You are on page 1of 15

Forward kinematic problem of 5-RPUR parallel mechanisms (3T2R) with

identical limb structures


Mehdi Tale Masouleh
a,
, Clment Gosselin
a
, Manfred Husty
b
, Dominic R. Walter
b
a
Laboratoire de robotique, Dpartement de gnie mcanique, Universit Laval, Qubec, QC, Canada, G1V 0A6
b
University Innsbruck, Institute for Basic Sciences in engineering, Unit Geometry and CAD, Technikerstrae 13, A-6020 Innsbruck, Austria
a r t i c l e i n f o a b s t r a c t
Article history:
Received 21 May 2010
Received in revised form 16 February 2011
Accepted 17 February 2011
Available online 13 March 2011
This paper investigates the formulation of the forward kinematic problem of the 5-DOF parallel
mechanisms performing three translation and two independent rotations with identical limb
structures. The mathematical framework used in this paper is based on algebraic geometry
where the forward kinematic problem is explored in a seven-dimensional kinematic space by
the means of the so-called Study parameters. In this paper, the algorithms proposed for
obtaining the forward kinematic expression is based on two approaches which differ by their
concept of eliminating passive variables. The first one is based on the application of the so-called
resultant method and the second one, as a novel approach referred to as linear implicitization
algorithm, uses an elimination procedure based on solving several systems of linear equations.
The forward kinematic problem is solved using the homotopy continuation and it is revealed
that it has 1680 nite solutions. Moreover, the forward kinematic problem is studied in three-
dimensional Euclidean space where a class of simplied designs are proposed which admit a
simpler formulation for the forward kinematic problem. Finally, based on the simplied designs,
for a nearly general design a univariate expression of degree 220 is obtained.
2011 Elsevier Ltd. All rights reserved.
Keywords:
5-DOF parallel mechanisms
Three translations and two rotations motion
pattern (3T2R)
Forward kinematic problem (FKP)
Study parameters
Elimination method
Linear implicitization algorithm
1. Introduction
The kinematic analysis of parallel mechanisms requires a suitable mathematical framework in order to describe both
translation and rotation in a most general way. This can be achieved by resorting to algebraic geometry. Algebraic geometry is the
study of objects dened by polynomial equations, using algebraic means [1]. The origin of algebraic geometry is attributed to
Descartes who had two brilliant ideas: (1) using coordinates to describe points in Euclidean space and (2) describing curves and
surfaces by algebraic equations. Recently, the advent of computer algebra systems made it possible to implement many theories of
algebraic geometry using the algorithmic approaches [2,3]. Our interest toward applying algebraic geometry to the kinematic
analysis is mainly due to two reasons:
1. Using the superabundance of variables which eliminates the need to resort to trigonometric expressions and produces
homogeneous equations;
2. Algebraic geometry leads more naturally to a global understanding of all the solutions of a system of equations, as opposed to
nding only some of the solutions.
In this paper, the main concern is with obtaining the Forward Kinematic Expression (FKE) which is the mathematical
expression that pertains to nding the rigid-body pose(s) that a given limb, which is referred to as the principal limb, can reach for
a given actuator coordinate. As dealing with parallel mechanisms having identical limb structures, in this paper, we rst focus our
Mechanism and Machine Theory 46 (2011) 945959
Corresponding author at: Universit Laval, Laboratoire de robotique, Dpartement de gnie mcanique, 1065 Avenue de la mdecine, Qubec, QC, Canada, G1V
0A6. Tel.: +1 418 656 3474; fax: +1 418 656 7415.
E-mail address: mehdi.tale-masouleh.1@ulaval.ca (M. Tale Masouleh).
0094-114X/$ see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechmachtheory.2011.02.005
Contents lists available at ScienceDirect
Mechanism and Machine Theory
j our nal homepage: www. el sevi er. com/ l ocat e/ mechmt
attention to the kinematic modelling of one kinematic chain. As mentioned previously, this kinematic chain is referred to as the
principal limb and the xed and mobile frame are attached respectively to the rst and last joints of this chain with the direction of
the frames based on the DenavitHartenberg (DH) convention [4]. More precisely, we are taking a step back from the Forward
Kinematic Problem (FKP) which pertains to nding the rigid-body pose(s) that the platform can reach for a given set of actuator
lengths. The general approach toward obtaining a univariate expression for the FKE which should be free of passive variables, i.e.,
non-actuated joint-coordinates, is based on elimination theory, and the use of techniques such as the resultant [5,6] and Grbner
bases [79]. It should be noted that the FKP is solved in polynomial form when one has an effective and reliable method for
computing solutions, e.g. an algorithm to reduce the set of polynomials to a univariate.
In this paper, two approaches are proposed which both stemfromthe concepts of using Study's parameters but that differ from
their concept of elimination. In the rst approach, the elimination of passive variables is carried out by applying a direct inspection
of the obtained expressions and is based on resultant method. The main problemwith this approach is that the degree of eliminates
tends to explode when resultants are used or Grbner bases cannot be computed when the input equations are complicated. Thus
for the sake of obtaining a systematic approach and to circumvent the problem of spurious roots, in this paper, a novel approach,
called the Linear Implicitization Algorithm(LIA), is proposed based on the generation of linear polynomials which is accompanied
by the computation of the Grbner basis.
In short, our interest toward introducing and applying the LIA for obtaining the FKE is twofold:
1. Other existing algorithms, based on the resultant method, which require inspection and intuition, do not preclude the existence
of an expression with a lower degree which can pass through the quadrics S
6
2
, called Study quadric, and the constraint
expression, and contains all points which can be generated by the parametric expressions generated by the principal limb, i.e.,
the FKE;
2. As compared to elimination based on resultants, the approach is more systematic and does not result in extraneous expressions.
The results of this paper open an avenue to an original direction of research in the eld of the FKP of parallel mechanisms which
is the development of a systematic algorithm for the symbolic computation of the resultant of polynomials. In fact, the symbolic
derivation of the resultant of polynomials can be an onerous task and requires a difcult and non uniable inspection procedure
for the elimination sequences and may tend to be unwieldy for complex expressions.
In this paper, the 5-DOF parallel mechanisms with identical limb structures performing three translations and two
independent rotations (3T2R) are considered in order to apply the two proposed approaches, elimination in Study coordinates and
LIA, to obtain the corresponding FKE.
In general, 5-DOF parallel mechanisms are a class of parallel mechanisms with reduced degrees of freedomwhich, according to
their mobility, fall into three classes: (1) three translational andtworotational freedoms (3T2R), (2) three rotational andtwoplanar
translational freedoms (3R2T
p
) and (3) three rotational and two spherical translational freedoms (3R2T
s
) [10]. Geometrically,
the 3T2R motion can be made equivalent to guiding a combination of a directed line and a point on it. Accordingly, the 3T2R
mechanisms can be used in a wide range of applications for a point-line combination including, among others, 5-axis machine tools
[11,12], welding and conical spray-gun. In medical applications that require at the same time mobility, compactness and accuracy
around a functional point, 5-DOF parallel mechanisms can be regarded as a very promising solution [13].
As far as 5-DOF mechanisms with identical limb structures are concerned, researchers have mainly worked on the type
synthesis [10,1418]. Recently, the type synthesis of 5-DOF parallel mechanisms with identical limb structures was reviewed
[10,17,19] and Huang and Li [20] and Liu et al. in [21] proposed a rst architecture. It is worth noticing that most existing 5-DOF
parallel manipulators are built using a 5-DOF passive leg which constrains some actuated 6-DOF limbs [22,23].
To the best of the knowledge of the authors, up to now, very few kinematic studies have been conducted on 5-DOF parallel
mechanisms with identical limb structures [2430]. This is probably due to their short history. Thus, in this research, the FKP of
symmetrical 5-DOF parallel mechanism, more precisely 5-RPUR, is addressed which can be regarded as one of the most challenging
topics in the kinematics of parallel mechanisms. In this notation, ineP stands for an actuated prismatic joint, R for a revolute joint
and U for a universal joint.
The remainder of this paper is organized as follows. The architecture and the general kinematic properties of the 5-DOF parallel
mechanism under study which originated from the type synthesis performed in[10,17] are broadly outlined. In order to lay down
the essential concept for the projective space, Study's kinematic mapping is given. The FKE of the principal limb is obtained by
means of Study's kinematic mapping using resultant method and the so-called linear implicitization algorithm. Then the systemof
equations for the FKP, obtained by the latter two approaches in terms of Study's parameters is solved via a Homotopy Continuation
algorithm. Finally, two distinct classes of simplied designs are proposed whose FKP admit a univariate solution and for a nearly
general design a univariate expression is found.
2. Architecture
Figs. 1 and 2 provide respectively a schematic representation of a RPUR limb and a schematic representation for a 5-DOF
parallel mechanism that can be used to produce all three translational DOFs, plus two independent rotational DOFs (3T2R) of the
end-effector, namely (x, y, z, , ). In the latter notation, (x, y, z) represent the translational DOFs with respect to the xed frame O,
illustrated in Fig. 1, and (, ) stand respectively for the orientation DOFs around axes x and y. The mapping sequence between the
desired orientation of the platformand angles and is the rst rotation, , about the x-axis followed by the second rotation about
the y-axis by angle .
946 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
More generally, this mechanism consists of an end-effector which is linked by 5 identical limbs of the RPRRR type to a base. The
input of the mechanism is provided by the ve linear prismatic actuators. From the type synthesis presented in [17], the geometric
characteristics associated with the components of each leg are as follows: The ve revolute joints attached to the platform(the last R
joint ineachof the legs) haveparallel axes, the ve revolutejoints attachedtothebase have parallel axes, therst tworevolute joints of
each leg have parallel axes and the last two revolute joints of each leg have parallel axes. It should be noted that the second and third
revolute joints in each leg are built with intersecting and perpendicular axes and are thus assimilated to U joints. This mechanical
simplicationleads toa 5-RPURtypemechanism. As mentionedabove, the 5-RPURis a limited-DOF parallel mechanismandcontains a
constrained rotational DOF whichis in a direction parallel to e
1
e
2
, where e
1
and e
2
are unit vectors respectively along the axes of the
rst (and second) and third (and fourth) revolute joints of the ith leg, as illustrated in Fig. 1. Further results regarding the kinematic
properties, such as the solution of the IKP and the determination of the constant-orientation workspace can be found in [24].
3. Two approaches for formulating the FKE in seven-dimensional kinematic space
3.1. Study's kinematic mapping
Study's kinematic mapping is a mapping of an element of the Euclidean displacement group SE(3) into a 7-dimensional
projective space, P
7
[31]. The homogeneous coordinate vector of a point in P
7
is X=[x
0
: x
1
: x
2
: x
3
: y
0
: y
1
: y
2
: y
3
] where x=
[x
0
: x
1
: x
2
: x
3
] and y=[y
0
: y
1
: y
2
: y
3
]. The kinematic pre-image of X is the displacement described by the transformation
matrix
=
1

