Dynamics Modeling of a Delta-type Parallel Robot

(ISR 2013)

2 3
Shi Baek Park I ,Han Sung Kim , Changyong Song, Kyunghwan Kim
[
Department of Mechanical Engineering, Yonsei University, Republic of Korea, amuro1988@yonsei.ac.kr
2
School of Mechanical Engineering of Kyungnam University, Republic of Korea, hkim@kyungnam.ac.kr
3
NT Research, Republic of Korea, kimk@ntresearch.net
(Tel: 82-55-249-2627, Fax: 82-55-249-2617)

Abstract-- This paper presents a simplified dynamics

modeling and hardware implementation of a Delta-type II. KINEMATIC ANALYSIS
parallel manipulator. Due to complex kinematics, the
Lagrangian equations of the first type is employed to derive
As shown in Fig. 1, the Delta parallel robot consists of
the inverse dynamic equations. Commercial Delta parallel three R-Pa (Revolute-spatial Parallelogram) legs
robot can achieve more than 10m/s and 15g. Accuracy and connecting in parallel from the base platform to the
fast calculation of dynamics are very essential typically in moving platform, which allows three translational
computed torque control of a Delta-type parallel motions to the moving platform. Since two S-S
manipulator for high speed applications. It is presented that (Spherical-Spherical) chains consisting of each Pa chain
the simplified dynamics equation matches very well with are subject to only axial load, Delta parallel robot has
ADAMS modeling and the calculation time of the inverse
small moving inertia.
kinematics and dynamics is less than O.04msec.
Figure 2 shows the vector-loop diagram with the
kinematic parameters; a and b denote the radii of the
Index Terms- Delta parallel robot, Dynamics modeling,
Lagrangian equation, computed torque control, ADAMS
fixed and moving platforms, and 1\ and 12 are the
actuating and connecting link lengths. The actuated joint
I. INTRODUCTION angles are denoted by eli and passive joint angles are
denoted by e2i and 83i. The vector-loop equation of each
In many sectors such as electronics, packing, food, leg is given by
pharmacy, and other light industries, Cartesian-type or
SCARA-type serial-kinematic manipulators with 3-DOF AiMi + MiB, = OP + PB, - OA, (1)
or 4-DOF are mainly used. However, serial-kinematic
For simplicity of expression, the local frame for the ith leg
structures suffer from large moving inertia and small ratio
of payload to weight. In order to overcome the above ( A, - XiYiZi) is introduced and the leading superscript
shortcomings, parallel manipulators have been
denotes the frame in which a vector is expressed. The
vector-loop equation is expressed in the ith local frame by
investigated. Since heavy actuators locate near or at the
fixed base and payload is distributed to several serial
chains, parallel manipulators can generate high speed, i i
I, U'i + 12 u2i=' (p + b - a) (2)
high stiffness and high accuracy.
Recently, many researchers have focused on the Delta­
where
type TPMs for high-speed applications in place of
SCARA-type serial manipulators. Delta parallel robot
[I�3] has very small moving inertia and the fastest
commercial Delta robot can achieve more than lOmls and
ISg.
Accuracy and fast calculation of dynamics are very
essential typically in computed torque control of a Delta­
type parallel manipulator for high speed applications. In
this paper, a simplified dynamic modeling of a Delta-type
parallel manipulator is presented. First, the position,
where (A = +lZ" /3, ¢, = lZ", (A = -lZ" /3 are defined as shown
velocity, and acceleration analyses are performed. Then,
the dynamics modeling is derived by using the in Fig. 3. Solving Eq. (2) for given p, two passive joint
Lagrangian equation of the first type due to the complex angles (eli' eli) and one active joint angle (eli) of the ith
kinematics. Finally, the derived simplified inverse leg are obtained by
dynamic equations are compared with full ADAMS
modeling and the calculation time is measured in the real­
time controller. (4)
 Fig. 2. Vector-loop diagram of the ih leg.

