Research Article

International Journal of Advanced

Robotic Systems
July-August 2019: 1–9
Stabilization of a double inverted rotary ª The Author(s) 2019
DOI: 10.1177/1729881419846741
pendulum through fractional order journals.sagepub.com/home/arx

integral control scheme

Ibrahim Mustafa Mehedi1,2, Ubaid M Al-Saggaf1,2,

Rachid Mansouri3 and Maamar Bettayeb1,4

Abstract
Rotary double inverted pendulum is a highly nonlinear complex system and requires a high performance controller for its
control. Using gain matrix which is obtained through state feedback technique may create complexity while removing the
steady-state error for all states. Incorporating an integral action can be an alternative for these errors. Therefore, a state
space–based fractional order controller with fractional integral action is designed and tested on a rotary double inverted
pendulum in this article. The fractional integral controller is designed based on Bode’s ideal transfer function. Two degree
of freedom is considered for tuning purpose as well. The integer order controller based on the state space approach is
also shown for comparison. Simulation and experimental results are presented for both controllers of this rotary double
inverted pendulum system.

Keywords
State space control, fractional order controller, two degree of freedom, Bode’s ideal transfer function, rotary double
inverted pendulum

Date received: 24 March 2018; accepted: 29 March 2019

Topic: Robot Manipulation and Control

Topic Editor: Andrey V Savkin
Associate Editor: Steven Su

Introduction the double inverted pendulum becomes an attractive tool to

A link rotating above its pivot point is called an inverted
pendulum. Controlling issues of this type of inherently
unstable systems are considered as a classic problem in
dynamics and control theory. This nonlinear system can
be implemented to justify the effectiveness and to proof
the enhanced performance of new control algorithms. The
double inverted pendulum is the next level complication of
classic single inverted pendulum problem. The double
inverted pendulum can move either linearly on a cart or
in a circular way mounted on a rotary arm while this arm
experiences torque produced by a servo motor. The rotary
double inverted pendulum is considered as an extremely
nonlinear unstable system and consequently to design its
controller is a challenging problem. Due to its complexity,
test the performance of different modern control tech-
niques. Moreover, the double inverted pendulum has sev-
eral critical applications in spacecraft and aircraft control.
1
Center of Excellence in Intelligent Engineering Systems (CEIES), King
Abdulaziz University, Jeddah, Saudi Arabia
2
Electrical and Computer Engineering Department, King Abdulaziz
University, Jeddah, Saudi Arabia
3
L2CSP Laboratory, Mouloud Mammeri University, Tizi Ouzou, Algeria
4
Electrical and Computer Engineering Department, University of Sharjah,
Sharjah, United Arab Emirates
Corresponding author:
Ibrahim Mustafa Mehedi, Center of Excellence in Intelligent Engineering
Systems (CEIES), King Abdulaziz University, Jeddah 21589, Saudi Arabia.
Email: imehedi@kau.edu.sa

