Computation of the conguration degree of freedom

of a spatial parallel mechanism by using

reciprocal screw theory
Jing-Shan Zhao
*
, Zhi-Jing Feng, Jing-Xin Dong
Department of Precision Instruments, Tsinghua University, Beijing 100084, China
Received 27 June 2005; received in revised form 15 November 2005; accepted 3 January 2006
Available online 9 March 2006
Abstract
In this paper, we propose a programmable algorithm to investigate the conguration degree of freedom (CDOF) of a
spatial parallel mechanism in one Cartesian coordinate system with reciprocal screw theory. According to the physical
meaning of reciprocal screws, we rst obtain the terminal constraints of every kinematic chain which connects the end-
eector with the xed base, and then gain the free motion(s) of the end-eector and its degree of freedom (DOF). Through
analyzing the controllability of the end-eector, we investigate the instantaneous screw expressions for the free motions of
the end-eector of a spatial parallel mechanism and the CDOF of the mechanism. The whole mathematical analysis pro-
cess can be directly embedded in a kind of CAD software, in which the imperfect Kutzbach Gru bler formulas or its amend-
ments are mostly utilized to analyze the DOF.
2006 Elsevier Ltd. All rights reserved.
Keywords: Conguration degree of freedom; End-eector; Spatial parallel mechanism; Terminal constraints; Reciprocal screw theory
1. Introduction
Determining the independent motions of an end-eector is the key problem of mechanism analysis, while
realizing the prescribed motion(s) through an end-eector is the focus of mechanism synthesis. Consequently,
DOF of a mechanism is the primary problem to be solved both in analysis and in synthesis of mechanisms.
However, the currently received formulas of DOF are those of Kutzbach Gru blers [1,2], which have been
found not to work for a certain mechanisms. In theories and applications, many scholars [25] have brought
forward kinds of amendments for Kutzbach Gru bler formulas, which are often divided into two types, one is
for planar and spherical mechanisms and the other is for spatial mechanisms, but the essence and geneses of
them are consistent.
0094-114X/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechmachtheory.2006.01.006
*
Corresponding author. Fax: +86 01 06 278 2351.
E-mail address: zjs01@mails.tsinghua.edu.cn (J.-S. Zhao).
Mechanism and Machine Theory 41 (2006) 14861504
www.elsevier.com/locate/mechmt
Mechanism
and
Machine Theory
There is the following Kutzbach Gru blers formula [6,7] for planar and spherical mechanisms:
M 3n j 1

j
i1
f
i
1
There is the following Kutzbach Gru bler formula [6,7] for spatial mechanisms:
M 6n j 1

j
i1
f
i
2
where n represents the number of the total members of the system, j represents the number of the total
kinematic pairs, f
i
represents the number of DOF of the ith kinematic pair.
Waldron [2] and Hunt [3, pp. 3435, p. 376, p. 382] proposed the following formula:
M dn g 1

g
i1
f
i
3
where d represents the order of the mechanism, and d = 6 k, k represents the number of the common con-
straints of a mechanism, n represents the total number of the members in the mechanism, g represents the total
number of the kinematic pairs of the mechanism, f
i
represents the DOF of the ith kinematic pair, l represents
the number of the independent loops of a mechanism.
Hunt also gave a more general formula in his treatise [3]:
M

g
i1
f
i

l
j1
d
j
4
where d
j
(j = 1, . . . , l) is the order of the jth (j = 1, . . . , l) loop.
As to a history of the calculation of the mobility of a mechanism, Gogu [8] made a relatively good historical
review. Details about the various methods presented in the literatures during the past 150 years can be found
in [8].
In [6,7], we have already analyzed the restrictions of the above methods from dierent aspects, and pre-
sented analytical methods about the DOF of an end-eector and the CDOF of a mechanism. In this paper,
we will concentrate on the computation of CDOF of a spatial parallel mechanism in a xed (absolute) Carte-
sian coordinate system with reciprocal screw theory [35,911]. In brief review the steps addressed in [6,7], we
rst analyze the terminal constraints of each kinematic chain of the end-eector, and then investigate its free
motion(s) at a xed Cartesian coordinate system. The free motion(s) of the end-eector and the constraints
exerted to it by the kinematic chains are all expressed in Plu cker coordinates at the same Cartesian coordinate
system.
2. Primary theory of reciprocal screws and the analytical degree of freedom method of terminal constraints
of an end-eector
According to reciprocal screw theory [25,911], a screw $ is dened by a straight line with an associated
pitch h and is conveniently denoted by six Plu cker homogeneous coordinates:
$ s s
0
hs 5
where s denotes direction ratios pointing along the screw axis, s
0
= r s denes the moment of the screw axis
about the origin of the coordinate system, r is the position vector of any point on the screw axis with respect to
the coordinate system. Consequently, the screw axis can be denoted by the Plu cker homogeneous coordinates
$
axis
s s
0
.
Assume
s L M N ;
s
0
hs P Q R
_
6
J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1487
Considering s (s
0
+ hs) = s s
0
+ ksk
2
h = ksk
2
h and presuming s 50, we obtain the instant pitch of a
screw:
h
s s
0
hs
ksk
2

