Content uploaded by Mark Cutkosky
Author content
All content in this area was uploaded by Mark Cutkosky on Oct 23, 2015
Content may be subject to copyright.
IEEE TRANSACTIONS ON ROBOTICS AND AUTOMATION,
VOL.
5,
NO.
3.
JUNE
1989
269
On Grasp Choice, Grasp Models, and the Design
of Hands for Manufacturing Tasks
MARK
R.
CUTKOSKY
Abstract-Current analytical models of grasping and manipulation
with robotic hands contain simplifications and assumptions that limit
their application to manufacturing environments.
Ta
evaluate these
models, a study was undertaken of the grasps used by machinists in a
small batch manufacturing operation. Based on the study, a taxonomy of
grasps was constructed. An expert system also was developed to clarify
the issues involved in human grasp choice. Comparisons of the grasp
taxonomy, the expert system and grasp quality measures derived from the
analytic models reveal that the analytic measures are useful for describing
grasps in manufacturing tasks, despite the limitations in the models.
In
addition, the grasp taxonomy provides insights for the design of versatile
robotic hands for manufacturing.
I. INTRODUCTION
S
MULTIFINGERED robotic hands begin to appear in
A
research laboratories, the design, analysis, and control of
such hands has become an active area of research. Numerous
analytic approaches have been proposed for characterizing
grasps and modeling the process of manipulation.
In
addition,
there have been significant advances in control strategies and
tactile sensing for hands. Yet it seems that we are still a long
way from building robots that can independently decide how to
pick up and manipulate objects to accomplish everyday tasks.
Part of the problem is that since hands and manipulation are
complex, attempts to model them require simplifying assump-
tions not usually valid outside of carefully structured experi-
ments in the laboratory. Consequently, we were lead to
compare analytic grasp models with the processes that people
use in choosing grasps and manipulating tools and workpieces
in a particular environment.
The work addresses a number of basic questions:
Can an order be imposed on human grasp selection and
can the process be codified?
How limiting are the assumptions made in today's
analytic grasp analyses and are the resulting grasp quality
measures practical?
How does human grasp selection compare with the
analytic approaches?
Are the results of studying human grasp selection useful
for the design of robot hands and for automating robotic
grasp selection?
Manuscript received April 20, 1988; revised July 6, 1988. Part of the
material in this paper was presented at the 1986 International Conference
on
Robotics and Automation, San Francisco, CA, April 1986. This work has
been supported by the National Science Foundation under Grants
DMC8552691 and DMC8602847.
The author is with the Department of Mechanical Engineering, Stanford
University, Stanford, CA 94305.
IEEE
Log
Number 8825229.
In retrospect, the most useful contribution of the study of
human grasps, from the standpoint of designing and control-
ling robot hands, has been
a
better appreciation of how task
requirements and object geometry combine to dictate grasp
choice. The study has resulted in a grasp taxonomy, which
makes it possible to identify particular grasps and to trace how
they derive from generic grasp types. The fact that both task
requirements (e.g., forces) and geometry are important is
clear from everyday experience. The grasp we use for picking
up a pencil is entirely different from the one we use for
writing, although the object geometry remains the same. On
the other hand, if we consider the task of filing a machined
part, the grasp we use for a flat file is different from the grasp
we use for a round one, although the forces and motions are
the same.
Our study of manufacturing grasps focused on tasks in
small-batch machining operations, or job-shops. Small-batch
machining is an increasingly important component of manu-
facturing (roughly
75
percent of machined items are produced
in batches of
50
parts or fewer
[IS])
and has spurred
considerable work
on
flexible, automated machining systems.
Where traditional small-batch operations counted on human
operators to adapt to minor variations in parts, fixtures, and
processes, automated systems rely on sensors, robots, and
computers. Unfortunately, the adaptability of human operators
has been difficult to duplicate, especially
in
handling, assem-
bling, and fixturing parts and tools. Thus it is common today
to see operations in which CNC machine tools cut parts but
humans fixture the parts on pallets and perform numerous
tasks using hand tools to assemble, inspect, and finish parts
and fixtures.
The study of grasps was confined to single-handed opera-
tions by machinists working with metal parts and hand tools.
I
The machinists were observed and interviewed and their grasp
choices were recorded as they worked. In addition, their
perceptions
of
tactile sensitivity, grasp strength, and dexterity
were recorded. Preliminary results of the study, and a
resulting partial taxonomy of manufacturing grasps, were
presented in
[4].
In subsequent work, a grasp expert system
has been developed, using the original results and taxonomy as
a starting point. The purpose of the codification exercise was
not to develop a program to predict what grasp a human would
adopt under particular circumstances (although it now appears
that this can be done in a limited context) but
to
have a
running, testable framework in which to try out hypotheses. In
'
However, we observe that two-handed tasks often
use
the same grasps as
found in
our
one-handed taxonomy, presented in Fig.
4.
1042-296X/89/06OO-0269$01
.OO
0
1989 IEEE
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
270
IEEE
TRANSACTIONS
ON
ROBOTICS
AND
AUTOMATION,
VOL.
5,
NO.
3,
JUNE
1989
addition, the codification exercise forces one to be more
careful about defining terms and organizing information.
While the expert system is not yet, and probably never will be,
complete (after all, how useful is an expert system that tells us
how we grip things?)
it
has forced a closer look at how grasps
are chosen and has resulted in modifications to the original
taxonomy in
[4].
The codification exercise has also lead us to
explore patterns or sequences among grasps, which provide
insights for controlling robotic hands to manipulate parts.
However, from the standpoint of hand design, we find that
while the expert system contains a great deal more information
than can be represented in a taxonomy, the taxonomy remains
more useful as a design aid since it allows one to see very
quickly where a particular grasp resides in the space of
possible grasps.
In
the following sections, we briefly review analytic grasp
models and examine the assumptions upon which these models
rest. We then present the results of our study
of
human grasp
selection
in
manufacturing tasks and describe the grasp expert
system that grew out of the study. Finally, we discuss the
results of the study and codification exercise in terms
of
their
ramifications for designing manufacturing hands.
11.
ANALYTIC APPROACHES
TO
GRASP MODELING
AND
GRASP
CHOICE
A.
Grasp Modeling
As Fig.
1
indicates, manipulation is complex, typically
involving combinations of open and closed kinematic chains,
nonholonomic constraints, redundant degrees of freedom, and
singularities.
In
addition, there are nonlinearities in the contact
conditions between soft, viscoelastic fingers and grasped
objects, and
in
the drive-train and actuator dynamics. To keep
the analysis tractable, early analyses (e.g.,
[l])
made the
following assumptions, many of which are also found in
current analyses of dextrous manipulation:
rigid-body models with point contacts between the
fingertips and the grasped object
linearized (instantaneous) kinematics
quasi-static analysis
(no
inertial or viscous terms)
no
sliding or rolling of the fingertips
no
cases with redundant degrees of freedom and no
Recent analyses, such as those by Nakamura
et al.