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 International Journal of Advanced Robotic Systems
Specially, the control concept of double inverted pendulum
is effectively applicable for attitude control with a double
gimbaled momentum wheel called gyropendulum.1 More-
over, the double inverted pendulum represents the building
block of humanoid robots and in the modeling of rockets in
liftoff.2,3
There are several robust controllers proposed in recent
years such as fuzzy controller,4 neural network controller,5
genetic algorithm controller,6 adaptive controller,7 and slid-
ing mode controller.8 These are all considered as integer order
controllers. However, recently, noninteger order calculus
became an attractive tool to apply in different disciplines of
science and technology. For the automation field, design and
implementation of fractional order controllers represent a
highly promising area of research. It is already proven that
the performance of integer order controllers can be improved
using fractional order calculus.9–11 In fact, the fractional order
controllers are more flexible than the integer order controllers
because of the availability of higher number of tuning para-
meters, although the tuning procedure increases the complex-
ity. Effective tuning techniques are investigated and can be
seen in published papers (see e.g. Das et al.12).
The challenge in this article is to use fractional order linear
controllers to control the rotary double inverted pendulum
system. Although, currently the fractional order controllers
are often implemented experimentally, very few implementa-
tions are addressed using the state space approach. The
designed controller using state space method–based augmen-
ted system approach can be represented using fractional order
dynamics.13,14 The pseudo-state feedback–based fractional
order system is also investigated with the help of system
augmentation.15,16 Recently, proportional–integral (PI)-
based state feedback approach was introduced using frac-
tional calculus.17–20 The control gains for the proportional
and integral compensator of the state feedback controller are
designed using the pole placement technique using the
closed-loop characteristic polynomial. The experiment was
also performed using the PI-state feedback approach for a
linear single inverted pendulum-cart system.
The state space–based fractional order controller is
another technique to regulate such a nonlinear scheme in
which the integral control–based state feedback is used but
a fractional integrator is introduced in the place of integer
integrator. In this case, the gain associated with this integra-
tor is substituted by a fractional integrator which must be
designed. This method is introduced in Al-Saggaf et al.21 and
implemented on a single inverted linear cart pendulum. But a
multiple inverted pendulum is more complicated and highly
nonlinear system. There are many uses of double or triple
rotary inverted pendulum in robotic arm and in space appli-
cations. In this article, the proposed method is implemented
to control a double inverted rotary pendulum. The illustra-
tion on a double inverted rotary pendulum system empha-
sizes the viability of the anticipated method to the control of
linear systems that can be stable or unstable. The remaining
portion of the article is organized as stated below. The
Figure 1. Fractional compensator–based state space control.
Bode’s ideal transfer function is explained briefly in the
second section. The integral control–based state feedback
control is also explained in the same section. The third sec-
tion describes the mathematical modeling of the rotary dou-
ble inverted pendulum. Simulation results are shown in the
fourth section. Experimental results are presented with a
brief discussion including the procedure in the fifth section.
Finally, the sixth section concludes this article.
Fractional integral control combined with
state feedback
A state feedback control with an integral action is a very
useful method for the purpose of tracking a specific set
point and rejecting a load disturbance.22 In the integer case,
the state space expression of linear time invariant (LTI)
system model is given by

