Scientia Iranica B (2011) 18 (2), 222230

Sharif University of Technology

Scientia Iranica
Transactions B: Mechanical Engineering
www.sciencedirect.com
Multi-aspect grasp index for robotic arms
F. Cheraghpour, S.A.A. Moosavian

, A. Nahvi
Department of Mechanical Engineering, K. N. Toosi University of Technology, Tehran, P.O. Box 19395-1999, Iran
Received 24 May 2010; revised 7 November 2010; accepted 11 December 2010
KEYWORDS
Object manipulation;
Grasp planning;
Performance index;
Cooperative manipulator;
Manipulability;
Optimization.
Abstract In this paper, a new Multi-Aspect Grasp (MAG) performance index is presented for evaluating
grasp quality during an object manipulation task. The position of grasp points, the configuration of
cooperative manipulators, and the kinetic aspects of manipulating arms and the grasped object are
considered in the MAG index. The MAG index is used to evaluate the candidate points to choose the best
grasp point and to select the most effective branch of the inverse kinematics solution, with respect to
the given task. Simulation results, which are validated with analytical solutions, show the merits of the
proposed index. According to these results, the MAG index indicates that in planar object manipulation
tasks without rotation, the best grasppoint is the object center of the mass, whichis physically meaningful.
2011 Sharif University of Technology. Production and hosting by Elsevier B.V. All rights reserved.
1. Introduction
Object manipulation is an interesting field in robotics
that involves many aspects. Grasp planning is one of the
most challenging and important problems. The cooperative
manipulation of objects provides versatility in task execution,
by allowing many alternatives in task mechanics. In particular,
multiple-arm cooperation is required if we are to assemble and
manipulate parts without the aid of fixtures or jigs, or if we
are to safely transfer heavy and large objects from one place
to another [15]. When an object is grasped by a system of
manipulators or a multi-fingered robotic hand, there are several
ways to find suitable grasps. Optimal grasp planning is required
to find the optimal grasp points for satisfying the objectives
of the given task. Also when there are several configurations
for a manipulator to reach a point within its workspace, a
criterion is required to select the appropriate branch of the
inverse kinematics solution.
Several performance indices have been defined to evaluate
the quality of grasp configuration. These measures are classi-

Corresponding author.
E-mail address: moosavian@kntu.ac.ir (S.A.A. Moosavian).
fied into three groups [6,7]. In the first group, indices consider
the position of contact points. Some of these indices are based
on the algebraic properties of the grasp matrix, G, such as the
smallest singular value and Grasp Isotropy Index [8], the deter-
minant [9] and the condition number [10]. Some indices are
based on the geometric relationships between the position of
contact points, such as the area of the grasp polygon [11,12],
the distance between the centroid of the contact polygon and
the center of mass of the grasped object [1315], and the stabil-
ity grasp index [1618]. Some indices consider limitations on
the End-Effector (EE) forces, such as indices based on the grasp
force distribution [19], indices based on a minmax optimiza-
tion in the Force Workspace (FW) [20], the total volume of the
resulting wrench space as a quality metric [21], the radius of
the largest wrench space ball [22,23], the optimal force-closure
graspbasedonthe Q-distance measure [24,25], the ability of the
grasp to reject disturbance forces [26,27] and, finally, the min-
imum volume bounding box [28]. The indices that considered
this aspect could analyze the quality of the grasp configuration,
statically.
In the second group, grasp indices consider kinematic char-
acteristics and configuration of the cooperative manipulators or
fingers of a multi-fingered robotic hand, such as the smallest
singular value of the Jacobian matrix [29], the condition num-
ber of the Jacobianmatrix [30], the volume of the manipulability
ellipsoid or Manipulability Index [3133], the Mobility Index [34]
and the Observability Index [35]. The indices that consider this
aspect couldonly evaluate the quality of the graspconfiguration
for kinematic tasks without force interaction with the environ-
ment.
In the third group, the indices consider the kinetic charac-
teristics of the manipulators and the grasped object, such as the
1026-3098 2011 Sharif University of Technology. Production and hosting by
Elsevier B.V. All rights reserved. Peer review under responsibility of Sharif
University of Technology.
doi:10.1016/j.scient.2011.03.017
F. Cheraghpour et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 222230 223
Figure 1: The relationship between the velocities in a grasping system.
Task Compatibility Index [36], performance analysis with the Dy-
namic Capability Equations [37,38], the use of energy consump-
tion as a cost function [39], the squared norm of the actuator
torque vector as a cost function [40], the volume of the dynamic
manipulability ellipsoid or Dynamic Manipulability [41,42] and
the Inertia Matching Ellipsoid (IME) that integrates the existing
dynamic manipulability and manipulating-force ellipsoids [43].
These indices could be used for the tasks that consist of force in-
teraction between the robotic systemand the environment, and
the grasped object with the robotic manipulators.
If the gripper grasps an object in an unsuitable condition,
the manipulator cannot perform the task efficiently. This may
occur in the manipulator kinematic singular condition or in a
saturation state of the actuators. In such conditions, the grasp
point should be changed with a regrasping strategy and a new
grasppoint shouldbe selected[44,45]. Althoughsome multiple-
aspect indices have been proposed for path planning [46,47],
motion planning [48,49] and trajectory planning [50], the
definition of such indices for grasp planning has not yet been
done.
In this paper, a new grasp performance index is proposed
for evaluation of the grasp configuration to perform an object
manipulation task. This index considers three aspects: The
first is the position of the grasp points, the second is the
kinematic characteristics of manipulators, and the third is the
corresponding kinetic characteristics. The goal is to evaluate
grasp configuration for a system of robotic arms that should
perform a task with suitable dexterity and minimum energy
consumption, individually or cooperatively.
2. Multiple-Aspect Grasp index
As mentioned above, the Multiple-Aspect Grasp (MAG)
index can be used to evaluate grasp configuration in object
manipulation tasks and consists of various static, kinematic and
kinetic aspects. Cooperative object manipulation tasks can also
be evaluated with the MAG Index. In this section, the proposed
index is developed.
A grasp matrix maps the object velocity into grasp point
velocity, as shown in Figure 1 [9]. The V
p
denotes the velocity of
grasp points and V
CM
denotes the velocity of the center of mass
of the object. Thus the grasp matrix is defined as:
V
p
= G
T
V
CM
. (1)
The grasp matrix for n grasp points will have the following
structure [51]:
G =
_
G
1
G
2
G
j
G
n
_
66n
,
G
j
=
_
1
3
.
.
.
.
.
.
0
3
. . . . . . . . . . . . . . . . . . . . . . .
S
T
obj
r