LP MQ NR
L
2
M
2
N
2
7
Therefore, from Eqs. (5)(7), the axis of the screw can also be denoted as
$
axis
L M N P Lh Q Mh R Lh 8
Assume that the vector of the projective point of the origin on the screw axis is represented by r
O
P
, there will
be s ? r
O
P
and
s r
O
P
s s sr
O
P
s r
O
P
s ksk
2
r
O
P
9
According to Eqs. (5) and (6), there are
s L M N ;
r
O
P
s P Lh Q Mh R Nh
_
10
which yields
r
O
P

s r
O
P
s
ksk
2

1
L
2
M
2
N
2
MR Nh NQ Mh
NP Lh LR Nh
LQ Mh MP Lh
_

_
_

_ 11
Consequently, if we know the Plu cker coordinates of a screw, we can easily obtain the unit direction vector,
s, the pitch, h, the screw axis and the vector of the projective point of the origin on the axis, r
O
P
, with Eqs. (5)
(11).
If the pitch of a screw equals zero, the screw coordinates reduce to be:
$ s s
0
12
which is just the Plu cker homogeneous coordinates of the screw axis.
In fact, formula (12) uniquely denes a line in a three-dimensional space. Assume that point O
P
is the
projective point of the origin on a line l and point A is any other point on the line. Then,
r
A
r
O
P
r
O
P
A
r
O
P

as
ksk
13
where s is a direction vector of line l, a is the length of line O
P
A.
The moment of line l about the origin at point A will be
s
0
r
A
s r
O
P

as
ksk
_ _
s r
O
P
s 14
From Eqs. (12) and (14), we know that the moment of a line about the origin is irrelevant to the points
selection on the line.
If a screw passes through the origin of the coordinate system, the screw coordinates can be denoted as
$ s hs 15
On the other hand, if the pitch of a screw is innite, the screw is dened as
$ 0 s 16
where 0 0 0 0 is a three-dimensional vector.
According to the above denitions, a screw associated with a revolute pair is a twist of zero pitch pointing
along the pair axis while a screw associated with a prismatic pair is a twist of innite pitch pointing in the
direction of the translational guide line of the pair.
1488 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504
From Eq. (6), we know that the kinematic screw is often denoted in the form of Plucker homogeneous
coordinates:
$ L M N P Q R 17
where the rst three components denote the angular velocity, the last three components denote the linear
velocity of a point in the rigid body that is instantaneously coincident with the origin of the coordinate system.
Similarly, $
r
is dened as
$
r
L
r
M
r
N
r
P
r
Q
r
R
r
18
where the rst three components denote the resultant force and the last three components denote the resultant
moment about the origin of the coordinate system.
Two screws, $ and $
r
, are called to be reciprocal if they satisfy the equation:
LP
r
MQ
r
NR
r
PL
r
QM
r
RN
r
0 19
Obviously, the free motions (general twists) $ and the prescribed constraints (general wrenches) $
r
of an
equilibrium rigid body should satisfy Eq. (19).
Eq. (19) is often written for short:
$ $
r
0 20
where denotes the reciprocal production of two screws. Kumar denote expression (20) in a form of matrix
multiplication [12]:
P$
T
$
r
0
where $ and $
r
are column vectors, P
0
3
I
3
I
3
0
3
_ _
, and I
3
and 0
3
are 3 3 identity and zero matrices, respec-
tively. But in this paper, for the sake of simplicity, we always denote the Plu cker homogeneous coordinates, $
and $
r
, in the form of row vectors.
Similarly, if we get a set of terminal constraints exerted to a rigid body, its free motion(s) can also be solved
through Eq. (20).
Therefore, based on the work of [6] and [7], we propose a programmable algorithm for investigating the
mobility properties of the end eector and the CDOF of spatial parallel mechanisms. Compared with [6]
and [7], the most distinctive merits of this paper are that the free motions of the end-eector are instanta-
neously expressed in a set of Plu cker homogeneous coordinates in one Cartesian coordinate system and this
processing is most adapt to programming. Below are the main steps.
2.1. Investigate the terminal constraints of the kinematic chains
In general, we can decompose the system into n (n P1) kinematic chains connecting the end eector with the
base, which should be satised to the spatial parallel mechanisms. In order to instantaneously analyze the
mobility properties of the end-eector, we herein only establish one absolute coordinate system while a local
coordinate system often needs to be created for every kinematic chain in [6,7]. After establishing the coordinate
system, the Plu cker homogeneous coordinates of all kinematic pairs in a chain can be obtained. Group all of the
kinematic screws of the same chain to be $
i
(i = 1, 2, . . . , n) and solve the terminal constraint(s) $
r
i
with Eq. (20).
In fact, if all of the terminal constraints of the kinematic chains were gained, the constraints exerted to the
end-eector, denoted by $
r
E
, should also be obtained. The dimension of constraint spaces spanned by the
terminal constraints of kinematic chains can be simplied as d Rank$
r
E
.
2.2. Solve the free motion(s), $
F
E
, of the end-eector with equation $
F
E
$
r
E
0
Naturally, the mobility properties of the end-eector is fully expressed by $
F
E
. Its number of DOF can be
expressed as
M Rank$
F
E
6 d 21
J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1489
2.3. Exert M actuations to the mechanism, and then investigate the CDOF of the mechanism by solving the free
motion(s) of the end-eector within its workspace
If the newly solved motion(s), denoted by $
F
ni
E
; i 1; 2; . . ., satisfy that $
F
ni
E
6 0 0 0 0 0 0 , then,
additional actuations are needed under this conguration and the actuation scheme. Of course, we can either
reselect the actuation scheme or add Rank$
F
ni
E
more actuation(s) under this conguration until
$
F
ni
E
0 0 0 0 0 0 .
In order to calculate the CDOF of a mechanism with a prescribed end-eector (output member), we
introduced a medi-formula in [6]:
F 6 d

n
i1
r
i
22
where F is the DOF of the mechanism with an end-eector, d is the number of dimensions that all the terminal
constraints can be spanned in the constraint spaces, r
i
, i = 1, 2, . . . , n is the rank deciency of the kinematic
screws of the ith chain.
But as a matter of fact, we can directly use M
j
Rank$
F
nj
E
; j 1; 2; . . . ; m 1 (where M
m
= 0) as the
medi-formula provided that the screws of each chain were replaced by their simplied screws (step (6) in
[6]). This is because the rank deciency of the kinematic screws of the ith chain will be compensated in the
second sum term of Eq. (30).
For instance, the kinematic screws of a simplied 5-R (revolute pair) kinematic chain shown in Fig. 1 are
$
ABCDE

