Research Article

International Journal of Advanced

Robotic Systems
September-October 2019: 1–9
Kinematic modeling and error ª The Author(s) 2019
DOI: 10.1177/1729881419878927
analysis of Delta robot considering journals.sagepub.com/home/arx

parallelism error

Yu Li1 , Deyong Shang1 and Yue Liu2

Abstract
Motion accuracy is an important indicator to evaluate the performance of robots. The driven arm of the Delta robot is
composed of a pair of parallelogram mechanisms. Because of the errors in components processing and installation, the
parallelogram-driven mechanism will tilt. In addition, the load will have greater inertia impact on the end-effector in high-
speed working condition. Therefore, the parallelogram-driven mechanism and loading effect will cause the parallelism
error of the moving platform and affect motion accuracy of the robot. Aiming at the error analysis and motion accuracy of
the Delta robot, currently, most studies mainly consider the influence of dimension error, clearance error, and driving
error. There are few research conducted on the kinematics and error analysis considering parallelism error of the
mechanism. The purpose of this article is to study the influence of parallelism error on motion accuracy of the robot.
Firstly, the space vector method in mathematics is used to calculate the forward solution of the robot, and the kinematic
equations of each kinematic chain are established. Secondly, the article analyzes the influence of the non-parallelism of the
driven mechanism on motion error of the robot. Taking the D3PM-1000 robot as an example, it is obtained from the
simulation results that with the increase in the tilt angle of the moving platform, the motion error caused by parallelism
error also increases and the variation value is obvious, especially in the z direction, meaning that the parallelism error has a
great influence on the motion error of the mechanism. The article effectively complements the research field of error
analysis and provides a theoretical reference for the error compensation of the robot.

Keywords
Delta robot, parallelism error, kinematic modeling, error analysis, motion accuracy

Date received: 6 May 2019; accepted: 3 September 2019

Topic: Robot Manipulation and Control

Topic Editor: Andrey V Savkin
Associate Editor: Michal Kelemen

Introduction
Delta parallel robots have the advantages of high moving 1
School of Mechanical Electronic and Information Engineering, China
speed, strong bearing capacity, and good dynamic perfor-
University of Mining and Technology (Beijing), Beijing, China
mance.1–4 They are widely used in assembly testing, packa- 2
Department of Mechanical Engineering, State Key Laboratory of
ging, sorting, precision positioning, and other fields.5,6 Tribology, Tsinghua University, Beijing, China
Motion accuracy is the key indicator to evaluate the per-
formance of robots. With the wide application of parallel Corresponding author:
Deyong Shang, School of Mechanical Electronic and Information
robots, people have higher and higher accuracy demand.7,8 Engineering, China University of Mining and Technology (Beijing), Beijing
Professor Sun Zhili9 and his team made a thorough study 100083, China.
on the error modeling and reliability analysis of the Delta Email: [email protected]

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 International Journal of Advanced Robotic Systems

