Robotica (2006) volume 24, pp. 125130.

2005 Cambridge University Press

doi:10.1017/S0263574705001670 Printed in the United Kingdom
Dynamic modelling of a 3-DOF parallel manipulator using
recursive matrix relations
Stefan Staicu*, Dan Zhang
1
** and Radu Rugescu***
(Received in Final Form: January 25, 2005)
SUMMARY
In this paper, a simple and convenient method Recursive
Matrix method is proposed for kinematic and dynamic
analysis of all types of complex manipulators. After
addressing the principle of the method, an example a
3-DOF parallel manipulator with prismatic actuators is
demonstrated for the efciency of the method in solving
kinematic and dynamic problems of complex manipulators.
With the inverse kinematic solutions, the inverse dynamic
problem is solved with the virtual powers method. Matrix
relations and graphs of the acting forces and powers for all
actuators are analysis and determined. It is shown that the
proposed method is an effective mean for kinematic and
dynamic modelling of parallel mechanisms.
KEYWORDS: Recursive matrix; Dynamic modelling;
Complex manipulators; Parallel mechanisms.
I. INTRODUCTION
Parallel mechanisms generally comprise two platforms
which are connected by joints or legs acting in parallel.
1
Over the past decades, parallel mechanisms have received
more and more attention from researchers and industries.
They can be found in several practical applications, such
as aircraft simulators,
2
positional tracker,
3
telescopes
4
and
micro-motion device.
5
More recently, they have been used in the development
of high precision machine tools
68
by many companies
such as Giddings & Lewis, Ingersoll, Hexel, Geodetic and
Toyoda, and others. The Hexapod machine tools are one of
the successful applications.
Parallel mechanisms have signicant advantages over
serial mechanisms and in particular possess reduced moving
mass and higher stiffness. Thus, parallel mechanisms can
work at higher velocities, and yet maintain sufcient rigidity
to deliver high levels of accuracy.
The analysis of parallel manipulators is usually imple-
mented through analytical method in classical mech-
anics,
911
in which projection and resolution of vector
1
Corresponding author: email: [email protected]

Department of Mechanics, University Politehnica of Bucharest

(Romania).

Faculty of Engineering and Applied Science, University of

Ontario Institute of Technology, Oshawa, Ontario, Canada, L1H
7K4.

Department of Aeronautics, University Politehnica of

Bucharest (Romania).
equations on the reference axes are written in a considerable
number of cumbersome, scalar equations and the solution
are rendered by large scale computations together with time
consuming computer codes.
11
In this paper, a new method
recursive matrix method is introduced. It has been proved to
reduce the number of equations and computation operations
signicantly by using a set of matrices for kinematics and
dynamic modelling.
12,13
A spatial 3-DOF parallel mechanism [8], which can be
used in several applications, including machine tools is
proposed in this paper. Existing 3-DOF parallel mechanisms
can be classied in terms of the types of actuated joints, and
the structures of the supporting frames. In regard to the
types of actuated joints, they can be either revolute or
prismatic. Since the prismatic joints can easily achieve
high accuracy and heavy loads, the majority of the 3DOF
parallel mechanisms in reality use actuated prismatic joints.
A prismatic joint can have an extensible length or a xed
length. The 3DOF parallel mechanism discussed in this
paper belongs to the type with extensible length and a
passive leg located in the centre to improve the stiffness.
In comparison with an approach for kinematic modelling
of robot manipulator, a dynamic modelling approach
could be systematized easily. Some systematic approaches
have been developed for general-purpose parallel mech-
anisms analysis.
1417
Meanwhile, quite a few of special
approaches have been conducted for dynamic modelling
of specic parallel mechanism congurations;
1822
Kane
and Levinson
23
obtained some vector recursive relations
concerning the equilibriumof generalized forces that are app-
lied to a serial manipulator. Sorli et al
24
conducted the
dynamic modelling for Turin parallel manipulator, though the
mechanism has three identical legs, it has 6-DOF. However,
to the best of our knowledge, there is no efcient dynamic
modelling approach available for parallel manipulator.
II. INVERSE KINEMATIC MODELLING
Let Ox
0
y
0
z
0
(T
0
) be a xed Cartesian frame. A spatial
manipulator with three degrees of freedom is moving with
respect to this frame; the mechanism consists of four
kinematical chains, including three variable length legs with
identical topology and one passive constraining leg, all
connecting the xed base to a moving platform (Fig. 1).
In this 3-DOF parallel mechanism, a kinematical chain,
associated with one of the three identical active legs, was
introduced between the base and the moving platform. It
consists of a xed Hook joint, characterized by the angular
126 Recursive matrix
A
B
C
O
Fig. 1. Parallel manipulator with prismatic actuators.
velocity
A
10
=
A
10
and the angular acceleration
A
10
=
A
10
, and
a moving link of length l
2
, mass m
2
and tensor of inertia