0 0 0
p x
2
0
+ x
2
1
x
2
2
x
2
3
2 x
1
x
2
x
0
x
3
2 x
1
x
3
+ x
0
x
2

q 2 x
1
x
2
+ x
0
x
3
x
2
0
x
2
1
+ x
2
2
x
2
3
2 x
2
x
3
x
0
x
1

r 2 x
1
x
3
x
0
x
2
2 x
2
x
3
+ x
0
x
1
x
2
0
x
2
1
x
2
2
+ x
2
3
_

_
_

_
1
where
p = 2 x
0
y
1
+ x
1
y
0
x
2
y
3
+ x
3
y
2
;
q = 2 x
0
y
2
+ x
1
y
3
+ x
2
y
0
x
3
y
1
;
r = 2 x
0
y
3
x
1
y
2
+ x
2
y
1
+ x
3
y
0
;
2
and =x
0
2
+x
1
2
+x
2
2
+x
3
2
. Note that x
i
, i =0, 3, are the so-called Euler parameters and they have to full an orthogonality
condition with the remaining four parameters y
i
, i =0, 3 [32]:
S
2
6
= x
0
y
0
+ x
1
y
1
+ x
2
y
2
+ x
3
y
3
= 0; 3
Fig. 1. Schematic representation of a RPUR limb.
947 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
and not all x
i
are zero. If these conditions are fullled [x
0
: : y
3
]
T
are called Study parameters of the displacement . Eq. (3) denes
a quadric S
6
2
P
7
and the range of the kinematic mapping is this quadric minus the three dimensional subspace dened by
E : x
0
= x
1
= x
2
= x
3
= 0: 4
S
6
2
is called Study quadric and E is the exceptional or absolute generator. One can normalize the parameters such that =1, then the
coordinate x
0
represents the cosine of the half rotation angle. Note that there are other possibilities to normalize, but these are not
used within this paper.
3.2. FKE via elimination in study coordinates
Now, we turn our attention to the kinematic modeling of a RPUR limb. As dealing with a parallel mechanismwith identical limb
structures, we rst focus our attention to the kinematic modeling of one kinematic chain, the principal limb. Then, having
determined the algebraic equations for the kinematic modeling of the principal limb, the kinematic modeling of its corresponding
parallel mechanism can be obtained.
Referring to Fig. 3 one can readily obtain the DH parameters of the principal limb which are given in Table 1 where
p
and l
p
stand for the prismatic elongation(assumed to be known) and leg length of the second moving link, respectively. Applying the DH
convention the two following transformation matrices are obtained:

i
=
1 0 0 0
0 cos u
i
sin u
i
0
0 sin u
i
cos u
i
0
0 0 0 1
_

_
_

_
i = 1; ; 4; 5

i
=
1 0 0 0
a
i
1 0 0
0 0 cos
i
sin
i
d
i
0 sin
i
cos
i
_

_
_

_
i = 1; 2; 3; 6
Table 1
DH parameters for a RPUR limb.
a
i
d
i

i
u
i
1
p
0 0 u
1
2 0 0

2
u
2
3 l
p
0
0
u
3
Fig. 2. Schematic representation of a 5-RPUR parallel mechanism.
948 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
where u
i
is the ith joint coordinate,
i
stands for the rotation about the x
i
-axis of the ith R joint and
i
represents the transformation
between successive local coordinate frames. In this paper, i is a local dummy variable. Thus the FKE for the principal limb with
respect to joint variables and the design variables can be expressed as follows:
F =
3
i =1

i
_ _

4
: 7
Reaching this step, the so-called kinematic mapping should be applied to F=[f
ij
]
i, j =1, , 4
which amounts to computing the
Study parameters, X, from matrix F. For that, Study in [33] presented a singularity-free procedure in which the homogeneous
quadruple x can be obtained from at least one set of four proportions [31]. Here, among the four proportions the following
proportion is chosen:
x
0
: x
1
: x
2
: x
3
= 1 + f
22
+ f
33
+ f
44
: f
43
f
34
: f
24
f
42
: f
32
f
23
: 8
Then, the four remaining Study parameters y=(y
0
: y
1
: y
2
: y
3
) can be computed from:
2y
0
= f
21
x
1
+ f
31
x
2
+ f
41
x
3
2y
1
= f
21
x
0
+ f
41
x
2
f
31
x
3
2y
2
= f
31
x
0
f
41
x
1
+ f
21
x
3
2y
3
= f
41
x
0
+ f
31
x
1
f
21
x
2
:
9
Upon substituting F into Eqs. (8) and (9), and applying the tangent half-angle substitution, t
i
= tan
u
i
2
_ _
, i =1, 2, 3, 4, it turns out
that all the computed Study parameters share a common factor which is:
1 + t
1
t
2
t
3
t
4
t
3
t
4
t
2
t
4
t
1
t
4
t
1
t
2
t
1
t
3
t
2
t
3
1 + t
2
1
_ _
1 + t
2
2
_ _
1 + t
2
3
_ _
1 + t
2
4
_ _ : 10
Using the fact that the above parameters are homogeneous allows to omit this common factor and one may obtain the
following for the rotational parameters of the Study parameters:
x
0
= 2 t
1
t
2
t
3
t
4
t
1
t
2
t
1
t
3
t
1
t
4
t
2
t
3
t
2
t
4
t
3
t
4
+ 1 ; 11
x
1
= 2 t
1
t
2
t
3
t
4
t
1
t
2
+ t
1
t
3
+ t
1
t
4
+ t
2
t
3
+ t
2
t
4
t
3
t
4
+ 1 ; 12
x
2
= 2 t
1
t
2
t
3
+ t
1
t
2
t
4
t
1
t
3
t
4
t
2
t
3
t
4
+ t
1
+ t
2
t
3
t
4
; 13
x
3
= 2 t
1
t
2
t
3
t
1
t
2
t
4
t
1
t
3
t
4
t
2
t
3
t
4
+ t
1
+ t
2
+ t
3
+ t
4
; 14
z
4
B
i
l
p
x
1
y
1
x
3
y
2
z
3
x
2
x
4
C
i

