Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/253730254

The Observer Adaptive backstepping Control for a Simple Pendulum

Article · June 2008


DOI: 10.1063/1.2953059

CITATIONS READS
4 117

3 authors, including:

Messaoud Mokhtari Noureddine Golea


University of Batna 2 Université Larbi Ben Mhidi
6 PUBLICATIONS   6 CITATIONS    77 PUBLICATIONS   615 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Control of a Permanent Magnet Synchronous Machine Using Neural Network Backstepping Approach View project

Adaptive Neural Network Control for Inverted Pendulum Using Backstepping with Uncertainties View project

All content following this page was uploaded by Messaoud Mokhtari on 01 May 2017.

The user has requested enhancement of the downloaded file.


The Observer Adaptive backstepping Control for a Simple Pendulum
M. Mokhtari, N. Golea, and S. Berrahal

Citation: AIP Conf. Proc. 1019, 85 (2008); doi: 10.1063/1.2953059


View online: http://dx.doi.org/10.1063/1.2953059
View Table of Contents: http://proceedings.aip.org/dbt/dbt.jsp?KEY=APCPCS&Volume=1019&Issue=1
Published by the AIP Publishing LLC.

Additional information on AIP Conf. Proc.


Journal Homepage: http://proceedings.aip.org/
Journal Information: http://proceedings.aip.org/about/about_the_proceedings
Top downloads: http://proceedings.aip.org/dbt/most_downloaded.jsp?KEY=APCPCS
Information for Authors: http://proceedings.aip.org/authors/information_for_authors

Downloaded 28 Jun 2013 to 10.0.50.139. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://proceedings.aip.org/about/rights_permissions
The Observer Adaptive backstepping Control
for a Simple Pendulum
M. Mokhtari*. N. Golea**. S. Berrahal***

* Department of electronics, Batna University, 05000 Batna, Algeria,


(e-mail : messaoud.mokhtari@yahoo.fr).
**EE Institute, Oum El-Bouaghi University, 04000 Oum El-Bouaghi, Algeria
(e-mail : n.golea@lycos.com)
*** Department of electronics, Batna University, 05000 Batna, Algeria,
(e-mail : razik61@yahoo.fr).

Abstract: In this paper, the adaptive backstepping control with an observer is presented for the class of
nonlinear systems. The problem of observability has a practical importance, because certain intern
variables are some times inaccessible to measurement or "expensive" to measure. Generally the physical
utilization of a sensor is impossible for reasons of implementation or cost, which do not allow the
measurement of all the states. We will see how we can start from measurements of inputs and outputs of
the process, to reconstruct (we also say to observe), the complex state vector. The subsystem, which
carries out this reconstructing, is called the observer. To clarify the approach further, an example of
simple pendulum is studied and the results of simulation clearly show the power of this approach.
Keywords: Nonlinear systems, Adaptive, Control, Lyapunov, backstepping, Observer

1. INTRODUCTION

In recent years, there was much progress concerning the


control of the nonlinear systems. Adaptive backstepping
control is considered as one of the principal results in this Output Y
Input U Process
field of research [1][2][3]. Its development is based on the
Lyapunov function which will guarantee the stability of the
system to be controlled. The design methods yields a force
control law with a parameter adaptation and update control
law. In this paper, we will shed the light on the technique of
control backstepping with observer and we will deduce its
efficacy.

2. THE OBSERVER ADAPTIVE BACKSTEPPING X̂


CONTROL (Estimate of X)
The observer has like inputs the inputs and the outputs of the Observer
real process and like output the estimated value (rebuilt) of
the state of this process (figure 1).
Thus, the problem consists in reconstructing this observer. Fig. 1. General diagram of the observer
For a given process, a system is defined by its equations of
state, which is the output that gives an estimate of the real
To achieve the aim that has already laid down in this part, we
state of the process. This estimate comprises an error which
must adopt all the assumptions in order to introduce the
must tend towards zero. When this property is satisfied, the
observer is known as asymptotic. observer. We will treat an example according to the usual
steps of the adaptive backstepping control.
ATTACHMENT
CREDIT LINE (BELOW) TO BE INSERTED The
ON THE
first FIRST PAGE
principle OF EACH
consists PAPERtwo diagrams that
in exposing
may clear up the difference between the non adaptive control
(figure 2) and the adaptive control (figure 3) with observer.

