Advances in Intelligent Systems Research, volume 148

International Conference on Information Technology and Management Engineering (ICITME 2018)

Algorithm and Application of Inverse Kinematics for

6-DOF Welding Robot Based on Screw Theory
Lei Hong1, Baosheng Wang2, Zhenqin Xu2 and Dongsheng Lv2
1
School of Automotive and Rail Transit, Nanjing Institute of Technology, Nanjing, China
2
Research Department of Intelligent Manufacturing Equipment, Nanjing Institute of Technology, Nanjing, China

Abstract—In order to achieve efficient and reliable inverse the axes of the 2 and 3 joints are parallel, and both are
kinematics for industrial robot, an inverse kinematics solution is perpendicular to the plane of the 1st joint axis. The structural
proposed for common 6 DOF industrial robots with the structure characteristics of the first three joints make the Paden-Kahan
characteristics that can be described as “the last three joints axes sub-problem can be solved directly. Aiming at the problem of
are intersect, joint 2 and 3 are parallel and non co-planar but solving the first three joint angle, Qian[4] used the classical
perpendicular to joint 1”.Taking the screw kinematics model, the elimination method to solve the problem, but this method has a
solution is established by using geometry relation method and large amount of calculation. Chen[5] extended it to a new sub-
basic known Paden-Kahan sub-problems. Firstly, the 1st joint problem, but the derivation process is more complex.
angle is solved by the geometrical relationship between the robot
wrist point and first three joint axes, and the rest five joint angles Based on the screw theory, this paper presents a new
are solved by transforming into three Panden-Kahan sub- method for solving the inverse kinematics of 6-DOF industry
problem 1 and one sub-problem 2, so that the complete closed welding robot, combining the geometric relation method with
form solutions for inverse kinematics are provided. The accuracy Paden-Kahan subproblem. This method decomposes the whole
and practicality of the algorithm are testified by the results in V- inverse problem into three steps: First, Solve joint angle 1 by
groove welding experiment. using geometric relation method. The plane constituted by axis
of joint 1 and point of wrist, which is always perpendicular to
Keywords—industry robot; screw theory; inverse kinematics;
the joints of 2 and 3, Thus the line connected between any
Paden-Kahan sub-problem
point on the axis of the joint 1 and the wrist point, satisfies the
vertical geometric relationship with the joints of 2 and 3, and
I. INTRODUCTION the joint angle 1 can be solved by this feature. Second, the
Robot inverse kinematics is to calculate the angles of each inverse solutions of joint angles of 2 and 3 are transformed into
joint of the robot based on the desired posture of the robot end- two Paden-Kahan sub-problems by using the parallel
effector, the inverse solution is the premise and basis for the geometrical features of joint 2 and 3 axes. Third, By the
robot trajectory planning and motion control. The first step of geometric feature that 4,5 and 6 joint axes intersect at the wrist
inverse kinematics is to establish a kinematic model of the point, and the joint angles of 4 and 5 are solved by
robot. At present, the kinematics model of the robot is mainly transforming into a Panden-Kahan sub-problem 2, and solution
composed of the Denavit-Hartenberg(D-H) parameter method of joint angle 6 is transformed into sub-problem [Link] the above
and the screw theory method. However, D-H parameter method method, the inverse kinematics closed solution is given through
is complex to establish the local coordinate system and requires the geometric characteristics of 6-DOF industry robot and
observation and experience to establish a constraint equation known Paden-Kahan sub-problem. The experiment is
with joint variables, and the analytical expression derivation performed at last to show the accuracy of this algorithm.
process is complicated. In screw method, the inverse problem
is solved by the Paden-Kahan sub problem, and avoids the II. MODELING AND ALGORITHM
singularity brought by the used of the DH coordinate method.
Sariyildiz[1] uses quaternions and known Paden-Kahan sub- A. Robot Kinematics Modeling
problems to solve the common structure of the robot
As shown in Figure.1, W o0  x0 y0 z0 and E o6  x6 y6 z6 are
kinematics inverse solution. Tan[2] applied the expanded
Paden-Kahan sub-problem to solve the 5-DOF robot set as the base and the end coordinate system respectively.
kinematics inverse solution with special structure. Chen For welding robot, the welding torch should be installed at
Weihai[3] solved the 7-DOF robot inverse kinematics problem the end, T oT  xT yT zT  is set as the welding torch tool
with Paden-Kahan sub-problems and velocity level inverse
solution algorithm. The disadvantage of the Paden-Kahan sub- coordinate system. In general, It is determined of the
problem is that the axes of all joints of the robot must have transformation matrix ETT , which is the coordinate system T
enough intersections. In particular, the first two joint axes are with respect to coordinate system E, and it can be precisely
intersected, which limits the generality of the algorithm. calibrated by the robot TCP calibration algorithm[6]. When the
desired position and pose of the welding torch is given, the
The typical structural feature of 6-DOF welding industrial
robot is that, the first 3 joints do not intersect each other, and transformation matrix WET which set as the end of the robot
relative to the base coordinate system can be expressed as,