p
A
i
Fig. 3. Local system attached to each R joint for the DH parameters.
949 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
which consequently leads to the following for y=(y
0
: y
1
: y
2
: y
3
):
y
0
=
p
+ l
p
_ _
t
1
t
2
t
3
t
4
t
1
t
4
+ t
2
t
3
+ 1 +
p
l
p
_ _
t
1
t
2
t
1
t
3
+ t
2
t
4
t
3
t
4
; 15
y
1
=
p
+ l
p
_ _
t
1
t
2
t
3
t
4
t
1
t
4
+ t
2
t
3
1 +
p
l
p
_ _
t
1
t
2
t
1
t
3
+ t
2
t
4
+ t
3
t
4
; 16
y
2
=
p
+ l
p
_ _
t
1
t
2
t
3
t
2
t
3
t
4
t
1
+ t
4
+
p
l
p
_ _
t
1
t
2
t
4
+ t
1
t
3
t
4
+ t
2
+ t
3
; 17
y
3
=
p
+ l
p
_ _
t
1
t
2
t
3
+ t
2
t
3
t
4
+ t
1
+ t
4
+
p
l
p
_ _
t
1
t
2
t
4
t
1
t
3
t
4
t
2
+ t
3
: 18
The objective now is to combine all the above expressions, Eqs. (1118), by eliminating passive variables, i.e., t
i
, i =1, , 4, in
order to obtain the FKE of the principal limb with respect to Study's parameters and design and input parameters,
p
and l
p
. In fact,
a rigid body is fully described by six constraint equations in space. In this case, xing the input, P joint, results in a RUR kinematic
arrangement which has four DOFs. Hence the aim of the whole elimination procedure is to nd two independent constraint
equations for the FKE of one limb. In this paper, two well-known algorithms of elimination theory are used, namely the resultant
method and the Grbner base algorithm.
Before proceeding to the elimination of the passive variables, matrix F is expanded and it reveals that f
44
=0 which amounts to

44
=0 which is a component of , Eq. (1):
C = x
2
0
x
2
1
x
2
2
+ x
2
3
= 0: 19
Thus, before any elimination, a quadric, called constraint equation, is found which is free of any design parameters and can be
regarded as the rst constraint equation. This result is consistent with results found for the rotational parameters, x, in Eq. (1114)
whichare all free of any designparameters, i.e., l
i
and
i
and are all expressions injoint space, t
i
. Inother words, ve RPURlimbs share a
common rotational constraint as conrmed by the type synthesis performed for such a mechanism in [10]: The mechanism has a
rotational constraint about the axis dened by e
1
e
2
. Eq. (19) is a statement equivalent to the latter written in terms of Study's
parameters and indicates that the rotational constraint of this mechanismlies ona quadric. The existence of this quadric together with
the homogeneous condition, =1, and the Study quadric in P
7
, results in two independent rotational parameters for the mechanism.
Returning now to the FKE and Eqs. (1118), skipping mathematical details, upon applying a step by step resultant method to
Eqs. (1118) and eliminating successively the passive variables t
3
, t
2
, t
4
and t
1
, leads to three independent sixth degree
polynomials, {T
1
, T
2
, T
3
}, with respect to the Study and design parameters. In fact, all the points describing the variety V generated
by the principal limb lie on the these expressions, together with C and S
6
2
. The expressions for {T
1
, T
2
, T
3
} are rather long and are
presented in Appendix A. According to the dimension formula, the degree and number of obtained expressions are coherent with
the dimensions of the eliminated variables, v
i
, which is four. In summary, the FKE of the principal limb can be dened by the
following ideal
1
:
I = b T
1
; T
2
; T
3
; S
2
6
; C >; 20
i.e., three sixth degree polynomials, T
i
, and two quadric S
6
2
and C, the Study and the constraint quadric, respectively. The above
formulation channels us to apply a Grbner base algorithm [1], a powerful algebraic geometry algorithm, to the ideal I. From an
algebraic geometry standpoint, Grbner bases canbe viewedas anorganizedway to generate newpolynomials inthe ideal selecting a
subset that retains exactly the same solution set as the original polynomials, and determining a valid set of monomial identities that
complete the denition of an eigenvalue problem[34]. Upon computing the Grbner basis of the ideal I with respect to the monomial
order x
0
x
1
x
2
x
3
y
0
y
1
y
2
y
3
, and after several mathematical manipulations, the following expression is found:
F
p
X = 8
2
p
+ l
2
p
_ _
y
1
x
1
+ y
2
x
2

2
+ 8
2
p
l
2
p
_ _
y
1
x
2
y
2
x
1
x
0
y
3
+ y
0
x
3

+ 16 y
2
2
+ y
2
1
_ _
y
2
0
+ y
2
3
_ _
+
2
p
l
2
p
_ _
2
x
2
2
+ x
2
1
_ _
2
= 0:
21
The above polynomial is of degree four, instead of six for T
{i =1, 2, 3}
, and together with Eq. (19) and the Study quadric
expression could be regarded potentially as the simplest expression describing the FKE of a RPUR limb in terms of the Study
parameters, X. The latter issue will be conrmed formally using the LIA. Thus, reaching this step two expressions are in place,
Eqs. (19) and (21), which fully constrain the mechanism in space. In order to ensure the validity of Eq. (21), the back-solving can
be applied for the Study parameters found in Eqs. (1118) into F
p
(X) and observe that F
p
vanishes. It can be shown that F
p
(X) is
valid for all limb kinematic arrangements of symmetrical 5-DOF parallel mechanisms performing the 3T2R motion pattern with
prismatic actuators, such as PRUR for the so-called Pentapteron [25]. These kinematic arrangements are the results of type
1
The ideal I(F) of a system of polynomials F={f
1
, f
2
, , f
n
} is the set of all polynomials that can be formed by multiplying each f
i
by a polynomial g
i
and
summing them up: I(F)={h|h=g
1
f
1
+g
2
f
2
++g
n
f
n
}, where g
i
can be any polynomial in the same variables as F [1].
950 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
synthesis performed for symmetrical 5-DOF parallel mechanisms and a comprehensive list of them can be found in [10]. The
geometrical meaning of Eq. (21) is difcult to assess. However, in [24], it is shown that in the three-dimensional kinematic space
the vertex space for a given orientation and input of a RPUR limb generates a Bohemian dome. This quartic surface can be generated
by moving a circle that remains parallel to a plane along a curve that is perpendicular to the same plane. Thus for a given x;
i
; l
i
,
F
p
y can be interpreted geometrically as the Study mapping of a Bohemian dome, Fig. 4. In this mapping one should take into
account C, for the feasible orientation, and S
6
2
, the Study quadric.
3.3. FKE via Linear Implicitization Algorithm (LIA)
Inthis section, we introduceanalternativetoeliminationbasedonimplicitizationwhichis calledtheLIA. It is well knownthat there
exists a one-to-one correspondence from all spatial transformations to the Study quadric, S
6
2
, which lives in CP
7
. Particularly, this
means that a tuple of Study parameters describing a transformation is a projective point and consequently always only unique up to
scalar multiples. For a givenparametrizedtransformation(e.g. Eqs. (1118)) one canobtainthe set of corresponding points inCP
7
and
searchfor the smallest variety V CP
7
(withrespect toinclusion) whichcontains all these points. What is inplace about this variety at
the earlier stage? Denitely, it can be conrmed that its ideal consists of homogeneous polynomials and contains S
6
2
, i.e. the equation
for the Study quadric. In the following it is shown howadditional equations can be computed which are necessary to describe V, Eqs.
(1118). Fromthe above analysis, the transformation depending on four parameters t
1
, t
2
, t
3
, t
4
, whichis given by the matrix F, Eq. (7),
is in hand. The parameters come from the tan-half-angle substitution which was made to get rid of the trigonometric functions. This
method leads in general to four different 8-tuples which can be used to represent the Study parameters. Fortunately they are not
essentially different, they are all multiples of one special 8-tuple. Using Study's method to compute the Study parameters out of the
entries of a Euclidean displacement matrix Eqs. (8) and (9), one can obtain four seemingly different 8-tuples.
The next step consists in searching for homogeneous polynomials which vanish on all points producible from this
parametrization, i.e. polynomials in [x
0
, x
1
, x
2
, x
3
, y
0
, y
1
, y
2
, y
3
] which are 0 when these expressions are substituted. One possibility to
nd such necessary polynomials is the following: The ansatz is g c; x =
j j j =n
c
j
x
j
, which after the substitution becomes
f c; t =
jJ

