1st IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

Robust Backstepping Control of Ball and Beam

System with Extemal Disturbance Estimator

l 2
Aditi Srivastava and Bhanu Pratap
IElectrical Engineering Department, National Institute of Technology Kurukshetra, Kurukshetra, India
2Member IEEE, Electrical Engineering Department, National Institute of Technology Kurukshetra, Kurukshetra, India
E-mail: [email protected]@gmail.com
Abstract-This paper presents a robust feedback Due to the nonlinear dynamics of the ball beam
controller for ball and beam system (BBS). The BBS is a system, various control techniques are studied and
nonlinear system in which a ball has to be balanced on a designed in the literature [1]-[7]. An adaptive dynamic
particular beam position. The proposed nonlinear controller
designed for the BBS is based upon Backstepping control
surface controller for BBS on the basis of Takagi-Sugeno
technique which guarantees the boundedness of tracking
(T-S) fuzzy logic is illustrated in [1]. Fuzzy based
error. To tackle the unknown disturbances, an external
feedback linearization controller is applied for the tracking
disturbance estimator (EDE) has been employed. The
and disturbance decoupling problem of AMIRA's BBS
stability analysis of the overall c10sed loop robust control [2]. A fuzzy based sliding mode control is designed for the
system has been worked out in the sense of Lyapunov theory. BBS in [3]. Also ant colony optimization has been used to
Finally, the simulation studies have been done to optimize the parameters of the controller. A robust and
demonstrate the suitability of proposed scheme. adaptive dynamic surface control for BBS with uncertain
Keywords-Backstepping Control; Ball and Beam System; disturbances is demonstrated in [4]. Here, the uncertainties
External Disturbance Estimator; Nonlinear System
of the system are approximated by interval type-2 fuzzy
I. INTRODUCTION neural networks. A nonlinear robust control of BBS is
implemented using state dependent Riccati equation
Ball and beam system is one of the most familiar (SDRE). SDRE is the extended form of linear quadratic
laboratory setup used for control experiments. It consists regulator which improves the robustness of the overall
of a long beam and a ball which is rolling back and forth system [5]. A modified ant colony optimization for fuzzy
on the top of the beam according to the changing angle of controller of BBS is applied to optimize the fuzzy rules,
the beam. The main function of this system is that the ball type of membership functions and parameters of
remains in contact with the beam and the rolling occur membership functions [6]. A virtual and remote laboratory
without slipping, which imposes the constraints on the for BBS is established using a leaming management
rotational acceleration of the beam. The mechanism of system, Moodle and Easy JAVA simulation (EJS) for the
BBS is under actuated. There are two degree of freedoms training of advanced techniques like robust, fuzzy and
reset control [7].
which are angle of beam which is actuated by servo motor The extemal disturbance estimator is implemented to
and the ball rolling on the beam which is not actuated. The estimate the unknown disturbances generated extemally
schematic diagram of BBS is depicted in Fig. 1[1]. which can be time varying. A generalized form of
disturbance estimator is reported to estimate the higher
order disturbances in the time series expansion [8]. The
robust observer based Backstepping controller for an
unknown nonlinear system with extemal disturbances [9].
Backstepping control approach is widely used to deal
the control issues of the systems having complex
nonlinearity. It is a recursive method on the basis of
Lyapunov theory of stability, which can properly handle
the systems with unknown model parameters. An adaptive
Backstepping controller for the position control of an
electro-hydrauJic actuator (EHA) is demonstrated in [10].
A 3-D path following nonlinear control of a six degree of
freedom (DOF) model scaled autonomous helicopter is
proposed using command fiItered Backstepping technique
Fig. 1: Ball and Beam System [11]. To control the exoskeleton for human upper Iimbs,
an adaptive Backstepping controller based on fuzzy
978-1-4673-8587-9/16/$31.00 ©2016 IEEE [1 )
 1st IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

