See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/260155231

Tuning and Digital Implementation of a Fractional-Order PD Controller for a

Position Servo

Article · January 2013

CITATIONS READS

21 1,140

3 authors:

Aleksei Tepljakov Eduard Petlenkov

Tallinn University of Technology Tallinn University of Technology
140 PUBLICATIONS 2,720 CITATIONS 253 PUBLICATIONS 3,474 CITATIONS

SEE PROFILE SEE PROFILE

Juri Belikov
Tallinn University of Technology
241 PUBLICATIONS 3,843 CITATIONS

SEE PROFILE

All content following this page was uploaded by Aleksei Tepljakov on 18 March 2015.

The user has requested enhancement of the downloaded file.

 1

Tuning and Digital Implementation of a

Fractional-Order PD Controller for a Position Servo
Aleksei Tepljakov, Eduard Petlenkov, Juri Belikov, Sergei Astapov
Department of Computer Control
Tallinn University of Technology
Ehitajate tee 5, 19086, Tallinn, Estonia
{aleksei.tepljakov, eduard.petlenkov, juri.belikov, sergei.astapov}@ttu.ee

Abstract—Fractional-order calculus offers flexible computa- a controller was obtained for a position servo model. In this
tional possibilities that can be applied to control design thereby work we complement these results by summarizing the design
improving industrial control loop performance. However, before and implementation methods for an equivalent controller,
theoretical results can be carried over to an industrial setting
it is important to study the effects of fractional-order control namely a fractional-order PDµ controller, using the FOMCON
by means of laboratory experiments. In this paper, we study toolbox control design module. Thereby, in addition, we ex-
the practical aspects of tuning and implementing a fractional- tend the results in [19], [20]. We also confirm the effectiveness
order PD controller for position control of a laboratory modular of these methods experimentally using a modular servo system
servo system using FOMCON (“Fractional-order Modeling and provided by INTECO [21].
Control”) toolbox for MATLAB. We provide an overview of the
tools used to model, analyze, and design the control system. The The paper is organized as follows. In Section II the reader is
procedure of tuning and implementation of a suitable digital introduced to fractional-order control, fractional-order system
fractional-order controller is described. The results of the real- implementation method, and the corresponding software tools
time experiments confirm the effectiveness of used methods. used to design and realize a digital fractional-order controller.
Index Terms—fractional calculus, position servo, pid controller, The description of the controlled servo system is provided in
digital control, control optimization Section III. In Section IV we provide the steps necessary to
design and realize a suitable controller for the position servo
system. An overview of the experimental platform and the
I. I NTRODUCTION
results of real-time closed loop control are given in Section V.
Fractional-order calculus is the generalization of conven- Some items for discussion are outlined in Section VI. Finally,
tional calculus, where the order α of integration or dif- conclusions are drawn in Section VII.
ferentiation is not restricted to integer numbers [1]. This
generalization offers interesting modeling possibilities. The II. F RACTIONAL - ORDER C ONTROL
number of applications where fractional-order calculus is used A. Introduction to Fractional-order Control
has been growing steadily in the last years [2].
In the heart of fractional-order modeling lies the generalized
Non-integer calculus is actively used in the field of control
non-integer order fundamental operator
system design [3], [4]. Novel modeling opportunities allow to 
α α
design efficient linear and nonlinear control strategies [2]. It is d /dt
 α > 0,
a Dt =
a well known fact, that PID-type controllers are ubiquitous in α
1 α = 0, (1)
the industry [5], [6]. However, a conventional PID controller ´ t
 −α
a
(dτ ) α < 0,
is inferior to a fractional-order PID controller due to extended
tuning flexibility of the latter. This was experimentally con- where a and t denote the limits of the operation. The case
firmed in, e.g., [7], [8], [9]. α ∈ Z corresponds to conventional differentiation or integra-
Computer Aided Control System Design (CACSD) tools tion. There exist several definitions of the generalized operator.
are readily available to assist engineers in the task of de- Next, we provide the Grünwald-Letnikov definition [2], [3]:
veloping suitable controllers for particular plants. Notable 1 X
k  
α
a Dt f (t) = lim α
α
examples include CRONE [10] and Ninteger [11] toolboxes (−1)j f (t − jh), (2)
h→0 h j
for MATLAB/Simulink software. The FOMCON (“Fractional- j=0
order Modeling and Control”) toolbox [12], [13] was recently where a = 0, t = kh, k is the number of computation steps
developed to further expand the existing toolset as well as to and h is step size. We assume zero initial conditions and thus
provide new features. the Laplace transform of the fractional α-order derivative is
In our previous work [14], [15], we focused primarily on ˆ ∞
the problem of extending an existing implementation technique e−st 0 Dtα f (t)dt = sα F (s), (3)
[16], [17] to achieve a frequency bounded approximation of a 0
fractional-order lead compensator similar to, e.g., Oustaloup’s where α ∈ R+ and s = σ + jω is the Laplace transform
method [4], [10], [18], and provided an example where such variable.
 2

