Direct Adaptive Control Based Sliding Mode Observer Designed For Fractional Nonlinear Systems

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Direct adaptive control based on sliding mode observer for

fractional nonlinear systems


D. Elleuch, T. Damak

Abstract— A combination between a direct adaptive control tracking trajectory and to estimate the state when the
and a sliding mode observer is designed for a canonical unknown dead zone is present or not. In [1], the authors
fractional nonlinear system. This combination guaranties the studied the performance of a direct adaptive controller
tracking of trajectory and the estimation of the unmeasured developed for a canonical fractional nonlinear system.
state. The simulation results prove the combination robustness
in presence of unknown parameters. In this work, we will study the performance of a sliding
mode observer combined to a direct adaptive controller. The
I. INTRODUCTION combination is designed to a canonical system having a
In general, the developed control laws depend on the fractional nonlinearity.
parameters and the states of system. When these parameters
and states are unknown, a problem is presented. Preliminary [1]
To remedy this problem, in literature, controller-based
observers are presented ([6], [7], and [8]). The output Consider the fractional non linear system having a
feedback control is widely studied for various classes of triangular form
nonlinear systems. The separation principle is to design and  x i  xi 1
study separately the performance of the controller and the 

observer. It isn’t usually verified and it is limited for a linear  x n 
1
 f x   g x u  ; i  1 n  1 (1)
system and some class of nonlinear systems.    x 
Controllers based on observers are especially designed for 
 y  x1
uncertain nonlinear systems ([9], [10], [11], [12]) and
nonlinear systems Lipchitz nonlinear systems ([13], [14], With:
[15], [16]). The combination designed for uncertain u : the control input, x : the state, y : the output,
nonlinear systems ensures the stabilization and tracking f ( x ), g( x ),  ( x ) : the nonlinear function
trajectory.
For the triangular nonlinear systems, the combination is f x    T w f x   f 0 x  ;  x    T w x    0 x 
possible only under restricted conditions ([10], [17]). In [14],
the author studied the duality of an observer and a controller
and defined a sufficient condition to guaranty the quadratic
Define the vectors x d , and a filtered error  s as:
( n 1 )
stability. xd  [ y d y d  y ]T ;
The combination is made between different architectures   x  xd  [   1  n ] T ,
of observers and controllers such as sliding mode control
d
and observer [18], high gain observer and proportional s  (   )n1  1  [ T 1 ] ,
derivative control [19], etc .... dt
Where the constant   0 and
In ([20], [21]), the authors studied the control of nonlinear
system, converted to a canonical form, in which the   [ n1 ( n  1 )n2 ( n  1 ) ] T
nonlinear function is an unknown dead zone function. The The derivative of  s is written as follow:
combination observer-controller proved robustness to
1
s  [ f ( x )  g( x )u ]  
( x )
D.Elleuch. is with University of Sfax, National Engineering School of    y d( n )  [ 0 T ]
Sfax (ENIS), Laboratory of Sciences and Techniques of Automatic control
& computer engineering (Lab-STA), Tunisia B.P. 1173, 3038 Sfax-Tunisia The state x n can be written as:
(e-mail: dorsafelleuch@ yahoo.fr).
T.Damak,., is with University of Sfax, National Engineering School of xn   s  y d( n1 )  [ T 0 ]
Sfax (ENIS), Laboratory of Sciences and Techniques of Automatic control
We note x  [ x1 x2  xn1 ] and
& computer engineering (Lab-STA), Tunisia B.P. 1173, 3038 Sfax-Tunisia T
(e-mail: tarak.damak@enis.rnu.tn).
 1  y d( n1 )  [ T 0 ] III. SLIDING MODE OBSERVER
The sliding technique is mainly used to construct a
Definition [1] nonlinear observer. The sliding mode observer is developed
to divers’ class of system such as Lipchitz nonlinear systems
For a bounded reference vector, a scalar function ([2]-[3]), and uncertain nonlinear systems ([4], [5]). In this
s part, a sliding mode observer designed for the fractional
V     x ,   1  d (2) nonlinear system (2) is described by:
0  x̂ i  x̂i 1  si
 ( x1 ,   1 )   ( x1 ,   1 ) ( xn ) 
Or 
 x̂ n   x̂   f x̂   g x̂  u   s n
1
(5)
the function V is a (WCLF) if: 
  ( x ) is a smooth function ; i  1 n  1
  ( x ) is positive definite in the filtered error  s With:
 V is a radially unbounded with respect to  s i.e e1  x1  x̂1 ; si  k i e1  i signe1  ;
V   as  s   and s n  k n e1  n signe1  ; f x̂    T w f x̂   f 0 x̂  ;
 V  0 ,   s  0  x̂    T w x̂    0 x̂ 
 ( x ) is called a weighting function (WF)
The sliding mode observer (5) is combined to the adaptive
controller ((3), (4)). The combination stability is studied in
II. DIRECT ADAPTIVE CONTROL
the next section.
Consider the fractional non linear system having a
triangular form IV. DIRECT ADAPTIVE CONTROLLER BASED ON SLIDING
 x i  xi 1 MODE OBSERVER
 The separation principal is not usually verified for all

