See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/341906680

Analytical dynamic modeling of Delta robot with experimental verification

Article in Proceedings of the Institution of Mechanical Engineers Part K Journal of Multi-body Dynamics · June 2020
DOI: 10.1177/1464419320929160

CITATIONS READS
0 65

2 authors:

Farshid Asadi Ali Heydari

5 PUBLICATIONS 14 CITATIONS
University of Sistan and Baluchestan
22 PUBLICATIONS 209 CITATIONS
SEE PROFILE
SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Developing a new mathematical model for designing a sustainable supply chain network View project

All content following this page was uploaded by Farshid Asadi on 11 June 2020.

The user has requested enhancement of the downloaded file.

 Analytical dynamic modeling of Delta robot
with experimental verification
Farshid Asadi and Ali Heydari

The content of the article is as the same as the final published

paper.
To refer to the original paper please cite as following,
For word:
Asadi F, Heydari A. Analytical dynamic modeling of Delta robot with experimental verification.
Proceedings of the Institution of Mechanical Engineers, Part K: Journal of Multi-body Dynamics. 2020
Jun 3:1464419320929160

For latex:
@article{asadi2020analytical,
title={Analytical dynamic modeling of Delta robot with experimental
verification},
author={Asadi, Farshid and Heydari, Ali},
journal={Proceedings of the Institution of Mechanical Engineers, Part K:
Journal of Multi-body Dynamics},
pages={1464419320929160},
year={2020},
publisher={SAGE Publications Sage UK: London, England}
}
Analytical dynamic modeling of Delta
robot with experimental verification

Farshid Asadi and Ali Heydari

Abstract
In this paper, an explicit dynamic model of Delta robot is obtained analytically. The main contribution of this work is that,
unlike existing prior work, the final dynamics model is given directly in the form of MX€ þ CX_ þ G ¼ JT , with explicit
expressions for M, C and G. This is of great importance, since many advanced control techniques like Optimal Control need
dynamic model in an explicit form, i.e. time derivative of state vector given explicitly in terms of the states and control
vectors. To this goal, first, velocity and acceleration analysis is done by differentiating robot’s geometrical loops directly.
Then, Jacobian matrices are calculated to have kinematic relations in a more compact form. After that, principle of virtual
work is implemented to derive the dynamic equations. In this part, Jacobian matrices are substituted into dynamic model.
This is unlike other referenced works on Delta robot dynamics that need to continue the derivation in symbolic software
or derive the model implicitly. Using Jacobians, dramatically simplifies the final explicit dynamic model. Therefore, the
final dynamic equations are calculated in a straightforward manner without any use of symbolic calculation
software. After all, the presented model is verified with an experimental setup. The model shows good accuracy in
terms of torque prediction.

Keywords
Delta robot, explicit dynamic model, analytical model, experimental

Introduction
Delta robot is a parallel manipulator that is mainly software. Based on these two studies, there is around
designed as a pick and place robot.1 However, 10% inaccuracy in predicting torques for the simpli-
because of its high force to weight ratio and simple fied model in software simulations; however, it is 38%
design, it can also be used for other purposes, e.g. more efficient in terms of computational cost.
machining, haptic interface, surgery.2,3 In model- To the best of our knowledge with the existing
based control techniques, one needs a dynamic models, one needs using symbolic calculation software
model of the system, even though there are similarities to calculate the dynamic model explicitly. While this is
in dynamic modeling procedures of parallel robots, possible, the resultant equations are too long and
due to their different structure, each case should be hard to track. Having an explicit model is a require-
resolved separately. Thus, several methods including ment of some modern control techniques. This moti-
Newton–Euler,4 Lagrange equation,5–11 Hamilton vated us to develop an approach to derive explicit
equation,12 principle of virtual work10,13,14 and recur- dynamic equations of Delta robot analytically, while
sive methods15,16 are used to derive dynamics equa- keeping it neat and compact without using computer
tions of this manipulator. In all the mentioned works, symbolic calculation. The calculated model can be
friction in spherical joints are neglected, which as dis- used directly for all control methods which require
cussed in Clavel,17 is reasonable because of using low the model of the dynamics to be in the form of
friction spherical joints. Also in some of the men-
tioned works, intermediate legs are modeled as point
masses distributed over two ends of the intermediary
legs.4,6,7,9,12 Effects of this simplification are generally Department of Mechanical Engineering, Southern Methodist University,
Dallas, USA
case dependent, but it is studied in Brinker et al.18,19
These two works compared accuracy and computa- Corresponding author:
tional cost of different dynamic modeling approaches. Farshid Asadi, Southern Methodist University, Dallas, TX 75205, USA.
Both of the studies are based on multi-body dynamics Email: [email protected]
2

