Mechanism and Machine Theory 87 (2015) 117

Contents lists available at ScienceDirect

Mechanism and Machine Theory

journal homepage: www.elsevier.com/locate/mechmt

Kinematic optimal design of a 2-degree-of-freedom

3-parallelogram planar parallel manipulator
Xin-Jun Liu a,b,, Jie Li a, Yanhua Zhou a,c
a
b
c

State Key Laboratory of Tribology & Institute of Manufacturing Engineering, Department of Mechanical Engineering, Tsinghua University, Beijing 100084, PR China
Beijing Key Lab of Precision/Ultra-precision Manufacturing Equipments and Control, Tsinghua University, Beijing 100084, PR China
Institute of Aircraft Engineering, Naval Aeronautical and Astronautical University, Yantai, Shandong Province, PR China

a r t i c l e

i n f o

Article history:
Received 7 December 2013
Received in revised form 18 December 2014
Accepted 25 December 2014
Available online 10 January 2015
Keywords:
Parallel manipulator
Optimal design
Performance atlas
Workspace
Transmissibility

a b s t r a c t
A 2-degree-of-freedom (2-DOF) 3-parallelogram planar parallel manipulator (PPM) can follow an
arbitrary planar curve and keep the end-effector in a denite posture. Such features are valuable
for spray-painting robots. Considering these advantages, authors proposed a new spray-painting
robot containing a 2-DOF 3-parallelogram PPM. In order to obtain a spray-painting robot with the
best performance, the 2-DOF 3-parallelogram PPM should possess the largest workspace and
most stable transmissibility. This study addresses the performance evaluation and kinematic optimization of this manipulator. First, the kinematics of the manipulator is analyzed, and performance indices that consider desirable workspace and transmissibility are proposed. Then, the
process to determine optimal geometry parameters by using performance atlases is presented. Finally, a 2-DOF 3-parallelogram PPM with desirable workspace and transmissibility is identied.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Parallel mechanisms (PMs) exhibit excellent characteristics, such as higher rigidity, better positioning accuracy, and higher load
capacity, compared with serial mechanisms [1]. As an important branch of PMs, the 2-DOF PPMs possessing both outstanding characteristics and simple structure have been widely used in the industry eld. PMs have been intensively studied since the 1980s and
still attract much attention up to the present [26]. For PMs one of the most important and challenging problems is kinematic optimization in which two issues are concerned: performance evaluation and dimensional synthesis [7,8].
The 2-DOF 3-parallelogram PPM, actuated by two coaxial revolute joints, is a typical 2-DOF PPM. It has a non-symmetrical structure composed of three parallelogram mechanisms. This PPM has a prominent function that does not only follow an arbitrary planar
curve precisely within the workspace but also keep the end-effector in a denite posture at all times. Given its outstanding advantages
and simple structure, this PPM has been widely applied in palletizing robots [9].
Regardless of how simple a 2-DOF PPM is, the optimal design is challenging. Many researchers have devoted themselves to this
issue for many years. Gao et al. [10] presented a physical model of the solution space of 2-DOF PPMs and systematically discussed
the comprehensive classication of this kind of manipulators. Huang et al. [11] presented a hybrid method for the kinematic optimal
design of 2-DOF PPMs with a mirror symmetrical geometry. Liu et al. [12] addressed the graphical representation of performance and
the optimum design issue of planar 5R symmetrical PPMs. However, all 2-DOF PPMs discussed in these articles have a mirror symmetrical structure. The 2-DOF 3-parallelogram PPM is another type of 2-DOF PPM with a non-symmetrical structure. Although many

Corresponding author at: Room 1502, Building 9003, Tsinghua University, Beijing 100084, PR China. Tel.: +86 10 6278 9211.
E-mail address: xinjunliu@mail.tsinghua.edu.cn (X.-J. Liu).

http://dx.doi.org/10.1016/j.mechmachtheory.2014.12.014
0094-114X/ 2014 Elsevier Ltd. All rights reserved.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

articles have investigated its application in palletizing robots, including structural optimization [13], optimum motion control [14] and
layout analysis [15], articles about the kinematic optimal design of this manipulator in spray-painting robots are seldom found.
The traditional kinematic optimization method for PMs is the objective function method that involves establishing an objective
function and obtaining a result via algorithms [1618]. Three unavoidable difculties are encountered in this method. (a) The highly
non-linear objective function is difcult to establish, and some simplication are inevitably involved in objective function. Thus, the
iterative solving process is not only time consuming, but the optimal result based on objective function method is sometimes unreliable. (b) This method may provide an optimal result, but users cannot know how optimal the result is because the relationship between design parameters and performance indices is unknown. (c) If the design conditions vary, then users have no choice but to
restart their work from the beginning [19]. Compared with the traditional objective function method, the performance atlases method
exhibits several obvious advantages. It not only graphically and globally shows the relationship between performance indices and design parameters but also provides all possible solutions to users [20]. Users can obtain the optimal results by comparing all possible
solutions in the same chart, which is convincing and convenient.
Performance evaluation and dimensional synthesis are two important issues in kinematic optimization. To conduct kinematic optimal design for a 2-DOF 3-parallelogram PPM, performance indices are required. The local conditioning index (LCI) [21] and the global conditioning index (GCI) [2] have been widely used by numerous researchers. However, a recent study [22] found serious
inconsistencies when these indices were applied to the kinematic optimal design of mixed-DOF PMs. Furthermore, the LCI cannot provide a mathematical distance between current position and singularity [7], which also limits the application of this index. Wu et al. [7]
proposed two simple but useful frame-free indices, namely, the local transmission index (LTI) and the global transmission index (GTI).
The LTI can judge the effectiveness of transmissibility in a single pose, whereas the GTI represents the average effectiveness of a series
of poses in the good transmission workspace (GTW). Although the GTI can provide a PM with good average transmissibility, how the
effectiveness of transmissibility in each pose varies around the average transmissibility remains unclear. This variation demonstrates
the uctuation of the LTI around the GTI, which is particularly important for the optimal design of driving motors. According to the
application requirements in the spray-painting robots, new indices that consider desirable workspace and transmissibility are introduced to comprehensively solve the kinematic optimization problems of the 2-DOF 3-parallelogram PPM. Then based on these performance indices, the kinematic optimization is achieved by using performance atlas method. Finally, a 2-DOF 3-parallelogram
PPM with desirable workspace and transmissibility is identied.
This paper is organized as follows. The kinematics of a 2-DOF 3-parallelogram PPM is analyzed in Section 2. Section 3 investigates
the design space of the manipulator, including its length and angular design parameters. The GTW of the manipulator is discussed in
Section 4. According to the application requirements in the spray-painting robots, performance indices based on desirable workspace
and transmissibility are proposed in Section 5. Performance atlases based on these performance indices are also illustrated in this
section. Section 6 introduces the use of performance atlases to determine the optimum parameters and provides an optimal manipulator. Conclusions are presented in Section 7.
2. Kinematics of a 2-DOF 3-parallelogram PPM
2.1. Architecture
Fig. 1(a) shows a 2-DOF 3-parallelogram PPM. Fig. 1(b) presents the physical model of a spray-painting robot proposed by authors
which contains a 2-DOF 3-parallelogram PPM. The 2-DOF 3-parallelogram PPM is widely used in palletizing robots, but seldom applied in the spray-painting robots. As shown in Fig. 1(a), the 2-DOF 3-parallelogram manipulator consists of three parallelogram
mechanisms, and each parallelogram mechanism can be congured as a parallelogram form or an anti-parallelogram form. Thus,
there can be several conguration modes for the end-effector. However, the anti-parallelogram mechanisms are easy to cause