Copyright © 2018, the Authors. Published by Atlantis Press.

This is an open access article under the CC BY-NC license ([Link] 159
 Advances in Intelligent Systems Research, volume 148

W
T  WTT ( ETT )1  n
gwe (θ )   eii gwe (0) 
E ˆ

i 1
By formula(1), the problem of inverse kinematics of robot
with a torch can be translated into that without welding torch,
namely coordinate E as the robot end for inverse solution. B. Robot Inverse Kinematics Algorithm
Therefore, the inverse problem of the robot without torch is As shown in Figure.2, suppose ξ1 , ξ 2 and ξ 3 as three zero
studied as follows, and according to the representation of
pitch, mutually disjoint unit motion screw, in which ξ 2 is
screw theory, WET is recorded as gwe ( ) .As shown in Figure
parallel to ξ 3 , ξ 2 and ξ 3 are all different plane with ξ1 , which
[Link] robot configuration with each joint angle zero is defined
as the reference configuration. The screw coordinate system is in the plane M rotated by ξ 2 and ξ 3 , In the initial
set up in this configuration. configuration, Point pw4 rotates  3 to pw3 around the axis of
d4 d6 ξ 3 , then rotates  2 to pw2 around the axis of ξ 2 , finally,
y6 6
rotates 1 to pw1 around the axis of ξ1 .
4 4
zT
a3 3  5  5  z6
x6
6
yT ξ2 ξ3
3 xT pw2 pw3 pw4
2 3
{E} {T }
a2
E ξ1
1
2 T T p1 M
1 2
pw1
1
FIGURE II. INVERSE KINEMATICS SUB-PROBLEM FOR THE FIRST
W
ET ( gwe ) THREE JOINTS
d1
W
z0 y0 T T Then the above sub-problem satisfies the following screw
motion formula:
x0
{W } a1 ˆ ˆ ˆ
eξ11 eξ22 eξ33 pw4  pw1 
FIGURE I. SCREW COORDINATE FRAMES OF 6-DOF ROBOT

gwe (0) is denoted as the transformation matrix between When pw1 , pw4 are given, 1 ,  2 and  3 can be solved.

1) Solve the joint angle 1

coordinate system E and W , which is,
The last three joints of 4, 5, and 6 intersect at one point,
0 0 1 a1  d 4  d 6  namely the robot wrist center point, set as pw4 , then:
0 1 0 0 
gwe (0)    
 1 0 0 a2  a3  d1  pw4  p5 = {a1  d4 ,0, d1  a2  a3 }
 