CP1019, Intelligent Systems and Automation, 1st Mediterranean Conference


edited by H. Arioui, R. Merzouki, and H. A. Abbassi
© 2008 American Institute of Physics 978-0-7354-0540-0/08/$23.00

85

Downloaded 28 Jun 2013 to 10.0.50.139. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://proceedings.aip.org/about/rights_permissions
These
These additional
additional conditions
conditions lead tolead to the form:
the form:
T
= x + ϕadditional
x&1 =x&1xThese T (y) .θ
2 + ϕ21 (y) 1.θ
conditions lead to the form:
Control
Control ϕT2.(y)
x&&2x3=+xϕ32+(y)
x& 2 =
T
θ T.θ
signal
signal Output
Output x1 = x 2 + ϕ 1 (y) .θ
InputInput . .
Control
Control Process
Process T
. .x& 2 = x3 + ϕ 2 (y) .θ (1) (1)
x& n-1x.&= x n=+xϕ n+−1ϕ
(y)T .(y)
θ T .θ
n-1 n n −1
x& n =. β(y).u + ϕ n (y)T .θ T (1)
x& n = β(y).u + ϕ n (y) T.θ
Estimate y = x&xn-1 1 = x n + ϕ n −1 (y) .θ
Estimate
state y = x1
state
Observer x& n = β(y).u + ϕ n (y)T .θ
x ϕ i : R→R is a vector of nonlinear functions of
P
Observer y =each
Where
1
θ ∈RϕPi :is R→R
P
Where
output, andeach a vectorisofaconstant
vector coefficients
of nonlinear functions of
which
Pϕ .
output, and θ ∈R iis a vector of constant coefficients which
scales the nonlinearity
scales
Where
In order tothe
each
design R→Rϕ i .is
ϕani : observer,
nonlinearity a vector
equation
P
(1) isofrewritten
nonlinear
as afunctions of
Fig. 2. General diagram of the no adaptive control with sum of:
Fig. 2. General diagram of thethe no adaptive control with
observer output,
In order and θ ∈Ranisobserver,
to design
P
a vectorequation
of constant
(1) iscoefficients
rewritten aswhich
a
Fig. 2. General diagram of the no adaptive control with
the observer -sumThe known
of:the
scales linear part, ϕ .
nonlinearity i
the observer - - TheThe
unknown nonlinear
known linear part,
part,and
In order to design an observer, equation (1) is rewritten as a
- The control function.
1
sum
- of:The unknown nonlinear part, and
These give us:
1 -- The
Thecontrol
knownfunction.
linear part,
x& = A. x + ϕ T (y) .θ + B. g(y).u (2)
These
- give
Theus:
unknown nonlinear part, and
2 Adaptation of
parameters T
Linear + ϕcontrol
x& =- A. xThe + B. g(y).u
function.
(y) .θControl
Nonlinear (2)
2 Adaptation of part part
-These give us:
Input parameters where :
Linear Nonlinear
Control
Output
-
A. x + ϕ Tpart
x& = part (y) .θ + B. g(y).u (2)
Control Process -
0 1 0 0 0 0
Input  where :
0 Linear
0 1 0 0 0
Estimate Output
0- 0 ...... Nonlinear 0  Control
........ 0
state Control Process A =  part
0 0 part
;
0- 0 0
0 1 01 00 0
1