j
c t
j
, where the coefcient functions f =
j
c are all linear. Due to the fact that f has to vanish for all values of the
t
i
, it has to be the zero polynomial. It follows that all coefcients of f themselves have to be 0. So after extraction of these
coefcients one obtains a system of linear equations where the unknowns are the
n + 7
n
_ _
coefcients from the general ansatz.
This system can be solved (assuming that the design parameters a
1
=
p
and a
3
=l
p
are generic) and the solution can be
substituted into the ansatz. The result is an expression r describing all homogeneous polynomials of degree n which vanish on the
points of V. An important point is that if the solution of the linear system is positive dimensional, the corresponding parameters
also appear in the nal expression, i.e. the expression r itself is parametrized. Now, we start with the smallest degrees.
n = 1 :
This case results in 16 linear equations in 8 unknowns. The solution of the systemis the zero vector and it follows that r=0, i.e.
there are no homogeneous linear polynomials which vanish on V.
n = 2 :
In this case, the problem is equivalent to 81 linear equations in 36 unknowns. The solution of the systemis now 2-dimensional
and for the nal expression r, one has:
r = x
0
y
0
+ x
1
y
1
+ x
2
y
2
+ x
3
y
3
p
1
+ x
2
0
x
2
1
x
2
2
+ x
2
3
_ _
p
2
: 22
2
1
0
5
2.5
0
-2.5
-5
-5
-2.5
0
2.5
5
-5
-2.5
0
2.5
5
Fig. 4. Study mapping of the vertex space of a RPUR limb, F (y), for a given orientation, (x
0
, x
1
, x
2
, x
3
)=(1, 1, 4, 4), and design parameters
p
=l
p
=1.
951 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
where p
i
are the parameters for the positive-dimensional solution of the linear system. The above means that all quadratic
expressions vanishing on V are a linear combinations of these two polynomials, q
1
and q
3
, and particularly that each of them alone
vanishes on V. Thus the rst two equations which are necessary to describe the constraint variety appear in this step.
n = 3 :
For this degree we get 256 linear equations in 120 unknowns. After solving the system we get a 16-dimensional solution
and:
r =
16
i =1
c
i
p
i
; 23
where the c
i
are again polynomials in the Study parameters. Most of them look like e.g. y
3
(x
0
y
0
+x
1
y
1
+x
2
y
2
+x
3
y
3
), i.e. they do
not introduce new information. To nd out which of these 16 polynomials lead to something new, all are reduced with a Grbner
basis computed from the equations from the previous case. The result is that all are reduced to 0, i.e. all cubic polynomials
vanishing in V can be generated from the known quadratic ones.
n = 4 :
Once again the same procedure with 625 linear equations in 330 unknowns. After solving the systemyields to a 72-dimensional
solution:
r =
72
i =1
c
i
p
i
; 24
and polynomials in the Study parameters as the coefcients c
i
. After reduction of them with the two quadrics, q
1
and q
2
, using the
Grbner basis from above exactly one non-zero polynomial remains. This means that we get here a new polynomial, q
3
, which is
not a combination of the known ones. Until now we have the following describing equations for the variety V:
q
1
= x
0
y
0
+ x
1
y
1
+ x
2
y
2
+ x
3
y
3
S
2
6
; 25
q
2
= x
2
0
x
2
1
x
2
2
+ x
2
3
C; 26
q
3
=8
2
p
+ l
2
p
_ _
y
1
x
1
+ y
2
x
2

2
+ 8
2
p
l
2
p
_ _
y
1
x
2
y
2
x
1
x
0
y
3
+ y
0
x
3

+ 16 y
2
2
+ y
2
1
_ _
y
2
0
+ y
2
3
_ _
+
2
p
l
2
p
_ _
2
x
2
2
+ x
2
1
_ _
2
F
p
:
27
Due to the fact that we had four parameters t
i
and the fact that the dimension of the variety generated by these equations is
now also 4, we stop the process here and use equations q
1
, q
2
, q
3
as the describing equations of the constraint variety.
4. Solving the general FKP expressed in seven-dimensional kinematic space
The rst step is to obtain the system of equations representing the FKP, i.e., the FKE corresponding to four other limbs, j =2,
, 5. Thus a transformation in the Study parameters X should be made in both base and moving frame and can be done by following
the procedure described in [31]. Consider b
j
= b
0j
: : b
7j
_ _
and m
j
= m
0j
: : m
7j
_ _
as the Study parameters describing the jth
limb placement in the base (xed frame) and in the moving platform (mobile frame), respectively. Since the axes of the R joints
xed to the base, e
1
, are all parallel thus b
j
consists of a pure translation. This has the following consequences for the Study
parameters of this transformation:
b
0j
= b
1j
= b
2j
= b
3j
= b
4j
= 0: 28
The same issue follows for axes of the R joints attached to the platform, e
2
, and leads to:
m
0j
= m
1j
= m
2j
= m
3j
= m
4j
= 0: 29
These particularities for the axes attached to the base and platformshould not be interpreted as mechanical simplications and
is originating from the type synthesis preformed for such mechanisms. Thus based on the transformation matrices presented in
[31], one may obtain:
j
T
b
=
I
44
0
44
B
j
I44
_ _
;
j
T
m
=
I
44
0
44
M
j
I44
_ _
; 30
952 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
so that B
j
and M
j
are the following skew-symmetric matrices:
B
j
=
0 b
5j
b
6j
b
7j
b
5j
0 b
7j
b
6j
b
6j
b
7j
0 b
5j
b
7j
b
6j
b
5j
0
_

_
_

_
; M
j
=
0 m
5j
m
6j
m
7j
m
5j
0 m
7j
m
6j
m
6j
m
7j
0 m
5j
m
7j
m
6j
m
5j
0
_

_
_

_
:
31
Thus the transformation becomes:
X
j
=
j
T
m
j
T
b
_ _
1
X; 32
which once expanded leads to:
X
j
=
x
T
b
5j
+ m
5j
_ _
x
1
+ b
6j
+ m
6j
_ _
x
2
+ b
7j
+ m
7j
_ _
x
3
+ y
0
b
5j
m
5j
_ _
x
0
+ b
7j
m
7j
_ _
x
2
+ b
6j
+ m
6j
_ _
x
3
+ y
1
b
6j
m
6j
_ _
x
0
+ b
7j
+ m
7j
_ _
x
1
+ b
5j
m
5j
_ _
x
3
+ y
2
b
7j
m
7j
_ _
x
0
+ b
6j
m
6j
_ _
x
1
+ b
5j
+ m
5j
_ _
x
2
+ y
3
_

_
_

_
: 33
Thus the FKE of the jth limb, F
j
, can be found as follows:
F
j
= F
p
XX
j
; pj
_ _
: 34
FromEq. (33) it follows that the Study rotational parameters, (x
0
: x
1
: x
2
: x
3
), remain the same for all limbs. In summary, the FKP
for the mechanism is equivalent to 8 equations and 8 unknowns:
G =
F
p
= 0
F
j
= 0 j = 2; ; 5
S
2
6
= 0
C = 0
= 1
_

