10.

1515/acsc-2015-0033
Archives of Control Sciences
Volume 25(LXI), 2015
No. 4, pages 513–527

Reference trajectory tracking for a multi-DOF

robot arm

RÓBERT KRASŇANSKÝ, PETER VALACH, DÁVID SOÓS, JAVAD ZARBAKHSH

This paper presents the problem of tracking the generated reference trajectory by the
simulation model of a multi-DOF robot arm. The kinematic transformation between task
space and joint configuration coordinates is nonlinear and configuration dependent. To
obtain the solution of the forward kinematics problem, the homogeneous transformation
matrix is used. A solution to the inverse kinematics is a vector of joint configuration coordi-
nates calculated using of pseudoinverse Jacobian technique. These coordinates correspond
to a set of task space coordinates. The algorithm is presented which uses iterative solution
and is simplified by considering stepper motors in robot arm joints. The reference trajec-
tory in Cartesian coordinate system is generated on-line by the signal generator previously
developed in MS Excel. Dynamic Data Exchange communication protocol allows sharing
data with Matlab-Simulink. These data represent the reference tracking trajectory of the end
effector. Matlab-Simulink software is used to calculate the representative joint rotations.
The proposed algorithm is demonstrated experimentally on the model of 7-DOF robot arm
system.
Key words: inverse kinematics, real-time reference tracking, signal generator, multi-DOF,
dynamic data exchange.

1. Introduction

Robotics represents one of the main disciplines in the industry. The synergy of ro-
botics with other applications like car assembly operations, vision systems or artificial
intelligence allows not only for innovations but also reduces the manufacture costs. In
the kinematic analysis of a robot arm position, two separate problems have to be solved:
forward kinematics and inverse kinematics. Forward kinematics includes solving the
forward transformation equation to find the location of the hand in terms of the angles
and displacements between the links of the robot arm.

R. Krasňanský, P. Valach and D. Soós are with Faculty of Electrical Engineering and IT, Slovak Uni-
versity of Technology in Bratislava, Ilkovicova 3, 81219 Bratislava, Slovak Republic. E-mails:(robert.kras-
nansky, peter.valach, david.soos)@stuba.sk. J. Zarbakhsh is with Carinthia University of Applied Science,
Europastrasse 4, 9524 Villach, Austria. E-mail: [email protected].
Received 16.07.2015.
This work has been supported by Grant N. 1/1241/12 of the Slovak Scientific Grant Agency.
514 R. KRASŇANSKÝ, P. VALACH, D. SOÓS, J. ZARBAKHSH

In order to find the corresponding sets of joint angle given the desired position and
orientation of the end effector it is necessary to solve the inverse kinematics problem.
Inverse kinematics, on the other hand, includes solving the inverse transformation equa-
tions in order to find the relationships between the links of the robot arm from its loca-
tion in space. There has been an increasing amount of work in recent years to derive the
inverse kinematics of the robot arm [1], [9], [6].
Many schemes as well as profound comparative analysis for the inverse kinematic
problem of redundant robot arm are proposed in previous robotics works.
Some schemes provide essential approach for manipulator geometries with unknown
inverse kinematic functions. However, for a continuous path motion control problem,
the inverse of the Jacobian matrix is required. A generalized inverse of Jacobian matrix,
so-called pseudoinverse is widely applied for a redundant robot.
The present work is concerned with the inverse kinematics solution for a seven-
DOF redundant manipulator. The trajectory planning of robot arms is a very active area
since many tasks require special characteristics to be satisfied. Using of MS Excel and
MS VBA (Visual Basic for Applications) software, the application for reference signal
generation has been developed. It allows to determine arbitrary reference trajectory and
to change it during runtime. By using of the Dynamic Data Exchange (DDE), the com-
munication between the application and Matlab-Simulink is secured in real-time.
The paper is organized as follows. In section 2, the general robot kinematics as well
as the inverse kinematics analysis is presented. In section 3, the kinematics model of
the seven-DOF robot arm is presented and the inverse kinematics solution is proposed.
Section 4 describes a real-time signal generator developed in MS Excel to generate an
arbitrary reference signal for the end effector. Subsequently, computer simulation and
experimental results of the proposed methodology are presented. Finally, the conclu-
sions are drawn in Section 5.

