# PARAMETER IDENTIFICATION OF INDUCTION MOTORS BY USING GENETIC

ALGORITHMS
Marián Jančovič, Milan Žalman, Ján Jovankovič
MicroStep-HDO s r.o.,
Faculty of Electrical Engineering and Information Technology, Slovak University of Technology,
Faculty of Electrical Engineering and Information Technology, Slovak University of Technology
mjancovic@gmail.com, milan.zalman@stuba.sk, jan.jovankovic@stuba.sk

Abstract
This paper aims at demonstrating the possibilities of
genetic algorithms application for parameter
identification of induction motor. The basic procedure to
develop a genetic algorithm is described and the
examples of its application for parameter identification
are introduced. Genetic algorithms (GA) present
universal optimization methods, which can be used to
solve a wide scope of optimization problems. One of
them is the identification of nonlinear systems. For this
ability, we can use GA for parameter identification of
induction motor. We need to know the values of these
parameters for control design, state observer design or
modeling of induction motor. Parameter identification of
induction motor by using genetic algorithm is based on
calculating the values of appropriate quality criterion
which compare input-output data from induction motor
and from simulation model. Minimization of this quality
criterion leads to obtain more precise parameters.

Keywords: induction motor, genetic algorithm,
parameter identification

1. INTRODUCTION
Genetic algorithms (GA) present universal optimization
methods, which use global stochastic search algorithms.
Owing to their universality these algorithms can be used
to solve a wide scope of optimization problems. One of
them is the identification of nonlinear systems. For this
ability, we can use GA for parameter identification of
induction motor. We need to know the values of these
parameters for control design, state observer design or
modeling of induction motor. The accuracy of designed
structures is also dependent on the accuracy of the motor
parameters. As their values can vary from various
reasons so the quality of motor control can be reduced.
Then, for the optimization of control we need to know
the new values of the parameters. Because a direct
measure some of these parameters is very complicated or
impossible, the various researches developed different
techniques for their identification.

Distribution of the identification methods starts from the
measured data set of motor. Under this we can classify
next categories [1]:
• with the measurement of voltage, current and
angular speed,
• with the measurement of current and angular speed –
for the structures with the measurement of angular
speed,
• with the measurement of voltage, current – for the
structures without direct measurement of angular
speed (sensorless control).

The methods of parameter estimation and identification
for servo-drive with induction motor can be divided also
into the following categories:
• Observers that are derived from the dynamic model
of induction motor
o Direct observers that use measured stator
voltages and currents
(MRAS)
o Feedback observers with the adaptation
(Kalman, Luenberger)
• Observers with artificial intelligence
o Systems based on fuzzy logic
o Neural networks
o Neuro-fuzzy systems
o Evolution algorithms (genetic algorithms)
• Observers that use the properties of motor
construction
o Salient motor poles, motor slots, asymmetry
of the air gap
o The third harmonic of the voltage that are
caused by the saturation of the magnetic
circuit
o Using the superposed signals

Parameter identification of induction motor using genetic
algorithm is based on calculating the values of
appropriate quality criterion which compare input-output
data from induction motor and from simulation model.
Minimization of this quality criterion leads to obtain
more precise parameters. The advantage of GA is that it
provides the identification of parameters with adequate
accuracy also for noised and quantized measured signals.
On the other hand, the system complexity and number of
searched parameters influence the number and time
demands on simulation of dynamical processes. These
properties also causes that genetic algorithms are
significantly time consuming.

2. PARAMETER IDENTIFICATION USING GA
For parameter identification of induction motor, we used
mathematical model of motor in stator coordinate system
described in [2]. The state equations for this model are as
follows:

+ +
′ +
= ω ψ ψ
β α α α r r r
r
r
s
s
s
k
T
k
u
s L R
i
1
1
(1)

− +
′ +
= ω ψ ψ
α β β β r r r
r
r
s
s
s
k
T
k
u
s L R
i
1
1
(2)

+
=
m
r
r s
r
m
r
L
T
i
sT
L
β α α
ωψ ψ
1
(3)