_
35
For the FKP, the above systemof equations should be solved for X with design values of: b
j
; m
j
; l
j
;
j
_ _
. Although the degree of
each of the FKE is reduced from 6 to 4, T
i
, i =1, 2, 3, versus F
p
, the contribution to the total degree of the system is reduced even
more dramatically, fro 6
3
to 4
1
. However, the degree is still too high to apply the resultant method to Ginorder to nd a univariate
expression for the FKP. This channels us to use numerical algebraic geometry which can be regarded as the intersection of
algebraic geometry and numerical analysis. A number of numerical algebraic algorithms have been implemented in Bertini
[34,35]. Bertini is a software for solving polynomial systems using the Homotopy Continuation approach. Using Bertini, it follows
that the FKP of a general 5-RPUR admits 1680 nite solutions, real and complex. Numerous randomexamples in solving the set of
Eqs. (35) reveal that the FKP of this 5-DOF parallel mechanism may admit up to 208 real solutions among the 1680 nite
solutions. However, this is not necessarily the upper bound for the number of real solutions because this result was obtained
through simple numerical trials. Providing an upper bound for the number of real solutions requires the development of a
univariate expression for the FKP which is an extremely complicated task. Obtaining 1680 nite solutions and also 208 real
solutions leads to conclude that arriving at a univariate expression for the FKP of this parallel mechanism should be extremely
difcult especially considering the complexity of deriving a univariate expression for the forty solutions of the FKP of the Gough
Stewart platform [36].
In order to ensure the validity of the solutions obtained by Bertini all the real solutions are back substituted into the IKP
expression found in [24] which is expressed in terms of the three-dimensional kinematic space. This test reveals that for all
solutions converted to Cartesian coordinates and Euler angles the same set of prismatic actuator elongations can be found. The
transformation of Study parameters into Cartesian coordinates and Euler angles is elaborated in [37].
953 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
Applying the so-called witness set method, Bertini has the ability to detect positive dimensional solution sets. Using this feature
of Bertini, it can be conrmed that regardless of the inputs and geometric parameters of the mechanism, the following complex
sets are always a solution of Eq. (35):
x
0
= 0 x
3
= 0 x
1
= ix
2
y
1
=iy
2
x
0
= 0 x
3
= 0 x
1
= ix
2
y
1
= iy
2
x
1
= 0 x
2
= 0 x
0
= ix
3
y
0
=iy
3
x
1
= 0 x
2
= 0 x
0
= ix
3
y
0
= iy
3
:
36
As the equations are linear, the dimension formula shows directly that each of the lines in Eq. (36) represents a three-
dimensional complex space. The above solutions lie in these four three-dimensional spaces and the solution set itself is then the
intersection of this three-dimensional linear space with the Study quadric. Because this set is always complex, it can never
contribute to the real solutions of the FKP regardless of the chosen design parameters. However, algebraically the above set of
solutions are a valid solution for the FKP.
In summary, using the Study kinematic mapping, the FKP of the general 5-RPUR mechanismis reduced to a systemof equations
G of degree 4 whose compactness and simplicity make it suitable to use Bertini. It should be noted that due to the fact that F
p
is of
degree 4 then this systemof equations, on the basis of Bezout's theorem
2
[38,39], has an upper bound for the number of solutions of
4
5
22=4096 (four times F
p
, C, S
2
6
, and =1) while using T
i ={1, 2, 3}
, instead of F
p
, leads to 6
5
22=31, 104. This upper
bound corresponds to the number of paths to be tracked by the homotopy continuation in Bertini and a reduction of this number
decreases the computation time for Bertini.
5. FKP in three-dimensional Euclidean space
Referring to Fig. 1, a xed reference frame O-xyz is attached to the base of the mechanismwith i, j and k as its unit vectors and a
moving reference frame or mobile frame, is attached to the moving platform. In this paper the superscript for a vector stands for a
representation in the mobile frame. In the ith leg, the motion of the actuated prismatic joint is measured with respect to a
reference point A
i
(x
Ai
, y
Ai
, z
Ai
), located on the axis of the rst revolute joint, by the joint coordinate
i
which is the signed distance
between point A
i
and a reference point B
i
(x
Bi
, y
Bi
, z
Bi
) on the U joint. Vector e
i
is in turn dened as a unit vector in the direction of
the prismatic joint and therefore the vector connecting point A
i
to point B
i
and can be written as
i
=
i
e
i
. Vector r
i
(a constant
vector) is dened as the position vector of point A
i
in the xed reference frame. Similarly, vector s
i
is the vector connecting point O
of the platform to a reference point C
i
(x
Ci
, y
Ci
, z
Ci
) on the axis of the last revolute joint of the ith leg. Point B
i
is dened as the
intersection of the axes of the second and third revolute joints of the ith leg. Vector v
i
is the vector connecting point B
i
to point C
i
.
Finally, the position of the platform is represented by vector p = x; y; z
T
connecting point O to point O and the orientation of the
moving frame with respect to the xed frame is given by a rotation matrix Q.
It should be noted that Q cannot be prescribed arbitrarily since the mechanism has only two degrees of rotational freedom.
Therefore, a rotation matrix consistent with the orientation capabilities of the mechanism must be chosen. In fact, the motion
capabilities of the mobile platformboil down to the position and orientation of a line attached to the mobile platform. Based on the
results presented in [24], this rotation matrix can be written as:
Q =
cos sin sin cos sin
0 cos sin
sin sin cos cos cos
_
_
_
_
: 37
For a given value of the angles and , matrix Q is readily computed and vectors s
i
and e
2
are then obtained as:
s
i
= Qs
0
i
; e
2
= Qe
0
2
: 38
With reference to Fig. 1, the following equations, arising from the kinematic constraint of each limb, can be written:
x
Bi
x
Ai

2
+ z
Bi
z
Ai

2
=
2
i
; 39
x
Ci
x
Bi

2
+ y
Ci
y
Bi

2
+ z
Ci
z
Bi

2
= l
2
i
; 40
x
Ci
x
Bi
cos z
Ci
z
Bi
sin = 0; i = 1; ; 5; 41
where the rst two equations represent, respectively, the magnitude of
i
and v
i
and the last one corresponds to the kinematic
constraints between e
2
and v
i
, i.e., e
2
v
i
.
As the complete solution of the problem in three-dimensional kinematic space is quite challenging, numerous approaches in
the literature and practice were proposed including the use of numerical procedures, simplifying the mechanism by the
2
The intersection of m homogeneous algebraic equations in n unknowns of degree n
1
, n
2
, , n
m
is constituted of at most
i = m
i = 1
n
i
points it points exist at all.
954 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
coalescence of some of the connection points on the platform or the base and, nally, to use some extra sensors. Similarly, in this
project, simplifying the mechanismby the coalescence of some of the connection-points is considered for solving the FKP with the
aim of obtaining a simpler design, reducing the mechanical interferences and increasing the workspace volume.
From the results obtained in [24], for the IKP and the FKP for a particular case, one could extend the latter study, for the FKP
resolution, to the following conclusion: Any mechanical simplications that provides the coordinate of two pairs of U joints
explicitly or a relation between them leads to a univariate expression for the FKP.
The foregoing issue remains central to the development of the simplied designs having a closed-form solution for the FKP.
Having the above conclusion in mind, consider two limbs, i and j, for which:
1. The connection points at the base, A
i
and A
j
, are in a plane with e
1
as normal or coincide;
2. Both second moving links have the same leg length, l
i
=l
j
, or coincide;
3. The connection points at the platform, C
i
and C
j
, are aligned with e
2
, or coincide.
From the IKP presented in [24], it follows that the design conditions result in a simplied arrangement in which both second
moving links are parallel, v
i
v
j
, or coincided, and the line connecting B
i
to B
j
is aligned with e
2
, or again coincided. Consequently, in
such a design there is a pair of U joints having a relation between them, i.e., the line connecting them is aligned with e
2
which
implies that:
x
Bj
x
Bi
= s
j
s
i
_ _
cos ; 42
z
Bj
z
Bi
= s
i
s
j
_ _
sin ; 43
where s
i
is a geometric parameter of the platform representing the component of s
i
along x, the vector connecting the limb to the
mobile frame, i.e., C
i
to O. Therefore, in a design for which two pairs of limbs fulll the above three conditions, on the basis of the
above conclusion, then the FKP admits a closed-form solution. There are three distinct situations, S={A
1
, A
2
A
3
}, in which the
above three conditions can occur and they are depicted in Fig. 5. Therefore, all second order subsets of S adopt a univariate
expression for their FKP:
A
1
A
1
f g; A
1
A
2
f g; A
1
A
3
f g; A
2
A
3
f g; A
2
A
2
f g; A
3
A
3
f g f g: 44
For instance, {A
1
A
2
} is a 5-DOF parallel mechanism which consists of two combined arrangements A
1
and A
2
plus a regular
limb of the RPUR type. Fig. 6 shows a CAD model of a A
1
A
1
parallel mechanism. As it can be observed, the U joints in A
1
arrangements are designed in such a way that there is an offset between the axes, which leads to a reduction in the mechanical
interferences and hence a large travel for the joints. In such a situation, although the joint is no longer a U joint per se, the
kinematic model given above remains valid. From Eq. (44) it follows that {A
3
A
3
}, Fig. 2, is free of any coalescence of connection
points and comparing with other designs presented above could be regarded as the most general design having a closed-form
solution for the FKP. The FKP of the designs belonging to Eq. (44) are treated in [29,30,40]. In the following section the FKP of a
design containing one arrangement of type A
i
is investigated.
x
v
i j
e
2
z

