2016 IEEE First International Conference on Control, Measurement and Instrumentation (CMI)

Optimal PID controlller design of an Inverted Pendulum

m Dynamics:
A Hybrid Pole-P
Placement & Firefly Algorithm Appproach
N. Surendranath Reddy and Srinivasaa Saketh M

Pikaso Pal and

a Rajeeb Dey

Department of Electronics & Communiccation Engg.

SMIT, Majitar, Sikkim, Indiaa
[email protected], [email protected]

Dept. Electriical Engineering

NIT Agartala and
a Silchar, India
[email protected],, [email protected]

Abstract The solution of control for inverrted pendulum is a

challenging problem and many control tech
hniques have been
tested on this bench mark problem to test the
t efficacy of the
designed control law. In this note an attemptt has been made to
stabilize a linear dynamics of inverted pendulu
um by designing an
optimal PID controller that guarantees the given
g
time-domain
performance specifications. To accomplish the
t
task a hybrid
pole-placement and firefly optimization conttrol algorithm has
been evolved. To apply the control algorithm, initially the upper
bound of the controller gain is computed usiing Routh-Hurwitz
criteria and later the lower bound of the contrroller gain is found
out using pole-placement technique. Oncce the ranges of
controller gains are obtained, the optimized vaalue of the gains are
found using firefly optimization technique.
Keywords- Inverted Pendulum Firefly optimizattion Pole-placement
PID control

I.

INTRODUCTION

An inverted pendulum system consistss of a pendulum,

which is free to oscillate around a fixed pivoot point attached to
a movable cart. The objective is to maintainn the pendulum in
the vertically upright position. However, thee system is highly
unstable and non-linear in nature; thereby offfering a variety of
control challenges. Different approaches have been used to
solve this problem. In [1], a Bang-Bang sttate type feedback
algorithm is used to swing up the penduluum to the inverted
position and Linear Quadratic control methhod to maintain it
thus. A high speed fuzzy controller that coould stabilize two
inverted pendulums simultaneously by usinng a set of control
rules is presented in [2]. In [3], a fuzzy-logic controller that
uses vision feedback for guiding the controlller in manipulating
motion of the cart to stabilize the system is presented. In [4],a
neural network is trained using reinforcem
ment and temporal
difference techniques to balance the invertted pendulum. An
Energy control method which utilizes a servoo design technique
to control the pivot acceleration for swing up control of an
inverted pendulum is presented in [5]. A nonn-linear small gain
theorem is used to show that non-linear, feedd forward systems
can be controlled and is applied to the inveerted pendulum in
[6]. The architecture in [7] employs a fuzzy
f
system that
compensates the plants nonlinearities annd stabilizes the
pendulum in inverted position.
To compensate for errors in the fuzzy system
m a sliding control
input is used. PID control is another commonly used
approach; however it involves a challenge of
o optimization of
controller parameters. The self-tuning method
m
for PID

978-1-4799-1769-3/16/$31.00 2016 IEEE

305

controllers proposed in [8] utiilizes the concept of Lyapunov

stability. In [9], a potentiometter is used for sensing and PD
controller, tuned by trial and error
e
method, for providing the
control signal. In the comparattive study of Sliding Mode and
PID control methods in [10], the classical Ziegler-Nichols
(ZN) approach is used to tune the PID controller. In [11], ZN
tuning method of PID and thee recent Firefly Algorithm (FA)
are compared; FA tuned syystem had better steady state
response and performance inddices. This paper amalgamates
the PID control design techhnique with a traditional pole
placement technique and the new heuristic based optimization
technique of Firefly Algorithm
m (FA) for optimization of PID
controller parameters. The polee placement technique is used to
achieve pre-defined perform
mance specification of the
considered system. The FA opptimization approach is inspired
by the natural biological phenoomenon of firefly flash, which is
an act of signaling, intendedd to attract other fireflies as
explained in [12-17].
a follows. In Section II, the
The paper is organized as
dynamics of the inverted penddulum system are discussed in
detail. In Section III, PID conntroller design and optimization
challenges are discussed. Sectioon IV addresses the optimization
of PID controllers using FA annd describes the obtained results.
Section V concludes this work
II.