The parallel form of the fractional PIλ Dµ controller is given where K is the static gain, |α| < 1 is the non-integer power.
in the following equation: Coefficients b and a are related to zero frequency ωz = 1/b
Ki and pole frequency ωh = 1/a for α > 0. Generally, this
+ Kd · sµ .
C(s) = Kp + (4) transfer function corresponds to a frequency bounded non-
sλ
integer differentiator (integrator) [15]. Therefore, the integral
In the frequency domain this controller offers more tuning flex-
and differential components of the FOPID controller in (4)
ibility. In general, by varying the order γ of a fractional-order
may be implemented using (5).
integrator (differentiator) a constant decrement (increment) in
the slope of the magnitude curve that equals 20γ dB/dec can
B. Fractional-order System Implementation
be achieved, as well as a constant delay in the phase plot
πγ/2 rad, where the values depend on the sign of γ. The In this work we turn our attention to the Oustaloup ap-
effects of control actions in the time domain corresponding to proximation method which is frequently used for practical
a fractional-order integrator and differentiator are illustrated in implementations of fractional-order systems and controllers
Fig. 1 and Fig. 2, respectively. [2], [3], [18]. A revised version of this method was proposed
in [27]. We restrict our attention to the original approximation
algorithm. In order to approximate a fractional differentiator
5
of order α or a fractional integrator of order (−α) one can
γ=0
4.5
γ=−0.5 use the following set of equations:
γ=−0.7
4 N
γ=−1 Y s + ωk0
3.5 sα ≈ K , (6)
s + ωk
3 k=1
where
u(t)

2.5

2
ωk0 = ωb · ωu(2k−1−α)/N , (7)
1.5
ωk = ωb · ωu(2k−1+α)/N , (8)
1 p
K = ωhα , ωu = ωh /ωb , (9)
0.5

0 and N is the order of approximation in the frequency range

0 1 2 3 4 5 6 7 8 9 10
Time [s] (ωb ; ωh ). The order of the resulting approximation is 2N + 1.
Taking a higher order N generally results in a more accurate
Fig. 1. Fractional integrator s−γ control actions approximation, though equations relating the parameters in
(6)–(9) to N exist as well and may be found in [18].
A suitable discretization method can be used to obtain a
2.5
discrete-time approximation from the continuous one in (6).
One possible method would be that of zero-pole matching
γ=0
2
γ=0.5 equivalents, where direct mapping of continuous zeros and
γ=0.7
1.5 γ=1
poles to discrete-time is done by means of the relation
1 z = esTs , (10)
where Ts is the desired sampling interval. The gain of the
u(t)

0.5

resulting discrete-time system H(z) must be corrected by a

0
proper factor. This implementation method has been success-
−0.5 fully used in our previous work [28], [29]. We remark, that
−1
for the synthesis of continuous zeros and poles in (6) with the
intent to obtain a discrete-time approximation the transitional
−1.5
0 1 2 3 4 5 6 7 8 9 10 frequency ωh may be chosen such that
Time [s]
2
ωh 6 . (11)
Fig. 2. Fractional differentiator sγ control actions Ts
After acquiring a set of discrete-time zeros and poles by
Various fractional-order PIλ Dµ controller tuning methods means of (10), the fractional-order controller may be imple-
have been proposed to date [9], [22], [23], [24], [25]. One mented in form of a IIR filter represented by a discrete-time
possible general approach is based on constrained optimization transfer function H(z −1 ). In general, one has two choices:
subject to particular control system design specifications and 1) Implement each fractional-order component approxima-
has been investigated in [26]. tion of the controller in (4) separately as H λ (z −1 ) and
Additionally, consider the transfer function corresponding H µ (z −1 ); this method offers greater flexibility, since
to a fractional-order lead (lag) compensator: the components may be reused in the digital signal
α
processing chain, but requires more memory and is