+
+
=
m
r
r s
r
m
r
L
T
i
sT
L
α β β
ωψ ψ
1
(4)
( )
α β β α
ψ ψ
s r s r
r
m
m
i i
L
L
p T − ′ =
2
3
(5)
dt
d
J T T
m
z m
ω
= − ,
m
p ω ω ′ = (6)
where
2
2
1
r
m
r s
L
L
R R R + = ,
1
1
R
L
T
s

= ,
r
r
r
R
L
T = ,
r
m
r
L
L
k = ,
s s
L L σ = ′ ,
r s
m
L L
L
2
1− = σ (7)
i

– real axis stator current,
i

– imaginary axis stator current,
u

– real axis stator voltage,
u

– imaginary axis stator voltage,
ψ

– real axis rotor flux,
ψ

– imaginary axis rotor flux,
ω – rotor electrical angular speed,
ω
m
– mechanical angular speed,
T
m
– motor torque,
T
z
p’ – number of the pole pairs,
R
s
– stator resistance,
R
r
– rotor resistance,
L
s
– stator self inductance,
L
r
– rotor self inductance,
L
m
– magnetizing inductance,
J – moment of inertia.

For this model we need to identify the values of stator
and rotor resistance, stator and rotor inductance,
magnetizing inductance and moment of inertia. The two
parameters L
s
and L
r
are linearly dependent [3] and the
differences between their values are small. Therefore we
used simplification
s r
L L = , and so reduced the number
of the searched parameters and thereby accelerated the
research process. Searched motor parameters represent
genes of GA and together create the chromosome, which
has a form:
( ) J L L R R r
m s r s
, , , , = (8)

To represent these parameters we choose real number
values. In using real number code in comparison with
binary code the procedure of the respective solution is
more stable, since the values of real numbers change
continuously, proportionally to the required value of
change.
For each gene of the chromosome we must define the
feasible values intervals, i.e. the search space of solution.
The search space must be large enough to contain the
global optimum, but if made too large, the GA might not
be able to find the global optimum at the required
parameter identification accuracy and in a reasonable
period of time (the convergence will be slow). For this,
we can start from catalogue values.
The speed of convergence is also influenced by our
choice of the population size. The population represents
group of chromosomes, i.e. potential solutions in the
time period. The size of it depends on a particular case.
In most cases it is recommended to choose the size
between 10 and 100, most frequently between 20 and 50
[4]. Small population does not provide enough space for
diversity of genetic information, too big population does
not provide better effect and the solution is much longer.

The GA tests each candidate – chromosome by an
objective function, which compare a set of measured
values with simulated data. The objective function
consists of two steps. The first one is the implementation
of parameters into simulation model and the following
simulation of the dynamic system with such input data as
for measured system. During the simulation, the system
output is recorded in each sampling period and this
output data are then returned into the quality criterion
with measured output data. The next step is the
calculation of the appropriate criterion. The identification
values of the system parameters approximate to their real
values by the minimalization of this criterion. For the
parameter identification, the integral criterion is used:
( )

− =
T
m
dt y y F
0
2
(9)
where y
m
is output of the simulation model and y is
output of the real system.

In our case, the real system is represented by the
induction motor whose input parameters are the
components of the stator voltage u

and u

and output
parameters are the components of the stator current i

a i

. Then criterion has a form:
( ) ( )
∑ ∑
= =
− + − =
N
k
i s i s
N
k
i s i s
i i i i F
1
2
1
2
ˆ ˆ
β β α α
(10)
where
i s
i
α
and
i s
i
β
are currents from real motor,
i s
i
α
ˆ

and
i s
i
β
ˆ
are estimated currents from model, N is the
number of samples and F is the fitness function value.

The GA implementation for motor parameter
identification is illustrated in Fig. 1.

usα
Frequency
inverter
Simulation
model
Objectives
function
Genetic
algorithm
isα
isβ
F
Induction
motor usβ

α s
u

u
r
z
T
~
z
T
i

i
~

i
~

Fig. 1 Principle of parameter identification of
induction motor using GA

On the basic of fitness values of all candidates, some of
them are selected to the new population and some to the
work group. The operations of crossover and mutations
are applied only to the work group. The copying of the
best individuals into to next population insures the
convergence of solution. The multiple selection of the
best individuals into the work group insures good values
for the crossover and mutation accelerates the
convergence to the optimal values. But it can lead to
local optimum deadlock, therefore we use random
selection of several individuals into the work group for
more diversity during the realization of genetic
operations.