x_ ¼ f ðx, uÞ, where x and u represent the states and

control vectors.
In what follows, first, velocity and acceleration ana-
lysis is done based on Codourey.4,10 Then, Jacobian
matrices of Tsai20 are utilised to have velocity and
accelerations in a more compact form. After that, vir-
tual work principle is employed to derive the dynamic
equations. Next, Jacobian matrices are substituted in
dynamic model to simplify it in an organised and trace-
able way. Furthermore, experimental verification of the
model is done and presented in the last part.

Kinematics
3-Revolute, spherical, spherical (3-RSS) Delta robot
is constructed by three serial chains connecting the
fixed platform to the moving platform as depicted in
Figure 1. To describe kinematics of the robot, two
frames are assigned for each serial chain. First, a
frame defined by a rotation about Z-axis of the
world frame (that is perpendicular to fixed platform
plane), whose rotation matrix is called Ri. Second, a
frame is assigned to motorised leg as a consecutive Figure 1. Delta robot structure.
body fixed rotation about the x-axis with rotation
matrix Oi (see Figures 1 and 2). The mentioned rota-
tion matrices can be written as
2 3
cosð’i Þ sinð’i Þ 0
6 7
Ri ¼ 4 sinð’i Þ cosð’i Þ 05 ð1Þ
0 0 1
2 3
1 0 0
6 7
Oi ¼ 4 0 cosðqi Þ sinðqi Þ 5 ð2Þ
0 sinðqi Þ cosðqi Þ

One can write the following equations

8
<XþR ¼ R þB ! !
ai bi i Ci þ Ci Ai
! T ! ð3Þ
:
Ci Ai Ci Ai ¼ l2K

where all vectors are stated in world frame.

! !
X, Rbi , Rai , Bi Ci , Ci Ai , and lK are moving platform
centre of mass position with respect to world frame Figure 2. Delta robot frame assignments for ith leg chain.
origin, fixed platform joint position with respect to
world frame origin, moving platform joint position
with respect to its centre of mass, motorised leg !
Bi Ci in the xi yi zi frame. So, one can write
vector, intermediary leg vector, and intermediary leg !
length. The first equation of (3) is called Loop Closure. Bi Ci ¼ Ri Oi L, where L is motorised leg vector stated
The second equation of (3) denotes the constant in its own xi yi zi frame. Using these in equation (3) and
length of the intermediary legs. after rearrangement, one can write
Note that vectors Rbi and Rai are constant vectors 
in their corresponding x0 i y0 i z0 i frame. Therefore, one Ui ¼ X  Ri ððRb  Ra Þ þ Oi LÞ
ð4Þ
can write Rbi ¼ Ri Rb and Rai ¼ Ri Ra , where Rb UTi Ui ¼ l2K
and Ra are fixed platform joint position and moving
!
platform joint position stated in x0 i y0 i z0 i frames. where Ci Ai is renamed to Ui for ease of notation. This
Also the same is true about motorised leg vector equation should be solved to obtain forward or
Asadi and Heydari 3

