You are on page 1of 30

Mechanism

and
Mechanism and Machine Theory 40 (2005) 10681097
Machine Theory
www.elsevier.com/locate/mechmt

Mobility of mechanisms: a critical review


Grigore Gogu *

Laboratory of Research and Applications in Advanced Mechanics, Department of Machines,


Mechanisms and Systems, French Institute of Advanced Mechanics and Blaise Pascal University,
Campus de Clermont-Ferrand l les Cezeaux, BP 265, Aubiere Cedex, France

Received 16 September 2003; received in revised form 10 September 2004; accepted 28 December 2004
Available online 2 March 2005

Abstract

This paper presents a critical review on the calculation of the mobility, the main structural and kinematic
parameter of a mechanism. We focus on a brief presentation and a critical analysis of various methods pre-
sented in the literature in the last 150 years, to clearly situate the dierent contributions to this very impor-
tant subject of the theory of mechanisms. Thirty ve approaches/formulas for mobility calculation are
presented and their genesis, similarities and limitations are investigated. The various methods proposed in
the literature for mobility calculation are grouped in two categories: approaches based on setting up the kine-
matic constraint equations and their rank calculation and formulas for a quick calculation of mobility with-
out need to develop the constraint equations. We emphasize on the limits of formulas for quick calculation
of mobility by applying them to a parallel robot with elementary legs and ascertaining that the results are
erroneous. In fact, the formulas for quick calculation of mobility known in the literature do not t for many
classical or modern mechanisms. We explain why these formulas do not work for certain mechanisms and we
propose a new formula for quick mobility calculation of parallel mechanisms with elementary legs.
 2005 Elsevier Ltd. All rights reserved.

Resume

Cet article presente une analyse critique de letat de lart dans le calcul de la mobilite, le plus important
parametre structural et cinematique dun mecanisme. Nous focalisons sur une breve presentation et une

*
Tel.: +33 4 73 28 80 22; fax: +33 4 73 28 81 00.
E-mail addresses: gogu@ifma.fr, grigore.gogu@ifma.fr

0094-114X/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechmachtheory.2004.12.014
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1069

analyse critique des dierentes methodes presentees dans la litterature dans les derniers 150 ans, pour bien
situer les dierentes contributions dans ce sujet tres important de la theorie des mecanismes. Trente cinq
approches/formules sont presentees et leur genese, similarites et limitations sont investiguees. Les methodes
proposees dans la litterature pour le calcul de la mobilite sont groupees en deux categories: approches fon-
dees sur le calcul du rang du systeme dequations des contraintes cinematiques et formules pour une deter-
mination rapide de la mobilite sans passer par les equations des contraintes. Nous soulignons les limites de
ces formules en les appliquant sur un robot parallele a jambes elementaires et en constatant que les resultats
sont errones. En fait, les formules pour une determination rapide de la mobilite connues dans la litterature
ne sont pas applicables a plusieurs mecanismes classiques ou modernes. Nous expliquons pourquoi ces
formules ne sont pas valables pour certains mecanismes et nous presentons une nouvelle formule pour le
calcul de la mobilite des mecanismes paralleles a jambes elementaires.
 2005 Elsevier Ltd. All rights reserved.

1. Introduction

Mobility is the main structural parameter of a mechanism and also one of the most fundamental
concepts in the kinematic and the dynamic modelling of mechanisms. IFToMM terminology denes
the mobility or the degree of freedom as the number of independent co-ordinates needed to dene the
conguration of a kinematic chain or mechanism [1]. Mobility (M) is used to verify the existence of a
mechanism (M > 0), to indicate the number of independent parameters in the both kinematic and
the dynamic models and to determine the number of inputs needed to drive the mechanism.
Earlier works on the mobility of mechanisms go back to the second half of the XIX century to
Chebychev [3], Sylvester [4], Grubler [5,6], Somov [7] and Hochman [8]. During the XX century,
sustained eorts were made to nd general methods for the determination of the mobility of any
rigid body mechanism. Various formulas and approaches were derived and presented in the liter-
ature by Koenigs [9], Grubler [10,11], Malytshe [12], Kutzbach [13], Dobrovolski [14,15], Art-
obolevski [16], Moroskine [17,18] Voinea and Atanasiu [19], Kolchin [20], Rossner [21], Boden
[22], Manolescu and Manafu [23], Ozol [24], Hunt and Phillips [25], Waldron [26], Manolescu
[27], Bagci [28], Antonescu [29,30], Freudenstein and Alizade [31], Hunt [32], Herve [33,34], Baker
[35,36], Gronowicz [37] , Davies [3840], Agrawal and Rao [41,43], Angeles and Gosselin [44],
Dudita and Diaconescu [45], Fanghella and Galletti [46,47], Fayet [4850] Tsai [51], McCarthy
[52]. Contributions have continued to emerge in the last years: Huang et al. [53], Rico and Ravani
[54], Rico, Gallardo and Rawani [55].
In the calculation of mechanism mobility the following key controlling parameters are generally
used: the number of joints (p) and the number of kinematic elements (n = m  1, by m is denoted
the total number of elements including the xed base), the mobility (f) and the degree of constraint
(c) of the joint, the motion parameter (b) and the constraint parameter (d = 6  b) of the mech-
anism, the number of independent closed loops of the mechanism (q).
The various methods proposed in the literature for mobility calculation of the closed loop
mechanisms can be grouped in two categories:

(a) approaches for mobility calculation based on setting up the kinematic constraint equations
and their rank calculation for a given position of the mechanism with specic joint location,
1070 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

(b) formulas for a quick calculation of mobility without need to develop the set of constraint
equations.

The approaches for mobility calculation based on setting up the kinematic constraint equations
and their rank calculation are valid without exception. The major drawback of these approaches
is that the mobility can not be determined quickly without setting up the kinematic model of the
mechanism. Usually this model is expressed by the closure equations that must be analyzed for
dependency. There is no way to derive information about mechanism mobility without perform-
ing position/velocity analysis by using analytical tools (screw system theory, linear algebra, ane
geometry, Lie algebra, etc). For this reason, the real and practical value of these approaches is
very limited in spite of their valuable theoretical foundations. The rank of constraint equations
is calculated in a given position of the mechanism with specic joint location. The mobility cal-
culated in relation with a given conguration of the mechanism is an instantaneous mobility
which can be dierent from general mobility (global mobility/gross mobility). Global mobility
has a single value for a given mechanism. This is a global parameter characterizing the mechanism
in all congurations excepting singular ones. Instantaneous mobility is a local parameter charac-
terizing the mechanism in a given conguration including singular ones. In a singular congura-
tion the instantaneous mobility is dierent from global mobility.
Formula for a quick calculation of mobility is an explicit relationship between structural
parameters of the mechanism: the number of links and joints, the motion/constraint parameters
of joints and of mechanism. Usually, these structural parameters are easily determined by inspec-
tion without need to develop the set of kinematic constraint equations.
Today we can note that all known formulas for a quick calculation of mobility do not t for
many classical mechanisms as, for instance, the mechanisms proposed by Roberval [56], Sarrus
[57], Delassus [5860], Bennett [61], Bricard [62], Myard [63], Goldberg [64], Altman [65], Baker
[66], Waldron [67], Baker et al. [68], or for many recent parallel robots. Special geometric condi-
tions play a signicant role in the determination of mobility of these mechanisms. When these
mechanisms were limited to special examples, considered as curiosities, they were called: para-
doxical mechanisms [62], paradoxical chains [69], [33], [34], [70], exceptions [71], special cases [72],
linkage with a paradox between practical degrees of freedom and computed degrees of free-
dom [73], overconstrained yet mobile linkage [7476], linkages with anomalous mobility [76].
The formulas known in the literature for the determination of mobility are not applicable to
many types of recent parallel robots, see for example the robots Delta [77], Star [78], H4 [79],
Orthoglide [80], CPM [81] and other parallel manipulators presented in the literature [82,83] or re-
cently proposed by the author [84]. We can not continue to consider the mechanisms that do not t
diverse methods or equations when determining mobility as having structural aws. We have to
review the formulas and to consider the aw to be that of the formula rather than of the mecha-
nism. The limits of these formulas are established and a new formula is proposed, in the last part
of the paper, for quick calculation of mobility of fully parallel mechanisms with elementary legs.

2. Critical review of mobility calculation

As presented in Section 1, several dozen of formulas/approaches have been proposed in the last
150 years for the calculation of the mechanism mobility. Many of these methods are reducible to
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1071

the same originated formula. The previous works on the state of the art in mobility calculation
[16,22,8588] have mentioned various formulas but they have not analysed their genesis and sim-
ilarities. In this section, we focus on a brief presentation and a critical analysis of these formulas/
approaches to clearly situate the dierent contributions in this very important subject of the
theory of mechanisms. Thirty ve contributions are critically reviewed. The original notation
of certain authors was not maintained to ensure a certain unity of presentation and to facilitate
the comparison between dierent formulas.
We apply each formula to determine the degree of mobility of a recent fully parallel manipu-
lator with elementary legs, the parallel Cartesian robotic manipulator CPM [8183], presented
in Fig. 1, which only contain revolute (R) and prismatic (P) joints. A serial kinematic chain is
associated with each elementary leg linking the mobile platform to the xed base. The mechanism
has n = 10 kinematic elements (links), p0 = 3 prismatic joints adjacent to the xed element, pn = 9
revolute joints non-adjacent to the xed element and q = 2 independent closed loops. Each loop
contains the same number and the same type of joints. The three legs are identical from a struc-
tural point of view. In each leg the direction of the prismatic joint and the axes of the revolute
joints are parallel (PkRkRkR). In [81] it is mentioned that this mechanism has three degrees of
freedom in spite of the fact that the general degrees-of-freedom equation presented by Tsai [89]
predicts that the mechanism has zero degrees of freedom. We analyse the results obtained by using
the dierent formulas to determine the degree of mobility of CPM robotic manipulator and nally
we will conclude on the limits to these formalisations.

