cm)
and
three
peg
sizes
(0.75
cm,
0.98
cm,
1.50
cm
diameter
pegs)
allowed
for
the
testing
of
insertion
trials
with
six
different
task
difficulties.
Table
1
shows
all
combinations
of
peg
size
and
motion
amplitude
and
lists
the
Index
of
Difficulty
for
each
task
as
dictated
by
Fitts'
Law
[7].
The
order
in
which
the
three
peg
sizes
were
presented
to
each
subject
was
randomized
to
ensure
that
mental
and
physical
fatigue
had
similar
effects
on
all
task
difficulties.
Table
1:
Index
of
Difficulties
(ID)
shown
for
all
combinations
of
peg
size
and
peg
motion
amplitudes
as
predicted
by
Fitts'
Law.
Motion
i
Amplitude
Diameter
5
to
6
(4
cm)
3
to
6
(16cm)
0.75
cm
2.68
bits/response
4.68
bits/response
0.98
cm
2.98
bits/response
4.98
bits/response
1.50
cm
4.00
bits/response
6.00
bits/response
22

Subjects
were
instructed
to
begin
each
test
period
with
a
3
to
6
peg
insertion
trial.
Subjects
would
then
perform
a
5
to
6
peg
insertion
trial,
then
a
3
to
6
peg
insertion
trial,
and
cycle
in
that
manner
throughout
the
36
trials.
Subjects
were
allowed
to
proceed
through
the
test
period
at
their
own
pace.
To
automate
the
testing
procedure
and
allow
subjects
to
proceed
through
the
test
with
little
operator
intervention,
a
number
of
simple
beeps
were
used
to
guide
subject
activity.
When
a
subject
inserted
a
peg
into
the
correct
starting
hole,
the
task
board
control
computer
would
emit
an
audible
tone
to
signal
that
the
peg
was
properly
positioned.
The
subject
would
then
keep
the
peg
in
the
start
hole
for
two
seconds
until
the
control
computer
produced
a
second
tone
which
signaled
that
the
task
could
now
be
performed
at
will.
The
reason
for
this
two
second
waiting
period
was
to
ensure
that
the
subject
maintained
a
steady
contact
with
the
microswitch
at
the
base
of
the
start
hole.
The
subjects
were
not
required
to
perform
the
insertion
task
as
soon
as
the
second
tone
was
heard;
this
second
tone
simply
meant
"you
are
free
to
go
whenever
ready."
Thus,
the
subjects
could
mentally
prepare
themselves
for
the
particular
insertion
task
and
begin
at
will.
When
ready,
the
subject
would
remove
the
peg
from
the
start
hole
and
insert
it
into
the
target
hole
as
fast
as
possible.
Upon
insertion
into
the
target
hole,
the
control
computer
would
emit
a
tone
to
signal
that
the
trial
had
been
successfully
completed.
The
subject
was
then
free
to
proceed
to
the
next
insertion
trial
at
will.
At
the
end
of
each
trial
the
control
computer
recorded
the
movement
time
along
with
the
peg
size
and
motion
amplitude.
After
the
completion
of
each
group
of
12
trials,
the
control
computer
would
emit
a
long-duration
tone
which
signaled
the
subject
to
rest
for
two
minutes
while
the
operator
exchanged
peg
sizes.
Post
testing
interviews
revealed
that
all
subjects
were
comfortable
with
the
use
of
audible
tones
to
automate
the
testing
procedure.
Subjects
were
tested
over
9
experimental
sessions,
each
lasting
45
to
60
minutes.
To
minimize
the
effects
of
fatigue
and
boredom,
no
subject
completed
more
than
one
experimental
session
during
a
single
day
of
testing.
Each
of
the
first
two
experimental
sessions
included
two
test
periods
of
36
trials.
The
first
period
of
each
session
was
performed
with
no
fixture
and
the
second
was
performed
with
Fixture
1.
These
initial
144
trials
were
treated
only
as
practice
during
which
the
subjects
familiarized
themselves
with
the
use
of
the
exoskeleton,
merlin
robot
arm,
vision
system,
and
fixture
table.
It
was
found
23

that
by
the
end
of
the
second
practice
session,
all
subjects
had
sufficiently
learned
the
task
that
variability
in
movement
times
for
trials
of
the
same
difficulty
had
fallen
below
20%
for
every
subject,
with
a
mean
variability
of
14%
for
all
subjects.
Once
learning
had
stabilized,
subjects
were
sequentially
tested
using
each
of
the
test
fixtures.
One
or
two
new
fixtures
were
tested
during
each
daily
session
in
addition
to
a
baseline
fixture
which
was
tested
during
every
session.
Fixture
1
was
the
baseline
fixture
and
was
used
to
track
performance
increases
resulting
from
day-to-day
learning
so
that
any
such
effects
could
be
compensated.
Table
2
lists
all
of
the
daily
sessions
along
with
the
fixtures
tested
during
that
session.
Table
2:
Testing
Schedule:
All
seven
tests
shown
along
with
corresponding
daily
sessions
and
fixtures
tested
on
that
day.
TEST
SESSION
FIXTURES
STUDIED
Practice
dav
1
No
Fixture
/
Fixture
1
Practice
day
2
No
Fixture
/
Fixture
1
Test
I
day
3
No
Fixture
/
Fixture
1
Test
II
day
4
Fixture
1
/
Fixture
2
/
Fixture
3
Test
III
day
5
Fixture
1
/
Fixture
4
Test
IV
day
6
Fixture
0
/
Fixture
5
Test
V
day
7
Fixture
1
/
Fixture
6
Test
VI
day
8
Fixture
1
/
Fixture
8
Test
VII
day
9
Fixture
1
/
Fixture
7
24

