International Journal of Control Systems and Robotics

Aymen Rhouma, Hafsi Sami http://www.iaras.org/iaras/journals/ijcsr

A Microcontroller Implementation of Fractional Order Controller

Aymen Rhouma and Hafsi Sami

Université de Tunis El Manar, Ecole Nationale d’Ingénieurs de Tunis, LR11ES20 Laboratoire Analyse,
Conception et Commande des Systèmes, 1002, Tunis, Tunisie;
Université de Tunis El Manar, Faculté des Sciences de Tunis, 2092, Tunis, Tunisie;
E-mail : [email protected], [email protected]

Abstract: The idea of this paper is to implement a Fractional Order proportional integral PI  on a real electronic
system by using the STM32 microcontroller. This controller is founded on an extended version of Hermite-Biehler
theorem to determine the complete set stabilizing. The STM32 keil starter kit based on a JTAG interface and the
STM32 board was used to implement the proposed fractional controller. So, this controller is implemented using Keil
development tools designed for ARM processor-based microcontroller devices and working with C/C++ langage. The
performances and the efficiency of the developed strategy are illustrated with simulation results.

Keywords: Fractional-order calculus, fractional order PI controller, STM32 microcontroller, Keil.

best PI  gains. In [11], the authors propose two sets of

1. Introduction tuning rules for fractional PID similar to those of the
first set of Ziegler–Nichols. A frequency approach for
Fractional order calculus is a mathematical discipline the auto-tuning of fractional-order PID is proposed in
with a 300-years-old history [1]. In recent years the [12], where PI is used to cancel the slope of the
fractional calculus has attracted the attention of curvephase of a position servo system with time delay
researchers in many fields such as biology, economics around a frequency point and the PI  D controller is
and engineering [2-5]. The fractional order calculus designed to fulfill the specifications of gain crossover
has been found that many physical systems have a frequency.
dynamic behavior non-integer order. The first non- A microcontroller is described as a computer on a chip
integer order system to be widely recognized is the because it contains all the features of a full computer
thermal system; other fractional order systems that are including central processor output and input ports with
known are the electromagnetic wave systems, special features, serial communication, digital-analog
electrode-electrolyte polarization, viscoelastic, and conversion, analog-digital conversion and signal
many others [6]. Also, the fractional dynamic system processing.
appears in the system industries, such as control In this paper, we propose a Fractional Order
application [7]. The first fractional order control to proportional integral PI  controller for a performed
regulator the physical systems is proposed by STMicroelectronics microcontroller (STM32). The
Oustaloup in 1988 [8, 9]. Later, Podlubny developed control application could benefit from the power
the fractional PID controller in 1994 [1]. Recently, features and flexibility of the STM32F103xB devices.
Rhouma is proposed the fractional model predictive The proposed framework was developed using Keil
control to fractional order system [8]. In [9], the development tools designed for ARM processor-based
authors propose a new PI  tuning method for first order microcontroller devices.
systems with time delay. Some results on the control The outline of this paper is organized as follows. In
of integrating systems with time delay using fractional section 2, a problem formulation and some definitions
order PD controllers were obtained. Recently, Monje of fractional order systems are introduced, and the G-L
et al. [10] give a new tuning method called F-MIGO definition used to approximate the fractional order
for PI  extended from the MIGO method, this tuning system is detailed. Section 3 states the hardware as
rules is used to determine the best fractional and the well as the software development tools used in this
application. The section 4 is reserved to focus on the

ISSN: 2367-8917 122 Volume 2, 2017

 International Journal of Control Systems and Robotics
Aymen Rhouma, Hafsi Sami http://www.iaras.org/iaras/journals/ijcsr

necessary steps in finding the design method proposed  al , bm   R 2 ,  al , bm   R2 , l=0,1, L,
for the PI  . Simulation results are given in the section m=0,1, M
5. Finally, a conclusion is given.

