32 Int. J. Mechanisms and Robotic Systems, Vol. 3, No.

1, 2016

A new approach for obtaining the dynamic

balancing conditions in serial mechanisms

André Garnier Coutinho* and

Tarcisio Antonio Hess Coelho
Department of Mechatronics and Mechanical Systems Engineering,
Escola Politecnica,
University of Sao Paulo, Brazil
Email: andre.garnier.coutinho@usp.br
Email: tarchess@usp.br
*Corresponding author

Abstract: Adaptive balancing means that the mechanical structure of the

manipulator is modified in order to achieve the decoupling of dynamic equations.
This work deals with a systematic formulation for the adaptive balancing.
Basically, two traditional balancing techniques are employed here: the addition
of counterweight and counter-rotating disks coupled to the moving links. In
addition, the feasibility of the dynamic decoupling for three distinct types of
serial manipulators is discussed regarding the achievement of such balancing
and the complexity level of the modified mechanical structure. The balancing
conditions are developed here for 3-dof spatial and planar open-loop kinematic
chain mechanisms, whose topologies are composed of revolute and prismatic
joints.

Keywords: adaptive balancing; balancing conditions; design; dimensional

synthesis; dynamic balancing; dynamic decoupling; Robotics; serial
mechanisms; spatial mechanisms.

Reference to this paper should be made as follows: Coutinho, A.G. and

Coelho, T.A.H. (2016) ‘A new approach for obtaining the dynamic balancing
conditions in serial mechanisms’, Int. J. Mechanisms and Robotic Systems,
Vol. 3, No. 1, pp.32–47.

Biographical notes: André Garnier Coutinho is currently pursuing his PhD

in Mechanical Engineering from Polytechnic School of Engineering of the
University of São Paulo (EPUSP), Brazil. He received his BEng in Mechanical
Engineering from the Polytechnic School of Engineering at the University of
São Paulo (EPUSP), Brazil in 2013. His main areas of research interest are
multibody dynamics, parallel robots and nonlinear control.

Tarcisio Antonio Hess Coelho is an Associate Professor at the University of

São Paulo since 2005. He received his BEng, MEng and PhD in Mechanical
Engineering from the Polytechnic School of Engineering at the University of São
Paulo (EPUSP), Brazil in 1987, 1990 and 1997, respectively. From 2001–2002,
he was a Post-Doctoral Fellow at Stanford University, USA. He is also a member
of two IFToMM Technical Committees (Mechatronics and Robotics, Multibody
Dynamics). His main areas of research interest are parallel robots, education and
biomechanics.

c 2016 Inderscience Enterprises Ltd.

Copyright
 A new approach for obtaining the dynamic 33

1 Introduction and literature review

Balancing can be considered as an important issue related to the design of any kind of
mechanical system in general, and also serial manipulators, in particular. As a matter of
fact, the performance of open-loop kinematic chain mechanisms associated to specific
applications depends on the choice of the balancing method, namely, either static Wang and
Gosselin (2000) or dynamic Wu and Gosselin (2005), either passive Van der Wijk (2013);
Wu and Gosselin (2005); Gosselin et al. (2004); Wang and Gosselin (2000, 1999); Alici and
Shirinzadeh (2003, 2006); Dehkordi et al. (2012); Russo, Sinatra and Xi (2005); Agrawal
and Fattah (2004) or active Arakelian and Smith (2008); Seo et al. (2013); Wang et al.
(2013); Briot, Arakelian and Le Baron (2012); Coelho, Yong and Alves (2004); Moradi,
Nikoobin and Azadi (2010).
Moreover, Coelho et al. Coelho, Yong and Alves (2004), Moradi et al. Moradi,
Nikoobin and Azadi (2010) and Arakelian and Sargsyan Arakelian and Sargsyan (2012)
use the adaptive balancing to achieve the decoupling of dynamic equations for open-
loop kinematic chain mechanisms. Consequently, this action simplifies the control of
manipulators because the actuators can be controlled independently. The necessary
modifications comprise the addition of either counterweights, or counter-rotating disks
or even both to the original kinematic chain of the manipulator. Consequently, the terms
associated to gravitational, centripetal and Coriolis efforts are completely eliminated from
the dynamic equations. As a matter of fact, the effective inertias for all the actuator axes
are constant and the mathematical expressions of the driving torques/forces become rather
simple. One of the main advantages of this approach concerns the reduction of computing
time for a closed-loop control of manipulators. Such reduction is really significant and it
constitutes in a great benefit for real-time applications.
The contributions of this work are the following: to present a systematic formulation
to obtain the balancing conditions for the adaptive balancing, to discuss the feasibility of
the dynamic decoupling for three distinct types of serial manipulators, not only in terms of
the possibility to achieve such balancing but also regarding the increase of the complexity
level of the modified mechanical structure. The analysed manipulators correspond to 3-dof
spatial and planar open-loop kinematic chain, whose topologies are composed of revolute
and prismatic joints.
This work is organised as follows. Section 2 describes the proposed formulation, while
Section 3 deals with the application of the formulation to three types of serial manipulators.
Finally, the conclusions are drawn in Section 4.