0 0 0 1 
By the Formula (3), it is can be obtained that,
The screw coordinates of each joint are:
ˆ ˆ ˆ
ξ1  [0,0,0,0,0,1] , e11 e22 e66  gst ( ) gst (0)1  g1 (5)
ξ2  [d1 ,0, a1 ,0,0,1] ,
Both sides of Formula(5) are right multiplied by pw4 ,
ξ3  [a2  d1 ,0, a1 ,0,1,0] ,
which locates on the axis of ξ 4 , ξ 5 and ξ 6 , According to the
ξ4  [0, a2  a3  d1 ,0,1,0,0] ,
principle of rigid body movement that “the position remains
ξ5  [a2  a3  d1 ,0, a1  d4 ,0,1,0] , unchanged when the point on the revolution axis”, then
ξ6  [0, a2  a3  d1 ,0,1,0,0] .
According to the exponential product formula of forward ˆ ˆ ˆ
eξ11 eξ22 eξ33 pw4  g1 pw4  pw1 (6)
kinematics for robot of n DOF, forward kinematics solution of
robot is denoted as gwe (θ) , which is,

160
 Advances in Intelligent Systems Research, volume 148

As shown in Figure.2, for ξ1 and pw4 are all located in the The above can be described as inverse kinematic sub-
Plane M , then the line p1 pw4 is perpendicular to ξ 2 . As problem of two parallel joints. pw4 rotates  3 to pw3 around
shown in Figure.3, suppose robot rotates angle 1 around ξ1 . ξ 3 , then rotates  2 to pw2 around ξ 2 , if given pw2 and pw4 ,
ˆ ˆ
from formula eξ22 eξ33 pw4  pw2 ,  2 and  3 can be solved.
pw2 First solve pw2 according to the its screw motion, then,

ξ1 1
M ˆ
pw2  eξ11  pw1 (10)
p1
pw1
ξ 2 If pw3 is solved, the following formula can be given,

FIGURE III. THE SCREW COORDINATE FRAMES OF THE FIRST TWO

1   ξˆ 
e 2 2  pw2  pw3
JOINTS AFTER ROTATION OF  ξˆ  (11)
 e  pw4  pw3
 3 3

Then pw4 is rotated to pw1 through pw2 and pw3 , ξ 2 is

rotated to ξ 2 , and Plane M is become to M' , position of p1 Equation(11) is actually two Panden-Kahan sub-problem 1,
remains constant， p1 pw1 is located in Plane M' , therefore, so the key issue is to solve pw3 . It is illustrated by Figure.4,
p1 pw1 is perpendicular to ξ 2 . r2  p2  pw 2 , r3  p3  pw4 , r23  p2  p3 , in pw3 p2 p3 ,
Set ω2 is the corresponding rotation vector for ξ 2 , then Suppose the angle between pw3 p2 and p2 p3 as 1 , from the
ˆ11 cosine theorem, then
ω2  e ω2 , it is calculated by formula of Rodrigues, then

r22  r232  2r2  r23 cos 1  r32 (12)

ω2  [ sin(1 ),cos(1 ),0] (7)
p2 , p3 , pw2 and pw4 are substituted into formula(12), then
pw1 is derived from the formula(6), suppose
pw1  [ pw1x , pw1 y , pw1z ]T , from p1 pw1  ω2 , then
 a 2  a 2  d 2  (q  a ) 2  (q  d ) 2 
1  acos  2 3 4 w2x 1 z 1
 (13)
(ω2 )T   p1  pw1   0  2 a  ( q  a ) 2
 ( q  d ) 2 
(8)  2 w2x 1 z 1 

From formula(8), 1 can be solved by In the reference configuration, plane M within points of
pw2 , pw3 and pw4 is coincides with the xz plane of the base
coordinate system W , suppose pw3  [ pw3x , pw3y , pw3z ]T , then
1 = arctan( pw1x ,  pw1 y ) (9)

2) Solve the joint angle  2 and  3  pw3x = pw2x  r2 cos(1 )

As shown in Figure.2, the joints of 2 and 3 constitute the 
 pw3y = 0 (14)
screw motion of two paralled joints, and the projection of the  p = p  r sin( )
screw motion trajectory of pw4 in the plane M is shown in  w3z w2z 2 1

