The Inverse Kinematics Solutions of Industrial Robot Manipulators
Serdar Kiiqiik Zafer Bingul
Electronic and Computer Education
Kocaeli University, Kocaeli, Turkey
e-mail :skucuk@kou.edu.tr
Abstract
The inverse kinematics problem of robot manipulators is
desired to be solved analytically in order to have complete
and simple solutions to the problem. This approach is also
called as a closed form solution of robot inverse
kinematics problem. In this paper, the inverse kinematics
of sixteen industrial robot manipulators classified by
Huang and Milenkovic [I] were solved in closed form.
Each robot manipulator has an Euler wrist (Figure 1)
whose three axes intersect at a common point. Basically,
five trigonometric equations were used to solve the
inverse kinematics problems. Robot manipulators can be
mainly divided into four different group based on the joint
structure. In this work, the inverse kinematics solutions of
SN (cylindrical robot with dome), CS (cylindrical robot),
NR (articulated robot) and CC (selectively compliant
assembly robot arm-SCARA, Type 2) robot manipulator
belonging to each group mentioned above are given as
an example. The number of the inverse kinematics
solutions for the other robot manipulator was also
summarized in a table.
1. Introduction
A robot manipulator is composed of a serial chain of
rigid links connected to each other revolute or prismatic
joints. A revolute joint rotates about a motion axis and a
prismatic joint slide along a motion axis. Each robot joint
location is usually defined relative to neighboring joint.
The relation between successive joints is described by 4x4
homogeneous transformation matrices that have
orientation and position data of robots. The number of
those transformation matrices determines the degrees of
fieedom of robots. The product of these transformation
matrices produces final orientation and position data of a
n degrees of fi-eedom robot manipulator. Given the sets of
joint angles, calculate the position and orientation of the
end-effector of the robot manipulator, is called forward
kinematics [Z]. Forward kinematics problem is
straightforward and there is no complexity deriving the
equations [ 5 ] . Hence, there is always a forward
kinematics solution of a robot manipulator. Tasks to be
02004 IEEE
0-7803-8599-3/04/$20.00 274
Department of Mechatronics Engineering
Kocaeli University Kocaeli, Turkey
e-mail:zaferb@kou.edu.tr
performed by a robot manipulator are in the Cartesian
space, whereas actuators work in joint space. Cartesian
space includes orientation matrix and position vector.
However, joint space is represented by joint angles. The
conversion of the position and orientation of a robot
manipulator end-effector from Cartesian space to joint
space is called as inverse kinematics problem [3,4]. This
relationship between joint space and Cartesian space is
illustrated in Figure 2.
A
Figure 1. The configuration of Euler wrist.
The inverse kinematics problem of six joint robot
manipulators has been studied for decades. The reason of
this study comes from widesprade use of six jointed robot
manipulators in industry. The complexity of the inverse
kinematics problem of industrial robot manipulators arises
from their geometry and including nonlinear equations.
Some other difficulties in inverse kinematics: kinematics
equations are coupled and multiple solutions and
singularities may exist. Matematical solutions for inverse
kinematics problem may not always correspond to
physical solutions and method of its solution depends on
the robot configuration [6-81.
There are three types of inverse kinematic solution:
complete analytical solution (closed form solution),
numerical solutions and semi-analitical solutions. In the
first type, all of the joint variables are solved analytically
according to given configuration data. Closed form
solution is preferable because in many applications where
the manipulator supports or is to be supported by a
sensory system, the results from kinematics computations
need to be supplied rapidly in order to have control
actions. In the second type of solution, all of the joint
variables are obtained iterative computational procedures.
There are four disadvantages in these: a) incorrect initial
estimations, b) before executing the inverse kinematics
algorithms, convergence to the correct solution can not be
guarantied, c) multiple solutions are not known, d) there
is no solution, if Jacobian matrix is singular [9]. In the
third type, some of the joint variables are determined
analytically in terms of two or three joints variables and
these joint variables computed numerically.
Joint
Figure 2. The schematic representation of forward and
inverse kinematics.
This paper is written following manner. In Section II,
the properties of homogenous transformation matrices
were identified and inverse kinematics problem
formulation was given. In Section 111, the two-Letter code
description of robot configurations were explained and
the fundamental robot manipulators were classified
according to their joint properties. In Section IV, the
inverse kinematics solutions of SN, CS, NR and CC robot
manipulator which were given in closed form and the
numbers of inverse kinematics solution of sixteen robot
manipulator were presented as table. Finally, the
contribution ofthis work is insisted in Section V.
2. Problem Formulation
For a six jointed robot manipulator, the position and
orientation of the end-effector with respect to the base is
given by,
where ,!,Is represent the rotational elements of
transformation matrix ( i a n d j = l , 2 and 3 ). p i , p v
and p _ are the elements of position vector. Using
complete analytical solution technique, it should be
started by equating the known transformation with the
product of the link transformations.
06T =?T(qi) h q 2 ) :T(q;) !T(qj) ' h q s ) iT(qg)
where q i is the joint variable (revolute joint or prismatic
joint) for joint i. To find the inverse kinematics solution,
it should he solved for q i as function of the known
elements of md&ET.To find q1 firstly as a h c t i o n
of the known elements, the link transformation inverses
are premultiplied as follows.
13)
~,
h3)
h2) 45m)
h4) im6)
where [qT(ql)rl q T ( q l ) = l , I is identity matrix. Then
equation 3 can be simplified as follows.
[pT(q,)ri :T= h 2 ) ?Aa):7'(q4) '$"(qs) h)(4)
To find the other joint variables, the following equations
are obtained in a similar manner.
[qT(41) :T(g2)ri %= :T(q3) : T ( 4 4 ) 'h,56T(%)
) (5)
[:%) h q 2 ) ?(93)f' !$= : T ( q j ) !T(qs) ? J ( q g ) (6)
[ q T ( a ) iT(q2) &) h 7 4 ) r i HT= !7(95) h k ) (7)
(qT(91) h) : T ( s j ) ?T(qs)ri :T= i T ( ' 1 6 ) (8)
There are 12 simultaneous set of nonlinear equations
to be solved. The only unknown on the left hand side of
equation 4 is q l . The 12 nonlinear matrix elements of
right hand side are either zero, constant or functions of
q2 through q 6 . If the elements of on the left hand side
that are a function of q1 are equated with elements on
the right hand side then the joint variable q , can be
solved as functions of rII , rL2 ..... r 3 3 , p x , p?, and
pI and fixed link parameters. Once 9 , is found, then
other joint variables are solved by same procedure as
before. There is no necessity that first equation will
produce q1 and second q 2 etc. To fmd suitable equation
to solve the inverse kinematics problem, any equation
defined above (equations 4-8) can be used arbitrary. Some
trigonometric equations used in the solution of inverse
kinematics problem were given as Table in Appendix A.
3. Description of Robot Configurations
B.Huang and V. Milenkovic [ l ] used a two-letter code
to classify a robot configuration. The first letter
characterizes the first joint and the first joint's
relationship to the second joint. The second letter
identifies the third joint and third joint's association to the
(2) second joint. The code letters and their meanings are:
275
 S : Slider
C : Rotary parallel to slider
N : Rotary perpendicular to rotary
R : Rotary perpendicular to rotary, or rotary
parallel to rotary