bs + 1
G(s) = K · , (5) generally more computationally expensive;
as + 1
 3

2) Compute a single LTI object approximating the whole • Fractional-order transfer function model identification
controller; this method is suitable when there is a need tool, calling sequence: fotfid. The graphical user in-
for a static description of a fractional-order controller, terface of the tool is presented in Fig. 4. Since fractional-
e.g., for a given control task. order calculus is viewed as a generalization of the usual
In this particular work we choose the second option, that is calculus operators in this context, the tool can also be
we seek a description of the controller in the form used to determine the parameters of classical, integer-
order systems. In particular, we are interested in identi-
b0 + b1 z −1 + b2 z −2 + · · · + bm z −m fying process models, a feature which the present tool
H(z −1 ) = K . (12)
a0 + a1 z −1 + a2 z −2 + · · · + an z −n fully supports.
For practical reasons, the equivalent IIR filter should be • Fractional PID controller design front-end, calling se-
comprised of second-order sections, since this allows to im- quence: fpid. A negative unity feedback connection is
prove computational stability, especially when the target signal assumed, i.e.
digital processing hardware has limited floating-point type C(s)G(s)
resolution and operation support [28], [30]. Thus, the discrete- Gc (s) = ,
1 + C(s)G(s)
time controller must be transformed to yield
where C(s) is the fractional-order PID controller and
N
Y 1 + b1k z −1 + b2k z −2 G(s) is the linear plant to be controlled. The tool has
H(z −1 ) = b0 . (13)
1 + a1k z −1 + a2k z −2 further links to the optimization and implementation
k=1
tools.
The form also easily lends itself to stability analysis re- • Optimization tool, calling sequence: fpid_optim. The
gardless of the method used to generate the coefficients of tool has been updated substantially since [12], the new
the second-order sections. Here we assume that computational graphical front-end illustrating the added features is given
stability is guaranteed with a specified precision. Then, in in Fig. 5. A thorough description of the tool is provided
order to determine whether a section is stable or not, we in [26]. One of the most notable features is the possibility
consider its discrete-time pole polynomial to use Simulink for simulation of nonlinear effects, such
p(z −1 ) = 1 + d1 z −1 + d2 z −2 . (14) as actuator saturation, and/or nonlinear plants. In this
regard, the nonlinear simulation option aims to fill the gap
A single section in (13) is stable [31], if the following between theoretical controller design results and practical
conditions, derived from the classical case of discrete-time control applications.
system stability analysis, are met in the (d1 , d2 )-plane (see • Implementation tools, calling sequence: impid,
Fig. 3 for a visual reference): d2sos(). The former allows to choose suitable
controller parameters for a discrete approximation, while
|d1 | < 1 + d2 , |d2 | < 1. (15) the latter may be used to directly obtain C language style
Note, that in case of the method in (6) for any non- IIR filter second-order arrays from the approximated
integer order α ∈ R, |α| < 1, the obtained approximation discrete controller.
is always stable [18] and after the subsequent discretization
procedure the rule (15) should also be satisfied for every III. D ESCRIPTION OF THE S ERVO S YSTEM
section comprising the filter in (13).
In this work we use the modular servo system provided
by INTECO [21]. This laboratory system is reconfigurable
and can be used for a variety of control experiments. We
consider the particular configuration depicted in Fig. 6. The
plant consists of the following modules: tachogenerator, which
is used to measure the rotational speed of the DC motor; inertia
load, backlash, incremental encoder, and gearbox with output
disk. Data acquisition and real-time experiments are done
using a specific PCI board, which connects to the servo system
power interface and also collects data from the tachogenerator,
incremental encoder and reference potentiometer (the latter is
not shown in the figure).
Fig. 3. The stability triangle The mathematical model of the servo system is based on
that of the DC motor. The first-order inertial system, where
static and dry kinetic friction and saturation are neglected, is
C. Fractional-order Controller Design Tools described by the equation
The tools in the identification, control design, and imple- Ts ω̇(t) = −ω(t) + Ksm v(t), (16)
mentation modules pertaining to the present work are briefly
described next. The corresponding MATLAB calling sequence where Ts is the time constant of the motor, ω(t) is the angular
is provided. velocity of the rotor, Ksm is the motor gain and v(t) is the
 4