Observer 0where
0 00 :
0 1 0 .......
0 
10 0 
2
Estimate
  
0 00 00 ......0 ........
0 
00 0
A =- 
[ ]
Estimate
state  ;
1
parameter ϕ(y) = ϕ10T0(y)01ϕ 20
T
0 0 ϕ i T1(y)
0(y) ......
T
0  ϕ n (y) ;
0 0.......
Observer B T = [0 000 0... 001... 10];0 .......0 10 
2  0  
x = [x1 0x02 00
.... 0 0 x n ]T 0; 0  
xi .............
...... 0 0
[ ]
Estimate
parameter Fig. 3. General diagram of the adaptive control A[ =
θ1 =θ02ϕ 1T0....
θ =ϕ (y) T
(y) θ0iϕ 2..... ] T  ; T
1 ϕ i 0(y) ....... ϕ n (y) ;
T
(y)0θ p......
 
with the observer B =[00 00 ...0 0 ...0 1];....... 1
T
 
x = [x10 x02 ....
0 xi 0 ..... x0n ]T ;0 
[ ]
3. BACKSTEPPING CONTROLLER DESIGN
= [θ=1 ϕ 1θT2FOR θ p ] ϕ i T (y)USING
Fig.
The 3. General
conditions diagram ofneeded
and assumptions the adaptive control
in order to synthesize T
Fig. 3. General diagram of the adaptive control ϕθ (y) ϕ 2θTi (y).....
(y)....SIMPLE PENDULUM
...... ϕ n T (y) ;
....... AN
an observer-based adaptive backstepping controller are [1]: OBSERVER
1.
with the observer
Full-state measurement is not available,
B = [0 0 ... 0 ... 1];
T

with the observer


2. An output function y=h(x) must be defined, and = [x1 of ax 2simple
3.1 xModel xi ..... x n ]T ;
.... pendulum
3. BACKSTEPPING TCONTROLLER DESIGN
3. All system nonlinearity must be a function of output
The conditions
only.and assumptions needed in order to synthesize It isθ about
= θ1 anθ 2unstable
.... θSIMPLE
FOR [
..... described
isystem θPENDULUM
p perfectly
USINGby aAN ]
Theanconditions
observer-based adaptive backstepping
and assumptions needed in controller are [1]: nonlinear model obtained by applying
order to synthesize the laws of physics.
OBSERVER
an observer-based adaptive backstepping controller are [1]: The simple pendulum is a traditional example [4].
1. Full-state measurement is not available,
1. 2.Full-state
An output function y=h(x)
measurement is not must be defined, and
available, 3.1 Model of a simple pendulum
3. BACKSTEPPING CONTROLLER DESIGN
2. An output function y=h(x) must befunction
3. All system nonlinearity must be a defined,ofand
output
It is about an FOR SIMPLE
unstable PENDULUM
system USING AN
described perfectly by a
3. Allonly.
system nonlinearity must be a function of output nonlinear model obtained by OBSERVER
applying the laws of physics.
only. The simple pendulum is a traditional example [4].
86
3.1 Model of a simple pendulum

It is about an unstable system described perfectly by a


nonlinear model obtained by applying the laws of physics.
The simple pendulum is a traditional example [4].
86

86

Downloaded 28 Jun 2013 to 10.0.50.139. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://proceedings.aip.org/about/rights_permiss
g 1
q&& = − . sin q + .u (7)
u l m.l 2
motor
While choosing the following state variables:

l x1=q : represent the angular position,


x2= q& : represent the angular velocity.
q
Then, the general model that we will use for this application
is as it follows:
m