J
2
,
which has a relative rotation with the angle
A
21
, so that

A
21
=
A
21
,
A
21
=
A
21
. An actuated prismatic joint is as well as
a moving link to the A
3
x
A
3
y
A
3
z
A
3
(T
A
3
) frame, having a relative
displacement
A
32
, velocity v
A
32
=

A
32
and acceleration
A
32
=

A
32
. It has the length l
3
, mass m
3
and tensor of inertia

J
3
.
Finally, a ball-joint is attached to the platform.
The fourth constraining chain has a different architecture.
It consists of a prismatic joint attached to the base and a
moving link of length l
1
and mass m
1
, having a purely vertical
displacement
D
10
, velocity v
D
10
=

D
10
and acceleration
D
10
=

D
10
. A Hook joint is attached to the moving platform. The
moving platform is an equilateral triangle with edge l, mass
m
4
and tensor of inertia

J
4
.
Rotations of the moving platformare dened by the angles

D
21
and
D
32
in the local coordinates (Fig. 2). The orientation
of the joints A, B, C on the xed platform (Fig. 1) is given
by

A
= 0,
B
=
2
3

C
=
2
3
. (1)
For convenience, the independent coordinates of the
platform have been chosen as z
G
0
,
1
,
2
, where z
G
0
is the
height of the platform and
1
,
2
are the angles of the Hook
joint.
Assuming the passive leg D on the OD
1
D
2
D
3
path, the
passing matrices are derived:
d
10
= e, d
21
= d

21
a
1
, d
32
= d

32
a
2
. (2)
Now, considering the OA
1
A
2
A
3
A
4
track of the limb A, the
transfer matrices are given by
a
10
= a

10
a
1
a
A

, a
21
= a

21
a

a
2
, a
32
= a
1
, (3)
Where:
25
a
1
=

0 0 1
0 1 0
1 0 0

, a
2
=

0 0 1
1 0 0
0 1 0

,
x
1
A
z
1
A
x
2
A
z
2
A
x
3
A
z
3
A
x
0
y
0
z
0
z
2
D
x
2
D
z
1
D
z
3
D
y
3
D
G
D
O
D
1
A
1
A
2
A
3
A
4
D
2
D
3

32

21
D

21
A

10
A

10
D
y
1
D

32
A
Fig. 2. Kinematical scheme of the manipulator.
a
A

cos
A
sin
A
0
sin
A
cos
A
0
0 0 1

cos sin 0
sin cos 0
0 0 1

,
a

k,k1
=

cos
A
k,k1
sin
A
k,k1
0
sin
A
k,k1
cos
A
k,k1
0
0 0 1

(4)
a
k0
=
k

j=1
a
kj+1, kj
, (k = 1, 2, 3).
The same equations can be written for the other two loops
O-B and O-C of the mechanism (Fig. 2).
Suppose the absolute motion of the platform is given by
z
G
0
= l
1
+l
6
+z
G
0