inverse kinematic of the manipulator, that is, to cal- By differentiating each row of equation (8) and
culate X based on q (where q ¼ ½q1 , q2 , q3 T ) and vice rearranging it one has
versa. While equation (4) is used for obtaining the
dynamical model in the next section, calculating the U_ Ti X_ þ UTi X€ ¼ ðU_ Ti Ti þ UTi T_ i Þq_ i þ UTi TTi q€i ð10Þ
mentioned kinematics is not the objective in this work
and therefore is skipped. See Williams21 and Hsu where T_ i ¼ Ri S1 S1 Oi Lq_ i . Equation (10) can be rear-
et al.22 for detailed solution and further analysis of ranged and written in matrix form as
the kinematics of the robot.
q€ ¼ JX€ þ J_X_ ð11Þ
Velocity, Acceleration and Jacobian where J_ in the above equation can be written as
In order to obtain a dynamic model of the manipula-
2 31
tor, velocity and acceleration analysis should be done. UT1 T1 0 0
Moreover, auxiliary Jacobians are helpful to make the _ q, X,
_ qÞ 6
_ ¼4 0 T 7
JðX, U2 T2 0 5
final equations simpler. First, the main Jacobians of
0 0 UT3 T3
the system should be calculated. This is done here, by 02 3 2 31
implementing the approach of Codourey.4,10 U_ T1 ðU_ T1 T1 þ UT1 T_ 1 ÞJ1
B6 7 6 7C
Differentiating the second equation of (4) results into  @ 4 U_ T2 5  4 ðU_ T2 T2 þ UT2 T_ 2 ÞJ2 5 A
: U_ T3 ðU_ T3 T3 þ UT3 T_ 3 ÞJ3
UTi Ui ¼ 0 ð5Þ ð12Þ
:
Calculation of Ui is straightforward through chain where Ji denotes ith row of J.
rule. The terms Ri , Rb , Ra , and L are constant and Now that the main Jacobians are calculated, leg
have zero derivatives. Therefore,
: one only needs to velocities and accelerations are easy to compute.
calculate the term Ri Oi L. Since Oi is a rotation The process done in Tsai20 is adopted here for this
matrix for an elementary rotation around x0i axis, its purpose.
derivative can be calculated as Ginsberg23 Angular velocity of motorised leg can be written in
its own frame, that is xi yi zi , as
:
Oi ¼ S1 Oi q_i ð6Þ 2 3
Ji
where S1 is the skew symmetric matrix related to the O
!Li ¼ 4 013 5X_ ð13Þ
derivative of rotation around a general x-axis, which 013
is defined as
2 3 For linear velocity of the motor leg COM one has
0 0 0 2 3
6 7 013
S1 ¼ 4 0 0 1 5 O
vLi ¼O !Li  ðrL j^ Þ ¼ rL lL 4 013 5X_ ð14Þ
0 1 0
Ji

Using above notes, one can write where j^ is defined as [0, 1, 0]T, lL is motor leg length
and rL is the ratio of motor leg length that shows the
: :
Ui ¼ X_  Ti qi ð7Þ distance between motor leg COM and corresponding
fixed platform joint, that is llL . For angular acceler-
where Ti ¼ Ri S1 Oi L. By substituting equations ation of motor leg COM one can write
(4) and (7) in equation (5), the result can be
combined as 2 3 2 3
Ji J_i
2 3 2 3 O
!_ Li ¼ 4 013 5X€ þ 4 013 5X_ ð15Þ
UT1 UT1 T1 0 0 013
6 T7 _ 6 7 013
4 U2 5X ¼ 4 0 UT2 T2 0 5q_ ð8Þ
UT3 0 0 UT3 T3
Also, linear acceleration of motor leg COM can be
calculated as
The above equation can be rewritten as q_ ¼ JX, _
where J is the Jacobian of the system and is defined as O
v_Li ¼ O !Li  ðO !Li  rL lL j^ ÞþO !_ Li  rL lL j^
2 3 2 3
2 31 2 3 013 013
UT1 T1 0 0 UT1 6 7 6 T 7 ð16Þ
6 7 6 7 ¼ rL lL 4 013 5X€ þ rL lL 4 X_ JTi Ji 5X_
JðX, qÞ ¼ 4 0 UT2 T2 0 5 4 UT2 5 ð9Þ
Ji J_i
0 0 UT3 T3 UT3
4