x_ ðtÞ ¼ A xðtÞ þ B uðtÞ þ E qðtÞ xð0Þ ¼ x 0
ð1Þ
yðtÞ ¼ C xðtÞ
where xðtÞ 2 Rn and yðtÞ 2 R are the state vector and the
output. uðtÞ 2 R and qðtÞ 2 R are denoted for control input
and disturbance input, respectively. A 2 Rnn is the system
matrix. B 2 Rn1 and C 2 R1n are the control and the out-
put gain vectors, respectively. While the integral action is
incorporated with this state space model, the augmented state
vector X ðtÞ ¼ ½xðtÞ xr ðtÞ0 is introduced, xr ðtÞ being the
output of the integral operator. The control law is written as
uðtÞ ¼ ½ K p K r X ðtÞ ð2Þ
There are some suitable methods to compute the state
feedback vector ½K p K r . Ackermann’s formula is used for
these gain computations.
It is known that the purpose of state feedback control is
to suitably change the poles position but to minimize the
influence of its zeros in the transient response, we can use a
fractional integrator 1=sa instead of the integer order inte-
grator 1=s. The static gain K r can also be replaced by a
compensator KðsÞ. Again, to stabilize the system, the state
feedback gain K s is used and to impose the transient
response of the closed-loop system, the compensator KðsÞ
is cascaded after the fractional order integrator (1=sa ). Fig-
ure 1 demonstrates the architecture fractional integrator–
based state feedback cascaded with a compensator. To
design KðsÞ, the Bode’s ideal transfer function23 is used
as an open-loop reference system model. Therefore, it
Mehedi et al.
entails a robust closed-loop system with respect to produce
variable gains and exhibits the property of iso-damping in
the step response.
The closed-loop transfer function using Bode’s ideal
method is
1 The inner-loop’s characteristic polynomial is denoted by
f ðsÞ ¼ 0<k<1
1 þ ic skþ1
Here, ic is responsible for the effectiveness of the
performance tracking system for transient conditions,
and the overshoot depends on the value of k. As shown
in Bettayeb and Mansouri,24 ic is related to the gain
crossover frequency !c and k is related to the phase
margin ’m
p  ’m 1 part of N ðsÞ inverted by KðsÞ and N þ ðsÞ is the part that
k¼ 1 and ic ¼
p=2 !kþ1
c
The step response of equation (3) shows the behavior of
an underdamped second-order system with damping ratio
between one and zero. Again, the value of k can be com-
puted from its relation with the overshoot of the step
response, M p ð%Þ.25 Equation (5) shows this relation.
5
k  6:75 10 M p2 þ 0:016 M p þ 0:066
Note that, due to the state feedback K s , this structure can
be used for both stable and unstable systems. The design
procedure of the state feedback gain (K s ) and the integer
filter (KðsÞ) can be summarized (see Al-Saggaf et al.21 for
more details). It is already mentioned that the integer filter
(KðsÞ) is connected in series with the fractional integrator
block as shown in Figure 1.
Let an integer order linear LTI system be given by equa-
tion (1). The control law is
uðtÞ ¼ K s xðtÞ þ KðtÞ  xr ðtÞ
where
D ðxr Þ ¼ r  C x
D is an operator which evaluates the fractional deriva-
tion. r is the reference parameters and xr is a new state
produced due to fractional integral operator. The tracking
error is integrated through the fractional integrator, which
produces xr . The impulse response of KðsÞ is represented
by KðtÞ. This KðtÞ is added in the control law through the
convolution product with K s . Here, the gain vector of the
state feedback, K s 2 R 1n , is generated to influence the
inner-loop’s characteristic polynomial. Further, a fractional
integrator of order  is associated with KðsÞ and used to
acquire in closed-loop the Bode’s ideal transfer function of
equation (3)
The n characteristic roots of inner-loop are arbitrarily
placed by the vector K s , which has the form
Dd ðsÞ ¼ sn þ a1 sn1 þ       þ an1 s þ an
3
and can be computed with the help of Ackermann’s tech-
nique,26 and the integer compensator KðsÞ is calculated by
Dd ðsÞ 1
KðsÞ ¼ ð9Þ
ic s N ðsÞ ð1 þ if sÞr
ð3Þ Dd ðsÞ. The numerator, N ðsÞ, is obtained from the transfer
function of the linear integer system shown in equation (1).
The time constant of the Bode’s ideal transfer function is
expressed by ic (equation (3)). In equation (9), KðsÞ is made
realizable in such a way that a low pass filter, ð1þi1f sÞr , is
cascaded with it.
N ðsÞ is in the form N ðsÞ ¼ N þ ðsÞ N  ðsÞ while system
(equation (1)) is non-minimum phase. Here, N  ðsÞ is the
ð4Þ KðsÞ does not attempt to inverse, it comprises the right-half
plane zeros. Moreover, N þ ðsÞ should be such that
N þ ð0Þ ¼ 1. Therefore, KðsÞ is given by
Dd ðsÞ 1
KðsÞ ¼ ð10Þ
ic s N  ðsÞ ð1 þ if sÞr
The details of the proof of this result can be found in Al-
Saggaf et al.21
ð5Þ
Mathematical modeling of rotary double
inverted pendulum
Model description
The schematic diagram of double inverted rotary pendulum
is shown in Figure 2.27 Lr is the length of the rotary arm, J r is
its moment of inertia, and the angle, q, changes positively
when the arm rotates in counterclockwise (CCW) direction.
ð6Þ The double pendulum system is attached at the end of this
rotary arm. Lp1 and lp1 are the length and center of mass for
the bottom pendulum, which is comparatively shorter. It has
ð7Þ a mass of M p1 and J p1 is the moment of inertia around the
center of mass. Similarly, Lp2 , lp2 , M p2 , and J p2 are the total
length, center of mass, mass, and moment of inertia for the
top medium-sized pendulum, respectively. a represents the
angle of the bottom pendulum and ’ represents the angle of
the top pendulum. Both angles become zero when the double
pendulum is clearly straight in the vertical position and they
increase positively when rotated CCW. M h is the mass of
hinge between both parts of the pendulum.
Nonlinear dynamics
By applying the Lagrange equation
@ 2L @L
 ¼ Qi ð11Þ
_
@t@ qi @qi
The variables qi are called generalized coordinates. L is
ð8Þ
the Lagrangian of the system. Let
4 International Journal of Advanced Robotic Systems