B,
Fig. I. Configuration of a Delta parallel robot

2
where K=(b l� +b �i +bJ'i _/, -/� ) /( 2IJ2s0JJ, bli =Pl '
b" =C(P,PI +Sr,p P2 + (b- a) , b2,=-Sr,p PI +Cr,p P2'

(5) x,

where gl i=/1 +/2c()2iS()3i g2i=/2s()2is()", (a) Fixed frame and i'h local frame (b) Moving frame
Fig. 3. Frame definitions.
The linear velocity of the moving platform is obtained
by taking derivatives of Eq. (2) with respect to time.
Using the principle of virtual works, the statics relation
(6) between force at the moving platform and actuator force
can be expressed as
i
whereiffili=Oli [ O,-l,O]', ffi2i =(Oli +02i) [0,-1,0]',
i F =JT T (10)
0" [s(O" + O,,), O, -c(O,, + 02JY . Tn order to eliminate
(03, =

passive joint rates, ffi2i and ffi3" taking dot-product at where F =L/; , /2' J; Y is the applied force vector at the
the both sides of Eq. (6) yields, moving platform and T =[ 7,, 72, 7J Y is the actuator
torque.
(7) The acceleration relation can be derived by taking
derivatives of Eq. (8) with respect to time.
Writing Eq. (7) for i=1,2,3 gives the velocity relation by
(11)
' '
Jx p =Jq 6 or 6=J p (8)
The derivation of the time derivatives of J, and Jq
where 6=[0,,02,OJ1 [OIP012,OllY for simple expression,
== is omitted.
l
J=J; Jx ' Jq =II diag(s()21s()ws()22s()12,s(),3s()3J,
m. DYNAMICS ANALYSIS
ilX [ i, ; iI, j },x=c (()li +(),,)s()3iCrp, -C()3iSrp,
The dynamic analysis can be performed by using only
Jx = x
' � i,y � " �:
, iV C(()li +()2JS()JiSrp, +C()Jjcrp, . three generalized coordinates since it is a 3-DOF
}3 x i3 ; }3 z },,- S(()li +()2i)S()3i manipulator. However, due to complex kinematics, three
redundant coordinates (Px' P" P') are included. Thus
Once 0, == 01, is solved, the two passive joint rates are the generalized coordinates become
obtained from Eq. (6),
(12)
OJ, =-' p, 1(l2s0li) ,
To simplifY the analysis, it is assumed that the mass of
02i =(-'P1C()"-'pJs ()li +12c()2ic()JAJ1 (l2s()2is()Jj) - 0i (9) each connecting rod, mz, in a parallelogram is distributed
evenly and concentrated at the two endpoints Mi and Bi. The Lagrangian function can be reduced to
Also, the acceleration of gravity points in the +Z
direction since the robot is installed in the downward. L=K - U
Using the Lagrange equations of the first type, the
closed-form dynamics equations can be derived by [4] (1 9)
1
+ ( m,!" + m,l,)g, L sB, + (mp + 3m2)g, p,
(1 3) i =l

Using Eq. ( 1 4) , the Lagrangian multipliers are

where n=6 is the number of generalized coordinates, determined by
k= 3 is the number of constraint functions, n - k= 3 is
the number of actuated joint variables, r, denotes the ith 2±Ai(p, +bC¢i- aC¢i-/,c¢,cB,)=(mp +3mJp, -.f;
i=!
constraint function, and Ai is the Lagragian multiplier.
2±Ai(P2 +bS¢i- as¢i-/,S¢iCB,)=(mp +3mJp2 -12 (2 0)
The first set of equations related to constraints can be j=1
written in the form
2±A, (P1 -/,sB,)=(mp +3m,KP1 -gJ - J;
j=1

l-�-Q'
�� or, � oL [ ) c
.or j=1,2,3 (1 4)
L.,, /l, =
,�, 'oq) dt oq) oq) Once the Lagrangian multipliers are obtained, the joint
torques are calculated by

where Q) denotes the generalized force contributed by 1'. =(y

2
Im +I. +m2/.')0. - ( m/Ie + m2/.)gccB.
an externally applied force. Once the Lagrangian - 2 /,A, [(p,c¢, +P2S¢, +b - a)sB, - p,cB,]
multipliers are found from Eq. ( 14), the second set of 2
1'2=('/ 1m +I. + m2/. )02 - ( m/Ie + m2/.)gccB2
equations related to actuation forces can be written as (2 1)
- 2 /.A2 [(P.C¢2 +P2S¢2 +b- a)sB2 - P3cB2]