Fig. 6. Configuration of the laboratory servo system

The backlash, found in many real mechanical systems [32],

introduces play into the system. In this plant the output dead-
zone of the backlash is close to 2π. This nonlinearity has to
be specifically considered when designing a suitable controller
Fig. 4. Graphical user interface of the fractional-order transfer function model for the system—the procedure which we will investigate in the
identification tool
next section.
Finally, the apparent dead-zone of the motor under the
voltage input has been identified as ud = (−0.05, 0.05).

IV. T UNING AND I MPLEMENTATION OF THE C ONTROLLER

In order to design the controller using the tools described in
Section II we first need to obtain the model of the plant given
by (19). This is possible by first identifying the model in (18)
in the time domain from a step experiment, and then simply
adding an integrator to arrive at the desired model. Using the
time-domain identification tool, described in Section II, we
obtain the following integer-order process model:
192.1638
G(s) = . (20)
s(1.001s + 1)
In what follows, this transfer function serves as the basis for
controller design using the optimization tool.
For a model of the form (19) it is natural to assume that
a lead compensator may be necessary for establishing the re-
quired closed-loop performance. Recall, that a fractional-order
Fig. 5. Graphical user interface of the fractional-order PID controller
optimization tool lead compensator corresponds, in principle, to a frequency-
bounded approximation of a PDµ controller, therefore our task
is to establish an appropriate set of parameters (Kp , Kd , µ).
input voltage. The input voltage in case of this laboratory plant The generic parameters for the parallel form of this con-
is normalized and is such that troller provided by INTECO are Kp = 0.1, Kd = 0.01. We
u(t) = v(t)/vmax , |u(t)| 6 1. (17) shall use these parameters as the initial ones for the design
of a suitable fractional-order PD controller, which we shall
Also let Ks = Ksm · vmax . Then, the corresponding velocity accomplish by means of constrained optimization. Following
transfer function is given by the controller design strategy applied in [26], [28], with design
Ks specifications of 1% set-point tracking accuracy and a mini-
Gv (s) = , (18) mum phase margin ϕm = 60◦ , we proceed to construct the
Ts s + 1
and the angle transfer function is obtained by adding an nonlinear Simulink model for time-domain simulation of the
integrator control system. Additionally, input disturbance is considered,
Ks as well as the motor control dead-zone and the backlash
Ga (s) = . (19) component. The resulting model is presented in Fig. 7.
s(Ts s + 1)
 5

Out1
The accuracy requirement as well as the phase margin
Impulse u=0.25
specifications have been satisfied. In the frequency domain
(30s to 30.5s)
the approximation of the PDµ controller corresponds to the
50 e u fpid_optimize_lti 1
Out1
fractional lead compensator in (5).
r Dead Zone Saturation Backlash
Fractional PID controller Plant LTI
2
Out2
Bode Diagram

Scope
40

Fig. 7. Simulink model used for controller optimization with added nonlin- 30

earities and input disturbance 20

Magnitude (dB)
10

−10

Other optimization parameters are set in the following −20

Ideal PDµ controller
−30 Approximated PDµ controller
way: performance metric is IAE (integral absolute error), −40

the linear model of the plant is used to compute the con- 90

straints in the frequency domain and the approximation is 60

Phase (deg)
obtained by means of the Oustaloup filter with parameters
ω ∈ [0.0001, 10000], N = 5. Actuator saturation is considered 30

and is such, that u(t) ∈ [−1, 1]. Simulation stop time is 60

0
seconds. Two optimization steps are considered: 10
−4 −3
10
−2
10
−1
10 Frequency 0
10 (rad/s)
1
10 10
2
10
3 4
10