$
A
$
B
$
C
$
D
$
E
_

_
_

_
23
where
$
A
0 0 1 0 0 0
$
B
0 0 1 y
B
x
B
0
$
C
0 0 1 y
C
x
C
0
$
D
0 0 1 y
D
x
D
0
$
E
0 0 1 y
E
x
E
0
o
x
y
End-Effector E
C
D
B
A( )
Fig. 1. A simplied 5-R kinematic chain.
1490 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504
The rank of the kinematic screws of the chain can be gained with algebra method:
R$
ABCDE
3
The rank deciency of the chain is
r
ABCDE
n
ABCDE
R$
ABCDE
5 3 2 24
where n
ABCDE
denotes the total number of the simplied screws [6] of the kinematic chain ABCDE.
The reciprocal screws (terminal constraints) of the kinematic chain can be obtained:
$
r
ABCDE

$
r
1
$
r
2
$
r
3
_

_
_

_ 25
where
$
r
1
0 0 1 0 0 0
$
r
2
0 0 0 1 0 0
$
r
3
0 0 0 0 1 0
The free motions the end eector can execute will be gained according to Eq. (20):
$
F
E

$
f
1
$
f
2
$
f
3
_

_
_

_ 26
where $
f
1
0 0 1 0 0 0 , and $
f
1
denotes one rotational constraint moment about z-axis;
$
f
2
0 0 0 1 0 0 , and $
f
2
denotes one translational constraint force along x-axis;
$
f
3
0 0 0 0 1 0 , and $
f
3
denotes one translational constraint force along y-axis.
As a matter of fact, $
F
E
is a subspace spanned by $
ABCDE
. If we add M Rank$
F
E
3 actuations to any
three revolute pairs, revolute pairs A, B, and C for example, the kinematic screws of the chain become:
$
DE

$
D
$
E
_ _
27
The reciprocal screws of (27) will be
$
r
DE

0 0 1 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
x
D
x
E

x
D
x
E

2
y
D
y
E

2
_
y
D
y
E

x
D
x
E

2
y
D
y
E

2
_ 0 0 0
x
E
y
D
x
D
y
E

x
D
x
E

2
y
D
y
E

2
_
_

_
_

_
28
Therefore, the free motions the end eector can execute under these actuations will be gained according to
Eq. (20) again
$
F
n1
E

0 0 1 y
D
x
D
0
0 0 0
y
D
y
E

x
D
x
E

2
y
D
y
E

2
_
x
D
x
E

x
D
x
E

2
y
D
y
E

2
_ 0
_

_
_

_
29
So, M
1
Rank$
F
n1
E
2. As a matter of fact, in this case, Eq. (29) is a subspace spanned by Eq. (27). Sim-
ilarly, we can go on adding actuations to the chain to analyze the eects on the free motions of the end-eec-
tor. We can nd that the number of free motions of the end-eector will not reduce to zero until ve actuations
are applied to the chain. Consequently, the term

m1
j1
Rank$
F
nj
E
will be ve and be two more than

m1
j1
F
j
M
in [6] which equals the result of Eq. (24).
J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1491
So the CDOF proposed in [6] can also be denoted as
F
conf
M

m1
j1
Rank $
F
nj
E
_ _
Rank $
F
E
_ _

m1
j1
Rank $
F
nj
E
_ _
6 d

m1
j1
Rank $
F
nj
E
_ _
30
Eq. (30) has the same form as that in [6] but is much simpler.
3. Applications and discussions
3.1. Analysis on the mobility properties of a spatial four-bar mechanism
A spatial four-bar mechanism, shown in Fig. 2 [6], consists of 2-RR (two revolute pairs) kinematic chains.
To analyze the free motion of the end-eector BC, we rst decompose the spatial parallel mechanism to form
two kinematic chains connecting link BC with the base and then study the reciprocal screws of each kinematic
chain.
Now, we rst analyze the terminal constraints of kinematic chain AB exerted to link BC.
The coordinate system oxyz is established as Figs. 2 and 3 show, the origin of the coordinate system is
superimposed with pair A, z-axis is along the axis of revolute pair A, and y-axis is perpendicular to the base
plane, and therefore, x-axis is naturally parallel to the base plane. If we presume the coordinates of pair B are
x
B
y
B
0 , the Plu cker coordinates of pairs A and B can be denoted as
Fig. 2. A spatial four-bar mechanism.
x
y
A (o)
B(C)
Fig. 3. The xed coordinate system.
1492 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504
$
A
0 0 1 0 0 0
$
B
0 0 1 y
B
x
B
0
Therefore, the kinematic screws of kinematic chain AB can be denoted as
$
AB

$
A
$
B
_ _
31
The reciprocal screws of $
AB
are
$
r
AB

$
r1
AB
$
r2
AB
$
r3
AB
$
r4
AB
_

_
_

_
32
where
$
r1
AB

x
B

x
2
B
y
2
B
_
y
B

x
2
B
y
2
B
_ 0 0 0 0
_ _
$
r2
AB
0 0 1 0 0 0
$
r3
AB
0 0 0 1 0 0
$
r4
AB
0 0 0 0 1 0
Similarly, the terminal constraints of kinematic chain DC in the same Cartesian coordinate system can be
solved:
$
r
DC