RESULTS
The
following
section
presents
all
results
of
fixture
performance
comparisons
in
tabular
form.
The
basic
quantities
presented
are
the
recorded
movement
times
and
the
coefficients
of
variation
of
movement
times
for
trials
of
the
same
difficulty.
The
coefficient
of
variation
is
a
measure
of
subject
consistency
at
a
particular
task
and
is
an
indication
of
the
repeatability
of
the
results.
Also
computed
is
the
slope
of
a
linear
regression
line
relating
movement
time
to
index
of
difficulty.
The
inverse
of
the
slope
has
units
(bits/sec)
and
is
a
measure
of
the
operator
information
processing
capacity.
Because
the
slope
is
susceptible
to
distortions
resulting
from
changes
in
strategy
between
tasks
of
different
difficulty,
a
more
robust
measure
of
processing
capacity
was
also
computed
called
the
Binary
Index
of
Performance.
This
value,
abbreviated
as
Ip,
describes
the
information
processing
capacity
required
of
the
operator
to
perform
a
task
[7].
Tables
3
through
5
list
the
mean
movement
time
results
for
each
fixture
tested.
Tables
6
through
8
list
the
computed
Index
of
Performance
(Ip)
for
each
fixture
tested.
Table
9
lists
the
coefficients
of
variation,
slopes,
and
the
inverse
of
the
slopes
for
each
fixture
tested.
Figure
8
graphically
compares
the
performance,
showing
the
mean
movement
time
versus
index
of
difficulty
curves
for
all
test
fixtures.
TABLE
3:
Mean
Movement
Times
for
each
Index
of
Difficulty
are
shown
for
8
subjects
using
no
fixture
as
well
as
using
Fixtures
1,
2,
3,
4
and
6.
Also
computed
is
the
mean
movement
time
across
all
trials
(across
task
difficulty).
2.68
IND
2.98
EX
OF
1
4.00
DIFFICU1
4.68
LTY
4.98
6.00
FIXTURE
mean
No
Fixture
1123ms
1280ms
1353ms
194
6ms
1866ms
2162ms
1622ms
Fixture
1
957ms
977ms
1118ms
1595ms
1656ms
1758ms
1344ms
Fixture
2
842ms
883ms
1030ms
1435ms
1527ms
1631ms
1225ms
Fixture
3
682ms
704ms
795ms
12
93ms
1350ms
1469ms
104
9ms
Fixture
4
692ms
713ms
815ms
137
6ms
1471ms
1543ms
1102ms
Fixture
6
667ms
722ms
835ms
1301ms
1385ms
14
95ms
1068ms
25

TABLE
4:
Mean
Movement
times
for
each
Index
of
Difficulty
are
shown
for
6
subjects
using
no
fixture
as
well
as
using
the
localizing
Fixture
5
with
the
unused
hand.
Also
computed
is
the
mean
movement
time
across
all
trials
(across
task
difficulty).
2.68
INDEX
OF
DIFFICULTY
6.00
FIXTURE
2.98
4.00
4.68
4.98
mean
No
Fixture
833ms
92
7ms
991ms
1617ms
1600ms
1666ms
1272ms
Fixture
5
656ms
662ms
807ms
1448ms
1396ms
1497ms
1078ms
TABLE
5:
Mean
Movement
Times
for
each
Index
of
Difficulty
are
shown
for
6
subjects
using
no
fixture,
using
purely
haptic
Fixtures
1
and
6,
as
well
as
using
auditory/haptic
Fixtures
7
and
8.
Also
shown
is
the
mean
movement
time
across
all
trials.
It
should
be
noted
that
Fixtures
1
and
7
are
identical
except
for
the
addition
of
auditory
information
in
7.
The
same
is
true
for
Fixture
6
and
Fixture
8.
2.68
IND
2.98
EX
OF
I
4.00
)IFFICU1
4.68
.TY
4.98
6.00
FIXTURE
mean
No
Fixture
1073ms
1177ms
1259ms
1845ms
1832ms
2058ms
1541ms
Fixture
1
840ms
910ms
969ms
1539ms
1569ms
1674ms
1250ms
Fixture
7
687ms
727ms
718ms
1310ms
1325ms
1369ms
1023ms
Fixture
6
641ms
691ms
811ms
1266ms
1361ms
1461ms
1039ms
Fixture
8
614ms
610ms
7
01ms
1136ms
1154ms
1270ms
915ms
TABLE
6:
Index
of
Performance,
Ip
=
(ID
/
mean
movement
time),
is
computed
and
shown
for
8
subjects
using
no
fixture
as
well
as
using
Fixtures
1,
2,
3,
4
and
6.
Ip
has
units
(bits/sec)
and
is
an
accepted
measure
of
human
information
processing
capacity.
2.68
INDEX
OF
DIFFICULTY
6.00
FIXTURE
2.98
4.00
4.68
4.98
mean
Ip
No
Fixture
2.34
2.33
2.96
2.40
2.67
2.77
2.58
Fixture
1
2.80
2.99
3.58
2.93
3.01
3.41
3.12
Fixture
2
3.18
3.37
3.88
3.26
3.26
3.68
3.44
Fixture
3
3.93
4.23
5.03
3.62
3.69
4.08
4.10
Fixture
4
3.87
4.18
4.91
3.40
3.39
3.89
3.94
Fixture
6
4.02
4.13
4.79
3.60
3.60
4.01
4.02
26