mg x&1 = x2
g 1
x& 2 = − . sin x1 + .u (8)
Fig. 4. General diagram for simple pendulum l m.l 2
y = x1
The interest of design is to stabilize the pendulum in its
position of unstable balance vertical. This system is treated We define the parameters as it follows:
like the model ( q, q& ), where q is the rod angle from the
g 1
vertical position as shows in the figure 4. θl = − ; θu =
l m .l 2
The Lagrangian is given by: These transformation lead to the form:
L = EC − E P x&1 = x2
1 x& 2 = θl .ϕ (x1 ) + θu .u (9)
EC = m.l 2 .q& 2 (3)
2 y = x1
E P = m.g .l .(1 − cos q )
what implies:
with the nonlinear function ϕ ( x1 ) = sin x1
1
L = m.l 2 .q& 2 − m.g .l .(1 − cos q ) (4) ¾ Observer
2
In what follows, it is considered that only the position x1 is
The differential equations are: measurable and it is supposed that the speed is constant.
∂L
= m.l 2 .q&
∂q& The observer is defined as:
d  ∂L  x̂ = ζ(t) + λ(t).θl + υ(t).θu (10)
  = m.l 2 .q&& (5)
dt  ∂q& 
∂L Where: ζ ∈ R2 , λ∈R2 and ν∈ R2xM. The terms θl and θ u ,
= −m.g .l . sin q
∂q used in the relation (10), are the “true” unknown parameters,
thus, we can not implement. In addition, the filter banks ζ , λ
According to the expression of Lagrange, the equation of the
and ν must be individually implemented as it follows:
system will be expressed by:

m .l 2 . q&&+ m . g .l . sinq = u (6)


ζ& (t ) = A.ζ − K .ζ 1 + K . y (11)

Where u is the controller applied to the pendulum rod, m is  0 


λ& = A.λ − K .λ1 +   (12)
the mass located at the end of the rod, g is the gravitational ϕ ( y )
constant and l is the half length of the rod.
0
υ& = A.υ − K .υ1 +   , (13)
3.2 Backstepping controller design assuming that the u  
nonlinear functions can be evaluated
0 1  k 
where: A=   and K=  1 
¾ Model
0 0  k 2 
To simplify the state presentation and to apply the algorithm
of backstepping, the equation 6 can be written as Given (10), (11), (12) and (13), the observer error dynamics
are described by the equation:

87

Downloaded 28 Jun 2013 to 10.0.50.139. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://proceedings.aip.org/about/rights_permissions
ε& = x& − x̂& Choosing the virtual control to add a stability term, we cancel
all the known terms except the z2 term and we dampen out
= x& − ( A .(ζ + λ .θl + υ .θu ) + K .( y − (ζ 1 + λ1 .θl + υ1 .θu ))
the unknown observer error:
0  
( )
 0  (14)
+
ϕ (y)  .θl + u  .θu  α1 = − c1 z1 − d 1 z1 − ζ 2 + λ2 .θ̂l (22)
    