2. Preliminary of fractional calculus 3. Fractional-order PI design

Fractional calculus is a generalization of
The aim of this section is to present the system which
differentiation and integration to non-integer orders
will be controlled by a fractional order PI controller
fundamental operator t 0 Dt f  t  which is defined as: and to present the design of the fractional controller.
 d This controller is based on the unit return
   0 configuration shown in the following figure.
 dt

t 0 Dt f  t   1

 =0 (1)
t
  (d )  <0
t 0
Where  is the order,   R , t and t0 are the upper and
lower limits of the operation, respectively. There are
several definitions for fractional order calculus, the
most popular definitions used are the Grunwald- Figure1: Closed-loop fractional system
Letnikov (G-L) and Riemann-Liouville (R-L)
definitions [13]. The R-L definition of function f (t ) is The first fractional order systems can be described by:
defined as: K
G ( p)  (7)
1 dn t f ( ) 1  Tp
t 0 Dt f  t  

 d , n  1    n (2)
(n   ) dt n t 0 (t   )  n 1
Our tuning strategy, is based on Hermite-Biehler
The G-L definition of function f (t ) is defined as:
theorem and the Pontryagin condition to determine the
k p and ki parameters.

(t t0 / h ) (  1)
t 0 Dt f (t )  lim h   (1) j f (t  jh) The fractional PI  controller transfer function C(s) is
h 0 j 0 (  j  1)( j  1)
given by the following equation:
(3) Ki
C ( p)  K p  (8)
Where ( ) is the Gamma function, h is the sampling p
period, n is an integer. The control input of the PI  controller is:
The laplace transform of R-L and G-L definitions for
u  t   K p  r  t   y  t    K i Dt  r  t   y  t  
zero initial conditions can be given as:
(9)
L  0 Dt f (t )  s L  f (t )  s F (s) (4)
  Where r(t) is the reference input or the setpoint signal,
In general, a fractional model can be described by a e(t) is the error, u(t) is the control, y(t) is the output
fractional differential equation characterized of the signal and Dt is the fractional differential/integral
following form:
L  al M  bm
operators.
 al D y (t )   bm D u (t ) (5) The control design method proposed in this paper is
l 0 m 0
based on a Hermite-Biehler and Pontryagin theorem
Using the laplace transform in equation (5), the
which consist on interlacement property of the real
fractional-order system can be represented by the
roots of the polynomial characteristic.
following transfer function: The closed-loop characteristic polynomial of a first
M

 bm s bm order system is given by:
Y ( s) m 0
G( s)   (6)  ( p )  ( KKi  KK p p  )  (1  Tp) p 
U ( s) L  al
 al s However, we present our theorem [29] useful to
l 0
Where compute the stability region of a first order system.
Based on the first property of Hermite-Biehler [30]

ISSN: 2367-8917 123 Volume 2, 2017

 International Journal of Control Systems and Robotics
Aymen Rhouma, Hafsi Sami http://www.iaras.org/iaras/journals/ijcsr

which consist that all the roots of the polynomial

characteristic of the closed loop equation are real.
Theorem1: [14]
We consider a first order plant given by the following
transfer function:
K  Ls
G( s )  e
1  Ts
where the parameters T, L and K are positive.
We can determine the set of all stabilizing ( k p , ki )
values for the given plant using the fractional order
controller PI 
Ki
C ( s)  K p 
s