energy for the top medium-sized pendulum. These para-

meters are derived in the literature.27 The total potential
energy combines both gravitational and elastic potential
energies as shown
V ¼ mp1 cosðaðtÞÞ þ mh gLp1 cosðaðtÞÞ
þ mp2 gðsinðaðtÞÞlp2 sinð’ðtÞÞ
þ cosðaðtÞÞlp2 cosð’ðtÞÞ þ lp1 cosðaðtÞÞÞ ð16Þ

Now, from equation (16), the Lagrangian is found in

order to get the nonlinear equations of motion (EOMs) and
various derivatives are computed. q is considered as a gen-
eralized coordinate vector q, therefore, it can be globalized
into a matrix form
DðqÞ€q þ Cðq; qÞ
_ q_ þ gðqÞ ¼ i ð17Þ
where, D, C, and gðqÞ are inertial matrix, damping matrix,
and gravitational vector, respectively. i is torque applied at
the base of the rotary arm produced by the servo motor and
it is defined by the equation
_
hg K g hm k t ðV m  K g k m qÞ
Figure 2. Schematic diagram of double inverted pendulum i¼ ð18Þ
Rm
system.25
See the literature26 for an explanation for corresponding
parameters of the rotary servo motor (SRV02) parameters
(e.g. constant of back-emf, k m ).
qðtÞT ¼ ½ qðtÞ aðtÞ ’ðtÞ  ð12Þ
The corresponding velocities are
" # Linearized model
T
@qðtÞ @aðtÞ @’ðtÞ
_
qðtÞ ¼ @t @t @t ð13Þ The linear state space model is

x_ ¼ A x þ B u
The Lagrangian of the system is described ð19Þ
y¼Cx
L¼T V ð14Þ
where x and u are the state and control input. A, B, and C
where T is the total kinetic energy and V is the potential are matrices for the state space representation. For the dou-
energy of the system. One can get T ¼ T r þ T b þ T h þ T t , ble inverted rotary pendulum system, the state vector is
where T r is the kinetic energy for rotary arm which is xT ¼ ½ q a ’ q_ a_ ’_ , the output is y ¼ q. The aim
 2 is to design a controller for stabilizing the servo to a desired
1 d
Tr ¼ Jr qðtÞ ð15Þ position, q, with a minimum deflection angle of the
2 dt
pendulum.
T b is the kinetic energy for the bottom short pendulum, Linearization of the nonlinear system gives following
T h is the kinetic energy for the hinge, and T t is the kinetic state space matrices27

2 3
0 0 0 1 0 0
60 0 0 0 1 07
6 7
6 7
60 0 0 0 0 17
A¼6
60 a
7 ð20Þ
6 a M þ p1lp1 Lr M p2 gðlp1 þ Lp1 Þ 0 0 07
7
6 7
4 0 ab M p2 gðLr2 Lp1 ÞM p1 þ Lr2 Lp1 M p1 þ J r Lp1 Þ 0 0 05
0 ac ad 0 0 0
Mehedi et al. 5