Q.)
=
[ l
� �
dt oq)
-
oL - �� . or,
L." /.
oq) ,�. ' oq)
c
lor j =
4 ,5, 6 (15)
2
1', =('/ 1m +I, + m,l, )0, - ( m,!" + m,lJg,cB,

- 2/.AJ(P.C¢3 +P2S¢3 +b - a)sB3 - P3cB,]

where Q) is the actuator torque. When the trajectories of the moving platform (p,p,p) are
The constraint functions are given by given, the procedure to calculate actuator torques (T)

--2
can be summarized as follows:
r, =MiB, -I:
(a) B3"B2,,8.,: inverse kinematics (Eqs. (4) and (5))
=(p, +bC¢i- ac¢i-/,c¢,cB,) '
( 16)
(b) B"B"A,: velocity analysis (Eqs. (8) and (9))
+(P2 +bS¢i- as¢i-/,s¢,cB,) '
+(p, -/,sB,) ' -I: fori=1,2,3 (c) B,: acceleration analysis (Eq. (II))
(d) Lagrangian multiplier (A,): Eq. ( 2 0)
The total kinetic energies can be obtained by (e) Actuator torque ( 1',): inverse dynamics (Eq. ( 2 1))
,
K=Kp +L (
i=l KIi +K2i)
(1 7)

[v. DYNAMICS SIMULATION AND [MPLEMENTATION

I 2 1 ,
where Kp = mp �.p, , K"= ( yIm +1, )ili , The kinematic parameters of the prototype Delta
'2 {;;( '2
parallel robot are given by Table I and from CAD data,
I �. 2 I 2 '2
K,., =-m,L."p' + -m,l, B' where y and 1m denote the mass properties are calculated by Table II. Line
- 2 - ,�, 2 2
trajectories with 7 m/sec and 100 m/sec along the x- and
gear ratio and its moment of inertia. y-axes are generated as shown in Fig. 4.
And total potential energies can be obtained by Figure 5 shows the joint torques for the trajectory and
, payload 0.5kg. [t is noted that the simplified dynamics
U=Up +L (UI i +U2,) ( 18) equation matches very well with ADAMS modeling. The
i=l small difference comes from the assumption on link 2 for
simple dynamics equation and thus fast calculation.
where Up =-mpg,p" UI; =-m,g)" sBi, and
U2i=-m2K (P3 +I"sB,) .
 TABLE I dynamics compensation takes about 0.04msec on Intel
KINEMATIC PARAMETERS OF A DELTA PARALLEL MANIPULATOR
CPU (2.16 GHz dual core).
Parameters I Values
Radius of the fixed base (a) 100 mm
Radius of the moving platform (b) 22.5 mm
Length of link I (h) 171 mm
Length of Iink2 (/2) 396 mm

TABLE II
MASS PROPERTIES OF A DELTA PARALLEL MANIPULATOR
Parameters I Values
Mass of link I (ml) 426 g
Half of mass of link2 (m2) 69 g
Mass of moving platform (m,)+payload 96 g + 500 g
Length of link1 's mass center (lie) 6l.84 mm
Link I 's mass moment of inertia (II) 39.8 x 10-4 kg·m"
Motor inertia with gear ratio (y'Im) 25" x 0.24 x 10-4 kg'm'

0.5�-�-�-�--,-----,--,-
I I [Parallel Robot] [AC Motor Drivers and Interface]
- -' L ___I _
Fig. 6. Configuration of control hardware with XPC Target
0.
- 50 0.02 0.04 0.00 0.08 0.1 0.12 0.14 0.16
10

__ L_ 1 ___ l ___1 ___

�.. '-:1""-------- �.

0
0 0.02 0.04 0.00 0.08 0.1 0.12 0.14 0.16
100
I
L ___1 ___ l ___1 ___
I I
_

100
- L
0 --0.-i.OCC2--=-0.�04C---;Oc':.06=- --'---:0.;;;:- 08:----;O;;-'.I:-- ---;0.�;- ' -;2--;;0�.14:-; ---;O.16;- ..
time[sec]
S.mp'w I;mw, '", 'Owll'LIC , ·

ro,,..- An"", ••O" 1l."crlp1lon �.h,.

_ 01 0'"'00 1 n=

Fig. 4. Line trajectory along the x- and y-axes

_
_
01
",
..
0 . 0, 2
[I. .""",, '
aWl
nO!