The combination of these rotary and prismatic joints

compose the sixteen robot configurations which are
named as

SS, SC, SN, CS, CC, CR, NS, NN, NR, RC, RN, RR,
RS, SR, CN and NC.

Table I gives the joint structure of sixteen robot

manipulator (P, R, F, S, T and M represent prismatic,
revolute, first, second and third joints and manipulator,
respectively).

Table 1. The joint structure of sixteen robot manipulator.

The fundamental robot manipulators can be divided in to

four types according to their joint properties as follows. In
the first Type, there are five orthogonal robot manipulator
(SS, SN, NS,.NN and RR) which have three joints are
perpendicular each other. In the second Type, there are
four robot manipulators.(CS, CR, RN and CN) whose the
first two joints are parallel to each other. In the third
Type, there are four robot manipulators (SC, NR, RC and
NC) whose the last two joints are parallel to each other. In
the fourth Type, there are three robot manipulators (CC,
RS and SR) whose three joints are parallel to each other.

3. The Inverse Kinematics

In this section, the inverse kinematics of SN, CS, NR
and CC robot manipulators illustrated in Figure 3 were
driven. The forward transformation matrices of SN, CS,
NR and CC robot manipulator used in inverse kinematics
- , --
problem were given in Appendix B.

3.1. The Inverse Kinematics for SN Robot

Revolute joint variable ff, can be determined by equating

(2,4) matrix elements each of sides (on the left hand side
and right hand side) in equation 4.
(d)

(9) Figure 3. (a) SN robot manipulator. (b) CS robot

manipulator. (e) NR robot manipulator. (d) CC robot
manipulator.