4. Microcontroller STM32 Star Kit

In this section, an overview of the hardware and the
software development tools is presented.
The choice to adopt STM32 is based on tradeoffs
between price and performance. So, the STM32
performance line family incorporates the high- Figure 3. STM32F4 architecture.
performance ARM Cortex-M4, 72 MHz frequency, an
extensive range of enhanced I/Os and high-speed Keil is a software development tool includes C/C++,
embedded memories. the devices of STM32 offer two simulation models, Real-Time Operating System,
12-bit ADCs. Moreover, the STM32 has configurable debuggers, integrated environments and evaluation
and flexible power management features that allow to boards for ARM, Cortex-M, Cortex-R4 families. The
choose the power option to fit application [15-16]. used version, in this paper, is the µVision 4. This
The STM32 starter kit presented in Figure 2 was used version screen supplies a menu bar for command entry,
to implement the Fractional PI program and The windows for source files, and a toolbar where we can
STM32F100 architecture is presented in Figure 3. choose command buttons dialog boxes.
The µVision 4 has two operating modes: Build Mode
and Debug Mode.
- Build Mode: It’s the standard working mode.
It lets us to convert all the application files and
to generate executable programs.
- Debug Mode: It supplies an effective debugger
fo testing applications.

5. Experiment results
In order to illustrate the effectiveness and
performances of the fractional order controller
developed in this paper, we make a practical
implementation of the proposed controller on an
electronic system by using an STM32 microcontroller.
The practical system used in this paper is a first order
system as shown in Figure 4.

Figure2. STM32F4 starter kit

ISSN: 2367-8917 124 Volume 2, 2017

 International Journal of Control Systems and Robotics
Aymen Rhouma, Hafsi Sami http://www.iaras.org/iaras/journals/ijcsr

The designed PI  parameters are fixed as follows:

K p  2, Ki  0.6 and   0.9
The evolutions of the output, the set point and the
control signal, obtained by applying the Fractional
Order proportional integral PI  controller to the real
system, are shown in Figures 6 and 7 respectively.
Based on these results, we note that the output signal
meets the desired requirements and the control signal
obtained provide a small variation.
Figure 4. First order system. 3

Output
Set point

In this work we choose R= 3.2K Ω and C=470µF. So, 2.5

by adopting these values, the system transfer function

2
will be
given by:

Tension (v)
1.5

Y ( p) 1
G( p)   (10)
U ( p) 1  0.01504 p 1

The output of the system will be injected into the

STM32 card through pin 6 of port A (PA6) which will 0.5

be used as the input of the analog converter "ADC".

This converter consist to converts the analog value of 0
0 10 20 30 40 50
Time (ms)
60 70 80 90 100

the output voltage to a digital value which can be used

in the program of the Fractional PI regulator already Figure 6. Set point and output signals with fractional
implemented on the STM32 board. Then, the control PI
delivered by the corrector passes through the pin PA4 3

of the STM32 card (figure 5).

Control Signal
2.5

2
Tension(v)

1.5

0.5

0
0 10 20 30 40 50 60 70 80 90 100
Time (ms)

Figure 7. Control signal with fractional PI

Figure 5. Connecting the STM32 with the system A comparison of the closed-loop performances of the
pro-posed approach is established with classic
To implement the fractional-order PI  of the thermal Proportional Integral (PI) controller.
system we have used the model given by equation The designed PI parameters are fixed as follows:
(26). So, the thermal system is defined as a first order K p  2, Ki  0.6
system with time delay. Therefore, we proceed the Figures 8 and 9 exhibits the output, the set point and
design of the controller by exploiting the approach the control signal obtained by applying the classic PI
exposed in section 4. controller.

ISSN: 2367-8917 125 Volume 2, 2017

 International Journal of Control Systems and Robotics
Aymen Rhouma, Hafsi Sami http://www.iaras.org/iaras/journals/ijcsr

3 REFERENCES
Output
Set point
2.5 [1] Podlubny I. Fractional Differential Equations,
Academie Press, New York, 1999.
2
[2] Mainardi F., Raberto M., Gorenflo R., Scalas E.
Fractional calculus and continuous-time finance II:
Tension (v)