[19],
Cutkosky and Wright
[5],
Ji
[
1
I],
and Li and Sastry
[
141,
have
relaxed some of these assumptions, although at the cost of
greater complexity. Moreover, even the most sophisticated
models involve the following simplifications:
overconstrained grasps.
idealized models of the fingertips (e.g., point-contact or
“soft finger” models with linear elastic deformation)
idealized friction models (e.g., Coulomb friction) that
ignore the effects
of
sliding velocity, material properties
of the “skin,” and the presence of dirt or moisture
simplified actuator and drive-train dynamics, ignoring
elasticity, backlash, and friction
simplified representations of the grasped objects, typi-
cally treating them as smooth, rigid geometric primitives
or
polyhedra.
particular
5
homogeneous
sol
ns
reflecled
inertla
propertles
stability
lmpedanceladmlltance
COdtUtfrr
rChUOM
Jalnt
5
llnk
compliance
flngerllp
deformations
canlacl
propertles
rzictlon
conditions
r-
obJect
stlflness
Fig.
1.
Issues
in analytic modeling
of
grasping and manipulation.
Based
on
the various analytic models of grasping and
manipulation, a number of quality measures have been
developed. For reference, these are summarized in Table
I.
We will return to the measures in Section
V
and compare them
with the empirically derived “grasp attributes” used in the
grasp expert system.
While the measures in Table
I
describe the properties of a
grasp, and are useful for assessing the suitability of a grasp for
a given task, there are clearly other factors involved in grasp
choice. For example,
if
an object is to be picked up from a
table, the grasp cannot place any fingers
on
the underside of
the object. Other considerations include the size, shape, and
location of the center of mass
of
the object and the work space
of the hand. Thus a number of investigators have proposed
geometric criteria for automated grasp selection
[2],
[
151,
1281.
B.
Analytic Grasp Choice
The problem of choosing a grasp, based
on
analytic grasp
models, quality measures, and constraints, is illustrated in Fig.
2.
There are three overlapping sets of constraints arising from
the task (e.g., forces and motions that must be imparted), from
the grasped object (e.g., the shape, slipperiness, and fragility
of the object), and from the hand
or
gripper (e.g., the
maximum grasp force and maximum opening of the fingers).
Within these constraints is a space of “feasible grasps.”
Choosing a grasp involves the definition of an objective
function, which is optimized, subject to the constraints. The
approach is conceptually straightforward, except that there is
little agreement
on
which of the measures in Table
I
(along
with additional geometric issues) should be included in the
objective function, and which should be used as constraints.
Kerr and Roth
[12]
establish a polyhedral region of “safe”
grasps, bounded by friction limitations at the contacts. They
define an optimal grasp as one that is furthest from the
boundaries of the friction polyhedron, while also satisfying
force-closure and constraints
on
internal forces and actuator
torques.
By contrast, Nakamura
et al.
[19]
search for a grasp that
minimizes internal forces (and consequently, grasping effort)
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
CUTKOSKY: DESIGN OF HANDS FOR MANUFACTURING TASKS
27
1
Compliance
Connectivity
Force closure
Form closure
Grasp
isotropy
Internal forces
Manipulability
Resistance to
slipping
Stability
TABLE
I
DEFINITIONS
OF
ANALYTICAL
MEASURES
USED
TO
DESCRIBE
A
GRASP
-~~____~~
~ ~~~ ~
__~~
~ ~~~~ ~~~ ~~~
-~
~~
__
~__
~~
What is the effective compliance (inverse of stiffness) of the grasped object with respect
to
the hand? The grasp
compliance matrix is
a
function of grasp configuration, joint servoing, and structural compliances in the links,
joints, and fingertips
[6].
How many degrees of freedom are there between the grasped object and the hand? Formally, how many
independent parameters are needed to completely specify the position and orientation of the object with respect to
the palm
[
17]?
Assuming that external forces maintain contact between the fingers and the object, is the object unable
to
move
without slipping when the finger joints are locked? Formally,
a
grasp satisfies force closure if the union of the
contact wrenches has rank
6 [17], [22].
Can external forces and moments be applied from any direction without moving the object, when the fingers are
locked? Formally, there is form closure,
or
complete kinematic restraint, if the intersection of
all
unisense contact
twists is a
null
set. Thus seven frictionless point contacts are in general required
to
achieve form closure
on
a
rigid
Does the grasp configuration permit the finger joints to
accurately
apply forces and moments to the object? For
example, if one of the fingers is nearly in
a
singular configuration, it will
be
impossible to accurately control force
and motion in
a
particular direction. Formally, the
grasp
isotropy is
a
function of the condition number of the
grasp
Jacobian matrix
[12], [17].
Li and Sastry
[14]
define similar grasp quality measures that are functions of the
singular values of the grasp Jacobian.
What kinds of internal
grasp
forces can the hand apply to the object? Formally, the internal
grasp
forces
are
the
homogeneous solution to the equilibrium equations of the object. Thus internal grasp forces can be varied without
disturbing the grasp equilibrium
[12],
[
171.
While not consistently defined
in
the literature, a useful definition is: Can the fingers
impart
arbitrary motions to
the object? Thus
a
manipulable grasp must have force closure and
a
connectivity of
6.
In addition, the rank space of
velocities due to the finger joints must span the space of velocities transmitted through the contacts
1121.
How large can the forces and moments on the object be before the fingers will start to slip? The resistance to
slipping depends on the configuration of the grasp, on the types of contacts, and on the friction between the object
and the fingertips
[5], [IO]-[12].
Will the grasp return
to
its initial configuration after being disturbed by
an
external force
or
moment? At low
speeds, the grasp is stable if the overall stiffness matrix is positive definite
[6], 1211.
At higher speeds,
dynamic
stability must be considered
(191.
body
1131,
U71.
Fig.
2.
Choosing
a
grasp that maximizes an objective function subject
to
task, object, and gripper constraints.
subject to constraints on force-closure, friction, and manipula-
bility. If a safety factor is used
in
setting the friction
constraints, this approach should give results sjmilar to the
approach that people seem to use, with forces a consistent
percentage above the minimum required to prevent slipping
[24], [29]. In a very different approach, Jameson and Leifer
[lo]
adopt a numerical hill-climbing technique
in
which a
simplified three-fingered hand searches for positions that are
most resistant to slipping, subject to constraints on joint
torques and geometric accessibility. However, they cast the
constraints as potential functions
so
that their effects are added
to those of the objective function. In still other work, Li and
Sastry
[14]
define a “task ellipsoid,” whose orientation and
relative dimensions depend on the expected magnitudes of
forces and moments during a task. Grasps are then compared
according to the largest diameter of the task ellipsoid that they
can encompass.
While there are numerous articles on grasp stability, force-
closure, and quality measures for comparing different grasps,
little has been proposed in the way
of
an overall strategy for
grasp planning. However, Ji
[l
11
outlines a sequence in which
the first step is to find “grasp planes” containing grasps that
satisfy form closure and have the ability to control internal
forces. Next, the grasps are checked for accessibility con-
straints (e.g., which parts of the object can the fingers actually
reach?) and finally, task requirements are checked, possibly
using a task-oriented quality measure such as that proposed by
Li and Sastry
[
141.