(b)

(a)

driving motor

spray gun

Fig. 1. The 2-DOF 3-parallelogram PPM: (a) kinematic structure and (b) application in a spray-painting robot.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

geometry interference problems which are inappropriate for both palletizing robots and the spray-painting robots. In order to concentrate the target, only the conguration mode that contains no anti-parallelograms is studied in this article. The end-effector of the manipulator is the link IP that is connected to the base through three parallelogram mechanisms, namely, OABC, ODEC, and CFGH. Link CH
in parallelogram mechanism CFGH and link CE in parallelogram mechanism ODEC are xed together. In addition, link CF in parallelogram mechanism CFGH and link CB in parallelogram mechanism OABC are the two edges of triangle link BFC. Link HI represents the
distance between the center of the end-effector (link IP) and that of joint H. Links IP and HI are xed together. All revolute joints in
this manipulator are perpendicular to plane XOY. Parallelogram mechanisms OABC, ODEC, and CFGH share a revolute center located
in point C. Parallelogram mechanisms OABC and ODEC share a revolute center located in point O. Link OA is xed on the serial manipulator assembly unit, as shown in Fig. 1(b). Two coaxial actuating revolute joints are located in point O, as shown in Fig. 2. When links
OC and OD are driven by actuating joints, the end-effector (link IP) can follow an arbitrary planar curve within the workspace and
maintain a denite posture at all times.
Geometric parameters in this manipulator can be described as follows. is the angle between xed link OA and axis OX in coordinate system XOY. is the vertex angle between edges CF and CB in triangle link CFB. Link OC is the arm of the manipulator, and
its length is R1. Link CH is the forearm, and its length is R3. The length of link HI is R, which represents the distance between the center
of joint H and that of the end-effector. The end-effector is a cylindrical spray gun assembled on a moving platform, as shown in
Fig. 1(a). R2 is the distance between the spray nozzle and the moving platform. This distance must satisfy the spray requirements
of the spray gun. Eight kinematic angular parameters are observed in this manipulator. 1 is the angle between link OD and axis OX
that is driven by the actuating joint. 2 is the angle between link OC and axis OX that is driven by another actuating joint. The other
six angular parameters are passive angles expressed by
1 DEC 1 2 ;
2 ABC 1 ;
3 FGH 2 ;

1 ODE 1 2 ;
2 BCO 1 ;
3 C FG 2 :

2.2. Kinematic analysis

To conduct performance evaluation and dimensional synthesis, the kinematics of the 2-DOF 3-parallelogram PPM must be investigated rst. According to Fig. 2, the position of the end-effector (point P) in coordinate system XOY can be expressed by the following
equation:

X P R1 cos 1 R3 cos 2 R
:
0
Y P R1 sin 1 R3 sin 2 R2

In practical applications, R should be as small as possible to obtain a compact structure. In general, the end-effector is a cylindrical
spray gun, shown in Fig. 1(a), and R can be regarded as the sum of the radius of joint H and that of the spray gun. Hence, R is a constant
value that cannot change the position range of the end-effector. Consequently, Eq. (1) can be reasonably simplied as follows:

X P R1 cos 1 R3 cos 2
:
Y p R1 sin 1 R3 sin 2 R2

According to Figs. 1 and 2, joint O and link OA are xed on the serial manipulator assembly unit and only the driven angles
above axis OX are meaningful for this spray-painting robot. Furthermore, according to the practical application of this 2-DOF 3paralelloggram manipulator, in order to avoid geometric interference with cars to be painted and singularity caused by out of control,

Fig. 2. Kinematic scheme of the common working mode for a 2-DOF 3-parallelogram PPM.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

the driven angles are always dened as 1 (90, 180) and 2 (0, 90). The ranges of the driven angles constitute a driving space
for the manipulator. When the driven angles vary in the driving space, the end-effector can reach different positions in coordinate system XOY. All possible positions of the end-effector constitute the workspace of the 2-DOF 3-parallelogram PPM.
In the practical application of a 2-DOF 3-parallelogram PPM, the working trace of the end-effector is determined by the driven angles (1 and 2) although the desirable workspace is given by a work task. To regulate the working trace, the values of the driven angles
that correspond with the positions of the end-effector are required. There are two solutions of 1 and 2 shown as Eq. (3) where i =
1, 2 to reach the same position (XP, YP) according to Eq. (2). Therefore, there can be theoretically two working modes for the 2-DOF 3parallelogram PPM to obtain the same pose. However, the aforementioned driven angles have been dened as 1 (90, 180) and
2 (0, 90). Only the solution in Eq. (3) where i = 1 can be used and the corresponding working mode of the 2-DOF 3parallelogram PPM is shown in Fig. 2.
(