parallel robot, mainly from three aspects: dimension error,

clearance error of the joints, and driving error. Ropponen Fixed platform
and Arai10 presented the relation between the joint displa-
cement errors of Stewart platform manipulators and the
Driving arm
end-effector accuracy. The position errors of the joints and
the actuation errors and backlash are included to the kine-
Driving motor
matic model. Chebbi et al.11 studied the 3-UPU parallel
robot and constructed a model of the mechanism. The Driven arm
model is presented in an analytical form, which allowed
us to predict easily the pose error for a given external load, Moving platform
a nominal pose, and the structural parameters of the 3-UPU
parallel manipulator. There are many studies on the error
and accuracy of parallel robots that consider joint Figure 1. The structure of the robot.
errors.12,13 In addition, Cui and Zhu14 studied the influence
of various error sources in the whole workspace, and the Aiming at the kinematics and error analysis of the
error model of the parallel industrial robot was established Delta parallel robot, currently, most studies mainly con-
by Jacobian matrix, then the accuracy analysis was carried sider the influence of dimension error of each component,
out. Chen et al.15 presented an approach to predict the the joint clearance error, and the influence of the driving
accuracy performance of the general planar parallel manip- angle on the motion output of the robot. There are few
ulators due to the input uncertainties and the joint clear- research conducted on the kinematics and error analysis
ance. Based on the theory of envelope, a geometric method considering the degree of non-parallelism of mechanism.
was employed to uniformly construct the indeterminate In this article, the parallelism error of the moving platform
influences of these two error sources on the pose (position is taken into account, the mathematical model is estab-
and orientation) deviation of the manipulators. Xu estab- lished. The kinematics analysis of the robot is performed
lished the kinematic model of Delta parallel robot and by establishing kinematic equations. Therefore, the influ-
studied the influence of each error source on the kinematic ence of parallelism error on the motion accuracy of the
reliability of the mechanism, mainly from the aspect of robot is obtained. The research complements the kine-
geometric dimension errors of each component.16 Zhang matics analysis and error modeling of the Delta robot. It
Xianmin established an error model considering joint clear- is of great significance to the structural optimization and
ance and analyzed the influence of joint clearance on error error compensation of the mechanism.23–27
distribution in repeated positioning.17 Yongjie Zhao intro-
duced a spatial synthesis method of 3UPS-PRU parallel
The description of the Delta robot
robot. Considering the constrained dimensions of the instal-
lation, the size of the robot, and the rotational angle of the and parallelism error
joints, a dimension synthesis method for parallel robot con- The description of the Delta robot
sidering the maximum value was proposed.18
In addition to the above, there is a lot of research on The Delta parallel robot is composed of a fixed platform, a
moving platform, and three identical kinematic chains. The
robots. Pavol Bezaka represented a simulated model of a
three chains are used to connect two platforms, and they are
multi-fingered robotic hand for grasping tasks. An intel-
distributed at 120 . Each chain is composed of a driving
ligent hand-object contact model was developed for a
arm and a driven arm connected by the revolute joint, and
coupled system assuming that the object properties were
the driven arm is composed of a parallelogram mechan-
known.19 In Pavol Bozek’s paper, the new method of cal-
ism.28 The structure of the Delta robot is shown in Figure 1.
culation based on the intrinsic properties of the plane was
proposed. The stated mathematical calculation was a suit-
able method for determining the trajectory of mobile The description of parallelism error
robots and industrial robots.20 In 2018, the article dealt Parallelism describes the degree of parallelism between
with the research of the supplementation of industrial two planes or lines.29,30 It refers to the maximum allowable
robot effector trajectory’s control systems by an inertial error of a plane (edge) parallel to another plane (edge).
navigation system.21 Ivan Kuric’s article dealt with mea- Parallelism evaluates the state of parallelism between
surement of pose repeatability of industrial robot FANUC straight lines, planes, or between straight lines and planes.
LR Mate 200iC (Japan). The measurement methodology One line or plane is the evaluation benchmark. The paral-
was based on the so-called ISO cube placed in the most lelism error between two ideal straight lines or planes is
used part of robot workspace. Individual steps of the solu- fixed,31 meaning that the results of parallelism error are
tion, the analysis of measurement data, and results were same no matter which line or plane is used as the bench-
presented in the article.22 mark. In this article, the transverse bar of the parallelogram
Li et al. 3

zo z1 Di1 Bi Di2
yO
φ
xo x1
O Ai ©i 1
z2
Bi

©i 2
x2

Ei2
Ci
Ci £
P γ
h
P γ
Ei1
Figure 2. Motion diagram of kinematic chain.

Figure 3. Schematic diagram of tilt angle of the moving platform.