As
we review the competing approaches in the literature,
and examine the serious simplifications upon which they rest,
we are lead to wonder how useful the analytic approaches to
grasp choice can be outside of carefully controlled laboratory
experiments.
To
be fair, any of the models may be a
reasonable approximation for a particular set of tasks. Thus
while point-contact is a poor approximation when human
fingers hold a small object,
it
is a fair approximation as long as
the contact areas are small compared to the characteristic
length of the object [5]. Nonetheless, we are motivated to look
at some actual manufacturing tasks and the characteristics of
the grasps adopted to accomplish them.
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
272
IEEE
TRANSACTIONS
ON
ROBOTICS
AND
AUTOMATION,
VOL.
5,
NO.
3,
JUNE
1989
I
Task
requirements
Object
attributes
Taxonomy
I
guidelines
for
hand
design
Fig.
3.
Task requirements and object attributes combine to dictate the grasp
choice. Viewing the grasp as part
of
a taxonomy permits
us
to draw
conclusions about designing robotic hands.
111. EXPLORING
THE
HUMAN GRASP SELECTION
PROCESS
If robot hands are going to succeed in small batch
manufacturing they will have to display some of the same
adaptability and sensitivity that human hands do. Thus it is
useful to analyze human grasps
in
machining, not necessarily
to imitate them, but to understand the relationship between
task requirements and the grasping “solution” adopted to
meet those requirements. This philosophy is summarized in
Fig.
3.
Of course, we have to be careful
in
drawing
conclusions based
on
a study of human hands. The hand has
evolved over millions
of
years as an organ used as much for
sensation and communication as for manipulation.
In
fact, for
many manufacturing tasks the human hand is less than ideal.
When a mechanic starts to work
on
a machine, the first thing
he reaches for is his toolbox, with pliers, wrenches, tweezers,
and work gloves to help him finish the job. This suggests that
by understanding the grasping and manipulation requirements
for tasks
in
a specific environment it should be possible to
design a hand that
exceeds
human performance.
A.
Previous Explorations of Human Grasps
The study of human grasping has long been an area of
interest
for
hand surgery, for designing prosthetic devices, and
for quantifying the extent of disability in individuals with
congenital defects
or
injuries. As a result, there
is
a substan-
tial, empirical, medical literature
on
the grasping capabilities
of the human hand. Much of the literature refers to six grasps
defined by Schlesinger [19] and summarized by Taylor and
Schwarz [27]
:
cylindrical, fingertip, hook, palmar, spheri-
cal,
and
lateral.
Such a categorization leads to associating
grasps with part shapes. Thus a sphere suggests a spherical
grip while a cylinder suggests a wrap grip. However, as the
pencil example cited earlier illustrates, when people use
objects in everyday tasks, the choice of grasp is dictated less
by the size and shape of objects than by the tasks they want to
accomplish. Even during the course of a single task with a
single object, the hand adopts different grips to adjust
to
changing forceltorque conditions. When unscrewing a jar lid,
the hand begins with a powerful grip in which the palm
is
pressed against the lid for extra torque. As the lid becomes
loose, torque becomes less important than dexterity and the
hand switches to a light grip in which only the fingertips touch
the jar lid. This suggests that grasps should first be categorized
according to function instead of appearance.
Napier
[20]
suggests a scheme in which grasps are divided
into
power grasps
and
precision grasps.
Where consider-
ations of stability and security predominate (as in holding a
hammer
or
getting a jar lid unstuck) a power grasp is chosen.
Power grasps are distinguished by large areas of contact
between the grasped object and the surfaces of the fingers and
palm and by little
or
no
ability to impart motions with the
fingers. Where considerations of sensitivity and dexterity
predominate a precision grasp is chosen.
In
precision grasps,
the object is held with the tips of the fingers and thumb.
In the following section we begin with the two basic
categories suggested by Napier [20] and develop a hierarchical
tree of grasps.
As
one moves down the tree, details
of
the task
and the object geometry become equally important
so
that
in
the final analysis, both task requirements and object shape play
a central role in determining the grasp.
B.
A
Taxonomy of Manufacturing Grasps
Once the basic choice between a power grasp and a
precision grasp has been made, a combination of task-related
and geometric considerations comes into play. Starting at the
top of Fig.
4,
let us suppose that a power grasp has been
chosen. The first question is: does the object need to be
clamped to sustain forces from a variety of directions, or does
it
merely need to be supported? If
it
merely needs to be
supported then a
nonprehensile
hook grasp (as used in
carrying a suitcase)
or
a palmar support (as used by a waiter
carrying a tray) may be adequate. If the object must be
clamped, a prehensile grip is chosen in which the fingers and
palm confine the object. At this stage some basic geometric
considerations become important: Is the object large? small?
flat? thin? These subsidiary choices are illustrated in Figs.
4
and 5. For example, if a power grip is needed, and the object
is
small and flat (as in turning a key in a lock) then a lateral pinch
(Grasp 15
in
Figs.
4
and 5) will probably be used. If the object
has a compact
or
approximately spherical shape then Grasp 11
is most likely. If the object is prismatic (i.e., a long shape with
nearly constant cross section, such as a cylinder
or
a hexagonal
prism), then a wrap is chosen. Since many objects, including
the handles of most tools, have prismatic shapes, the power
wrap represents a large family
of
manufacturing grips.
Fig.
6
shows several precision grasps from the right side of
the taxonomy. While the different precision grasps appear to
be motivated by part geometry, the decision to use one
precision grasp instead of another may actually be task-related
since many objects have several gripping surfaces with
different shapes. For example, a light cylindrical object can be
gripped either using the thumb and four fingers as
in
Grasp
6,
or
it
can be gripped by one end, like the hollow cylinder shown
for Grasp 12
in
Fig.
6.
C.
Trends in the Taxonomy
Moving from left to right in Fig.
4,
the grasps become less
powerful and the grasped objects become smaller. Thus the
Heavy Wrap grips are the most powerful and least dextrous
(all manipulation must be done with the wrist and even the
wrist is restricted to a limited range of motions) while the
Tripod (Grasp
14)
and Thumb-Index Finger (Grasp
9)
grips
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
CUTKOSKY: DESIGN
OF
HANDS
FOR
MANUFACTURING TASKS
273
emphasis
on
Grasp
emphasis
on
Securfty,Stablllty
dexter@.
sensltluuy
I
Power
clnmplng
required
Non-Ihehensile Prehknsile
thin
Hook.
Uatfam.
push
15
htdpineh
Prismatic Circ'ular
(wrap
aynu".
Bngen lradlal
symmetly.
surround
part)
kgcm
surround
pad
Precision
Specltlc
Grasp8
italic
labels
-
grasp
attributes
boldI.oo~-grp~
4
b
Increasing Power Increasing
Dexterity.
and
Object
Size
Decreasing Object Size
Fig.
4.