2 3 formula, the static state feedback gain vector, K is

0 s
6 0 7 computed as
6 7
6 7
6 0 7 K s ¼ ½ 0:0142 0:0415 2:5538 0:0369 0:2319 0:2828 
6 7
6 M L 2
þ M L 2 7
B¼6 h p1 p1 p1 7 ð26Þ
6 7
6 L ðM L þ M L Þ 7
6 r h p1 p1 p1 7 Now, the new transfer function is obtained as follows
6 Lr 7
4 2
ðM p1 Lp1 þ M p1 Lp1 Lp2 þ M h Lp1 Lp2 þ M p1 Lp1 Lp1 Þ 5
Lp2 qðsÞ
GIL ðsÞ ¼
vðsÞ
ð21Þ
479:1375ðs  13:56Þðs  5:948Þðs þ 5:852Þðs þ 11:06Þ
¼
The position of the servo that actuates the pendulum link ðs þ 20:95Þðs þ 10:81Þ 2 ðs þ 5:645Þ 2 ðs þ 0:4562Þ
angles is considered as a measured output. Therefore, the ð27Þ
matrix, C, of the output equation is
where vðsÞ is the control input of KðsÞ and qðsÞ is the output
C ¼ ½1 0 0 0 0 0 ð22Þ need to be controlled. To avoid the complexity, SVD (sin-
The velocities of the servo link and the angles of pen- gular value decomposition)-based method (singular pertur-
dulum can be obtained by computing the derivative and bation approximation) is used to reduce the order of the
passing through a high-pass filter. equation (27). The reduced model thus obtained is given by
3:0502ðs 2  12:12s þ 39:38Þ
Gq ðsÞ ¼ ð28Þ
ðs þ 3:728Þðs þ 0:4583Þ
Simulation results
Now, Gq ðsÞ is decomposed into G þ
q ðsÞG q ðsÞ due to its
The proposed method described in the third section is 
non-minimum phase, where Gq ðsÞ ¼ ðsþ3:728Þðsþ0:4583Þ
3:0502
and
implemented to test the stability and check the set point Gq ðsÞ ¼ s  12:12s þ 39:38. In addition, G
þ 2
ðsÞ has the
q
tracking performance of the rotary double inverted pen-
relative degree of 2. Equation (10) is used to calculate the
dulum system. The Process block of the diagram shown in
filter KðsÞ. To obtain a reasonable value of the control law,
Figure 1 characterizes the linear model of the double
the value of ic is suitably selected. Again, the overshoot is
inverted rotary pendulum. A fractional integral control
minimized by taking noninteger order k for the closed-loop
with state feedback is incorporated in the model. The
step response. Therefore, the suitably selected values are
linearized pendulum model is numerically described as
ic ¼ 100 and k ¼ 0:07. Hence, the obtained transfer func-
follows tion of KðsÞ is given by
2 3
0 0 0 1 0 0
60 7 0:3278ðs þ 3:728Þðs þ 0:4583Þ
6 0 0 0 1 0 7 KðsÞ ¼ ð29Þ
6
60
7
7
sðs þ 100Þ
0 0 0 0 1
A¼6
60
7
6 368:1561 9:0635 1:1499 1:3285 1:4407 7
7 For the performance evaluation of the designed controller,
6 7
40 477:8131 40:7331 1:3285 1:8262 2:3306 5 a set point tracking is observed between t ¼ 0 s and t ¼ 25 s.
0 518:1773 143:5827 1:4407 2:3306 4:1084 The proposed fractional order control scheme is evalu-
ð23Þ ated by changing the set point at t ¼ 0 s and t ¼ 25 s. The
2 3 simulation results are demonstrated in Figures 3 to 6 of the
0 linearized model described in equations (23) to (25). These
6 1 7 figures entail that, after some transition period, the pendulum
6 7
6 7 angles are stabilized and the rotary position of the arm tracks
6 0 7
6
B¼6 7 ð24Þ the predefined set point. The fractional order controller per-
7
6 479:1375 7 forms better than the integer state feedback controller com-
6 7
4 553:5338 5 puted using the LQR (Linear Quadratic Regulator) method.
600:2946 Figures 4 and 5 show that, due to the variable set point, the
angles of the bottom and top pendulum are distorted but the
C ¼ ½1 0 0 0 0 0 ð25Þ control law reacts immediately and reactuates to return
toward its stable position a ¼ 0 and q ¼ 0.
The characteristic roots of the open-loop system are
located at 0, 0, 25.3110, 21.1906, 9.8158, 9.0252,
and 0.1263; therefore, the system is unstable. To
stabilize it, K s is used to relocate the poles of the
Experimental results
inner-loop at 20.9506, 10.8062, 10.8062, After simulating the designed control law on the linear
5.6445, 5.6445, and 0.4562. Using Ackermann’s model of a rotary inverted pendulum, it is implemented
6 International Journal of Advanced Robotic Systems