2.  Robot kinematics

2.1.  Homogeneous transformation matrix

The analysis the motion, a robot arm can be arranged by assigning certain coor-
dinate frames to each link, in order to obtain a transformation relating joint to space
coordinates as the position and orientation of the end effector with respect to a reference
frame [3]. The relative position and orientation of translation and rotation of the robot
arm in 3D space can be represented by the 4x4 homogeneous transformation matrix in
the following form:
R T 
Q=  . (1)
 0 1 

(Notation and symbols are given in the end of this paper).
 REFERENCE TRAJECTORY TRACKING FOR A MULTI-DOF ROBOT ARM 515

Homogeneous transformation is used to calculate the new coordinate values for

each robot part. The rotation for each axis in the space is represented by the 3x3 matrix,
which may change with respect to rotation value:

 r1 r2 r3 ∆x 
 3x 3 3x1   r r r ∆y 
 rotation matrix translation  4 5 6 .
Q =  (2)
 r7 r8 r9 ∆z 
1x3 perspective global scale   
 0 0 0 1 

The resulting motion is generated by composition of elementary motions of each

link with respect to the previous one. The joints are supposed to be controlled individu-
ally. The translation matrix has dimension 3x1 and introduces the changing value be-
tween the coordinate systems. Using rotation matrices brings a compact representation
of the rotational movement of a rigid body.
Consider the rotation of a given reference frame with respect to the world reference
frame (rotation about z-axis) as depicted in Fig. 1.

Figure 1. Rotation of frame (x’, y’, z’) respect to the (x,y,z) reference frame

Since the both frames have the common origin, the relationship between them fol-
lows only from the rotation. Rotation is then represented by the following matrix:

cos(θ ) - sin(θ ) 0 
Rz (θ ) =  sin(θ ) cos(θ ) 0  .
(3)
 0 0 1 

In the same way, we can define other rotations with respect to the axis y and x as
follows:
516 R. KRASŇANSKÝ, P. VALACH, D. SOÓS, J. ZARBAKHSH

 cos(θ ) 0 sin(θ ) 
Ry (θ ) =  0 1 0  , (4)
 - sin(θ ) 0 cos(θ ) 

1 0 0 
=Rx (θ ) 0 cos(θ ) - sin(θ )  .

(5)
0 sin(θ ) cos(θ ) 

The translation about x, y, and z axes of l is then in the form:

l  0  0
     
Tx = 0, Ty =  l , Tz = 0  .
(6)
     l 
0 0

Consider the 2-DOF robot arm as depicted in Fig. 2.

Figure 2. 2-DOF robot arm

Equation representing the dependence of the end effector (EE) coordinates on joint
coordinates and vice versa can be obtained according to the basic trigonometry in the
following way:
 x 0
 y 0
= =
EE  Rz (θ1 ) R y (θ 2 )Tz (l1 ) R y (θ 3 )   (7)
z  l2 
   
1 1
 REFERENCE TRAJECTORY TRACKING FOR A MULTI-DOF ROBOT ARM 517

where
cos(θ1 ) - sin(θ1 ) 0 0
 
 sin(θ1 ) cos(θ1 ) 0 0
Rz (θ1 ) =   (8)
 0 0 1 0
 
 0 0 0 1

 cos(θ 2 ) 0 sin(θ 2 ) 0
 
 0 1 0 0
R y (θ 2 ) =  
(9)
- sin(θ 2 ) 0 cos(θ 2 ) 0
 
 0 0 0 1

1 0 0 0 
 
0 1 0 0 
Tz (l1 ) =  
0 0 1 l1  (10)
 
0 0 0 1 

 cos(θ3 ) 0 sin(θ3 ) 0
 0 1 0 0 
Ry (θ3 ) =  . (11)
 - sin(θ3 ) 0 cos(θ3 ) 0
 
 0 0 0 1

Resulting position of the end effector is described by the following equations

x = l1 cos θ1 sin θ 2 + l2 cos θ1 sin(θ 2 + θ 3 ) (12)

y = l1 sin θ1 sin θ 2 + l2 sin θ1 sin(θ 2 + θ 3 ) (13)

z =l1 cos θ 2 + l2 cos(θ 2 + θ3 ). (14)