A
partial taxonomy of manufacturing grasps, modified from a taxonomy presented in
141.
The drawings of hands were
provided by M.
.I.
Dowling and are reprinted with permission of the Robotics Institute, Carnegie-Mellon University.
are the most precise. However, the trend is not strictly
followed. A Spherical Power grasp may be either more or less
dextrous than a Medium Wrap, depending
on
the size of the
sphere. Moving from top to bottom, the trend is from general
task considerations, such as whether clamping is required,
to
details of geometry and sensing. Again, the trend is not strictly
observed. For example, a small, flat object may provoke the
choice of a Lateral Pinch grip near the top of the tree.
The role of task forces and torques
on
grip choice is most
apparent when the hand shifts between grips during a task. For
example, in unscrewing a knob the hand shifts from Grasp
11
to Grasp
13.
Similarly, when holding a tool, as
in
Grasp
3,
the
hand shifts to Grasp
5
as the task-related forces decrease and
may adopt Grasp
6,
a precision grasp, if the forces become
still smaller. The role of object size is most apparent when
similar tasks are performed with different tools. For example,
in light assembly work Grasps
12
and
13
approach Grasp
14,
and finally Grasp
9,
as the objects become very small.
A
related observation, brought out more clearly in developing
the grasp expert system discussed in Section
IV,
is that
sequences can be traced in the taxonomy, corresponding
to
adjustments that the machinists make in response to shifting
constraints.
D.
Limitations
of
the
Taxonomy
While the taxonomy
in
Fig.
4
has proven to be a useful tool
for classifying and comparing manufacturing grasps, it suffers
from a number of limitations.
To
begin with, it is incomplete.
For example, there are numerous everyday grasps, such as the
grasp that people use in writing with a pencil or in marking
items with a scribe (Figs.
7
and
8)
that are not included. It was
also found that the machinists in our study adopted numerous
variations
on
the grasps in Fig.
4,
partly in response to
particular task
or
geometry constraints and partly due to
personal preferences and differences in the size and strength of
their hands. Such individual grasps could usually be identified
as "children" of the grasps in Fig.
4.
To examine such issues
further, and to clarify the roles of dexterity, sensitivity and
stability in grasp choice, an expert system was constructed for
choosing grasps from initial information about the task
requirements and object shape.
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
274
IEEE
TRANSACTIONS ON ROBOTICS AND AUTOMATION,
VOL.
5,
NO.
3,
JUNE
1989
Grasp
2
Heavy
Wrap
-
Fmgm
and
palm
map
around
a
heavy objecl
FrlcUon pmvldes
much
oi
thc
lone
balance
Here. loading
a
lathe
chuck
1s
a
heavy Fg-In-hole assembly
Ysk
using
compliance
md
force
isdback
The
-si
prrforms
fine
Grasp
5.
Ltghl
Tool
--
Fingers
panidly swround
the
objecl.
bu1
there
IS
also
up
prehension This
15
the
pan.
but hac
some
atmbuies
of
praismn
p\pr
good
lactile
~cnstuvlly
and an ability
10
adpsl
the
grasp
stiffness.
1
power
grasp
snce
!k
fingers
do
Mt
manlpulale
Grasp
16
LaleraI
Rnch
--
A thin object
IS
clamped
between
Ihe
lhumb
an0
ride
of
Ihe
mder
finger
Ta,k
and
grasp
iones
are
high
Thls
graw
IS
used
lo
tum
a
key
tn
a
lock
and
IE
closesi
to
the
grrp
achieved
wi!h
Z~fingcr
tmdusbd
gnppere
Fig.
5.
Several power grasps used
Grasp numbers mati
Grasp
3
Mehum
Wrap
abut
a
ml
The
object
IS
smaller
and
lighter
lhan
in
Crap
2
ai
left
bur mk
and
grip
force(
are
high
Large
~OUO~E
are
made
wi!h
lhc
arm
and
line
mmions wilh
the
wnsr
The
thumb
may
he
adduced
(Gmp 4)
for
contn,l
of
the
Uxd
up
Fingers
and
thumb
curl
Grasp
15
(nxdified)
Hook
The
hook
IS
one
01
d
class
of one
sided
grasps
(a
pure
hook
would
not
ur
Ihc
thumb
as
Shown
hcrc)
Ihai
includes
a
waierr
platom
fur
supwnmg
a
my
Force\
are
largely
,none dimlion A
hmk
in
which !he
hqen
JR.
uehily
cwlcd
komer
I)
Medium
Wup
Gracp
6
Vwsed
Thumb
4
Finger
A
pefision
Grasp
9
Opposed
Thumblndcr
Fmger
A
grasp
grasp
for
long
objects
Fingerups
and thumb
for
small
obFe
Fmger
and thumb manipulate
rnmLp"la1e
Ihe
pan
Wllh
be31
Cmml
m
mlbng
11
abut
1c1
axis A
penon
may
use
thls
grasp
lum
he
pa
wiIh
high mobility
along
two
axes
FlngeNps conform
Lo
and
pamally
enoap
lhe
pan
.I"
ad,u\ung
screw
Wllh
d
"211
screwdriver
in
a small-batch machining environment.
:h
the numbers
in
Fig.
4.
Grasp
12
Rec~sm
Disk
--
A radlally
rpmemc
grasp
with
the
fingerups
Fingers
can
adj~i
onsnuuon
01
Ole pan.
but
larger
lwisung
mouons
arc
made
wilh
!he
wnst
Grasp
IO
uill
converge
10
Ul,r
gracp
a\
'CqUlred
mrqm
dsw
Grasp
14
Tripod
--ne
"CIaUIC"
gmsp
ai
3-
fingered
kitrow
manipulauon.
Thc
ulrrc
mor1
mdependen,
lingen
are
"sd. wlb
nsulung
mobilily
I
all
hrccuons.
Fig.
6.
Several precision grasps used
in
small-batch machining. Grasp
numbers match the numbers
in
Fig.
4.
IV.
GRASP-EXP:
AN
EXPERT SYSTEM
FOR
MANUFACTURING
GRASPS
An object-oriented expert system provides features not
found
in
a tree-like taxonomy. For example, it allows
individual "child" grasps to inherit the properties
of
more
than one parent. It also makes
it
possible
to
assign a
combination
of
qualitative and quantitative attributes to grasps
so
that comparisons may be made with the analytic grasp
I
Circular
(radial
symmetry.
3
'virtual
fingers')
I
I
Prismatic
(opposed
thumb.
2
'Wrtual
fingers')
i/
Thumb2
Finger
Thumbbidex
Finger
//
9
J/
rhMge
Jorces
and
onenfation
Fig.
7.
Trends
in the grasp taxonomy
measures discussed in Section
11.
More importantly, the expert
system makes it easy to consider extra constraints (e.g., only
three fingers will fit on the handle of a particular tool) and to
ask "what if" questions (e.g., "what if
I
only had three
fingers and could not oppose my thumb?").