2.1. Chebychevs contribution

In the second half of the XIX century, the mathematician Chebychev published an article (con-
cerning 4-bar linkage) in two parts in the proceedings of the Russian Royal Academy of Science of
Saint-Petersburg. Chebychev was the rst scientist who proposed a mathematical formalisation
for the calculation of mechanism mobility. The rst part of the article [2] was published under

Fig. 1. Parallel Cartesian robotic manipulator (PCM): (a) kinematic chain; (b) associated graph.
1072 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

the impulse of Watts invention, known at the end of the XVIII century. In this article, Chebychev
applied his method of polynomial function approximation to nd the lengths of the elements of
the 4-bar mechanism capable of describing a given couple curve with the least error. In the second
part of the article [3], Chebychev proposed the rst formula for the calculation of the number
of independent variables in a mechanism. This contribution was quickly integrated in the rst
treatises on the theory of mechanisms published at the end of the XIX century [90,91].
Chebychev expressed his formula in the form:
3n  2p0 pn 1; 1
where 3n represents the number of variables required to describe the position and the orientation
of the n kinematic (mobile) bars in the plane and 2(p0 + pn) is the number of constraint equations
imposed by the p(p0 + pn) revolute joints of the mechanism that can be adjacent (p0) or non-adja-
cent (pn) to the xed base. It is known that each revolute joint introduces two constraint equations
in a planar mechanism. Chebychev applied this formula to elementary planar mechanisms with
p0 = 2 and to complex planar mechanisms (p0 = 3) having only revolute joints and one degree
of mobility. He did not extend the use of Eq. (1) to other types of mechanisms. This formula could
be extended in the form
M 3n  2p0 pn 2
to any planar mechanism with one degree of freedom joints (helical, prismatic and revolute
joints).
The well-known k 4-bar mechanisms were obtained by Chebychev, to describe couple curves
containing a straight line segment, by using Eq. (1).
Supposing that we try to extrapolate Eq. (2), we obtain M = 6 for the mechanism presented in
Fig. 1 (with n = 10, p0 = 3 and pn = 9). This result is obviously erroneous, Chebychevs formula (2)
being limited to the planar mechanisms (mechanisms in which the axes of all the revolute joints
are parallel and perpendicular to the plane containing the directions of the prismatic pairs).

2.2. Sylvesters contribution

In 1874, Sylvester [4] presented a modied form of Eq. (1) as a structural condition for one
degree of freedom pin-connected planar mechanisms:
3m  2p  4 0: 3
In Eq. (3), m is the total number of elements of a mechanism including the xed element and the
kinematic elements. Sylvester stated Eq. (3) in the following words: In order for a combination of
links to full this so to say fatalistic condition of constrained motion and to entitle it to the name
of a linkage in the speakers sense, a numerical relation must be satised between the number of
links and the number of joints, viz., three times the number of links must be four greater than
twice the number of joints. In applying this rule it must be understood that, if three links are
jointed together, the junction counts for two joints; if four are jointed together, for three joints;
and so on. We can see that Eq. (3) can be obtained by simply replacing n by m  1 in the Cheby-
chevs formula and in the general case joints connecting a links count as a  1 joints.
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1073

2.3. Grublers contribution

In 1883, Grubler [5,6] presented a structural condition for one degree of freedom planar mech-
anisms identical with the Eq. (3). Later, Grubler proposed a structural condition for one degree of
freedom spatial complex mechanisms with helical joints [10,11]:
5h  6m 7 0; 4
where h is the total number of helical joints.

2.4. Somovs contribution

In 1887, Somov proposed the following structural condition for one degree of freedom
mechanisms [7]:
m  qb  1 2: 5
Somov introduced for the rst time the motion parameter (b), called mobility number, for planar
(b = 3) and for spatial mechanisms (b = 6).

2.5. Hochmans contribution

Hochman made an important contribution to the structural analysis of mechanisms in his book
of kinematics of machinery [8] published in 1890. A detailed presentation of Hochmans contribu-
tion can be found in [45] and in [87].
He proposed a relation between the total number of elements m, the total number of joints p
and the number of independent closed loops q in a complex mechanism (m + q  p = 1)Euler
formula in the theory of graphs.
He also proposed various structural conditions for the existence of one degree of freedom
mechanisms:

bm  1  C 1; 6

F  qb 1; 7

bp  q  C 1 8
with
X
b1
C iC bi ; 9
i1

X
b
F 1iF bi ; 10
i1

F C pb; 11
1074 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

where C is the total number of constraints imposed by the joints and Fthe total number of de-
grees of freedom of the joints. C bi represents the number of joints having the degree of constraint
cbi i and F bi the number of pairs having the degree of connectivity fib i. Hochman dened the
motion parameter b as the mechanism category given by the number of elementary existing mo-
tions between any kinematic element of the mechanism and the xed element or between any two
kinematic elements (b < 6). He marked the dierence between degree of constraint of a joint in the
motion space with dimension b, denoted by cb = b  fb, and the degree of constraint of a joint in
the motion space with dimension b = 6, also called joint class and denoted by c = 6  f.
We can see that Eq. (6) extends Eq. (1) proposed by Chebychev. For the general case of a mech-
anism having M degrees of freedom, Eqs. (6)(8) become
M bm  1  C; 12

M F  qb; 13

M bp  q  C: 14
Hochman considered that a mechanism in which M = 1 has determined motions, if M < 1 the
mechanism is a xed structure without relative motions and if M > 1 the motions of the mecha-
nisms are not determined. We can see that Hochman limited the mechanisms with determined
motions to one degree of freedom mechanisms.
Hochman considered that Eqs. (12)(14) are applicable to any kind of mechanism having va-
rious types of joints and the same b for each independent loop (b = 3 for planar and spherical
mechanisms, b = 6 for spatial mechanisms).
The greatest contribution of Hochman consisted in including the mechanism category between
the fundamental structural parameters of a mechanism. Unfortunately, he did not indicate any
method to determine the mechanism category b.
By applying Eqs. (12)(14) to the mobility calculation of the mechanism presented in Fig. 1,
with m = 11, q = 2, p = 12, we obtain: M = 6 (if we presume b = 3 and consequently
C = 12 2 = 24), M = 4 (if b = 4 and C = 12 3 = 36), M = 2 (if b = 5 and C = 12 4 = 48) and
M = 0 (if b = 6 and C = 12 5 = 60).
We can see that the results obtained for this mechanism are erroneous for all values presumed
for b(b = 3, . . . , 6). It is obvious that the mechanism presented in Fig. 1 has the same b for all the
closed loops, but the results obtained by using the formulas proposed by Hochman are erroneous.
So, we can say that Eqs. (12)(14) do not work for all mechanisms even if the independent loops
have the same motion coecient.

2.6. SomovMalytshes formula

SomovMalytshe formula [7,12]


X
5
M 6n  iC i 15
i1

represents a particular case of Eqs. (12) and (9) proposed by Hochman (with the notations
n = m  1 and C 6i C). Eq. (15) is applicable only to spatial mechanisms when b = 6.
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1075

2.7. Koenigs formula

In 1905, Gabriel Koenigs, published an introduction to a new theory of mechanisms [9] in


which he presented an equation for the calculation of the degrees of freedom of a mechanism
M 6m  1  C; 16
where 6(m  1) is the number of parameters necessary to dene the relative positions of the
n = m  1 kinematic elements of the mechanism and C is the number of independent constraint
equations given by the joints. We can see that Eq. (16) represents the particular case of
Eq. (12), proposed by Hochman, in which b = 6.

2.8. Kutzbachs mobility equation

The mobility equation proposed by Kutzbach [13]


X
p
M 6  dm  1  6  d  fi 17
i1

also represents the particular case of Eq. (12) for a spatial mechanism with b = 6  d, fi6 fi and
X
p
C 6  d  fi : 18
i1

In the literature, Eq. (17) is considered to be applicable to all mechanisms having the same d for
each independent loop.
By applying Eqs. (17) to the mobility calculation of the mechanism presented in Fig. 1 (for
which m = 11, p = 12 and fi = 1, i = 1, 2, . . . , 12), we obtain: M = 6 (if we presume d = 3),
M = 4 (if d = 2), M = 2 (if d = 1) and M = 0 (if d = 0). This shows that the results obtained for
the mechanism analysed are also erroneous for all values presumed for d(d = 0, . . . , 3). So, we
can say that Eq. (17) does not work for all mechanisms even if the independent loops have the
same motion coecient.

2.9. Dobrovolskis mobility equation

Dobrovolski named the constraint parameter d as the mechanism family and indicated the
following mobility equation [14,15]:
X
5
M d 6  dn  i  dC i ; 19
id1

where C i C 6d
i represents the number of joints having the degree of constraint c6d
i i  d. We
can see that Eqs. (19) and (12) are similar by taking into consideration that n = m  1 and
X
5
C i  dC i : 20
id1
1076 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

Dobrovolski considered d = 6  b = 0, . . . , 4, but he did not indicate any method for the deter-
mination of d.
Eq. (19) written in the form
X
6d1
M d 6  dm  1  6  d  iF i 21
i1

is also known as the ArtobolevskiDobrovolski mobility equation [28]. In Eq. (21), F i F 6d i is
the number of pairs having the degree of connectivity fi6d i. Eqs. (19) and (21) are also consid-
ered in the literature to be applicable to any mechanism having the same d for each independent
loop.
By applying Eqs. (19) and (21) to the mobility calculation of the mechanism presented in Fig. 1
we obtain the same erroneous results that were obtained by using Eqs. (17) and (12): M = 6 (if we
presume d = 3), M = 4 (if d = 2), M = 2 (if d = 1) and M = 0 (if d = 0). We can see that Eqs. (19)
and (21) do not work for all mechanisms, even if the independent loops have the same motion
coecient, as with Eqs. (12) and (17) from which they can be derived.