TABLE
7:
Index
of
Performance,
Ip
=
(ID
/
mean
movement
time),
is
computed
and
shown
for
6
subjects
using
no
fixture
as
well
as
using
the
localizing
Fixture
5
with
the
unused
hand.
Ip
has
units
(bits/sec)
and
is
an
accepted
measure
of
human
information
processing
capacity.
2.68
INDEX
OF
DIFFICULTY
6.00
FIXTURE
2.98
4.00
4.68
4.98
mean
Ip
No
Fixture
3.22
3.21
4.03
2.89
3.11
3.60
3.60
Fixture
5
4.09
4.50
4.96
3.23
3.58
4.01
4.01
TABLE
8:
Index
of
Performance,
Ip
=
(ID
/
mean
movement
time)
,
is
computed
and
shown
for
6
subjects
using
no
fixture,
using
purely
haptic
Fixtures
1
and
6,
as
well
as
using
auditory/haptic
Fixtures
7
and
8.
Ip
has
units
(bits/sec)
and
is
an
accepted
measure
of
human
information
processing
capacity.
It
should
be
noted
that
Fixtures
1
and
7
are
identical
except
for
the
addition
of
auditory
information
in
7.
The
same
is
true
for
Fixture
6
and
Fixture
8.
2.68
INDEX
OF
DIFFICULTY
6.00
FIXTURE
2.98
4.00
4.68
4.98
mean
Ip
No
Fixture
2.50
2.67
3.18
2.54
2.72
2.92
2.76
Fixture
1
3.19
3.27
4.12
3.04
3.17
3.58
3.40
Fixture
7
3.90
4
.01
5.57
3.57
3.76
4.38
4.20
Fixture
6
4.18
4.31
4.93
3.70
3.66
4.11
4.15
Fixture
8
4.36
4.88
5.71
4.12
4.32
4.72
4.69
Table
9:
Mean
Coefficient
of
Variation
for
all
subjects
performing
tasks
of
the
same
difficulty
with
each
fixture
is
shown.
Also
shown
is
the
slope
and
standard
error
of
a
linear
regression
analysis
of
movement
time
against
index
of
difficulty.
The
reciprocal
of
the
slope
is
also
shown
and
has
units
(bits/sec).
This
value
is
an
accepted
measure
of
information
processing
capacity.
FIXTURE
MEAN
c.o.v.
SLOPE
[ms/bit]
Std
Err
of
Coef
1/SLOPE
[bits
/
sec]
No
Fixture
13.7%
±2.5
317
43
3.15
Fixture
1
11.9%
+2.0
276
39
3.62
Fixture
2
12.8%
±3.2
267
37
3.75
Fixture
3
11.1%
±2.7
271
47
3.69
Fixture
4
8.6%
±1.9
300
56
3.33
Fixture
5
13.0%
±1.3
302
62
3.31
Fixture
6
9.6%
±2.8
280
44
3.57
Fixture
7
11.2%
+1.9
244
61
4.10
Fixture
8
9.2%
±1.5
228
40
4.39
27

Performance
Curves
for
Each
Virtual
Fixture
Tested
2000
-
E
CO
CD
E
c
CO
E
CD
>
o
1500
-
1000
-
'
No
Fixture
""""
™
^~
Fixture
1
~
"
~
~
~
Fixture
2
Fixture
3
..........
Fixture
4
Fixture
5
Fixture
6
"
■
Fixture
7
"
■
"
»""
■
Fixture
8
500
-
2.5
—I—
3.5
i
4.5
5.5
i
6.5
Index
of
Difficulty
Figure
8:
Performance
using
each
fixture
shown
for
comparison
as
best
line
fit
to
Fitts'
Law
relation
between
mt
and
ID
In
order
to
quantify
performance
increase
due
to
fixture
use
in
a
meaningful
way
that
is
resistant
to
variations
across
task
difficulty,
percentage
changes
in
movement
times
and
Index
of
Performance
were
computed
for
each
fixture
with
respect
to
the
no-fixture
case.
The
percentage
decrease
in
movement
times
is
a
unitless
measure
of
the
effectiveness
of
a
fixture
to
speed
operator
performance
in
a
task.
The
percentage
increase
in
Index
of
Performance
is
a
unitless
measure
of
the
increase
in
the
information
processing
capacity
of
the
operator
or
the
decrease
in
processing
requirements
of
the
task
resulting
from
fixture
use.
Tables
10
and
11
list
the
percentage
decrease
in
movement
times
corresponding
to
the
use
of
each
fixture.
Tables
12
and
13
list
the
percentage
increase
in
processing
capacity
Ip
associated
with
the
use
of
each
fixture.
28

TABLE
10:
Percent
Decrease
in
Movement
Time
for
each
fixture
with
respect
to
the
no-fixture
case.
Values
are
computed
for
each
index
of
difficulty.
Also
shown
is
the
mean
percentage
decrease
in
movement
time
for
each
fixture.
This
mean
value
represents
the
effectiveness
of
a
fixture
to
enhance
operator
performance.
'£
2.68
INDEX
OF
DIFFICULTY
2.98
4.00
4.68
4.98
6.00
FIXTUF
mean
Fixture
1
15.96%
26.85%
19.02%
19.82%
11.93%
20.61%
19.03%
±4.9
Fixture
2
28.60%
36.71%
27.11%
30.23%
19.98%
28.00%
28.44%
+5.4
Fixture
3
48.86%
58.06%
51.96%
40.32%
32.09%
38.17%
44.91%
+9.1
Fixture
4
47.49%
56.90%
49.63%
34.32%
23.67%
33.41%
40.90%
±11.9
Fixture
5
23.77%
33.35%
20.47%
11.03%
13.62%
10.69%
18.82%
±7.9
Fixture
6
50.95%
55.74%
47.35%
39.73%
2
9.5
9%
36.48%
43.31%
±9.2
Fixture
7
43.86%
47.27%
54.73%
33.91%
32.12%
40.21%
42.02%
±8.4
Fixture
8
54.42%
63.46%
56.94%
47.57%
45.41%
47.36%
52.52%
±6.5
TABLE
11:
Percentage
Decreases
in
Movement
Times
when
using
Fixtures
7
and
8
as
compared
to
Fixtures
1
and
6
respectively.
Since
Fixtures
7
and
8
are
identical
to
1
and
6
in
all
ways
except
for
the
addition
of
auditory
information,
the
percentage
decrease
in
movement
times
reflects
the
performance
advantage
associated
with
overlaying
auditory
information.
INDEX
OF
DIFFICULTY
FIXTURE
2.68
2.98
4.00
4.68
4.98
6.00
mean
Fixture
7
20.04%
22.36%
29.76%
16.08%
16.86%
20.05%
20.89%
±4.9
Fixture
8
4.30%
12.45%
14.55%
10.82%
16.4
6%
13.99%
12.10%
±4.2
TABLE
12:
Percent
Increase
in
Processing
Capacity
Ip
for
each
fixture
with
respect
to
the
no-fixture
case.
Values
are
computed
for
each
index
of
difficulty.
Also
computed
is
the
mean
percentage
increase
in
processing
capacity
associated
with
each
fixture.
RE
2.68
INE
2.98
)EX
OF
4/00
DIFFK
4.68
:ULTY
4.98
6.00
FIXTU
mean
Fixture
l
19.66%
28.33%
20.95%
22.08%
12.73%
23.10%
21.14%
±5.0
Fixture
2
35.90%
44.64%
31.08%
35.83%
22.10%
32.85%
33.73%
±
7.4
Fixture
3
67.95%
81.55%
69.93%
50.83%
38.20%
47.29%
59.29%
15.3
Fixture
4
65.38%
79.40%
65.88%
41.67%
26.97%
40.43%
53.29%
±18.9
Fixture
5
27.02%
40.19%
23.08%
11.76%
15.11%
11.39%
20.31%
+10.8
Fixture
6
71.79%
77.25%
61.82%
50.00%
34.83%
44.77%
56.74%
±15.3
Fixture
7
56.00%
50.19%
75.16%
40.55%
38.24%
50.00%
51.69%
±13.2
Fixture
8
74.40%
82.77%
79.56%
62.20%
58.82%
61.64%
69.90%
±
9.5
29