mechanism is regarded as an ideal straight line and the
moving platform is regarded as an ideal plane, namely, the
straightness error of the transverse bar and the flatness error
of the moving platform are zero. transverse bar connected to the moving platform and
The driven arm of each chain of the robot is com- the horizontal direction is b, and the angle between the
posed of a pair of parallelogram mechanisms. Due to moving platform and the horizontal plane is g. The distance
the machining error and installation error of the com- between the connecting point (the middle point of the lower
ponent, the parallelogram-driven mechanism will not transverse bar) and the horizontal line of the platform is h,
be parallel. In addition, when the Delta robot is running as shown in Figure 3.
at high speeds, the load will produce greater inertia
impact,32 which will cause the moving platform of the
robot not to be parallel to the horizontal plane, result- Forward kinematic model of the mechanism
ing in a slight tilt angle. Therefore, the non-parallelism The forward solution is calculated by space vector
of the parallelogram will lead to the tilt of the moving method. Taking a kinematic chain as an example, accord-
platform and affect motion accuracy of the robot. The ing to the geometric method, the closed-loop equation of
component and plane are regarded as ideal state in this the ith chain is
article, and then the kinematic equations and error
! ! ! ! !
modeling considering the parallelism error of the AB þ BC ¼ OP þ PC  OA ð1Þ
mechanism are established.
! !
Firstly, OP and OA are transformed from coordinate
system xo yo zo to the base coordinate system x 1 y 1 z 1 . That is,
Kinematic modeling considering the angle ’ is rotated around the z o -axis, where
parallelism error ’¼  ð180 þ ai Þ (the angle ’ is a negative value due to
the clockwise rotation). The coordinate of the end point P is
The coordinate system of the mechanism !
ðPx ; Py ; Pz Þ. Therefore, in the base coordinate system, OP
The kinematic chain of the Delta parallel robot is shown in !
Figure 2. Point O is the center point of the static platform. and OA can be expressed as follows
2 32 3
Point P is the center point of the moving platform. OAi is cos’ sin’ 0 Px
the radius of the static platform, and OAi ¼ R. PC i is the ! 6 76 7
radius of the moving platform, and PC i ¼ r. The length of OP ¼ 6 4 sin’ cos’ 0 54 Py 5
76 7
the driving arm Ai Bi is a, and the length of the parallelo- 0 0 1 Pz
gram mechanism (the driven arm Bi C i ) is c. ai is the dis- 2 32 3
cosai sinai 0 Px
tribution angle of three kinematic chains. ai is 0 , 120 , and 6 76 7
240 , respectively. The input angle (driving angle) is qi1 ¼64 sinai cosai 0 7 6 7
54 Py 5 ð2Þ
and the rotation angle of driven arm is qi2 . The upper and
0 0 1 Pz
the lower transverse bar of the parallelogram mechanism 2 3
are Di1 Di2 and Ei1 Ei2 , respectively, and the length of the Px cosai  Py sinai
6 7
upper and lower transverse bar is l, where i ¼ 1; 2; 3. ¼6 7
4 Px sinai  Py cosai 5
When considering the non-parallelism of the
parallelogram-driven mechanism, the angle between the Pz
4 International Journal of Advanced Robotic Systems

2 32 3 system. Similarly, the vector ½ c 0 0  T rotates the angle

cos’ sin’ 0 R
q 11 around the y 1 -axis, and since the component BC is con-
! 6 76 7
!
OAi ¼ 6 76
4 sin’ cos’ 0 54 0 5
7
nected to the joint B, BC is obtained by rotating q 12 around
0 0 1 0 the y 2 -axis at the same time. Therefore, in the base coordi-
2 32 3 ! !
cosai sinai 0 R nate system, AB and BC can be expressed as follows
6 76 7 2 32 3 2 3
¼64 sinai cosai 0 7 6
54 0 5
7 ð3Þ cosq 1 0 sinq 1 a acosq1
! 6 76 7 6 7
0 0 1 0 AB ¼ 4 0 1 0 54 0 5 ¼ 4 0 5 ð6Þ
2 3 sinq1 0 cosq1 0 asinq 1
Rcosai
6 7 2 32 32 3
¼6 7
4 Rsinai 5 cosq1 0 sinq 1 cosq2 0 c
sinq 2
! 6 76 76 7
0 BC ¼ 6
4 0 1 0 76
54 0 1 0 7 6 7
54 0 5
The tilt angle of the moving platform in the horizontal sinq1 0 cosq1 sinq2 0 cosq2 0
2 3
direction is g, so PC is rotated clockwise by the angle g ðcosq1 cosq 2  sinq 1 sinq 2 Þc
around the y-axis, namely the rotation angle is g. Thus, in 6 7
! ¼6
4 0 7
5
the base coordinate system, PC is written as
2 32 3 ðsinq 1 cosq2  cosq1 sinq 2 Þc
cosðgÞ 0 sinðgÞ r 2 3
! 6 76 7 ccosðq1 þ q 2 Þ
PC ¼ 4 6 0 1 0 7 6 7 6 7
54 0 5 ¼6 0 7
4 5
sinðgÞ 0 cosðgÞ 0
2 32 3 csinðq1 þ q 2 Þ
cosg 0 sing r
6 76 7 ð7Þ
¼6 4 0 1 0 7 6 7
54 0 5 ð4Þ
Then we can get
sing 0 cosg 0 2 3 2 3
2 3 acosq1 ccosðq1 þ q 2 Þ
rcosg
6 7 ! ! 6 7 6 7
AB þ BC ¼ 6 0 7þ6 0 7
¼6 4 0 5
7 4 5 4 5
asinq 1 csinðq1 þ q 2 Þ
rsing 2 3 ð8Þ
acosq1 þ ccosðq 1 þ q2 Þ
From the above formulas, it can be obtained as 6 7
¼6
4 0 7
5
! ! !
OP þ PC  OA
2 3 2 3 2 3 asinq 1  csinðq1 þ q 2 Þ
Px cosai  Py sinai rcosg Rcosai
6 7 6 7 6 7 From the above formulas, it can be obtained as
¼6 7 6 7 6
4 Px sinai  Py cosai 5 þ 4 0 5  4 Rsinai 5
7
2 3
Px cosai  Py sinai þ rcosg  Rcosai
Pz rsing 0 6 7
2 3 6 7
Px cosai  Py sinai þ rcosg  Rcosai 4 Px sinai  Py cosai  Rsinai 5
6 7
¼64 Px sinai  Py cosai  Rsinai 7
5
Pz þ rsing
2 3 ð9Þ
acosq 1 þ ccosðq1 þ q 2 Þ
Pz þ rsing 6 7
ð5Þ ¼64 0 7
5
! asinq 1  csinðq 1 þ q2 Þ
AB can be regarded as the vector, ½ a 0 0  T rotates
the angle q 11 around the y 1 -axis in the base coordinate The three equations of kinematic chain can be written as