$
r1
DC
$
r2
DC
$
r3
DC
$
r4
DC
_

_
_

_
33
where
$
r1
DC

x
B
x
D

x
B
x
D

2
y
B
y
D

2
_
y
B
y
D

x
B
x
D

2
y
B
y
D

2
_ 0 0 0
x
D
y
B
x
B
y
D

x
B
x
D

2
y
B
y
D

2
_
_ _
$
r2
DC
0 0 1 0 0 0
$
r3
DC
0 0 0 1 0 0
$
r4
DC
0 0 0 0 1 0
Therefore, the terminal constraints exerted to link BC are
$
C
BC

x
B

x
2
B
y
2
B
_
y
B

x
2
B
y
2
B
_ 0 0 0 0
x
B
x
D

x
B
x
D

2
y
B
y
D

2
_
y
B
y
D

x
B
x
D

2
y
B
y
D

2
_ 0 0 0
x
D
y
B
x
B
y
D

x
B
x
D

2
y
B
y
D

2
_
0 0 1 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
_

_
_

_
34
J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1493
As a result, the free motion of the end-eector can be solved with Eq. (34) according to reciprocal screw
theory again
$
BC
0 0 1 y
B
x
B
0 35
With Eq. (7), we can nd the pitch of the screw (twist):
h
s s
0
hs
ksk
2

LP MQ NR
L
2
M
2
N
2
0 36
With Eq. (11), we can nd the projecting point of the origin on the axis is
r
O
P

s r
O
P
s
ksk
2

1
L
2
M
2
N
2
MR Nh NQ Mh
NP Lh LR Nh
LQ Mh MP Lh
_

_
_

_
x
B
y
B
0
_

_
_

_ 37
Therefore, Eq. (35) indicates that the end-eector BC can make a free rotation about a line passing through
point x
B
y
B
0 with a direction 0 0 1 , which is just its self-rotation axisz
0
-axis shown in Fig. 2.
So, the end-eector only has one rotational DOF about z
0
-axis; and as a result, we can only either select
pair B or select C as the actuation. In application, this mechanism can be utilized as an adjustable bearing.
3.2. Analysis on the mobility properties of the end-eector in a 4-PUU spatial parallel mechanism
A spatial parallel mechanism, shown in Fig. 4 [6], is made up of 4-PUU (one prismatic pair and two uni-
versal pairs) kinematic chains. To investigate the mobility properties of the end-eector, we rst decompose
the spatial parallel mechanism to be four kinematic chains connecting the end-eector with the xed guides
and then analyze the reciprocal screws (terminal constraints) of each kinematic chain according to the above
analysis steps.
Assume that the distance between the two guides is 2L, the x-axis of the xed Cartesian coordinate system
(xed coordinate frame) shown in Fig. 4 is superimposed with the middle line of the two guides, and y-axis and
z-axis are all perpendicular to the guides. Therefore, the Cartesian coordinates of slider P
i
(i = 1, 2, 3, 4) can be
denoted as
P
1
x
1
L 0 ; P
2
x
2
L 0 ; P
3
x
3
L 0 ; P
4
x
4
L 0
We rst analyze kinematic chain P
1
B
1
in the xed Cartesian coordinate system shown in Fig. 4.
In Fig. 5, x
0
, y
0
and z
0
are parallel to the corresponding axes of the xed one shown in Fig. 4. B
1P
is the
projection of B
1
in xoy plane, and OB
1P
is the projective line of OB
1
in xoy plane. b
1
is the angle from line
OB
1P
to line OB
1
, a
1
is the angle from x axis to line OB
1P
.
Fig. 4. A spatial parallel mechanism made up of 4-PUU.
1494 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504
The Plu cker coordinates of slider P
1
are
$
1
0 0 0 1 0 0
The universal pair P
1
can be decomposed as two orthogonal rotational pairs, the direction vectors of which
are denoted by e
1
u
and e
2
u
individually. Where e
1
u
is parallel to z-axis and as a result has the form of
e
1
u
0 0 1 , e
2
u
can be found by the following formula:
e
2
u

P
1
B
1P
!
P
1
B
1P
!
_
_
_
_
_
_
o
0
z
0
!
cos a
1
sin a
1
0 0 0 1 sin a
1
cos a
1
0
Therefore, the individual Plucker coordinates corresponding to universal pair P
1
are
$
2
0 0 1 L x
1
0
$
3
sin a
1
cos a
1
0 0 0 Lsin a
1
x
1
cos a
1

And the same are true for universal pair B

1
. The coordinates of B
1
are
B
1
x
1
l
1
cos a
1
cos b
1
L l
1
sin a
1
cos b
1
l
1
sin b
1

The Plu cker coordinates can be denoted as

$
4
$
1
4
; $
0
4
; $
5
$
1
5
; $
0
5

where
$
1
4
0 0 1 ; $
1
5
sin a
1
cos a
1
0
$
0
4
r
B
1
$
1
4
L l
1
sin a
1
cos b
1
x
1
l
1
cos a
1
cos b
1
0
$
0
5
r
B
1
$
1
5
l
1
cos a
1
sin b
1
l
1
sin a
1
sin b
1
L sin a
1
x
1
cos a
1
l
1
cos b
1

Therefore,
$
4
0 0 1 L l
1
sin a
1
cos b
1
x
1
l
1
cos a
1
cos b
1
0
$
5
sin a
1
cos a
1
0 l
1
cos a
1
sin b
1
l
1
sin a
1
sin b
1
Lsin a
1
x
1
cos a
1
l
1
cos b
1

Fig. 5. PUU kinematic chain.

J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1495
As a result, the kinematic screws of the kinematic chain P
1
B
1
can be expressed as
$
P
1
B
1

$
1
$
2
$
3
$
4
$
5
_

_
_

_
38
The terminal constraint of kinematic chain P
1
B
1
can be gained:
$
r
P
1
B
1
0 0 0 cos a
1
sin a
1
0 39
According to the physical meaning of the reciprocal screws, $
r
P
1
B
1
denotes a pure constraint moment that is
perpendicular to the universal pair plane.
With similar process, we can nd the terminal constraints of the other three kinematic chains of the end-
eector:
$
r
P
2
B
2
0 0 0 cos a
2
sin a
2
0 40
$
r
P
3
B
3
0 0 0 cos a
3
sin a
3
0 41
$
r
P
4
B
4
0 0 0 cos a
4
sin a
4
0 42
where a
i
(i = 1, 2, 3, 4) is the angle from x-axis of the absolute coordinate system to the projective line of the
ith kinematic chain in the plane xoy.
Therefore, the terminal constraints exerted to the end-eector are four pure constraint moments, which are
all perpendicular to the direction of z-axis shown in Fig. 6.
The terminal constraints exerted to the end-eector are
$
r
E