The convergence speed problems can also be at the last
phase of the algorithm, when the differences between
estimated and optimal values are small. For this we used
the multiplicative mutation, which multiplies the value of
the chosen gene by a random number from the defined
range.

The GA procedure can be described as follows:
1. Initially, a population of 50 chromosomes generated
at random in the given interval.
2. Successively, all 50 chromosomes are applied into
the simulation model. The sinusoidal voltages u

and u

with the same amplitude and frequency are
lead to the model input. After simulation taking 2
second a value of chosen criterion is evaluated. The
GA aims at its minimization.
3. The chromosomes with the smallest and second
smallest fitness value are copied into a new
population.
4. We copy into the working group: three times the
chromosome with the smallest fitness value, twice
the chromosome with the second smallest and once
the chromosome with the third smallest fitness
value.
5. 32 chromosomes are randomly copied into the
working group.
6. The crossover operator is applied to the working
group, where the choice of parents for the crossover
is random. This operator realizes exchange of genes
between parent strings.
7. The multiplicative mutation operator is applied to
the working group, where the value of the chosen
gene is multiplied by a random number from the
range (0, 2), with a mutation probability 0.1.
8. The intermediate crossover operator is applied on
the working group, where the choice of parents for
the crossover is random with parameter 25 , 1 = α .
This operator can be used for real number values. It
represents the combination between crossover and
mutation. New string will be created according to
equation:
) (
1 2 1
r r r p − + = α (11)
where p is a new string, r
1
and r
2
are the parent
strings and α is a random value from interval
25 , 1 25 , 0 < < α .
9. Then, the work group is moved to the new
generation.
10. The algorithm is repeated with the next generation
population until the 500 generations is reached.
Then the best string is assessed as the best solution.

3. IMPLEMENTATION OF GA AND
SIMULATION RESULT
To verify our proposed genetic algorithm the
Matlab/Simulink environment is used. We used the
simulation model instead of real induction motor because
of checking algorithm. We realized simulations with
sample time 0.25 ms and we defined the space of
solutions by the following intervals:
10 ; 1 ∈
s
R , 5 ; 1 ∈
r
R , 1 ; 1 . 0 ∈
s
L , 1 ; 1 . 0 ∈
m
L ,
1 , 0 ; 0001 . 0 ∈ J

We tested the application of designed GA for the
different amplitudes and frequencies of the input
voltages. The results of several identifications are
presented in Tab. 1, where the percentage errors of this
results and reference values are calculated. From the
table, you can see that the increase of the frequency
decreases the accuracy of parameter identification.

Tab. 1 Results of identification for the different
amplitudes and frequencies of the input voltages
Rs Rr Ls Lm J

Reference
value
7.608 3.7 0.6015 0.5796 0.0017
Identified
value
7.5954 3.7093 0.60227 0.58047 0.0017059
7.5V, 5Hz
F=1.5062x10
-3
Relative
error
0.166% 0.251% 0.127% 0.150% 0.350%
Identified
value
7.5587 3.7269 0.60204 0.58028 0.0017174
15V, 10Hz
F=1.6439x10
-2
Relative
error
0.648% 0.726% 0.090% 0.117% 1.021%
Identified
value
7.4066 3.8124 0.62926 0.60741 0.0017529
30V, 20Hz
F=1.8171 Relative
error
2.647% 3.039% 4.616% 4.799% 3.111%
Identified
value
7.2382 3.9168 0.70633 0.68436 0.0018073
45V, 30Hz
F=15.3852 Relative
error
4.861% 5.858% 17.43% 18.07% 6.311%

In the case of parameter identification of real motor, the
results are influenced by the noise of the measured
signals and by the quantization of the signals. Therefore
we tested the ability of the parameter identification by
GA also in this case. As the source of the noise we used
white noise with the gain 1.10
-6
. From the result of the
identifications in Tab. 2 you can see that the noise and
quantization didn’t have a great effect on the accuracy of
parameter identification compared with the results in
Tab. 1. The Fig. 2 and 3 describe the confrontation
between reference and identified currents for the noised
and quantized signals.