8
< Px cosai  Py sinai þ rcosg  Rcosai ¼ acosq 1 þ ccosðq 1 þ q 2 Þ
>
Px sinai  Py cosai  Rsinai ¼ 0 ð10Þ
>
:
Pz þ rsing ¼ asinq1  csinðq1 þ q 2 Þ

For the first kinematic chain, a 1 ¼ 0 . The input angle (driving angle) is q 11 , and the rotation angle of driven arm is q12 .
Thus, the kinematics equation of the first kinematic chain is written as
Li et al. 5

8   
< P1x cosð0 Þ  P1y sinð0 Þ þ rcosg  Rcosð0 Þ ¼ acosq21 þ ccosðq21 þ q22 Þ
>
P1x sinð0 Þ  P1y cosð0 Þ  Rsinð0 Þ ¼ 0 ð11Þ
>
:
P1z þ rsing ¼ asinq1  csinðq21 þ q22 Þ

It can be obtained as
8
< P1x ¼ rcosg  R  acosq11  ccosðq11 þ q12 Þ
>
P1y ¼ 0 ð12Þ
>
:
P 1z ¼ asinq 11  csinðq11 þ q12 Þ  rsing

For the second kinematic chain, a 2 ¼ 0 . The input angle (driving angle) is q 21 , and the rotation angle of driven arm is
q 22 . Thus, the kinematics equation of the second kinematic chain is written as
8   
< P2x cosð120 Þ  P2y sinð120 Þ þ rcosg  Rcosð120 Þ ¼ acosq 21 þ ccosðq 21 þ q 22 Þ
>
  
P2x sinð120 Þ  P2y cosð120 Þ  Rsinð120 Þ ¼ 0 ð13Þ
>
:
P2z þ rsing ¼ asinq 1  csinðq 21 þ q 22 Þ

Namely,
8 pffiffiffi
>
> 1 3 1
>
> P2x  P2y þ rcosg þ R ¼ acosq21 þ ccosðq21 þ q22 Þ
>
> 2 2 2
>
< pffiffiffi pffiffiffi
3 1 3 ð14Þ
>
> P 2x þ P2y  R¼0
>
> 2 2 2
>
>
>
:
P2z þ rsing ¼ asinq 21  csinðq21 þ q22 Þ

It can be obtained as
8
>
> 1
>
> P 2x ¼ ½R þ rcosg  acosq21  ccosðq21 þ q22 Þ
>
> 2
>
< pffiffiffi
3 ð15Þ
>
> P2y ¼ ½ðR þ rcosg  acosq 21  ccosðq 21 þ q 22 Þ
>
> 2
>
>
>
:
P2z ¼ asinq21  csinðq 21 þ q 22 Þ  rsing