1 2 tan ui i 1; 2
;
1
2 2 tan vi i 1; 2

where
a ib
i 1; 2;
c

a 2R1 R2 Y p ;

2 
2
2
c R1 X p R2 Y p R3 ;

2 
2
2
e X p R3 R2 Y p R1 ;

2
l R2 Y p ;

ui

n R1 R3 ;

d ib
vi
i 1; 2;
e
q

b k lmk ln;


d 2R3 R2 Y p ;
2

k Xp ;
2

m R1 R3 ;
1 1;

2 1:

3. Design space
According to Fig. 2, both length parameters (R1, R2 and R3) and angular parameters ( and ) are involved to determine a 2-DOF 3parallelogram PPM. All these geometric parameters constitute a design space for the manipulator. According to Eq. (2), when the driven angles are given, the position of point P is determined by length parameters. Hence, in the same driving space, the workspace of this
manipulator is determined by the length parameters, including R1, R2 and R3. In other words, the workspace of this manipulator has no
relationship with the angular parameters. Meanwhile, as link OA is xed, the angle between link CB and axis OX remains constant via
parallelogram mechanism OABC, thus indicating that the posture of triangle link CFB is constant. Similarly, the end-effector (link IP)
maintains a dened posture via parallelogram mechanisms CFGH and OABC. Thus, the working posture of the end-effector has no relationship with the length parameters and is determined only by angular parameters and . That is, the workspace and the working
posture are uncoupled in the design space. Therefore, we can divide the design space into two independent subspaces, namely, the
length design space and the angular design space.
3.1. Length design space
The length design space consists of three length parameters R1, R2, and R3. Theoretically, each length parameter can vary from zero
to innite. Thus, it is impractical to conduct kinematic optimization using these length parameters directly. To illustrate the relationship between performance indices and length parameters in a performance atlas, these length parameters should be nondimensionalized. Dividing all the length parameters by their average is an efcient method, as used in [23] and [24]. Let
D R1 R2 R3 =3:

Then length parameters Ri (i = 1, 2, 3) are divided by D to dene three non-dimensional parameters as follows:
r 1 R1 =D; r 2 R2 =D;

and r 3 R3 =D:

Obviously,
r 1 r 2 r3 3 0 b r 1 b 3; 0 b r 2 b 3 and 0 b r 3 b 3:

Therefore, Eq. (2) can be rewritten in non-dimensional space as follows:

xP r 1 cos 1 r 3 cos 2
yp r 1 sin 1 r 3 sin 2 r 2

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

where (xp, yp) is the non-dimensional position of the end-effector in the workspace corresponding with (Xp, Yp). According to Eq. (6),
the distribution of the non-dimensional parameters (r1, r2, and r3) can be illustrated as triangle UVW in the non-dimensional design
space in Fig. 3.
Although three non-dimensional length parameters are indicated in Eq. (7), only two of them are independent according to
Eq. (6). A linear map can be made to convert r1, r2, and r3 in the spatial coordinate system to a planar coordinate system dened by
s and t, thus allowing performance evaluation and dimensional synthesis based on performance atlases to be conducted in a planar
coordinate system conveniently. Let
t r3

0 b t b 3 ;

p
p  p
p
s r 3 = 3 2r 2 = 3
t= 3 b s b 6t = 3 :

Then, Eq. (7) can be rewritten as follows:



8
p
< xP 3 3s=2t=2 cos 1 t cos 2



p
pp :
: y 3 3s=2t=2 sin t sin st= 3 3=2
p
1
2

10

Meanwhile, based on the linear map shown in Eqs. (8) and (9), the non-dimensional length design space shown as triangle UVW in
Fig. 3 can be converted to triangle UVW in Fig. 4. According to Eq. (5), the value of each non-dimensional parameter (r1, r2, or r3) in
triangle UVW has a variation range from 0 to 3 and there can be 12 kinds of relationships between them according to their values.
The manipulators with the length parameters under the same relationship have the similar workspace. Thus, these 12 relationships
can be regarded as the boundary condition to classify a manipulator. Three non-dimensional parameters under the same relationship
are located in the same region in the triangle. Then, this triangle UVW can be divided into 12 regions, as shown in Fig. 4. Finally, According to Fig. 4, we can clearly recognize the relationship of the parameters and identify the workspace properties of the manipulators in each region.
3.2. Angular design space
Two angular parameters determine the posture of the end-effector, namely, (the angle between xed link OA and axis OX) and
(the vertex angle between the edges CF and CB in triangle link CFB). According to the practical application of this manipulator in
Fig. 1(a), in order to avoid geometric interference between link AB and the driving motor, generally, joint A is always xed above
axis OX shown as in Fig. 2. Thus, the angle between xed link OA and axis OX can be dened as

0 b b 180 :

11

Meanwhile, is an inner angle of triangle CFB and it should naturally satisfy the condition as follows

0 b b 180 :

12

Fig. 3. Non-dimensional design space of the length parameters.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

Fig. 4. Design space of s and t.