276
Taking square of (l,4), (2,4) and (3,4) matrix elements of Extracting cos0, and sin$, from (1,3) and (3,3),
each sides and adding the results then, prismatic joint cos8, and sin 8, from (2,2) and (2, I) matrix elements of
variable d, can be computed.
each sides in equation 7, revolute joint variables 8, and
8, can he computed, respectively.
(lo)
d, = p , - d d , 2 + 1 2 2 - 2 d 2 d 4 ~ 8 ~ - p ~ 2 - I ~ 2 + Z P r ~ ~-Py2

0, = Atan2(cB1cB4r,;+sB2cS,r,; +s8,r3, ,
Revolute joint variable 8, can be found by equating (1,3) (19)
s@i3 - 4 r 2 ; )
matrix elements of each sides in equation 4.
. . 0, = Atan2(-~B~sB,r,~
-sf3,s8,r2, +cB4r31,
B2 = Atan2(d4sO3J2) (20)
-cBlsB4r,2 -sB1sB4r22 +c84r32)
(1 1)
f A tan2(dl: +d2s2& -(pz - d,I2.pz - 4)

Revolute joint variable 8, can be determined by equating 3.3. The Inverse Kinematics for NR Robot
(2,3) matrix elements of each sides in equation 6
Revolute joint variable 8, can be determined by
equating (2,4) matrix elements of each sides in equation
0, = Atan2(i-dl -(cS,s8;q3 -cB3r2, + s B 2 s B , r 3 ~ ) ’ , 4.
(12)
cB2s&r,, -cB,r2; +sS2s0,r3;)
01=Atan2(-py . - P ~ ) Or 81=Atm2(py PI) (21)
Extracting cos.9, and sin$, from (1,3) and (3,3), cos8,
and sin8, from (2,l) and (2,2) matrix elements of each Taking square of (1,4) and (3,4) matrix elements of each
sides in equation 6, revolute joint variables 8, and 8, sides in equation 4 and adding the results then, revolute
joint variable 8, can be computed.
can be computed, respectively.

8, = Atan2([-sB2r,; + ~ . 9 ~ q ~ ] ,/ s 8 ~
(13)
[cS2cB;r,; + sB3r2;+ss2c03r3,]/s0,)
2 2
( c 8 , p , + s 8 , p , . ) 2 + ( ~ , - h 1 ) 2 - d q-12
0, = Atan2([-cB2sbjq2 -s82s6jr32l/s0j ,
(14)
* I-(
2d412
1
[ C B ~ S B ;-c&r2,
~ , ~ +sB&r,~l/sQ,)
Revolute joint variable 8, can be determined by equating
3.2. The Inverse Kinematics for CS Robot
(1,4) matrix elements of each sides in equation 4.
Revolute joint variable 8, can be determined by B2=Atan2(d4c8~,d4s03
+12)
equating (1,4) matrix elements of each sides in equation
f Atan2(,/d: +I: +2I2d4sB3- ( c B I p r +rqp,,)’ , (23)
4.
ce,px + S B ~ P J
= A t a n 2 ( p y , p x ) + A t a n 2 ( J m , I , ) (15)
Revolute joint variable 8, can be determined by equating
Prismatic joint variables d2 and d,can he attained by (2,3) matrix elements of each sides in equation 6.
equating (3,4) and (2,4) matrix elements of each sides,
respectively in equation 4.

d2 =P: (16)
d3 = sBI px -CBI p v -d, (17)
Extracting cos@, and from (1,3) and (3,3),
Revolute joint variable 8, can be determined by equating cos@, and sine6 from (2,l) and (2,2) matrix elements of
(2,3) matrix elements of each sides in equation 7. each sides in equation 3, 8, and 8, revolute joint
variables can be computed, respectively.
8, = ~ t a n z ( - r ;, ~- C B , Y ~ ~-selr2;) or
8, = Atan2(1j~. c 0 1 r 1 3 + s 0 , r 2 ; ) (18)

277
 e, = ~ t a n 2 ( [ s e , q ,-celr23iisej , Table 2. The numbers of inverse kinematics solution for
[~ew~,r,,+ s w 2 3 r 2 3+ s e 2 3 r 3 3 ~ / s ~ j(25)
j sixteen robot manipulator.

4 = A t a n 2 ( [ ~ @ 8 ~+se1se2,r2,
~r,~
[-ce,sOZ3q I
- ~ 0 2 ~ r ~ ~ l ,i . s O ~
-.sOl.~023r21
+c823r3~lis8~j
M S S S C C C
S C N S C R S N
N I 2 4 1 8 1 4 1 81 8
:
N