For the third kinematic chain, a 3 ¼ 0 . The input angle (driving angle) is q31 , and the rotation angle of driven arm is
q 32 . Thus, the kinematics equation of the third kinematic chain is written as
8   
< P3x cosð240 Þ  P3y sinð240 Þ þ rcosg  Rcosð240 Þ ¼ acosq 31 þ ccosðq 31 þ q 32 Þ
>
P3x sinð240 Þ  P3y cosð240 Þ  Rsinð240 Þ ¼ 0 ð16Þ
>
:
P3z þ rsing ¼ asinq 31  csinðq31 þ q32 Þ

Namely,
8 pffiffiffi
>
> 1 3 1
>
> P3x þ P3y þ rcosg þ R ¼ acosq31 þ ccosðq31 þ q32 Þ
>
> 2 2 2
>
< pffiffiffi pffiffiffi
3 1 3 ð17Þ
>
>  P3x þ P3y þ R¼0
>
> 2 2 2
>
>
>
:
P3z þ rsing ¼ asinq 31  csinðq31 þ q32 Þ
6 International Journal of Advanced Robotic Systems

It can be obtained as From the above analysis, the position error of the end-
8 effector caused by parallelism error in x, y, and z directions
>
> 1
>
> P3x ¼ ½acosq 31 þ ccosðq 31 þq 32 Þ  rcosg þ R is DPiq . i ¼ 1; 2; 3, q ¼ x; y; z. So it can be obtained as
>
> 2
>
< pffiffiffi DPiq ¼ Kqiq Dg ð24Þ
3
>
> P 3y ¼ ½acosq31 þ ccosðq31 þq32 Þ  rcosg  R
>
> 2 In equation (24), Kqiq is the error sensitivity coefficient
>
>
>
: of Dg. The value of Kqiq can quantitatively reflect the
P3z ¼ asinq 31  csinðq 31 þ q 32 Þ  rsing
degree of the influence of error Dg on motion accuracy
ð18Þ of the robot.
For the first kinematic chain, the error sensitivity coeffi-
From Figure 3, according to the geometric relationship,
cient of the parallelism error in x, y, and z directions is as
it can be concluded that
follows
h 2h 8
sinb ¼ ¼ ð19Þ > @DP1x
ðl=2Þ l >
> K 1x ¼ @g ¼ rsing
>
>
>
>
Because h ¼ lsinb h lsinb >
>
2 , sing ¼ r , so we can get sing ¼ 2r . < @DP1y
Therefore, the relationship between the tilt angle of the K 1y ¼ ¼0 ð25Þ
> @g
moving platform and the tilt angle of the parallelogram >
>
>
>
mechanism is as follows >
> @DP1z
>
  : K 1z ¼ @g ¼ rcosg
>
lsinb
g ¼ arcsin ð20Þ
2r
For the second kinematic chain, the error sensitivity
Motion error analysis considering coefficient of the parallelism error in x, y, and z directions
is as follows
parallelism error 8
>
> @DP2x 1
Ideally, the moving platform is parallel to the horizontal > K 2x ¼
> ¼ rsing
plane, namely g ¼ 0 . From the previous analysis results, >
> @g 2
>
>
we can calculate the error value of the three chains in three >
> pffiffiffi
< @DP 2y 3
directions. K 2y ¼ ¼ rsing ð26Þ
>
> @g 2
For the first kinematic chain, the error value in x, y, and z >
>
>
>
directions is as follows: >
> @DP 2z
8 >
> K ¼ ¼ rcosg
: 2z @g
< DP1x ¼ r  rcosg ¼ rð1  cosgÞ
>
DP1y ¼ 0  0 ¼ 0 ð21Þ For the third kinematic chain, the error sensitivity coef-
>
:
DP1z ¼ 0  ðrsingÞ ¼ rsing ficient of the parallelism error in x, y, and z directions is as
follows:
For the second kinematic chain, the error value in x, y, 8
and z directions is as follows >
> @DP3x 1
8 > K 3x ¼
> ¼  rsing
>
> @g 2
>
> 1 1 >
>
>
> DP2x ¼ ðr  rcosgÞ ¼ rð1  cosgÞ >
> pffiffiffi
>
> 2 2 < @DP3y 3
>
< K 3y ¼ ¼ rsing ð27Þ
pffiffiffi > @g 2
3 ð22Þ >
>
>
> DP2y ¼ rð1  cosgÞ >
>
>
> 2 >
> @DP3z
> >
>
>
>
: > K 3z ¼ @g ¼ rcosg
:
DP2z ¼ 0  ðrsingÞ ¼ rsing
For the third kinematic chain, the error value in x, y, and
z directions is as follows Numerical example
8
>
> 1 1 This article takes the robot of Chen Xing (Tianjin) automa-
>
> DP 3x ¼ ðr  ðrcosgÞÞ ¼  rð1  cosgÞ
>
> 2 2 tion equipment company in our laboratory as an example.
>
< pffiffiffi pffiffiffi The model of this robot is D3PM-1000. It is mainly com-
3 3 posed of a fixed platform, moving platform, driving arm,
>
> DP 3y ¼ ðr  ðrcosgÞÞ ¼  rð1  cosgÞ
>
> 2 2 and driven arm. The robot’s workspace covers a diameter
>
>
>
: of 500–1400 mm, and its grasping speed is 75–150 times
DP3z ¼ 0  ðrsingÞ ¼ rsing
per minute. Its maximum acceleration and maximum velo-
ð23Þ city are 100 m/s2 and 8 m/s, respectively. It has the
Li et al. 7