50 2 Statefeedback control
Fractional order control
40
1.5
30
20 1

10 0.5
θ (deg)

Vm (V)
0
0
−10
−20 −0.5
−30 −1
Reference input
−40 Statefeedback controller
Fractional order controller −1.5
−50
0 10 20 30 40 50 60 70 80 90 100
−2
time (s) 0 10 20 30 40 50 60 70 80 90 100
time (s)

Figure 3. Double inverted pendulum rotary arm position

tracking. Figure 6. Control voltage input.

3
Statefeedback Control
Fractional order control

1
α (deg)

−1

−2

−3
0 10 20 30 40 50 60 70 80 90 100
Sample time (s)
Figure 7. Experiment of rotary double inverted pendulum
Figure 4. Angle of bottom pendulum and a set point change. system.

purposes. A single channel linear voltage amplifier (Volt-

2 State feedback control
PAQ-X1) is used to adjust the supply voltage to the servo
1.5
Fractional order control
system in order to produce required torque. Further, a data
acquisition board (DAQ) is used to establish digital con-
1
nection with personal computer as an interface between
0.5 hardware and software. Matlab software is used to design
φ (deg)

0 the fractional order controller in Simulink environment. A

QUARC Version 2.3.355 software is used as the bridge
−0.5
connection between Matlab and hardware in real time.
−1 By implementing a controlled voltage to SRV02 unit, both
−1.5 of the pendulums are being balanced keeping them verti-
−2
cally in the upright position. To implement the fractional
0 10 20 30 40 50 60 70 80 90 100 integral control scheme based on state feedback technique,
Sample time (s) the linearized model of the rotary dual inverted pendulum
(equations (23) to (25)) is considered again.
Figure 5. Angle of top pendulum and a set point change. Experimental results are shown in Figure 8 for fractional
order controller Further, the results of integer order con-
on a similar physical system in a practical experiment. The troller are also produced for comparison as shown in Figure
Quanser27,28 experimental setup is shown in Figure 7. In 9. The closed-loop reference mode parameters are used for
this experimental setup, a double inverted pendulum is the the design of fractional order controller such as, ic ¼ 100
main component. A Rotary Servo unit (SRV02) is used to and k ¼ 0.07. All the results are shown from t ¼ 30 s and
create angular motion of the pendulum for tracking the beginning portion of time is ignored because the double
Mehedi et al. 7

Rotary arm angle (deg)

10
5
0
−5
−10
30 35 40 45
time (s)
Bottom pendulum (deg)

5 5

φ (deg)
0
0
−5

−10 −5
30 35 40 45 30 35 40 45
time (s) time (s)

4
Control voltage (V)

2
0
−2
−4
30 35 40 45
time (s)

Figure 8. Experimental results with fractional order controllers.