• Fixing µ = 1 we obtain optimized integer-order PD

controller parameters Kp and Kd .
Fig. 9. Frequency domain characteristics of an ideal PDµ controller vs. its
• Fixing the gains at the obtained values, we search for an frequency-bounded approximation
optimal order µ.
This method minimizes the number of optimization variables Finally, we can obtain a digital implementation of this
thereby improving the optimization speed. The results of controller. The target hardware is a simple 8 bit Atmel AVR
optimization are such, that after 100 iterations the gains of ATmega8A based microcontroller prototype. The microcon-
the PD controller have been found as Kp = 0.055979 and troller connects to external 12 bit analog-to-digital and digital-
Kd = 0.025189. After fixing the gains and setting the initial to-analog converters by means of the I2 C interface. Basic
value of µ to 0.5, the optimized PDµ controller is obtained signal conditioning circuits are employed.
with µ = 0.88717. Phase margin of the open loop control Continuous Oustaloup filter approximation parameters are
system is ϕm = 65.3◦ . the same as were used during controller optimization. The
The fractional-order PDµ controller has been found to ex- ’matched’ method in 10 is used to obtain a discrete-time
hibit superior performance than the initially obtained integer- approximation of the controller with sampling interval Ts =
order PD controller, especially in the presence of aforemen- 0.01. The d2sos() function is used to directly obtain the
tioned nonlinearities. The comparison of simulated transient IIR filter second-order section coefficients, provided next:
responses of the servo control system with initial generic
integer-order PD controller and the optimized fractional PDµ b = {+1.0000000000, −0.9647855878, +0.0000000000} ,
controller is given in Fig. (8).
{+1.0000000000, −0.0209224276, +0.0000000000} ,
70
{+1.0000000000, −1.3493207288, +0.4180066451} ,
60
Integer−order PD controller
{+1.0000000000, −1.9807306143, +0.9807890156} ,
50 {+1.0000000000, −1.9991305017, +0.9991306026} ,
Angle [rad]

40
{+1.0000000000, −1.9999692428, +0.9999692429} ,
30

20

10
Integer−order PD controller a = {+1.0000000000, −0.0000000000, +0.0000000000} ,
Fractional−order PDµ controller

0
0 10 20 30 40 50 60
{+1.0000000000, −0.0409802515, +0.0000000016} ,
1
{+1.0000000000, −1.4434599048, +0.4912545169} ,
Control law u(t)

0.5

0
{+1.0000000000, −1.9752697983, +0.9753515564} ,
−0.5 {+1.0000000000, −1.9991239831, +0.9991240851} ,
−1
0 10 20 30 40 50 60
{+1.0000000000, −1.9999692318, +0.9999692319} ,
Disturbance

0.4

0.2

0
b0 = 1.5336084022.
0 10 20 30 40 50 60
Time [s]
These IIR filter coefficient arrays are hard-coded into the
Fig. 8. Performance of the initial integer-order PD controller vs. the optimized microcontroller memory. We remark, that the presented coeffi-
PDµ controller in the presence of an input disturbance cient resolution will not be utilized in full by a single-precision
 6

floating number format used in DSP operations running on the B. Real-time Control Results
microcontroller. Three experiments are considered:
In the next section we describe the experimental platform
• Evaluation of performance of the controller implemented
and provide the results of real-time control experiments that
in Simulink;
verify the proposed implementation.
• Evaluation of performance of the external digital con-
troller;
V. E XPERIMENTAL R ESULTS • Evaluation of external controller set-point tracking.
A. Description of the Experimental Platform The first two experiments are grouped so that a comparison
In order to validate the performance of the digital controller, can be made. Additionally, a similar experiment is conducted
the configuration depicted in Fig. 10 is used. Apart from the with the initial integer-order PD controller for reference. The
servo system, we use an updated version of the serial com- third experiment is done for the external controller. Set-point
munication based DAQ board, discussed in [14] in connection changing in this case is done by means of the potentiometer
with controller prototyping. It offers two input and two output disk of the servo system.
channels with 12 bit sample resolution and 2.5kSPS theoretical The results of the first set of experiments are presented in
full-duplex real-time sampling rates on both channels with at Fig. 12. The control system responses obtained from using the
most a single sample delay. Unfortunately, said performance fractional PDµ controller implemented as a Simulink block
will vary depending on the hardware configuration of the and the external controller match up. A small discrepancy is
personal computer used. caused by a voltage offset error. The control law exhibits limit
cycles due to, one hand, quantization [33] caused by finite
word length of the A/D and D/A converters, and on the other
hand by measurement noise. Some noise is naturally present
in the analog circuit. However, the amplitude of these limit
cycles falls inside the dead zone of the control signal, so they
do not have any major effect on the control system.