$
r
P
1
B
1
$
r
P
2
B
2
$
r
P
3
B
3
$
r
P
4
B
4
_

_
_

_
43
Obviously, 1 6 Rank$
r
E
6 2. With Eq. (20), we can nd the free motion(s) of the end-eector as the
following two cases:
Fig. 6. The terminal constraints exerted to the end-eector B
1
B
2
B
3
B
4
.
1496 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504
(1) d Rank$
r
E
1
In this case, there must be
cos a
i
0;
sin a
i
1;
_
i 1; 2; 3; 4 or
cos a
i
1;
sin a
i
0;
_
i 1; 2; 3; 4 44
In the applicable structure shown in Fig. 4, only the rst solution set of Eq. (44) is possible. Therefore, we
have
a
1
a
4

p
2
a
2
a
3

3p
2
_
45
In fact, this case is corresponding to one kind of singularities of the mechanism.
The free motions the end-eector gains in this case can be obtained by solving $
r
E
$
F
1
E
0
$
F
1
E

$
1
E
$
2
E
$
3
E
$
4
E
$
5
E
_

_
_

_
46
where
$
1
E
1 0 0 0 0 0
$
2
E
0 0 1 0 0 0
$
3
E
0 0 0 1 0 0
$
4
E
0 0 0 0 1 0
$
5
E
0 0 0 0 0 1
(2) d Rank$
r
E
2
The free motions the end-eector obtained in this case can be similarly solved:
$
F
2
E

$
2
E
$
3
E
$
4
E
$
5
E
_

_
_

_
47
From the above analysis, we can nd that the end-eector at least has the motions represented by $
F
2
E
. When
a
1
a
4

p
2
and a
2
a
3

3p
2
, the end-eector gains one more rotation $
1
E
.
That is, the end-eector at least has M Rank$
F
2
E
4 DOFs, whose mobility properties are depicted by
$
F
2
E
one rotational DOF about z-axis and three independent translational DOFs along the coordinate axes in
a real three-dimensional space.
If we select the four sliders shown in Fig. 4 as the actuators, the $
F
n1
E
can be analyzed as follows:
When the actuations are given, the four sliders will be seen as xed pairs. Therefore, the kinematic screws of
kinematic chain P
1
B
1
can be rewritten as
$
actu
P
1
B
1

$
2
$
3
$
4
$
5
_

_
_

_
48
J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1497
Therefore, the reciprocal screws of kinematic chain P
1
B
1
can be gained:
$
r1
P
1
B
1
0 0 0 cos a
1
sin a
1
0
$
r2
P
1
B
1
cos a
1
cos b
1
sin a
1
cos b
1
sin b
1
L sin b
1
x
1
sin b
1
x
1
sin a
1
L cos a
1
cos b
1

_
49
With similar processes, we can nd the terminal constraints of the other three kinematic chains of the end-
eector in the same Cartesian coordinate system:
$
r1
P
2
B
2
0 0 0 cos a
2
sin a
2
0
$
r2
P
2
B
2
cos a
2
cos b
2
sin a
2
cos b
2
sin b
2
L sin b
2
x
2
sin b
2
x
2
sin a
2
Lcos a
2
cos b
2

_
50
$
r1
P
3
B
3
0 0 0 cos a
3
sin a
3
0
$
r2
P
3
B
3
cos a
3
cos b
3
sin a
3
cos b
3
sin b
3
L sin b
3
x
3
sin b
3
x
3
sin a
3
Lcos a
3
cos b
3

_
51
$
r1
P
4
B
4
0 0 0 cos a
4
sin a
4
0
$
r2
P
4
B
4
cos a
4
cos b
4
sin a
4
cos b
4
sin b
4
L sin b
4
x
4
sin b
4
x
4
sin a
4
L cos a
4
cos b
4

_
52
Therefore, when the actuations are assigned to the four sliders, the terminal constraints exerted to the end-
eector are four pure constraint forces along the individual axes of the limbs, and four pure constraint moments,
which are all perpendicular to the direction of z-axis in the absolute coordinate system shown in Fig. 7.
The terminal constraints exerted to the end-eector under the given actuations are
$
r
E

$
r1
P
1
B
1
$
r1
P
2
B
2
$
r1
P
3
B
3
$
r1
P
4
B
4
$
r2
P
1
B
1
$
r2
P
2
B
2
$
r2
P
3
B
3
$
r2
P
4
B
4
_

_
_

_
53
Fig. 7. The terminal constraints exerted to the end-eector B
1
B
2
B
3
B
4
.
1498 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504
Obviously, the number of DOFs of the end-eector can be calculated with M 6 d 6 Rank$
r
E
.
From the above analysis, we know that the four pure moments can prevent the end-eector from rolling about
x-axis and y-axis in the absolute coordinate system provided that Eq. (45) did not hold.
Therefore, b
i
(i = 1, 2, 3, 4) will keep invariable during moving. Assume b
i
= b, we can simplify the analysis
of free motion(s) of the end-eector under the given actuations with Eq. (20).
Assume the free motion(s) that the end-eector gains under the given actuations is denoted by
$
F
ni
E
l m n p q r 54
which can be obtained by solving the following linear equation set:
0 0 0 cos a
1
sin a
1
0
0 0 0 cos a
2
sin a
2
0
0 0 0 cos a
3
sin a
3
0
0 0 0 cos a
4
sin a
4
0
cos a
1
cos b sin a
1
cos b sin b L sin b x
1
sin b x
1
sin a
1
L cos a
1
cos b
cos a
2
cos b sin a
2
cos b sin b L sin b x
2
sin b x
2
sin a
2
L cos a
2
cos b
cos a
3
cos b sin a
3
cos b sin b L sin b x
3
sin b x
3
sin a
3
L cos a
3
cos b
cos a
4
cos b sin a
4
cos b sin b L sin b x
4
sin b x
4
sin a
4
L cos a
4
cos b
_