1 cos
2
3
t

i
=

1 cos
2
3
t

(i = 1, 2). (5)
To solve the inverse kinematic problem, the passive leg can
be taken as a serial 3-DOF mechanism with the coordinates
determined by the conditions
d
T
30
d
30
= a, u
T
3

r
D
10
+
D
10
u
3
+d
T
10
r
D
21

= z
G
0
, (6)
Recursive matrix 127
where a = a
2
a
1
is a rotation matrix from the frame Ox
0
y
0
z
0
to Gx
G
y
G
z
G
, and r
D
10
= l
6
u
3
, r
D
21
= l
1
u
3
d
T
30
=

1 0 0
0 0 1
0 1 0

, u
3
=

0
0
1

, u
3
=

0 1 0
1 0 0
0 0 0

(7)
a
1
=

1 0 0
0 cos
1
sin
1
0 sin
1
cos
1

, a
2
=

cos
2
0 sin
2
0 1 0
sin
2
0 cos
2

It results

D
10
= z
G
0
l
1
l
6
,
D
21
=
1
,
D
32
=
2
. (8)
Once the solution of the inverse kinematic of this 3-DOF
serial chain is found, the complete position and orientation
of the limbs A, B, C can be determined by the following
geometric conditions
r
A
10
+
3

k=1
a
T
k0
r
A
k+1,k
d
T
30
r
A
4
3
= r
B
10
+
3

k=1
b
T
k0
r
B
k+1,k
d
T
30
r
B
4
3
= r
C
10
+
3

k=0
c
T
k0
r
C
k+1,k
d
T
30
r
C
4
3
= z
G
0
u
3
, (9)
where
r
A
10
= l
0
a
AT

u
1
, r
A
21
=

0,
r
A
32
= l
5
u
1
+
A
32
a
T
32
u
3
, r
A
43
= l
3
u
3
.
u
1
=

1
0
0

, u
2
=

0
1
0

, r
A
4
3
=

l
4
cos
A
0
l
4
sin
A

(10)
III. Velocities and accelerations
The motion of all links is characterized by the following skew
symmetric matrices [23]

A
k0
= a
k,k1

A
k1,0
a
T
k,k1
+
A
k,k1
u
3
. (11)
These matrices are associated with the absolute angular
velocities, given by the recurrence relations

A
k0
= a
k,k1

A
k1,0
+
A
k,k1
u
3
. (12)
The velocity v
A
k0
of the joint A
k
can be obtained as
v
A
k0
= a
k,k1

v
A
k1,0
+
A
k1,0
r
A
k,k1

+v
A
k,k1
u
3
,
v
A
,1
=

0 ( = 1, 2, 4). (13)
The following matrix conditions of connectivity constitute
the inverse kinematical model

A
10
u
T
i
a
T
10
u
3
a
T
21

r
A
32
+a
T
32
r
A
43

+
A
21
u
T
i
a
T
20
u
3

r
A
32
+a
T
32
r
A
43

+v
A
32
u
T
i
a
T
30
u
3
= v
D
10
u
T
i
u
3
+ u
T
i
d
T
30

D
30
r
A
4
3
. (14)
These relations generate Jacobian matrix of the mecha-
nism, dening the robot workspace. From equation (14),
relative velocities
A
10
,
A
21
, v
A
32
result as functions of
platforms characteristic velocities
v
D
10
= z
G
0
,
D
21
=
1
,
D
32
=
2
(15)
Assume that the robot has a virtual motion determined
by the angular velocities v
Av
32a
=1, v
Bv
32a
=0, v
Cv
32a
=0. The
characteristic virtual velocities, expressed as function of
manipulators position, are given by new connectivity
conditions of relative velocities
u
T
i
a
T
30

v
Av
30a
+
Av
30a
r
A
43

= u
T
i
d
T
30

v
Dv
30a
+
Dv
30a
r
A
4
3

. (16)
Other two compatibility relations of the loops O-B and
O-C can be obtained, if one considers v
Bv
32b
=1 and v
Cv
32c
=1.
Some new connectivity relations of the angular
accelerations
A
10
,
A
21
,
A
32
of the links can be obtained