180

160

140

120

100
Angle [rad]

80

60
System with integer−order PD controller
40 System with PDµ controller (in Simulink)
20 System with PDµ controller (external)
Reference angle
0

−20
0 10 20 30 40 50 60
1

Fig. 10. Real-time closed loop control evaluation platform 0.5

Control law u(t)

For the experiment it is assumed, that the controller receives

−0.5
the error encoded in a voltage signal of amplitude 0 . . . 5V.
−1
Virtual ground with +2.5V reference is used to encode the 0 10 20 30 40 50 60

0.4
Disturbance

negative error and control signal. The values are scaled ac- 0.2
cordingly. The voltage supply is reasonably well filtered. 0
0 10 20 30 40 50 60
The general Simulink diagram for experiments with the ex- Time [s]

ternal controller is given in Fig. 11. In case of all experiments,

the PC is running MATLAB/Simulink and specific Real-Time Fig. 12. Comparison of control systems with different controllers and
configurations
Windows Target drivers for the serial DAQ board and the
modular servo system. The result of manual set-point change experiment with
the external controller is presented in Fig. 13. The experi-
Out1 ment confirms the expected controller performance. We can
Impulse u=0.25
(20s to 20.5s)
Angle
conclude, that apart from small discrepancies, the hardware
e u
Angle (encoder) Velocity implementation of the PDµ controller is working correctly
Control law
Step
Discrete fractional
PID controller
Saturation
1
0.02s+1
within the desired performance specifications.
[External] Disturbance
Simple filter Scope

Reset Manual angle setting

Reset
encoders
0 Modular servo system
VI. D ISCUSSION
Disable manual
Manual Switch set−point setting
Gain The methods of digital controller design and implementation
10
provided in this paper were successfully verified by real-time
Fig. 11. Real-time closed loop control Simulink diagram control loop experiments. During this process some issues
were identified.
 7

[5] K. Åström and T. Hägglund, PID Controllers: Theory, Design, and

250
Tuning, 2nd ed. The Instrumentation, Systems, and Automation Society
200 (ISA), 1995.
150 [6] ——, Advanced PID control. The Instrumentation, Systems, and
Angle [rad]

100 Automation Society (ISA), 2006.

50
[7] M. Čech and M. Schlegel, “The fractional-order PID controller outper-
Measured angle forms the classical one,” in Process control 2006. Pardubice Technical
0
Reference angle University, 2006, pp. 1–6.
−50
0 20 40 60 80 100 120 [8] D. Xue, C. Zhao, and Y. Chen, “Fractional order PID control of a DC-
1 motor with elastic shaft: a case study,” in Proc. 2006 American Control
Conference (ACC), 2006.
0.5
[9] C. Monje, B. Vinagre, V. Feliu, and Y. Chen, “Tuning and auto-
Control law u(t)