2 Formulation

2.1 Dynamic model

The dynamic model of a serial mechanism can be written as follows:

M# (q# )q̈# + v# (q# , q̇# ) + g# (q# ) = u (1)

where q# is a column-matrix of independent generalised coordinates, whose entries are

relative displacements of the joints, and u is a column-matrix of the generalised actuators’
efforts in the directions of the independent quasi-velocities p# = q̇# .
34 A.G. Coutinho and T.A.H. Coelho

In order to perform the dynamic balancing of a serial mechanism, it is required to

obtain the dynamic model of the unbalanced mechanism. Once in a serial mechanism
it is possible to express all the absolute velocities of the links’ centres of mass and
all the absolute angular velocities of the links as functions of q# and q̇# , the dynamic
model can be obtained without major difficulties by using analytical mechanics techniques,
like Lagrange Chen, Chen and Tsai (1990) and Kane Kane and Levinson (1985)
formalisms, Orsino’s method Orsino and Coelho (2015), and Boltzmann-Hamel equations
Altuzarra et al. (2015), allied to programs or libraries of programming languages that are
capable of using symbolic manipulation, such as Mathematica and SymPy.

2.2 Static balancing

After obtaining the dynamic model, the static balancing is performed by determining
the links’ centres of mass positions that annul the term g# . That can be achieved in
mechanisms whose revolute joints axes are in any direction and whose prismatic joints axes
are orthogonal to the gravity. The positioning of the centres of mass is done mechanically
by extending the mechanism’s bars and adding counterweights.

2.3 Dynamic balancing

The dynamic balancing is performed by attaching counter-rotating disks to the statically
balanced mechanism. One can take into account this structural modification in the dynamic
model by using the coupling subsystems technique of Orsino’s method 1 Orsino, Coutinho
and Coelho (2016); Orsino and Coelho (2015).
Let M0 be a mechanical subsystem composed of a statically balanced serial
mechanism, whose equation of motion is given by (1), with g# = 0. Let Mi be a
mechanical subsystem composed of a counter-rotating disk that will be coupled to the
mechanism, whose equation of motion is given by:

M# # # #
i ṗi + vi + gi = ui (2)

where p# i is a set of independent quasi-velocities, whose elements are non-null components

of the absolute angular velocity vector of the disk, expressed in a reference frame fixed
to the disk, and v# #
i = gi = ui = 0. In this model, only the rotational effect of the disk
inertias is considered. The translational effects of the disk inertias are taken into account in
the model of bar where the disk is attached.
By assuming that n counter-rotating disks will be attached to the mechanism, the
following definitions are made:
 
M# 0 . . . 0
 0 M#1 ... 0 
M′ =  . . . (3)
 
..
 . . ..
. . 
. 
0 0 . . . M#
n

h iT
T
v′ = v# T v# #T
1 ... vn
(4)

h iT
T
g′ = g# T g#
1 ... g#T
n (5)
 A new approach for obtaining the dynamic 35
h iT
T #T
p◦ = p#
1 ... pn (6)
T
p = p# T p◦ T

(7)

Let p◦ be the vector p◦ expressed as function of q# e p# , i.e.,

p◦ = p◦ (q# , p# ) (8)

The kinematic constraints matrix is defined as

 
1
C =  ∂p ◦
(9)
∂p#
The dynamic model of the serial mechanism coupled with the counter-rotating disks is
given by
# # #
M′ (q# )q̈# + v′ (q# , q̇# ) + g′ (q# ) = u (10)

where

M′ = CT M′ C
#
(11)

v′ = CT (M′ Ċq̇# + v′ )
#
(12)

g′ = CT g′
#
(13)

The dynamic balancing conditions are achieved by determining the system parameters that
# #
make M′ diagonal and v′ null.