5
Fixed platform 4.5 x direction
y direction

The error in x, y, and z direction (mm)

4 z direction
Driving arm
3.5

Driving motor 2.5

Driven arm 1.5

0.5
Moving platform
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
The tilt angle of the moving platform (rad)

Conveyor belt Figure 5. Error curve of the first kinematic chain.

Figure 4. The diagram of D3PM-1000 robot.

5

4.5 x direction
y direction
The error in x, y, and z direction (mm)

Table 1. Structural parameters of the robot. 4 z direction

Structural parameters Numerical value 3.5

3
r1 (mm) 125
r2 (mm) 50 2.5

a (mm) 400 2
c (mm) 950
1.5

r1: radius of the fixed platform; r2: radius of the moving platform; a: the 1
length of the driving arm; c: the length of the driven arm.
0.5

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
characteristics of moving in x, y, and z directions and is The tilt angle of the moving platform (rad)
widely used in the field of high-speed sorting and packa-
ging. The robot is shown in Figure 4. Figure 6. Error curve of the second kinematic chain.
Through the product instruction manual and actual mea-
surement, the structural parameters of this robot are pre-
sented in Table 1.
The distribution angles of the driving arm are a 1 ¼ 0 ,
a 2 ¼ 120 , and a 3 ¼ 240 , respectively. 5
4.5 x direction
Since the parallelism error and tilt angle of the robot are y direction
4
The error in x, y, and z direction (mm)

small, the range of the tilt angle g between the moving z direction
3.5
platform and the horizontal plane can be taken as
3
g 2 ½0; 0:1 rad. 2.5
The simulation analysis is carried out with MATLAB 2
2014a. It can be concluded that on three kinematic chains, 1.5
the relationship in three directions between the motion 1
error caused by the tilt angle g, and the tilt angle is shown 0.5
in Figures 5 to 7. 0
From equation (21), we can get the relationship between -0.5
the motion error of the first kinematic chain and the tilt -1
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
angle of the platform.
The tilt angle of the moving platform (rad)
From equation (22), we can get the relationship between
the motion error of the second kinematic chain and the tilt
Figure 7. Error curve of the third kinematic chain.
angle of the platform.
8 International Journal of Advanced Robotic Systems

0.14
It can be seen that b decreases with the increase in l, so
when the parallelism error caused by the component error is
determined, namely the value of h is fixed, the length l of
The tilt angle of the moving platform (rad)

0.12
the transverse bar can be increased appropriately to reduce
0.1
the tilt angle b of the driven mechanism. Moreover, we can
0.08
also improve motion accuracy by controlling the error of
driven components.
0.06