Note that replacing rL with one in equation (14) moment of inertias is roughly equal. Therefore, lon-
and equation (16) gives linear velocity and acceler- gitudinal moment of inertia is neglected here and then
ation of point Ci in the corresponding motor leg the intermediary leg is replaced with three masses,
frame. located at points C, A and intermediary leg COM,
Having velocity and acceleration of moving plat- so that the result is dynamically equivalent to the
form and point Ci, motion properties of intermediary intermediary leg (considering that longitudinal
legs can be determined easily in motor leg frames. For moment of inertia is neglected). To guarantee
linear velocities and accelerations, one can see that the dynamic equivalence, mass distribution between
following equations hold. Note that every point of three points can be calculated through solving the fol-
moving platform moves with the same linear velocity lowing linear system of equation
and acceleration. Therefore, one can write 2 32 3 2 3
1 1 1 mC MK
O
vKi ¼ O vCi þ rK ðO vAi O vCi Þ 6 76 7 6 7
4 rK ðrK  1Þ 0 54 mA 5 ¼ 4 0 5 ð22Þ
0 2 3 1
013 r2K l2K ðrK  1Þ2 l2K 0 mK IK
B 6 7 C ð17Þ
¼ @ ð1  rK ÞlL 4 013 5 þ rK OTi RTi AX_
Ji where mC, mA and mK are masses distributed over
points A, C, and intermediary leg COM. MK and IK
O
v_Ki ¼ O v_Ci þ rk ðO v_Ai O v_Ci Þ are also intermediary leg mass and dominant moment
0 2 3 1 of inertia with respect to its COM. In equation (22), the
013
B 6 7 C first row shows mass equivalence, the second row rep-
¼ @ ð1  rK ÞlL 4 013 5 þ rK OTi RTi AX€ resents COM equivalence, and the third row denotes
Ji ð18Þ inertial equivalence. This process results in the same
2 3
013 equivalence that is derived in Codourey10; however, it
6 T 7 is in a form more suitable for applying virtual work
þ ð1  rK ÞlL 4 X_ JTi Ji 5X_
method. Note that this simplification is different than
J_i the mentioned simplification in literature.4,6,7,9,12 In the
referenced simplified models, mass of intermediary leg
where rK is the intermediary leg length parameter that is distributed over point C and A; thus, there is no such
denotes the distance between point Ci and intermedi- dynamic equivalence as discussed here.
ary leg COM, that is equal to lkK . To construct equation of motion of the mechanism
Finally, auxiliary Jacobians can be defined from by the principle of virtual work, one needs to first
the calculated velocities for our points of interest, calculate applied and inertial active wrenches (forces
including motor leg and intermediary leg COMs and and momentums). This basically needs to be calcu-
point Ci, as lated at motor, motor leg COM, point Ci, intermedi-
2 3 ary leg COM, point Ai and moving platform COM.
013 For the moving platform, there is no rotational
6 7 term. So, one has
JLi ¼ rL lL 4 013 5 ð19Þ
Ji
WP ¼ FP þ mP g  mP X€ ð23Þ
0 2 3 1
013
B 6 7 C where FP, mP and g are external force applied to
JKi ¼ @ ð1  rK ÞlL 4 013 5 þ rK OTi RTi A ð20Þ moving platform, moving platform mass and gravity
Ji vector, respectively. Motors are fixed and there is no
translational term; also note that motors only have
2 3
013 active wrenches along their own rotational axis direc-
6 7 tion. Therefore, one can combine motors active
JCi ¼ lL 4 013 5 ð21Þ
wrenches and write
Ji
Wmo ¼ Imo q€  fric þ  ð24Þ
Dynamics where Imo , fric and  are motor inertia, friction in
Before proceeding to the main dynamic model, a motor joint and motor torques, respectively. Active
useful modeling simplification can be done. Because translational wrenches of motor leg can be written
of the structure of intermediary legs in Delta robot, as the following
product of inertias of intermediary legs is negligible.
Also, moment of inertia around its longitudinal axis is
WLi ¼ mL OTi RTi g  mL O v_Li ð25Þ
much less in comparison with other two principal
moments of inertia. Furthermore, the dominant
Asadi and Heydari 5