50
[Line move along the x-axis] [Line move along the y-axis1
50,--,--,--,
':�
'::-- H H,brid

=
fJaI """ ; '_

Fig. 7. Control program with Simulink and XPC Target

E E
� 0 - - + --I - - - +- - - � 0 - - -+ ----
1 - - +- - -
2 2
� �

0.05 ,--.-----.---,--,---,-,
-50 -50 I I I I I
a 0.05 0.1 0.15 0.2 a 0.05 0.1 0.15 0.2 -- 1- - - +- - - -+ - - ---1 ---1----
0.045 1 ---
1 - -J- -- 1- --
50 20 I I I I I I
I I I
0.04
E I E
z - - 1- ---I - - - t- - - Z --t --- I -- t- --
a
� � 0.035
2 2

-50 -20 0.03

a 0.05 0.1 0.15 0.2 a 0.05 0.1 0.15 0.2 �
_

I I I I I I
� 0.025 -- j- - - t- - - --t - - ---1 ---1---1---- 1 - - j- -- j- --
50 50 r- I I I 1 1 I
w
E E I- 0.02 � - - _ - __1 ______ 1 L __
I I 1 1
1 I
- - 1- --I - - t- - - - - --t ---- --
_

� 0 � 0 1 - - t-
I I I I I I 1
0.015 -- r - - - I -- ---, ---1---1----
- - - r -- r --
II II T 1
I I I I I I 1 1 I
-50 -50 0.01 - - L __ l __ � � -: - -:- --
-L: --�
0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2 __ __ _ _

I
_ __

time[sec] time[sec] I
r
I I I I I 1
--
1
- - r --
0.005 - - - - T - - I -- ---, ---1---1---- 1 I
Fig. 5. Inverse dynamics simulation results I I I I I I

(solid: dynamics equation, dotted: ADAMS modeling) °OL---L---4�--�---;�� 1�0--� 12�� 14�� 16��18��20
time [msec]
As shown in Fig. 6, the control hardware consists of Fig. 8. TET of the Simulink control program
host PC, real-time target PC with DAQs and AC servo
drivers. The host PC running on Windows OS plays a
role in Simulink programming, user interface, etc. The V. CONCLUSIONS
target PC running on xPC target provides the real-time
In this paper, a simplified dynamics modeling and
control of AC servo motors. Figure 7 shows the control
hardware implementation of a Delta-type parallel
program made by Simulink including PID) �int controll:r
manipulator are presented. For real-time computed torque
and dynamics compensation. The PID Jomt control IS
control, the derivation of the closed form solution of
performed in every 0.2msec. The inverse kinematics and
inverse dynamics is essential. The position, velocity, and
dynamics compensation are calculated in every 1.0 mse�.
. . acceleration analyses are performed. Due to complex
Figure 8 shows the TET (Task ExecutIOn Time). It IS
kinematics, the inverse dynamics is derived by using
shown that the PID control takes about 0.03msec and the
Lagrangian equation of the first type. The numerical
simulations with ADAMS are performed to verify the
accuracy of the analytical dynamic modeling. It is shown
that numerical and analytical results match very well. The
real-time controller is developed with XPC Target and it
is shown that the calculation time of the inverse
kinematics and dynamics is less than 0.04msec on Intel
CPU (2. 16 GHz dual core).

ACKNOWLEDGMENT
This work is supported by the project (No. 1004 1965)
entitled as Development of a parallel link robot with 6kg
payload and 0.43 sec cycling time funded by the Korean
Ministry of Trade, Industry and Energy.

REFERENCES
[I] R. Clavel, "Delta, a Fast Robot with Parallel Geometry,"
18th Int. Sym. on Industrial Robots, Sydney, Australia, 91-
100. 1988.
[2] L. W. Tsai. "Kinematics of a Three-DOF Platform
Manipulator with Three Extensible Limbs," Advances in
Robot Kinematics, 401-410, 1996.
[3] X. J. Liu, "Optimal Kinematic Design of a Three
Translational DoFs Parallel Manipulator," Robotica, Vol.
24, No. 2, 239-250, 2006.
[4] L. W. Tsai, Robot Analysis: The Mechanics of Serial and
Parallel Manipulators, New York, Wiley, 1999.