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
CUTKOSKY: DESIGN OF HANDS FOR MANUFACTURING TASKS
275
Opposed
thumb/
2
finger
gnp
Scnk
pp
Pencil
gnp
hecision. light
10
Precision, but
large
Precision,
but
less Precision, light forces,
moderate forces. high forces, some modon mobility
and
mort
high mobility. Axis of
mobdny.
Tool
=is
with
wnst.
Tool
axis suppon. tml
@le1
IO
fingers
perpendicular
io
fingers.
Fig.
8.
largely perpendicular to
fingers
A sequence of
grips
in response
to
changing
task
requirements
An expert system, Grasp-Exp, was written in POGIE,2 a the system (e.g., to associate numerical values with the grasp
framework developed by
K.
Ishii for his work on knowledge-
sensitivity) without affecting the syntax of many rules. In fact,
based design
of
mechanical systems
[9].
POGIE is written in some care is required
in
defining grasp attributes. For
Common Lisp and supports forward and backward chaining, example, the easiest way to identify a precision grasp would be
numerical values, and fuzzy measures. The basic syntax of to ask whether the fingers actively manipulate the part or
statements in
POGIE
is predicate logic:
tool-but this is hardly fair, since part of the grasp choice
(Rule
37
(if (grasp is precision-grasp)
(*provable (requires object-size small) fail)
(requires rough-object-shape compact))
(then (grasp is precision-circular-grasp)))
;IF the grasp is a precision grasp
;AND the object is not small
;AND the object has a compact shape
;THEN the grasp is a “precision circular grasp”
It is also straightforward to include numerical criteria
in
POGIE:
(Rule
07
(if (requires stability $num)
(*
>
$num
0.75
t)
(requires security $num)
(*
>
$num
0.75
t)
(then (grasp is power grasp)))
;IF the grasp requires a stability rating
;of greater than
75%
;AND a security rating
;of greater than
75%
;THEN the grasp is a “power grasp”
Several versions
of
Grasp-Exp were developed as we experi-
mented with different approaches to interacting with the user
and presenting information about the task and the grasped
object. The first attempt was to put the rules for grasp
selection directly into the framework that Ishii had developed
for interactive systems design. A procedural front-end would
ask questions of the user and record the answers
in
a list of
facts. Grasp-Exp would then
try
to draw conclusions about the
grasp. Unfortunately, the question-asking procedure tended to
ask unnecessary questions which irritated users. For example,
the system might ask about the delicacy of touch required for a
“heavy wrap” grasp. In addition, we had not distinguished
carefully between
grasp-types
and
grasp-attributes
in the
knowledge base. The
grasp-type
is the classification
in
the
taxonomy. For example,
heavy- wrap-grasps
are a subset of
wrap-grasps, prehensile-grasps,
and
po wer-grasps.
The
grasp-attributes
are the characteristics required of the grasp
(e.g., sensitivity, stability), established by interrogating the
user. As a result of this confusion, it was difficult to modify
*
Portable Generic Inference Engine.
~
exercise is to decide whether the object can be manipulated
with the fingers or whether
it
should be manipulated with the
wrist.
The latest version of Grasp-Exp has the structure shown
in
Fig.
9.
The user begins by entering any number of facts about
the grasp; for example, that the grasp requires large forces and
involves a large, cylindrical workpiece. The initial facts are
then acted upon by the Grasp-Exp in either a forward chaining
or backward chaining mode. In the former mode, the system
tries to prove that one or more of the individual grasps will
satisfy all the requirements, and
in
the latter mode, Grasp-Exp
uses the initial facts to trigger rules. In either case, Grasp-Exp
asks questions only when it cannot prove or deduce further
results without additional information. However, due to the
branching nature of the taxonomy, backward chaining
is
more
efficient. Thus when trying to “prove” that a grasp might be
manipulated with the fingers the system chains backward,
looking for supporting evidence (e.g., are the part and the task
forces relatively light? how important is sensitivity? etc.). In
this way, Grasp-Exp has become closer to the classic
consultant who lets the
USF
r
lay out some initial facts and then
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
276
Interface
(question-
asking
procedure)
IEEE
Inferences
4
‘facts’
rules
for
selection
gasp
choke
predlction
Fig.
9.
The architecture
of
Grasp-Exp.
(WHAT IS THE DEXTERITY FUCQUIREIIENT
?
SELECT
FROM
(YES
NO
U”))
>no
(WHAT IS THE STABILITY FUCQIJIREIIENT
?
SELECT
FRW
(YES
NO
UNKNOWN))
>yes
(HOW IMPORTANT IS SECURITY IN YOUR GRASP? SELECT FROM
([O
TO
11
U”0IIw))
>0.7
(WHAT IS THE CLAMPING REQUIREMENT
?
SELECT
FROH
<YES
NO
-OWN))
>yes
(WHAT IS THE OBJECT-THICKNESS REQU1”T
?
SELECT
FROM
(THIN NOT-THIN))
>not-thin
(WHAT
IS THE OBJECT-SIZE ReQUIReneNT
?
SELECT
FROM
>medium
(WHAT
IS
THE ROUGH-OBJECT-SHAPE REQUIRF.MF.NT
?
SELECT
FROM
>compact
(WHICH
OF
THE FOLLOWING
XS
CLOSEST TO THE SHAPE
OF
THE OBJECT? SELECT
(SMALL MEDIUM WGE
UNKNOWN))
(COMPACT PRISlUTIC
UHKNOWN))
FROM
(SPHERE DISK RECTANGLE CYLINDER UNKNOWN))
>disk
REQUIRED GRASP
is
of
type:
POWER DISK-GRASP [GRASP
101
Other
classifications
are
as
follows:
POWER-GRASP
PREHENSILE-GRASP
PENTADACTYL-GRASP
CUPPED-GRASP
DISK-GRASP
Fig.
IO.
A
short
session with Grasp-Exp-answering with “unknown”
would have caused Grasp-Exp
to
ask more detailed questions.
tries to form conclusions
or
diagnoses, asking questions along
the way. Fig.
10
shows a short session with Grasp-Exp.
While Grasp-Exp is still unfinished (there are currently
about
50
rules specifically involved
in
grasp choice, along
with lists of templates for object-attributes and grasp-attrib-
Utes), it seems that a total of about
100
rules would be adequate
for predicting how people will grasp parts and tools
in
a
particular environment. As discussed
in
the following section,
most of the additional rules would make Grasp-Exp more
“friendly”
in
interrogating the user and would permit more
detailed descriptions of the task and the grasped object.
A.
Lessons from Grasp-Exp
Grasp-Exp’s knowledge of the factors involved
in
human
grasp choice is far from complete. Indeed, Grasp-Exp may
never be complete, for the point of constructing Grasp-Exp
was not to create a reliable predictor
Df
how humans will grasp
tools
or
parts but to raise and clarify additional issues
in
the
study of manufacturing grasps.
In
this section, we discuss a
TRANSACTIONS ON
ROBOTICS
AND AUTOMATION,
VOL.
5,
NO.
3.
JUNE
1989
number of the issues that surfaced during the work
on
Grasp-
Exp.