TABLE
13:
Percentage
Increases
in
Processing
Capacity
Ip
when
using
Fixtures
7
and
8
as
compared
to
Fixtures
1
and
6
respectively.
Since
Fixtures
7
and
8
are
identical
to
1
and
6
in
all
ways
except
for
the
addition
of
auditory
information,
the
percentage
increase
in
Ip
reflects
the
performance
advantage
associated
with
overlaying
auditory
information.
2.68
INDEX
OF
DIFFICULTY
6.00
FIXTURE
2.98
4.00
4.68
4.98
mean
Fixture
7
22.26%
22.63%
35.19%
17.43%
18.61%
22.35%
23.08%
±6.3
Fixture
8
4.31%
13.23%
15.82%
11.35%
18.03%
14.84%
12.93%
+4.7
30

DISCUSSION
Analysis
of
Each
Test
Fixture
The
following
section
sequentially
addresses
the
change
in
performance
associated
with
each
fixture
and
discusses
the
implications.
The
objective
of
this
analysis
is
to
gain
general
insight
into
the
fundamental
properties
of
virtual
fixtures
in
order
to
facilitate
the
design
and
implementation
of
an
interactive
perceptual
workstation.
Before
comparing
the
performance
results
for
task
completion
with
each
fixture
configuration,
the
issue
of
repeatability
and
reliability
of
the
results
should
be
addressed.
As
shown
in
Table
9,
the
coefficients
of
variation
for
task
completion
is
under
14%
for
all
fixture
configurations,
with
a
mean
variability
of
10%.
This
means
that
for
any
given
task
of
a
particular
difficulty
performed
by
any
given
subject,
the
variation
in
performance
for
all
such
trials
was
on
average
10%.
Such
consistency
for
human
performance
is
surprisingly
good
and
suggests
that
the
results
of
these
tests
are
highly
reliable.
Virtual
Fixture
1
As
shown
in
Figure
6,
virtual
Fixture
1
is
a
rigid
impedance
plane
oriented
perpendicularly
to
the
plane
of
the
task
board.
When
interacting
with
this
fixture,
the
subject's
hand
motion
is
restrained
to
move
only
in
the
plane
of
the
peg
holes.
Thus,
this
fixture
reduces
the
Cartesian
degrees
of
freedom
on
hand
motion
from
three
to
two.
An
interesting
point
about
this
fixture
is
that
the
end-effector
of
the
slave
robot
is
restricted
by
software
to
planar
two
degree
of
freedom
motion
regardless
of
the
operator's
commands.
Thus,
the
degree
of
freedom
which
is
eliminated
by
the
use
of
Fixture
1
has
no
effect
on
the
position
of
the
slave
robot
arm.
Since
this
degree
of
freedom
plays
no
part
in
performing
the
task,
it
would
seem
that
Fixture
1
would
have
little
effect
on
the
performance
of
the
operator.
As
shown
in
Tables
3
and
10,
the
mean
movement
times
for
peg
insertions
performed
with
the
aid
of
virtual
Fixture
1
was
19%
faster
than
with
the
no-fixture
case.
Comparing
Indexes
of
Performance
(Ip)
as
shown
in
31

Tables
6
and
12,
the
human
information
processing
capacity
when
using
virtual
Fixture
1
was
21%
greater
than
with
no
fixture
at
all.
This
comparison
suggests
that
a
virtual
fixture
can
enhance
performance
by
simplifying
the
perception
of
the
task
even
though
the
task
itself
is
unchanged.
In
both
cases,
the
peg
insertion
task
was
a
2
degree
of
freedom
operation.
Without
the
fixture,
the
subjects
were
free
to
move
in
all
three
Cartesian
directions
even
though
they
were
well
aware
that
the
slave
robot's
position
was
locked
in
the
y
direction.
With
the
fixture,
the
subjects
were
guided
to
move
only
in
the
plane.
This
implies
that
even
after
the
144
trials
of
practice,
the
subjects
were
unable
to
completely
ignore
the
irrelevant
degree
of
freedom
and
were
wasting
processing
capacity
on
it.
This
result
suggests,
first,
that
use
of
the
single
rigid
impedance
plane
as
a
virtual
fixture
was
effective
in
significantly
increasing
operator
performance.
Second,
this
result
suggests
that
when
developing
a
presence
system
which
interfaces
a
human
limb
to
a
robot
arm
with
fewer
degrees
of
freedom,
the
unused
degrees
of
freedom
may
degrade
performance
and
thus
should
be
restrained
whenever
possible.
Think
back
to
the
example
given
earlier
which
proposed
the
use
of
a
virtual
fixture
similar
to
Fixture
1
to
enhance
performance
in
a
telepresence
system
for
surgery.
A
single
rigid
impedance
plane
was
described
which
could
be
located
at
some
depth
below
the
tissue
of
a
patient
such
that
interaction
of
the
surgeon's
scalpel
with
this
rigid
surface
could
assure
that
the
incision
only
reached
a
particular
depth,
thus
protecting
vital
organs.
The
results
of
testing
with
Fixture
1
suggest
that
not
only
would
such
a
fixture
limit
motion
to
the
plane
and
thus
maintain
a
constant
incision
depth,
it
would
also
enhance
operator
ability
to
perform
in
free
plane
and
increase
information
processing
capacity.
Thus,
not
only
would
the
surgeon
be
safe
from
encountering
vital
organs
located
beneath
the
protective
surface,
the
surgeon
would
have
enhanced
dexterity
above
the
surface.
32