2.2.  Inverse kinematics
Inverse kinematics uses the kinematics equations of a robot to determine the joint
parameters providing a desired position of the end effector. For example, using these
formulas allows calculation of the joint parameters which set a robot arm to pick up an
object. Similar formulas might determine the positions of the skeleton of an animat-
518 R. KRASŇANSKÝ, P. VALACH, D. SOÓS, J. ZARBAKHSH

ed character which is supposed to move in a particular way. However, the solution of

the inverse kinematics problem is not always unique, since the same end effector pose
might be reached within several configurations corresponding to distinct joint position
vectors.
There are many methods of modeling and solving inverse kinematics problems.
Besides of analytic methods, most flexible methods typically rely on iterative opti-
mization in order to find an approximate solution. The main reasons are the difficulty
of inverting the forward kinematics equation and the possibility of an empty solution
space [5], [8], [2].
The inverse Jacobian technique is a simple yet effective way of implementing in-
verse kinematics. The velocities in joint space are being mapped to velocities in Car-
tesian space. It contains the first order partial derivatives of the joint angle parameters:

 ∂x ∂x 
  
 ∂θ1 ∂θ n 
 ∂y
∂y 
J =  . (15)
 ∂θ1 ∂θ n1 
 ∂z ∂z 
  

 ∂θ1 ∂θ n 

Calculation the inverse Jacobian allows the inverse kinematics to be directly calcu-
lated for a small time step as follows:

∆
=θ J -1 (θ )∆x. (16)

According to (16), the real function can be linearly approximated for sufficiently
small steps. For kinematically redundant robots with n > 6 DOF, the Jacobian matrix is
non-square and its inverse can be obtained using a pseudoinverse J+ [4], [10]

J + = ( J T J ) -1 J T . (17)

However, by using the pseudoinverse, the problems of singularities in the Jaco-

bian still remain. These singularities might occur for instance, when multiple links be-
come aligned in the same direction and subsequently, identical derivatives for several
joints of the robot arm are obtained. Inversion of nearly singular matrices results in
excessively large velocities and causes unrealistic behavior with oscillations around
a  singular configuration. There are many alternative approaches which address the
inverse kinematics problem, such as Jacobian transpose methods [13], quasi-New-
ton and conjugate gradient methods [12], [14] or Levenberg-Marquardt damped least
squares methods [11], [7].
 REFERENCE TRAJECTORY TRACKING FOR A MULTI-DOF ROBOT ARM 519

3.  Modeling and control of multi-DOF robot arm

3.1.  Model of 7-DOF robot arm

The 7-DOF robot arm is composed of four rotational joints and five links with
lengths l1 to l5 as depicted in Fig. 3. We consider seven stepper motors for rotation and
translation of individual links. Joints two and four can rotate about two axes. All the
others can rotate just about one axis. We introduce the following notation:
θi – joint variable (generalized coordinates), i = 1, 2,…,7;
li – length of i-th link, i = 1, 2,…,5.

Figure 3. 7-DOF robot arm

3.2.  Inverse kinematics analysis

The control task was to determine the joint parameters that provide a desired posi-
tion of the end effector. This position is determined by the reference values of x, y, z
coordinates and angles θ6, θ7, which determine the rotation settings of the end link. The
end effector position was determined according to (7) in the following way:

EE = Tz (l1 ) Rz (θ1 ) R y (θ 2 )Tz (l2 ) R y (θ 3 )Tz (l3 ) Rz (θ 4 )

(18)
R y (θ5 )Tz (l4 ) R y (θ 6 ) Rx (θ 7 ) [0 0 l5 1]
T
520 R. KRASŇANSKÝ, P. VALACH, D. SOÓS, J. ZARBAKHSH

where
1 0 0 0 1 0 0 0
   
0 1 0 0 0 1 0 0
Tz (l1 ) =  , Tz (l2 ) =   (19)
0 0 1 l1  0 0 1 l2 
   
0 0 0 1  0 0 0 1 

1 0 0 0 1 0 0 0
   
0 1 0 0 0 1 0 0
Tz (l3 ) =  , Tz (l4 ) =  
0 0 1 l3  0 0 1 l4  (20)
   