_
_

_
p
q
r
l
m
n
_

_
_

_
0 55
In reality, b 50 and b 6
p
2
, therefore, Eq. (55) can be simplied to be
0 0 0 cos a
1
sin a
1
0
0 0 0 cos a
2
sin a
2
0
0 0 0 cos a
3
sin a
3
0
0 0 0 cos a
4
sin a
4
0
cos a
1
sin a
1
tan b Ltan b x
1
tan b x
1
sin a
1
L cos a
1
cos a
2
sin a
2
tan b Ltan b x
2
tan b x
2
sin a
2
L cos a
2
cos a
3
sin a
3
tan b Ltan b x
3
tan b x
3
sin a
3
L cos a
3
cos a
4
sin a
4
tan b Ltan b x
4
tan b x
4
sin a
4
L cos a
4
_

_
_

_
p
q
r
l
m
n
_

_
_

_
0 56
In order to investigate the instantaneous controllability of the end-eector under the four given actuations
within its workspace, we have to nd the forward displacement solutions of the system. For the sake of sim-
plicity and feasibility for any cases, we adopt a natural coordinate method, which was originally proposed by
Javier Garc a de Jalo n and Eduardo Bayo [13], to build the forward displacement model. Because it is beyond
the scope of this paper, we herein do not intend to make a thorough discussion of the method and only give
the natural coordinate equations of the model. More details about the method can be found in [13].
Let us select the Cartesian coordinates of the four pairs, B
i
x
B
i
y
B
i
z
B
i
; i 1; 2; 3; 4, on the end-eector
as the natural coordinates. From the above analysis, the end-eector has one rotational DOF about z-axis and
three translational DOFs along the orthogonal coordinate frames. Therefore, the relationship
z
B
1
z
B
2
z
B
3
z
B
4
z
B
holds. So, the number of the natural coordinates can be reduced from twelve to nine,
and the coordinates can be denoted by B
i
x
B
i
y
B
i
z
B
; i 1; 2; 3; 4 individually.
The four equations for actuations are
x
B
1
x
1

2
y
B
1
L
2
z
B
0
2
l
2
1
x
B
2
x
2

2
y
B
2
L
2
z
B
0
2
l
2
2
x
B
3
x
3

2
y
B
3
L
2
z
B
0
2
l
2
3
x
B
4
x
4

2
y
B
4
L
2
z
B
0
2
l
2
4
_

_
57
where l
i
(i = 1, 2, 3, 4) is the length of each limb. In this model, l
i
= l.
J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1499
The rigid body constraint equations can be given below:
x
B
2
x
B
1

2
y
B
2
y
B
1

2
a
2
x
B
4
x
B
1

2
y
B
4
y
B
1

2
b
2
x
B
4
x
B
2

2
y
B
4
y
B
2

2
c
2
_

_
58
where a B
1
B
2
!
_
_
_
_
_
_, b B
1
B
4
!
_
_
_
_
_
_, c B
2
B
4
!
_
_
_
_
_
_.
The rigid geometric constraint equations are
x
B
3
x
B
1
k
1
x
B
2
x
B
1
k
2
x
B
4
x
B
1

y
B
3
y
B
1
k
1
y
B
2
y
B
1
k
2
y
B
4
y
B
1

_
59
where k
j
(j = 1, 2) is determined uniquely by the rigid geometric relationships of the end-eector. For example,
k
1
= k
2
= 1 for a parallelogram end-eector in this model.
The forward displacement model can be built up by grouping Eqs. (57)(59).
F x
B
i
; y
B
i
; z
B

x
B
1
x
1

2
y
B
1
L
2
z
2
B
l
2
1
x
B
2
x
2

2
y
B
2
L
2
z
2
B
l
2
2
x
B
3
x
3

2
y
B
3
L
2
z
2
B
l
2
3
x
B
4
x
4

2
y
B
4
L
2
z
2
B
l
2
4
x
B
2
x
B
1

2
y
B
2
y
B
1

2
a
2
x
B
4
x
B
1

2
y
B
4
y
B
1

2
b
2
x
B
4
x
B
2

2
y
B
4
y
B
2

2
c
2
x
B
1
x
B
2
x
B
3
x
B
4
y
B
1
y
B
2
y
B
3
y
B
4
_

_
_

_
0 60
The Jacobian matrix of (60) is
oF x
Bi
; y
B
i
; z
B

ox
B
i
; y
Bi
; z
B

2
x
B
1
x
1
y
B
1
L 0 0 0 0 0 0 z
B
0 0 x
B
2
x
2
y
B
2
L 0 0 0 0 z
B
0 0 0 0 x
B
3
x
3
y
B
3
L 0 0 z
B
0 0 0 0 0 0 x
B
4
x
4
y
B
4
L z
B
x
B
2
x
B
1
y
B
2
y
B
1
x
B
2
x
B
1
y
B
2
y
B
1
0 0 0 0 0
x
B
4
x
B
1
y
B
4
y
B
1
0 0 0 0 x
B
4
x
B
1
y
B
4
y
B
1
0
0 0 x
B
4
x
B
2
y
B
4
y
B
2
0 0 x
B
4
x
B
2
y
B
4
y
B
2
0
1
2
0
1
2
0
1
2
0
1
2
0 0
0
1
2
0
1
2
0
1
2
0
1
2
0
_