pj
.
.
.
.
S
T
obj
_
66
, (2)
where 1
3
and 0
3
are the unit and zero 33 matrix, respectively,
r
pj
is the position vector of the jth end-effector with respect to
Figure 2: Coordinate frames and the position vector of the jth grasp point with
respect to the C.G. of the object.
the object center of mass, as shown in Figure 2, and r

pj
is the
cross operator of vector r
pj
defined as:
r

pj
=
_
0 r
zj
r
yj
r
zj
0 r
xj
r
yj
r
xj
0
_
33
. (3)
Matrix S
obj
is the map between the time derivative of the Euler
angles array and the velocity vector of the object and is defined
as:

obj
= S
obj

obj
. (4)
To evaluate the grasp condition, we need to determine the
condition number of the grasp matrix. The grasp matrix, G, is
not dimensionally homogeneous, because the first row entries
of G in Eq. (3), are dimensionless, but entries in the second
row have a dimension of length. Some methods to resolve this
dimensional inconsistency were proposed, such as Scaling [29],
and Characteristic Length [52]. Here, we define the normalized
grasp matrix

G, which is a homogenous grasp matrix:

G =
_

G
1

G
2


G
j


G
n
_
66n
,

G
j
=
_

_
G
j
if r
pj
= 0
_
r
pj
1
3
.
.
.
0
3
. . . . . . . . . . . . . . .
0
3
.
.
.
1
3
_
G
j
_
r
pj

1
1
3
.
.
.
0
3
. . . . . . . . . . . . . . . . . . . . . . . . . . .
0
3
.
.
.
r
pj

1
1
3
_
if r
pj
= 0
(5)
where r
pj
is the normof r
pj
and G
j
was defined in Eq. (2). Note
that when r
pj
= 0, the grasp point is the center of gravity
and the terms in the second row of G are normalized. When
r
pj
= 0, the pre and post multiplying matrices will make the
second row elements of G
j
dimensionless. To show this and to
focus on the dimensions, we may substitute S
obj
= 1
3
in G
j
:

G
j
=
_
r
pj
1
3
.
.
.
0
3
. . . . . . . . . . . . . . .
0
3
.
.
.
1
3
__
1
3
.
.
.
0
3
. . . . . . . . . . .
r
pj

.
.
.
1
3
_

_
r
pj

1
1
3
.
.
.
0
3
. . . . . . . . . . . . . . . . . . . . . . . . . . .
0
3
.
.
.
r
pj

1
1
3
_
=
_
r
pj
1
3
.
.
.
0
3
. . . . . . . . . . . . . . .
0
3
.
.
.
1
3
__
r
pj

1
1
3
.
.
.
0
3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
r
pj

1
r
pj

.
.
.
r
pj

1
1
3
_
=
_
1
3
.
.
.
0
3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
r
pj

1
r
pj

.
.
.
r
pj

1
1
3
_
. (6)
224 F. Cheraghpour et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 222230
Therefore, to evaluate the graspcondition, the inverse condition
number of the normalized grasp matrix

G is considered:
C
N
=

min
(

G)

max
(

G)
, (7)
where
min
(

G) and
max
(

G) are the minimum and maximum

singular values of the normalized grasp matrix,

G, respectively.
If C
N
is close to zero, the grasp is singular. This means that
at least one of the cooperating manipulators is incapable of
moving the object in a direction. On the other hand, if C
N
is close
to 1, we have an isotropic grasp matrix, and all end-effectors are
capable of moving the object in any arbitrary direction. Next, D
i
denotes the dexterity measure of the ith manipulator to grasp
the object, as:
D
i
=
_
det(J

i
J
T
i
), (8)
where J

i
denotes the Jacobian matrix between the actuators
velocity domain and the object velocity domain of the ith
manipulator, which is computed as follows:
J

i
= (G
T
)
#
J
i
, (9)
where J
i
denotes the Jacobian matrix of the ith manipulator
and also (.)
#
is the pseudo-inverse of the matrix. The term D
i
denotes the volume of the manipulability ellipsoid. The greater
the volume of this ellipsoid, the greater is the dexterity of
the grasp. When this ellipsoid becomes a sphere, its volume
is maximized. When the ith manipulator is in a singular
configuration, the determinant of J
i
approaches zero. Note that
D
i max
is the maximum value of D
i
among all candidate grasp
points. If D
i
= D
i max
, the term D
i
/D
i max
becomes one, which
results in the most dexterous grasp. Finally, this definition of
D
i
allows us to use the index for manipulators with non-square
Jacobian matrices, e.g. redundant manipulators.
Finally, suppose that P
i
denotes the consumed power of
actuators for the ith manipulator:

P
i
dt =

_
|

T
i
|.|

i
|
_
dt, (10)
where
i
denotes the joints angle array, and

i
denotes the
corresponding actuator torque array of the ith manipulator
with an augmented mass matrix. The augmented mass matrix
is the manipulator mass matrix where the object inertial
parameters are added to the inertial parameters of the last
link. In a cooperation condition, suppose that the inertial
parameters of the grasped object are distributed into all
cooperative manipulators with equal portions. Note that P
i max
is the maximum value of P
i
among the candidate grasp points.
Our desired condition is that the term P
i
/P
i max
is smaller and,
therefore, the term 1 (P
i
/P
i max
) is near to 1. In a weak
condition, where P
i
= P
i max
, the actuators of the manipulator
consume maximum power and, so, are near to the saturation
limit.
The newgraspperformance index, MAG, considers the above
three characteristics of grasp in a weighted format. Therefore,
the MAG index evaluates grasp configuration during the task
execution for cooperating arms:
MAG =
1
t
f
t
0

t
f
t
0
_
W
1
C
N
+ W
2
1
n
n

i=1
_
D
i
D
i max
_
+W
3
1
n
n

i=1
_
1
P
i
P
i max
_
_
dt, (11)
Table 1: Choosing the weighting factors.
Type of the Task W
1
W
2
W
3
Default choice 0.33 0.33 0.34
Type-I: static tasks 0.50 0.25 0.25
Type-II: kinematic tasks 0.25 0.50 0.25
Type-III: dynamic tasks 0.25 0.25 0.50
Table 2: The geometric and inertial parameters of the manipulator.
Link
no.
L (m) m(kg) I
xx
(kg m
2
)
I
yy
(kg m
2
)
I
zz
(kg m
2
)
Motor
inertia
Gear
ratio
1 1.04 17.4 0.130 0.524 0.539 210
4
100
2 1.04 17.4 0.130 0.524 0.539 210
4
100
3 0.92 6.1 0.0154 0.212 0.192 210
4
100
where t
0
and t
f
are the initial and final times, respectively. The
first termof MAGin Eq. (11) deals with the position of the grasp
points, which considers the static characteristics of the grasp,
the second term of MAG in Eq. (11) considers the kinematics
characteristics of the cooperative manipulators, and the third
term evaluates the kinetic characteristics of the manipulators
and the grasped object. Each termof the MAG index has a value
between zero and 1. The goal is to maximize the MAG index.
Also, weighting factors, W
1
, W
2
and W
3
, are included to put
a different emphasis on each term, while the total sum of these
three factors should be equal to 1. As a rule of thumb, the
weighting factors can be chosen according to Table 1.
As shown in Table 1, the tasks of Type-I involve static tasks
like holding an object, or quasi-static tasks like those with very
slow velocity. The tasks of Type-II involve kinematic tasks like
painting, arc welding, etc. where the end-effector and grasped
object do not have force interaction with the environment. The
tasks of Type-III involve dynamic tasks like moving an object
with high velocity, in particular, for mobile robots that have an
onboard limited power.
Finally, the normalizing factor, 1/(t
f
t
0
), puts the total
index between 0 and 1. Therefore, a good grasp has a MAGindex
of 1, and a poor grasp has a MAG index of zero.
3. Using MAG for 3R manipulator: a case study
In this section, the details of formulation will be presented
for a system that includes a 3R manipulator and a grasped
object. Figure 3 shows a 3R robotic manipulator in global
motion, performing an object manipulation task. The task is
moving the object on a given trajectory in the plane. The object
has been grasped with a solid grasp condition, i.e. its orientation
cannot change with respect to the end-effector of the robot.
The geometric and inertial parameters of the manipulator are
shown in Table 2.
3.1. The object desired path
The object path is a straight line along the x-axis, coordinate
y is constant and coordinate z is zero. Also, the orientation of
the object is fixed at 10 during the task.
The trajectories of the joint angles are quintic functions as
follows:
x(t) = a
0
+ a
1
t + a
2
t
2
+ a
3
t
3
+ a
4
t
4
+ a
5
t
5
,
y(t) = 1, (t) = 10,
(12)
F. Cheraghpour et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 222230 225
Figure 3: A 3R arm is performing an object manipulation task.
Figure 4: The defined path of the grasped object in the workspace and
trajectory of x(t) and its derivatives.
where the coefficients are presented in [53]. At the initial time,
the object is at rest in x
0
= 0. Also at the final time, the object is
at rest in x
f
= 1.5 (m). Figure 4 shows the defined path in the
workspace, and the trajectory of x(t) and its derivatives.
3.2. Calculation of MAG terms
For calculating various terms of the MAG index in Eq. (11),
we consider the velocity of the grasped object in two ways.
First, with the transformation of the actuators velocity and,
then, with the task predefined trajectory. The equality of these
relations gives us the configuration of the manipulator to reach
the defined task. The velocity of the grasp points is computed
from the velocity of the actuators as follows:
V
p
= J

, (13)
where V
p
denotes the velocity array of the grasp points on the
grasped object as follows:
V
p
=
_
x
p
y
p
z
p

x

y

z
_
T
. (14)
x
p
, y
p
and z
p
are the linear components and
x
,
y
, and
z
are the angular components of the grasp point velocity array.
The V
CM
denotes the object velocity array containing the linear
velocity and time derivative of the Eulers angles array as
follows:
V
CM
=
_
x
obj
y
obj
z
obj