First, it is necessary to quantify terms like “precision,”
“dexterity,” and “sensitivity”
so
that different grasps may be
ranked.
For
example, the left branch of the grasp taxonomy
in
Fig.
4
includes power grasps that “emphasize security,
stability.” But must a grasp manifest both security and
stability to qualify
or
will either one do? Further, suppose that
a grasp has stability and security, but also displays some
sensitivity to vibrations? After some experimentation, it was
decided that there were scales of dexterity, sensitivity, power,
and stability such that precision grasps tend to be at one end of
the spectrum and power grasps at the other. For example, the
Light Tool Grasp (Grasp
5)
is distinguished from other power
wraps primarily by the ability to sense forces and vibrations (a
characteristic of precision grasps) although it is classified as a
power grasp because the fingers and palm surround the part
and do not manipulate it. As another example, the precision
Thumb-Index Finger grasp (Grasp
9)
has some of the
characteristic stability of a power grip since the soft fingertips
partially encompass a small object.
Assigning quantities to terms like dexterity and sensitivity is
not easy. Thus it was necessary to have the expert system ask
additional questions about the force requirements, approxi-
mate object weight, :he importance of sensing vibrations at the
tool tip, and
so
forth,
so
that the relative importance of
stability and dexterity could be assessed. Often, it is easiest
to
ask such questions in terms of analogies: “Would you classify
the task as most like a prying task? a tapping
task?
a pushing
task?
,..”
A related difficulty with Grasp-Exp, common to
many expert systems
[30],
is the need to quantify subjective
terms like “heavy,” “large,” and “thin.” Clearly, it is
necessary to be consistent in using such terms; yet, people are
uncomfortable with assigning numerical values. For example,
if the system asks for the object size (small, medium, large),
how smsll is “small?” As in assigning measures of dexterity
and precision, the solution lies in asking different questions,
e.g., “is the object smaller than your fist?”
In
fact, this is the
best way to ask such questions because it is really the
relative
size
of
the object with respect to the hand that matters.
In
experimenting with Grasp-Exp
it
also became clear that
the very approximate geometric descriptions (compact, thin,
prismatic)
in
the taxonomy were too vague. These descriptions
were extended to include the
rough-object-shape
and detailed
object shape
so
that one could ask whether objects were long,
thin, disk-shaped, rectangular, and
so
forth. Although not
currently implemented, Grasp-Exp should be working with
geometric
features
of the parts and tools, stored
in
a database.
The features should not be neutral descriptions of the part
geometry (e.g., cubes, cylinders), but should emphasize
elements of the geometry that are important for grasping. Thus
a cup
or
a hammer would be described largely in terms of its
handle. With a feature-based description of objects, Grasp-
Exp would ultimately resemble rule-based planning systems
for setup and fixturing of machined parts, such as
GAR1
[7].
Feature-based descriptions of parts have also been explored
for automatic robot grasp planning
[15],
[23].
Finally, in
addition to describing the shapes and grasping features of
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
CUTKOSKY:
DESIGN
OF
HANDS
FOR
MANUFACTURING
TASKS
277
parts, Grasp-Exp should be extended to understand the details
of the part orientation with respect to the hand; for example,
the orientations of the principal axes of a tool with respect to
the fingers.
Experiments with Grasp-Exp also revealed sequences
among the grasps in the original taxonomy. As a person
proceeds with a task, the grasp shifts in response to changing
forcehorque requirements or geometric constraints. For exam-
ple, consider pulling hard
on
the handle of a wrench to loosen
a large bolt. Initially, the Hook Grasp (a subgrasp under Grasp
15)
may be used for maximum pulling force. But as the bolt
starts to loosen, and the required force
on
the handle is less
predictable, the hand switches to a Medium Wrap (Grasp
3)
or
Adducted Thumb grasp (Grasp
4),
in which the fingers entrap
the handle to keep it from slipping. As the following, slightly
more complicated example reveals, subgrasps (one level
below those shown in the taxonomy) may not have a single
parent. A person picks up a pen or a scribe and starts to mark
with it, bearing hard upon the writing surface. Upon finding
that it is not necessary to press hard, the person shifts
to
a
more standard writing grasp. As the task is completed, the
person shifts to a tripod grasp using the scribe as a pointer to
indicate the markings to a colleague. The sequence involves a
progression of grasps from Grasp
8
to Grasp
14.
Grasp
8
is
useful for picking up an object. Mobility is mostly confined to
rolling the object about its central axis. The grasp choice is
dictated largely by the need to lift the object from a flat
surface; any of the other opposed-thumb grasps would also
suffice. The Scribing Grasp involves a shift
of
the fingers to
produce larger forces and to orient the axis of the scribe
slightly more parallel to the fingers
so
that mobility in rolling
the object is reduced but mobility perpendicular to the object
axis is increased. The Writing Grasp further shifts the object
axis
so
that it becomes parallel with the fingers resulting
in
slightly greater mobility, at the expense of force. Finally, the
Tripod can be seen as a Writing Grasp
in
which the hand has
slid to one end of the object (losing support from the side of the
palm)
so
that mobility is improved in all directions, but only
small forces can be applied.
V.
DISCUSSION
In summarizing the results of the study of one-handed
manufacturing grasps, we return to the questions raised in the
introduction:
1)
Can the human grasp process be codified?
Under
limited circumstances it now appears that an expert system can
predict how people will grasp parts and tools. Moreover, in
experiments with Grasp-Exp we found that where the expert
system failed to identify the particular grasp that a person
used, it picked a close relative that could also have been used
to accomplish the task.
2)
How
accurate are the analytic models
of
grasps, with
their numerous simplifications?
Under particular circum-
stances, any one of the analytic grasp models may be a good
approximation. For example, the point-contact models are
reasonably accurate for the precision Disk and Sphere grasps,
where the contact areas are small compared to the diameter of
the grasped object.
On
the other hand, a very-soft-finger
model
[5]
more accurately approximates the Tripod and
Thumb-Index Finger precision grasps, where the finger pads
conform to, and even partially entrap the object. For the power
grasps, most of the theoretical analyses are irrelevant since the
fingers do not manipulate the part. Perhaps the best solution
for power grasps is to assume complete kinematic coupling
(with compliance) between the hand and the object, and to
assign a set of friction limitations to the grasp.
3)
How
does human grasp selection compare with the
analytic measures?
In
terms of the analytic measures dis-
cussed
in
Section
11,
the power grasps are stiffer, more stable,
and have a larger resistance to slipping than the precision
grasps. In addition, the power grasps for which clamping
is
required are form-closure grasps (assuming extra contact
wrenches due to friction). The nonclamping grasps (Grasps
15)
are force-closure, provided that external forces do not
cause the fingers to detach from the object. Finally, the power
grasps have a connectivity of
0
since the fingers do not
manipulate the part. Like the power grasps, the precision
grasps satisfy form and force-closure. However, the connec-
tivity between the grasped object and the hand
is
always at
least
3
and often
6.