LINEARIZED DYNAMICS
D
OF SYSTEM

In this work we consider thee experimental set up of inverted

pendulum of Feedback make [119]. The cart-pendulum set up of
[19] is shown in Fig. 1. We will operate the pendulum in
upright position as highlighted in previous section. To compute
the dynamics of this experimeental set up we analyze various
forces acting on the system undder equilibrium and be found in
[18] and [19].

Fig. 1. Cart-pendulum
m experimental set up.

2016 IEEE First International Conference on Control, Measurement and Instrumentation (CMI)

The parameter values of invertted pendulum in (12) & (13) can

be found in [19]. Substitution of
o parameter values specified in
Table I in equation (12) & (13) give (14) & (15) respectively,

pendulum has its CG at the center of the penndulum bob and is

movable in one plane only. The equatioon describing the
horizontal motion of the cart is:
f ( t ) = M 
x + b x + N
(1)
Where f (t ) is the force applied on the cart, b x is the restoring
force due to friction, N is reaction force of the
t pendulum, Mx
is the inertial force due to mass of the cart. Vertical
V
motion of
the cart is not assumed.
The rotational motion of the pendulum aboutt its mass is:
I 
+ d  + N l cos = P l sin
(2)
The horizontal motion of CG of the pendulum
m is:
2



(3)
N = m x + m l cos - m l sin
The vertical motion of CG of the pendulum is
i
(4)
P = mg - m 
l sin - m  2 l cos
Substituting for in the equation (1),
f ( t ) = ( M + m ) 
x + b x + m  l cos - m  2 l sin
(5)
Substituting for N and P in the equation (2)),
I + ml 2 
+ d  + m l 
x cos - mg l sin = 0

( - 0 )

( - 0 )

2!

f ( 0 )

The Laplace transform of equations (8) & (9)) yields

F ( s ) = X ( s ) s 2 ( M + m ) + sb + ( s ) s 2 ml

( s ) s 2 I + ml 2 + sd - mgl + X ( s ) s 2 ml = 0

Fig. 3. Displacement of COG (penduulum) on application of unit step input

III.

(9)
(10)
(11)

The transfer function of the pllant shown in Fig.5. From (15),

one can write,
s
(16)
G(s) =
3
2
-4.008s - 0.23s + 25.79s +0.490
The transfer function of a PID controller is given by
K
F (s)

(17)
= Gc ( s ) = K p + i + K d s
s
E ( s)

The closed loop transfer functiion of the system can be written

as:
G ( s ) Gc ( s )
(s)
= T (s) =
(18)
R( s )
1+G ( s ) Gc ( s )

( t ) are the outputs. Transfer function anallysis is performed

considering both outputs independently as
a done in [10].
F (s)

+ ds - mgl

{[( I + ml )( M + m) (ml ) 2 ]s 4 + [b( I + ml 2 ) +

2

(12)

d ( M + m)]s 3 + [bd mgl ( M + m)]s 2 m glb s}

( s)

F ( s)

PID CON
NTROLLER DESGN

A PID controller is used too stabilize this system and is

introduced in the feed forward path as shown in Fig. 4 [9].
(s)
Let
= G ( s)
F (s)

In the inverted pendulum system, f ( t ) is the

t input, x ( t ) and

( I + ml ) s

(s)
s
(15)
=
F ( s ) -4.008s 3 - 0.236s 2 + 25.79s + 0.490
Fig. 2 & 3 show the opeen-loop response of the system
models (14) & (15) respectivvely, to a unit step input. The
system is unstable under open-lloop condition as the response is
unbounded.