0 0 0 1  0 0 0 1 

cos(θ1 ) - sin(θ1 ) 0 0 cos(θ 4 ) - sin(θ 4 ) 0 0

   
 sin(θ1 ) cos(θ1 ) 0 0  sin(θ 4 ) cos(θ 4 ) 0 0
Rz (θ1 ) =  , Rz (θ 4 ) =   (21)
 0 0 1 0  0 0 1 0
   
 
 0 0 0 1  0 0 0 1

 cos(θ 2 ) 0 sin(θ 2 ) 0  cos(θ 3 ) 0 sin(θ 3 ) 0

   
 0 1 0 0  0 1 0 0
R y (θ 2 ) =  , R y (θ 3 ) =   (22)
- sin(θ 2 ) 0 cos(θ 2 ) 0 - sin(θ 3 ) 0 cos(θ 3 ) 0
   
 
 0 0 0 1  0 0 0 1

 cos(θ 5 ) 0 sin(θ 5 ) 0  cos(θ 6 ) 0 sin(θ 6 ) 0

   
 0 1 0 0  0 1 0 0
R y (θ 5 ) =  , R y (θ 6 ) =   (23)
- sin(θ 5 ) 0 cos(θ 5 ) 0 - sin(θ 6 ) 0 cos(θ 6 ) 0
   
 
 0 0 0 1  0 0 0 1

1 0 0 
Rx (θ 7 ) 0 cos(θ 7 ) - sin(θ 7 )  .
=  (24)
0 sin(θ 7 ) cos(θ 7 ) 

 REFERENCE TRAJECTORY TRACKING FOR A MULTI-DOF ROBOT ARM 521

Since the rotation of the end link of the robot arm is determined by the reference
trajectory, it stays in determined position until the reference values of angles θ6, θ7 will
change. The calculation of joint parameters is realized according to Algorithm 3.1. Re-
sulting Jacobian matrix has a size of 3 x 5.

Algorithm 3.1 Inverse kinematics

Inputs: length of links l1 – l5, reference coordinates x, y, z, reference angles θ6, θ7.
Outputs: joints parameters Δθi.
Step 1: Determine the position of the end effector’s according to (18) – (24)
Step 2: Determine the Jacobian matrix according to (15)
Step 3:  for k = 1: n do
1: Evaluate Jacobian with actual parameters θi and the inverse Jacobian
using pseudoinverse J+ (17);
2: Obtain the vector Δθ according to (16);
3: Compute the joints parameters θi(k+1)= Δθi + θi(k);
4: Check the position error of the end effector according to reference tra-
jectory point err = xref – xi; if err satisfy certain condition, go to step 4,
otherwise go back to step 3;
    end for
Step 4: Update θi(k) ← θi(k+1).
In this way, we obtained suitable joint rotations in order to achieve the desired posi-
tion of the end effector.

3.3.  Real-time reference trajectory generator

In the motion control we consider the trajectory of the end effector which is updated
through the time, so these information can be stored in a table or in a text file. Conse-
quently, they are sent to Matlab-Simulink software in real-time. In Matlab-Simulink we
are able to compute exact joint rotation to achieve required reference trajectory by us-
ing the inverse kinematics. In this paper, we consider the trajectory data obtained from
table stored in Microsoft Excel (MS Excel), where the so-called signal generator (SG)
is developed. Thus, the signals for the end effector motion in x, y, z coordinates as well
as the exact rotation of the end effector using the angles θ6 and θ7 (Fig. 3) are generated
in real-time.
The communication between Microsoft Excel and Matlab-Simulink needed for data
exchange is established by using DDE (Dynamic Data Exchange) communication de-
veloped by Microsoft. Thus, the goal of using Microsoft Excel is to generate analogue
processing variables in real-time, which are used to generate trajectory of the end ef-
fector.
In order to generate different kinds of signals in real-time, it was needed to develop
the automatic generator on-line. By using appropriate functions in MS Excel, it was
possible to obtain the value of real system time. In order to update data sheet continu-
ously, it was needed to create a macro using Visual Basic software, which would per-
522 R. KRASŇANSKÝ, P. VALACH, D. SOÓS, J. ZARBAKHSH

form these updates in the shortest time interval. The macro can be run by using the start/
stop button, as can be seen in Fig. 4 (only a part of the main window of the application
is shown in Fig. 4).