_
T
. (15)
For relating the time derivative of the Euler angles and angular
velocity vector as in Eq. (4), matrix S
obj
R
33
can be expressed
in terms of the Euler angles as:
S
obj
=
_
1 0 S

0 C

0 S

_
, (16)
where C

and S

denotes cos() and sin(), respectively, and

and are the pitch and roll Euler angles. For a single grasp point,
the grasp matrix is computed as:
G =
_
_
_
_
1
33
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
0 r
z
r
z
C

+ r
y
S

r
x
S

r
z
C

r
y
C

r
z
S

+ r
x
C

.
.
. 0
33
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
r
y
r
x
C

r
y
S

r
x
C

.
.
.
.
.
.
.
.
.
.
1
0
S

0
C

0
S

_
. (17)
3.3. Planar case
For planar rigid grasping, the relationship between the
velocity components of the grasp point and the velocity of the
center of mass of the object is as follows:

z
obj
=
z
,
_
x
obj
y
obj
_
=
_
x
c
y
c
_
+
_

k
_

_
r
x
r
y
_
, (18)
where
z
obj
is the angular velocity of the center of mass of
the object about the z-axis. The Jacobian matrix of the 3R
manipulator is as follows:
J =
_
_
(L
1
S
1
+ L
2
S
12
+ L
3
S
123
) (L
2
S
12
L
3
S
123
) L
3
S
123
L
1
C
1
+ L
2
C
12
+ L
3
C
123
L
2
C
12
+ L
3
C
123
L
3
C
123
1 1 1
_
_
,
(19)
where S
12
denotes sin(
1
+
2
) and so on. L
i
denotes the length
of the ith link. Moreover, for the planar case, the grasp matrix is
computed as:
G =
_
1 0 0
0 1 0
r
y
r
x
1
_
. (20)
The solution for the inverse kinematics of the considered 3R
manipulator is discussed in the Appendix. Next, the terms of
the MAG index for each candidate grasp point in each branch of
the inverse kinematics solution, can be computed.
3.4. Various grasped objects
The first grasped object is a nonsymmetrical, L-Shape,
1.5 m0.9 mblock. The secondobject is a rectangular, 1.25 m
0.5 m block. And the third object is a 2 m 0.4 m bar. The
geometric and inertial parameters of the grasped objects are
presented in Table 3. The numbering of the candidate grasp
points on each object is shown in Figure 5.
226 F. Cheraghpour et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 222230
Figure 5: Numbering of grasp points. (a) The L-shape object; (b) the rectangle;
and (c) the bar.
Table 3: The geometric and inertial parameters of the grasped objects.
Object m
(kg)

(kg/m
3
)
Thickness
(m)

I
xx
(kg m
2
)

I
yy
(kg m
2
)

I
zz
(kg m
2
)
L-shape 23.7 7200 0.045 1.135 4.701 5.793
Rectangle 22.27 7200 0.045 0.5651 2.9035 3.4611
Bar 19.44 7200 0.045 0.1491 6.4833 6.6258
4. Obtained results and discussions
The values of the MAG index for candidate grasp points on
the object, which are calculated based on Eq. (11), are shown in
Figure 6. These values are plotted by MATLAB
TM
as the Object
Grasping Map. In this map, different values of the MAG index
have different colors and equal values located on the same
contour. The values of the MAG index of the interior points are
interpolated linearly. Figure 6 shows the Object Grasping Map
of the single manipulator, for grasping three different objects.
The weighting factors are selected equal to each other, W
1
=
W
2
= W
3
. These maps indicate that the points near to the object
center of mass have greater values of the MAG index.
Figure 7 shows the comparison between four grasp indices,
the grasp isotropy index (C-index), the normalized manipu-
lability index (D-index), the energy index as a cost function
of grasp evaluation (R-index), and the MAG index, for three
grasped objects (i.e. the L-shape, the rectangle and the bar).
Figure 6: The object grasp quality map for a single manipulator with
equal weighting factors. (a) L-shape object in negative branch of the inverse
kinematics solution; (b) rectangle in a positive branch of the inverse kinematics
solution; and (c) bar in a positive branch of the inverse kinematics solution.
As shown in Figure 7a, because of the static nature of the C-
index, it could not distinguish the quality of grasping between
two branches of the inverse kinematics solutions. The weakness
of the R-index, with respect to the MAG index, is more obvi-
ous at specific points, e.g. point No. 3 in Figure 7a, point No. 54
in Figure 7b and point No. 29 in Figure 7c. Since the value of
the R-index means that the grasp configuration at these points
leads to less energy consumption by the joint actuators, these
points, in the R-index diagram, are introduced as better points
among the other candidate points. But the C-index and D-index
diagrams show that these points are one of the weakest can-
didate points, because the grasp with these points could not
transform the velocity of the actuators to the center of mass
of the object effectively. Also, the configuration of the manip-
ulator, when grasping the object at these points, would not be
suitable considering the required dexterity for performing the
task. However, the MAG index could distinguish this condition
clearly and evaluates the performance of the system in a multi-
aspect manner.
The effect of weighting factors for the L-shape graspedobject
is showninFigure 8. As showninthis figure, the best grasppoint
is different, depending on the importance of each aspect that
corresponds to the weighted factor. Similar results are obtained
for the other grasped objects.
5. Results validation
To validate the obtained simulation results, we will compare
them with an analytical solution to the problem. For this
purpose, we should find the grasp point that maximizes the
MAG index. Therefore, we find the condition in which three
terms of MAG in Eq. (11) are maximized. First, we present a
proposition as follows.
F. Cheraghpour et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 222230 227
Figure 7: Comparison between four grasp indexes: the MAG index with equal
weighting factors, the grasp isotropy index (C), the normalized manipulability
index (D), and the energy index as a cost function of grasp evaluation (R) for
a single manipulator. (a) L-shaped object in a negative branch of the inverse
kinematics solution; (b) rectangle in a positive branch of the inverse kinematics
solution; and (c) bar in a positive branch of the inverse kinematics solution.
Proposition 1. The reciprocal of the condition number of matrix
A becomes maximum if:

max
(A) =
min
(A), (21)
where
max
(A) and
min
(A) are maximum and minimum singular
values of A, respectively [54].
For the grasp matrix G, which is presented in Eq. (20) for
the planar case study, with symbolic calculations of MATLAB
TM
and using the SVD(G) command, we would have singular values
Figure 8: The object grasp quality map for L-shaped object with different
weighting factors. (a) W
1
= 0.8, W
2
= 0.1, W
3
= 0.1; (b) W
1
= 0.1, W
2
= 0.8,
W
3
= 0.1; and (c) W
1
= 0.1, W
2
= 0.1, W
3
= 0.8.
of G, as the function of grasp point coordinates, r
x
and r
y
, with
respect to the object center of mass as:

1
= 1,

2
=

2
2
_
2 + (r
2
x
+ r
2
y
) +
_
4(r
2
x
+ r
2
y
) + (r
2
x
+ r
2
y
)
2
,

3
=

2
2
_
2 + (r
2
x
+ r
2
y
)
_
4(r
2
x
+ r
2
y
) + (r
2
x
+ r
2
y
)
2
. (22)
Introducing r = r
2
x
+ r
2
y
, we have:

1
= 1,

2
=

2
2
_
2 + r +
_
4r + r
2
,

3
=

2
2
_
2 + r
_
4r + r
2
. (23)
228 F. Cheraghpour et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 222230
According to Proposition 1, C
N
is maximum if
max
=
min
,
therefore:
_
4r + r
2
= 0 r
2
= 4r. (24)
For r 0, this leads to r = 0. It means that
1
=
2
=

3
= 1 and the grasp point is the grasped object center of mass.
Therefore, Proposition 2 will be concluded.
Proposition 2. For planar cases, if the object center of mass is
selected as the candidate grasp point, then the reciprocal of the
condition number of grasp matrix G, C
N
(G), becomes its maximum,
which is equal to one.
For the second term of Eq. (11), we should calculate the
maximum value of D in Eq. (8) as a function of grasp point
coordinates. Based on Eqs. (8) and (9), we should calculate the
maximum of the term below:
det
_
J

J
T
_
= det
_
J

_
det
_
J
T
_
= det
_
_
G
T
_
#
J
_
det
_
_
_
G
T
_
#
J
_
T
_
= det
_
_
G
T
_
#
_
det (J) det
_
J
T
_
det
_
_
_
G
T
_
#
_
T
_
= det
_
_
G
T
_
#
_
det (J) det
_
J
T
_
det
_
_
_
G
#
_
T
_
T
_
= det
_
_
G
T
_
#
_
det (J) det
_
J
T
_
det
_
G
#
_
. (25)
For grasp matrix, G, which is presented in Eq. (20) for a planar
case, we have:
det(G) = det(G
#
) = 1. (26)
Therefore:
max
_
det
_
J

J
T
__
= max
_
det
_
JJ
T
__
. (27)
This shows that in the planar case, the term max{det(JJ
T
)} does
not depend on the grasp matrix and corresponding grasp point
coordinates.
For the third term of Eq. (11), we should calculate the
minimum value of P in Eq. (10) as a function of grasp point
coordinates. To this end, the object equations of motion can be
considered:
_
M
obj

X
G
+ F

= F
e
I
G
+ I
G
= r
e
F
e
(28)
where M
obj
and I
G
are the mass and moment of inertia of the
object, respectively, F

and F
e
are nonlinear terms and the
force of the end effector, is the angular velocity of the object,
and r
e
is the position vector of the grasp point, with respect
to the object center of mass. The first equation governs the
linear motion of the object, and the second one shows the
angular motion of the object. In this system of equations, the
term r
e
is the only term that is affected by the grasp point.
In the task defined in this case study the object orientation
remains constant during the task. Therefore, to minimize the
consuming energy in the manipulator actuators, the moment
that is generated by the object on the end effectors must be
minimized and, so, r
e
= 0.
Therefore, in the case study task, all three terms of MAG in
Eq. (11) are maximized when the manipulator grasps the object
at its center of mass, which was obtained before in Figure 5.
6. Conclusions
In this paper, a new performance index was proposed as the
Multiple-Aspect Grasp (MAG) index, which could be efficiently
used for grasp planning in object manipulation tasks. This in-
dex considers three main aspects of grasp planning for object
manipulation tasks, i.e. grasp points, grasp dexterity, and the
power consumption of the manipulator actuators. Also, the co-
operation of the manipulators can be considered. The proposed
MAG index was used to find the best grasp point in a planar
task. Results show that the object center of mass is the best
grasp point, which maximizes the MAG index. Simulation re-
sults were discussedandvalidatedwithanalytical solutions. For
future work, this index is to be applied to cooperative tasks and
mobile robotic manipulations.
Appendix. Inverse kinematics and augmented mass matrix
computations
For computing the inverse kinematics solution of the manip-
ulator, first, we compute the transformation matrix for the base
to end effector, as follows:
0
T
3
=
_
_
_
C
123
S
123
0 L
1
C
1
+ L
2
C
12
+ L
3
C
123
S
123
C
123
0 L
1
S
1
+ L
2
S
12
+ L
3
S
123
0 0 1 0
0 0 0 1
_