Many of the detailed grasp attributes in Grasp-Exp can also
be correlated with the analytic measures. However, since the
terms that people use for describing grasps are subjective, and
depend
on
many subtle factors, the correspondence
is
rarely
exact. For example, consider the following grasp attributes
and corresponding analytic measure (see Fig.
11):
Sensitivity-A term that depends
on
many factors but is
primarily related to how accurately the fingertips can pick up
small vibrations and small changes in force and position. Thus
sensitivity is a function of grasp isotropy (if the fingers can
impart forces with accuracy then they can also measure forces
with accuracy) and stiffness (a more compliant grasp is more
sensitive to small changes
in
force).
Precision-A measure of how accurately the fingers can
impart small motions or forces to the object. Thus precision
requires light grasp forces, full manipulability, and isotropy.
Dexterity-Dexterity is similar to precision but implies
that larger motions can be imparted to the object. Thus
dexterity depends both
on
manipulability and the kinematic
work space of the hand.
Stability-When people speak of a stable grasp they
include both the definition
in
Section 11,
in
which a stable
grasp will return to its nominal position after being disturbed,
and the ability of the grasp to resist external forces without
slipping.
Security-In common use, grasp security is related to
stability, but is most closely associated with resistance to
slipping.
4)
Are the results
of
studying human grasps useful for
the design
of
robot hands?
For designing robot hands, we
need to turn from the details of human grasp choice to a
general consideration of how grasps satisfy geometric and task
requirements. For this purpose, we have found that although
the expert system contains more information and is more
flexible than the taxonomy, the taxonomy is more useful as a
design aid since it allows one to see very quickly where a set of
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
278
IEEE TRANSACTIONS ON ROBOTICS AND AUTOMATION,
VOL.
5,
NO.
3,
JUNE
1989
grasp
’
analytlc
measures
attributes
j
dexterity
nianipulabtlity
isotropy
st,ffness/compiiance
force.
farnr
c~~~~~e
1
secunty
fnction
constrants
I
Fig. 11,
Human grasp attributes in terms of analytic grasp measures.
grasps lies
in
the space of all possible grasps and to see how a
specific grasp descends from the generic grasp types.
An effective way to extend the grasp taxonomy is to
eonsider grasps
in
terms of “virtual fingers’’ that do not
necessarily have a one-to-one correspondence with fingers of
the human hand [8], [16]. Iberall [8] argues that
in
most grasps
the object is held between two virtual fingers and that the type
of opposition (e.g., trapping an object between the fingers and
the palm, or between the pads of the thumb and the index
finger) is of central importance. Iberall therefore recognizes
three basic type of grasps:
1) encompassing grasps (grasps with palm opposition)-
Grasps
1-4
and 11 are the most obvious examples of
encompassing grasps;
2) lateral grasps (grasps with side opposition)-the Lateral
Pinch, Grasp 16, is a grasp with side opposition;
3)
precision grasps (grasps with pad opposition)-the preci-
sion grasps on the right-hand side of the taxonomy
display pad opposition.
Iberall provides a table mapping the original taxonomy of
Cutkosky and Wright [4] to her categorization of grasps with
two virtual fingers. While we recognize the usefulness of
virtual fingers for generalizing the taxonomy of Fig. 4, our
own interpretation is slightly different and therefore we have
added “virtual finger” numbers to the revised taxonomy of
Fig.
4.
The Opposed-Thumb (Grasps 6-9) and Lateral Pinch
(Grasp
16)
are two-fingered grasps since there are two
independently controllable gripping surfaces. Even the Op-
posed Thumb4 Finger grasp is basically a two-fingered grasp
since the four fingers act in unison. However, the disc and
tripod grasps are more accurately thought of as grasps with
three virtual fingers since they have three independently
controllable contacts-three points define the object orienta-
tion. At the other end of the spectrum, power grasps
1-3
and
11 are difficult to describe
in
terms of virtual fingers since they
completely envelope the part with something approaching
uniform radial symmetry, but have no independent contact
areas. Finally, the nonclamping grasps (almost nongrasps)
such as the Platform and Hook grasps have one virtual finger.
It is also possible to examine industrial gripper design
in
light
of
the taxonomy
in
Fig.
4.
For the most part, today’s
commercial grippers achieve particular instances of the power
grasps on the left-hand side of Fig. 4. For example, a two-
fingered parallel-jaw gripper (Fig. 12(a)) is capable of pushing
objects (a subcategory under Grasp
15)
and of a grasp that
resembles the Lateral Pinch (Grasp
16),
in
which a small
object is clamped securely between two strong fingers.
As
(b)
Fig.
12.
Industrial grippers in terms
of
the grasp taxonomy. The typical two-
fingered gripper in
(a)
(reprinted from Com-Pick Grippers
(31)
executes the
equivalent
of
a Lateral Pinch. The pneumatic gripper in
(b)
(reprinted from
(261)
has a “palm” and is designed
for
wrap grasps. By contrast, most
dextrous robotic hands are designed for precision grasps with three
independent fingers.
another example, the commercial gripper of Fig. 12(b) is
capable of Power Wrap Grasps (Grasps 1-3).
Increasingly, general-purpose grippers are becoming inade-
quate for the variety
of
part shapes and tasks encountered in
flexible manufacturing systems. A common solution is to
provide an array of special-purpose grippers for each part
style. Although this method leads to difficulty in routing
power and sensory information from the fingers through
connections into the robot arm, and increases cycle times as
grippers are swapped, it is attractive
to
manufacturing
engineers since the grippers can be much less complicated than
a universal hand. The taxonomy in Fig.
4
suggests, however,
that if several grippers are to be used, they should be designed
for
classes
of grasps and tasks-not for different part styles.
To design a gripper for a part style is to design a tool, not a
hand. Thus like a Phillips-head screwdriver which can only be
used with Phillips-head screws, the gripper is a special-
purpose device.
A
better approach is to start with basic task requirements
and let those requirements dictate the design. For example,
one might construct a gripper for precision grasps with
opposed fingers and a second gripper for power wrap grasps.
Another possibility is to construct a hand for two types of tasks
with a single object. For example, a manufacturing hand used
for picking up small power tools and then working with them
could shift between the Opposed Thumb4 Finger grasp,
Grasp
6,
and the Light Tool grasp, Grasp
5.
Such generic
designs can be adjusted to
fit
a variety of part shapes and
finger adaptors may be used for specific constraints encoun-
tered with exceptional parts. It is also unnecessary
to
achieve
all of the different grasps
in
Fig.
4.
For example, the task
shown
in
Fig.
6
for Grasp 12 could easily be achieved with just
three fingers, as
in
Grasp 14. While
it
suits the machinist with
his human hand to bring out a full repertoire of grasps, Grasp
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
CUTKOSKY: DESIGN OF
HANDS
FOR MANUFACTURING TASKS
279
12
may be unnecessary for a robot. Similarly, the hammering
task shown in Fig.
5
for Grasp
3
could be achieved with
Grasps
2
or
4.