2.10. Contribution of Moroskine

To calculate the mobility of a planar or spatial mechanism, Moroskine [17,18] proposed the
formulas:
M N  r; 22

X
5
M iC i  r; 23
i1

where N is the total number of scalar kinematic parameters of the mechanism, rthe rank of the
linear homogeneous set of equations dening the kinematic constraints, Cithe number of joints
of class i (joints with the degree of freedom 6  i). The two formulas proposed by Moroskine are
valid without exception, but these formulas do not give to us a quick calculation of mobility with-
out the need to develop the kinematic equations and to calculate their rank. These calculations
will be presented in Section 3 of this article.

2.11. Contribution of Voinea and Atanasiu

Voinea and Atanasiu [19] proposed the following formulas for mobility calculation of single-
loop closed mechanisms:
M N  r N  r1 r2  we ; 24
where N is the total number of degrees of freedom in the joints given by the total number of kine-
matic joints with a connectivity of one (revolute, prismatic and helical joints) equivalent to the
joints of the mechanism, r is the rank of the screw system equivalent to the joints of the single-
loop mechanism, r1 and r2 are the ranks of the screw systems equivalent to the two open subchains
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1077

connecting a kinematic element e to the xed element. In the single-loop mechanism the kinematic
element e has the connectivity we related to the xed base.
To calculate the mobility of a complex mechanism, Voinea and Atanasiu [19] proposed the
formula:
X
q
M N rj pp ; 25
j1

where rj is the rank of the screw system of joints with connectivity one equivalent to the joints of
the jth independent closed loop of the mechanism and pp is the total number of passive joints in
mechanism. To calculate the rank rj, Voinea and Atanasiu used the theory of the instantaneous
screw axis proposed by Ball [92]. An important contribution was set up by Voinea and Atanasiu
by identifying the geometrical congurations of screw systems with 1 6 r 6 6 and by proposing an
analytical method for computing the motion parameter bj = rj [19,93].
By applying Eq. (25) to the mobility calculation of the mechanism presented in Fig. 1 (for which
N = p = 12, r1 = r2 = 5 and pp = 0) we obtain M = 2, which is an erroneous result. We can see that
all the closed loops of the mechanism are identical and have the same rank (rj = 5).
Eq. (25) is the rst mobility formula applicable to the mechanisms having independent closed
loops with dierent values for the motion coecient bj = rj, but this formula does not work for all
complex mechanisms, even if rj is the same for all independent loops.

2.12. Kolchins mobility equation

Kolchin [20] modied ArtobolevskiDobrovolski mobility equation (21) and expressed it in the
form:
X
6d1
M 6m  1  6  iF i dq: 26
i1

Kolchins contribution was also summarised in [94].


By applying Eq. (26) to the mobility calculation of the mechanism presented in Fig. 1, we
obtain the same erroneous results that were obtained by using Eqs. (12), (17), (19) and (21).

2.13. Rossners contribution

The mobility equation proposed by Rossner [21]


Xp
M fi  6p  m  6 27
i1

also represents a particular case of Eq. (13) for the spatial mechanism with b = 6 (by taking into
consideration Eulers formula q = p  m + 1).
Rossner also mentioned the existence of so called half open kinematic chain containing at
least one link which must have at least four joints. The half open kinematic chain is also known
in the literature as kinematic chain with fractionated freedom.
1078 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

2.14. Bodens mobility equation

Boden [22] extended Eq. (27) proposed by Rossner to make it applicable to spatial mechanisms
(b = 6) containing planar loops (b = 3):
Xp
M fi  6p  m  6 D; 28
i1

where
X
z
D 3pEk  mEk 3: 29
k1

z is the number of distinct directions of the joint axes belonging to planar loops, pEk and mEk are
the number of joints and the number of elements in the planar loops having the axes of the rev-
olute joints in kth direction. D takes into consideration the planar loops existing in the spatial
mechanism.

2.15. Manafus formula

In order to take into consideration the fact that the degree of mobility of a complex kinematic
chain depends on the choice of the independent closed loops, Manolescu and Manafu [23]
proposed the following mobility equation, presented as Manafus mobility equation [27]:
!
Xq
M max Mi  Mc ; 30
i1

where Mi is the mobility of the elementary closed loop i and Mc is the mobility of the connecting
mechanism (the mechanism formed with the common joints between dierent closed loops). To
calculate the degree of mobility of an elementary closed loop, Manolescu and Manafu used
Dobrovolskis mobility equation (19). The mobility of the connecting mechanism is given by
the sum of the degrees of freedom of the common joints.
Each elementary closed loop of the mechanism presented in Fig. 1 has d = 1 and Mi = 3
(i = 1,2). We can see that any two elementary closed loops have four common joints (Mc = 4).
So, by using Eq. (30) we obtain the erroneous value of M = 2 for the mobility of the mechanism
presented in Fig. 1.

2.16. Ozols formula

Ozols formula [24]


X
5
M 6  cC c  6q 31
c1

represents a particular
P case of Eqs. P (13) and (10) proposed by Hochman (for b = 6 and the nota-
tion F 6i F i and 5c1 6  cC c 5i1 iF i ) where Cc represents the number of joints of class c
and Fi the number of pairs having the degree of connectivity fi = i.
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1079

2.17. Contribution of Waldron

To calculate the mobility of a closed loop, Waldron [26] proposed the formula:
M F  b; 32
which also represents a particular case of Eqs. (13) and (10) proposed by Hochman (for q = 1).
Waldron called b the order of the equivalent screw system of the closed loop and he dened it
as the order of the screw system of a series chain containing the same joints as the loop in the same
geometrical relationship. To calculate b, Waldron used the theory of the instantaneous screw axis
proposed by Ball [92] and developed by Voinea and Atanasiu [19,93] and Phillips and Hunt
[25,95].
Waldron [26] delineates the series and parallel laws for screw system theory and shows how to
determine both the relative freedom between particular links and the mobility of certain class of
multi-loop linkages in which some members (elements) are shared by two or more loops. He
shows that such linkages may be handled by considering one loop to be a complex joint in the
other. In such cases the values of b applicable to each closed loop may vary with the order
in which the closed loops are considered. He emphasised on the importance of relative freedom
between two particular members of a linkage in mobility calculation. An example is given in
[26] for the mechanism known as the De Roberval Scale.

2.18. Contribution of Manolescu

Manolescu and Manafu also proposed [23] the mobility equation:


X
5 X
q
M 6n  iC i max dj 33
i1 j1

that extends formula (15), proposed by SomovMalytshe, by adding a supplementary term. In


this term, dj is the family of the elementary closed loop j. By applying Eq. (33) to thePmobility
calculation of the mechanism presented in Fig. 1, with n = 10, q = 2, C 5i1 iC i
12 5 60, we obtain: M = 0 (if we presume d1 = d2 = 0), M = 2 (if d1 = d2 = 1), M = 4 (if
d1 = d2 = 2), M = 6 (if d1 = d2 = 3), M = 8 (if d1 = d2 = 4) and M = 10 (if d1 = d2 = 5).
It is obvious that the mechanism presented in Fig. 1 has the same dj for all the closed loops, but
the results obtained by using the formula proposed by Manolescu are erroneous even if the inde-
pendent loops have the same motion coecient.
In [27] Manolescu proposed the mobility formula
X
5
M 6  cC c  6  dq 34
cd1

that extends formula


6
P5 P5 by Ozol. By taking into consideration the notation 6  d = b,
(31), proposed
F i F i and c1 6  cC c i1 iF i , where Cc represents the number of joints of class c and Fi
the number of pairs having the degree of connectivity fi = i, we can note that Manolescus formula
(34) is identical to Eq. (13) proposed by Hochman.
1080 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

Manolescu [96] described three kinds of kinematic chains having total, partial and fractionated
mobility. A multi-loop kinematic chain has a total mobility M > 0 if each of its loops have a
mobility Mi P M and a partial mobility if it has at least one loop of mobility Mi < M. The mul-
ti-loop kinematic chain of mobility M is said to have a fractionated mobility if it contains a
so-called separation link which when cut into two split the chain into two separate subchains
of mobilities M1 and M2 such that M = M1 + M2.

2.19. Contribution of Bagci

Bagci [28] proposed the following mobility equation to calculate the degrees of freedom of
motion in a mechanism of m links and q loops:
M Mi Mr M0 D Mc  Mp 35
Pm1
in which Mi is the number of inputs required to drive the mechanism M r j1 frj is the total
number of redundant freedoms in the mechanism; frj is the number of redundant freedoms of
the link j.
X
r
M 0 bm  1  b  iF i ; 36
i1

where r = b  dmin  1 is the maximum number of degrees of freedom that a joint can have in the
motion space of the mechanism and dmin is the smallest number of general constraints in the loops
of the mechanism. By taking into consideration the notation b = 6  d, Eq. (36) represents
ArtobolevskiDobrovolskis
Pq mobility equation (21) when dmin = d.
D j1 d j is the total number of general constraints in the mechanism. This term is similar to
the supplementary term introduced by Manolescu and Manafu in Eq. (33) to extend the formula
proposed by SomovMalytshe.
P
M c Ng j1 qcj  1 is the total number of overclosing constraints in the mechanism, Ngthe
number of groups of loops introducing overclosing constraints, qcjthe number of loops in the
group j introducing overclosing constraints.
P
M p pj1 fpj is the total number of passive mobilities in the joints of the mechanism, fpjthe
number of passive mobilities in joint j.
The mechanism presented in Fig. 1 have Mc = Mp = 0. By applying Eq. (35) to the mobility
calculation of this mechanism, we obtain the same erroneous results that were obtained by using
Eqs. (33).