_
=
0
T
E
. (A.1)
On the other hand, if we compute the transformation for the
object frame to the base frame, we have:
0
T
obj
=
_
_
_
C

0 x
obj
S

0 y
obj
0 0 1 0
0 0 0 1
_

_. (A.2)
Substituting the object center of mass coordinate, with respect
to the end-effector, the transformation of end-effector to the
base frame is computed, as follows:
0
T
E
=
_
_
_
_
C

0
.
.
.
.
.
.
.
.
.
.
r
x
C

r
y
S

+ x
obj
S

0 r
x
S

r
y
C

+ y
obj
0 0 1 0
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
0 0 0
.
.
.
.
1
_

_
. (A.3)
Equating Eqs. (A.1) and (A.3), a nonlinear system of equa-
tions for computing the inverse kinematics solutionis obtained:
_

_
C

= C
123
S

= S
123
r
x
C

r
y
S

+ x
obj
= L
1
C
1
+ L
2
C
12
+ L
3
C
123
r
x
S

+ r
y
C

+ y
obj
= L
1
S
1
+ L
2
S
12
+ L
3
S
123
.
(A.4)
Solving Eq. (A.4),
1
,
2
and
3
are obtained:

1
= A tan 2
_

Y,

X
_
A tan 2
__

X
2
+

Y
2
c
2
, c
_
, (A.5)

2
= (1/L
2
)A tan 2(

Y L
1
S
1
,

X L
1
C
1
)
1
, (A.6)

3
= A tan 2(S

, C

) (
1
+
2
) (A.7)
where

X = X
obj
L
3
C

, and

Y = Y
obj
L
3
S

, also, c =
(1/2L
1
)(

X
2
+

Y
2
+ L
2
1
+ L
2
2
). We name the solution with the
positive sign in Eq. (A.5) as the positive branch, and the solution
with the negative sign as the negative branch.
F. Cheraghpour et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 222230 229
For the power term in Eq. (11), we must compute the
augmented mass matrix of the manipulator, as follows:
M