Virtual
Fixture
2
As
shown
in
Figure
6,
virtual
Fixture
2
is
the
same
as
Fixture
1
except
a
second
rigid
planar
impedance
surface
is
added
parallel
to
the
surface
of
the
task
board.
Like
Fixture
1,
Fixture
2
restricts
operator
hand
motion
to
the
horizontal
plane.
The
added
surface
also
restricts
how
far
back
from
the
board
the
operator
can
draw
the
peg.
As
shown
in
Tables
3
and
10,
the
mean
movement
times
for
peg
insertions
performed
with
the
aid
of
virtual
Fixture
2
were
28%
faster
than
with
the
no-fixture
case.
Comparing
Indexes
of
Performance
(Ip)
as
shown
in
Tables
6
and
12,
the
human
information
processing
capacity
when
using
virtual
Fixture
2
was
34%
greater
than
with
no
fixture
at
all.
This
comparison
clearly
shows
that
the
use
of
Fixture
2
significantly
enhanced
performance
over
the
no-fixture
case
and
that
use
of
the
additional
surface
in
Fixture
2
as
compared
to
Fixture
1
made
a
significant
improvement.
The
question
thus
remains,
why
should
the
additional
surface
in
Fixture
2
increase
performance?
It
is
believed
that
the
additional
surface
serves
a
number
of
purposes
by
restricting
how
far
back
the
peg
can
be
drawn
from
the
board:
First,
this
surface
guides
operator
ballistic
motion,
preventing
unnecessarily
wide
trajectories.
This
should
reduce
the
information
processing
required
of
the
task
because
the
distance
from
the
task
board
is
no
longer
a
parameter
the
operator
needs
to
be
constantly
concerned
with.
Secondly,
this
surface
acts
as
a
kinesthetic/proprioceptive
localizing
agent,
giving
the
user
a
better
sense
of
hand
position
in
the
remote
workspace
by
providing
a
haptic
indication
of
proximity
from
the
task
board.
Thirdly,
by
providing
depth
information
haptically,
the
fixture
reduces
the
demand
on
the
visual
system
to
gage
depth
and
frees
up
that
modality
for
other
uses,
like
tracking
the
target
hole.
Virtual
Fixture
3
As
shown
in
Figure
6,
virtual
Fixture
3
is
the
same
as
Fixture
2
except
the
second
rigid
planar
impedance
surface
is
not
parallel
to
the
surface
of
the
task
board,
but
rather
is
diagonal.
Like
Fixture
1,
Fixture
3
restricts
operator
hand
motion
to
the
plane.
Like
Fixture
2,
the
additional
surface
restricts
how
far
back
from
the
board
the
operator
can
draw
the
peg.
The
unique
thing
33

about
Fixture
3
is
that
it
does
not
uniformly
restrict
how
far
back
the
peg
can
be
drawn,
but
rather
converges
hand
motion
toward
the
board
as
the
target
hole
is
approached.
Although
this
fixture
is
composed
of
2
intersecting
planes,
it
can
be
thought
of
as
a
cone
projected
into
the
two
dimensional
workspace
which
funnels
ballistic
motion
towards
the
target
hole.
As
shown
in
Tables
3
and
10,
the
mean
movement
times
for
peg
insertions
performed
with
the
aid
of
virtual
Fixture
3
were
45%
faster
than
with
the
no-fixture
case.
Comparing
Indexes
of
Performance
(Ip)
as
shown
in
Tables
6
and
12,
the
human
information
processing
capacity
when
using
virtual
Fixture
3
was
59%
greater
than
with
no
fixture
at
all.
Like
Fixture
2,
this
fixture
provides
haptic
depth
cues
and
kinesthetic/proprioceptive
localization
to
the
remote
workspace.
The
primary
advantage
of
this
fixture
is
that
it
guides
ballistic
motion
more
efficiently,
funnelling
the
gross
motion
towards
the
target
hole.
It
is
believed
that
this
perceptual
aid
in
target
convergence
reduces
the
demand
upon
kinesthetic
and
visual
feedback
and
thus
reduces
the
information
processing
required
for
the
task.
Virtual
Fixture
4
As
shown
in
Figure
6,
virtual
Fixture
4
is
the
same
as
Fixture
3
except
for
the
addition
of
a
third
rigid
impedance
surface
which
crosses
the
diagonal
surface.
This
additional
surface
was
placed
such
that
the
operator
would
contact
this
surface
when
the
peg
was
directly
in
front
of
the
target
hole.
Whereas
Fixture
3
aided
ballistic
motion
by
converging
hand
position
towards
the
target
hole,
Fixture
4
was
intended
to
provide
further
trajectory
shaping
by
halting
hand
motion
when
the
peg
was
aligned
with
the
hole.
As
shown
in
Tables
3
and
10,
the
mean
movement
times
for
peg
insertions
performed
with
the
aid
of
virtual
Fixture
4
were
41%
faster
than
with
the
no-fixture
case.
Comparing
Indexes
of
Performance
(Ip)
as
shown
in
Tables
6
and
12,
the
human
information
processing
capacity
when
using
virtual
Fixture
4
was
53%
greater
than
with
no
fixture
at
all.
The
results
for
Fixture
4
are
not
significantly
different
from
those
for
the
Fixture
3
case,
showing
that
the
addition
of
the
third
surface
did
not
enhance
performance
in
these
trials.
It
is
thus
concluded
that
the
additional
surface
added
no
important
information
for
task
completion.
An
interesting
side
note
is
that
in
a
time
delay
study
currently
being
completed,
Fixture
4
was
found
to
provide
significant
34