i
B
i j
C
i j
A
j
A
i
(a)
1
z
x
e
2
e
2
A
i j
C
i
C
j
B
j
B
i
v
j
e
2
v
i
(b)
2
x
z
B
i
A
i A
j
e
2
C
j
C
i
B
j
v
j
v
i
(c)
3
Fig. 5. Simplied kinematic arrangements.
955 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
5.1. FKP for a design containing one arrangement of type A
i
In such a design only one simplied arrangement belonging to S is considered. No matter which one is chosen the approach
toward obtaining a univariate expression for the FKP remains similar. Let us consider A
1
in order to describe the approach and
Fig. 7 depicts, notationally, such a design. As it can be observed from the latter gure, the xed frame is attached to O
(x, y, x)
B
12
,
whose coordinates are known for the FKP, and the direction of the principal axes are such that the x-axis is in the direction of e
2
,
the y-axis remains in the direction of e
1
and obviously the z-axis becomes in the direction of e
1
e
2
. The origin of the mobile
frame coincides with C
12
. For this case, the constraint equations, Eqs. (3941), are re-formulated with respect to the new frames.
The constraint equations for A
1
can be expressed as:
F
1
= y
2
+ z
2
l
2
12
= 0; 45
x
B12
= z
B12
= x = 0: 46
For the three regular limbs, i =3, 4, 5, which are free of mechanical simplications, one has:
y
Ci
y
Bi

2
+ z
Ci
z
Bi

2
l
2
i
= 0; 47
x
Bi
x
Ai

2
z
Bi
z
Ai

2

2
i
= 0; 48
x
Ci
x
Bi
= 0: 49
The coordinates of the passive variables C
i
can be computed as follows:
x
Ci
; y
Ci
; z
Ci

T
= p + Q

i
; 50
with:
Q

=
1 0 0
0 cos sin
0 sin cos
_
_
_
_
: 51
Due to special considerations for the xedframe, i.e., OB
12
, the coordinates of the xedRjoint, A
i
, for the three regular limbs are
no longer knownparameters and depend on the orientationof the xed frame around the y-axis, namely . This implies that:
x
Ai
; y
Ai
; z
Ai

T
= Q
T

r
B
i
; i = 3; 4; 5; 52
i = 1
i =3 i = 4
i = 5
y
i = 2
z
x
Fig. 6. Schematic representation of a {A
1
A
1
} parallel mechanism.
956 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
where
Q

=
cos 0 sin
0 1 0
sin 0 cos
_
_
_
_
: 53
In the above, r
B
i
stands for the vector connecting B
12
to A
i
expressed in a frame similar to the one used in Fig. 1 but originating
from B
12
.
The rst step is to nd the FKE of each limb. For the arrangement A
1
the FKE is already known, Eq. (45). Moreover, from
Eq. (46) it follows that x=0. Thus, one unknown for the FKP is readily in place. For the sake of obtaining the simplest expression
describing the FKE of the three others limbs, i =3, 4, 5, one should subtract Eq. (45) from Eq. (47). Then, taking the resultant of the
latter expression with Eq. (48), by substituting Eq. (49), with respect of z
Bi
yields a second degree expression, called F
i
, i =3, 4, 5, for
y and z. This expression is of degree 4 in degree with respect to t = tan

2
_ _
and u = tan

2
_ _
and as a whole the nal expression is
of degree 10. Although it seems that the degree of the expressions is too high to proceed to the resultant method, upon considering
an appropriate sequence of elimination together with some simplications and factoring in each step of elimination it is possible to
obtain a univariate expression for t, namely F
t
:
F
34
= Res F
3
; F
4
; u = 0;
F
45
= Res F
4
; F
5
; u = 0;
F
341
= Res F
34
; F
1
; y = 0;
F
451
= Res F
45
; F
1
; y = 0;
F
t
= Res F
451
; F
341
; z = 0:
54
In the above, Res(a, b, c) stands for the resultant of polynomials a and b with respect to the common variable c. Thus, F
t
is the
univariate expression with respect to t which consists of 7 polynomial factors. Only one of these factors corresponds to the FKP and
this can be found using Bertini. To do so, the systemof equations constituted by Eq. (45) and Eqs. (4749) with respect to variables
(y, z, t, u, z
B3
, z
B4
, z
B5
) is given to Bertini, which reveals that all solutions found by Bertini correspond to the solution of only one of
the seven factors of F
t
. This expression is of degree 220 which allows us to conclude that the FKP for this design can be expressed by
a univariate expression of degree 220 with respect to t. Moreover, using the FKE expressed in term of Study parameters, Eq. (21),
for this nearly general design results in 1680 nite solutions which appears to be inconsistent with the univariate expression of
degree 220. This inconsistent number of solutions is due to the fact that Eq. (21) does not take into account the fact that there is a
A
1
arrangement, i.e., a coalescence for two U joints in B
12
. Thus all the 1680 solutions, complex and real, should be veried in such
a way that the two limbs belonging to the A
1
arrangement, here i =1, 2, have the same working modes. Upon applying the latter
condition the number of solutions reduces from 1680 to 220 solutions which all correspond to the roots of the univariate
polynomial of degree 220 found above. Handling the value of t, nding other unknowns is just a question of back substitution.
6. Conclusion
This paper investigatedthe FKP of 5-DOF parallel mechanisms (3T2R) with a leg kinematic arrangement of type RPUR. By means of
Study's kinematic mapping and based on the two proposed approachesresultant method and the so-called linear implicitization
Fig. 7. Schematic representation for a design containing one arrangement of type A
1
.
957 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
algorithmthe FKP is explored in a seven-dimensional projective space. Study's kinematic mapping made it possible to obtain a
simpler expressionfor theFKEwhichis of degree4, F
p
(X) withtwo quadratic expressions, S
6
2
andC, as constraint expressions. Basedon
the LIAit was conrmed that these three polynomials are the simplest expressions whichdescribe the FKE. Using Bertini, a package for
solving polynomial systems using homotopy continuation, the FKP was solved for a general design and it revealed that it has 1680
nite solutions where for a given design 208 real solutions were found. Using the three-dimensional Euclidean space, six simplied
designs were proposed whose FKP can be expressed by a univariate expression. Moreover, the FKP of one of those designs was solved
in its most general case by reducing the mechanical simplications to one design condition, Fig. 7. The principles of this paper can
be applied equally well to the other types of symmetrical 5-DOF parallel mechanisms developed through the type synthesis, such as
the 5-PRUR, in order to obtain similar results for the FKP. Ongoing work includes the solution of the FKP in a univariate form for a
general design of 5-RPUR parallel mechanism and the optimum synthesis of the mechanism by considering singularities.
Acknowledgement
The authors would like to acknowledge the nancial support of the Natural Sciences and Engineering Research Council of
Canada (NSERC) as well as the Canada Research Chair program. The authors would also like to thank Ian Tremblay for his help in
the preparation of the CAD graphical.
Appendix A. Three Expressions T
i =1, 2, 3
Three expressions for the FKE of a RPUR limb:
T
1
= 16y
2
0
x
2
2
y
2
1
+ 16y
2
0
x
2
1
y
2
1
+ 16y
2
3
x
2
2
y
2
1
+ 16y
2
0
x
2
3
y
2
3
+ 16y
2
3
x
2
1
y
2
1
+ 16y
2
0
y
2
3
x
2
0
+ 32y
3
0
y
1
x
0
x
1
+ 32y
3
0
y
3
x
0
x
3
+ 32y
3
3
x
3
x
1
y
1
+ 32y
3
3
y
0
x
0
x
3
+ 8y
2
0
x
3