where mL is motor leg mass. Motor legs have active where FP is the external force exerted on moving
rotational wrenches only along their x-axis (they platform.
rotate around a fixed axis); thus, rotational wrenches Principle of virtual work as used in Tsai20 can be
of motor legs can be collected in vector form and written for the mechanism as
written as
X
3
WL ¼ IL q€ ð26Þ ðxTCi WCi þ XWAi þ xTLi WLi þ xTKi WKi Þ
i¼1
where IL is moment of inertia of motor leg around þ qT ðWmo þ WL Þ þ XWP ¼ 0
x-axis of motor leg frame, passing through its
ð37Þ
COM. Finally, active wrenches of point Ci, Ai and
COM of intermediary leg can be computed as
where  terms denote virtual displacements. All virtual
WCi ¼ mC OTi RTi g  mC v_CiO
ð27Þ displacements can be stated in terms of moving plat-
form virtual displacement, i.e. X, as
WAi ¼ mA g  mA O v_Ai ð28Þ
8
> q ¼ JX
WKi ¼ mK OTi RTi g  mk v_KiO
ð29Þ >
>
< xLi ¼ JLi X
ð38Þ
By substituting acceleration terms (equations (11), > xCi ¼ JCi X
>
>
:
(16) and (18)) into equations (23) to (29), active xKi ¼ JKi X
wrenches can be rewritten in terms of generalised
coordinate X and its derivatives. The summary is as
follows Finally, by substituting acceleration terms (equa-
tions (11), (16) and (18)) and Jacobians (equations
WP ¼ mP X€ þ FP þ mP g ð30Þ (9) and (19) to (20)) into equation (37), factoring
out X and rearranging the equation, one can write
Wmo ¼ Imo JX€  Imo J_X_  fric þ  ð31Þ the dynamic equation of the mechanism in the form of
2 3
013 MðX, qÞX€ þ CðX, q, X,
_ qÞ
_ X_
6 7
WLi ¼mL rL lL 4 013 5X€ þ GðX, qÞ  FP ¼ JðX, qÞT ð  fric Þ
Ji ð39Þ
2 3 ð32Þ
013
6 T 7 MðX, qÞ ¼ ðImo þ IL ÞJT J þ ðmP þ 3mB ÞI3
mL rL lL 4 X_ JT Ji 5X_ þ mL OTi RTi g
i
X
3
J_i þ ðmK rK JTKi OTi RTi Þ
i¼1
0 2 31
WL ¼ IL JX€  IL J_X_ ð33Þ T T ! 013
X3
B m l J
C L Ci þ m r l J
L L L Li 6 7C
2 3 þ @ T
4 013 5A
013 i¼1 þ mK ð1  rK ÞlL JKi
Ji
6 7
WCi ¼mC lL 4 013 5X€
ð40Þ
Ji
2 3 ð34Þ _ qÞ
013 CðX, q, X, _ ¼ ðImo þ IL ÞJT J_
6 T 7 0 2 31
mC lL 4 X_ JTi Ji 5X_ þ mC OTi RTi g T T ! 013
X3
B m l J
C L Ci þ m r l J
L L L Li 6 T 7C
J_i þ @
T
4 X_ JTi Ji 5A
i¼1 þ mK ð1  rK ÞlL JKi
J_i
WAi ¼ mA X€ þ mA g ð35Þ ð41Þ
0 2 3 1
013 GðX, qÞ ¼ ðmP þ 3mA Þ g
B 6 7 C
WKi ¼mk @ ð1  rK ÞlL 4 013 5 þ rK OTi RTi AX€ X
3  
Ji  ðmC JTCi þ mL JTLi þ mK JTKi ÞOTi RTi g
2 3 i¼1
013
6 7 ð42Þ
mK ð1  rK ÞlL 4 X_ T JT Ji 5X_ þ mK OT RT g
i i i i
J_i where matrices M, C and G are summarised in equa-
tions (40) to (42). I3 in these equations denotes to
ð36Þ
6

3  3 identity matrix. Also keep in mind that equation no verification or it has been done only on multi-body
(39) can be easily written in state-space form as dynamics software. In the present work, to verify the
" # accuracy of the dynamic model, torques required for
 