3 Applying the technique

In this section, the proposed formulation will be applied in three different 3-dof serial
mechanisms. First, some definitions, valid for these mechanisms, are required:
 
D11 D12 D13
M# = D12 D22 D23  (14)
D13 D23 D33
   2   
D111 D122 D133 q̇1 D112 D113 D123 q̇1 q̇2
v# = D211 D222 D233  q̇22  + 2 D212 D213 D223  q̇1 q̇3  (15)
D311 D322 D333 q̇32 D312 D313 D323 q̇2 q̇3
 ⊤
g# = D1 D2 D3 (16)
 ⊤
q# = q1 q2 q3 (17)
 ⊤
u = u1 u2 u3 (18)

In order to employ the traditional notation, qi = θi and ui = τi for revolute joints, and
qi = di and ui = fi for prismatic joints.
36 A.G. Coutinho and T.A.H. Coelho

The Denavit-Hartenberg convention is used here not only to define the coordinate
system in each link, but also to enumerate the links and joints of the mechanism.

3.1 3-dof RRR planar serial mechanism

The entries of g# for the unbalanced mechanism are given by



 D1 = g [(m1 lg1 + m2 l1 + m3 l1 ) c(θ1 ) + (m2 lg2 + m3 l2 ) c (θ1 + θ2 )

 +m3 lg3 c (θ1 + θ2 + θ3 )]
(19)


 D2 = g[(m2 lg2 + m3 l2 )c(θ1 + θ2 ) + m3 lg3 c(θ1 + θ2 + θ3 )]
D3 = g[m3 lg3 c(θ1 + θ2 + θ3 )]


By performing the static balancing,

l1 (m2 +m3 )
 
 D1 = 0
  lg 1 = −
 m1
D2 = 0 ⇒ lg2 = − l2mm23 (20)
 
D3 = 0 lg 3 = 0
 

Substituting (20) in the mechanism model, the terms of the dynamic model of the statically
balanced mechanism are obtained

l12 (m2 +m3 )2 l22 m23




 D11 = Jz1 + Jz2 + Jz3 + m2 l12 + m3 (l12 + l22 ) + m1 + m2
l 2 m2
D22 = Jz2 + Jz3 + m3 l22 + 2m23






D33 = Jz3


D12 = D22 (21)

D13 = D23 = D33





v# = 0





g =0
 #

To perform the dynamic balancing, four counter-rotating disks are coupled to the
mechanism, as shown in Figure 1. Once the disks rotate in a single plane, their
corresponding dynamic models are as follows:
   
M# #
i = Jzi+3 ; pi = ωzi+3 , i = 1, 2, 3, 4 (22)

The counter-rotating disks 1 and 2 (rigid bodies 4 and 5) are coupled to link 1. The angular
displacement of disk 1 with respect to link 1 is θ2 , due to the belt transmission of the
motor 2 motion, while the angular displacement of disk 2 with respect to link 1 is βθ2 , with
β < 0, due to the gear transmission of disk 1 motion.
The counter-rotating disks 3 and 4 (rigid bodies 6 and 7) are coupled to link 2. The
angular displacement of disk 3 with respect to link 2 is θ3 , due to the belt transmission of
the motor 3 motion, while the angular displacement of disk 4 with respect to link 2 is γθ3 ,
with γ < 0, due to the gear transmission of disk 3 motion.
 A new approach for obtaining the dynamic 37
Figure 1 Dynamically balanced RRR planar serial mechanism

Thus, the following quasi-velocities constraints are obtained:

 

 ωz4 = ωz1 + θ̇2 
 ωz4 = θ̇1 + θ̇2 
θ̇1 + θ̇2


ω = ω + β θ̇ 
ω = θ̇ + β θ̇
z5 z1 2 z5 1 2  θ̇1 + β θ̇2 
⇒ ⇒ p◦ = 
 θ̇1 + θ̇2 + θ̇3  (23)



 ω z 6
= ω z 2
+ θ̇ 3 

 ω z 6
= θ̇ 1 + θ̇ 2 + θ̇ 3

ωz7 = ωz2 + γ θ̇3

ωz7 = θ̇1 + θ̇2 + γ θ̇3 θ̇1 + θ̇2 + γ θ̇3

 
100
 0 1 0 
 

1 0 0 1 
 
∴ C =  ∂p◦  = 
1 1 0 
 (24)
∂p # 
 1 β 0 

1 1 1 
11γ

By applying (21), (22) and (24) in (11), (12) and (13), the mechanism’s statically balanced
model coupled with the counter-rotating disks is obtained:

 ′

 D11 = D11 + Jz4 + Jz5 + Jz6 + Jz7
+ Jz5 β 2 + Jz6 + Jz7

D 22 = D22 + Jz4

 ′


D33 = D33 + Jz6 + Jz7 γ 2

 ′

D12′
= D12 + Jz4 + Jz5 β + Jz6 + Jz7 (25)