 x n 
1
 f x   g x u  ; i  1 n  1 systems. For that a stability analysis of the direct adaptive
  x  controller and the sliding mode observer is the object of this

 y  x1
section.
The stability and robustness combination is studied under
With: the following assumptions:
u : the control input, x : the state, y : the output, Assumptions
f ( x ), g( x ),  ( x ) : the nonlinear function 1 ) W ( x )  W f ( x )
f x    T w f x   f 0 x  ;  x    T w x    0 x  , 2 )Wf ( x ) 
In [1], the authors developed a direct adaptive control 3 ) The two vectors are known sign
designed for the system (1). A Weighted Control Lyapunov g( x )
Function (WCLF) (2) is developed and used to eliminate the 4) 0
parameter nonlinearity. The direct adaptive control has the ( x )
following architecture: Let (5) the sliding mode observer designed to system (1)
 The control law is: Considering the following variable:
1  k s ˆ T   
v   y dn   0 T  ; v1  y dn 1  T 0  ; 
u    wz   hz 

g x    x   x̂ n   s  v1 ;   x̂    x̂  x̂  ;
 
(3)
x̂ n   s  y dn 1  T 0  ; x̂  x̂1 x̂ 2 x̂ n 1 
T
 The adaptation law is:

 s 
1
 f x̂   g x̂ u   x̂ sn   v
ˆ   x̂ wz  s (4)  x̂ 
The global asymptotic stability and the tracking trajectory Let the Lyapunov function
s
 
are proved. These results are tested throughout two
examples. V     x̂ ,  v1 d ;
The controller used all the system states but these are not 0
all measured, to ensure the controller performance, an
observer is combined to the controller.
Then the V derivative is

s
  x̂ ,  v1  
V    x̂ ,  v1  s  s      x̂ d  
  k s 
k
 s 


V   x̂  s  x̂   x̂  
0  x̂  
s

   x̂ ,  v1   
 T
w  z   g  x̂ u  h  z 
0 
 
v1

v 1


d
s n x̂  s  T w x̂    0 x̂   
~
Let     ˆ the parameter error.
Or v  v1 The Lyapunov function of the direct adaptive
s

   x̂ ,  v1   controller with sliding mode observer is
  
0  v1
v1 d 

~ ~ ~ 
V  V   T  1  V  V   T  1ˆ
We have:
s
 
 
v   sj   x̂ ,  v1     x̂ ,  v1 d    k 2  k s


 ˆ T w z   g x̂ u 
  V   s
   x̂  s  x̂  
 x̂   
0

The V expression becames:  hz  


V   x̂    v  
  s s n s 
s   x̂  ˆ T w x̂    x̂  
 0 
s
   x̂ ,  v1      d  x̂  s T wz   s n x̂  s T w x̂    T  1ˆ
~ ~ ~ 
  x̂
x̂  v  x̂ ,  v1

0
k 2
  x̂  V   s  0
  s  f x̂   g x̂ u   x̂ s n    x̂ 
 x̂  If
s
   x̂ ,  v1    
  u
1  k s ˆ T


  wz   hz 
0 

x̂
x̂  v  x̂ ,  v 1  d


g x̂    x̂  

This is implying: ˆ   x̂ wz  s

V   x̂  s  T wz   g x̂ u  hz     
s n x̂  s ˆ T w x̂    0 x̂  

s n x̂  s  T w x̂    0 x̂   s n x̂  s T w x̂   0
~
With z  x̂ xd y dn T
;
Or s n  0 then t  
s
 w x̂ ,  v1     In sliding mode we have:
 x̂  s n  k n e1  n signe1   0
w z   w f  x̂  
1
 
 s x̂  0 
x̂  d
 vw x̂ ,  v1 
    kn 
n signe1 

n

  0 x̂ ,  v1  
s   e1 e1

h z   f 0  x̂  
1  x̂   t  e 0 k 
 s x̂  0   d
x̂ When 1 this is implies n
v 0 x̂ ,  v1   For that we conclude this theorem:
Theorem:
The control law and the sufficient condition for stabilizing
Which s   then: the direct coupling of the adaptive controller and sliding
mode observer for this class of system are defined by:
vw  x̂   The control law is:
w z   w f x̂  
 x̂  1  k s ˆ T 
u     wz   hz 
v 0 x̂  g x̂    x̂  
hz   f 0 x̂  
 x̂   The adaptation law is:

ˆ   x̂ wz  s
The combination performance is guarantied if the
observer gain satisfies:
ki  R pour i  1n  1 et k n   m  0.2kg ; M  1kg ; l  0.5 m ;ˆ 1 0   1 ;
ˆ 0   0.01;ˆ 0   0.3;  1 ; k  10 ;
The control and the adaptation laws have the same form
as those obtained by [1]. This means that if the observer and 2 3
the system have the same fractional form we will have the x1 0   x 2 0   0 ; x̂1 0   x̂ 2 0   1; L1  30 ;
same laws and therefore it is sufficient to study the adaptive
control based on the equation of the observer. 
L2  500 ;  1  10 4 ;  2  0.01 ; y d  sin( t );
6
  diag 0.04 ,0.02 ,0.03;
The performance of this coupling is tested on the example of
the pendulum treated in [1].
Although the parameters do not converge to their desired
value (Fig.1.e, f, g), the direct adaptive controller coupled to
V. SIMULATION a sliding mode observer has a good perofrmance to tracking
trajectory (Fig.1. a) and state estimation (Fig.1.c, d).
The inverse pendulum plane is described by the following
The output y
equations: 2

 x1  x 2

 1 g sinx1    2 x 22 sinx1  cos x1   cos x1  u
1.5
the reference yr
 the output y
x
 2 
  3   2 cos 2 x1  1

0.5

 
 1  m  M  0

 
Where    2  ml  ; -0.5

 
 3  l m  M 
4
-1
 3  0 5 10 15
time(s)
20 25 30

v1  y d   x1  y d  ; v   yd   x2  y d  ;


1.a. The tracking trajectory
 s  x̂1  y d    x1  y d  ;  x̂   1
d
The control law u
dt 20

 g sinx1   15

 
wz   v1 x 2 sinx1  cos x1   v cos x1  ;
2
10

v 
  5

h z   0 0

 Control law: -5

u
1
cos x1 

 k s  ˆ T wz   -10

-15
 Adaptation law:

ˆ  wz  s
-20
0 5 10 15 20 25 30
time(s)

 The sliding mode observer :


1.b. The control signal
 x̂ 1  x̂ 2  L1 e1   1 signe1 

 ˆ 1 g sinx̂1   ˆ 2 x̂ 22 sinx̂1  cos x̂1   cos x̂1  u
 2
x̂  
 ˆ  ˆ cos 2 x̂ 
 3 2 1

 L e
2 1   2 sign e 1 
 The parameter simulation
Estimate the state x1 Estimate the 2nd parameter
2 1
desired state desired parameter
estimated state 0.8 estimated parameter
1.5
0.6

1 0.4

0.2
0.5
0

-0.2
0
-0.4

-0.5 -0.6

-0.8
-1
0 5 10 15 20 25 30
-1
temps(s) 0 5 10 15 20 25 30
time(s)

1.c. Estimate of the state x1


1.f. Estimate of the parameter 2

Estimate the 3rd parameter


1
Estimate the state x2
10 0.8
desired state
8 estimated state 0.6

6 0.4
desired parameter
4 0.2 estimated parameter

2 0

0 -0.2

-2 -0.4

-4 -0.6

-6 -0.8

-8 -1
0 5 10 15 20 25 30
-10 time(s)
0 5 10 15 20 25 30
time(s)
1.g. Estimate of the parameter  3

1.d. Estimate of the state x2 Fig1. Tracking trajectory, state and parameter estimation by
combining the direct adaptive control and the sliding mode
2
Estimate the 1st parameter observer with unknown constant parameter

VI. CONCLUSION
1.5
In this paper, a combination between an adaptive direct
1
control and a sliding mode observer is studied for a
canonical non linear system having a fractional nonlinearity.
0.5
The combination stability is studied throughout the quadratic
Lyapunov function. The stability and the robustness are
guarantied under defined condition. The simulation results
0
desired parameter prove a good performance to tracking trajectory and estimate
estimated parameter the state.
-0.5