_
_

_
61
Therefore, the forward displacement solutions can be found by numerical method, such as Newton
Raphson method.
Assume X
j
x
j
B
1
y
j
B
1
x
j
B
2
y
j
B
2
x
j
B
3
y
j
B
3
x
j
B
4
y
j
B
4
z
j
B
_ _
T
; j 1; 2; . . ., then the iteration formula of
NewtonRaphson method is
X
j1
X
j

oF X
j

oX
j
_ _
1
F X
j
62
1500 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504
After obtaining the forward displacement solutions, we can calculate the coecient matrix of Eq. (56) with
cos a
i

x
B
i

x
2
B
i
y
2
B
i
_
sin a
i

y
B
i

x
2
B
i
y
2
B
i
_
tan b
z
B

x
2
B
1
y
2
B
1
_
_

_
63
Therefore, the free motion(s) the end-eector obtains under the given actuations can be solved with Eq.
(56). If $
F
n1
E
0 0 0 0 0 0 , the end-eector must be stable; otherwise, additional actuations should
be added under this conguration. So, the CDOF of the mechanism is a function of the conguration and
instantaneously depicts the actuations needed to control the end-eector.
Obviously, if Rank$
r
E
6, then $
F
n1
E
0 0 0 0 0 0 and the end-eector is controllable; other-
wise, there will be $
F
n1
E
6 0 0 0 0 0 0 , and we have to add additional Rank $
F
n1
E
_ _
actuations or rese-
lect the initial four actuations to control the end-eector under this conguration. In fact, these congurations
are the traditional singularities and the CDOF of the mechanism is
F
conf
6 d

m1
j1
Rank $
F
nj
E
_ _
6 2

m1
j1
Rank $
F
nj
E
_ _
P4 64
All of the above analysis process can be programmed to instantaneously compute the CDOF of a concept
mechanism at a computer. The CDOF of the mechanism under any actuation schemes can be searched
throughout its whole workspace by a computer. Besides, the CDOF and the forward displacement problems
can be simultaneously cracked by embedding the algorithm proposed in this paper into the CAD software
directly.
3.3. Analysis on the mobility properties of the end-eector in a 3-UPU spatial parallel mechanism
A spatial parallel mechanism shown in Fig. 8 [14] consists of three identical UPU kinematic chains.
In order to investigate the terminal constraints of the kinematic chains connecting the mobile end-eector
with the xed base, we establish a xed Cartesian coordinate system as Fig. 8 shows. The origin of the coor-
dinate system oxyz is superimposed with the center of universal pair B
1
, and y-axis is along the midline of the
base triangle, x-axis passes through one of the two orthogonal axes of universal pair B
1
and the z-axis is per-
pendicular to the base plane. Assume the length of limb B
1
P
1
M
1
is l
1
. The screws of kinematic chain B
1
P
1
M
1
can be obtained:
The universal pair B
1
can be decomposed as two orthogonal revolute pairs, one of which is about x-axis.
From Fig. 8, we can nd that the normal line vector of the universal pair plane can be denoted as
Fig. 8. The architecture of a spatial parallel manipulator.
J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1501
e
B
1
0 cos b
1
sin b
1
65
Therefore, the other revolute pair of B
1
can be expressed as
x e
B
1
0 sin b
1
cos b
1
66
The individual Plu cker coordinates are
$
1
1 0 0 0 0 0
$
2
0 sin b
1
cos b
1
0 0 0
The Plu cker coordinates of slider P
1
are
$
3
0 0 0 x
M
1
y
M
1
z
M
1

The Plu cker coordinates for the two revolute pairs of M
1
can be denoted as
$
4
$
1
4
; $
0
4
; $
5
$
1
5
; $
0
5

where
$
1
4
1 0 0 ; $
1
5
0 sin b
1
cos b
1

$
0
4
r
B
1
$
1
4
0 z
M
1
y
M
1

$
0
5
r
B
1
$
1
5
y
M
1
cos b
1
z
M
1
sin b
1
x
M
1
cos b
1
x
M
1
sin b
1

Therefore
$
4
1 0 0 0 z
M
1
y
M
1

$
5
0 sin b
1
cos b
1
y
M
1
cos b
1
z
M
1
sin b
1
x
M
1
cos b
1
x
M
1
sin b
1

Therefore, the kinematic screws of the limb B

1
P
1
M
1
can be expressed as
$
B
1
P
1
M
1

$
1
$
2
$
3
$
4
$
5
_

_
_

_
67
) The reciprocal screw of the kinematic chain B
1
P
1
M
1
can be gained:
$
r
B
1
P
1
M
1
0 0 0 0 cos b
1
sin b
1
68
Similarly, we can nd the terminal constraints of the rest two kinematic chains:
$
r
B
2
P
2
M
2
0 0 0 cos a
2
cos b
2
sin a
2
cos b
2
sin b
2
69
$
r
B
3
P
3
M
3
0 0 0 cos a
3
cos b
3
sin a
3
cos b
3
sin b
3
70
In fact, from Fig. 8 we can nd that there are the following equations:
a
2

5p
6
a
3

p
6
_
71
Therefore, the terminal constraints exerted to the end-eector are
$
r
E

$
r
B
1
P
1
M
1
$
r
B
2
P
2
M
2
$
r
B
3
P
3
M
3
_

_
_

_ 72
1502 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504
The free motions the end-eector has, denoted by $
F
E
, can be solved with Eq. (20):
$
F
E
$
r
E
0
If d Rank$
r
E
6 3, it is corresponding to the singular cases which were discussed thoroughly in [14].
If d Rank$
r
E
3, the number of DOFs of the end-eector is M = 6 d = 6 3 = 3. And we can easily
solve the free motions of the end-eector:
$
F
E

0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
_

_
_

_ 73
Obviously, the DOFs of the end-eector shown in Fig. 8 include three independent translations along the
orthogonal coordinate axes.
If we introduce three actuations to the prismatic pairs shown in Fig. 8, the three prismatic pairs can be seen
as xed ones. Therefore, the kinematic screws of kinematic chain B
1
P
1
M
1
can be rewritten as
$
actu
B
1
P
1
M
1