=
_
h
11
h
12
h
13
h
21
h
22
h
23
h
31
h
32
h
33
_
33
. (A.8)
The values of the elements of the augmented mass matrix of the
3R planar manipulator are computed as follows:
h
11
= I
1
+ m
1
r
2
1
+ m
2
_
L
2
1
+ r
2
2
_
+ 2m
2
L
1
r
2
C
2
+ I
2
+m
3
_
L
2
1
+ L
2
2
+ r
2
3
+ 2L
1
L
2
C
2
+ 2L
1
r
3
C
23
+ 2L
2
r
3
C
2
) + I
3
,
h
12
= h
21
= m
2
_
r
2
2
+ L
1
r
2
C
2
_
+ I
2
+m
3
_
L
2
2
+ r
2
3
+ L
1
r
3
C
23
+ L
1
L
2
C
2
+ 2L
2
r
3
C
3
_
+ I
3
,
h
13
= h
31
= m
3
(r
2
3
+ L
1
r
3
C
23
+ L
2
r
3
C
3
) + I
3
,
h
22
= m
2
r
2
3
+ m
3
(L
2
2
+ r
2
3
+ 2L
2
r
3
C
2
) + I
2
+ I
3
,
h
23
= h
32
= m
3
(r
2
3
+ L
2
r
3
C
3
) + I
3
,
h
33
= m
3
r
2
3
+ I
3
.
References
[1] Moosavian, S.A.A. and Papadopoulos, E. Cooperative object manipulation
with contact impact using multiple impedance control, International
Journal of Control, Automation, and Systems, 8(2), pp. 314327 (2010).
[2] Eslamy, M. and Moosavian, S.A.A. Dynamic and cooperative object
manipulation control of suspended mobile manipulators, Journal of
Intelligent and Robotic Systems, 60, pp. 181199 (2010).
[3] Moosavian, S.A.A. and Rastegari, R. Multiple-armspace free-flying robots
for manipulating objects with force tracking restrictions, Journal of
Robotics and Autonomous Systems, 54, pp. 779788 (2006).
[4] Masoud, A. Kinodynamic motion planning, IEEE Robotics & Automation
Magazine, 17(1), pp. 8599 (2010).
[5] Korayem, M.H., Haghighi, R., Nikoobin, A., Alamdari, A. and Korayem, A.H.
Determining maximum load carrying capacity of flexible link manip-
ulators, Scientia Iranica, Transactions B: Mechanical Engineering, 16(5),
pp. 440450 (2009).
[6] Suarez, R., Roa, M. and Cornella, J. Grasp Quality Measures, Institute
of Industrial and Control Engineering, Technical University of Catalonia,
Tech. Rep. (2006).
[7] Cutkosky, M.R., Howe, R.D. and Provancher, W.R. Force and tactile
sensors, In Springer Handbook of Robotics, B. Siciliano and O. Khatib, Eds.,
pp. 465466, Springer-Verlag, Berlin Heidelberg (2008).
[8] Kim, B., Oh, S., Yi, B and Suh, I.H. Optimal grasping based on non-
dimensionalized performance indices, Proceeding of IEEE/RSJ Int. Conf. on
Intelligent Robots and Systems, pp. 949956 (2001).
[9] Murray, R.M., Li, Z. and Sastry, S., A Mathematical Introduction to Robotic
Manipulation, CRC Press, Boca Raton, Florida (1994).
[10] Shimoga, K.B. Robot grasp synthesis algorithms: a survey, International
Journal on Robotics Research, 15(3), pp. 230266 (1996).
[11] Chinellato, E., Fisher, R.B., Morales, A. and del Pobil, A.P. Ranking planar
grasp configurations for a three-finger hand, Proc. IEEE Int. Conf. Robotics
and Automation, pp. 11331138 (2003).
[12] Popovic, M., Kraft, D., Bodenhagen, L., Baseski, E., Pugeault, N., Kragic, D.,
Asfour, T. and Kruger, N. A strategy for grasping unknown objects
based on co-planarity and colour information, Robotics and Autonomous
Systems, 58(5), pp. 551565 (2010).
[13] Chinellato, E., Morales, A., Fisher, R.B. and del Pobil, A.P. Visual quality
measures for characterizing planar robot grasps, IEEE Transactions on
Systems, Man and Cybernetics-Part C: Applications and Reviews, 35(1),
pp. 3041 (2005).
[14] Speth, J., Morales, A. and Sanz, P.J. Vision-based grasp planning of
3D objects by extending 2D contour based algorithms, Proc. IEEE/RSJ
International Conference on Intelligent Robots and Systems, pp. 22402245
(2008).
[15] Ding, D., Liu, Y.H. and Wang, B. Computation of 3-Dform-closure grasps,
IEEE Transactions on Robotics and Automation, 17(4), pp. 515522 (2001).
[16] Park, Y.C. and Starr, G.P. Grasp synthesis of polygonal objects using a
three-fingered robotic hand, International Journal on Robotic Research,
11(3), pp. 163184 (1992).
[17] Xiong, C. and Xiong, Y. Stability index and contact configuration planning
for multifingered grasp, Journal of Robotic Systems, 15, pp. 183190
(1998).
[18] Xiong, Y. and Xiong, X. Algebraic structure and geometric interpretation
of rigid complex fixture systems, IEEE Transactions on Automation Science
and Engineering, 4(2), pp. 252264 (2007).
[19] Dai, J.S. and Kerr, D.R. Analysis of force distribution in grasps using aug-
mentation, Journal of Mechanical Engineering Science, 210(1), pp. 1522
(1996).
[20] Papadopoulos, E. and Gonthier, Y. A framework for large-force task
planning of mobile and redundant manipulators, Journal of Robotic
Systems, 16, pp. 151162 (1999).
[21] Ciocarlie, M., Miller, A. and Allen, P. Grasp analysis using deformable
fingers, IEEE/RSJ International Conference on Intelligent Robots and Systems,
pp. 41224128 (2005).
[22] Miller, A.T. and Allen, P.K. GraspIt! A versatile simulator for robotic
grasping, IEEE Robotics and Automation Magazine, 11(4), pp. 110122
(2004).
[23] Xue, Z., Zoellner, J.M. and Dillmann, R. Automatic optimal grasp planning
based on found contact points, Proc. of the IEEE/ASME Int. Conf. on
Advanced Intelligent Mechatronics, pp. 10531058 (2008).
[24] Zhu, X. and Wang, J. Synthesis of force-closure grasps on 3-D objects
based on Q distance, IEEE Transactions on Robotics and Automation, 19(4),
pp. 669679 (2003).
[25] Zhu, X. and Ding, H. Optimality criteria for fixture layout design:
a comparative study, IEEE Transactions on Automation Science and
Engineering, 6(4), pp. 658669 (2009).
[26] Strandberg, M. and Wahlberg, B. A method for grasp evaluation based on
disturbance force rejection, Robotics, IEEE Transactions on Robotics, 22(3),
pp. 461469 (2006).
[27] Suarez, R. and Roa, M. Finding locally optimum force-closure grasps,
Robotics and Computer-Integrated Manufacturing, 25(3), pp. 536544
(2009).
[28] Huebner, K., Ruthotto, S. and Kragic, D. Minimum volume bounding
box decomposition for shape approximation in robot grasping, IEEE
International Conference on Robotics and Automation (ICRA 2008), pp.
16281633 (2008).
[29] Nahvi, A., Hollerbach, J. and Hayward, V. Calibration of a parallel robot
using multiple kinematic closed loop, IEEE International Conference on
Robotics and Automation, ICRA 1994, pp. 407412 (1994).
[30] Gosselin, C. and Angeles, J. A global performance index for the kinematic
optimization of robotic manipulators, Journal of Mechanical Design,
113(3), pp. 220227 (1991).
[31] Yoshikawa, T. Manipulability of robotic mechanism, International Journal