(7)

B. Transfer Function

(14)

f '' ( 0 )

( I + ml )  + d  + m l x - m g l = 0

X ( s)

0.1288s 2 + 0.005s - 0.8122

0.3319s 4 + 0.01955s 3 - 2.136s 2 - 0.0406s

Fig. 2.Angular displacement (penduulum) on application of unit step input

+ HOD
3!
Neglecting the higher order derivatives, thee equations (8) &
(9) are obtained:
f ( t ) = ( M + m ) x + b x + m l 

(8)

F (s)

(6)

A. Linearization
Considering small angles of deviation , linnearize the system
as follows:
The Taylors series expansion of f ( ) is giveen by:

f ( ) = f ( 0 ) + ( - 0 ) f ( 0 ) +

X (s)

-mls
(13)
{[( I + ml 2 )( M + m) (ml )2 ]s 3 + [( I + ml 2 )b +

From equations (16), (17) & (18), we can write:

( M + m)d ]s 2 + [bd mgl ( M + m)]s m glb}

306

2016 IEEE First International Conference on Control, Measurement and Instrumentation (CMI)

parameter can be tuned to varry with the iteration counter t

during iterations as given by t = t , where 0 < < 1 .
Fig. 4 Block diagram of closed loop system withh PID controller

T(s) =

K p s + Ki + K d s 2

B. Proposed Firefly optimizedd PID Contrller algorithm

In this paper, among various performance criteria, ITSE is
used for evaluating the PID controller parameters. The pseudo
A is given below,
code for implementation of FA
Step1: Initialize the algorithm parameters (number of fireflies,
alpha, beta and gamma).
Step2: Specify the lower annd upper bounds of the three
controller parameters. The upper bound is determined using
the Routh-Hurwitz criteria, and the lower bound is
p
technique as per the
approximated from the pole placement
design requirements.
Step3: Define the transfer funnction of the closed loop system
and calculate the error.
Step4: Define ITSE ass the objective function
f ( x ) , x = ( x1 , x2 ,, xd ) .

(19)

[ - 4.008s 3 - ( 0.236 - K d ) s 2 -25.79 - K p s

- ( -0.490 - K i )]
The characteristic equation obtained from eqquation (19) is:
-4.008s 3 - ( 0.236 - K d ) s 2

- -25.79 - K p s - ( -0.490 - K i ) = 0

(20)

It is a necessary condition for the stability off a system that the

coefficients of its characteristic equatiion be positive.
Therefore, from (21), (22) & (23) we obbtained following
ranges of K p ,K i & K d respectively:
(21)

K p < -25.79, K i < -0.490, K d < 0.236

(
index).
The aim is to minimize f ( x ) (performance

MAIN RESULTS

IV.

Step5: Generate
xi ( i = 1, 2, , n ) .

A. Firefly Optimization
FA wasformulated by Xin-She Yang in 20077 by assuming:
All fireflies are unisexual, so that one
o firefly will be
attracted to all other fireflies.
Attractiveness is proportional to theeir brightness, and
for any two fireflies, the duller onee will be attracted
by (and thus move to) the brighter one; however, the
brightness can decrease as their disttance increases.
If there are no fireflies brighter thann a given firefly, it
will move randomly.

an

initiaal

population

of

n fireflies

Step6: Determine the light inteensity Ii of the fireflies.

Step7: While ( t < Maximum generation)
g
For i = 1: n (all n fireflies), Foor j = 1: i (all n fireflies)
if ( I j > I i )
Move fireflies i and j accordinng to their attractiveness.
Evaluate new solutions and uppdate the light intensity for the
next iteration. Check whetherr the new updated fireflies lie
within the limits/bounds.
End for j
End for i
Sort the fireflies to find the present best
End while. Begin post process on best results obtained as
indicated in [11, 13, 14, 17].

The brightness should be associated wiith the objective

function so that it may be optimized.. As a fireflys
attractiveness is proportional to the light inteensity seen by the
adjacent fireflies, the variation of attractiiveness with the
2

distance r is given by, = e r where is the attractiveness

at r = 0 . The distance r or rij between any two fireflies i and

j
rij =

at

xi and

positions

y
x j respectively

is

given

To apply FA its parameters need to be initialized and its

I Application of FA requires a
values are as shown in Table II.
range of K p ,K i & K d to be choosen. The upper bound has been

by

determined using Routh-Hurw

witz stability criteria in equation
(21). The lower bound will bee determined by pole-placement
technique and further the optiimized value of controller gain
will be obtained using Firefly optimization
o
algorithm.