Rotary arm angle (deg)

10
5
0
−5
−10
30 35 40 45
time (s)
Bottom pendulum (deg)

Top pendulum (deg)

5 4

2
0
0

−5 −2
30 35 40 45 30 35 40 45
time (s) time (s)

4
Control voltage (V)

2
0
−2
−4
30 35 40 45
time (s)

Figure 9. Experimental results with integer order controllers.

inverted pendulum is initially manually erected unlike the
self swing up control available for the single inverted pen-
dulum. The results obviously tell that the integer order
controller is better in terms of rotary arm position and
pendulum angles deflection. However, we can see that the
control voltage behaves more precise and smoothly for the
fractional order controller over the integer order controller.
Figure 10 shows the performance on disturbance rejec-
tion. To produce these disturbances, the top pendulum is
tapped lightly arbitrarily. In this investigation, the top pen-
dulum is tapped from t ¼ 32 s to t ¼ 37 s. The results
produced by this experiment prove the effectiveness of
fractional order controller in terms of external disturbance
rejection. Hence, the proposed controller is robust enough
8 International Journal of Advanced Robotic Systems

20

Bottom pendulum (deg) Rotary arm angle (deg) 10

−10

−20

−30
30 35 40 45
time (s)

Top pendulum (deg)

10 10
5
0 5
−5
0
−10
−15 −5
30 35 40 45 30 35 40 45
time (s) time (s)
4
Control voltage (V)

2
0
−2
−4
30 35 40 45
time (s)

Figure 10. Experimental results with disturbance rejection for fractional order controllers.

40

20
θ (deg)

−20
15 20 25 30
time (s)
10 10
5
0
α (deg)

φ (deg)

0
−10
−5
−20 −10
15 20 25 30 15 20 25 30
time (s) time (s)

5
Vm (V)

−5
15 20 25 30
time (s)

Figure 11. Complete experimental results for the square reference signal.