0.04 Conclusions
0.02
In this article, the influence of the non-parallelism of the
driven mechanism on the parallelism of the moving plat-
0 form is analyzed. The kinematic equation and error model-
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
The tilt angle of the parallelogram mechanism (rad)
ing of the mechanism considering parallelism error are
established. The influence of parallelism error on motion
Figure 8. The relationship between the tilt angle of the moving
error is analyzed. The simulation results show that the par-
platform and the tilt angle of the parallelogram mechanism. allelism error has a great influence on motion accuracy of
the robot, especially in the z direction. When studying the
motion error of parallel mechanisms that are similar to
Delta robots, it is necessary to consider the parallelism
From equation (23), we can get the relationship between
error. The results of this article can be applied to other
the motion error of the third kinematic chain and the tilt high-speed and high-precision mechanisms, such as high-
angle of the platform. speed parallel machine tools. To improve the manufactur-
It can be seen from Figures 5 to 7 that with the increase ing precision, the influence of parallelism error should be
in the tilt angle of the moving platform, the motion error considered and controlled. The research complements the
caused by the parallelism error also increases and the var- motion error analysis and provides a theoretical basis for
iation value is obvious, especially in the z direction, mean- the high-precision calibration and error compensation of
ing that the parallelism error has a great influence on the robots. Besides, what factors can cause non-parallelism and
motion error of the mechanism. Among these, the influence how these factors affect motion accuracy of motion are my
degree in z direction (i.e. vertical direction) is significantly next research work.
greater than that in x and y direction (i.e. horizontal direc-
tion). Therefore, it is necessary to control the component Declaration of conflicting interests
error of the parallelogram-driven mechanism and reason- The author(s) declared no potential conflicts of interest with
ably increase the rigidity of the moving platform, thereby respect to the research, authorship, and/or publication of this
effectively reducing the parallelism error and achieving the article.
purpose of improving motion accuracy.
According to the calculation results, it is obtained that Funding
 
g ¼ arcsin lsinb
2r . The author(s) disclosed receipt of the following financial support
By measurement, the length of the transverse bar is l, for the research, authorship, and/or publication of this article: This
l ¼ 120 mm and r ¼ 50 mm are also known. Similarly, the research was funded by the National Natural Science Foundation
range of the tilt angle g between the moving platform and of China (Science Foundation) [nos 51834006 and 11802035] and
the Fundamental Research Funds for the Central Universities [nos
the horizontal plane can be taken as g 2 ½0; 0:1 rad. Sub-
2019QJ002 and 2011YJ01].
stituting the numerical values, it is obtained that
 
g ¼ arcsin 120sinb
100 ¼ arcsinð1:2sinbÞ. ORCID iD
So the relationship between the tilt angle of the moving Yu Li https://orcid.org/0000-0001-6746-0747
platform and that of the parallelogram mechanism is shown
in Figure 8. References
Figure 8 reflects the relationship between the paralle-
1. Der-Ming K. Direct displacement analysis of a Stewart plat-
lism error of the moving platform and the non-parallelism
form mechanism. Mech Mach Theory 1999; 34(3): 453–465.
of the parallelogram-driven mechanism. As the parallelism
2. Mahmood M, Mostafa T, and Najafi MR. Kinematic analysis
error of the driven mechanism increases, the tilt angle of
and design of a 3-DOF translational parallel robot. Int J
the moving platform also increases. Thus, the motion error
Autom Comput 2017; 14(04): 432–441.
of the end-effector is increasing. 3. Hunt KH. Structural kinematics of in- parallel-actuated robot
h
From the formula sinb ¼ ðl=2Þ ¼ 2hl , we can get: - arms. ASME J Mech Trans Autom Des 1983; 105 (4):
 2h
b ¼ arcsin l 705–712.
Li et al. 9