xi x j . The movement of a firely i , attracted to a

brighter

xit +1 = xit + e

j is

firefly
rij2

(x

t
j

deteermined

by,

xit + t ti , where thee second term is

due to the attraction. The third term is randoomization with t

TABLE I. PARA
AMETERS OF FA

Number of
fireflies (n)
5

t
i

as the randomization parameter, and is a vector of random

numbers drawn from a gaussian distribuution or uniform
distribution at time t . If = 0 , it becomess a simple random
walk. The determines the variation of attractiveness
a
with
increasing distance from communicatedd firefly. = 0
corresponds to no variation and reduces to a variant of particle
swarm optimization. As t controls the randomness, this

Max
generation
100

Alpha

Beta

Gamma

0.5

0.2

0.5

C. Hybrid Pole-placement & FA

F for optimization of PID
parameters for Angular conntrol
The upper bound of Kp, Ki & Kd values of PID for angular
control has been determinedd in equation (21). The lower
bound of controller gain is prooposed to be determined by the
pole-placement technique, acccording to the given design

307

2016 IEEE First International Conference on Control, Measurement and Instrumentation (CMI)

specifications as follows. The closed loop characteristic

equation is given by:
( 0.236 - K d ) 2 -25.79 - K p
( -0.49 - Ki )
s3 +
s +
s+
= 0 (24)
4.008
4.008
4.008
The desired characteristic equation of closed loop system is:
s3 + ( 2 + m ) n s 2 + (1 + 2m 2 )n 2 s + mn3 = 0
(25)

Comparing equation (24) and (25), we get:

( 0.236 - K d ) = 2 + m
(
) n
4.008

4.008
( -0.49 - K i )

2 .2
2 .0

(26)

C ase
C ase
C ase
C ase
C ase
C ase

1 .8
1 .6
1 .4

(27)

Angle

( -25.79 - K ) = (1 + 2m )
p

D. Hybrid Pole-placement & FA for optimization of PID

parameters for Cart-Position control
The upper bound has been determined using Routh-Hurwitz
stability criteria in equation (21). The lower bound can be
determined by the pole-placement technique, according to the
design requirements as follows:

1 .2

s tu d y : 1
s tu d y : 2
s tu d y : 3
s tu d y : 4
s tu d y : 5
s tu d y : 6

1 .0
0 .8

= mn 3

(28)

0 .6

4.008
The design specifications to meet the unit step response of the
system are assumed as:
A settling time less than 5s.
An Overshoot less than 25% and
A Rise time less than 1s.
Hence, choosing some desired values of , n & m , the tuned
values of K p , K i & K d can be obtained from the equations

0 .4
0 .2
0 .0
0

X ( s)
F (s)

(26), (27) & (28), as per pole-placement technique. The values

thus obtained are shown in Table III. A case study was done
with six different combinations of , n & m values. The
angular displacement of pendulum for these cases is shown in
Fig. 5. Values of time response specifications were obtained
from the simulation results of the case study and are shown in
Table IV. From Table III, two ranges of K p , K i & K d values
were chosen and FA was applied for these ranges by
considering ITSE the performance criteria. Table V. shows the
optimal values of K p , K i & K d . After optimizing the values
of K p , K i & K d using FA, a unit step input is given to the
closed loop system shown in Fig. 5. The time response plot is
shown in Fig. 7.
TABLE II. THE VALUES OF PID PARAMETERS OBTAINED FROM
POLE-PLACEMENT

S
N

Closed loop parameters

= 0.8, n = 1.5, m = 9

-202.90 -97.88

= 0.8, n = 1.5, m = 5

-92.52

= 0.8, n = 1, m = 15

-106.75 -48.59

= 0.5, n = 1.2, m = 20

-89.28

-69.75

= 0.5, n = 1.5, m = 9

-75.39

-61.36

= 0.5, n = 1, m = 9

-47.83

-18.53

Kp

Ki

-11.31

10

15

T im e ( s )