From such observation,
it
is expected that a
grasp taxonomy will allow the streamlining
of
hand design,
construction, and control. Thus
in
a form-follows-function
sense, robotic hands will be capable of a specified and
necessary subset of tasks
in
a small-batch manufacturing cell
but will not be overdesigned and hence overly expensive.
ACKNOWLEDGMENT
The author wishes to express his appreciation to
P.
Wright
at NYU who participated in much
of
this work and was a
source of inspiration throughout. He also wishes to thank J.
Jourdain and M. Nagurka at CMU for their assistance and
numerous suggestions
on
the grasp taxonomy and to K. Ishii at
OSU
for his assistance in developing Grasp-Exp. Mary-Jo
Dowling
of
the Robotics Institute at CMU provided the
original drawings for Fig.
4
and machinists J. Dillinger, D.
McKeel, and
S.
Klim provided their expert advise and the use
of
their hands in Figs.
5
and
6.
REFERENCES
H.
Asada, “Studies on prehension and handling by robot hands with
elastic fingers,” Ph.D. dissertation, Kyoto University, Apr. 1979.
R. C. Brost, “Automatic grasp planning in the presence
of
uncer-
tainty,”
Int.
J.
Robotics Res.,
vol. 7, no.
1,
pp. 1-17, 1988.
Compact Air Products Inc., “Com-pick grippers,” in
Catalog SC100,
Westminster, SC, 1983.
M. R. Cutkosky and P. K. Wright, “Modeling manufacturing grips and
correlations with the design of robotic hands,” presented at the IEEE
Int. Conf. on Robotics and Automation, San Francisco, CA, Apr.
1986.
~-
,
“Friction, stability and the design of robotic fingers,”
Int.
J.
Robotics Res.,
vol. 5, no. 4, pp. 20-37, 1987.
M. R. Cutkosky and I. Kao, “Computing and controlling the
compliance of a robotic grasp,”
IEEE Trans. Robotics Automat.,
vol. 5, no. 2, pp. 151-165, Apr. 1989.
Y.
Descone and J.-C. Latombe, “Making compromises among
antagonistic constraints in a planner,”
Artificial Intell.,
Dec. 1984.
T. Iberall, “The nature of human prehension: Three dextrous hands in
one,” in
Proc.
I987
IEEE Int. Conf.
on
Robotics and Automation
(Raleigh, NC, Mar. 1987), pp. 396-401.
K. Ishii, “Knowledge-based design of complex mechanical systems.”
Ph.D. dissertation, Stanford University, Stanford, CA, Oct. 1987.
J. Jameson and L.
J.
Leifer, “Automatic grasping: An optimization
approach,”
IEEE Trans. Syst., Man, Cybern.,
vol. SMC-17, no.
5,
pp. 806-813, Sept. 1987.
Z. Ji, “Dexterous hands: Optimizing grasp by design and planning,”
Ph.D. dissertation, Stanford University, Stanford, CA, 1987.
J.
Kerr and B. Roth, “Analysis
of
multifingered hands.”
Int.
J.
Robotics Res.,
vol.
4,
no. 4, pp. 3-17, 1986.
K. Lakshminarayana, “Mechanics of form closure,” ASME Tech.
Paper,
78-DET-32.
Z.
Li and
S.
Sastry, “Task oriented optimal grasping by robotic
hands,” in
Proc.
1987
IEEE
Int. Conf.
on
Robotics and Automa-
tion
(Raleigh, NC, Mar. 1987), pp. 389-394.
T. Lozano-Perez,
J.
L. Jones, E. Mazer,
P.
O’Donnell, and E. L.
Grimson, “Handey: A robot system that recognizes, plans and
manipulates,” in
Proc.
I987
IEEE Int. Conf.
on
Robotics and
Automation
(Raleigh, NC, Mar. 1987), pp. 843-849.
D. Lyons, “A simple set of grasps
for
a dextrous hand,” COINS Tech.
Rep. 85-37, Dep. Comput. Informat. Sci., Univ. of Mass., Amherst,
1985.
M. T. Mason and J. K. Salisbury,
Robot Hands and the Mechanics
of
Manipulation.
Cambridge, MA: MIT Press, 1985.
M. E. Merchant, “World trends in flexible manufacturing systems,”
The FMS Magazine, vol.
1,
no.
1,
pp. 4-5, Oct. 1985.
Y. Nakaniura,
K.
Nagai, and T. Yoshikawa, “Mechanics
of
coordinate
manipulation by multiple robotic mechanisms,”
in
Proc.
I987
IEEE
Int.
Conf.
on
Robotics and Automation
(Raleigh, NC, Mar. 1987),
pp. 991-998.
J. Napier, “The prehensile movements of the human hand,”
J.
Bone
Joint Surg.,
vol. 38B, no. 4, pp. 902-913, Nov. 1956.
V.-D. Nguyen, “Achieving stable grasps in 3D,” in
Proc. 1987IEEE
Int.
Conf.
on
Robotics andAutomation
(Raleigh, NC, Mar. 1987),
pp. 240-245.
M.
S.
Ohwovoriole and B. Roth, “An extension of screw theory,’’
J.
Mech. Des.,
vol.
103,
pp. 725-735, 1981.
R.
J.
Popplestone, “The Edinburgh designer system as
a
framework
for
robotics,” presented at the 1987 IEEE Int. Conf. on Robotics and
Automation, Raleigh, NC, Mar. 1987.
N. D. Ring and D. B. Welbourn, “A self-adaptive gripping device: Its
design and performance.”
Proc.
Inst.
Mech. Eng.,
vol.
183,
pp. 45-
G. Schlesinger, “Der Mechanische Aufbau der Kunstlichen Glieder,”
in
Ersatzglieder
und
Arbeitshilfen,
M. Borchardt
et al.,
Eds.
Berlin, Germany: Springer, 1919.
Simrit Corp.,
Pneumatically Operating Gripping and Clamping
Elements,
Arlington Hts., IL. 1984.
C.
L.
Taylor and R. J. Schwarz, “The anatomy and mechanics of the
human hand,”
Artificial Limbs,
vol.
2,
pp. 22-35, May, 1955.
R. Tomovic, G. A. Bekey, and W. A. Karplus, “A strategy for grasp
synthesis for multifingered robot hands.” in
Proc.
I987
IEEE Int.
Conf.
on
Robotics
and
Automation
(Raleigh. NC, Mar. 1987). pp.
G. Westling and R.
S.
Johansson,
“Factors influencing the force
control during precision grip,”
Exper. Brain Res.,
vol. 53, pp. 277-
284, 1984.
A. C. Zimmer,
“A
model for the interpretation
of
verbal predictions,”
in
Developments
in
Expert Systems,
M.
J.
Coombs, Ed. New York:
Academic Press, 1984, pp. 235-248.
49, 1968-1969.
83-89.
Mark
R.
Cutkosky,
for a photograph and biography, please see page 165 of
the April 1989 issue of this TRANSACTIONS.
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 20, 2009 at 12:34 from IEEE Xplore. Restrictions apply.