approximations is reported in [12]. A fuzzy-neural based

adaptive Backstepping control of a hybrid magnetic (3b)
levitation system for the levitated and propulsive control
[13]. A nonlinear Backstepping direct power control (3c)
approach is applied to the doubly fed induction generator
(DFIG) driven by wind turbine [14]. (3d)
In this paper, a nonlinear robust controller is proposed
for BBS in the presence of external disturbances. Since y=x 1 (3e)
very less number of papers are available on disturbance
observer based control of ball and beam system. So, it is where, [x i e er = [x1 X2
novel work from the application view. Using coordinate state vector, =V, is the input and Y is the output.
U
change the BBS is transformed into an approximate Normal Form Modeling of the Nominal Plant:
normal form to apply the proposed schemes. An external Y
disturbance estimator has been developed to observe the Differentiating the output of ball and beam system
unknown disturbance. Further Backstepping control has (3), the robust relative degree is obtained as {4} at x = 0 .
been implemented on the basis of disturbance estimator. Now, considering z [h(x) LJh(x) L�h(x) L�h(x)r,
=

The reminder of the paper is as folIows. Section 11 the dynamics of BBS (3) can be rewritten in the normal
contains dynamics of the BBS along with the parameters. form given as [15],
Problem statements of the proposed work are expressed in Ljh(x)
Section III. In Section IV, design of external disturbance
estimator is explained. Robust Backstepping controller for L�h(x) (4)
BBS is illustrated in Section V. ResuIts obtained through Ejh(x) J;
the simulation studies are demonstrated in Section VI and J(x)+ g(x)u J(x)+ g(x)u
concluding remarks are given in Section vrr.
11. MODELlNG OF BALL AND BEAM SY S TEM
where,
Consider a ball and beam system, where a servo
motor is coupled to the ball and beam model. The servo
3 (x ) =---gsm
LJ h
5r . (r 3 ) 4'
7L
-x
L
x

motor drives a lever arm which is coupled to a track upon

which a rolling ball rests. The dynamics of a ball and -()x = --5g(r)
f
2. (r )x.2+-COS
- 5rg
sm (Lr
-X3 -X 3
)(K;KmKbJ x4
beam system with servo motor is given as follows [15], 7 L L 7L JeqR",
.. 5
x=-"7gsm . (Lr )
e (1) and -g(x) = --
5 g cos r (r ) ( 1 -x3
KK
g
--.
m
7L JeqRm
L
.. K K (
e =---.!L....!!!.. V-KKe
' ) (2)
b g
J eq R m I
III. PROBLEM FORMULATIONS
where, x is the ball position and e is the angular Modeling of the Plant with Uncertainty:
displacement of the servo motor. The parameters of the The dynamics of ball and beam system in the normal
system are given below in the Table 1 [15]. form (4) can be rewritten as
21 = Z2
TABLE I: PARAMETERS OF
BALL AND BEAM SYSTEM (5a)
Parameter Value 22 = Z3
Gear ratio ( Kg ) 70 : 1 (5b)
23 = Z4
Back EMF constant ( Kb) 0.00767 V/(rad/s) (5c)
Motor torque constant ( K", ) 0.00767 Nm!A (5d)
Beam length ( L) 43.18 cm (5e)
r
Lever radius ( )
Gravitational constant ( g)
2.5 4 cm
9.8 m!s'
where, [ z
,
] is the state vector after
z
,
Z
3
Z
4

Equivalent inertia ( J,,, ) 0.0029kg-m' coordinate transformation, 7(z) and g ( ) are the z

Motor resistance ( R", ) 2.60

nonlinear functions, de is the unknown external
o:!

disturbances which are bounded.

The dynamics of the ball and beam system (1) and (2) This work is framed with following objective:
can be represented in the state space form as follows [16], l. Design of nonlinear robust controller for BBS,
(3a) where the actual ball position Y converges to the
[21
 1st IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

desired ball position Yd and the tracking error Zu = Yd-KISI (13)