Therefore, the ranges of these two angular parameters, i.e., conditions (11) and (12), constitute the angular design space for the 2DOF 3-parallelogram PPM.
4. Presentation of the GTW
All reachable positions of the end-effector constitute the workspace of a PM when the theoretical driven angles vary from minimum to maximum. The denition of theoretical driven angles usually does not consider singularity and geometric interference in
practical applications. Once a PM has been assembled, it only works in one continuous workspace without any singularity and interference. To strengthen the working ability of a PM, the largest continuous workspace generally is selected, namely, the usable
workspace (UW). With respect to this 2-DOF 3-parallelogram PPM, singularity and possible geometric interference are considered
when we dene the driving space according to its application in the spray-painting robot. Thus, the UW is the workspace under
the driven space: 1 (90, 180) and 2 (0, 90).
Although the UW dened in the driven space is the largest continuous workspace without singularity, it is bounded by singular
loci. The end-effector can reach every point in the UW, but the transmissibility of the manipulator becomes poor when the endeffector is near the singular loci. To avoid this phenomenon, the GTW is dened. The transmission angle is widely used to evaluate
the transmissibility of a link mechanism. If the transmission angle is near 90, then the end-effector is far from the singular loci,
and the mechanism works at high efciency. If the transmission angle is near 0 or 180, the end-effector is near the singular loci,
and the transmissibility of the link mechanism becomes poor. For high-speed and high-quality motion transmission, the most widely
accepted range of appropriate transmission angle is [45, 135] [25]. For the 2-DOF 3-parallelogram PPM in the spray-painting robot,
as the end-effector is a light spray gun that works at a slow speed, the range [45, 135] are too conservative for the spray-painting
task. Thus, we can reasonably widen the range of the acceptable transmission angle. However, no uniform standard regarding the appropriate transmission angle for the spray-painting robots yet exists up to now. To our knowledge, the range [20, 160] is widely accepted and used by engineers in the spray-painting industry. For the application of the 2-DOF 3-parallelogram PPM in the spraypainting robot, the range [20, 160] can be the acceptable transmission angle. Thus, the GTW can be dened as a workspace in
which all transmission angles involved belong to [20, 160]. The GTW that can keep the end-effector at a certain distance from the
singular loci is a subspace of the UW.
According to Fig. 2, the position of the end-effector is determined by parallelogram mechanism ODEC. To analyze the GTW of the
2-DOF 3-parallelogram PPM, all transmission angles in the parallelogram mechanism ODEC are observed. 1 can be regarded as the
direct transmission angle, and 1 can be regarded as the inverse transmission angle [26]. To make this manipulator work in the GTW,
1 and 1 should belong to [20, 160]. According to Fig. 2, the relationship of 1, 2, 1, and 1 can be expressed by
1 2 1 ;

13

1 1802 1 :

14

Then, we can obtain

20 2 1 160 :

15

Thus, when the two driven angles satisfy condition (15), the 2-DOF 3-parallelogram PPM works in the GTW.
Numerous properties of the GTW are associated with the performance of the 2-DOF 3-parallelogram PPM. The appropriate transmission angle is only the condition to dene the GTW. To investigate the GTW fully, non-dimensional parameters (s and t) that determine the area and length of the GTW must be considered. According to Eq. (10) and condition (15), the GTWs of all possible

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

manipulators can be illustrated in Fig. 5, which considers all possible combinations of s and t (r1, r2, and r3, shown in the 12 regions in
Fig. 4).
In Fig. 5, each point in triangle UVW represents a denite 2-DOF 3-parallelogram PPM. Each of the 12 charts around triangle UVW
represents a set of all GTWs from all possible manipulators within the same region. As we divide the length design space into 12 regions
according to the relationships of the length parameters, the GTWs in the same region have similar shapes. Thus, the sum of all the GTWs
in the same region can reect the common properties of manipulators in this region. The GTW of a manipulator indicates the shape and
area of the workspace, so several qualitative suggestions based on the workspace can be obtained from Fig. 5 when we select the values
of s and t (r1, r2, and r3).
On one hand, the area of the workspace can reect the working ability and exibility of a mechanism. In detail, a larger workspace
indicates better working ability and exibility, thus we always prefer a larger GTW to a smaller one. Obviously, if the best working
ability and exibility of the manipulator are required, we should choose the values of s and t in region 1 or 11 that possesses the largest
GTW.
On the other hand, for the GTW of the 2-DOF 3-parallelogam PPM in a spray-painting robot, a longer length in the X direction and
more symmetrically distribution along axis X indicate a larger painting region with fewer heaving motion in the Y direction. Thus, in
order to ensure that the spray-painting robot own the largest painting region and the most stable working performance we should
choose the values of s and t in region 8 or 9 that possesses both the longer length in the X direction and the more symmetrically distribution along axis X.
Besides, as a denite manipulator works only in a single continuous workspace, and too many discontinuous areas are distributed
in region 2, we usually do not select the parameters in region 2. Therefore, regions 1, 8, 9, and 11 can be regarded as the qualitative
optimal regions in which the optimum parameters can be obtained. Moreover, these four qualitative optimal regions can be picked
out from the design space and regarded as the boundaries for quantitative optimal design such that the optimization efciency can
be improved signicantly.

Fig. 5. The GTWs of all possible manipulators.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

5. Performance indices
When we identify a 2-DOF 3-parallelogram PPM, we expect that the identied manipulator exhibits the best performance. For example, the manipulator has the largest area of the GTW, thus reecting its working ability and exibility; the largest horizontal length
of the GTW, thus indicating the largest painting region for the spray-painting robot; the relatively symmetrical distribution of the
GTW along axis X, thus indicating the least heaving motion of the spray-painting robot; the highest transmission efciency; and
the most stable transmissibility. However, the aforementioned performances always contradict each other in the design space. To
solve contradiction problems among different performance requirements, performance indices are proposed. The performance indices within the same design space can then be illustrated in one performance atlas, such that the intersection region of these performance indices can be selected. The intersection region that can satisfy each performance index efciently is the optimal parameter
region of the manipulator.
For the 2-DOF 3-parallelogram PPM in this study, both performances based on desirable workspace and transmissibility should
be given attention. According to the analysis in Section 3, the workspace and working posture are uncoupled in the design space.
Thus, two kinds of performance indices can be proposed, namely, the performance indices based on desirable workspace and
transmissibility.
5.1. Performance indices based on desirable workspace
All manipulators illustrated in triangle UVW shown in Fig. 5 work in their GTWs. However, the GTW of each manipulator is different. Although several qualitative suggestions can be obtained when the 12 regions are compared with each other, determining
which point in the same region is better than the others remains infeasible. Thus, new indices based on the property of the GTW
are necessary for selecting the optimum length parameters.
5.1.1. Area of the GTW (AGTW)
The area of the workspace can reect the working ability and exibility of a mechanism. In a non-dimensional design space, a 2DOF 3-parallelogram PPM with a larger AGTW is better than that with a smaller AGTW. Based on Eq. (10), the denition of the AGTW
can be specically given as
Z Z
AGTW

xp yp d1 d2 ;