D13 = D13 + Jz6

 ′
+ Jz7 γ


 D′ = D13 ′
 ′ #23



v =0
38 A.G. Coutinho and T.A.H. Coelho

To perform the dynamic balancing, the values of β and γ as functions of the mechanism’s
#
parameters that makes M′ diagonal are found. Thus,

m2 2

( Jz2 +Jz3 +Jz4 +Jz6 +Jz7 +m3 l22 + 3 l2
D12
′
=0 
β=− m2

⇒ Jz5 (26)
D13
′
=0 J +J
γ = − z3Jz z6
7

By applying (26) in (25), the mechanism’s dynamic balanced model is as follows:


τ1 = k1 θ̈1

τ2 = k2 θ̈2 (27)

τ3 = k3 θ̈3


where


 k1 = Jz1 + Jz2 + Jz3 + Jz4 + Jz5 + Jz6 + Jz7 + m2 l12 + m3 (l12 + l22 )

l2 (m +m )2 l 2 m2

+ 1 2m1 3 + 2m23




l 2 m2

k2 = Jz2 + Jz3 + Jz4 + Jz6 + Jz7 + m3 l22 + 2m23

(28)
m2 l2 2



 Jz2 +Jz3 +Jz4 +Jz6 +Jz7 +m3 l22 + m3 2



 + Jz5
2

k = (Jz3 +Jz6 )(Jz3 +Jz6 +Jz7 )



3 Jz 7

3.2 3-dof RRR spatial serial mechanism

The entries of g# for the unbalanced mechanism are given by


 D1 = 0

D2 = g[(m2 lg2 + m3 l2 )c(θ2 ) + m3 lg3 c(θ2 + θ3 )] (29)

D3 = g[m3 lg3 c(θ2 + θ3 )]


By performing the static balancing,

( (
D2 = 0 lg2 = − l2mm23
⇒ (30)
D3 = 0 lg 3 = 0

Substituting (30) in the mechanism model, the terms of the dynamic model of the statically
balanced mechanism are obtained.
 A new approach for obtaining the dynamic 39

D = Jx2 s2 (θ2 ) + Jx3 s2 (θ2 + θ3 ) + Jz1 + Jy2 c2 (θ2 ) + Jy3 c2 (θ2 + θ3 )
 11


2
+m3 (l1 + l2 c(θ2 ))2 + (m2 l1 −m 3 l2 c(θ2 ))



m2


l 2 m2

D22 = Jz2 + Jz3 + m2 l22 + 2m23





D33 = Jz3





D12 = D13 = 0





D23 = D33 





D211 = − 1 Jx − Jy
s(2θ2 )

 2 2 2
 (31)
+ Jx3 − Jy3 − m3 l22 (1 + m





3
m2 ) s(2θ 2 + 2θ 3 )

 
1


D = J − J + 2θ 3)

 s(2θ
 311 2 y3 x3 2



D111 = D122 = D133 = D222 = D233 = D322 = D333 = 0







 D112 = −D211
D113 = −D311





D123 = D212 = D213 = D223 = D312 = D313 = D323 = 0





g =0
 #

To perform the dynamic balancing, two counter-rotating disks are coupled to the
mechanism, as shown in Figure 2. Their respective dynamic models are as follows

   
Jxi+3 0 0 ωxi+3
M#
i =
 0 Jyi+3 0  ; p#
i = ωyi+3 , i = 1, 2
  (32)
0 0 Jzi+3 ωzi+3

The counter-rotating disks 1 and 2 (rigid bodies 4 and 5) are coupled to link 2. The angular
displacement of disk 1 with respect to link 2 is θ3 , due to the belt transmission of the motor
3 motion, while the angular displacement of disk 2 with respect to link 2 is βθ3 , with
β < 0, due to the gear transmission of disk 1 motion.
Thus, the following quasi-velocities constraints are obtained:

   

 0 
 ωx 4 = (θ̇1 s(θ2 ))c(θ3 ) + (θ̇1 c(θ2 ))s(θ3 )
4 ]B4 = [1]B4 | B2 [ω2 ]B2 +  0 

[ω

ω = −(θ̇1 s(θ2 ))s(θ3 ) + (θ̇1 c(θ2 ))c(θ3 )
   





 y4
 
 θ̇3 ω
z4 = θ̇2 + θ̇3
  ⇒ (33)


 0  ωx 5

 = (θ̇1 s(θ2 ))c(βθ3 ) + (θ̇1 c(θ2 ))s(βθ3 )
[ω5 ]B5 = [1]B5 | B2 [ω2 ]B2 + 
 
 0  ωy = −(θ̇1 s(θ2 ))s(βθ3 ) + (θ̇1 c(θ2 ))c(βθ3 )
  