2
p
x
2
x
1
x
0
+ 32y
0
y
1
y
2
3
x
0
x
1
8
2
p
x
3
2
y
3
x
0
y
1
8y
3

2
p
x
2
x
2
1
y
1
x
0
+ 32y
2
0
x
3
y
3
x
1
y
1
+ 16y
4
0
x
2
0
+ 16y
4
3
x
2
3
8
2
p
x
1
y
0
y
3
x
2
0
x
2
+
4
p
x
4
2
x
2
3
+
4
p
x
4
2
x
2
0
+ x
2
0
l
4
p
x
4
2
+ l
4
p
x
4
2
x
2
3
8y
2
0

2
p
x
2
2
x
2
0
8
2
p
x
2
2
x
2
3
y
2
3
+ x
2
1

4
p
x
2
3
x
2
2
+
4
p
x
2
2
x
2
0
x
2
1
+ x
2
0
l
4
p
x
2
1
x
2
2
+ l
4
p
x
2
1
x
2
2
x
2
3
8l
2
p
y
2
0
x
2
0
x
2
2
8l
2
p
y
2
3
x
2
2
x
2
3
2l
2
p

2
p
x
4
2
x
2
3
2
2
p
x
2
0
l
2
p
x
4
2
+ 8x
3
2
x
3

2
p
y
1
y
0
+ 8
2
p
x
2
1
y
0
x
3
x
2
y
1
8y
2
3
x
3

2
p
x
2
x
1
x
0
16x
2
2
x
3

2
p
y
3
x
0
y
0
+ 8
2
p
x
1
y
0
x
2
3
x
2
y
3
8l
2
p
y
2
0
x
0
x
1
x
2
x
3
8l
2
p
y
1
y
0
x
2
1
x
2
x
3
8l
2
p
y
1
y
0
x
3
2
x
3
+ 8l
2
p
y
3
y
0
x
2
0
x
1
x
2
+ 8l
2
p
y
3
y
1
x
0
x
3
2
+ 8l
2
p
y
3
y
1
x
0
x
2
1
x
2
+ 8l
2
p
y
2
3
x
0
x
1
x
2
x
3
16l
2
p
y
3
y
0
x
0
x
2
2
x
3
8l
2
p
y
3
y
0
x
1
x
2
x
2
3
2l
2
p

2
p
x
2
1
x
2
2
x
2
3
2
2
p
x
2
0
l
2
p
x
2
1
x
2
2
A:1
T
2
= 16y
2
2
x
2
1
y
2
1
+ 32y
3
y
2
y
2
1
x
2
x
3
+ 8
2
p
x
3
0
y
3
x
1
y
2
8y
2
1
x
3

2
p
x
2
x
1
x
0
+ 8y
2
2
x
3

2
p
x
2
x
1
x
0
16
2
p
x
1
x
2
0
x
2
y
1
y
2
8
2
p
x
3
y
2
x
2
2
x
0
y
1
+ 8
2
p
x
3
y
2
x
2
1
y
1
x
0
8x
3
0

2
p
y
3
x
2
y
1
8
2
p
x
2
3
y
3
x
0
x
2
y
1
+ 8
2
p
x
2
3
y
2
y
3
x
1
x
0
+ 8l
2
p
y
2
1
x
0
x
1
x
2
x
3
+ 16y
2
2
x
2
2
y
2
1
+ 16y
2
1
y
2
3
x
2
0
+ 16y
2
1
x
2
3
y
2
3
+ 16y
2
2
x
2
3
y
2
3
+ 16y
2
2
y
2
3
x
2
0
+ x
4
0

4
p
x
2
2
+ x
4
0

4
p
x
2
1
+ x
4
0
l
4
p
x
2
1
+ x
4
0
l
4
p
x
2
2
+ 16y
4
1
x
2
1
+ 16y
4
2
x
2
2
8l
2
p
y
3
y
2
x
3
0
x
1
2x
2
0

2
p
l
2
p
x
2
2
x
2
3
2x
2
0

2
p
l
2
p
x
2
1
x
2
3
8l
2
p
y
2
2
x
0
x
1
x
2
x
3
16l
2
p
y
1
y
2
x
2
0
x
1
x
2
2x
4
0

2
p
l
2
p
x
2
1
2x
4
0

2
p
l
2
p
x
2
2
+ 8l
2
p
y
1
y
2
x
0
x
2
2
x
3
8l
2
p
y
1
y
2
x
0
x
2
1
x
3
+ 8l
2
p
y
3
y
1
x
3
0
x
2
+ 8l
2
p
y
3
y
1
x
0
x
2
x
2
3
8l
2
p
y
3
y
2
x
0
x
1
x
2
3
+ 32y
1
y
3
2
x
1
x
2
+ 32y
3
1
y
2
x
1
x
2
+ 32y
3
1
x
3
y
3
x
1
+ 32y
3
y
3
2
x
2
x
3
+
4
p
x
2
3
x
2
0
x
2
1
+
4
p
x
2
3
x
2
2
x
2
0
8y
2
1

2
p
x
2
0
x
2
1
8y
2
2

2
p
x
2
2
x
2
0
+ x
2
0
l
4
p
x
2
1
x
2
3
+ x
2
0
l
4
p
x
2
2
x
2
3
8l
2
p
y
2
1
x
2
0
x
2
1
+ 32y
2
2
x
3
y
3
x
1
y
1
8x
2
0
l
2
p
y
2
2
x
2
2
A:2
T
3
= 16y
2
0
x
2
3
y
2
3
+ 16y
2
0
y
2
3
x
2
0
+ 32y
3
0
y
3
x
0
x
3
+ 32y
3
3
y
0
x
0
x
3
8y
2
0

2
p
x
2
0
x
2
1
8y
2
0
x
3

2
p
x
2
x
1
x
0
+ 32y
2
y
3
0
x
0
x
2
+ 16y
4
0
x
2
0
+ 16y
4
3
x
2
3
+ 8
2
p
x
1
y
0
y
3
x
2
0
x
2
+ 8l
2
p
y
2
y
0
x
1
x
2
2
x
3
+ 32y
2
y
2
3
y
0
x
0
x
2
+ 32y
2
0
x
2
y
2
x
3
y
3
8x
2
2
y
2
x
1
x
3

2
p
y
0
8
2
p
x
3
1
y
0
y
2
x
3
16
2
p
x
2
1
y
0
x
3
y
3
x
0
+ 8x
3
1

2
p
y
3
x
0
y
2
+ 8
2
p
x
2
2
y
3
x
0
x
1
y
2
+ 8l
2
p
y
2
y
0
x
3
1
x
3
8l
2
p
y
2
y
3
x
0
x
3
1
16l
2
p
y
3
y
0
x
0
x
2
1
x
3
8l
2
p
y
2
y
3
x
0
x
1
x
2
2
2l
2
p

2
p
x
4
1
x
2
3
8l
2
p
y
2
3
x
2
1
x
2
3
8y
2
0
l
2
p
x
2
0
x
2
1
8x
2
1

2
p
x
2
3
y
2
3
+ x
2
1

4
p
x
2
3
x
2
2
+
4
p
x
2
2
x
2
0
x
2
1
+ x
2
0
l
4
p
x
2
1
x
2
2
+ l
4
p
x
2
1
x
2
2
x
2
3
+ 8y
2
3
x
3

2
p
x
2
x
1
x
0
8
2
p
x
1
y
0
x
2
3
x
2
y
3
+ 8l
2
p
y
2
0
x
0
x
1
x
2
x
3
8l
2
p
y
3
y
0
x
2
0
x
1
x
2
8l
2
p
y
2
3
x
0
x
1
x
2
x
3
+ 8l
2
p
y
3
y
0
x
1
x
2
x
2
3
2l
2
p

2
p
x
2
1
x
2
2
x
2
3
2
2
p
x
2
0
l
2
p
x
2
1
x
2
2
2
2
p
x
2
0
l
2
p
x
4
1
+ 32y
3
3
x
2
y
2
x
3
+ 16y
2
0
x
2
1
y
2
2
+ 16y
2
0
x
2
2
y
2
2
+ 16y
2
3
x
2
1
y
2
2
+ 16y
2
3
x
2
2
y
2
2
+ x
4
1