0
tuning of fractional order controllers for industry applications,” Control
Engineering Practice, vol. 16, no. 7, pp. 798–812, 2008.
−0.5 [10] A. Oustaloup, P. Melchior, P. Lanusse, O. Cois, and F. Dancla, “The
CRONE toolbox for Matlab,” in Proc. IEEE Int. Symp. Computer-Aided
−1
0 20 40 60 80 100 120
Control System Design CACSD 2000, 2000, pp. 190–195.
Time [s] [11] D. Valério. (2005) Toolbox ninteger for MatLab, v. 2.3. [Online].
Available: http://web.ist.utl.pt/duarte.valerio/ninteger/ninteger.htm
[12] A. Tepljakov, E. Petlenkov, and J. Belikov, “FOMCON: a MATLAB
Fig. 13. External PDµ controller set-point tracking evaluation toolbox for fractional-order system identification and control,” Interna-
tional Journal of Microelectronics and Computer Science, vol. 2, no. 2,
pp. 51–62, 2011.
• Limit cycles exist in the control law of the hardware [13] ——. (2011) FOMCON toolbox. [Online]. Available:
http://www.fomcon.net/
controller due to noise and quantization effects. However,
[14] ——, “Implementation and real-time simulation of a fractional-order
in an industrial application these will be minimized if controller using a MATLAB based prototyping platform,” in Proc. 13th
a suitable data acquisition configuration tailored to the Biennial Baltic Electronics Conference, 2012, pp. 145–148.
specific task is used. [15] ——, “Application of Newton’s method to analog and digital realization
of fractional-order controllers,” International Journal of Microelectron-
• While the control system is capable of tracking the set ics and Computer Science, vol. 2, no. 2, pp. 45–52, 2012.
angle with 1% accuracy, in a particular position tracking [16] G. Carlson and C. Halijak, “Approximation of fractional capacitors
situation a more strict accuracy requirement may be (1/s)1/n by a regular Newton process,” IEEE Transactions on Circuit
Theory, vol. 11, no. 2, pp. 210–213, 1964.
demanded. Proper calibration must ensure the absence [17] B. M. Vinagre, I. Podlubny, A. Hernández, and V. Feliu, “Some
of offset errors. approximations of fractional order operators used in control theory and
applications,” Fractional Calculus & Applied Analysis, vol. 3, pp. 945–
950, 2000.
VII. C ONCLUSIONS [18] A. Oustaloup, F. Levron, B. Mathieu, and F. M. Nanot, “Frequency-band
In this paper, we presented a method for design and im- complex noninteger differentiator: characterization and synthesis,” IEEE
Transactions on Circuits and Systems—Part I: Fundamental Theory and
plementation of a digital controller for a position servo using Applications, vol. 47, no. 1, pp. 25–39, 2000.
tools available in the FOMCON toolbox for MATLAB. The [19] I. Petráš, S. Grega, and L. Dorcak, “Digital fractional order controllers
digital controller was implemented in form of a IIR filter realized by PIC microprocessor: Experimental results,” in Proc. of the
ICCC’2003 conference, High Tatras, Slovak Republic, 2003, pp. 873–
running on an Atmel AVR microprocessor. The results of 876.
controller design were verified during real-time closed loop [20] I. Petráš, “Practical aspects of tuning and implementation of fractional-
experiments. Some problems were identified and possible order controllers,” ASME Conference Proceedings, pp. 75–84, 2011.
solutions provided. Future work should be devoted to further [21] (2013) Official website of INTECO Sp. z o. o. [Last access time:
11.03.2013]. [Online]. Available: http://www.inteco.com.pl/
improving the controller design process, e.g. implementation [22] D. Xue and Y. Chen, “A comparative introduction of four fractional order
of auto-tuning directly on the embedded device. controllers,” in Proceedings of the 4th World Congress on Intelligent
Control and Automation, 2002, pp. 3228–3235.
[23] C. Yeroglu, C. Onat, and N. Tan, “A new tuning method for PIλ Dµ
ACKNOWLEDGMENT controller,” in Proc. Int. Conf. Electrical and Electronics Engineering
The work was supported by the European Union through ELECO 2009, vol. II, 2009, pp. 312–316.
[24] C. Zhao, D. Xue, and Y. Chen, “A fractional order PID tuning algorithm
the European Regional Development Fund, the target funding for a class of fractional order plants,” in Proc. IEEE Int. Mechatronics
project SF0140113As08 of Estonian Ministry of Education and Automation Conf., vol. 1, 2005, pp. 216–221.
and the ESF grant N8738, and the Estonian Doctoral School [25] C. A. Monje, B. M. Vinagre, A. J. Calderon, V. Feliu, and Y. Chen,
“Auto-tuning of fractional lead-lag compensators,” in Proceedings of
in Information and Communication Technology through the the 16th IFAC World Congress, 2005.
interdisciplinary project FOMCON. [26] A. Tepljakov, E. Petlenkov, and J. Belikov, “A flexible MATLAB tool for
optimal fractional-order PID controller design subject to specifications,”
in Proceedings of the 31st Chinese Control Conference, W. Li and
R EFERENCES Q. Zhao, Eds., Hefei, Anhui, China, 2012, pp. 4698–4703.
[1] K. Miller and B. Ross, An introduction to the fractional calculus and [27] D. Xue, Y. Chen, and D. P. Atherton, Linear Feedback Control: Analysis
fractional differential equations. Wiley, 1993. and Design with MATLAB (Advances in Design and Control), 1st ed.
[2] C. A. Monje, Y. Chen, B. Vinagre, D. Xue, and V. Feliu, Fractional- Philadelphia, PA, USA: Society for Industrial and Applied Mathematics,
order Systems and Controls: Fundamentals and Applications, ser. Ad- 2008.
vances in Industrial Control. Springer Verlag, 2010. [28] A. Tepljakov, E. Petlenkov, J. Belikov, and M. Halás, “Design and imple-
[3] Y. Q. Chen, I. Petráš, and D. Xue, “Fractional order control - a tutorial,” mentation of fractional-order PID controllers for a fluid tank system,”
in Proc. ACC ’09. American Control Conference, 2009, pp. 1397–1411. in Proc. 2013 American Control Conference (ACC), Washington DC,
[4] A. Oustaloup, J. Sabatier, and P. Lanusse, “From fractal robustness to USA, June 2013, pp. 1780–1785.
the CRONE control,” Fractional Calculus and Applied Analysis, vol. 2, [29] A. Tepljakov, E. Petlenkov, J. Belikov, and J. Finajev, “Fractional-order
no. 1, pp. 1–30, 1999. controller design and digital implementation using FOMCON toolbox
 8