16
N
R
16
R R
C N
81 16 : i
R R S C N
R S R N C
81 4 4 18: 8

N represents the number of solutions of sixteen robot

3.4. The Inverse Kinematics for CC Robot
Prismatic joint variable d2can be determined by
equating (3,4) matrix elements of each sides in equation
4.
d2 = pi
Taking square of (1,3) and (2,3) matrix elements of each
sides in equation 4 and adding the results then, revolute
joint variable 8, can he computed.
Revolute joint variable 8, can be determined by equating
(1,3) matrix elements of each sides in equation 4.
e, = Alan 2(p, ,p,)-t A tan 2(dp,2 + p; - (+e3 + l2 j2
dqse3+i2j
Revolute joint variable 8, can be found by equating (2,3)
matrix elements ofeach sides in equation 7.
manipulators.
4. Conclusion
The inverse kinematics solutions of sixteen industrial
robot manipulators were found in closed form. These
(27) manipulator designs have efficient closed form solutions,
because they allow for the decoupling of the position and
orientation kinematics. The geometric feature that
generates this decoupling is the intersection of joint axes.
Having closed form solutions of robot manipulators
avoids from computation cost and time consuming, while
inverse kinematics solution data are evaluated in the
microprocessor. Therefore, it is desirable to find robot
structure that gives closed form solution. In literature,
there are no criteria to test which robot structure gives
closed form solution or to obtain closed form solution
what kind of robot structure should be had. It can be said
that six degree of freedom robot manipulators with Euler
wrist always produce closed form solutions.
References
2 (29) [I] V.Milenkovic and B.Huang, “Kinematics of Major Robot
Linkages”, Robotics /nternationo[ of SME, April. 1983,vo1.2,
pp. 16-3 1.
[2] 1. Craig, hhoduction to Robotics: Mechanics and Control,
USA, 1989.

0, =Atan2(-r3, ,-ce13r13-sOI3rz3j or [3] K. S. Fu, R. C. Gonzalez, and C. S. G. Lee, Robotics:

Control Sensing and Intelligence, New York, Mc Grow-Hill,
8, =Atan2(r33,c43r13+ s 4 3 ~ 3 ) (30) 1987.

Extracting cos@j and sinOj from (3,3) and (1,3), [4l V. D. Towissis, “Principles and Design of Model-based
Controllers,” I n / . J of Contr.,1988, vol. 47(5j, pp.1267-1275.
cos8, and sine6 from (2,Z) and (2,l) matrix elements of
each sides in equation 7, revolute joint variable O5 and [5] Robert J. Schilling, Fundamentals of Robotics, Prentice
Hall, USA, 1990.
0, can be computed respectively.
[6] Phillip John McKerrow, Introduetion to Robotics, Addison-
us = A t a n 2 ( - ~ 6 ~ c Q-seI3c8,r2,
~~q~ -se4rj3, Wesley Publishing, USA, 1991.
(31)
se13q3 -c813r23)
[7] H. Asada, J.J.E. Slotine, Robot Anolysir ond Control, Wiley
Interscience Publication. USA. 1986.
e, = Atan2(-cfi3sQqq1-s8,,s04r2,+ C B , I ~ ~ ,
(32) [SI William A. Volovich, Roborzcs- Bmic Analym ond Design,
- ~ e , ~ s e ,-se13se4r22
q~ +ce4r3,) CBS Collage Publication, USA, 1987.

Table 2 gives the numbers of inverse kinematics solution [Y] V. D. Tourissis and Marcel0 H. Ang, “A Modular
for sixteen robot manipulator. Architecture for Inverse Robot Kinematics”, IEEE Transaction
on Robotics andAuromation, 1989, Vol. 5( 5 ) .

278
 Appendix A
T
- Ill IV
Table 3. Some trigonometric equations and solutions cos0, -sin@ 0 0
used in inverse kinematics.
!T

I
;j
Equations Solutions
osinO+bcos8=c 1 O=AtanZ(o,b)

osinO+bcos8=0 1
f AtanZ(d-.
O=Atan2(-b, a) or
c)
:T
cos8,
[s:?i
-sine2
co;S2
0 ;
0
-1
0
0

B = A t a n 2 ( b ,-a)
case = a and sin 8 = b I 8 = Aian 2 j b . a I cos8, -sineJ 0 1,
sine, cosb', 0 0
0 0 1 0

sinO=a I
I

B=Atan2[ U, fm) 0 0 0 1

1
cos8, -sin.!?, 0 0 cos.9, -sin.!?, 0 0
0 -I -d4
0 -I -d4
Appendix B :r
cosls, .;, ;
Table 4. The forward transformation matrices of (I) SN: c o d j -sine5 0 0
(11) CS, (111) NR and (IV) CC robot

1 :]
0
manipulators.
4
ST
I I1

- s i i08 6
0
-sine, -cosS6 0 0

T represents the index of transformation matrices.

1 II] 0
0
-cosS,
-sinBJ -sileJ
cos04 -sine4 0
-1
0 0

0
-d4
cosl?, -sine4 0 0

a'
1 3 cos0, -sin@, 0 0
0 [si;8, co;06
0
-sins, -cos@, 0 0

279