$
1
$
2
$
4
$
5
_

_
_

_
74
Therefore, the terminal constraints of kinematic chain B
1
P
1
M
1
under the actuations can be gained:
$
r1
B
1
P
1
M
1
0 0 0 0 cos b
1
sin b
1

$
r2
B
1
P
1
M
1
x
M
1
y
M
1
z
M
1
0 0 0
_
75
With similar process, we can nd the terminal constraints of the rest two kinematic chains of the end-
eector when the four sliders are selected as the actuators in the same Cartesian coordinate system:
$
r1
B
2
P
2
M
2
0 0 0 cos a
2
cos b
2
sin a
2
cos b
2
sin b
2

$
r2
B
2
P
2
M
2
x
M
2
x
B
2
y
M
2
y
B
2
z
M
2
z
B
2
0 0 0
_
76
$
r1
B
3
P
3
M
3
0 0 0 cos a
3
cos b
3
sin a
3
cos b
3
sin b
3

$
r2
B
3
P
3
M
3
x
M
3
x
B
3
y
M
3
y
B
3
z
M
3
z
B
3
0 0 0
_
77
Therefore, the terminal constraints exerted to the end-eector under the given actuations are
$
r
E

$
r2
B
1
P
1
M
1
$
r2
B
2
P
2
M
2
$
r2
B
3
P
3
M
3
$
r1
B
1
P
1
M
1
$
r1
B
2
P
2
M
2
$
r1
B
3
P
3
M
3
_

_
_

x
M
1
y
M
1
z
M
1
0 0 0
x
M
2
x
B
2
y
M
2
y
B
2
z
M
2
z
B
2
0 0 0
x
M
3
x
B
3
y
M
3
y
B
3
z
M
3
z
B
3
0 0 0
0 0 0 0 cos b
1
sin b
1
0 0 0

3
p
2
cos b
2

1
2
cos b
2
sin b
2
0 0 0

3
p
2
cos b
3

1
2
cos b
3
sin b
3
_

_
_

_
78
If we know the conguration of the mechanism (forward displacement) at any time, we can get a $
r
E
at this
conguration. As a result, the free motion, $
F
E
, of the end-eector at this conguration can be solved with
$
F
n1
E
$
r
E
0. If $
F
n1
E
0 0 0 0 0 0 , then the mechanism is controllable under this conguration;
otherwise, additional actuations should be added to steadily control the end-eector. Therefore, the CDOF
instantaneously depicts the actuations needed to steadily control the end-eector. In addition, the structure
parameters can also be optimized if we let $
F
n1
E
0.
J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504 1503
4. Conclusion
In this paper, we propose a programmable algorithm to investigate the CDOF of a spatial parallel mech-
anism in one xed Cartesian coordinate system with reciprocal screw theory. Based on the work of [6], we here
give a concrete algorithm to investigate the free motions and controllability of an end-eector with respect to
the congurations of the mechanism. Compared with [6] and [7], the most distinctive contributions of this
paper are that the free motions of the end-eector are instantaneously expressed in a set of Plu cker homoge-
neous coordinates in one Cartesian coordinate system and are most adapted to programming. This algorithm
can be utilized not only for mechanism analysis but also for mechanism synthesis and structure optimization.
The whole mathematical analysis process for CDOF can be directly embedded in advanced CAD software, in
which the imperfect Kutzbach Gru bler formulas are often utilized to analyze the DOF of a CAD mechanism.
References
[1] K.J. Waldron, G.L. Kinzel, Kinematics, Dynamics, and Design of Machinery, Hamilton Printing, John Wiley & Sons, New York,
1999, pp. 142.
[2] K.J. Waldron, The constraint analysis of mechanisms, Journal of Mechanisms 1 (1966) 101114.
[3] K.H. Hunt, Kinematic Geometry of Mechanisms, Oxford University Press, Oxford, 1978.
[4] J. Phillips, Freedom in machineryIntroducing Screw Theory, Vol. 1, Cambridge University Press, Cambridge, 1984.
[5] J. Phillips, Freedom in machineryScrew Theory Exemplied, Vol. 2, Cambridge University Press, Cambridge, 1990.
[6] J.-S. Zhao, K. Zhou, Z.-J. Feng, A theory of degrees of freedom for mechanisms, Mechanism and Machine Theory 39 (6) (2004) 621
643.
[7] J.-S. Zhao, K. Zhou, D.-Z. Mao, Y.-F. Gao, Y. Fang, A new method to study the degree of freedom of spatial parallel mechanisms,
The International Journal of Advanced Manufacturing Technology 23 (34) (2004) 288294.
[8] G. Gogu, Mobility of mechanisms: a critical review, Mechanism and Machine Theory 40 (9) (2005) 10681097.
[9] R.S. Ball, A Treatise on The Theory of Screws, Cambridge University Press, Cambridge, 1990 (First Paperback Edition 1998).
[10] J.R. Phillips, K.H. Hunt, On the theorem of three axes in the spatial motion of three bodies, Australian Journal of Exerted Science 15
(1964) 267287.
[11] Z. Huang, Q.C. Li, General methodology for type synthesis of symmetrical lower-mobility parallel manipulators and several novel
manipulators, The International Journal of Robotics Research 21 (2) (2002) 131145.
[12] V. Kumar, Instantaneous kinematics of parallel-chain robotic mechanisms, ASME Journal of Mechanical Design 114 (1992) 349358.
[13] J. Garc a de Jalo n, E. Bayo, Kinematic and dynamic simulation of multibody systems: the real-time challenge, Springer-Verlag,
NewYork, 1994.
[14] J.-S. Zhao, Z.-J. Feng, K. Zhou, J.-X. Dong, Analysis of the singularity of spatial parallel manipulator with terminal constraints,
Mechanism and Machine Theory 40 (3) (2005) 275284.
1504 J.-S. Zhao et al. / Mechanism and Machine Theory 41 (2006) 14861504