-'>oo (ji) = 0 .
ji = Y-Yd converge to zero, i.e., tlim where, KI is positive constant.
2. Design an external disturbance estimator for Substituting Z2d from (13), (12) becomes,
BBS, where the unknown disturbance estimation SI=2 S -KISI (14)
error dett(z,/)= dett(Z,/)-dett(z,/) converges to Differentiating Z2d from (13), gives
zero, i.e., !�� {dext(z,t)} = 0 . Z2d=Yd-KI (S2-KII S ) (15)
IV. EXTERNAL DISTURBANCE ESTIMATOR DESIGN S tep 2: Using (10) and (15), the time derivative of S2
is given by
The external disturbances present in BBS model are
assumed to be unknown, which can't be directIy used to S =S 3 +Z 3 d -z2d
2 (16)
design the robust Backstepping controller (RBC). Thus, to to stabilize the system (16) virtual control Z3d can be
eliminate this problem, a nonlinear external disturbance chosen as,
estimator (EDE) has been developed for BBS. The Z3d = Z2d-K22 S -I S (17)
estimate of external disturbance dett (z, I ) is designed by Where, K 2 is a positive constant. On substituting the
[8], value of Z3d from (17), (16) becomes
dett(Z,/)= rO(z4 -z) (6) S2=S 3 -K22 S -I S (18)
where, ro 0, and
>
Differentiating (17) results,
Z = 1(z)+ g(z)u +dett(z,/) (7)
Z 3d=Z2d-K22 S -I S (19)
The estimation error between unknown external
disturbance dex1 (z,t) and its estimate dett (z,/) can be where, 2 z d=Yd-K]{S 3 -(K]+K2)S2 -(1- KnSI} .
expressed as S tep 3: On using (10) and (19), the time derivative of
dett(z,/)= dett(Z,/)-dett(z,/) (8) S 3 can be written as
The disturbance error dynamics is given by S 3 =S 4 +Z 4d-Z 3d (20)
dett (z,/)= dew (z,/)-rodew (z,/) (9) virtual control Z4d to stabilize the system (20) can be
taken as
Also, Idew (z,/)1 � e-tolldew (O)I+�po ( I ) , Z4d = Z 3 d-K 3 S 3 -2 S (21)
ro where, K 3 is a positive constant. On substituting Z4d
where, Po ( I ) :?: dett (z,/) is an envelope function, from (21) into (20), resuIts
Vt :?: O. Thus, characteristics of external disturbance S 3 =S 4-K 3S 3 -2 S (22)
estimation are exponentially stable, where the steady state Differentiating (21),
error depends upon dett (z, I ) .
Z 4d=Z 3d-K 3S 3 -2 S (23)
V. ROBUST BACKSTEPPING CONTROLLER DESIGN Where,
The robust Backstepping controller is designed in the Z 3d="Y� -(K]+K2)8 3 + {K](K]+K2)+K� -I} 2 8
following four steps [9]: +{K](I-Kn+K]+K2}8].
S tep J: Define the sliding surfaces,
Sj = Zj -Zjd (10) S tep 4: Using (10) and (23), the time derivative of S 4
where, Zjd represents the desired value of state Zj for can be expressed as,
i = 1, ... ,4 . Now from (10), we have S 4 = 1(z)+ g(z)u -Z 4d (24)
(11) To satisfy the stability of the system (24), actual
Let Zld = Yd and using (10) and (11), the time control effort can be selected as
derivative of I S can be found as u = [g(z)J' [-I(z)+24d -K4S4 -S3 - Risl Sgn(S4)] (25)
SI=2 S +Z2d-Yd (12) where, K 4 and are the positive constants. The
a

where, Z2d can be regarded as a virtual control to robust term -K' I SJ sgn( S4 ) is taken for the rejection of
stabilize the dynamic system represented in (12) by external disturbance. On substituting u from (25), (24)
seIecting becomes

[31
 1st IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

S4=-K4S4-S3-KIS Jsgn (S4 ) (26) E

,s. 0
Stability Analysis:
Consider a Lyapunov function candidate, gQ) -0.05
g> -0.1
V= S S2 +.!.S23 +.!.S24
.!.2 +.!.2 (27) :.i2
� -0.15
21 2 2 2 � -0.2
Differentiating (27), o
� -0.25
�=��+��+��+�� O� 8. -0.3
co
Putting the values of SI' S2' S 3 and S 4 from (14), co
o 5 10 15 20 25
(18), (22), and (26) in (27), Time (sec)
V = S, (S, - K,S,)+S, (S3 - K,S, - S,)+S3 (S4 - K3S3 - S,) Fig. 3: Tracking Error ofBall Position with Disturbances