Tab. 2 Results of identification for the different
amplitudes and frequencies of the input voltages for
the noised and quantized reference currents
Rs Rr Ls Lm J

Reference
value
7.608 3.7 0.6015 0.5796 0.0017
Identified
value
7.6558 3.7203 0.59966 0.57804 0.0016947
7.5V, 5Hz
F=65.798 Relative
error
0.628% 0.548% 0.305% 0.270% 0.310%
Identified
value
7.5915 3.7133 0.5987 0.57701 0.0017053
15V, 10Hz
F=66.0976 Relative
error
0.217% 0.359% 0.465% 0.446% 0.312%
Identified
value
7.5184 3.7514 0.61277 0.59093 0.001728
30V, 20Hz
F=66.0636 Relative
error
1.178% 1.389% 1.873% 1.955% 1.644%
Identified
value
8.1683 3.3823 0.49211 0.47021 0.0015378
45V, 30Hz
F=97.9766 Relative
error
7.364% 8.587% 18.19% 18.87% 9.543%

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-1.5
-1
-0.5
0
0.5
1
1.5
2
t [s]
i
s
α

[
A
]

reference model
identified model

Fig. 2 Stator current i

responses for the noised and
quantized reference signals
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-1.5
-1
-0.5
0
0.5
1
1.5
t [s]
i
s
β

[
A
]

reference model
identified model

Fig. 3 Stator current i

responses for the noised and
quantized reference signals

The accuracy of parameter identification can depend on
the sample period. The size of the sample period depends
on the available computing resources. The increase of the
frequency of the input voltage needs the decrease of the
sample period. The results of parameter identification for
different sample periods are presented in the Tab. 3. For
the frequency of the input voltage equal to 5 Hz we
obtained the error less than 1% also with sample period
equal to 2.5 ms. For frequency 10 Hz we needed for this
accuracy the sample period 2 ms and for frequency 20
Hz we need the sample period 0.5 ms. For frequency 30
Hz we obtained the error less than 5% with the sample
period 0.2 ms.

Tab. 3 Results of identification for the different
amplitudes and frequencies of the input voltages
and for the different sample periods
Rs Rr Ls Lm J

Reference
value
7.608 3.7 0.6015 0.5796 0.0017
Identified
value
7.5974 3.684 0.59917 0.5772 0.0016948 7.5V, 5Hz
Tvz=2.5ms
F=1.3542x10
-3

Relative
error
0.139% 0.432% 0.388% 0.414% 0.305%
Identified
value
7.597 3.7073 0.60399 0.58205 0.0017019 15V, 10Hz
Tvz=2ms
F=1.5665x10
-3

Relative
error
0.145% 0.197% 0.414% 0.423% 0.109%
Identified
value
7.628 3.6868 0.59853 0.57663 0.0016941 30V, 20Hz
Tvz=0.5ms
F= 8.2013x10
-3

Relative
error
0.263% 0.356% 0.494% 0.512% 0.344%
Identified
value
7.8813 3.5953 0.59221 0.57026 0.0016332 45V, 30Hz
Tvz=0.2ms
F=2.8285
Relative
error
3.593% 2.829% 1.545% 1.612% 3.932%

4. EXPERIMENTAL RESULTS
After the simulation verification we applied the designed
genetic algorithm to the real system. The real system is
represented by a computer with development software
frequency inverter and induction motor. The computer is
connected with the frequency inverter by interface
DS1104 from company dSPACE. Analogous to the
simulations experiments we realized identification for the
different amplitudes and frequencies of the input
voltages. The results of several identifications are
presented in Tab. 4, where the percentage errors of this
results and values of identification by software for
frequency inverter are calculated. Fig. 4 to Fig. 7 show
the comparison of stator currents i

and i

responses
from the real system and identified model for the input
voltage with frequency 25 Hz. You can see in Tab. 4.,
that the results of the identification of the stator
resistance are situated at the edge of the interval. The
differences between the values of the other parameters
may be caused by non-linearity of the frequency inverter
and motor structures and also by unknown load torque.
Therefore we repeat the experiments with the addition of
the parameters of Coulomb friction (T
z0
) and viscous
friction (B