16

where xp and yp should observe Eq. (10). Meanwhile, 1 and 2 should follow condition (15). The AGTWs of all possible manipulators
in triangle UVW can then be illustrated in one atlas. Several representative values of the AGTWs are shown in Fig. 6.
Based on Figs. 4 and 6, the AGTW curves in triangle UVW are symmetrically distributed along line r1 = r3. Meanwhile, as r2 increases, the value of the AGTW decreases. Thus, if the largest AGTW is required, the value of r2 should be as small as possible, and the
values of r1 and r3 should be as close as possible.
5.1.2. Length of the GTW along axis X (LGTW)
In application of the 2-DOF 3-parallelogram PPM in a spray-painting robot, the length of the workspace requires much concern. A
long working trace in the X direction indicates a large painting region. To avoid too much heaving motion, the painting trace should

Fig. 6. The AGTW of the 2-DOF 3-parallelogram PPM.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

work efciently along axis X. Thus, the LGTW is proposed. The LGTW is dened as the length of the GTW along axis X in the coordinate
system XOY. The denition of the LGTW can be given specically as follows:
LGTW jxp

max xp min jyp 0 :

17

The LGTWs of all possible manipulators in triangle UVW can then be illustrated in one atlas. Several representative values of the
LGTWs are shown in Fig. 7.
According to Figs. 4 and 7, the LGTW curves in triangle UVW are mainly distributed on the left side of line r1 = r2. Meanwhile, as
r1 increases, the value of the LGTW also increases. To obtain a large painting region for the spray-painting robot, r1 should rstly be
larger than r2 and the value of r1 should secondly be as large as possible.
5.2. Performance indices of transmissibility
Aside from its planar pointing function, the 2-DOF 3-parallelogram PPM can also keep the end-effector in a denite posture within
the entire workspace. During the working process, parallelogram mechanism ODEC changes its status to reach different positions,
whereas parallelogram mechanisms OABC and CFGH change theirs to keep the end-effector in a denite posture. Therefore, when
we investigate the transmissibility of the manipulator, all transmission angles in the aforementioned three parallelogram mechanisms
should be considered.
5.2.1. Denition of the comprehensive GTW (CGTW)
The denition of the GTW only concerns the transmission angles involved to determine the position of the end-effector. However,
the transmission angles associated with the posture of the end-effector are also related to the transmissibility of the manipulator. In
detail, 2 and 3 can be regarded as the direct transmission angles of parallelogram mechanisms OABC and CFGH. 2 and 3 can be
regarded as the inverse transmission angles of parallelogram mechanisms OABC and CFGH [26]. The CGTW is dened as the workspace
that considers all these transmission angles (i and i) that belong to [20, 160]. According to Eq. (10), the CGTW and the GTW dene
the same workspace. The difference is that the CGTW considers all transmission angles associated with both the working position and
the working posture, whereas the GTW only concerns the working position of the manipulator. Thus, the CGTW is an extension of the
GTW, and it enables the entire mechanism to maintain good transmissibility during the working process.
5.2.2. The LTI
To quantitatively evaluate the transmissibility of the 2-DOF 3-parallelogram PPM in a denite position, the LTI is proposed based
on the CGTW. The LTI is dened as
LTI minf sin 1 ; sin2 ; sin3 ; sin1 ; sin2 ; sin3 g:

18

Then, the maximum and minimum LTI can be obtained as follows:

LTImax sin90 1;
LTImin

19


sin20 or sin160 0:342:

20

Fig. 7. The LGTW of the 2-DOF 3-parallelogram PPM.

10

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

When the LTI is near the LTImax, all transmission angles are near 90. The 2-DOF 3-parallelogram PPM exhibits superior transmissibility. By contrast, when the LTI is near the LTImin, at least one transmission angle is near 20 or 160. The 2-DOF 3-parallelogram PPM
then exhibits poor transmissibility. The LTI is suitable to evaluate the transmissibility of the 2-DOF 3-parallelogram PPM in a denite
position and posture.
5.2.3. The GTI
A set of positions are found in the CGTW. Calculating the LTIs of all possible positions in the CGTW and then judging their transmissibility are impractical. To evaluate the transmissibility of the 2-DOF 3-parallelogram PPM conveniently, the GTI is proposed.
The GTI is dened as the average transmissibility of the 2-DOF 3-parallelogram PPM in the CGTW. As the GTI concerns all the transmission angles in all possible positions in the CGTW, it is quite a comprehensive index to evaluate the transmissibility of the manipulator. The specic denition of the GTI is given by
Z
GTI

W
Z

LTIdW
;

21

dW
W

where W is the area of the CGTW, and the LTI is the local transmissibility index of the manipulator in a denite position and posture.
Given that the value of the LTI ranges from the LTImin to LTImax, the GTI also ranges from the LTImin to LTImax. A large GTI indicates that
the LTIs of the manipulator in all possible positions are relatively large, which indicates a good average transmission level for this
manipulator.
For the 2-DOF 3-parallelogram PPM in this study, the length design space and the angular design space are uncoupled. Once the
desirable workspace is determined by length parameters, the transmissibility of this manipulator is only associated with the angular
design space. Namely, and determine the GTI of the manipulator. According to conditions (11) and (12), each point in the angular
design space indicates a denite 2-DOF 3-parallelogram PPM. Numerous combinations of the angular parameters ( and ) are available in the angular design space, thus leading to numerous manipulators with various GTIs. The GTIs of all possible manipulators can
then be illustrated in one atlas. Several representative values of the GTIs are shown in Fig. 8.
According to Fig. 8, the GTI curves have monotonically increasing features toward the common center of the curves, thus indicating
that the center exhibits better transmissibility than the edges. To obtain the best transmissibility for the 2-DOF 3-parallelogram PPM,
we should select the values of and within the innermost curve with the value of 0.75.