A
10
u
T
i
a
T
10
u
3
a
T
21

r
A
32
+a
T
32
r
A
43

+
A
21
u
T
i
a
T
20
u
3

r
A
32
+a
T
32
r
A
43

+
A
32
u
T
i
a
T
30
u
3
=
D
10
u
T
i
u
3
+ u
T
i
d
T
30

D
30

D
30
+
D
30

r
A
4
3

A
10

A
10
u
T
i
a
T
10
u
3
u
3
a
T
21

r
A
32
+a
T
32
r
A
43

A
21

A
21
u
T
i
a
T
20
u
3
u
3

r
A
32
+a
T
32
r
A
43

2
A
10

A
21
u
T
i
a
T
10
u
3
a
T
21
u
3

r
A
21
+a
T
32
r
A
43

2
A
10
v
A
32
u
T
i
a
T
10
u
3
a
T
21
a
T
32
u
3
2
A
21
v
A
32
u
T
i
a
T
20
u
3
a
T
32
u
3
.
(17)
When the other two kinematical chains are pursued, the
similar relations can be easily obtained.
The following relations are for the angular accelerations

A
k0
and the linear accelerations
A
k0
of the joints

A
k0
= a
k,k1

A
k1,0
+
A
k,k1
u
3
+
A
k,k1
a
k,k1

A
k1,0
a
T
k,k1
u
3

A
k0

A
k0
+
A
k0
=a
k,k1

A
k1,0

A
k1,0
+
A
k1,0

a
T
k,k1
+
A
k,k1

A
k,k1
u
3
u
3
+
A
k,k1
u
3
+2
A
k,k1
a
k,k1

A
k1,0
a
T
k,k1
u
3

A
k0
= a
k,k1

A
k1,0
+a
k,k1

A
k1,0

A
k1,0
+
A
k1,0

r
A
k,k1
,

A
,1
=

0 ( = 1, 2, 4). (18)
The equations (14), (17) are the inverse kinematical model
of the manipulator.
128 Recursive matrix
IV. Motion simulation
For the inverse dynamic problem, the virtual power method is
applied. Graphs and recursive matrix relations for the forces
and powers of the three actuators are obtained.
The three actuators forces are

f
A
32
= f
A
32
u
3
,

f
B
32
= f
B
32
u
3
,

f
C
32
= f
C
32
u
3
. (19)
The force of inertia and the resultant moment of the forces
of inertia of the rigid body T
k
are determined in respect to
the centre of the joint A
k
. On the other hand the characteristic
vectors

f

k
and m

k
evaluate the inuence of the weight action
m
k
g and all other external and internal forces applied to the
same T
k
manipulator link.
Given the absolute motion of the platform by equation (5),
the position, velocity and acceleration of each joint can be
determined, and the wrench about A
k
can also be determined.
Assuming that friction forces at the joints are negligible, we
can then calculate the actuating forces.
In the virtual-velocities method, the dynamic equilibrium
condition of the mechanism requires that the virtual power
of all external, internal and inertia forces, which is developed
during a general virtual displacement, to be set null. Applying
the fundamental equations of the parallel robots dynamics,
13
the following matrix equation is obtained
f
A
32
= u
T
3

F
A
3
+
Av
21a

M
A
2
+
Bv
21a

M
B
2
+
Cv
21a

M
D
2
+v
Dv
10a

F
D
1
+
Dv
32a

M
D
3

, (20)
where

F
A
k0
= m
A
k

A
k0
+m
A
k

A
k0

A
k0
+
A
k0

r
CA
k
+9.81m
A
k
a
k0
u
3

M
A
k0
= m
A
k
r
CA
k

A
k0
+

J
A
k

A
k0
+
A
k0

J
A
k

A
k0
+9.81m
A
k
r
CA
k
a
k0
u
3

F
A
k
=

F
A
k0
+a
T
k+1,k

F
A
k+1
,

M
A
k
=

M
A
k0
+a
T
k+1,k

M
A
k+1
+ r
A
k+1,k
a
T
k+1,k

F
A
k+1
,
(k = 1, 2, 3). (21)
The equations (20) and (21) represent the inverse dynamic
model of the-3-DOF parallel manipulator with prismatic
actuators.
An example is given with the following parameters