improvement
over
Fixture
3
when
communication
delays
were
introduced
into
the
system.
This
makes
sense
because
additional
localizing
references
become
more
important
when
time
delays
distort
the
relation
between
kinesthetic
sense
of
position
and
visual
feedback
of
position.
These
results
will
be
presented
formally
in
a
companion
paper
addressing
time
delay
issues.
Virtual
Fixture
5
As
shown
in
Figure
6,
virtual
Fixture
5
is
composed
of
a
single
rigid
impedance
plane
which
is
parallel
to
the
task
board
surface
and
is
located
approximately
0.5"
in
front
of
the
board.
Whereas
all
fixtures
thus
far
described
were
designed
for
interaction
with
the
operator's
right
hand
(the
hand
that
manipulates
the
peg),
Fixture
5
is
designed
for
interaction
only
with
the
operator's
unused
hand.
Subjects
place
the
palm
of
their
left
hand
upon
the
surface
while
they
perform
the
task
with
their
right
hand.
The
purpose
of
testing
this
fixture
was
to
isolate
the
effect
of
localization
upon
performance
because
this
fixture
can
only
influence
performance
through
pure
localization.
The
idea
for
this
fixture
was
developed
while
observing
operators
perform
the
peg
insertion
task
in
person
(not
telepresent).
All
subjects
rested
their
unused
left
hand
upon
the
task
board
while
performing
the
task
with
their
right
hand.
This
suggested
that
the
operators
gathered
useful
localizing
information
from
the
unused
hand.
As
shown
in
Tables
3
and
10,
the
mean
movement
times
for
peg
insertions
performed
with
the
aid
of
virtual
Fixture
5
were
19%
faster
than
with
the
no-fixture
case.
Comparing
Indexes
of
Performance
(Ip)
as
shown
in
Tables
6
and
12,
the
human
information
processing
capacity
when
using
virtual
Fixture
5
was
20%
greater
than
with
no
fixture
at
all.
These
results
suggest
that
kinesthetic/proprioceptive
localization
to
the
remote
site
plays
an
important
part
in
the
effectiveness
of
virtual
fixtures
to
enhance
performance.
These
results
further
suggest
the
importance
of
telepresence
systems
which
allow
for
bilateral
(dual
arm)
interaction
with
the
workspace
even
if
for
tasks
which
only
require
one
hand.
35

Virtual
Fixture
6
As
shown
in
Figure
6,
virtual
Fixture
6
is
identical
to
Fixture
3
with
the
addition
of
two
angled
surfaces
which
guide
hand
motion
directly
into
the
target
hole.
While
Fixture
3
guides
ballistic
motion
by
converging
near
the
target
hole,
Fixture
6
also
guides
fine
motion
by
converging
hand
position
directly
into
the
center
of
the
target
hole.
Although
this
fixture
is
composed
of
4
intersecting
planes,
it
can
be
thought
of
as
two
cones
projected
into
the
two
dimensional
workspace.
One
cone
funnels
ballistic
motion
across
the
length
of
the
board;
the
other
cone
funnels
fine
motion
into
the
target
hole.
As
shown
in
Tables
3
and
10,
the
mean
movement
times
for
peg
insertions
performed
with
the
aid
of
virtual
Fixture
6
were
43%
faster
than
with
the
no-
fixture
case.
Comparing
Indexes
of
Performance
(Ip)
as
shown
in
Tables
6
and
12,
the
human
information
processing
capacity
when
using
virtual
Fixture
6
was
57%
greater
than
with
no
fixture
at
all.
The
results
for
Fixture
6
are
not
statistically
different
from
those
for
the
Fixture
3
case,
thus
showing
that
the
addition
of
the
fine
positioning
surfaces
did
not
enhance
performance
in
these
trials.
This
result
suggests
that
ballistic
motion
accounted
for
the
most
significant
part
of
movement
times.
An
interesting
side
note
is
that
in
a
time
delay
study
currently
being
completed,
Fixture
6
was
found
to
provide
significant
improvement
over
Fixture
3
when
communication
delays
were
introduced
into
the
system.
This
makes
sense
because
fine
position
becomes
more
difficult
when
time
delays
are
present.
These
results
will
be
presented
formally
in
a
companion
paper
addressing
time
delay
issues.
Virtual
Fixture
7
As
shown
in
Figure
6,
virtual
Fixture
7
is
identical
to
Fixture
1
with
the
addition
of
field-like
auditory
feedback.
The
sonic
information
is
modeled
as
a
series
of
parallel
planes
such
that
crossing
of
each
plane
results
in
a
tone
of
a
given
frequency.
As
planes
are
crossed
from
left
to
right,
the
frequency
associated
with
each
plane
increases.
Both
the
density
of
auditory
pulses
and
the
pitch
of
each
pulse
are
useful
cues.
The
result
is
a
texture-like
sonic
field
which
provides
rich
position,
velocity,
and
acceleration
feedback
to
the
operator.
As
shown
in
Tables
5,
10
and
11,
the
mean
movement
times
for
peg
insertions
performed
with
the
aid
of
virtual
Fixture
7
were
20%
faster
than
the
36
<

Fixture
1
case
and
42%
faster
than
with
the
no-fixture
case.
Comparing
Indexes
of
Performance
(Ip)
as
shown
in
Tables
8,
12
and
13,
the
human
information
processing
capacity
when
using
virtual
Fixture
7
was
23%
greater
than
the
Fixture
1
case
and
52%
greater
than
with
no
fixture
at
all.
The
over
20%
improvement
in
performance
when
using
Fixture
7
over
Fixture
1
(which
differ
only
by
the
addition
of
auditory
information
in
7)
strongly
suggests
that
the
use
of
multiple
sensory
modalities
increases
information
processing
capacity
and
is
thus
a
powerful
tool
in
fixture
design.
It
further
suggests
that
overlaying
an
auditory
gradient
field
on
top
of
the
workspace
to
provide
position,
velocity,
and
acceleration
cues
is
a
simple
means
of
enhancing
operator
performance.
Post
testing
subject
interviews
revealed
that
most
subjects
were
unaware
that
their
performance
increased
with
the
additional
auditory
information,
but
all
reported
having
a
better
sense
of
the
shape
of
their
peg
insertion
trajectories.
Virtual
Fixture
8
As
shown
in
Figure
6,
Fixture
8
is
identical
to
Fixture
6
with
the
addition
of
an
auditory
compliant
surface
modeled
such
that
compression
of
the
surface
corresponds
to
increasing
pitch
of
the
auditory
feedback.
The
surface
is
positioned
such
that
first
contact
occurs
just
before
the
peg
should
be
inserted
into
the
target
hole.
This
auditory
surface
provides
a
well
positioned
localizing
reference
to
act
as
a
warning
signal
to
indicate
that
the
peg
has
traveled
far
enough
and
now
should
be
inserted.
As
shown
in
Tables
5,
10
and
11,
the
mean
movement
times
for
peg
insertions
performed
with
the
aid
of
virtual
Fixture
8
were
12%
faster
than
the
Fixture
6
case
and
53%
faster
than
with
the
no-
fixture
case.
Comparing
Indexes
of
Performance
(Ip)
as
shown
in
Tables
8,
12
and
13,
the
human
information
processing
capacity
when
using
virtual
Fixture
8
was
13%
greater
than
the
Fixture
6
case
and
70%
greater
than
with
no-fixture
at
all.
The
over
12%
improvement
in
performance
when
using
Fixture
8
over
Fixture
6
(which
differ
only
by
the
addition
of
auditory
information
in
8)
suggests
that
the
use
of
the
additional
modality
was
useful
in
increasing
information
processing
capacity.
It
should
also
be
noted
that
of
all
the
fixtures
tested,
the
use
of
Fixture
8
resulted
in
the
greatest
improvement
in
movement
times
and
Index
of
Performance
over
the
no-fixture
case.
This
suggests
that
the
use
of
multiple
modalities
in
fixture
design
is
a
powerful
37