for MATLAB,” in Proc. of the 2013 IEEE Multi-Conference on Systems

and Control conference, 2013, pp. 340–345.
[30] A. V. Oppenheim, R. W. Schafer, and J. R. Buck, Discrete-time signal
processing (2nd ed.). Upper Saddle River, NJ, USA: Prentice-Hall,
Inc., 1999.
[31] S. Mitra, Digital Signal Processing: A Computer-Based Approach, ser.
Connect, learn, succeed. McGraw-Hill Education, 2011.
[32] M. Nordin, “Controlling mechanical systems with backlash—a survey,”
Automatica, vol. 38, no. 10, pp. 1633–1649, Oct. 2002.
[33] S. Kuo and B. Lee, Real-time digital signal processing: implementations,
applications, and experiments with the TMS320C55X. Wiley, 2001.

Aleksei Tepljakov was born in 1987 in Tallinn.

He received his B.Sc and M.Sc in Computer and
Systems Engineering from Tallinn University of
Technology in 2009 and 2011, respectively. He is
currently a PhD student working at the Depart-
ment of Computer Control at Tallinn University
of Technology. His main research interests include
fractional-order control of complex systems and
fractional-order filter based analog and digital signal
processing.

Eduard Petlenkov was born in 1979. He received

his B.Sc, M.Sc and PhD degrees in computer
and systems engineering from Tallinn University of
Technology. He is an Associate Professor in the De-
partment of Computer Control at Tallinn University
of Technology. His main research interests lie in the
domain of nonlinear control, system analysis and
computational intelligence.

Juri Belikov was born in 1985. He received his B.Sc

degree in mathematics from Tallinn University, and
his M.Sc and PhD degrees in computer and systems
engineering from Tallinn University of Technology.
At present he holds the positions of researcher in
the Institute of Cybernetics and Associate Professor
in the Department at Computer Control of Tallinn
University of Technology. His main research inter-
ests lie in the domain of nonlinear control theory.

Sergei Astapov was born in 1988. He received

his M.Sc degree in the field of Computer System
Engineering at the Tallinn University of Technology
in 2011. He continues his education as a PhD stu-
dent at the Department of Computer Control at the
Tallinn University of Technology and is a member of
the Department’s Research Laboratory for Proactive
Technologies. His research interests include object
tracking using wideband signal analysis, classifica-
tion tasks and distributed computing in embedded
multi-agent systems. His recent research concerns
object localization and identification in open environments and acoustic signal
based diagnostics of industrial machinery.

View publication stats