5.2.4. The uctuant transmission index (FTI)

The GTI can well reect the average transmissibility that is important for selecting an optimal manipulator. However, we are usually confused with the circumstance that there can be several options that possess the same large GTI. In fact, it is not wisdom to conclude that all manipulators with the same large GTI are optimal for some differences indeed exist among these manipulators. For
example, there are two manipulators and both possess the same GTI as 0.8. However, the LTI of the rst manipulator varies stably
in the CGTW while the LTI of the second manipulator varies dramatically in the CGTW. Obviously, the transmissibility of these two
manipulators is quite different, and the rst one is more reliable in practical application. Thus, when we identify a manipulator, we

Fig. 8. The GTI of the 2-DOF 3-parallelogram PPM in the angular design space.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

11

generally aim to obtain more information on the transmissibility than the GTI, particularly the uctuation of the LTI around the GTI,
which is vital for designing driving motors. Thus, the FTI is proposed. The FTI is dened specically as
Z
FTI

0Z
12
2
LTI dW B
LTIdW C
WZ
W
C ;
B
@ Z
A
dW
dW
W

22

where W is the area of the CGTW, and the LTI is the local transmissibility index of the 2-DOF 3-parallelogram PPM in a denite position
and posture. The FTI can be regarded as the variance of the LTI that reects the uctuation of the LTI around the GTI. A small FTI indicates that the LTI of the manipulator varies stably in the CGTW, thus, a stable working performance can be obtained. On the contrary,
a large FTI indicates a dramatic variation working condition that should be avoided.
For the manipulator in this study, numerous combinations of and are available, thus leading to different manipulators with various FTIs. The FTIs of all possible manipulators can then be illustrated in one atlas. Several representative values of the FTIs are shown
in Fig. 9.
According to Fig. 9, three concentrated center regions are found for all curves. The FTI has a larger value than 0.04 within the upper
and lower curves with a value of 0.04. The FTI has a smaller value than 0.025 within the middle curve with a value of 0.025. Although
the FTI curves at the corners of the angular design space have small values, the GTIs of these corner curves are too small according to
Fig. 8. Thus, if a small FTI is required under the condition of a large GTI, only one usable region from which the optimal angles can be
selected is available, namely, the region within the middle curve with a value of 0.025.
5.3. Summery of the performance indices
Two kinds of performance indices are proposed in this section, including the performance indices based on the workspace and
transmissibility. Both of them are necessary for the kinematic optimization of the 2-DOF 3-parallelogram PPM, but they are different
in the application range. As the AGTW and LGTW are dened according to the practical application requirements, they are specic to
the 2-DOF 3-parallelogram PPM in a spray painting robot. However, the LTI, GTI and FTI are dened based on the general properties of
the transmission angles, thus they can also be used in the optimization of other PMs. Although the LTI and GTI based on GTW are proposed and used by other researchers [7], the LTI and GTI proposed in this paper considering the CGTW bring more comprehensive
meanings. Furthermore, the FTI proposed in this study has never been used before that reveals more deep transmissibility of a PM.
In conclusion, the FTI, together with the LTI and GTI in this study can make the transmissibility evaluation of a PM more efcient
and make kinematic optimization in transmission more targeted.
6. Optimal kinematic design based on performance atlases
6.1. Requests of kinematic optimal design

As shown in Fig. 10, the spray-painting robots are always assembled on each side of a vehicle in a production line. Thus, one spraypainting robot only needs to paint half of the surface of the vehicle. Fig. 11 shows a vehicle that must be painted. The largest continuous painting area is half of the top region on the vehicle. This area can be regarded as a rectangle with a length of 2000 mm (BD) and a

Fig. 9. The FTI of the 2-DOF 3-parallelogram PPM in the angular design space.

12

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

Fig. 10. Layout of the spray-painting robots.

width of 900 mm (AE). If the spray-painting robot can nish painting this rectangular area, then it can certainly complete the painting
task for the vehicle.
In Fig. 11, point O is the center of arc BCD and is the original point of the 2-DOF 3-parallelogram PPM where it has been assembled.
The line segment OA 300 mm is the radius of the dead zone for the manipulator. Arc BCD is the smallest painting arc that can cover
the rectangular painting area. Therefore, the minimum painting radius for the spray-painting robot is line segment AC, which can be
given by
AC OCOA OBOA:

23

Meanwhile, line segment OB in triangle OBE can be expressed as follows:

OB

q q

2
2
2
2
OE BE OE BD=2 :

24

According to Eqs. (23) and (24), we can obtain

AC 1262 mm:

25

Fig. 11. The largest painting region on the vehicle.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

13

Thus, to nish the painting task, the radius of the painting trace should not be smaller than 1262 mm; that is, the practical value of
the LGTW, i.e., PLGTW, should satisfy the condition
P LGTW 1262 mm:

26

In the practical application of a spray-painting robot, the spray gun includes an electrostatic spinning cup, thus no metallic object
should be present within the range of 300 mm. Therefore, in the 2-DOF 3-parallelogram PPM shown in Fig. 2, R2 should satisfy the
condition
R2 300 mm:

27

The requirements of the optimal kinematic design are to nd the optimum geometric parameters that will provide the manipulator with good exibility, high transmission efciency, and stable transmissibility, which should satisfy conditions (26) and (27).
6.2. Determining the length parameters
To obtain the best working exibility and the largest painting region, the AGTW and the LGTW should have the largest values. The
intersecting region of the AGTW and the LGTW curves with the largest values is presented as the black shaded region in Fig. 12.
Theoretically, the value of r2 should be as small as possible to obtain the largest values of the AGTW and the LGTW. However, r2 in
the non-dimensional design space should coincide with condition (27) in the practical design space. Meanwhile, the relationship between the practical values of the parameters (R1, R2, R3, pAGTW, and pLGTW) and the non-dimensional parameters (r1, r2, r3, vAGTW, and
vLGTW) can be expressed by the following equations [10]:
R1 Dr1 ;

R2 Dr2 ;

R3 Dr3 ;

28

pAGTW D vAGTW ;