20

25

30

Fig. 5. Angle of deviation for different values of closed loop parameters

( I + ml 2 ) s 2 + ds mgl
{[( I + ml )( M + m) (ml )2 ]s 4 + [b( I + ml 2 )
2

+ d ( M + m)]s 3 + [bd mgl ( M + m)]s 2 - m glb s}

For position control,
X ( s ) 5.841
(29)
= 2
F ( s)
s

The closed loop characteristic equation is given by

s3 + 5.841s 2 K d + 5.841sK p + 5.841K i = 0

(30)
The desired closed loop characteristic equation is given by
(31)
s 3 + (2 + m)n s 2 + (1 + 2m 2 )n 2 s + mn 3 = 0
Comparing equation (30) and (31), we get
(2 + m)n = 5.841K d
(32)

(1 + 2m 2 )n 2 = 5.841 K p

(33)

m n 3 = 5.841 K i

(34)

TABLE III. TIME DOMAIN PERFORMANCE SPECIFICATIONS FOR

DIFFERENT DESIRED VALUES FOR THE CLOSED LOOP SYSTEM

Kd
52.67
33.43
54.27
52.67
32.83
21.80

308

Case
Param Study:
1

Case
Stud
y:2

Case
Stud
y:3

Case
Study:
4

Case
Study:
5

Case
Study:
6

R.T

0.33

0.53

0.56

0.60

0.42

0.48

S.T

4.84

5.24

8.20

21.62

47.30

57.27

86.86

US
Peak
P.T

0
1.23
1.04

5.35
36.5
1
0
1.38
1.92

6.42

%OS

8.74
22.7
5
0
1.40
1.89

0
1.48
1.75

0
1.58
1.40

0
1.96
1.98

2016 IEEE First International Conference on Control, Measurement and Instrumentation (CMI)
TABLE IV. TIME DOMAIN PERFORMANCE SPECIFICATIONS
OBTAINED USING FA FOR DIFFERENT RANGES

Range: 1
[ K p = -100 to - 25.79,

Range: 2
[ K p = -200 to - 25.79,

K i = -100 to - 0.49,

K i = -100 to - 0.49,

K d = -100 to 0.236]

K d = -100 to 0.236]

Kp

-99.9987

-190.6896

Ki

-86.7029

-99.6066

Param

The ITSE is considered as the performance criteria for FA and

is applied to two ranges of K p , K i & K d values obtained from
Table VI. The optimal values of K p , K i & K d obtained by
minimizing the ITSE are shown in Table VIII.
V.

Considering the same design specification in section IVC after

(28) we vary the length of the inverted pendulum by 10% than
that of the nominal value an attempt was made to stabilize the
system by meeting the desired criteria using the proposed
method as well as LQR technique. The Table IX below depicts
that the proposed method could meet the desired specification
from the ranges of the gain calculated by satisfying the RouthHurwitz criteria and pole-placement technique. Also for this
case we have compared the transient response obtained by the
proposed method as well as LQR technique which is placed in
Fig. 8.
VI. CONCLUSION
In this paper an optimal PID controller is designed adopting a
hybrid control structure. The design technique combines the
traditional techniques in control theory and evolutionary
algorithm to stabilize a cart-inverted pendulum dynamics. The
result obtained for the system is convincing using the
proposed hybrid control law for a preliminary robustness
study. As a future work the control algorithm will be modified
to develop a robust PID controller capable to tackle parametric
uncertainties and inclusion of delay in the system.

Kd
-46.3397
-66.6712
R.T
0.4553
0.4328
S.T
5.2691
4.7834
Peak
1.4569
1.2454
P.T
1.4530
1.4067
%OS
44.8910%
23.9302%
US
0
0
Legend: R.T = Rise time, S.T = Settling time, P.T = Peak
time, OS = Overshoot, US = Undershoot.
1 .6

1 .4

R a n g e:1
R a n g e:2

1 .2

1 .0

Angle

ROBUSTNESS STUDY

0 .8

0 .6

0 .4

0 .2

TABLE V: CLOSED LOOP PARAMETERS FROM POLE-PLACEMENT

0 .0
0

10

15

20

25

30

T im e ( s )

Fig. 6 Angular displacement of pendulum with for the two ranges of

K p , K i & K d values specified in TABLE V.