 5

 


β θ̇3 ω z5 = θ̇2 + β θ̇3

 
θ̇1 s(θ2 + θ3 )
 θ̇ c(θ + θ ) 
 1 2 3 
 θ̇2 + θ̇3 
 
⇒p =
◦
(34)
 θ̇1 s(θ2 + βθ3 ) 

 
θ̇1 c(θ2 + βθ3 )
θ̇2 + β θ̇3
40 A.G. Coutinho and T.A.H. Coelho

Figure 2 Dynamically balanced RRR spatial serial mechanism

 
1 00

 0 1 0
  
 0 0 1
1  s(θ2 + θ3 ) 0 0 
 
C =  ∂p◦  = 
 c(θ2 + θ3 ) 0 0 
 (35)
∂p# 
 0 1 1
 s(θ2 + βθ3 ) 0 0 
 
c(θ2 + βθ3 ) 0 0 
0 1β

By applying (31), (32) and (35) in (11), (12) and (13), the mechanism’s statically balanced
model coupled with the counter-rotating disks is obtained:




 D11
′
= D11 + Jx4 s2 (θ2 + θ2 ) + Jx5 s2 (βθ2 + θ2 ) + Jy4 c2 (θ2 + θ2 )
+Jy5 c2 (βθ2 + θ2 )





D22 = D22 + Jz4 + Jz5
′




= D33 + Jz4 + Jz5 β 2

D33
′








 D12 = D13
′ ′
=0

D23 = D23 + Jz4 + Jz5 β
 ′

D211
′
= D211 (36)


D 311 = D311

 ′



D111 = D122 = D133 = D222
= D233 = D322 = D333 = 0


 ′ ′ ′ ′ ′ ′ ′
 

D112
′
= D112 + 14 Jx4 − Jy4 s(2θ2 + 2θ3 ) + Jx5 − Jy5 s(2βθ2 + 2θ3 )



  

1
D = D + (J − J )s(2θ + 2θ ) + (J − J )s(2βθ + 2θ )
 ′
113 2 3 2 3
 x y x y
 113 4 4 4 5 5



D ′ = D ′ = D ′ = D ′ = D ′ = D ′ = D ′ = 0
123 212 213 223 312 313 323
 A new approach for obtaining the dynamic 41
# #
To perform the dynamic balancing, M′ becomes diagonal for a specific value of β and v′
is null for a given set of the mechanism’s parameters. Thus,

J +J
 

 D23
′
=0 
 β = − z3Jz z4
  5

 211 = 0
D
 ′ 
 J x 2 = J y2

 

D311
′
=0 ⇒ Jx3 = Jy3 + m3 l22 (1 + m3 (37)
  m2 )
D112 = 0
′
 J x 4 = J y4

 


 

D ′ = 0 
J x 5 = J y5

113

By applying (37) in (36), the mechanism’s dynamic balanced model is as follows


τ1 = k1 θ̈1

τ2 = k2 θ̈2 (38)
f3 = k3 d¨3



where


2 2 2 l12 m23
k1 = Jz1 + Jy2 + Jy3 + Jy4 + Jy5 + m22l1 2+ m3 (l1 + l2 ) +

 m2
l m
k2 = Jz2 + Jz3 + Jz4 + Jz5 + m3 l22 + 1m23 (39)
k3 = (Jz3 +Jz4 )(Jz3 +Jz4 +Jz5 )


Jz 5

Note that the necessary conditions for the dynamic balancing of this mechanism require
very high longitudinal moments of inertia for the bars 2 and 3, which is not practically
feasible for an industrial manipulator. Among the five conditions of (37), let us consider
the following two:

(
J x 2 = J y2
m3 (40)
Jx3 = Jy3 + m3 l22 (1 + m2 )

According to Denavit-Hartenberg convention, the x-axis is the longitudinal direction while

the y-axis and the z-axis are the corresponding transversal directions, for the bars 2
and 3. In this case, typically, the Jx moment of inertia is quite low in a comparison
with Jy and Jz . In order to satisfy the balancing conditions, the values of Jx should
be equal to Jy for bar 2 and higher than Jy for bar 3. Consequently, the link section
should be increased according, which would lead to an inconvenient extremely large cross
section area.
In the following example, which corresponds to another 3-dof spatial serial mechanism,
such inconvenience will not occur.
42 A.G. Coutinho and T.A.H. Coelho

3.3 3-dof RRP spatial serial mechanism (SCARA)