2.20. Contribution of Antonescu

Antonescu proposed two mobility formulas applicable to the mechanisms having independent
closed loops with dierent values for the motion coecient bi [29,30]. The rst formula [29]
X
5
M da 6  d a m  1  i  d a C i 37
i1
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1081

represents a generalisation of Eq. (18) proposed by Dobrovolski, in which m  1 = n and da is the


apparent family calculated as the average of the constraint parameters di(i = 1, 2, . . . , q) of the
independent loops
1X
q
da d i: 38
q i1

By applying Eq. (37) to the mobility calculation of the mechanism presented in Figs. 1 and 2, we
obtain the same erroneous results that were obtained by using Eq. (33).
The second formula proposed by Antonescu [30]
X
5 X
6
M iF i  jRj 39
i1 j2

represents a particular
P5 case
Pof P6 proposed by Voinea and Atanasiu where pp = 0. We can
Eq. (25)
q
see that N i1 iF i and j1 rj j2 jRj , where Fi represents the number of joints having the
degree of connectivity fi = i and Rj represents the number of closed loops having the motion coef-
cient bj = j. By applying Eq. (39) to the mobility calculation of the mechanism presented in Fig.
1, we obtain obviously the same erroneous results that were obtained by using Eq. (25) from
which Eq. (39) can be derived.

2.21. Contribution of Freudenstein and Alizade

Freudenstein and Alizade proposed a mobility equation that applies to mechanisms without
exception as the authors stated in [31]:
X
E X
q
M ei  bi 40
i1 k1

in which ei is ith independent, scalar, displacement variable of mechanism (associated with the rel-
ative displacements of the joint), and E is the total number of independent, scalar, displacement
variables. The authors named bi as mobility number and is given by the number of independent,
scalar, dierential loop-closure equations for the ith independent loop. In the general case (ab-
sence of metric restrictions) this number is equal to the degree of freedom of the space within
which the mechanism operates. If the value of bi is not identical in each independent loop, the
mechanism is said to have mixed mobility.
Three special cases of Eq. (40) are also presented in [31] when:

(a) the displacement variables are in 1:1 correspondence with the degrees of freedom of the
relative motion in joints
Xp Xq
M fi  bk ; 41
i1 k1

(b) the number of independent scalar loop-closure equations is identical in each independent
loop
1082 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

X
E
M ei  qb; 42
i1

(c) general kinematic chains satisfying simultaneously the two previous conditions.
Xp
M fi  qb: 43
i1

We can see that Eqs. (40)(43)


P representPparticular cases of Eq. (25) proposed by Voinea and
Atanasiu where pp = 0, N Ei1 ei (N pi1 fi ) and rj = bk.P
Eq. (43) is identical to Eq. (13) pro-
posed by Hochman, by taking into consideration that F pi1 fi .

By applying Eqs. (40)(43) to the mobility calculation of the mechanism presented in Fig. 1, we
obtain obviously the same erroneous results that were obtained by using Eq. (25) from which
these equations can be derived.
In a recent work [76], Waldron and Kinzel presented a formula similar to Eq. (41) for the
mobility calculation that accommodates overconstrained closures of arbitrary type. They critically
analysed the utility of this equation by mentioning: Unfortunately, unless the value of bk asso-
ciated with the dierent closures can be identied by inspection, such expressions have no value.
The reason is that the mobility equation gives a quick check of the number of position variables
and independent equations without the need to develop these equations. However, the only way
to verify an overconstraint closure of type not identiable by inspection is to develop the closure
equations and analyze them for dependency. Therefore the quick-check advantage of the mobility
equation disappears, and there is no way to derive information about the linkage without
performing a complete position analysis.
By applying Eqs. (41) and (43) to the mobility calculation of the mechanism presented in Fig. 1,
we obtain obviously the same erroneous results that were obtained by using Eq. (21) from which
these equations can be derived.

2.22. Hunts contribution

The formula proposed by Hunt [32]


Xp
M bm  p  1 fi 44
i1
P
is identical to Eq. (13) proposed by Hochman, by taking into consideration that F pi1 fi and
q = p  m + 1.
Hunt also presented in [32] a formula similar to Eq. (41). A warning is added in [32] that this
equation poses further questions of how to be sure of identifying suitable independent loops and
how to assign the correct values to bk.

2.23. Herves contribution

Herve [33,34] proposed a mobility formula based on the algebraic group structure of the
displacement set
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1083

X
p
M bm  1  b  fi ; 45
i1

where b is given by the dimension of the displacement subgroup associated with the kinematic
chain and fi by the dimension of the displacement subset associated with the kinematic pair be-
tween the elements i and i + 1. This formula is identical to Kutzbachs mobility equation (17)
by taking into consideration that b = 6  d. As with Eq. (17), Eq. (45) also represents a particular
case of Eq. (12).
Herve [33,34] considered the kinematic chains that obey to Eq. (45) as trivial chains and a
warning is added concerning the application of this formula to exceptional chains and to par-
adoxical chains. The exceptional chain is a kinematic chain which involves several intersecting
subgroups with dierent dimensions. The dimension of the displacement subgroup associated with
a paradoxical chain depends on specic geometric constraints dened by link-length and twist
angle relationships. In a more recent paper Herve [69] considered that paradoxical chains are still
an enigma.
By applying Eq. (45) to the mobility calculation of the mechanism presented in Fig. 1, we ob-
tain obviously the same erroneous results that were obtained by using Eqs. (12) and (17) from
which Eq. (45) can be derived. We note that the mechanism presented in Fig. 1 does not t Herves
denitions of exceptional and paradoxical chains.

2.24. Gronowiczs contribution

Gronowicz [37] presented a method for identication of mobility properties for multi-loop
kinematic chains and he proposed the following formula for mobility calculation:
X
q X
q1 X
q
M Mi  F ij ; 46
i1 i1 ji1

where Mi is the mobility of the loop i and Fij is the mobility of the connecting mechanism made of
the joints common between any two loops i and j (the mobility of the joints which have been
counted twice in the q loops of the mechanism).
We can see that
Pq1 PEq. (46) can be derived from Eq. (30) proposed by Manafu, by taking into ac-
q
count that i1 ji1 F ij M c and ignoring the fact that the degree of mobility of a complex
kinematic chain depends on the choice of the independent closed loops, as in [23].
By applying Eq. (46) to the mobility calculation of the mechanism presented in Fig. 1, we
obtain obviously the same erroneous results that were obtained by using Eq. (30) from which
Eq. (46) can be derived.

2.25. Bakers contribution

Baker [35] proposed an algorithm for determining the freedom between links which are sepa-
rated by cross-jointing. The algorithm incorporates quantitative relationships and embraces earlier
results, particularly the series and parallel laws of screw system theory expounded by Waldron [26]
and Davies and Primrose [98], by examining relations among graph theory, the AronholdKenedy
1084 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

[99,100] theorem for planar linkages, the more general PhillipsHunt theorem [95] of three axes in
spatial motion of three bodies, Plucker co-ordinates for screws and vector and projective spaces.
Baker [35] used the motor notations of screw system algebra to represent all six co-ordinates of
a screw, thereby resulting in relative magnitudes of screws becoming available for various relative
freedoms investigated.
Baker [36] placed the notion of mobility in perspective with the algorithm developed in [35] and
he pointed out that the determination of gross mobility without regard to relative freedom be-
tween specic links is not possible by the known techniques (without hazard). By placing the cal-
culation of gross mobility as a secondary stage after the study of relative freedom, Baker shows
that the number of independent parameters of the linkage arose by consideration of the relative
motion alone between individual members, for any pair of which we can write down the appro-
priate screw motor. He did not present any explicit formulas for the calculation of mechanism
mobility by considering that we may nd the chains mobility as a consequence simply by count-
ing the number of independent parameters.

2.26. Daviess contribution

Davies [97] adapted Kirchhos circulation law for potential dierence to the purpose of nding a
set independent instantaneous screw motors associated with any two links in a kinematic chain when
the conguration of the kinematic chain is given. The procedure leads to a constraint matrix formu-
lation and the rank calculation and is applicable to any kinematic chain: it does not require special
cases to be identied. Davies [3840] used the procedure based on the constraint matrix to determine
mobility, passivity and redundancy of mobile and immobile assemblies named mechanical
networks.
Davies proposed the following formulas for the calculation of the degree of mobility [38,39]
M e  r; 47

X
q
M e bk ; 48
k1

where e represents the number of joints with one degree of freedom associated to the mechanical
network, r is the rank of the coecient matrix of constraint equations and bk is the order of the
screw system of the loop (circuit) k.
As Davies mentioned in [39], Eq. (48) is identical to Eq. (22) proposed by Moroskine (by taking
into consideration that e = N) and Eq. (48) is identical to Eq (41) proposed by Freudenstein and
Alizade [29] and by Hunt [32].
A warning is added in [39] that Eq. (48) poses further questions of how to assign the correct
values to bk. The danger lies in the interpretation of the words bk is the order of the screw system
of the circuit k. Davies stated that the interpretation that most would assume, and evidently the
one intended in [32] is that bk is the order of the screw system to which the joints of loop (circuit) k
belong when no other loop is closed. Using this interpretation Eq. (48) underestimates the value of
M for some mechanical networks whichever set of independent loop is chosen. To give Eq. (48)
universal applicability, Davies mentioned that bk must be dened as the order of the screw system
of loop k when account has been taken of the constraints imposed by all previously closed loops.
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1085

The loops can be considered in any sequence but account must be taken of constraints imposed by
loops closed earlier in the sequence. He also noticed in [39] that this interpretation of bk is unfor-
tunately rather more dicult to apply but there may be circumstances in which this stepwise
approach proves easier than nding the rank of the constraint matrix of the mechanical network
in its entirety.