Figure.3, where p2 and p3 are the centers of rotation. Vectors are defined as follows,
x
pw3 pw4
pw2 u = pw2  p2 , v = pw4  p3
3  (15)
2  p23  p3  p2 , p32  p2  p3
r2 1 2 r3
p2 r23 p3 z By formula(11), then

pw3
  2 = arctan(2 ( p23  u), p23 u)
 T T

 (16)
3 = arctan(3 (v  p32 ), v p32 )

T T
FIGURE IV. SCREW MOTION OF TWO PARALLEL JOINTS

161
 Advances in Intelligent Systems Research, volume 148

3) Solve the joint angle  4 ,  5 and  6 1) Determine the transformation matrix ETT
After solving the above three joint angles, the inverse E
T is calibrated by the robot TCP calibration, which is,
solution of the last 3 joints can be solved by using the known T

Paden-Kahan sub-problem. By formula(5), then

 0.7038 0.0299 0.7098 3.119 
 0.0151 0.9995 0.0271 1.102 
 
TT 
E
ˆ ˆ ˆ ˆ ˆ ˆ
eξ44 eξ55 eξ66  eξ11 eξ22 eξ33 g1  g2 (17)  0.7103 0.0083 0.7039 325.494 
 
 0 0 0 1 
Take point pw 6 on axis of ξ 6 but not on ξ 4 and ξ 5 , Both
2) Determine the position and pose matrix WET
sides of formula(17) are right multiplied by pw 6 , then W
T is determined by the expected planning points along
E
welding seam. Take the starting point of V-groove seam in
ˆ ˆ
eξ44 eξ55 pw6  g2 pw6  q6  [q6 x , q6 y , q6 z ]T (18) bottom welding as example, which is,
 0.6873 0.0400 0.7253 940.044 
By Paden-Kahan sub-problem 2,  4 ,  5 can be solved as  0.0443 0.9989 0.0131 52.026 
W
T   
E
 0.7250 0.2232 0.6883 743.756 
 
 4   arctan((q6 z  a2  a3  d1 ), q6 y )
  0 0 0 1 
 (19) 3) Determine the optimum solution of robot joint angles
5  arctan(q6 x  a1  d 4 , d6  (a1  d 4  q6 x ) )
2 2
 The inverse kinematics algorithm is used to solve the robot
joint angles according to WET , and a total set of eight solutions
By formula(5), then
is obtained, as shown in Table 1.
ˆ ˆ ˆ ˆ ˆ ˆ
eξ66  eξ55 eξ44 eξ11 eξ22 eξ33 g1  g3 (20) TABLE I. EIGHT SOLUTIONS OF INVERSE KINEMATICS

No. 1 / ( ) 2 / ( ) 3 / ( ) 4 / ( ) 5 / ( ) 6 / ( )
Take pt as any point not on the sixth axis, such as pt  p1 ,
1 3.32 126.74 172.67 178.29 -104.10 179.32
Both sides of formula(20) are right multiplied by pt , then 2 3.32 126.74 172.67 -1.71 104.10 -0.68
3 3.32 4.96 26.25 172.28 -12.42 -172.72
4 3.32 4.96 26.25 -7.72 12.42 7.285
ˆ
eξ66 pt  g3 pt  [qtx qty qtz ]T (21) 5 -176.68 -32.48 -152.25 -2.64 -38.83 -178.20
6 -176.68 -32.48 -152.25 177.36 38.83 1.80
7 -176.68 -112.26 -8.83 -1.70 -102.42 179.37
By Paden-Kahan sub-problem 1, we can get 8 -176.68 -112.26 -8.83 178.30 102.42 -0.63
In practical application, an optimal solutions can be
selected by applying the optimal conditions such as the
6  arctan(a2 +a3  d1  qtz , qty ) (22)
shortest path principle. In this example, the fourth group
solution is optimal.
So far, all the closed solutions for 6-DOF inverse
kinematics of robot have been obtained. 4) Offline programming and Experiment
Four layers with ten passes are planned in off-line
programming, which is shown in Figure 5 as follows:
III. ALGORITHM APPLICATION IN WELDING EXPERIMENT
In order to verify the correctness of the inverse kinematics,
a V-groove butt-jointed seam welding experiment is
performed, and robot ABB IRB1410 is taken as welding robot,
As shown in Figure.1, the link parameters of the robot are as
follows(unit: mm):