Choosing some desired values of , n & m , the tuned values

of K p , K i & K d can be obtained from the equations (32), (33)
& (34), as per pole-placement technique. The values thus
obtained are shown in Table VI. The simulation results viz.
position of COG of pendulum with application of step input,
for the case study of different values of closed loop parameters
are shown in Fig. 6. Measurements of time response
specifications were obtained from the simulation results
shown in Fig. 7 and are presented in Table IV.
C a se
C a se
C a se
C a se
C a se
C a se
C a se

1 .4
1 .2

Position

1 .0

0 .6
0 .4
0 .2
0 .0
1

4
5
T im e ( s )

Closed loop system

parameters
= 0.8, n = 1.2, m = 9

Kp

Ki

Kd

3.09

2.13

1.80

2.

= 0.8, n = 3, m = 20

40.00

9.039

3.

= 0.8, n = 4, m = 20

72.86

4.

= 0.8, n = 4, m = 10

37.80

73.95
175.3
1
87.66

5.

= 0.7, n = 4, m = 10

29.58

6.

= 0.7, n = 5, m = 10

46.22

76.70
149.8
0

12.05
6.57
5.75
7.19

TABLE VI. TIME DOMAIN PERFORMANCE SPECIFICATIONS FOR

DIFFERENT DESIRED VALUES FOR THE CLOSED LOOP SYSTEM

s tu d y :1
s tu d y :2
s tu d y :3
s tu d y :4
s tu d y :5
s tu d y :6
s tu d y :7

0 .8

S
no.
1.

10

Fig. 7 Position of COG of pendulum for various values of PID parameters

309

Par
am

Case
Stud
y:1

Case
Study
:2

Case
Stud
y:3

Case
Stud
y:5

Case
Study
:6

Case
Stud
y:7

0.21

0.17

0.18

1.31
19.4
0

Case
Stud
y:4
0.20
05
1.35
20.7
0

R.T

0.67

0.27

0.20

S.T
OS
US
Pea
k
P.T

4.51
20.0
0

1.747
19.45
0

1.30
24.0
0

1.04
23.9
0

1.58
33.3
0

1.21

1.19

1.19

1.20

1.24

1.24

1.33

1.86

0.74

0.54

0.54

0.56

0.479

0.51

2016 IEEE First International Conference on Control, Measurement and Instrumentation (CMI)
TABLE VII. TIME DOMAIN PERFORMANCE SPECIFICATIONS
OBTAINED USING FA FOR DIFFERENT RANGES

1.6
1.5

Range: 2
[ K p = 0 to 100,

K i = 0 to 100,

K i = 0 to 100,

K d = 0 to 100]

K d = 0 to 10]

Kp

79.8383

60.0933

Ki

13.6195

59.1724

0.5

Kd
RT
ST
Peak value
PT
% OS
US

97.6556
2.1426
8.4746
1.1198
5.1932
5.7259
0

10
0.2494
2.3167
1.1064
0.7267
10.6351
0

0.3

1.3
1.2
1.1
1.0

Angle

Parameters

Kp

Range: Arbitrary
range selection
[ = -100 to -25.8,
-100 to -0.4905,
= -100 to 0.2185]
-99.9975

Range selection by
Proposed method
[ = -250 to -25.8,
-120 to -0.4905,
= -80 to 0.2185]
-192.2073

Ki

-65.7356

-99.0493

Kd
R.T
S.T
Peak
value
P.T
% OS
US

-66.1587

-42.2962

0.6602
7.5948
1.432

0.2782
4.4850
1.2392

2.046
41.3378
0

0.9161
23.2577
0

0.9
0.8
0.7
0.6

0.4

0.2
0.1
0.0
0

Time (sec)