) into the motor model:
ω ω B T T
z z
′ + = sgn
0
(12)
and into the chromosome:
( ) B T J L L R R r
z m s r s
′ = , , , , , ,
0
(13)
For new parameters we defined the following intervals:
1 . 0 ; 0001 . 0
0

z
T , 1 . 0 ; 0001 . 0 ∈ ′ B

Tab. 5 shows the results of identifications of (13) for the
different amplitudes and frequencies of the input
voltages and Fig. 8 to Fig. 11 show the comparison of
stator current i

and i

responses from the real system
and identified model for the input voltage with frequency
25 Hz. We can see that the results of the identification of
the stator resistance are better than the results of
identifications of (8) and the differences between stator
current i

responses are smaller. But the differences
between the results of the identification of Coulomb and
viscous friction are too large.

Tab. 4 Experimental results of identification of (8)
for the different amplitudes and frequencies of the
input voltages
Rs Rr Ls Lm J

Identification
by frequency
inverter
6.3783 3.9558 0.3796 0.3502 0.0068
Identi fied
value
10 2.8509 0.37274 0.31881 0.00325
15V, 5Hz
F=54.8578 Relative
error
56.78% 27.93% 1.81% 8.96% 52.21%
Identi fied
value
10 4.2446 0.42614 0.38074 0.00411
30V, 10Hz
F=313.1876 Relative
error
56.78% 7.30% 12.26% 8.72% 39.56%
Identi fied
value
10 4.4441 0.41365 0.37274 0.00485
35V, 15Hz
F=559.2084 Relative
error
56.78% 12.34% 8.97% 6.44% 28.68%
Identi fied
value
10 4.5360 0.40418 0.36589 0.00474
50V, 20Hz
F=639.4507 Relative
error
56.78% 14.67% 6.47% 4.48% 30.29%
Identi fied
value
9.2844 4.8792 0.38927 0.35272 0.00483
70V, 25Hz
F=647.8662 Relative
error
45.72% 23.34% 2.55% 0.72% 28.97%

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-5
-4
-3
-2
-1
0
1
2
3
4
5
t [s]
i
s
α

[
A
]

real system
identified model

Fig. 4 Stator current i

responses for the input
voltage with frequency 25 Hz
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
t [s]
e
r
r
o
r

i
s
α

[
A
]

Fig. 5 Error of stator current i

responses for the
input voltage with frequency 25 Hz
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-5
-4
-3
-2
-1
0
1
2
3
4
5
t [s]
i
s
β

[
A
]

real system
identified model

Fig. 6 Stator current i

responses for the input
voltage with frequency 25 Hz
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
t [s]
e
r
r
o
r

i
s
β

[
A
]

Fig. 7 Error of stator current i

responses for the
input voltage with frequency 25 Hz

Tab. 5 Experimental results of identification of (13)
for the different amplitudes and frequencies of the
input voltages
Rs Rr Ls Lm J Tz0 B´

Identification
by frequency
inverter
6.3783 3.9558 0.3796 0.3502 0.0068 - -
Identified
value
7.7701 4.1915 0.40205 0.36616 0.00569 0.00484 0.02276 15V,
5Hz
F=26,35
Relative
error
21.82% 5.96% 5.91% 4.56% 16.32% - -
Identified
value
8.2009 4.7236 0.44727 0.40803 0.00471 0.02623 0.01077 30V,
10Hz
F=44,17
Relative
error
28.58% 19.41% 17.83% 16.51% 30.74% - -
Identified
value
7.6785 5.011 0.45112 0.4138 0.00521 0.01970 0.00715 35V,
15Hz
F=34,85
Relative
error
20.38% 26.67% 18.84% 18.16% 23.38% - -
Identified
value
7.4173 5.3356 0.43607 0.39994 0.00543 0.03162 0.00559 50V,
20Hz
F=36,99
Relative
error
16.29% 34.88% 14.88% 14.20% 20.15% - -
Identified
value
7.7299 5.276 0.40314 0.36761 0.00504 0.01806 0.00479 70V,
25Hz
F=75,63
Relative
error
21.19 33.37% 6.20% 4.97% 25.88% - -