REFERENCES
-1
0 5 10 15 20 25 30 [1] Ge S.S., Hang C.C., T-Zhang, “A direct adaptive controller for
time(s) dynamic systems with a class of nonlinear parametrizations” Automatica
vol 35 pp 741-747. 1999.
1.e. Estimate of the parameter  1 [2] Elleuch.D, Damak.T, “Adaptive sliding mode observer for Lipchitiz
nonlinearsystems with varying parameters”, Proceedings 10th international
conference on Sciences andTechniques of Automatic Control and Computer
Engineering, STA'09, Hammamet , December 20-22, 2009, (STA 2009)
[3] Cheng C-F., “Design of adaptive observation schemes for lipschtiz
nonlinear systems” JSME international journal, series C, Vol 44, N° 3, 2001
[4] Raoufi R., Zinober A. S. I., “Smooth adaptive sliding mode
observers in uncertain chaotic communication” International Journal of
Systems Science, Vol. 38, No. 11, November 2007, 931–942
[5] Veluvolu K. C., Soh Y. C., Cao W., Liu Z. Y., “Observer with
Multiple Sliding Modes for a Class of Nonlinear Uncertain Systems”
American Control Conference June 8-10, USA, 2005
[6] Gaaloul A., M’sahli F. “High gain observer based higher order
sliding mode controllers” proceedings of the 7th Asian control
conference,China,August 27-29,2009
[7] Milman R., Bortoff S.A., “Observer based adaptive control of a
variable reluctance motor:experimental results” IEEE transaction on
control systems technology vol 7,n°7,September 1999
[8] Tkackev S.B., Alexeenkov S.G., “Numerical algorithms for
nonlinear observer based control” IEEE, phys con 2003, Russia
[9] Daly J.M., Wang D.W.L., “Output feedback sliding mode control in
the presence of the unknown disturbances” Systems & control letters, vol°
58, 2009, pp: 188-193
[10] Qian C., Lin W., “Output feedback control of a class of nonlinear
systems: a nonseparation principle paradigm” IEEE transactions on
automatic control, Vol°47, N°10, October 2002
[11] Gibbens P.W., Fu M., “Output feedback control for output
tracking of nonlinear uncertain systems” Proceeding of the 30th conference
on decision and control, Brighton, England, December 1991
[12] Choi H.H., Yi H.K., Chung M.J., “ Sliding mode control of
nonlinear uncertain systems using a sliding mode observer” Proceedings on
10 IEEE conference region, Computer, Communication, Control and Power
Engineering, 1993, Beijing
[13] Areak M., Kokotovic P., “Observers-based-control of systems
with slope-restricted nonlinearities” IEEE transaction on automatic control,
vol°46, N°7, July 2001
[14] Song B., Hedrick J. K., “Observer-based dynamic surface control
for Lipschitz nonlinear systems” Proceeding of the 42nd IEEE conference
on decision and control, Maui, Hawaii USA, December 2003
[15] Yu F., Su H., Chu J., “H control for Lipschitz nonlinear
systems via output feedback” Proceedings of the 3rd word congress on
intelligent control and automation, China, June 28- July 2, 2000
[16] Pagilla P.R., Zhu Y. “controller and observer design for Lipschitz
nonlinear systems” Proceeding of the American control conference, Boston
Massachusetts, June 30-July 2, 2004
[17] Alimhan K., Inaba H., “Output feedback control for a class of
nonlinear systems” International Journal of automation and computing,
Vol° 3, 2006, pp: 215-221
[18] Nollet F., Floquet T., Perruquetti W., “Observer-based second
order sliding mode control laws for stepper motors” Control engineering
practice, vol°16, 2008, pp:429-443.
[19] Heredia J.A., Yu W., “A high-gain observer based PD control for
robot manipulator” Proceeding of the American control conference,
Illinois, June, 2000
[20] Jasim I. F. “Stable Robust Adaptive Controller and Observer
Design for a Class of SISO Nonlinear Systems with Unknown Dead Zone”
World Academy of Science, Engineering and Technology, Vol°62, 2010
[21] Wu X.L., Wu X.J., Luo X.Y., Zhu Q.M., “Design of observer-
based adaptive controller for nonlinear systems with unmodeled dynamics
and actuator dead zone” International journal of automation and
computing, 2010

You might also like