Figure 4. Real-time signal generator

The task of the macro is to run the procedure containing infinite update cycles in
MS Excel data sheets. In Fig. 4 it can be also seen how the cells are generated in the
real-time. In order to create required trajectory of the motion of the end effector, differ-
ent kinds of mathematical functions and operations for the coordinates and angles were
used and stored in other cells of the data sheet.
In this way, we obtained the vector of desired reference values in the following
form:

r = ( x, y , z , θ 6 , θ 7 ) (25)

The vector r is updated continuously in MS Excel (Fig. 4). The values of the vector
r are consequently sent via DDE to Simulink software, where the inverse kinematics
problem is solved according to (15)–(17) and the resulting joint parameters are ob-
tained. From these parameters the position of each robot link are calculated according
to (18) for simulation purposes. The calculated position of the end effector is sent for
plotting as depicted in Fig. 5.
Obtained data can be used to control of the real robotic system by simple intercon-
nection of Arduino platform and Matlab-Simulink software. In the next section, the
simulation results of the motion of individual robot links are shown.
 REFERENCE TRAJECTORY TRACKING FOR A MULTI-DOF ROBOT ARM 523

DDE [A] u fcn y -18.04

0.5526
GotoA
Theta6 Jpseudo Matrix 1 1.862
Multiply s
3.805
multiply Integrator
1.075
DDE [B]
Angles_rad
GotoB
Theta7 End_Efector
1.434 XYZ_End_Efector FromA
MATLAB [A]
0.8379
Function
DDE 0.5
FromB

X joint 4 [B]
1.301 XYZ_4
0.9808 MATLAB
DDE
Function
0.4568
Y XY_ref

joint 3
DDE
0.8178 XYZ_3
x
0.864 MATLAB
Z Function
0.4042

joint 1
joint 2
0 XYZ_1 0.3608 XYZ_2
XY_EE
0 0.3812 MATLAB
[0;0;l1]
Function y
0.3 1.151
0.5

End_Efector_Z

Figure 5. Real-time control scheme in Matlab-Simulink

4.  Simulation results

The simulations were carried out in the Matlab-Simulink software. The link length
parameters of the 7-DOF robot arm are as follows: l1 = 0.3 m, l2 = 1 m, l3 = 1 m,
l4 = 0.5 m, l5 = 0.2 m. The rotational range of motion of individual joints is as follows:
θ1Î<0, 270°>, θ2Î<-60°, 120°>, θ3Î<-120°, 150°>, θ4Î<-180°, 180°>,
θ5Î<-90°, 90°>, θ6Î<-90°, 90°>, θ7Î<-90°, 90°>, where the values of the angles θ6
and θ7 are determined by the user via signal generator. According to these parameters
we were able to evaluate and plot the workspace of the end effector in 3D space as de-
picted in Fig. 6.
The simulation results of the robot arm tracking the desired circle-shape trajectory
generated in real-time via signal generator in MS Excel are presented in Figs 6–8. Each
link is represented by a unique color.
524 R. KRASŇANSKÝ, P. VALACH, D. SOÓS, J. ZARBAKHSH

Figure 6. Workspace of the end effector in 3D space

The circle with the radius of 0.5 m is situated in the workspace of the end effector
considering the midpoint coordinates: (x, y, z) = (1, 1, 1). Simulated trajectories pro-
jected on the XY and XZ planes are shown in Fig. 7.

Figure 7. Trajectory of the end effector‘s motion (blue dots) and reference trajectory (red dots) in A) XY
plane B) XZ plane

The robot arm′s configuration during the simulation is depicted in Fig. 8. For better
illustration of the motion of each link, every sixth step of motion is considered.
 REFERENCE TRAJECTORY TRACKING FOR A MULTI-DOF ROBOT ARM 525

x z

y
Figure 8. Robot arm configuration

The time responses of the end effector′s position in X and Y axis are shown in Fig. 9.
These results verify the sufficient tracking of the reference trajectory by the end effector.