The entries of g# for the unbalanced mechanism are given by


D1 = g[(m1 lg1 + m2 l1 + m3 l1 )c(θ1 ) + m2 lg2 c(θ1 + θ2 )]

D2 = g[m2 lg2 c(θ1 + θ2 )] (41)

D3 = 0


By performing the static balancing,

( (
D1 = 0 lg1 = − l1 (mm
2 +m3 )

⇒ 1 (42)
D2 = 0 lg 2 = 0

Substituting (42) in the mechanism model, the terms of the dynamic model of the statically
balanced mechanism (Figure 3) are obtained.

l12 (m2 +m3 )2


 D11 = Jz1 + Jz2 + Jz3 + m2 l12 + m3 l12 + m1




 D22 = Jz2 + Jz3

D33 = m3


D12 = D22 (43)

D13 = D23 = 0





v# = 0





g# = 0

To perform the dynamic balancing, two counter-rotating disks are coupled to the
mechanism. Once the disks rotate in a single plane, their corresponding dynamic models
are as follows:
   
M# #
i = Jzi+3 ; pi = ωzi+3 , i = 1, 2 (44)

The counter-rotating disks 1 and 2 (rigid bodies 4 and 5) are coupled to link 1. The angular
displacement of disk 1 with respect to link 1 is θ2 , due to the belt transmission of the motor
2 motion, while the angular displacement of disk 2 with respect to link 1 is βθ2 , with
β < 0, due to the gear transmission of disk 1 motion.
Thus, the following quasi-velocities constraints are obtained:
( (  
ωz4 = ωz1 + θ̇2 ωz4 = θ̇1 + θ̇2 θ̇1 + θ̇2
⇒ ⇒p = ◦
(45)
ωz5 = ωz1 + β θ̇2 ωz5 = θ̇1 + β θ̇2 θ̇1 + β θ̇2

 
  100
1 0 1 0
 
C =  ∂p◦  = 
0 0 1
 (46)
∂p #  1 1 0
1β0
 A new approach for obtaining the dynamic 43

Figure 3 Statically balanced RRP spatial serial mechanism

By applying (43), (44) and (46) in (11), (12) and (13), the mechanism’s statically balanced
model coupled with the counter-rotating disks is obtained:

 ′

 D11 = D11 + Jz4 + Jz5
D ′ = D + J + J β 2



 22 22 z4 z5
D = D

 ′
 33
 33
D12′
= D12 + Jz4 + Jz5 β (47)

D13 ′
=0




 D =0
′
 ′ #23



v =0

To perform the dynamic balancing, the values of β as function of the mechanism’s

#
parameters that make M′ diagonal are found. Thus,

Jz2 + Jz3 + Jz4

′
D12 =0⇒β=− (48)
Jz5
44 A.G. Coutinho and T.A.H. Coelho

By applying (48) in (47), the mechanism’s dynamic balanced model is as follows:


τ1 = k1 θ̈1

τ2 = k2 θ̈2 (49)
f3 = k3 d¨3



where

2 2 l12 (m2 +m3 )2
 k 1 = J z 1 + J z 2 + J z 3 + J z 4 + J z 5 + m2 l 1 + m3 l 1 +

 m1
(Jz2 +Jz3 +Jz4 )(Jz2 +Jz3 +Jz4 +Jz5 )
k2 = Jz5
(50)


k = m
3 3

Note that the necessary conditions for the dynamic balancing of this mechanism do not
require restrictions on the mechanism inertia parameters, as in the previous example. Thus,
as the mechanism in the previous example and the mechanism of this example are both
spatial serial mechanisms that perform end-effector translations in three axes, it can be
said that the mechanism in question is a good alternative to Section 3.2 mechanism in
applications in which the dynamic balancing is advantageous to the system.

4 Conclusions

This work dealt with a systematic formulation for the adaptive balancing. This special
formulation uses the dynamic coupling between subsystems in order to derive the equations
of motion of the whole system explicitly. Consequently, this feature allows the automatic
generation of the adaptive balancing conditions. Two traditional balancing techniques were
employed here: the addition of counterweight and counter-rotating disks coupled to the
moving links. In addition, the feasibility of the dynamic decoupling for three distinct types
of serial manipulators was discussed regarding the achievement of such balancing and
the complexity level of the modified mechanical structure. The balancing conditions were
developed for 3-dof spatial and planar open-loop kinematic chain mechanisms, whose
topologies are composed of revolute and prismatic joints. By analysing the necessary
conditions, one can notice that the adaptive balancing brings great benefits for the planar
RRR and the spatial RRP. However, for the spatial RRR, in spite of the achievement of
the adaptive balancing, the modifications in the mechanical structure require very high
longitudinal moments of inertia for the second and third bars of the mechanism, which
would lead to bars with extremely large cross-section areas. Consequently, the authors
believe that the discussion provided here might help the designer to choose an adequate
topology for a specific application taking advantage of the adaptive balancing whenever it
brings no further consequences in terms of the added inertias.
 A new approach for obtaining the dynamic 45