(29) 0.2 �--�--�--�-�

(5 0.1
(30) G.
t:::

Hence, the � is negative definite. Therefore, the gQ) 0

controller errors I S ' S2' S 3 ' and S 4 converge e
c
exponentially to zero. Next section presents the simulation 8 -0.1
resuIts to validate the proposed robust control scheme in
the presence of external disturbances. -O.2 �--�--�--�-�
o 5 10 15 20 25
VI. SIMULATION RESULTS Time (sec)
The initial conditions of the BBS plant and external Fig. 4: ControI Effort AppIied to BBS with Disturbances

disturbance are selected as [ 0 . 5 0 0 Of and 0.5 -

x
Q)
:2.
respectively. "" 10 ----- Extemal disturbances
C/)
The robust controller design parameters are chosen as, Q)
u -- Estimate of disturbances
c 8
K, = K2 = K3= K4= K= 2 and a=0.5. The EDE gain CI)
-e
:::> 6
ro is selected as 20 . 1i5
i5
The desired trajectory and extemal disturbance is co
E
4
chosen as x
Q)
2
Q)
Yd =0.3 sin(0.2m)+0.2cos(0.1m)
'0 0
dex1 =-sin(IX )-cos(x2) +sin(2m) . E0
z 0 5 10 15 20 25
Time (sec)
A. Tracking Performance with Disturbances
Fig. 5 : Estimation ofUnknown ExternaI Disturbances
0.6 r-r---�--�----,
'9
x 2
Q)
0.4 g
E 0
,s. 0.2 gQ)
c c
0 -2
:;:; 0
'(i;
0
0 �
Cl. E -4
co
�Q)
-0.2
co Q) -6
u
c
CI)
-0.4 -- Actual ball post ion -e -8
:::>
(i)
o 5 10 15 20 25 i5 -10
Time (sec) '0 0 5 10 15 20 25
E Time (sec)
Fig. 2: Tracking ofBall Position with Disturbances
Fig. 6: Estimation Error of Unknown ExternaI Disturbances

[41
 1st IEEE International Conference on Power Electronics, Intelligent Control and Energy Systems (ICPEICES-2016)

B. Tracking Performance without Disturbances VII. CONCLUSION

0.6 �----�----� Robust Backstepping controller design for BBS is
0.4
proposed in this paper. Stability of the overall system is
E
defined by the Lyapunov theory of stability. Actual ball
� 0.2 position properly tracks the desired trajectory and the
c:

�Cf)
tracking error between these converges to zero. The EDE
0
0 is augmented to estimate the unknown external
a.
disturbances of the BBS. The tracking error between the
(ij -0.2
III external disturbance and its estimate converges to zero.
-0.4
--..-- Desired trajectory Finally, robustness issue also has been addressed. The
-- Actual ball postion real-time implementation of the proposed robust control
o 5 10 15 20 25 has been worked out, it may be reported in near future.
Time (sec)
REFERENCES
Fig. 7: Tracking ofBall Position without Disturbances
[1] Y-H Chang, W-S Chan, and C-W Chang, "T-S fuzzy model-based
E adaptive dynamic surface control for ball and beam system, " IEEE
� 0 Transactions on Industrial Electronics, vol. 60, no. 6, pp. 2251-2263,
�
2013.
g -0.05
Q) [2] T. L. Chien, C. C. Chen, M. C. Tsai, and Y. C. Chen, "Control of
0)
c: -0.1 AMlRA's ball and beam system via improved fuzzy feedback
� linearization approach, " Applied Mathematical Modelling, vol. 34, pp.
<.> -0.15 3791-3804,2010.
jg
c: -0.2 [3] Y. H. Chang, C. W. Chang, C. W. Tao, H. W. Lin, and J. S. Taur,