(a) Bottom welding (b) Cover welding

d1  475 , a1  150 , a2  600 , a3  120 , d4  720 , d6  85 .
FIGURE V. OFF-LINE PROGRAM PLANNING OF WELDING

The workpiece is composed of two Q345 low carbon steel, The cooresponding program is generated and the
the thickness of the plate is 50mm, and the angle of the groove experiment is performed as shown in Figure 6, the results
is , The diameter of the welding wire is 1.0mm, the deposition show that the welding gun can tracking the weld seam
coefficient is 0.9, the wire feeding speed of the backing accurately.
welding is 2.5mm/s, the filler welding and the cover welding
are both 7.5mm/s. The experimental steps are as follows:

162
 Advances in Intelligent Systems Research, volume 148

FIGURE VI. THE EXPERIMENT OF MULTI-LAYER AND MULTI-PASS

WELDING

IV. CONCLUSIONS
This paper aims at the inverse kinematics for 6-DOF
industrial robots. Based on screw theory, a new method is
proposed for inverse solution of kinematics according to its
structural characteristics. Firstly, the joint angle 1 is solved by
the geometrical relationship between the robot wrist point and
first three joint axes, and the rest angles are decomposed into
several known Panden-Kahan sub-problems. The accuracy and
reliability of the method is proved by the results in welding
experiment. It has the value of promotion and application in
the actual motion control of 6-DOF industrial robots.

ACKNOWLEDGMENT
We are grateful to the reviewers and the editors for their
constructive suggestions. The research of this paper is
supported by the National Natural Science Funds of China
(Granted No. 61703200), Scientific Research Fund Project of
Nanjing Insititute of Technology (Granted No. PTKJ201606),
Scientific Research Fund Project of Nanjing Insititute of
Technology (Granted No. PTKJ201701), Scientific Research
Fund for the introduction talent of Nanjing Insititute of
Technology (Granted No. YKJ201741), and sponsored by
Qing Lan Project.
REFERENCES
[1] SARIYILDIZ E, TEMELTAS H. Solution of inverse kinematic problem
for serial robot using quaternions[C]. IEEE Inter Conf. on Mechatronics
and Automation. Changchun: IEEE, 2009: 26-31.
[2] Tan Yue-sheng, Xiao Ai-ping. Extension of the second paden-kahan sub
problem and its’ application in the inverse kinematics of a
manipulator[C]. IEEE Inter Conf. on Robotics, Automation and
Mechatronics. Chengdu: IEEE, 2008: 379-381
[3] Chen wei-hai, Chen Quan-zhu, Zhang Jian-bin. Inverse kinematic
analysis for cable-driven humanoid-arm manipulator[J]. Chinese Journal
of Mechanical Engineering, 2007, 43(4):12-20.
[4] [10] Qian Dong-hai, Wang Xin-feng, Zhao Wei. Algorithm for the
inverse kinematic calculation of 6-DOF robots based on screw theory
and Paden-Kahan sub-problems[J]. Chinese Journal of Mechanical
Engineering, 2009, 45(9):72-76.
[5] [11] Chen Qing-cheng, Zhu Shi-qiang, Wang Xuan-yin. Inverse
kinematics sub-problem solution algorithm for serial robot based on
screw theory[J]. Journal of Zhejiang University (Engineering Science),
2014, 48(1): 8-14.
[6] [12] Hong Lei, JI Bao-jian, SHEN Jian, etc. Research on an algorithm
for industrial robot TCP calibration[J]. Journal of Machine Design, 2017,
34(3): 81-85

163