perceptual
aid
and
should
be
expanded
to
visual
and
tactile
information
as
well.
Post
testing
interviews
revealed
that
the
use
of
the
auditory
fixtures
caused
subjects
to
alter
their
conceptualization
of
the
task
such
that
a
successful
trial
no
longer
just
looked
and
felt
a
certain
way,
but
also
sounded
a
certain
way.
We
can
further
compare
the
advantage
of
using
multiple
modalities
in
fixture
design
by
analyzing
the
slopes
of
the
Fitts'
Law
curve
relating
movement
times
to
Index
of
Difficulty
for
each
fixture.
The
inverse
of
the
slope
of
the
Fitts'
Law
curve,
computed
by
linear
regression
and
listed
in
Table
9,
has
units
(bits/sec)
and
is
an
accepted
measure
of
overall
information
processing
capacity
of
the
operator.
This
information
is
also
represented
in
graphical
form
in
Figure
8.
Comparing
each
fixture
tested,
the
two
fixtures
whose
use
resulted
in
the
greatest
overall
increase
in
information
processing
capacity
were
Fixtures
7
and
8,
the
only
two
fixtures
to
introduce
auditory
information.
This
analysis
further
supports
the
idea
that
using
virtual
fixtures
to
introduce
abstract
sensory
information
from
multiple
modalities
is
a
viable
method
of
enhancing
operator
performance.
38

CONCLUSIONS
The
results
of
this
study
confirm
that
overlaying
abstract
sensory
information
in
the
form
of
virtual
fixtures
on
top
of
the
sensory
feedback
from
a
remote
environment
can
greatly
enhance
performance
in
telemanipulation
tasks.
Virtual
fixtures
composed
of
simple
combinations
of
rigid
impedance
surfaces
and
abstract
auditory
information
increased
operator
performance
by
up
to
70%
in
a
standard
peg
insertion
task.
It
should
be
noted
that
a
peg
insertion
task
was
chosen
for
this
study
because
it
offered
an
effective
means
of
quantifying
operator
performance
and
not
because
it
was
thought
to
be
a
particularly
good
application
for
virtual
fixtures.
Although
virtual
fixtures
proved
to
be
an
effective
means
of
improving
peg
insertion
performance,
it
is
likely
that
the
fixtures
would
have
had
a
much
greater
impact
upon
the
performance
of
a
more
challenging
telemanipulation
task.
Because
effective
virtual
fixtures
were
developed
from
very
basic
elements
like
rigid
impedance
surfaces
and
simple
gradient
fields,
the
development
of
an
interactive
perceptual
workstation
that
allows
an
operator
to
build
virtual
fixtures
from
basic
building
blocks
seems
like
a
feasible
endeavor.
Such
an
interactive
workstation
could
allow
a
teleoperator
to
develop
powerful
virtual
fixtures
upon
first
encounter
with
an
unfamiliar
task
in
an
unstructured
environment.
Such
a
workstation
might
also
allow
an
operator
to
quickly
modify
a
fixture
as
task
conditions
change.
Analysis
of
some
basic
perceptual
elements
tested
in
this
study
suggests
that
virtual
fixtures
enhance
operator
performance
in
the
following
ways:
1.
Virtual
fixtures
simplify
the
operator's
interaction
with
the
remote
workspace
by
restricting
unnecessary
kinematic
freedoms
and
preventing
unwanted
actions
from
occurring.
This
simplification
limits
the
alternatives
the
operator
has
to
consider
and
reduces
the
information
that
needs
to
be
processed.
This
simplification
also
allows
the
operator
to
relax
criteria
for
success
and
failure
by
eliminating
some
modes
of
failure
and
by
providing
assistance
in
achieving
some
aspects
of
a
successful
task
completion.
39

2.
Virtual
fixtures
alter
the
operator's
conceptualization
of
the
remote
task
by
introducing
abstract
sensory
information
into
the
workspace.
Without
the
aid
of
any
virtual
fixturing,
the
remote
peg
insertion
task
was
primarily
a
visual
operation.
Thus,
the
operator
was
likely
to
conceptualize
the
peg
insertion
task
as
a
manual
procedure
that
"looks
a
certain
way."
With
virtual
fixtures,
additional
haptic
and
auditory
information
is
provided
to
the
operator.
This
additional
sensory
information
allows
the
operator
to
conceptualize
the
peg
insertion
task
as
not
just
"looking
a
certain
way"
but
also
as
"feeling
a
certain
way"
and
"sounding
a
certain
way."
3.
Virtual
fixtures
were
shown
to
provide
a
localizing
reference
to
the
remote
worksite
by
introducing
haptic
and
auditory
cues
that
can
be
coupled
to
the
operator's
kinesthetic
sense
of
workspace
geometry.
This
was
demonstrated
dramatically
by
test
Fixture
5
which
allowed
a
20%
performance
improvement
by
providing
haptic
localizing
information
to
the
operator's
unused
hand.
4.
Virtual
fixtures
can
displace
the
burden
from
taxed
sensory
modalities
by
providing
information
through
alternative
sensory
pathways.
Without
the
aid
of
virtual
fixtures,
the
peg
insertion
task
primarily
taxed
the
visual
modality.
With
the
fixtures,
haptic
and
auditory
sensations
could
also
be
relied
upon.
Had
this
task
required
additional
use
of
the
visual
modality
for
scanning
the
scene
or
monitoring
some
other
aspect
of
the
task,
the
beneficial
effect
of
displacing
the
burden
from
the
visual
modality
would
likely
have
been
even
more
pronounced.
40