29

pLGTW DvLGTW :

30

Based on Fig. 12 and Eq. (28), R1, R2, and R3 can be determined as follows. First, the intersection point formed by the AGTW curve
with a value of 4 (vAGTW = 4) and the LGTW curve with a value of 2.5 (vLGTW = 2.5) is identied in the black shaded region. The coordinates (s and t) of the intersection point are determined according to Fig. 12. The value of r2 is then calculated according to
Eqs. (8) and (9). Second, the value of non-dimensional factor D is calculated according to equation D AC=vLGTW . Third, the value
of R2 is calculated according to Eq. (28) and an evaluation is made. If R2 satises condition (27), then we can calculate the values of
r1 and r3 according to Eqs. (8) and (9), determine the values of R1 and R3 by Eq. (28), and obtain PAGTW via Eq. (29). The optimal process
is then nished. If R2 does not satisfy condition (27), then we should abandon this intersection point and nd another one that is
formed by a smaller LGTW and the same AGTW used in the former process. Steps one to three are repeated to calculate the value
of R2, and the same judgment as in step three is made. If the smallest value of the LGTW, i.e., vLGTW = 0.3, is used but no optimal intersection point is found, then we should reduce the value of the AGTW and repeat the processes until the optimal intersection point is
found.

Fig. 12. The intersection of the AGTW and the LGTW.

14

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

According to the qualitative analysis in Fig. 5, the optimum length parameters should be found in the qualitative optimal regions,
i.e., regions 1, 8, 9, or 11. The intersection point of the LGTW and the AGTW curves will exceed the qualitative optimal regions when
vAGTW b 1.6. If no optimal intersection point can be found under conditions vLGTW N 0.3 and vAGTW N 1.6, we consider the optimization
to be unsuccessful. The process to determine the optimum parameters R1, R2, and R3 is illustrated in Fig. 13.
Based on the performance atlas in Fig. 12 and the optimization owchart in Fig. 13, we can observe that at the intersection point of
the LGTW curve with a value of 2.1 and the AGTW with a value of 2.8, the value of R2 satises condition (27). The value of nondimensional factor D can then be obtained as
D AC=pLGTW 1262=2:1 600:95:
Meanwhile, the coordinates of this intersection point can be obtained from Fig. 12 as
s 1:085; t 0:879:
Thus, according to Eqs. (6), (8) and (9), we can obtain
r 1 1:621; r 2 0:5 and r 3 0:879:
According to Eq. (28), we can then obtain
R1 974:14 mm; R2 300:475 mm; and R3 528:24 mm:
Based on Fig. 12 and the non-dimensional factor (D = 600.95), we can obtain the practical value of the AGTW as
2

pAGTW D vAGTW 600:95  2:8 1011194:5 mm :

6.3. Determining the angular parameters

The GTI and the FTI are important for the 2-DOF 3-parallelogram PPM if high transmission efciency and stable transmissibility are
required. A large GTI indicates high transmission efciency. A small FTI indicates stable transmission performance. Therefore, we

Start:vAGTW =4 and vLGTW =2.5

Find r2 of the intersection point

vLGTW =2.5

AC / vLGTW
R2

Reduce vAGTW

Reduce vLGTW

Dr2

Yes

Yes

R2

300 mm

No

vLGTW >0.3

No

Yes
Calculate R1 and R3
Calculate

pAGTW

End
Fig. 13. Flowchart for determining the length parameters.

vAGTW >1.6
No
Failure

15

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

Fig. 14. The intersection of the GTI and the FTI of the 2-DOF 3-parallelogram PPM.

should rst nd out the region with the largest GTI and then the subregion with the smallest FTI. This subregion is the optimum angular parameter region wherein the optimum angular parameters can be selected, as shown in the shaded region in Fig. 14.
According to Fig. 14, the center of the shaded region should be selected as the optimal point with the largest GTI and the smallest
FTI, namely, = 90 and = 45.

6.4. Establishing a 2-DOF 3-parallelogram PPM

The optimum length parameters, i.e., R1 = 974.14 mm, R2 = 300.475 mm, and R3 = 528.24 mm, are obtained. The optimum angular parameters, i.e., = 45 and = 90, are also identied by the performance atlas in Fig. 14. All optimum values of the geometric
parameters in the design space are determined. However, other geometric parameters not mentioned earlier are still necessary to
build the 2-DOF 3-parallelogram PPM. These geometric parameters include the lengths of links OA, OD, HI (R), and GH. These parameters do not affect the workspace and transmissibility of the manipulator and should be determined based on practical circumstance
or the experience of the designer. An optimum 2-DOF 3-parallelogram PPM is established by the authors, as shown in Fig. 15.
Based on the optimum geometric parameters in Fig. 15, the optimum workspace and transmissibility of the optimum 2-DOF 3parallelogram PPM is illustrated in Fig. 16.
Furthermore, based on Fig. 14, we can determine that the GTI of the 2-DOF 3-parallelogram PPM is greater than 0.75, and that the
FTI of the manipulator is lower than 0.025.

Fig. 15. An optimum 2-DOF 3-parallelogram PPM.

16

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

Fig. 16. Workspace and transmissibility of the optimum 2-DOF 3-parallelogram PPM.