10

Fig.8: Comparison of results under robustness study

TABLE VIII. CONTROLLER GAINS CORRESPONDING TO ROBUSTNESS STUDY

Paramet
er

Firefly (Range:2)
Firefly (Range:1)
Pole-placement

1.4

Range: 1
[ K p = 0 to 100,

REFERENCES:
[1]. Furuta. K, Yamakita. M , Kobayashi. Swing Up Control of Inverted
Pendulum, IEEE International Conference on Industrial Electronics,
Control and Instrumentation, 1991. Proceedings of IECON '91, 1991.
[2]. Yamakawa, T. Stabilization of an inverted pendulum by a high speed
fuzzy logic controller hardware system. Fuzzy sets and Systems, 32(2),
1989, 161-180.
[3]. Magana, M.E. (1998), Fuzzy-logic control of an inverted pendulum with
vision feedback, IEEE Transactions on Education, 41(2).
[4].Anderson, C. W. Learning to control an inverted pendulum using neural
networks. Control Systems Magazine, IEEE, 9(3), 1989, 31-37.
[5]. Yoshida.K(1999,June), Swing up Control of an Inverted Pendulum by
Energy-Based Methods IEEE American Control Conference, 1999.
Proceedings of the 1999 (Volume:6 ), pp. 4045 4047.
[6]. Teel, A. R.. A nonlinear small gain theorem for the analysis of control
systems with saturation. Automatic Control, IEEE Transactions
on, 41(9), 1996, pp. 1256-1270.
[7]. Chen, C. S., & Chen, W. L. Robust adaptive sliding-mode control using
fuzzy modeling for an inverted-pendulum system. Industrial Electronics,
IEEE Transactions on, 45(2), 1998, pp. 297-306.

310

[8]. Chang, W. D., Hwang, R. C., & Hsieh, J. G, A self-tuning PID control for
a class of nonlinear systems based on the Lyapunov approach. Journal
of Process Control, 12(2), 2002, pp. 233-242.
[9]. M. Hasan, C. Saha, Md. M. Rahman, Md. R. I. Sarker&S. K. Aditya.
Balancing of an Inverted Pendulum Using PD Controller, Dhaka Univ.
J. Sci. 60(1), 2012, pp. 115-120.
[10]. A. N. K. Nasir, R. M. T. Raja Ismail & M. A. Ahmad. Performance
Comparison between Sliding Mode Control (SMC) and PD-PID
Controllers for a Nonlinear Inverted Pendulum System, 2010.
[11]. D. Kumanan& B. Nagaraj. (2015, June). Tuning of proportional integral
derivative controller based on firefly algorithm.Systems Science &
Control Engineering: An Open Access Journal, 1:1, 52-56, DOI:
10.1080/21642583.2013.770375.
[12]. X. S. Yang &X. He. (2013). Firefly Algorithm: Recent Advances and
Applications. Int. J. Swarm Intelligence, Vol. 1, No. 1, pp. 3650.
[13]. X. S. Yang,Firefly algorithm, Levy flights and global optimization,
in: Research and Development in Intelligent Systems XXVI, Springer
London, pp. 209-218 (2010).
[14]. O. Roeva& T. Slavov. Firefly Algorithm Tuning of PID Controller for
Glucose Concentration Control during E. coli Fed-batch Cultivation
Process, in Proc. FedCSIS.,Wroclaw, 2012, pp. 455-462
[15]. Dos Santos Coelho, L.; De Andrade Bernert, D.L. Mariani, V.C.
(2011, June). "A chaotic firefly algorithm applied to reliabilityredundancy optimization," Evolutionary Computation (CEC), 2011
IEEE
Congress
on ,
vol.
5
no.
8
pp.517-521.DOI:
10.1109/CEC.2011.5949662.
[16]. L. D. S. Coelho, V. C.Mariani. Firefly algorithm approach based on
chaotic Tinkerbell map applied to multivariable PID controller tuning.
Computers & Mathematics with Applications, 64(8), 2012, pp. 23712382.
[17]. Sh. M. Farahani, A. A. Abshouri, B. Nasiri& M. R. Meybodi..A
Gaussian Firefly Algorithm. International Journal of Machine Learning
and Computing.1 (5), 2011, pp. 448-453.
[18]. K. Ogata, Mathematical Modeling of Mechanical Systems and
Electrical Systems in Modern Control Engineering, 5th ed. New Jersey,
U.S.A Pearson Education, Inc.
[19] Feedback Instruments Ltd., Operating Manual of Digital Control
Experimental Set up 33-200. East Sussex, UK, 2009.