4. Korayem MH and Tourajizadeh H. Maximum DLCC of spa- 17. Zhang XM and Liu W. Kinematic calibration and error com-
tial cable robot for a predefined trajectory within the work- pensation of 3-RRR parallel robot with clearance. J South
space using closed loop optimal control approach. J Intell China Univ Technol (Nat Sci Ed) 2014; 42(07): 97–103.
Robot Syst 2011; 63(1): 75–99. 18. Zhao YJ and Cheng G. Dimensional synthesis of a 3UPS-
5. Dai ZW, Sheng XJ, Hu J, et al. Design and implementation of PRU parallel robot. Robotica 2017; 35(12): 2319–2329.
Bezier curve trajectory planning in Delta parallel robots. 19. Bezak P, Bozek P, and Nikitin Y. Advanced robotic grasping
Berlin: Springer International Publishing, 2015. system using deep learning, Procedia Eng 2014; 96: 10–20.
6. Qiu H, Li Y, and Li YB. A new method and device for 20. Bozek P, Pokorny P, Svetlik J, et al. The calculations of
motion accuracy measurement of NC machine tools. Part Jordan curves trajectory of the robot movement. Int J Adv
1: principle and equipment. Int J Mach Tool Manuf 2001; Robot Syst 2016; 13: 1–7.
41(4): 521–534. 21. Pivarciova E, Bozek P, Turygin Y, et al. Analysis of control
7. Tsai LW., Walsh GC, and Stamper RE. Kinematics of novel and correction options of mobile robot trajectory by an iner-
three DOF translational platforms. In: Proceedings of IEEE tial navigation system. Int J Adv Robot Syst 2018; 15(1):
international conference on robotics and automation, Min- 1–15.
neapolis, MN, USA, 22–28 April 1996, pp. 3446–3451. Min- 22. Kuric I, Tlach V, Sagova Z, et al. Measurement of industrial
nesota: IEEE (Robotics&Automation Society) Staff. robot pose repeatability. MATEC Web Conf 2018; 244:
8. Huang M. Research on the working performance and pose 01015.
error with sphere joint clearance of the Delta robot for pick- 23. Qu WW, Dong HY, and Ke YL. Pose accuracy compensation
ing operation. South China University of Technology, technology in robot-aided aircraft assembly drilling process.
Guangzhou, 2015. J Aeronaut 2011; 32(10): 1951–1960.
9. Yang Q, Sun ZL, Yan M, et al. Kinematic reliability of 24. Wang SM and Ehmann KF. Error model and accuracy anal-
improved delta parallel mechanism. J Aeronaut 2008; ysis of a six-DOF Stewart platform. J Manuf Sci Eng 2002;
29(02): 487–491. 124(2): 286–296.
10. Ropponen R and Arai T. Accuracy analysis of a modified 25. Zhou W, Liao WH, and Tian W. Theory and experiment of
Stewart platform manipulator. In: Proceedings of the 1995 industrial robot accuracy compensation method based on spa-
international conference on robotics and automation, tial interpolation. J Mech Eng 2013; 49(03): 42–48.
Nagoya, Aichi, Japan, 21–27 May 1995, pp. 521–525. New 26. Zhou W, Liao WH, Tian W, et al. Method of industrial robot
York: IEEE. accuracy compensation based on particle swarm optimization
11. Chebbi AH, Affi Z, and Romdhane L. Prediction of the pose neural network. China Mech Eng 2013; 24(02): 174–179.
errors produced by joints clearance for a 3-UPU parallel 27. Wang D, Bai Y, and Zhao J. Robot manipulator calibration
robot. Mech Mach Theory 2009; 44(9): 1768–1783. using neural network and a camera-based measurement sys-
12. Mohan S. Error analysis and control scheme for the error tem. Trans Inst Meas Control 2012; 34(1): 105–121.
correction in trajectory-tracking of a planar 2PRP-PPR par- 28. Zolfa A, Parnyan A, and Mehdi TM. Collision-free work-
allel manipulator. Mechatronics 2017; 46: 70–83. space and kinetostatic performances of a 4-DOF delta parallel
13. Fu JX, Gao F, Chen WX, et al. Kinematic accuracy research robot. J Braz Soc Mech Sci 2019; 41(2): 1–7.
of a novel six-degree-of-freedom parallel robot with three 29. Yu J, Yu W, Wu CY, et al. Error modeling and sensitivity
legs. Mech Mach Theory 2016; 102: 86–102. analysis of linear driven parallel robot. Trans Chinese Soc
14. Cui HL and Zhu ZQ. Industrial robotics: theory, modelling Agric Eng 2017; 48(07): 383–390.
and control. Germany: Pro Literatur Verlag, 2006. 30. Peter V and Reymond C. Kinematic calibration of the parallel
15. Chen GL, Wang H, and Lin ZQ. A unified approach to the delta robot. Robotica 1998; 16(02): 207–218.
accuracy analysis of planar parallel manipulators both with 31. Wen XQ. Study on assessment strategies and measuring
input uncertainties and joint clearance. Mech Mach Theory methods for double-line parallelism. Harbin University of
2013; 64: 1–17. Technology, Harbin, 2018.
16. Xu DT. Kinematic reliability and sensitivity analysis of the 32. Zheng KM and Zhang QJ. Elastic dynamics and analysis of
modified Delta parallel mechanism. Int J Adv Robot Syst vibration characteristics of Delta robot with joint clearance.
2018; 15(1): 1–8. Trans Chinese Soc Agric Eng 2015; 31(14): 39–48.