4
p
x
2
0
+ x
4
1

4
p
x
2
3
+ l
4
p
x
2
0
x
4
1
+ l
4
p
x
4
1
x
2
3
A3:
958 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959
References
[1] D.A. Cox, J.B. Little, D. O'Shea, Using algebraic geometry, Springer Verlag, 2005.
[2] K. Brunnthaler, Synthesis of 4R Linkages Using Kinematic Mapping, Ph.D. thesis, Institute for Basic Sciences in Engineering, Unit Geometry and CAD,
Innsbruck, Austria (December 2006).
[3] M. Pfurner, Analysis of Spatial Serial Manipulators Using Kinematic Mapping, Ph.D. thesis, University Innsbruck, Institute for Engineering Mathematics,
Geometry and Computer Sciences, Innsbruck, Austria (October 2006).
[4] J. Denavit, R. Hartenberg, A kinematic notation for lower-pair mechanisms based on matrices, transaction of ASME, Journal of Applied Mechanics 22 (1)
(1955) 215221.
[5] C. Innocenti, Forward kinematics in polynomial form of the General Stewart platform, ASME, Journal of Mechanical Design 123 (2) (2001) 254260.
[6] T. Lee, J. Shim, Forward kinematics of the General 66 Stewart platform using algebraic elimination, Mechanism and Machine Theory 36 (9) (2001)
10731085.
[7] D. Gan, Q. Liao, J.S. Dai, S. Wei, L. Seneviratne, Forward displacement analysis of the General 66 Stewart mechanism using Grobner bases, Mechanism and
Machine Theory 44 (9) (2009) 16401647.
[8] D. Lazard, On the representation of rigid-body motions and its application to generalized platform manipulators, Computational Kinematics, CK, 1, 1993,
pp. 175181.
[9] D. Lazard, Stewart platform and Grbner basis, Advances in Robot Kinematics, Kluwer Academic Publishers, 1992, pp. 136142.
[10] X. Kong, C. Gosselin, Type Synthesis of Parallel Mechanisms, Vol.33, Springer, Heidelberg, 2007.
[11] F. Gao, B. Peng, H. Zhao, W. Li, A Novel 5-DOF Fully Parallel Kinematic Machine Tool, The International Journal of Advanced Manufacturing Technology 31 (1)
(2006) 201207.
[12] ParallelMic, http://www.parallemic.org/.
[13] O. Piccin, B. Bayle, B. Maurin, M. de Mathelin, Kinematic modeling of a 5-DOF parallel mechanism for semi-spherical workspace, Mechanism and Machine
Theory 44 (8) (2009) 14851496.
[14] Z. Huang, Q.C. Li, General methodology for type synthesis of symmetrical lower-mobility parallel manipulators and several novel manipulators, International
Journal of Robotics Research 21 (2) (2002) 131145.
[15] Y. Fang, L.W. Tsai, Structure synthesis of a class of 4-DoF and 5-DoF parallel manipulators with identical limb structures, International Journal of Robotics
Research 21 (9) (2002) 799810.
[16] Z. Huang, Q.C. Li, Type synthesis of symmetrical lower-mobility parallel mechanisms using the constraintsynthesis method, International Journal of Robotics
Research 22 (1) (2003) 5979.
[17] X. Kong, C. Gosselin, Type synthesis of 5-DOF parallel manipulators based on screw theory, Journal of Robotic Systems 22 (10) (2005) 535547.
[18] S.J. Zhu, Z. Huang, Eighteen fully symmetrical 5-DoF 3R2T parallel manipulators with better actuating modes, The International Journal of Advanced
Manufacturing Technology 34 (3) (2007) 406412.
[19] J.P. Merlet, Parallel robots open problems, 9th Int. Symp. of Robotics Research, 1999, pp. 2732.
[20] Z. Huang, Q. C. Li, A Decoupled 5-DoF Symmetrical Parallel Mechanism, Patent pending, China, No 01122274.3.
[21] Q. Jin, T.L. Yang, A.X. Liu, H.P. Shen, F.H. Yao, Structure synthesis of a class of 5-DOF parallel robot mechanisms based on single opened-chain units,
Proceedings of the 2001 ASME Conferences DETC2001/DAC21153, Pittsburgh, PA, 2001.
[22] J. Wang, C. Gosselin, Kinematic analysis and singularity representation of spatial ve-degree-of-freedomparallel mechanisms, Journal of Robotic Systems
14 (12) (1997) 851869.
[23] T. Mbarek, I. Barmann, B. Corves, Fully parallel structures with ve degree of freedom: systematic classication and determination of workspace, Proceedings
Mechatronics & Robotics, 2004, pp. 990995.
[24] M. Tale Masouleh, C. Gosselin, Kinematic analysis and singularity representation of 5-RPRRR parallel mechanisms, Fundamental Issues and Future Research
Directions for Parallel Mechanisms and Manipulators, 2008, pp. 7990, Montpellier, France.
[25] C. Gosselin, M. Tale Masouleh, V. Duchaine, P.L. Richard, S. Foucault, X. Kong, Parallel mechanisms of the multipteron family: kinematic architectures and
benchmarking, IEEE International Conference on Robotics and Automation, 2007, pp. 555560, Roma, Italy.
[26] M. Tale Masouleh, M. H. Saadatzi, C. Gosselin, H. D. Taghirad, A Geometric Constructive Approach for the Workspace Analysis of Symmetrical 5-PRUR Parallel
Mechanisms (3T2R), in: Proceedings of the 2010 ASME Design Engineering Technical Conferences, DETC2010-28509, 2010.
[27] M. Tale Masouleh, C. Gosselin, Singularity Analysis of 5-RPRRR Parallel Mechanisms via Grassmann Line Geometry, in: Proceedings of the 2009 ASME Design
Engineering Technical Conferences, DETC2009-86261, 2009.
[28] M. Tale Masouleh, M. Husty, C. Gosselin, A general methodology for the forward kinematic problemof symmetrical parallel mechanisms and application to
5-PRUR parallel mechanisms (3T2R), Proceedings of the 2010 ASME Design Engineering Technical Conferences, ASME, Montreal, Canada, 2010,
DETC2010-28222.
[29] M. Tale Masouleh, Kinematic Analysis of Five-DOF (3T2R) Parallel Mechanisms with Identical Limb Structures, Ph.D. thesis, Laval University, Quebec, QC,
Canada (October 2010).
[30] M. Tale Masouleh, C. Gosselin, M. Saadatzi, X. Kong, H. Taghirad, Kinematic Analysis of 5-RPUR (3T2R) Parallel Mechanisms, Meccanica 46(1) (2011) 131146.
[31] M.L. Husty, H.-P. Schrcker, Algebraic geometry and kinematics, in: I.Z. Emiris, F. Sottile, T. Theobald (Eds.), Nonlinear Computational Geometry, 2007,
pp. 85106.
[32] M. Husty, A. Karger, Self-motions of GrifsDuffy type parallel manipulators, IEEE International Conference on Robotics and Automation, Vol. 1, 2002,
pp. 712.
[33] E. Study, Von den Bewegungen und Umlegungen, Mathematische Annalen 39 (1891) 441566.
[34] A. Sommese, C. Wampler, The Numerical Solution of Systems of Polynomials Arising in Engineering and Science, World Scientic Pub. Co. Inc, 2005.
[35] D.J. Bates, J.D. Hauenstein, A.J. Sommese, C.W. Wampler, Bertini: software for numerical algebraic geometry, Available at http://www.nd.edu/-sommese/bertini.
[36] M.L. Husty, An algorithm for solving the direct kinematics of general StewartGough platforms, Mechanism and Machine Theory 31 (4) (1996) 365379.
[37] M. Tale Maosouleh, M. Husty, C. Gosselin, Forward kinematic problem of 5-PRUR parallel mechanisms using study parameters, Advances in Robot
Kinematics: Motion in Man and Machine, Springer, 2010, pp. 211221.
[38] R. Walker, Algebraic Curves, Vol.2, SpringerVerlag, 1950.
[39] J.-P. Merlet, Algebraic-Geometry Tools for the Study of Kinematics of Parallel Manipulators, in: J. Angeles, G. Hommel, P. Kovacs (Eds.), Computational
Kinematics, Kluwer Academic Publishers, 1993, pp. 183194.
[40] M. Tale Masouleh, C. Gosselin, M.H. Saadatzi, H.D. Taghirad, Forward kinematic problem and constant orientation workspace of 5-RPRRR (3T2R) parallel
mechanisms, 18th Iranian Conference on Electrical Engineering (ICEE), IEEE, 2010, pp. 668673.
959 M. Tale Masouleh et al. / Mechanism and Machine Theory 46 (2011) 945959

You might also like