2.27. Contribution of Agrawal and Rao

Agrawal and Rao [41] developed analytical tests for fractionated degree of freedom kinematic
chains based on path loop connectivity matrix. They used loop connectivity properties of multi-
loop kinematic chains to develop a seven-step hierarchical classication scheme of kinematic
structures [42]. A formula is also proposed to calculate the mobility of a multi-loop mechanism
with simple or multiple joints in terms of the mobility of its independent loops and connecting
mechanism [43]
Xq X
q1 X q XN1
1 2 XN2
1 2
M Mi  F ij mi  mi  2F mi n  3ni 2F ni ; 49
i1 i1 ji1 i1
2 i1
2 i
where mi is the number of links forming the ith internal multiple joint, nithe number of links
forming the ith external multiple joint, N1the total number of internal multiple joints, N2
the total number of external multiple joints, Fmi - the mobility of simple joint forming the ith inter-
nal multiple joint, Fnithe mobility of simple joint forming the ith external multiple joint.
A multiple joint formed by mi links is equivalent to (mi  1) simple joints. An internal multiple
joint is completely inside the mechanism, i.e. it is common to independent loops, and not peri-
pheral loop. An external multiple joint is on the peripheral loop of the kinematic chain and hence
it is common to both independent loops and the peripheral loop.
We can see that Eq. (49) extends Eq. (46) proposed by Gronowicz to multiple jointed mecha-
nisms. Agrawal and Rao [43] considered that Eq. (49) is applicable to any general mechanism with
constant or variable general constraints with simple or multiple joints. In spite of this statement,
by applying Eq. (49) to the mobility calculation of the mechanisms presented in Fig. 1 (which has
only simple joints) we obtain the same erroneous results that were obtained by using Eq. (46).

2.28. Contribution of Angeles and Gosselin

Angeles and Gosselin [44] addressed the problem of nding the mobility of an overconstraint
mechanism, which is done by using the Jacobian matrix of a simple or multi-loop closed kinematic
chain. They have shown that, once the Jacobian matrix (J) of a kinematic chain coupled by either
revolute or prismatic pairs is suitably dened, the mobility of the chain can be uniquely computed
as the dimension of the nullspace of J (the nullity of J):
M nullityJ : 50
The authors mentioned that the idea developed in [44] was originally suggested by Freudenstein
[101] in a very general manner. We also can see that Eq. (50) can be derived from Eq. (22) pro-
posed by Moroskine. We take into account that in a linear transformation (with a Jacobian
matrix J) of a nite-dimensional vector space V, the following equation can be written:
1086 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

rankJ nullityJ dimV 51


and the fact that r = rank(J) and N = dim(V).
The Jacobian matrix produces the mobility of the kinematic chain, regardless of its type (single
or multi-loop). Five fully examples, including Bennet mechanism and a planar three degrees of
mobility parallel robotic manipulator are given in [44]. As with Moroskines equation, Eq. (50)
are valid without exception, but it does not give to us a quick calculation of mobility without
the need to develop the kinematic equations and to calculate the nullity of the Jacobian matrix.
These calculations will be presented in Section 3.

2.29. Contribution of Dudita and Diaconescu

Dudita and Diaconescu [45] proposed mobility formulas applicable to elementary or complex
mechanisms. They demonstrated that the passive mobilities in joints do not inuence the mobility
of an elementary (single loop) mechanism; consequently the following two formulas give the same
results:
X
p
M fi  b; 52
i1

X
p
M fie  be ; 53
i1

where fie denotes the eective (active) mobilities in ith joint and be is the dimension of the active
motion space of the mechanism (the motion parameter obtained by taking into consideration only
the active mobilities of the joints). We can see that Eq. (52) also represents a particular case of
Eqs. (13) and (10) proposed by Hochman (for q = 1). Dudita and Diaconescu called b the kine-
matic rank of the mechanism (the dimension of the motion space). To calculate the kinematic
rank, they used the screw co-ordinates and the system of generators from the screw algebra.
To calculate the degree of mobility of a complex (multi-loop) mechanism, Dudita and Diacone-
scu [45] proposed the following formulas:
Xp Xq
M fi  bj ; 54
i1 j1

X
p X
q
M fie  bej ; 55
i1 j1

X
q
M M i  M c; 56
i1

with
X
e
Mc qj  1fcom:j ; 57
j
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1087

e
where Mc is the mobility of the common joints between dierent closed loops, fcom:j is the active
degree of mobility of the jth common joint and qj is the number of loops to which the common
joint j belongs.
We can see that Eq. (54)
P representsP a particular
P case of Eq. (25) proposed by Voinea and Ata-
nasiu when pp = 0 (N pi1 fi and qj1 rj qj1 bj ). By applying Eq. (54) to the mobility calcu-
lation of the mechanisms presented in Fig. 1, we obtain obviously the same erroneous results that
were obtained by using Eq. (25) from which Eq. (54) can be derived. The same erroneous results
are also obtained
P by usingP Eq. (55),Pin the mechanisms
P presented in Fig. 1 all the joint mobilities
are eective ( pi1 fi pi1 fie and qj1 bj qj1 bej ). By applying Eq. (56) to the mobility calcu-
lation of the mechanisms presented in Fig. 1, we obtain the same erroneous results that were
obtained by using Eq. (30) proposed by Manafu.

2.30. Contribution of Fanghella and Galletti

Fanghella and Galletti [46] developed a systematic approach to determining the mobility pro-
perties of single-loop kinematic chains based on displacement groups introduced by Herve [33] in
1978. Such properties are dened by: the mobility of the chain, the number of degrees of freedom
of the relative motion between any two links in the chain (link connectivity), the type of relative
displacements between any two links in the chain (the displacement group containing all their
relative displacements), the set of invariant properties [47] of every displacement subgroup.
The algorithm uses the following equation for determining mobility of single-loop kinematic
chains:
X p
M fi  mincoii ; i 1; . . . ; m; 58
i1

where coii is the loop connectivity dened as the connectivity of the open chain obtained by cut-
ting the link i of the closed loop. Any link i (i = 1, . . . , m) of the closed chain can be cut, and, due
to non-associative characteristics of the kinematic constraints composition, dierent values of the
connectivities coii can be obtained. The minimum value coii gives the correct loop connectivity.
The approach is applicable to single-loop kinematic chains which can be dened by the com-
position of ten displacement groups dened in [46]. Any mechanism not complying with this
scheme cannot be considered. The approach being independent of local singularities can not be
used to determine the local or permanent changes in mechanism mobility due to singular
congurations.
By taking into account that min(coii, i = 1, . . . , m) = b, we can see that Eq. (58) also represents a
particular case of Eq. (13) proposed by Hochman (for q = 1). The main contribution of Fanghella
and Galletti [46] is to give a systematic approach to determine the loop connectivity b based on the
displacement groups.

2.31. Fayets contribution

Fayet [48] proposed a general iterative process to obtain the spaces of twists between any two
bodies of a multi-loop mechanism. A linear application, represented by a matrix called genera-
tive matrix, is set up by a triangular projective method for the generalisation of the sum of
1088 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

dependent spaces in a multi-loop mechanism. At each addition of a kinematic chain, in order to


obtain a new face in the graph of the mechanism, the space of the twists (consistent with the new
mechanism) between the solids at the end of the chain is a subspace of the space consistent with
the previous one. After each reconguration, the previous generative matrix is multiplied by a
connexion matrix to obtain the new generative matrix relative to the solids taking place on the
summits of the graph. By using the triangular projective method in the edges of the graph, the
generative matrix is extended to other solids. The sum of dependent spaces all the spaces of twists
for the whole mechanism are obtained.
The iterative method proposed for obtaining the space of twists between any two bodies pro-
posed in [48] is used by Fayet and Bonnet [49] to obtain the rank of constraint equations of a
multi-loop mechanism. They bring out a general proof to the stepwise approach put forward in
1983 by Davies [39] to determine the order of the screw system of loop k when account has been
taken of the constraints imposed by all previously closed loops. The proof is based on a lemma
about the rank of a partitioned system of linear equations. In their iterative approach, the rank
of the constraint equations of a mechanism A 0 showing q independent loops is equal to the rank
of the constraint equations a mechanism A included in A 0 with q  1 independent loops plus the
rank of a following virtual loop. The graph of this virtual loop is composed of the edge {i, . . . , j}
which is not included in A and a virtual chain generating the space of twists consistent with A.
Three examples are given in [49], including the mechanism analysed by Davies [39], to emphasize
the applicability of the iterative approach proposed. This method has universal applicability but it
gives the instantaneous mobility (mobility in a given position of the mechanism). The rank of the
constraint equations is obtained by the rank of the twists consistent with the iterative approach
dened for the given position.

2.32. Tsais formula

To take into consideration the fact that passive (internal) degrees of freedom cannot be used to
transmit motion or torque about an axis, Tsai [51] proposed to subtract the number of passive
degrees of freedom, fp, from the degree of freedom equation:
X
p
M bm  p  1 fi  fp : 59
i1

By applying Eq. (59) to the mobility calculation of the mechanisms presented in Fig. 1 having
fp = 0, we obtain the same erroneous results that were obtained by using Eq. (13) from which
Eq. (59) can be derived.

2.33. McCarthys formula

To calculate the mobility of the platform linkages, in which the platform is connected to the
ground by serial chains, McCarthy [52] proposed the mobility formula
X
pc
M b b  M ci ; 60
i1
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1089

where pc represents the number of serial open chains connecting the platform to the ground and
M ci is the mobility of the ith chain including its part of the platform. By applying the formula (60)
to the mobility calculation of the mechanism presented in Fig. 1, having M ci 4, we obtain the
erroneous results M = 0 (if we consider the motion parameter b = 6), M = 2 (if we consider
b = 5), M = 4 (if we consider b = 4).

2.34. Contribution of Huang, Kong and Fang

Huang, Kong and Fang proposed the formula


X
p
M bm  p  1 fi  t 61
i1

considered as a correct application of the GrublerKutzbach criterion in the case of parallel


mechanisms [53]. In Eq. (61) b = 6  d is called order of the mechanism and d is the number of
independent common constraints in the mechanism. A common constraint is dened in [53] as
a screw reciprocal to the unit twists associated with all kinematic pairs in the mechanism. The cor-
rective term t represents the number of redundant constrains that are linearly dependent with
other constraints. Eq. (61) gives the instantaneous mobility (mobility in a given position of the
mechanism). The order of the mechanism and the number of redundant constraints are calculated
for any given position.