for the highly nonlinear double inverted rotary pendulum Conclusion

system.
For further investigation, the experimental results are ana-
lyzed for the fractional order controller influencing the square
reference with frequency 0.1 Hz and amplitude gain 5, in
Figure 11. In these results, it is observed that the fractional
order control law is following the pattern of the reference
signal. Hence the designed controller is robust as well.
In this article, the fractional order state feedback with inte-
gral controller is designed based on the state space repre-
sentation. The designed controller is simulated for the
highly nonlinear rotary double inverted pendulum system.
Results show better performance than the integer order
state feedback controller. Practical implementation is also
Mehedi et al.
successfully conducted using the designed fractional order
controller. Although implementing the fractional order
controller on such inherently unstable system is a big chal-
lenge, yet the experimental results show a better perfor-
mance in terms of precise and smooth control voltage as
well as external disturbances applied on the pendulum.
Disturbance rejection and robustness performances are also
observed for the proposed fractional order controller.
Acknowledgment
The authors would like to acknowledge the technical and financial
support of KAU.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) disclosed receipt of the following financial support
for the research, authorship, and/or publication of this article: This
project was funded by the Center of Excellence in Intelligent
Engineering Systems (CEIES), King Abdulaziz University, under
grant no. CEIES-16-12-01.
References
1. Jerry G. Engineering dynamics. Cambridge: Cambridge Uni-
versity Press, 2008.
2. Ahmed E and Alexandre P. Humanoid robot balance control
using the spherical inverted pendulum mode. Front Robot AI
2015; 2(21): 1–13.
3. Jensen JS. Nonlinear dynamics of the follower-loaded double
pendulum with added support-excitation. Fuzzy Set Syst
1998; 79(3): 315–321.
4. Fuyan C, Guomin Z, Youshan L, et al. Fuzzy control of a
double-inverted pendulum. J Sound Vib 1996; 215(1):
125–142.
5. Ashwani K. Anti-swing and position control of double
inverted pendulum (dip) on cart using hybrid neuro-fuzzy
controllers. Int J Inform Technol Comput Sci 2016; 7: 16–21.
6. Bingbing W, Chaoying L, Xueling S, et al. Design and imple-
mentation of the inverted pendulum optimal controller based
on hybrid genetic algorithm. In: International conference on
automation, mechanical control and computational engineer-
ing, China, 24–26 April 2015, pp. 623–629.
7. Shubhobrata R and Ranjit KB. Robust adaptive backstepping
control of inverted pendulum on cart system. Int J Control
Autom 2012; 5(1): 13–26.
8. Ilhami Y. Model free sliding mode stabilizing control of a
real rotary inverted pendulum. J Vib Control 2017; 23(10):
1–18.
9. Oustaloup A. La commande CRONE. Paris: Hermès Editeur,
1991.
10. Oustaloup A and Mathieu B. La commande CRONE du sca-
laire au multivariable. Paris: Hermès Editeur, 1999.
9
11. Poblubny I. Fractional-order system and PI k D controllers.
IEEE Trans Autom Control 1999; 44(1): 208–214.
12. Das S, Saha S, Das S, et al. On the selection of tuning meth-
odology of FOPID controllers for the control of higher pro-
cesses. ISA Trans 2011; 50: 376–388.
13. Nia SHH, Sierociuk D, Calderon AJ, et al. Augmented system
approach for fractional order SMC of a DC-DC Buck con-
verter. In: Proceedings of the 4th IFAC workshop on frac-
tional differentiation and its applications Badajoz, Spain,
October 2010, pp. 18–20.
14. Sierociuk D and Vinagre BM. State and output feedback
fractional control by system augmentation. In: Proceedings
of the 4th IFAC workshop on fractional differentiation and its
applications, Badajoz, Spain, October 2010, pp. 18–20.
15. Farges C, Moze M, and Sabatier J. Pseudo-state feedback
stabilization of commensurate fractional order systems. In:
Proceedings of the 10th European control conference, Buda-
pest, Hungary, August 2009, pp. 23–26.
16. Farges C, Moze M, and Sabatier J. Pseudo-state feedback
stabilization of commensurate fractional order systems. Auto-
matica 2010; 46: 1730–1734.
17. Bettayeb M, Boussalem C, Mansouri R, et al. Stabilization of
an inverted pendulum-cart system by fractional PI-state feed-
back. ISA Trans 2014; 53: 508–516.
18. Mansouri R, Bettayeb M, Boussalem C, et al. Linear integer
order system control by fractional PI-state feedback, Chapter
4. In: RAZ D (ed) Fractional calculus: Applications. New
York: Nova publishers, 2015.
19. Bettayeb M and Mansouri R. Fractional IMC-PID-filter con-
trollers design for noninteger order systems. J Process Contr
2014; 24(4): 261–271.
20. Al-Saggaf UM, Mehedi IM, Mansouri R, et al. Fractional-
order controller design for a heat flow process. Proc Insti-
tut Mech Eng Part I, J Syst Control Eng 2016; 230(7):
680–691.
21. Al-Saggaf UM, Mehedi IM, Mansouri R, et al. State feedback
with fractional integral control design based on the Bode’s
ideal transfer function. Int J Syst Sci 2016; 47(1): 149–161.
22. Ogata K. Modern control engineering. Englewood Cliffs:
Prentice-Hall, 1990.
23. Bode HW. Network analysis and feedback amplifier design.
New York: Van Nostrand, 1945.
24. Bettayeb M and Mansouri R. Fractional IMC-PID-filter con-
trollers design for integer order systems. ISA Trans 2014; 53:
1620–1628.
25. Bettayeb M, Mansouri R, Al-Saggaf U, et al. Smith predictor
based fractional-order-filter PID controllers design for long
time delay systems. Asian J Control 2017; 19(2): 587–598.
26. Dorf RC and Bishop RH. Modern control systems. 13th ed,
London: Pearson, 2017.
27. Quanser innovative Edutech laboratory guide—rotary dou-
ble inverted pendulum experiment. Markham: Quanser Inc,
2012.
28. Quanser SRV02 user manual. Markham: Quanser Inc, 2009.