�Cf) -0.25
"Fuzzy sliding-mode control for ball and beam system with fuzzy ant
colony optimization, " Expert Systems with Applications, vol. 39, pp
0
a. 3624-3633,2012.
-0.3
(ij [4] Y. H. Chang and W. S. Chan, "Adaptive dynamic surface control for
III uncertain nonlinear systems with interval type-2 fuzzy neural networks,
o 5 10 15 20 25 " IEEE Transactions on Cybernetics, vol. 44,no. 2, pp. 293-304,2014.
Time (sec) [5] E. V. Kumar, J. Jerome, and G. Raaja, "State dependent Riccati
equation based nonlinear controller design for ball and beam System, "
Fig. 8: Tracking Error ofBall Position without Disturbances Procedia Engineering, vol. 97, pp. 1896-1905,2014.
[6] O. Castillo, E. Lizimaga, J. Soria, P. Melin, and F. Valdez, "New
approach using ant colony optimization with ant set partition for fuzzy
0.15 control design applied to the ball and beam system, " Information
Sciences, vol. 294, pp. 203-215,2015.
'0 0.1 [7] L. Torre, M. Guinaldo, R. Heradio, and S. Dormido, "The ball and
G
t:: beam system: A case study of virtual and remote lab enhancement with
0.05
g
Q)
moodle, " IEEE Transactions on Industrial Informatics, vol. 11, no. 4,
pp. 934-945,2015.
0
e [8] K. S. Kim, K. H. Rew, and S. Kim, "Disturbance observer for
'E -0.05
0
estimating higher order disturbances in time series expansion, " IEEE
ü Transactions on Automatic control, vol. 55, no. 8, pp. 1905-1911,2010.
-0.1 [9] W. Kim and C. C. Chung, "Robust output feedback control for
unknown nonlinear systems with external disturbance, " lET Control
Theory and Applications, vol. 10, no. 2, pp. 173-182,2016.
o 5 10 15 20 25 [10] K. K. Ahn, D. N. C. Nam, and M. Jin, "Adaptive Backstepping control
Time (sec) of an electrohydraulic actuator, " IEEE/ASME Transactions on
Mechatranics, vol. 19, no. 3, pp. 987-995,2014.
Fig. 9: Control Effort Applied to BBS without Disturbances
[11] B. Zhu and W. Huo, "3-D path-following control for a model-scaled

The tracking of ball position with and without autonomous helicopter, " IEEE Transactions
Technology, vol. 22, no. 5, pp. 1927-1934,2014.
on Contral Systems

disturbances is given in Fig. 2 and Fig. 7 respectively. Fig. [12] Z-J Li, C-Y Su, G. Li, and H. Su, "Fuzzy approximation-based adaptive

3 and Fig. 8 represents the tracking error of the ball backstepping control of an exoskeleton for human upper limbs, " IEEE

position with and without disturbances. It can be seen that [13]

Transactions On Fuzzy Systems, vo1.23, nO.3, pp. 555-566,2015.
R-J Wai, J-X Yao, and J-D Lee, "Backstepping fuzzy-neural-network
the ball position converge to the desired trajectory very control design for hybrid maglev transportation system, " IEEE
quickly. The control effort applied to the BBS with and Transactions on Neural Networks and Learning Systems, vol. 26, no. 2,

without disturbances is plotted in Fig. 4 and Fig. 9. The [14]

pp. 302-317,2015.
P. Xiong and D. Sun, "Backstepping-based DPC strategy of a wind
estimation of external disturbances is depicted by Fig. 5 turbine-driven DFIG under normal and harmonic grid voltage, " IEEE
and estimation error of external disturbances is shown in Transactions on Power Electronics, vol. 31, no. 6, pp. 4216-4225,2016.

Fig. 6. It can be observed that the error in estimation of the [15] L. Behera and l.
University Press, 2009.
Kar, Intelligent Systems and Control, Oxford

external disturbance is approximately equals to zero. [16] Aditi Srivastava and Bhanu Pratap, "Nonlinear Robust Observers for
These simulation results demonstrate the robustness of the Ball and Beam System: A Comparative Analysis, " 2n
d
International

proposed control scheme. Conference on Recent Advances in Engineering and Computational

Sciences, Chandigarh, lndia, Dec. 21-22,2016.

[51