1.5
the waiting-time distribution. Physica A: Statistical
Mechanics and its Applications, 2000, 287:468–
1
481.
0.5
[3] Yuste S., Abad E., Lindenberg K. Application of
fractional calculus to reaction-subdiffusion
0
0 10 20 30 40 50 60 70 80 90 100
processes and morphogen gradient formation.
Time (ms)
Arxiv preprint arXiv 2010, 1006.2661.
Figure 8. Set point and output signals with classic PI [4] Bagley R., Calico R. Fractional order state
equations for the control of viscoelastically
3
damped structures. Journal of Guidance, Control,
Control signal
and Dynamics, 1991, 14 :304–311.
2.5
[5] Zhang Y., Tian Q., Chen L., Yang J. Simulation of
a viscoelastic flexible multibody system using
2
absolute nodal coordinate and fractional derivative
Tension (v)

1.5
methods. Multibody System Dynamics, 2009,
21 :281–303.
1 [6] Sun H.H., Charef A., Tsao Y., Onaral B. Analysis
of polarization dynamics by singularity
0.5 decomposition method. Annals of Biomedical
Engineering, 1992, 20:321-335
0
0 10 20 30 40 50
Time (ms)
60 70 80 90 100 [7] Shantanu D. Functional fractional calculs for
system identification and controls. Springer-
Figure 9. Control signal classic PI
verlag, Berlin, 2008.
Rhouma, A. and Bouani, F. Robust model
Comparing the results obtained by the proposed PI 
[8]
predictive control of uncertain fractional systems:
and the classic PI , we deduce that the first controller
a thermal application IET Control Theory and
reaches the desired reference despite the change of the
Applications. 2014, Vol. 8, Iss. 17, pp. 1986–1994.
system dynamic whereas the second controller
presents oscillations at the set point variations. We [9] Sundaravadivu K., Arun B., and Saravanan K.
have also remark that the control law obtained by the Design of fractional order PID controller for liquid
PI  is smoother than one obtained by PI .
level control of spherical tank. IEEE Int. Conf.
Control Syst. Comput. Eng., Penang, Malaysia,
2011, 522–525.
6. Conclusion [10] Monje C, A., Chen Y. Q. Vinagre, B. M, Dingyu
The implementation of frcational PI controller by the use of
X. and Vicente F. Fractional-order Systems and
the STM32 microcontroller has been the object of this work. Controls: Fundamentals and Applications
This controller is designed so as to ensure certain closedloop (Advances in Industrial Control). 2010. London:
performances. The classic PI controller has been tested with Springer.
the same microcontroller and compared with the fractional [11] Valério D, and Costa J. S. Tuning of fractional
PI controller. Practical results show the effectiveness of the PID controllers with Ziegler-Nichols-type rules.
latter. Signal Processing, 2006, 86: 2771–2784.
[12] Sundaravadivu K., Arun B., and Saravanan K.
Design of fractional order PID controller for liquid
level control of spherical tank. IEEE Int. Conf.
Control Syst. Comput. Eng., Penang, Malaysia,
2011, 522–525.

ISSN: 2367-8917 126 Volume 2, 2017

 International Journal of Control Systems and Robotics
Aymen Rhouma, Hafsi Sami http://www.iaras.org/iaras/journals/ijcsr

[13] Oustaloup A. From fractality to non-integer [15] Ben Hariz M. Bouani F. Implementation of a fixed
derivation through recursivity, a property common low order controller on STM32 microcontroller,
to these two concepts: a fundamental idea from a International Conference on Control, Engineering
new process control strategy. 12th IMACS World & Information Technology (CEIT’14), Tunisie
Conf., Paris, 1988, 203–208. 2014.
[14] Hafsi.S, Laabidi.K and Farkh.R. A New Tuning [16] Kherigi A. Bouani F. Ksouri M. ben Ahmed M. A
Method for Stabilization Time Delay Systems Microcontroller Implementation of Model
Using PI  D Controllers. Asian Journal of Predictive Control. International Journal of
Control, 2015 Vol. 17, No. 3, pp. 1-11. Electrical, Computer, Energetic, Electronic and
Communication Engineering Vol:5, No:5, 2011.

ISSN: 2367-8917 127 Volume 2, 2017