2.35. Contribution of Rico, Gallardo and Ravani

Rico and Ravani [54] extends the works of Herve [33,34,69], Fanghella and Galletti [46,47] and
Angeles [102] in application of group theory to analysis of kinematic chains. They proposed the
following formula to calculate the mobility of a single-loop closed mechanism:
X
p
M fi  dimH c i; j  dimH cc i; j dimH a i; j; 62
i1

where Hc(i, j) and Hcc(i, j) are the clockwise and the counterclockwise composite subgroups asso-
ciated with the two open chains connecting two links i and j in the single-loop mechanism and
Ha(i, j) is the absolute composite subgroup between i and j (the connectivity between links i
and j). Dim stands for the dimension of the respective subgroup of the Euclidean group. We
can see that Eq. (62)
Pp is equivalent to Eq. (24) proposed by Voinea and Atanasiu by taking into
account that N i1 fi , r1 = dim(Hc(i, j)), r2 = dim(Hcc(i, j)), and we = dim (Ha(i, j)).
Rico, Gallardo and Ravani [55] demonstrated that mobility formula can be obtained either by
using arguments of nite kinematics or by using arguments of innitesimal kinematics. In [55] the
authors translated mobility criterion (62) from the nite kinematics language of group theory into
innitesimal kinematics language of Lie algebra. The instantaneous form of the mobility criterion
presented here is based on the theory of subspaces and subalgebras of the Lie algebra of the
Euclidean group and their possible intersections.
1090 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

3. Why the formulas for quick calculation of mobility do not work for any mechanism

Let us consider a mechanism with p joints and at least one closed loop. Each joint i with a de-
gree of mobility (connectivity) fi, introduces fi independent motion parameters. The closed loops
cancel the independence of a part of the joints independent parameters. By taking into consider-
ation the mobility denition given by IFToMM [1], we can say that the number of independent
co-ordinates (parameters) needed to dene the conguration of a mechanism with closed
Pp loops is
the dierence between the number of independent motion parameters of the joints ( i1 fi ) before
loop closures provide further constraints and the number of joint parameters that lost their inde-
pendence after loop closures (r):
Xp
M fi  r: 63
i1

Eq. (63) is valid without exception, as like Eq. (22) from which it also can be derived. Moros-
kine[17] indicated that the number of joint parameters that lost their independence in the closed
loops of the mechanism (r) is given by the rank of the homogeneous linear set of velocity equa-
tions of the mechanism. To calculate this rank we need to set up the velocity constraint equations
of the mechanism. These calculations are not easy applied in the case of complex mechanisms. The
numerical calculation of the rank in the proximity of the singular positions introduces certain
ambiguities in the interpretation of the results. It is therefore very dicult to make the dierence
between a real zero and a very small numerical value close to zero in the calculation of the deter-
minants. As we have mentioned in Section 1, this drawback is specic to all methods for mobility
calculation based on rank calculation of constraint equations in a given position of the mecha-
nism with joint location dened numerically. Symbolic calculation of the rank could overcome
these limits if the constraint equations are dened symbolically in a generic position, as it will
be presented in this section.
The closure equations of the parallel Cartesian robotic manipulator presented in Fig. 1 can be
established by the condition that the velocity of point H (expressed in the reference system
O0x0y0z0see Fig. 1) must be the same in the three legs (H HA HB HC):
0  0  0 
vHA vHB vHC
0
0 0 ; 64
xHA xHB xHC
or
0  0 
vHA vHB
0
 0
0 65
xHA xHB
and
0  0 
vHA vHC
0
 0
0: 66
xHA xHC
Eqs. (65) and (66) represents the closure equations of the closed loops A  B (0A  1A 
2A   4A 4B   2B  1B  0B 0A) and AC (0A  1A  2A   4A =
4C   2C  1C  0C 0A). These loops can be considered two structurally independent
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1091

closed loops of the Cartesian robotic manipulator, presented in Fig. 1. As we have seen, Eulers
formula known in the theory of graphs and proposed by Hochman [8] to calculate the number of
independent closed loops (q) of a complex mechanism (q = p  m + 1 = p  n) gives
q = 12  10 = 2. By calculating the velocity of point H in the three legs of the parallel robotic
manipulator, Eqs. (65) and (66) lead to the following sets of linear equations:
 T
D6 8 d_ 10A u_ 21A u_ 32A u_ 43A d_ 10B u_ 21B u_ 32B u_ 43B 08 1 ; 67
 T
E6 8 d_ 10A u_ 21A u_ 32A u_ 43A d_ 10C u_ 21C u_ 32C u_ 43C 08 1 ; 68
 T
F 12 12 d_ 10A u_ 21A u_ 32A u_ 43A d_ 10B u_ 21B u_ 32B u_ 43B d_ 10C u_ 21C u_ 32C u_ 43C 012 1 :
69

By symbolic calculation of the rank of the matrices [D]68, [E]68, [F]1212 with MAPLE
we obtain: r1 = rAB = rank(D68) = 5, r2 = rA C = rank(E68) = 5 and r = rABC = rank-
(F1212) = 9. We can see that r 5 r1 + r2. This result indicates that the structurally independent
closed loops A-B (0A  1A  2A   4A 4B   2B  1B  0B 0A) and AC (0A 
1A  2A   4A 4C   2C  1C  0C 0A) are not independent from a kinematic point
of view. The motion parameters of these two closed loops are b1 = r1 = 5 and b2 = r2 = 5.
In the general case, the rank (r) of the homogeneous linear set of kinematic constraint equations
of a mechanism with q structurally independent closed loops is less than, or equal to, the sum of
the ranks (ri, i = 1, 2, . . . , q) of the linear sets of kinematic constraint equations of the q loops
X
q
r6 ri : 70
i1

We can establish the following property of structural limitation of the independence of the
closed loops: Eulers formula known in the theory of graphs and proposed by Hochman [8] to cal-
culate the number of independent closed loops (q) of a multi-loop mechanism (q = p  m + 1 =
p  n) must be restricted, in the general case, to the structural independence (in the sense of
the theory of graphs). Structural independence of q closed loops does not always involve the kine-
matic independence of the q closed loops.
By using Eq. (63) we get the right value of mobility of the parallel Cartesian robotic manipu-
lator presented in Fig. 1 (M = 12  9 = 3). The same result is obtained by using Eq. (50) proposed
by Angeles and Gosselin [44] (M = nullity([F]1212) = 3).
We recall that the rank of the matrices [D]68, [E]6 8 and [F]1212 can be calculated numerically
or symbolically. The numerical calculation gives to us the instantaneous rank in a given position
of the mechanism dened by numerical values of joint variables and geometric parameters. The
symbolic calculation gives to us the global rank in a non-dened position of the mechanism with-
out indicating numerical values of joint variables and geometric parameters. We just use the sym-
bolic denition of these matrices. The general rank represents the maximum value of the
instantaneous ranks. The mobility calculation based on the instantaneous rank gives to us the
instantaneous mobility of the mechanism. By using symbolic calculation of the rank we calculate
global or full cycle mobility of mechanisms.
1092 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

All formulas for quick calculation of mobility of multi-loop mechanisms based on the motion
parameter (bi) or on the constraint parameter (di6  bi) associated to the loop i independently of
any other closed loop do not have universal applicability. Their applicability is limited to the
mechanisms that have the rank (r) of the homogeneous linear set of kinematic constraint equa-
tions equal to the sum of the motion parameters (bi) associated with the q structurally independent
closed loops (i = 1, 2, . . . , q)
X
q
r bi : 71
i1

The mechanisms that do not obey Eq. (71) do not t various formulas for quick calculation of
mobility presented in Section 2. These mechanisms have nothing enigmatic or paradoxical, as
the literature considered. They just do not obey Eq. (71). This is the case of the parallel Cartesian
robotic manipulator presented in Fig. 1. Paradoxical is the fact that the formulas for quick cal-
culation of mobility are used inappropriately for these mechanisms. This can be explained by
the lack of more appropriate formulas for quick calculation of mobility of multi-loop mechanisms
with a wider applicability.
The restricted applicability of all formulas for quick calculation of mobility of multi-loop mech-
anisms is due to the fact that the motion parameter (bi) or the constraint parameter (also called
family di = 6  bi) is associated to the loop i independently of any other closed loop and to the fact
that the motion parameter (bi) is considered to be order of the screw system to which the joints of
the independent loop i belong when no other loop is closed.
Critical analysis of some formulas for the calculation of mobility of multi-loop mechanisms was
previously discussed by Waldron [26], Davies[39] and Fayet and Bonnet [49] as mentioned in
Section 2, but they did not formalised the applicability limitation of various formulas for quick
calculation of mobility.