References
Agrawal, S.K. and Fattah, A. (2004) ‘Gravity-balancing of spatial robotic manipulators’, Mechanism
and Machine Theory, Vol. 39, pp.1331–1344.
Alici, G. and Shirinzadeh, B. (2003) ‘Optimum force balancing with mass distribution and a single
elastic element for a five-bar parallel manipulator’, Proceedings of the IEEE International
Conference on Robotics and Automation, 14–19 September, Taipei, pp.3666–3671.
Alici, G. and Shirinzadeh, B. (2006) ‘Optimum dynamic balancing of planar parallel manipulators
based on sensitivity analysis’, Mechanism and Machine Theory, Vol. 41, pp.1520–1532.
Altuzarra, O., Eggers, P.M., Campa, F.J., Roldan-Paraponiaris, C. and Pinto, C. (2015) ‘Dynamic
modelling of lower-mobility parallel manipulators using the Boltzmann-Hamel equations’,
Mechanisms, Transmissions and Applications, Vol. 31, pp.157–165.
Arakelian, V. and Sargsyan, S. (2012) ‘On the design of serial manipulators with decoupled
dynamics’, Mechatronics, Vol. 22, No. 6, pp.904–909.
Arakelian, V.H. and Smith, M.R. (2008) ‘Design of planar 3-dof 3-RRR reactionless parallel
manipulators’, Mechatronics, Vol. 18, pp.601–606.
Briot, S., Arakelian, V. and Le Baron, J-P. (2012) ‘Shaking force minimization of high-speed
robots via centre of mass acceleration control’, Mechanism and Machine Theory, Vol. 57,
pp.1–12.
Chen, J., Chen, D.Z. and Tsai, L.W. (1990) A systematic methodology for the dynamic analysis of
articulated gear-mechanisms.
Cheng, H. and Yiu, Y.-K. (2003) ‘Dynamics and control of redundantly actuated parallel
manipulators’, IEEE/ASME Transactions on Mechatronics, Vol. 8, No. 4, pp.483–491.
Coelho, T.A.H., Yong, L. and Alves, V.F.A. (2004) ‘Decoupling of dynamic equations by means of
adaptive balancing of 2-dof open-loop mechanisms’, Mechanism and Machine Theory, Vol. 39,
pp.871–881.
Dehkordi, M.B., Frisoli, A., Sotgiu, E. and Bergamasco, M. (2012) ‘Modelling and experimental
evaluation of a static balancing technique for a new horizontally mounted 3-UPU parallel
mechanism’, International Journal of Advanced Robotic Systems, Vol. 9, pp.193–205.
Gosselin, C.M., Vollmer, F., Cote, G. and Wu, Y. (2004) ‘Synthesis and design of reactionless three-
degree-of-freedom parallel mechanisms’, IEEE Transactions on Robotics and Automation,
Vol. 20, No. 2, pp.191–199.
Kane, T.R. and Levinson, D.A. (1985) Dynamics, Theory and Applications. McGraw-Hill series in
mechanical engineering. McGraw Hill, New York, USA.
Moradi, M., Nikoobin, A. and Azadi, S. (2010) ‘Adaptive decoupling for open chain planar robots’,
Transaction B: Mechanical Engineering, Vol. 17, No. 5, pp.376–386.
Orsino, R.M.M. and Coelho, T.A.H. (2015) ‘A contribution on the modular modelling of multibody
systems’, of the Royal Society A: Mathematical, Physical and Engineering Science, Vol. 471,
No. 2183, p.20150080.
Orsino, R.M.M., Coelho, T.A.H. and Pesce, C.P. (2015) ‘Analytical mechanics approaches in the
dynamic modelling of delta mechanism’, Robotica, Vol. 33, No. 4, pp.953–973.
Orsino, R.M.M., Coutinho, A.G. and Coelho, T.A.H. (2016) ‘Dynamic modelling and control of
balanced parallel mechanisms’, Dynamic Balancing of Mechanisms and Synthesizing of Parallel
Robots, (Eds.): D. Zhang, and B. Wei, Springer, Cham, Switzerland, pp.403–445.
Russo, A., Sinatra, R. and Xi, F. (2005) ‘Static balancing of parallel robots’, Mechanism and Machine
Theory, Vol. 40, pp.191–202.
Seo, J.-T., Woo, J.H., Lim, H., Chung, J., Kim, W.K. and Yi, B.-J. (2013) ‘Design of an
antagonistically counter-balancing parallel mechanism IEEE/RSJ International Conference on
Intelligent Robots and Systems (IROS), 3-7 November, Tokyo, pp.2882–2887.
Van der Wijk, V. (2013) ‘Shaking moment balancing of mechanisms with principal vectors and
moments’, Frontiers of Mechanical Engineering, Vol. 8, No. 1, pp.10–16.
46 A.G. Coutinho and T.A.H. Coelho