7. Conclusions
The kinematic optimal design of a 2-DOF 3-parallelogram PPM is addressed in this study by considering desirable workspace and
transmissibility. The kinematics of common working mode for the 2-DOF 3-parallelogram PPM is conducted, and the capable solution
is selected for the working pattern. The design space of the manipulator is analyzed and divided into the uncoupled length design
space and the angular design space. The GTW is dened by considering the transmission angles involved to determine the position
of the end-effector. All possible manipulators are qualitatively investigated by considering their GTWs, and four qualitative optimal
regions are obtained. Based on the GTW, the LGTW is proposed to evaluate the painting region of the spray-painting robot. The
AGTW is also proposed to evaluate the working exibility. The CGTW is dened by considering all the transmission angles associated
with both working position and working posture. Based on the CGTW, the LTI is proposed to investigate the transmissibility of the
manipulator in a denite position. Based on the LTI, the GTI is introduced to reect the average transmissibility in the GTW. The FTI
is presented to describe the dynamic wave of the LTI around the GTI. Based on the performance atlas of the LGTW and the AGTW,
the optimum length parameters are obtained as R1 = 974.14 mm, R2 = 300.475 mm and R3 = 528.24 mm. According to the performance atlases of the GTI and the FTI, the optimum angular parameters are obtained as = 45 and = 90. Finally, an optimum
2-DOF 3-parallelogram PPM is identied. The optimum workspace and transmissibility of this manipulator are described. Although
the performance indices based on the workspace are specic to the 2-DOF 3-parallelogram PPM, the performance indices based on
the transmissibility and the optimal design method used in this study are general and can also be applied to other PPMs.
Acknowledgments
This project is supported by the National Natural Science Foundation of China (NSFC) (Grant Nos. 51425501, 51375251) and the
National Basic Research Program (973 Program) of China under Grant No. 2013CB035400.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.mechmachtheory.2014.12.014.
References
[1] D. Zeng, Z. Huang, W. Lu, Performance analysis and optimal design of a 3-DOF 3-PRUR parallel mechanism, ASME J. Mech. Des. 130 (2008) 042307.
[2] C. Gosselin, J. Angeles, A global performance index for the kinematic optimization of robotic manipulators, ASME J. Mech. Des. 113 (1991) 220226.

X.-J. Liu et al. / Mechanism and Machine Theory 87 (2015) 117

17

[3] R. Kurtz, V. Hayward, Multiple-goal kinematic optimization of a parallel spherical mechanism with actuator redundancy, IEEE Trans. Robot. Autom. 8 (1992)
644651.
[4] K.E. Zanganeh, J. Angeles, Kinematic isotropy and the optimum design of parallel manipulators, Int. J. Robot. Res. 16 (1997) 185197.
[5] L.W. Tsai, S. Joshi, Kinematics and Optimization of a Spatial 3-UPU Parallel Manipulator, ASME J. Mech. Des. 122 (2000) 439446.
[6] X.J. Liu, Optimal kinematic design of a three translational DoFs parallel manipulator, Robotica 24 (2006) 239250.
[7] C. Wu, X.J. Liu, L.P. Wang, Optimal design of spherical 5R parallel manipulators considering the motion/force transmissibility, J. Mech. Des. 132 (2010) 031002.
[8] F.G. Xie, X.J. Liu, L.P. Wang, Optimum kinematic design of the 4R 2-DOF parallel mechanism, Tsinghua Sci. Technol. 14 (2009) 663668.
[9] X.G. Li, J.H. Liu, Research and application situation, problems and solutions of palletizing robots, Packag. Eng. 32 (2011) 96102.
[10] F. Gao, X.Q. Zhang, Y.S. Zhao, et al., Physical model of the solution space and the atlas of the reachable workspace for 2-DOF parallel planar manipulators,
Pergamon 31 (1996) 173184.
[11] T. Huang, M. Li, Z. Li, et al., Optimal kinematic design of 2-DOF parallel manipulators with well-shaped workspace bounded by a specified conditioning index, IEEE
Trans. Rob. Autom. 20 (2004) 538543.
[12] X.J. Liu, J.S. Wang, G. Pritschow, Performance atlases and optimum design of planar 5R symmetrical parallel mechanisms, Mech. Mach. Theory 41 (2006)
119144.
[13] J.Q. Li, S.H. Ding, B.L. Duan, et al., Dynamic analysis and structural optimization of a novel palletizing robot, J. Beijing Inst. Technol. 19 (2010) 274278.
[14] L. Nan, H.Q. Zhang, S.G. Tong, Optimum motion control of palletizing robots based on iterative learning, Ind. Robot. Int. J. 39 (2012) 162168.
[15] L.G. Zhang, J.P. Mei, X.M. Zhao, et al., Layout analysis and path planning of a robot palletizing production line, Proceedings of the IEEE International Conference on
Automation and Logistics, Qingdao, China, September 2008, 2008, pp. 24202425.
[16] M. Stock, K. Miller, Optimal kinematic design of spatial parallel manipulators: application to linear delta robot, J. Mech. Des. 125 (2003) 292301.
[17] J. Ryu, J. Cha, Volumetric error analysis and architecture optimization for accuracy of HexaSlide type parallel manipulators, Mech. Mach. Theory 38 (2003)
227240.
[18] D. Chablat, P.H. Wenger, Architecture optimization of a 3-DoF parallel mechanism for machining applications: the Orthoglide, IEEE Trans. Robot. Autom. 19
(2003) 403410.
[19] X.J. Liu, J.S. Wang, A new methodology for optimal kinematic design of parallel mechanisms, Mech. Mach. Theory 42 (2007) 12101224.
[20] X.J. Liu, J.S. Wang, G. Pritschow, On the optimal kinematic design of the PRRRP 2-DoF parallel mechanism, Mech. Mach. Theory 41 (2006) 11111130.
[21] C. Gosselin, J. Angeles, The optimum kinematic design of a spherical three-degree-of-freedom parallel manipulator, ASME J. Mech. Transm. Autom. Des. 111
(1989) 202207.
[22] J.P. Merlet, Jacobian, manipulability, condition number, and accuracy of parallel robots, ASME J. Mech. Des. 128 (2006) 199206.
[23] F. Gao, X.J. Liu, W.A. Gruver, Performance evaluation of two-degree-of-freedom planar parallel robots, Mech. Mach. Theory 33 (1998) 661668.
[24] X.J. Liu, The relationships between the performance criteria and link lengths of the parallel manipulators and their design theory(Ph.D. Thesis) Yanshan University, Qinhuangdao, China, 1999.
[25] D.C. Tao, Applied Linkage Synthesis, Addison-Wesley, Reading, MA, 1964. 712.
[26] J.S. Wang, X.J. Liu, C. Wu, Optimal design of a new spatial 3-DOF parallel robot with respect to a frame-free index, Sci. China Ser. E Technol. Sci. 52 (2009)
986999.