2
reference (y-axis)
1.5 end effector (y-axis)
1 end effector (x-axis)
position

reference (x-axis)
0.5

-0.5

-1
0 50 100 150 200 250 300
time [s]

Figure 9. Time response of the end effector position in X, Y axes

As can be seen from the figures, the simulation experiments have shown desired
results.

5. Conclusions

This paper presents a complete solution to the inverse kinematics of a 7-DOF ro-
bot arm. In order to obtain the solution, the iterative procedure based on calculating
of pseudoinverse Jacobian was proposed. The procedure requires that the Jacobian is
computed numerically. The main contribution of this work has been the development of
the signal generator in order to generate the reference trajectory in real-time. This tool
has been developed using macros in MS Excel and it enables modulation of the shape of
the reference trajectory continuously during runtime. The simulation results prove the
526 R. KRASŇANSKÝ, P. VALACH, D. SOÓS, J. ZARBAKHSH

efficiency of the proposed solution with the developed tool for the real-time reference
trajectory tracking by the robot arm.

Notation and symbols

DOF degree of freedom

DDE dynamic data exchange
VBA visual basic for applications
EE end effector
SG signal generator
k time instant
R rotation matrix
T translation matrix
Q homogeneous transformation matrix
q rotation angle between the link and the respective reference frame
l link of the robot arm
J Jacobian
J+ pseudoinverse of Jacobian
JT transpose of Jacobian
x, y, z coordinates in Cartesian coordinate system
r desired reference value
Δ time derivative

References

 [1] J. Angeles, F. Ranjbaran and R.V. Patel: On the design of the kinematic struc-
ture of seven-axes redundant manipulators for maximum conditioning. Proc. of the
IEEE Int. Conf. Robotics and Automation, Nice, France,(1992), 494-499.
 [2] J.G.P. Barnes: An algorithm for solving non-linear equations based on the secont
method. Computer J., 8, (1965), 66-72.
 [3] J. Denavit and R.S. Hartenberg: A kinematic notation for lower-pair mecha-
nisms based on matrices. Journal of Applied Mechanics, (1955), 215-221.
 [4] R.G. Fenton, B. Benhabib and A.A. Goldenbereg: Optimal point-to-point mo-
tion control of robots with redundant degrees of freedom. J. of Manufacturing
Science and Engineering, 108(2), 1986, 120-126.
  [5] R. Fletcher: Generalized inverse methods for the best least squares solution of
systems of non-linear equations: Computer J., 10 (1968), 392-399.
 REFERENCE TRAJECTORY TRACKING FOR A MULTI-DOF ROBOT ARM 527

 [6] C.A. Klein and C.H. Hung: Review of pseudoinverse control for use with kin-
ematically redundant manipulators. IEEE Trans. on Systems, Man, and Cybernet-
ics, 13, 245-250.
 [7] Y. Nakamura and H. Hanafusa: Inverse kinematics solutions with singularity
robustness for robot manipulator control. J. of Dynamic Systems, Measurement,
and Control, 108 (1986), 163-171.
 [8] P. Rabinowitz: Numerical Methods for Non-linear Algebraic Equations. New
York: Gordon and Breach, 1970, 87-114.
 [9] J. De Schutter, T. De Laet and J. Rutgeerts: Constraint-based task specifica-
tion and estimation for sensor-based robot systems in the presence of geometric
uncertainty. The Int. J. Robotics Research, 26(5), (2007), 433-455.
[10] B. Siciliano: Kinematic control of redundant robot manipulators: A tutorial. J. of
Intelligent and Robotic Systems, 3(3), (1990), 201-212.
[11] C.W. Wampler: Manipulator inverse kinematic solutions based on vector formu-
lations and damped least squares methods. IEEE Trans. on Systems, Man and Cy-
bernetics, 16 (1986), 93-101.
[12] L.C.T. Wang and C.C. Chen: A combined optimization method for solving the
inverse kinematics problem of mechanical manipulators. IEEE Trans. on Robotics
and Automation, 7 (1991), 489-499.
[13] W. A. Wolovichand and H. Elliot: A computational technique for inverse kin-
ematics. 23rd IEEE Conf. on Decision and Control, (1984).
[14] J. Zhao and N. I. Badler: Inverse kinematics positioning using nonlinear pro-
gramming for highly articulated figures. ACM Trans. on Graphics, 13 (1994),
313-336.