1
= /12,

2
= /18, z
G
0
= 0.1 m
l = 0.75 m, D
1
D
2
= l
1
= 0.65 m,
l
2
= 1.2 m, A
3
A
4
= l
3
= 1.25 m
GA
4
= l/

3, A
2
A
3
= l
5
= 0.25 m, OD
1
= l
6
= 0.1 m
sin = (l
1
+l
6
)/(l
3
+l
5
), l
0
= (l
3
+l
5
) cos +l
4
m
1
= 1 kg, m
2
= 2.5 kg, m
3
= 1.5 kg, m
4
= 5 kg.
Finally, one obtains the graphs of the forces f
A
32
(Fig. 3),
f
B
32
(Fig. 4), f
C
32
(Fig. 5) and powers p
A
32
(Fig. 6), p
B
32
(Fig. 7),
p
C
32
(Fig. 8) of the three prismatic actuators.
0 0.5 1 1.5 2 2.5 3
75
80
85
90
95
100
105
110
115
120
125
t(s)
f
3
2
a
(
N
)
Fig. 3. Force f
A
32
of the rst actuator.
80
70
60
50
40
30
20
10
0
0 0.5 1 1.5 2 2.5 3
f
3
2
b
(
N
)
t(s)
Fig. 4. Force f
B
32
of the second actuator.
0 0.5 1 1.5 2 2.5 3
78
79
80
81
82
83
84
85
t(s)
f
3
2
c
(
N
)
Fig. 5. Force f
C
32
of the third actuator.
Recursive matrix 129
0 0.5 1 1.5 2 2.5 3
-8
-6
-4
-2
0
2
4
6
8
t(s)
p
3
2
a
(
w
)
Fig. 6. Power p
A
32
of the rst actuator.
0 0.5 1 1.5 2 2.5 3
-15
-10
-5
0
5
10
15
t(s)
p
3
2
b
(
w
)
Fig. 7. Power p
B
32
of the second actuator.
0 0.5 1 1.5 2 2.5 3
-6
-4
-2
0
2
4
6
t(s)
p
3
2
c
(
w
)
Fig. 8. Power p
C
32
of the third actuator.
V. CONCLUSIONS
In the paper, the matrix relations for the real time computation
of position, velocity and acceleration of each link of a 3-DOF
parallel manipulator have been established. The analytical
calculus involved in the Lagrange equation is very tedious,
thus presenting an elevated risk for errors. Furthermore, the
duration of numerical computations is getting longer when
the number of components of the mechanism is increased. In
this case, the new method showed its advantage over others.
With the example of the 3-DOF parallel mechanism, the new
method illustrated an efcient way to determine the real time
variation of the forces and powers of all the actuators of
a parallel manipulator. In a context of automatic control,
the iterative matrix relations (20) and (21) given in this
dynamic model can be easily transformed into a robust model
for the computerised control of the most general parallel
manipulators.
References
1. J. P. Merlet, Parallel Robots (Kluwer Academic Publishers,
2000).
2. D. Steward, A platform with six degrees of freedom,
Proceedings of the Institution of Mechanical Engineers 180,
371378 (1965).
3. J. A. Carretero, R. P. Podhorodeski, M. N. Nahon and C. M.
Gosselin, Kinematic analysis and optimization of a new
three degree-of-freedom spatial parallel manipulator, ASME
Journal of Mechanical Design 122, 1724 (2000).
4. G. R. Dunlop and T. P Jones, Position analysis of a two DOF
parallel mechanism Canterbury tracker, Mechanism and
Machine Theory 34, 599614 (1999).
5. K. M. Lee and S. Arjunan, A three-degrees-of-freedom
micromotion in-parallel actuated manipulator, IEEE Trans-
actions on Robotics and Automations 7, No. 5, 634641 (1991).
6. D. Fedewa, M. G. Mehrabi, S. Kota, N. Orlandea and
V. Gopalakrishran, Parallel structures and their applications in
recongurable machining systems, Proceedings of the 2000
Parallel Kinematic Machines. International Conference, Sept.
1315, 2000, Ann-Arbor, Michigan (2000) pp. 8797.