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-5
-4
-3
-2
-1
0
1
2
3
4
5
t [s]
i
s
α

[
A
]

real system
identified model

Fig. 8 Stator current i

responses for the input
voltage with frequency 25 Hz

5. CONCLUSION
The results of our experiment show that genetic
algorithms can be used for the parameter identification of
induction motor. The GA obtains more accurately values
of searched parameters using serial reproduction of the
new solutions and by using the objective function that
compares the current components from the simulation
and real systems.

The accuracy of identification depends on the frequency
of the input voltage, the sample period, the number of
generations and the accuracy of simulation model. Our
simulation results where very good but the experimental
results were influenced by the differences between
simulation model and real system. It’s better if we can
exactly describe the real system by the simulation model.
It’s possible for identification by GA. The simulation
model can consider the non-linearity of frequency
inverter and pulse wide modulation and so we can obtain
better parameters of the real system.

The disadvantage of the genetic algorithms is their time
demand that follows mainly from a number of required
simulations.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
t [s]
e
r
r
o
r

i
s
α

[
A
]

Fig. 9 Error of stator current i

responses for the
input voltage with frequency 25 Hz
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-5
-4
-3
-2
-1
0
1
2
3
4
5
t [s]
i
s
β

[
A
]

real system
identified model

Fig. 10 Stator current i

responses for the input
voltage with frequency 25 Hz
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
t [s]
e
r
r
o
r

i
s
β

[
A
]

Fig. 11 Error of stator current i

responses for the
input voltage with frequency 25 Hz

ACKNOWLEDGEMENT
This work has been carried out under the financial
support of the KEGA project 3/4196/06 and RLZ VVA
project no. 13120200015 (Europe Social Program).

REFERENCES
[1] Jančovič, M., Žalman, M.: Parameter identification of
induction motor by genetic algorithm. AT&P Journal.
2006, no. 2, pp. 90-96.

[2] Žalman, M.: Actuating devices. STU Publishing house,
Bratislava, 2003.

[3] Ursem, R. K.: Models for Evolutionary Algorithms and
Their Application in System Identification and Control
Optimization. PhD thesis, EVALife, Department of
Computer Sience, Univerzity of Aarhus, 2003.

[4] Sekaj, I.: Evolution computations and their practice using.
IRIS Publishing house, Bratislava, 2005.

[5] Chrzan, P., J., Klaassen, H.: Parameter identification of
vector-controlled induction machines. Electrical
Engineering, vol. 79 (1996), pp. 39-46.

[6] Chung, P. Y., Dölen, M., Lorenz, R. D.: Parameter
Identification for Induction Machines by Continuous
Genetic Algorithms. ANNIE 2000 Conference, St. Louis,
November 2000.

[7] Huang, K., S., Wu, Q., H.: Effective Identification of
Induction Motor Parameters Based on Fewer
Measurements. IEEE Transaction on Energy Conversion,
vol. 17, no. 1, March 2002, pp. 55-60.

THE AUTHORS

Marián Jančovič (Ing) was born in Piešťany, in 1976.
He graduated from the Faculty of Electrical Engineering
and Information Technology, Slovak University of
Technology (FEI STU), Bratislava, in 2001. Since 2001
he has been a PhD student in the Department of Control
and Automation FEI STU. His research interests include
neural networks, genetic algorithms and the control of
electrical drive.

Milan Žalman (Prof, Ing, CSc) was born in Necpaly, in
1945, graduated from the Faculty of Electrical
Engineering of the Slovak University of Technology,
Bratislava, in 1968. At the same university he received
his CSc (PhD), in 1981. Since 1987, appointed Associate
Professor, and from 1999 as a Full Professor, is at the
Department of Control and Automation FEI STU
Bratislava. His research and teaching interests include
the control of electrical drive and motion control
systems.

Jan Jovankovič (Ing, PhD) was born in 1973 in Novi