4. A new formula for mobility calculation of parallel mechanisms with elementary legs

The lack of formulas for quick calculation of mobility for multi-loop mechanism with
wider applicability represents a stumbling block in structural (type) synthesis of spatial mecha-
nisms. To overcome this, the author of this article have presented in [84] new formulas for
quick calculation of mobility of multi-loop mechanisms that do not obey Eq. (71). We have
demonstrated formulas to calculate r for various types of parallel mechanisms without calcu-
lating the rank of the homogeneous linear set of constraint equations associated with loops clo-
sure or without calculating the rank of the complete screw system associated to the joints of the
mechanism. We have found our demonstrations on algebra of matrix transformations. The
parameters used in these new formulas can be easily obtained by inspection. An analyti-
cal method to compute these parameters has also been presented just for verication and for a
better understanding of the meaning of these parameters. These formulas were applied in struc-
tural synthesis of new parallel robotic manipulators [84]. The formula for quick calculation of
mobility for parallel mechanisms with t P 2 elementary legs proposed by the author of this paper
in [84] is
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1093

X
p X
t
M fi  Sj Sp; 72
i1 j1

where fi represents the degrees of mobility of the ith kinematic pair (i = 1, . . . , p), Sj is the spati-
ality of the leg j and Sp is the spatiality of the mobile platform in the parallel mechanism. Leg spa-
tiality Sj is dened by the connectivity between the extreme links (the mobile platform and the
xed base) in the leg j considered isolated of the rest of the mechanism. Platform spatiality Sp
is dened by the connectivity between the mobile platform and the xed base in the parallel mech-
anism. Leg spatiality is given by the maximum number of independent motion parameters be-
tween the mobile platform and the xed base in the serial open kinematic chain associated
with the elementary leg j considered isolated of the rest of the mechanism. The platform spatiality
is given by the maximum number of independent motion parameters between the mobile platform
and the xed base in the parallel mechanism. All parameters from Eq. (72) can be easily deter-
mined by inspection. We need just to observe the independent motion parameters dening Sj
and Sp.
For the parallel Cartesian robotic manipulator presented in Fig. 1 we can easily obtain by
inspection p = 12, fi = 1 (i = 1, . . . , 12), Sj = 4 (j = 1, . . . , 3) and Sp = 3 that gives M =
12  12 + 3 = 3. In future paper we will demonstrate that Eq. (72) represents the particular case
(for parallel mechanism with elementary legs) of more general formula applicable to parallel
mechanisms with complex legs. We recall that a serial open kinematic chain is associated with
each elementary leg. One or more closed loops can be integrated in the structure of a complex leg.

5. Conclusions

In the last 150 years, sustained eorts have been made to nd general formula for a quick cal-
culation of mobility of any rigid body mechanism but the magic formula was not yet been
found. Several dozen formulas/approaches have been proposed for the calculation of the mecha-
nism mobility, but many of them are reducible to the same originated formula. Thirty ve contri-
butions have been critically reviewed in this paper by setting up their genesis, similarities and
limitations. Today, we note that formulas for a quick calculation of mobility do not t for many
classical mechanisms or recent parallel robots. We have shown that no formula for quick calcu-
lation of mobility presented in the literature is applicable to the mobility calculation of the recent
parallel robot CPM. We have explained why the previous formulas do not work for some multi-
loop mechanisms and we have delimited the applicability eld of these formulas. We have pre-
sented a new formula for quick calculation of mobility applicable to any parallel mechanisms with
t P 2 elementary legs. In future paper we will demonstrate that this formula represents the
particular case of more general formula applicable to parallel mechanisms with complex legs.

Acknowledgement

This work was supported by the French National Council of Scientic Research (CNRS) in the
framework of the project ROBEA-MAX, 20022003.
1094 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

References

[1] T.G. Ionescu, Terminology for mechanisms and machine science, Mech. Mach. Theory 38 (2003).
[2] P.A. Chebychev, Theorie des mecanismes connus sous le nom de parallelogrammes, 1ere partie, Memoires
presentes a lAcademie imperiale des sciences de Saint-Petersbourg par divers savants, 1854.
[3] P.A. Chebychev, Theorie des mecanismes connus sous le nom de parallelogrammes, 2eme partie, Memoires
presentes a lAcademie imperiale des sciences de Saint-Petersbourg par divers savants, 1869.
[4] J.J. Sylvester, On recent discoveries in mechanical conversion of motion, Proc. Roy. Inst. Great Britain 7/5 (1874)
179198.
[5] M. Grubler, Allgemeine Eigenschaften der Zwanglaugen ebenen kinematischen Ketten, Part I, Zivilingenieur 29
(1883) 167200.
[6] M. Grubler, Allgemeine Eigenschaften der Zwanglaugen ebenen kinematischen Ketten, Part II, Verh. Ver. Bef.
Gew. 64 (1885) 179223.
[7] P.I. Somov, On the degree of freedom of motion of kinematic chains, J. Phys. Chem. Soc. Russia 19 (9) (1887)
443477 (in Russian).
[8] K.I. Hochman, Kinematics of machinery (in Russian), Odesa, 1890.
[9] G. Koenigs, Introduction a Une Theorie Nouvelle Des Mecanismes, Librairie Scientique A. Hermann, Paris,
1905, pp. 2728.
[10] M. Grubler, Das Kriterium der Zwanglaugkeit der Schraubenkelten, Festschrift, O. Muhr. Zum 80, Gubertstag,
Berlin, 1916.
[11] M. Grubler, Getriebelehre: Eine Theorie Des Zwanglaufes Und Der Ebenen Mechanismen, Springer, Berlin,
1917.
[12] A.P. Malytshe, Analysis and synthesis of mechanisms with a viewpoint of their structure (in Russia), Izvestiya
Tomskogo of Technological Institute, 1923.
[13] K. Kutzbach, Mechanische Leitungsverzweigung, ihre Gesetze und Anwendungen, Maschinenbau. Betrieb. 8
(1929) 710716.
[14] V.V. Dobrovolski, Dynamic analysis of statically constraint mechanisms, Akad. Nauk. SSSR, Trudy Sem. Teorii
Masin i Mekhanizmov 30 (8) (1949) (in Russian).
[15] V.V. Dobrovolski, Theory of mechanisms (in Russian), Moscow, 1951.
[16] I.I. Artobolevskii, Theory of mechanisms and machines (in Russian), Moscow, 1953.
[17] Y.F. Moroskine, General analysis of the theory of mechanisms, Akad. Nauk. SSSR, Trudy Sem. Teorii Masin i
Mekhanizmov 14 (1954) 2550 (in Russian).
[18] Y.F. Moroskine, On the geometry of complex kinematic chains, Sov. Phys.Dokl. 3/2 (1958) 269272.
[19] R. Voinea, M. Atanasiu, Contribution a letude de la structure des chanes cinematiques, Bull. Inst. Politechnic
Bucuresti XXII (1960) 2977.
[20] I. Kolchin, Experiment in the construction of an expanded structural classication of mechanisms and a structural
table based on it. Trans. 2nd All-Union Conf. Basic Problems of the Theory of Machines and Mechanisms (in
Russian), Moscow, 1960.
[21] W. Rossner, Zur strukturellen Ordnung der Getriebe, Wissenschaft. Tech. Univ., Dresden, vol. 10, 1961, pp.
11011115.
[22] H. Boden, Zum Zwanglauf gemischt raumlich-ebener Getriebe, Maschinenbautechnik (Getriebetechnik) 11 (1962)
612615.
[23] N. Manolescu, V. Manafu, Sur la determination du degre de mobilite des mecanismes, Bull. Inst. Politechnic
Bucuresti 25 (1963) 4566.
[24] O.T. Ozol, On a new structural formula of mechanisms, Izv.Vuzov. Maschinostroenie 2 (1963) (in Russian).
[25] K.H. Hunt, J.R. Phillips, Zur Kinematic mechanischer Verbindung fur raumliche Bewegung, Maschinenbau.
Getriebe. 14 (1965) 657664.
[26] K.J. Waldron, The constraint analysis of mechanisms, J. Mech. 1 (1966) 101114.
[27] N. Manolescu, For a united point of view in the study of the structural analysis of kinematic chains and
mechanisms, J. Mech. 3 (1968) 149169.
[28] C. Bagci, Degrees of freedom of motion in mechanisms, ASME J. Eng. Industry 93B (1971) 140148.
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1095

[29] P. Antonescu, Extending of structural formula of Dobrovolski to the complex mechanisms with apparent family,
in: Proceedings of the SYROM, Bucharest, 1973.
[30] P. Antonescu, General formula for the d.o.f. of complex structure manipulators and robots, Tenth World
Congress on the Theory of Machines and Mechanisms, Oulu, 1999.
[31] F. Freudenstein, R. Alizade, On the degree-of-freedom of mechanisms with variable general constraint, Fourth
World Congress on the Theory of Machines and Mechanisms, Newcastle upon Tyne, 1975.
[32] K.H. Hunt, Kinematic Geometry of Mechanisms, Oxford University Press, Oxford, 1978.
[33] J.M. Herve, Analyse structurelle des mecanismes par groupe des deplacements, Mech. Mach. Theory 13 (1978)
437450.
[34] J.M. Herve, Principes fondamentaux dune theorie des mecanismes, Rev. Roum. Sci. Tech.Mec. Appl. 23/5
(1978) 693709.
[35] J.E. Baker, On relative freedom between links in kinematic chains with cross-jointing, Mech. Mach. Theory 15
(1980) 397413.
[36] J.E. Baker, On mobility and relative freedoms in multiloop linkages and structures, Mech. Mach. Theory 16
(1981) 583597.
[37] A. Gronowicz, Identizierungs-Methode der Zwanglaufbedingungen von kinematischen ketten, Mech. Mach.
Theory 16 (1981) 127135.
[38] T.H. Davies, Mechanical networksI: Passivity and redundancy, Mech. Mach. Theory 18 (1983) 95101.
[39] T.H. Davies, Mechanical networksII: Formulae for degrees of mobility and redundancy, Mech. Mach. Theory
18 (1983) 103106.
[40] T.H. Davies, Mechanical networksIII: Wrenches on circuit screws, Mech. Mach. Theory 18 (1983) 107112.
[41] V.P. Agrawal, J.S. Rao, Fractionated freedom kinematic chains and mechanisms, Mech. Mach. Theory 22 (1987)
125130.
[42] V.P. Agrawal, J.S. Rao, Structural classication of kinematic chains and mechanisms, Mech. Mach. Theory 22
(1987) 489496.
[43] V.P. Agrawal, J.S. Rao, The mobility properties of kinematic chains, Mech. Mach. Theory 22 (1987) 497504.
[44] J. Angeles, C. Gosselin, Determination du degre de liberte des chanes cinematiques, Trans. CSME 12/4 (1988)
219226.
[45] F. Dudita, D. Diaconescu, Optimizarea structurala a mecanismelor, Tehnica, Bucuresti, 1987, pp. 3645, 229254.
[46] P. Fanghella, C. Galletti, Mobility analysis of single-loop kinematic chains: an algorithmic approach based on
displacement groups, Mech. Mach. Theory 29 (1994) 11871204.
[47] P. Fanghella, Kinematics of spatial linkages by group algebra: a structure-based approach, Mech. Mach. Theory
23 (1988) 171183.
[48] M. Fayet, Mecanismes multi-bouclesI: Determination des espaces de torseurs cinematiques dans un mecanisme
multi-boucles quelqonque, Mech. Mach. Theory 30 (1995) 201217.
[49] M. Fayet, P. Bonnet, Mecanismes multi-bouclesII: Processus de determination du rang des equations de
liaisondistribution des mobilites, Mech. Mach. Theory 30 (1995) 219232.
[50] M. Fayet, Mecanismes multi-bouclesIII: Hyperstatisme au sens de la dynamique et au sens de la cinematique
dualite, Mech. Mach. Theory 30 (1995) 233252.
[51] L-W. Tsai, Robot Analysis: the Mechanics of Serial and Parallel Manipulators, John Wiley, 1999.
[52] J.M. McCarthy, Geometric Design of Linkages, Springer-Verlag, New York, 2000, pp. 38.
[53] Z. Huang, L.F. Kong, Y.F. Fang, Mechanism Theory and Control of Parallel Manipulators, China Machine
Press, 2003, quoted by Z. Huang, Q.C. Li, Type synthesis of symmetrical lower-mobility parallel mechanisms
using the constraint-synthesis method, Int. J. Robotics Res. 22 (2003) 5979.
[54] J.M. Rico Martinez, B. Ravani, On mobility analysis of linkages using group theory, Trans. ASME, J. Mech.
Des. 135 (2003) 7080.
[55] J.M. Rico, J. Gallardo, B. Ravani, Lie algebra and the mobility of kinematic chains, J. Robotics Syst. 20 (2003)
477499.
[56] P. De Roberval, Nouvelle maniere de balance, Memoires de mathematiques et de physique, Tirez des registres de
lAcademie Royale des Sciences depuis 1666 jusqua 1699, Tome X, Institut de France, Academie des sciences,
1670, pp. 494496.
1096 G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097