REFERENCES
[I]
Akin,
D.L.,
Minsky,
M.L.,
Thiel,
E.D.,
and
Kurtzman,
C.R.,
"Space
Applications
of
Automation,
Robotics,
and
Machine
Intelligence
Systems,"
(ARAMIS)
-
Phase
II,
Vol.
3:
Executive
Summary,
M.I.T.,
Cambridge
MA,
1983.
[2]
Bejczy,
A.,
Kim,
W.,
and
Venema,
S.,
"The
Phantom
Robot:
Predictive
Displays
for
Teleoperation
with
Time
Delay,"
IEEE
Robotics
and
Automation,
1990.
[3]
Brooks,
T.,
Ince,
I.,
and
Lee,
G.,
"Operator
Vision
Aids
for
Space
Teleoperation
Assembly
and
Servicing,"
Space
Operations,
Applications,
and
Research
Symposium
1991.
[4]
Burdea,
G.,
and
Zhuang,
J.,
"Dexterous
Telerobotics
with
Force
Feedback
-
an
overview.
Part
I:
Human
Factors,"
Robotica,
vol.
9,
pp.
171-178,
1991.
[5]
Burdea,
G.,
and
Zhuang,
J.,
"Dexterous
Telerobotics
with
Force
Feedback
-
an
overview.
Part
II:
Control
and
implemenation,"
Robotica,
vol.
9,
pp.
291-298,
1991.
[6]
Fischer,
R.D.,
and
Siva,
K.V.,
"Specification
and
Design
of
Input
Devices
for
Teleoperation,"
IEEE,
1990.
[7]
Fitts,
P.M.,
"The
Information
Capacity
of
Human
Motor
Systems
in
Controlling
the
Amplitude
of
a
Movement,"
Journal
of
Experimental
Psychology,
vol.
47,
pp.
381-391,
1954.
[8]
Hill,
J.W.,
"Study
of
Modeling
and
Evaluation
of
Remote
Manipulation
Tasks
with
Force
Feedback,"
Final
Report,
SRI
Project
7696,
JPL
Contract
95-5170,
1979.
[9]
Jacobsen,
S.C.,
Smith,
F.M.,
Backman,
D.K.,
and
Iversen,
E.K.,
"High
Performance,
High
Dexterity,
Force
Reflective
Teleoperator
II,"
ANS
Topical
Meeting
on
Robotics
and
Remote
Systems,
New
Mexico,
1991.
[10]
Julian,
R.G.,
and
Anderson,
T.R.,
"Robotics
Telepresence:
Applications
of
Human
Controlled
Robots
in
Air
Force
Maintenance,"
Aerospace
Simulation,
ISBN
0-911801-28-6,
1988.
[II]
Kim,
W.,
Tendick,
F.,
and
Stark,
L.,
"Visual
Enhancements
in
Pick
and
Place
Tasks:
Human
Operators
Controlling
a
Simulated
Cylindrical
Manipulator,"
IEEE
Journal
of
Robotics
and
Automation,
RA-3,
1987
(b).
[12]
MB
Associates,
"Operations
and
Maintenance
Manual,
Manipulator
Arms
Systems,"
San
Ramon
CA,
April
1976.
[13]
McGovern,
D.E.,
"Factors
Effecting
Control
Allocation
of
Augmented
Remote
Manipulation,"
Doctoral
dissertation,
Stanford
University,
Stanford
CA,
1974.
41

[14]
Oxenberg,
S.,
Landell,
B.,
and
Kan,
E.,
"Geometric
Database
Maintenance
Using
CCTV
Cameras
and
Overlay
Graphics,"
SPIE
Vol.
1006
-
Space
Station
Automation
IV,
1988.
[15]
Pepper,
R.L.,
and
Hightower,
J.D.,
"Research
Issues
in
Teleoperator
Systems,"
28th
Annual
Human
Factors
Society
Meeting,
San
Antonio
TX,
1984.
[16]
Pepper,
R.L.,
and
Kaomea,
P.K.,
"Teleoperation:
Telepresence
and
Performance
Assessment,"
Annual
International
Ergonomics
Society
Meeting,
Teleoperation
and
Control,
July
1988.
[17]
Remis,
S.J.,
and
Repperger,
D.W.,
"Quantifying
Performance
Degradation
Due
to
the
Human-Machine
Interface
of
Telemanipulators,"
Armstrong
Laboratories
internal
report
AAMRL-SR-90-510,
Dec
1990.
[18]
Repperger,
D.W.,
and
Goodyear,
C,
"Active
Controllers
and
Time
Duration
to
Learn
a
Task,"
Conference
on
Manual
Control,
1985.
[19]
Repperger,
D.W.,
Remis,
S.J.,
and
Merrill,
G.,
"Performance
Measures
of
Teleoperation
Using
an
Exoskeleton
Device,"
Proc.
1990
IEEE
Int.
Conf.
Robot.
Automat.,
vol.
1,
pp.
552-557,
1990.
[20]
Russell,
R.,
and
D'Arcy,
A.,
"Video-Based
Satellite
Attitude
Determination,"
SPIE
Vol.
729
-
Space
Station
Automation
II,
1986.
[21]
Sheridan,
T.,
"Human
Supervisory
Control
of
Robot
Systems,"
IEEE
Robotics
and
Automation,
Vol.
2,
1986.
[22]
Sheridan,
T.,
"Telerobotics,"
Automatica,
25(4),
1989.
[23]
Sheridan,
T.B.,
and
Ferrell,
W.R.,
"Man-Machine
Systems,"
The
MIT
Press,
1981.
[24]
Spain,
Edward,
H.,
"Peg-In-Hole
Taskboard
Documentation,"
Final
Report
for
Task
FQ7624-88-00024,
March
1989.
[25]
Stark,
L.,
et
al.,
"Telerobotics:
Display,
Control,
and
Communications
Problems,"
IEEE
Robotics
and
Automation,
1989.
[26]
Thompson,
D.A.,
"The
Development
of
a
Six
Degree
of
Freedom
Robot
Evaluation
Test,"
In
Proceedings
of
13th
Annual
Conference
on
Manual
Control,
Cambridge
MA,
MIT,
1977.
42