Wang, J. and Gosselin, C.M. (1999) ‘Static balancing of spatial three-degree-of-freedom parallel
mechanisms’, Mechanism and Machine Theory, Vol. 34, pp.437–452.
Wang, J. and Gosselin, C.M. (2000) ‘Static balancing of spatial four-degree-of-freedom parallel
mechanisms’, Mechanism and Machine Theory, Vol. 35, pp.563–592.
Wang, K., Luo, M., Mei, T., Zhao, J. and Cao, Y. (2013) ‘Dynamics analysis of a three-DOF planar
serial-parallel mechanism for active dynamic balancing with respect to a given trajectory’,
International Journal of Advanced Robotic Systems, Vol. 10, pp.23–33.
Wu, Y. and Gosselin, C.M. (2005) ‘Design of reactionless 3-dof and 6-dof parallel manipulators
using parallelepiped mechanisms’, IEEE Transactions on Robotics, Vol. 21, No. 5, pp.821–833.

Nomenclature

a, b, . . . Scalars, components of column-matrices, components of matrices or

indexes
A, B, . . . Scalars, components of column-matrices or components of matrices
a, b, . . . Column-matrices
A, B, . . . Matrices
a, b, . . . Vectors
A, B, . . . Coordinate systems
A , B, . . . Sets or multibody mechanical systems
Bi Coordinate system fixed in the ith rigid body of the mechanical
system
C Kinematic constraints matrix
c(.) Shorthand notation for cos(.)
g Gravitational acceleration
g# Generalised gravitational forces column-matrix of a serial mechanism
g#i Generalised gravitational forces column-matrix of a counter-rotating
disc
g′ Generalised uncoupled gravitational forces column-matrix of a serial
mechanism coupled with counter-rotating discs
g′
#
Generalised gravitational forces column-matrix of a serial
mechanism coupled with counter-rotating discs
Jxi , Jyi , Jzi Principal moments of inertia of the ith rigid body of the mechanical
system
li Length of the ith bar of a serial mechanism
l gi Position of the mass center of the ith bar relative to the ith joint and
of a serial mechanism
mi Mass of the ith rigid body of the mechanical system
M# Generalised inertia matrix of a serial mechanism
M#
i Generalized inertia matrix of a counter-rotating disc
M′ Generalised uncoupled inertia matrix of a serial mechanism coupled
with counter-rotating discs
M′
#
Generalised inertia matrix of a serial mechanism coupled with
counter-rotating discs
N Inertial reference frame
p# Independent quasi-velocities column-matrix
p◦ Redundant quasi-velocities column-matrix
p Quasi-velocities column-matrix
qi Generalised coordinate
q# Independent generalised coordinates column-matrix
s(.) Shorthand notation for sin(.)
ui Effort made by the ith actuator of a serial mechanism
u Generalised actuators’ efforts column-matrix
 A new approach for obtaining the dynamic 47

v# Generalised coupled gyroscopic inertia forces column-matrix of a

serial mechanism
v#
i Generalised coupled gyroscopic inertia forces column-matrix of a
counter-rotating disc
v′ Generalised uncoupled gyroscopic inertia forces column-matrix of a
serial mechanism coupled with counter-rotating discs
v′
#
Generalised coupled gyroscopic inertia forces column-matrix of a
serial mechanism coupled with counter-rotating discs
[ωi ]Bj Angular velocity of the ith rigid body of the mechanical system
measured relatively to a inertial reference frame N , written in the
basis of Bj
ωx i 1st component of [ωi ]Bi
ωyi 2nd component of [ωi ]Bi
ω zi 3rd component of [ωi ]Bi
0 Null column-matrix or null matrix
1 Identity matrix
1 Identity tensor
[1 ]B i | Bj Change of basis matrix, i.e. [v]Bi = [1]Bi | Bj · [v]Bj
[·]T Matrix transposition