d
X X_ performing a typical trajectory are compared on both
¼   the proposed model and the experimental setup (see
dt
X_ M 1 _
CX þ G  FP
  ð43Þ Figure 4); this is motivated by Miller.12 Here, the
033 robot end effector is controlled (with a simple joint
þ ð  fric Þ:
M1 JT space PD control7) to move along a circle in z ¼ 450
(mm) plane, with x ¼ 250cos(t) (mm) and y ¼ 250sin(t)
The procedure for calculating model parameters at (mm). The real trajectory travelled by robot can be
every instance is summarised here: seen in Figure 3. For torque measurement, it is
assumed that the current–torque relation of the
1. Calculate X and q from kinematics. motors is linear. It is also assumed that the current
2. Calculate J(X, q) from equation (9). commanded to motors is provided instantly, which is
3. Calculate X_ and q_ from Jacobian J. a reasonable assumption comparing time constant of
4. Calculate J_ from equation (11). the current control loop (around 0.5 ms) to the time
5. Calculate JLi , JKi and JCi from equations (19) to constant of the robot control loop (around 120 ms).
(21). Furthermore, motor friction is identified through
6. Calculate matrices M, C and G from equations Stribeck model24 and subtracted from experimental
(40) to (42). results (Figure 4 (solid line)). Predicted torques of the
7. Use matrices M, C, G and J in equation (39). model (Figure 4 (dashed line with star marks)) are also
calculated by giving the real trajectory travelled by
robot (figure 3) to the proposed dynamic model in
equation (39). As it can be seen in Figure 4, the
Evaluation dynamic model prediction closely matches experimen-
Verification of the model is an important task to be tal results. Also note that there are discontinuous
done. This has been done experimentally in bumps in experimental torques when it reaches to its
Codourey,4,10 where the verification is done through extremum values. This happens when motors change
checking tracking error of closed loop control (which from acceleration to deceleration and vice versa, and is
cannot explicitly show the accuracy of the model itself), attributed to motor-gearbox backlash.
and in Miller12 through examining the predicted torque Other than accuracy of the model, its computa-
explicitly. In the rest of the cited prior works, there is tional cost is another important aspect. To analyse

Figure 3. Trajectory passed by end effector.

Asadi and Heydari 7

Figure 4. Torques.

this, average calculation time of the presented model model is verified experimentally and good accuracy in
is compared to the same property of the simplified experimental evaluation is observed.
implicit dynamics of the Delta robot based on
Lagrange equations, given in Brinker et al.18 On Declaration of Conflicting Interests
average, the proposed method takes 105% longer The author(s) declared no potential conflicts of interest with
to be calculated. This matches the comparison done respect to the research, authorship, and/or publication of
in Brinker et al.18 and shows that the current this article.
explicit derivation does not add much computational
cost to existing implicit virtual work-based models. Funding
It is also good to mention that the proposed The author(s) disclosed receipt of the following financial
model has minimum level of simplification in its support for the research, authorship, and/or publication
derivation. of this article: This research was partially supported by
Properties of the Delta robot used in experiments the United States National Science Foundation through
are given in Appendix. Grant 1745212.

Conclusion
In this paper, an explicit dynamic model of the Delta
robot is calculated analytically in a very compact form References
in comparison to the models in the referenced works. 1. Clavel R. Device for the movement and positioning of
Friction in spherical joints and longitudinal moments an element in space. Patent 4,976,582, USA, 1990.
of inertia of intermediary leg are neglected in the 2. Grange S, Conti F, Rouiller P et al. The delta haptic
device. Technical report, Ecole Polytechnique Fdrale de
model. The model is obtained without any need for
Lausanne.
computer-based symbolic calculation. This is done
3. Bonev I. Delta parallel Robot - the story of success.
using auxiliary Jacobians and implementing principle Newsletter, www.parallelmic.org (2001, accessed April
of virtual work. Using the auxiliary Jacobians simpli- 2019).
fied the final model effectively. The resulted model is 4. Codourey A. Dynamic modelling and mass
derived explicitly and this makes it suitable for apply- matrix evaluation of the delta parallel robot for axes
ing modern model-based control techniques. Also decoupling control. In Proceedings of IEEE/RSJ inter-
because of the well-organised structure of the model, national conference on intelligent robots and systems.
contribution of each part of the robot to the dynamics IROS’96, Osaka, Japan, vol. 3. pp.1211–1218.
can be understood easily. Furthermore, the proposed Piscataway: IEEE.
8