on Robotics Research, 4(2), pp. 39 (1985).
[32] Melchiorri, C., Chiaccio, P., Chiaverini, S., Sciavicco, L. and Siciliano, B.
Comments on global task space manipulability ellipsoids for multiple-
arm systems and further considerations, (with reply) P. Chiacchio et al.,
IEEE Transactions on Robotics and Automation, 9(2), pp. 232236 (1993).
[33] Elkady, A.Y., Mohammed, M. and Sobh, T. A newalgorithmfor measuring
and optimizing the manipulability index, Journal of Intelligent and Robotic
Systems, 59, pp. 7586 (2010).
[34] Rimon, E. and Burdick, J.W. Mobility of bodies in contact. I. A 2nd-order
mobility index for multiple-finger grasps, IEEE Transactions on Robotics
and Automation, 14(5), pp. 696708 (1998).
[35] Sun, Y. and Hollerbach, J. Observability index selection for robot calibra-
tion, IEEE International Conference on Robotics and Automation (ICRA2008),
pp. 831836 (2008).
[36] Chiu, S.L. Task compatibility of manipulator postures, International
Journal on Robotics Research, 7(5), pp. 1321 (1988).
[37] Bowling, A. and Khatib, O. The dynamic capability equations: a new
tool for analyzing robotic manipulator performance, IEEE Transactions on
Robotics, 21(1), pp. 115123 (2005).
[38] Bowling, A. and Kim, C. Velocity effects on robotic manipulator dynamic
performance, ASME Journal of Mechanical Design, 128, pp. 12361245
(2006).
[39] Kim, J.H., Yang, J. and Abdel-Malek, K. Planning load effective dynamic
motions of highly articulated human model for generic tasks, Robotica,
27, pp. 739747 (2009).
[40] Kim, J.H., Xiang, Y., Bhatt, R., Yang, J., Chung, H.J., Arora, J.S. and Abdel-
Malek, K. Generating effective whole-body motions of a human-like
mechanism with efficient ZMP formulation, International Journal on
Robotics and Automation, 24(2), pp. 125136 (2009).
[41] Yoshikawa, T. Erratumto dynamic manipulability of robot manipulators,
Journal of Robotic Systems, 17, p. 449 (2000).
[42] Yokokohji, Y., San Martin, J. and Fujiwara, M. Dynamic manipulability of
multifingered grasping, IEEE Transactions on Robotics, 25(4), pp. 947954
(2009).
[43] Kurazume, R. and Hasegawa, T. A new index of serial-link manipulator
performance combining dynamic manipulability and manipulating force
ellipsoids, IEEE Transactions on Robotics, 22(5), pp. 10221028 (2006).
[44] Roa, M.A. and Suarez, R. Regrasp planning in the grasp space using
independent regions, IEEE/RSJ International Conference on Intelligent
Robots and Systems, IROS 2009, pp. 18231829 (2009).
[45] Phoka, T. and Sudsang, A. Regrasp planning of three-fingered hand for a
polygonal object, IEEE International Conference onRobotics and Automation
(ICRA 2010), pp. 43284333 (2010).
[46] Vannoy, J. and Xiao, J. Real-time tight coordination of mobile manipula-
tors inunknowndynamic environments, IEEE/RSJ International Conference
on Intelligent Robots and Systems, IROS 2007, pp. 25132519 (2007).
230 F. Cheraghpour et al. / Scientia Iranica, Transactions B: Mechanical Engineering 18 (2011) 222230
[47] Korayem, M.H. and Nikoobin, A. Maximum payload path planning
for redundant manipulator using indirect solution of optimal control
problem, International Journal of Advanced Manufacturing Technology,
44(78), pp. 725736 (2009).
[48] Watanabe, T. and Beetz, M. Grasp motion planning for box opening
task by multi-fingered hands and arms, IEEE International Symposium on
Computational Intelligence in Robotics and Automation, CIRA 2009, pp. 17
(2009).
[49] Yoshida, E., Esteves, C., Belousov, I., Laumond, J.-P., Sakaguchi, T. and
Yokoi, K. Planning 3-D collision-free dynamic robotic motion through
iterative reshaping, IEEE Transactions on Robotics, 24(5), pp. 11861198
(2008).
[50] Saravanan, R. and Ramabalan, S. Evolutionary minimum cost trajectory
planning for industrial robots, Journal of Intelligent and Robotic Systems,
52, pp. 4577 (2008).
[51] Cheraghpour, F., Moosavian, S.A.A. and Nahvi, A. Multiple aspect grasp
performance index for cooperative object manipulation tasks, IEEE/ASME
International Conference on Advanced Intelligent Mechatronics, pp. 386391
(2009).
[52] Angeles, J., Fundamentals of Robotic Mechanical Systems: Theory, Methods,
and Algorithms, 2nd Edn., Springer-Verlag, New York Inc. (2003), pp.
180188.
[53] Cheraghpour, F., Moosavian, S.A.A. and Nahvi, A. Robotic grasp planning
by multiple aspects grasp index for object manipulation tasks, 18th
Iranian Conference on Electrical Engineering (ICEE), pp. 635640 (2010).
[54] Nakamura, Y., Advanced Robotics: Redundancy and Optimization, Addison-
Wesley, Reading, MA (1991).
Farzad Cheraghpour Samavati was born in Tehran, Iran, in 1979. He received
a B.S. degree from the Central Branch of the Azad University in Tehran, in 2003,
and obtained his M.S. degree fromK.N. Toosi University of Technology, in 2006,
both in Mechanical Engineering. Now, he is a Ph.D. candidate at KNTU. His
current research interests include: Grasp Planning, Cooperative Manipulators,
Object Manipulation, Mobile Robots and Force Control.
S. Ali A. Moosavian received his B.S. degree in 1986 from Sharif University of
Technology, his M.S. degree in 1990 from Tarbiat Modaress University (both in
Tehran) andhis Ph.D. degree in1996fromMcGill University (Montreal, Canada),
all in Mechanical Engineering. Since 1997, he has been with the Mechanical
Engineering Department at K. N. Toosi University of Technology (Tehran);
currently, as a Professor. He teaches courses in areas of robotics, dynamics,
automatic control, analysis and synthesis of mechanisms. His research interests
include areas of: Automatic Control Systems, Dynamics and Control of Robotic
Systems. He has published a number of books in these fields and more than 150
articles in peer-reviewed journals and conference proceedings. He is one of the
founders of the ARAS Research Center for Design, Manufacturing and Control of
Robotic Systems, and Automatic Machineries.
Ali Nahvi received a Ph.D. degree in Mechanical Engineering from the
University of Utah in 2003. He is currently a faculty member at K.N. Toosi
University of Technology in Iran. He was the recipient of the Anton Philips Paper
Award in 1998. His research interests include areas of: Virtual Reality, Design
and Control of Driving Simulators, Haptic Interfaces and Virtual Laboratories.