7. T. Huang, Z. Li, M. Li, D. Chetwynd and C. M. Gosselin,
Conceptual Design and Dimensional Synthesis of a Novel
2-DOF Translational Parallel Robot for Pick-and-Place
Operations, ASME Journal of Mechanical Design 126, 449
455 (2004).
8. D. Zhang, Kinetostatic Analysis and Optimization of Parallel
and Hybrid Architectures for Machine Tools, Ph.D. Thesis,
(Laval University, Canada, 2000).
9. L. W. Tsai, Robot Analysis: The Mechanics of Serial and
Parallel Manipulators (John Wiley & Sons, Inc., New York,
1999).
10. B. Dasgupta and T. S. Mruthyunjaya, A Newton-Euler
formulation for the inverse dynamics of the Stewart platform
manipulator, Mechanismand Machine Theory 33, 11351152
(1998).
11. Y. W. Li, J. Wang, L. P. Wang and X. J. Liu, Inverse dynamics
and simulation of a 3-DOF spatial parallel manipulator, Pro-
ceedings of the IEEE International Conference on Robotics &
Automation, Taipei, Taiwan (2003) pp. 40924097.
12. S. Staicu, Mod` ele dynamique en robotique, Scientic
Bulletin, Series D (Mechanical Engineering, University
Politehnicaof Bucharest) 61, 34, 519 (1999).
13. S. Staicu, M ethodes matricielles en dynamique des
m ecanismes, Scientic Bulletin, Series D (Mechanical
Engineering, University Politehnica of Bucharest) 62, 3, 3
12 (2000).
130 Recursive matrix
14. Z. Ji, Dynamics decomposition for Stewart platforms, ASME
Journal of Mechanical Design 116, No. 1, 6769 (1994).
15. B. Dasgupta and P. Choudhury, A general strategy based on
the Newton-Euler approach for the dynamic formulation of
parallel manipulators, Mechanism and Machine Theory 34,
No. 6, 801824 (1999).
16. Z. Gengand and L. S. Haynes, On the dynamic model and
kinematic analysis of a class of Stewart platforms, Robotics
and Autonomous Systems 9, 237254 (1992).
17. L. C. T. Wang and C. C. Chen, On the dynamic analysis of a
general parallel robotic manipulators, International Journal
of Robotics and Automation 9, No. 2, 8187 (1994).
18. K. M. Lee and D. K. Shah, Dynamic analysis of a
three-degree-freedom in-parallel actuated manipulator, IEEE
Journal of Robotics and Automation 4, No. 3, 361367 (1988).
19. L. W. Tsai, Solving the inverse dynamics of a Stewart-Gough
manipulator by the principle of virtual work, ASME Journal
of Mechanical Design 122, No. 1, 39 (2000).
20. H. Pang and M. Shahinpoor, Inverse dynamics of a parallel
manipulator, Journal of Robotic Systems 11, No. 8, 693702
(1994).
21. S. Bhattacharya, D. N. Nenchv and M. Uchiyama, Arecursive
formula for the inverse of the inertia matrix of a parallel
manipulator, Mechanismand Machine Theory 33, No. 7, 957
964 (1998).
22. K. E. Zanganeh, R. Sinatra and J. Angeles, Kinematics
and dynamics of a six-degree-of-freedom parallel mani-
pulator with revolute legs, Robotica 15, part. 4, 385394
(1997).
23. T. R. Kane and D. A. Levinson, Dynamics: Theory and
Applications (McGraw-Hill, NewYork, 1985).
24. M. Sorli, C. Ferraresi, M. Kolarski. B. Borovac and
M. Vukobratovic, Mechanics of Turin parallel robot,
Mechanism and Machine Theory 32(1), 5177 (1997).
25. S. Staicu, Mecanica eoretica (Edit. Didactica & Pedagogica,
Bucharest 1998).