5. Miller K and Clavel R. The Lagrange-based model 21. Williams II RL. The delta parallel robot: kinematics
of Delta-4 robot dynamics. Robotersysteme 1992; 8: solutions. www.ohio.edu/people/williar4/html/pdf/
49–54. DeltaKin.pdf (2016, accessed April 2019).
6. Stamper RE. A three degree of freedom parallel manipu- 22. Hsu K, Karkoub M, Tsai MC, et al. Modelling and
lator with only translational degrees of freedom. index analysis of a delta-type mechanism. Proc
Dissertation, University of Maryland, USA, 1997. IMechE, Part K: J Multi-body Dynamics 2004; 218:
7. Tsai LW. The mechanics of serial and parallel manipu- 121–132.
lators. Robot Analysis 1999; 505. 23. Ginsberg J. Engineering dynamics, vol. 10. Cambridge:
8. Park SB, Kim HS, Song C et al. Dynamics modeling of Cambridge University Press, 2008.
a delta-type parallel robot (isr 2013). In: IEEE ISR 24. Olsson H, Åström KJ, De Wit CC, et al. Friction
2013, Seoul, South Korea, pp.1–5. Piscataway: IEEE. models and friction compensation. Eur J Control
9. Tho TP and Thinh NT. Analysis of kinematics and 1998; 4: 176–195.
dynamics of 4-dof delta parallel robot. In: Robot intel-
ligence technology and applications 2: results from the
2nd international conference on robot intelligence tech-
nology and applications. New York: Springer Appendix
International Publishing, 2014, pp.901–910.
Notation
10. Codourey A. Dynamic modeling of parallel robots for
computed-torque control implementation. Int J
Geometrical parameters
Robotics Res 1998; 17: 1325–1336. rb ¼ 0.2; m fixed platform radius
11. Kuo YL. Mathematical modeling and analysis of the ra ¼ 0.2; m moving platform radius
delta robot with flexible links. Comp Math Appl 2016; ’1 ¼  1st leg position angle
71: 1973–1989. ’2 ¼  3 2nd leg position angle
12. Miller K. Experimental verification of modeling of delta ’3 ¼ 3 3rd leg position angle
robot dynamics by direct application of Hamilton’s lL ¼ 0.2;m motor leg length
principle. In Proceedings of 1995 IEEE international rL ¼ 0.3933 motor leg COM length
conference on robotics and automation, Nagoya, Japan, ratio
vol. 1, pp.532–537. Piscataway: IEEE. lK ¼ .52; m intermediary leg length
13. Codourey A and Burdet E. A body-oriented method for
rK ¼ .5 intermediary leg COM
finding a linear form of the dynamic equation of fully
parallel robots. In: Proceedings of international confer-
length ratio
ence on robotics and automation, Albuquerque, NM, T
Rb ¼ 0 rb 0
USA, vol. 2, pp.1612–1618. Piscataway: IEEE.
T
14. Wang J and Gosselin CM. A new approach for the Ra ¼ 0 ra 0
dynamic analysis of parallel manipulators. Multibody
Syst Dyn 1998; 2: 317–334.
15. Staicu S. Recursive modelling in dynamics of delta par- Inertial parameters
allel robot. Robotica 2009; 27: 199–207.
dK ¼ 0:08; m intermediary leg distance
16. Carp-Ciocardia D et al. Dynamic analysis of clavel’s
delta parallel robot. In: 2003 IEEE international confer-
from each other at every
ence on robotics and automation (Cat. No. 03CH37422), joint
vol. 3, pp.4116–4121. Piscataway: IEEE. lK ¼ .5.74769459 intermediary leg pairs
17. Clavel R. Conception d’un robot parallèle rapide à 4  10–3; kg.m2 dominant inertia
degrés de liberté. Technical report, EPFL, 1991.
18. Brinker J, Corves B and Wahle M. A comparative study
of inverse dynamics based on Clavels Delta robot. In
2 3
Proceedings of the 14th world congress in mechanism and 2840617 0 0
machine science, Taipei, Taiwan. pp.25–30, Germany: IKmatrix ¼4 0 2840617 0 5  109 ; kg:m2
RWTH Publications. 0 0 1016
19. Brinker J, Ingenlath P and Corves B. A study on sim-
plified dynamic modeling approaches of Delta parallel single intermediary leg inertia matrix
robots. In: Jadran L and Jean PM (eds) Advances in IL ¼ 6:4345319  104 ; kg:m2 motor leg inertia
robot kinematics 2016. New York: Springer, 2018.
pp.119–128.
20. Tsai LW. Solving the inverse dynamics of a Stewart-
Imo ¼ 0:0465475; kg:m2 motor inertia
Gough manipulator by the principle of virtual work. mK ¼ 2  0:05788; kg intermediary leg mass
J Mech Des 2000; 122: 3–9. mL ¼ 0:116; kg motor leg mass
mP ¼ 1:055; kg moving platform mass

View publication stats