[57] P.F. Sarrus, Note sur la transformation des mouvements rectilignes alternatifs en mouvements circulaires et
reciproquement, Academie des Sciences, CR Hebd. Sci., Paris 36 (1853) 10361038.
[58] E. Delassus, Sur les systemes articules gauches, Premiere partie,, Ann. Sci. Ecole Norm. Superieure, Paris, 3
Series 17 (1900) 455499.
[59] E. Delassus, Sur les systemes articules gauches, Deuxieme partie, Sur les systemes articules gauches, Deuxieme
partie 19 (1902) 119152.
[60] E. Delassus, Les chanes articules fermees et deformables a quatre membres, Bull. Sci. Math., Paris 46 (1922) 283
304.
[61] G.T. Bennett, A new mechanism, Engineering 76 (1903) 777778.
[62] R. BricardLecons De Cinematique, vol. 2, Gauthier-Villars, Paris, 1927, pp. 712.
[63] F.E. Myard, Contribution a la geometrie des systemes articules, Bull. Soc. Math. France 59 (1931) 183210.
[64] M. Goldberg, New 5-bar and 6-bar linkages in three dimensions, ASME J. Mech. 65 (1943) 649661.
[65] F.G. Altman, Sonderformen Raumlieher Koppelgetriebe und Grenzen Ihrer Verwendbarkeit, Konstruktion,
Werkstoe Versuchswesen 4 (1952) 97106.
[66] J.E. Baker, Overconstrained 5-bars with parallel adjacent joint axes, Mech. Mach. Theory 13 (1978) 213218.
[67] K.J. Waldron, Overconstrained linkages, Environ. Plan. B 6 (1979) 393402.
[68] J.E. Baker, T. Duclong, P.S.H. Khoo, On atempting to reduce undesirable characteristics of the Schatz
mechanism, ASME, J. Mech. Des. 104 (1982) 192205.
[69] J.M. Herve, The Lie group of rigid body displacements, a fundamental tool for mechanism design, Mech. Mach.
Theory 34 (1999) 719730.
[70] R.L. Norton, Design of Machinery: an Introduction to the Synthesis and Analysis of Mechanisms and Machines,
McGraw-Hill, Boston, 1999, pp. 3738.
[71] D.H. Myszka, Machines and Mechanisms: Applied Kinematic Analysis, Prentice Hall, 2001, p. 21.
[72] H.H. Mabie, C.F. Reinholtz, Mechanisms and Dynamics of Machinery, John Wiley & Sons, New York, 1987, pp.
1315.
[73] H.D. Eckhardt, Kinematic Design of Machines and Mechanisms, McGraw-Hill, New York, 1998, pp. 2037.
[74] J. Phillips, Freedom in machineryIntroducing Screw Theory, vol. I, Cambridge University Press, Cambridge,
1984, p. 5.
[75] J. Phillips, Freedom in machineryScrew Theory Exemplied, vol. II, Cambridge University Press, Cambridge,
1990, pp. 147166.
[76] K.J. Waldron, G.L. Kinzel, Kinematics, Dynamics, and Design of Machinery, John Wiley & Sons Inc., 1999,
pp.1726.
[77] R. Clavel, Delta, a fast robot with parallel geometry, in: 18th International Symposium on Industrial Robots,
Lausanne, 1988.
[78] J.M. Herve, F. Sparacino, STAR, a new concept in robotics, in: 3rd International Workshop on Advances in
Robot Kinematics, Ferrara, 1992.
[79] F. Pierrot, O. Company, H4: a new family of 4 dof parallel robot, in: IEEE/ASME International Conference on
Advanced Intelligent Mechatrionics, Atlanta, 1999.
[80] P. Wenger, D. Chablat, Kinematic analysis of a new parallel machine tool: the orthoglide, in: J. Lenarcic, M.L.
Stanisic (Eds.), Advances in Robot Kinematics, Kluwer Academic Publishers, 2000, pp. 305314.
[81] H.S. Kim, L.-W. Tsai, Evaluation of a Cartesian parallel manipulator, in: J. Lenarcic, F. Thomas (Eds.),
Advances in Robot Kinematics, Kluwer Academic Publishers, 2002, pp. 2128.
[82] X. Kong, C.M. Gosselin, Type synthesis of linear translational parallel manipulators, in: J. Lenarcic, F. Thomas
(Eds.), Advances in Robot Kinematics, Kluwer Academic Publishers, 2002, pp. 453462.
[83] M. Carricato, V. Parenti-Castelli, Singularity-free fully isotropic translational parallel mechanism, Int. J.
Robotics Res. 21 (2) (2002) 161174.
[84] G. Gogu, Structural synthesis of parallel robotic manipulators with decoupled motions. Report ROBEA MAX
CNRS, 2003.
[85] F. Wittenbauer, Grapische Dynamik, Springer, Berlin, 1923.
[86] K. Federhofer, Graphische Kinematik Und Kinestostatik, Springer, Berlin, 1932.
[87] A.N. Bogolyubov, History of the Mechanics of Machines (in Russian), Kiev, 1964.
G. Gogu / Mechanism and Machine Theory 40 (2005) 10681097 1097

[88] T.S. Mruthyunjaya, Kinematic structure of mechanisms revisited, Mech. Mach. Theory 38 (2003) 279320.
[89] L.-W. Tsai, Mechanism Design: Enumeration of Kinematic Structures According to Function, CRC Press, Boca
Raton, Florida, 2000, pp. 6672.
[90] Dwelshauvres-Dery, Programme du cours de mecanique applique et de physique industrielle, professe a lEcole
des Mines de Liege, Mons, 1876.
[91] C. Laboulaye, Traite de cinematique: theorie et pratique ou Theorie des mecanismes, Troisieme edition, Librairie
du Dictionnaire des Arts et manufactures, Paris, 1878, pp. 955961.
[92] R.S. Ball, A Treatise on the Theory of Screws, Cambridge University Press, 1990.
[93] R. Voinea, M. Atanasiu, Contributions a la theorie geometrique des vis, Bull. Polytech. Inst. Bucharest XXI (3)
(1959) 6990.
[94] A.G. Erdman, Modern Kinematics: Developments in the Last 40 Years, John Wiley & Sons, Inc., New York,
1993, pp. 5455.
[95] J.R. Phillips, K.H. Hunt, On the theorem of the three axes in the spatial motion of three bodies, Aust. J. Appl.
Sci. 15 (1964) 267287.
[96] N.I. Manolescu, Une methode unitaire pour la formation des chanes cinematiques et des mecanismes plans
articules avec dierents degres de liberte et mobilite, Rev. Roum. Sci. Tech.Mec. Appl. 9/6 (1964) 12631273.
[97] T.H. Davies, Kirchos circulation law applied to multi-loop kinematic chains, Mech. Mach. Theory 16 (1981)
171183.
[98] T.H. Davies, E.J.F. Primrose, An algebra for screw systems of pairs of bodies in a kinematic chain. in:
Proceedings of the 3rd World Congress for the Theory of Machines and Mechanisms, Kupari, Yugoslavia, vol. D,
Paper D-14, 1971, pp. 199212.
[99] S.H. Aronhold, Outline of kinematic geometry, Verh. Ver. Gew. Fleiss. 51 (1872) 129155.
[100] A.B.W. Kenedy, Mechanics of Machinery, Macmillan, London, 1886.
[101] F. Freudenstein, On the variety of motions generated by mechanisms, Trans. ASME J. Eng. Industry 84 (1962)
156160.
[102] J. Angeles, The qualitative synthesis of parallel manipulators, in: C.M. Gosselin, I. Ebert-Upho (Eds.),
Proceedings of the workshop on Fundamental Issues and Future Research Directions for Parallel Mechanisms
and Manipulators, Quebec City, Quebec, Canada, October 34, 2002.

You might also like