= A . ε − K . ε1 This leads to the Lyapunov function derivative:
2
 1  1
V&1 = − c1 z1 2 + z1 . z 2 − d 1 . z1 − ε 2  + ε2 2
 2. d  4. d
0 1  ε  k  1 1
where: A =   , ε =  1  et K =  1 
0 0  ε 2  k 2  1 T ~ 1 &  ~  1 & 
− ε . ε + θl  z1 . λ2 − θ̂l  + θu T  z1 .υ2 T − θ̂u 
d1  gl   g u 
which then leads to: (23)
3 T
≤ − c1 z1 2 + z1 . z 2 − ε .ε +
ε − k .ε  4. d 1
ε& =  2 1 1  = A0 .ε (15)
 − k 2ε 1  ~ 1 &  ~  1 & 
θl  z1 . λ2 − θ̂l  + θu T  z1 .υ2 T − θ̂u 
 gl   g u 
 − k 1
where A0=  1  , and K is chosen such that A0 is
 − k 2 0
Hurwitz (the solutions of equation s2+k1.s+k2=0 with ¾ Step 2
negative real parts ).
In the second step, the first Lyapunov function is augmented
¾ Step 1 with a z2 term along with an additional observer error term,
We adopt the following coordinate transformation: which gives us:
z1 = y − y r (16) 1 2 1 T
V2 = V1 + z2 + ε .P.ε (24)
2 d2
z 2 = υ2 .θ̂u − y& r − α1 (17)
Its derivative is written like this way:
Where α1 is the virtual control as yet undefined.
1 T
Knowing that the first step consists in identifying the virtual V&2 = V&1 + z 2 .z& 2 − ε .ε (25)
d2
control, we choose a P∈R2x2 such that P>0 and PT=P where
P.A0+A0T.P= -I. The first Lyapunov function is chosen as: Differentiating V2, we get:
1 1 ~2 1 ~T ~ 1 T 3 T 1 T
V1 = z12 + θl + θ u .θ u + ε . P .ε (18) V&2 ≤ − c1 z1 2 + z 2 (z1 + z& 2 ) − ε .ε − ε .ε
2 2. g l 2. g u d1 4. d 1 d2
(26)
Its derivative can be written: ~ 1 &  ~  1 & 
+ θl  z1 . λ2 − θ̂l  + θu T  z1 .υ2 T − θ̂u 
 gl   g u 
~ 1 &  ~  1 &  1 T
V&1 = z1 . z&1 + θl  − θ̂l  + θu T  − θ̂u  − ε .ε
 gl   gu  d1 For compactness of notation, we define c1∗ = c1 + d 1 and
u = α 2 /θ̂u , then, we can develop the term (z1 + z& 2 ) by using
~ 1 &  ~  1 &  1 T
= z1 .( y& − y& r ) + θl  − θ̂l  + θu T  − θ̂u  − ε .ε the following expression:
 gl   gu  d1
(19) (z1 + z& 2 ) = α 2 + z1 − k 2 .(ζ 1 + λ1θ̂l + υ1 .θ̂u )
Let us note that: ( ) ~
(
+ c1∗ . ζ 2 + λ2 θ̂l + υ2 .θ̂u + c1∗ . λ2 θl + υ2 .θu
~
)
& &
y& = x&1 = x2 = x̂2 + ε 2 = ζ 2 (t) + λ2 (t).θl + υ2 (t).θu + ε 2 (20) − c1∗ . y& r + k 2 y − &y&r + c1∗ . ε 2 + λ2 θ̂l + υ2 .θ̂u + ϕ (y).θ̂l

We will have then: (27)


By defining the control u, such that:
( ) ~ 1 &
V&1 = z1 . z 2 + α1 + ζ 2 + λ2 .θ̂l + z1 .ε 2 + θl  z1 . λ2 − θ̂l 
 gl 
(21)
{ (
α 2 = − c 2 . z 2 − d 2 .(c1∗ )2 . z 2 − z1 − k 2 . ζ 1 + λ1θ̂l + υ1 .θ̂u )
~ 
+ θu T  z1 .υ2T −
1 &  1 T
θ̂u  − ε .ε (
α 2 = + c1∗ . ζ 2 + λ2 θ̂l + υ2 .θ̂u )
}
 g u  d1 (28)

− c1 . y& r + k 2 y − &y&r + ϕ (y).θ̂l + λ2 . g l . τ l + υ2 . g u . τ u

88

Downloaded 28 Jun 2013 to 10.0.50.139. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://proceedings.aip.org/about/rights_permissions
Where τ l and τ u will be defined later, we obtain: 4. SIMULATION RESULTS

(z1 + z&2 ) = − c2 . z 2 − d 2 .(c1∗ )2 . z 2 + c1∗ .(λ2 θ~l + υ2 .θ~u + ε2 ) 4.1 Regulation: yr =1


& &
+ λ2 θ̂l + υ2 .θ̂u − (λ2 . g l . τ l + υ2 . g u . τ u ) In this observer-based experiment, the reference of real rod
angle is equal to 0.7 rad. The selected parameters of the
(29)
pendulum are: m=10 kg, l=1m and g=10 ; θl = -10 ;
and expression (26) will have the following structure:
θu =10 ; θ̂l 0 =-9,3 ; θ̂u 0 =9,5.
3 T 3 T  1 & 
V&2 ≤ − c1 z1 2 − c 2 z 2 2 − ε .ε − ε . ε − λ2 . z 2 g l  τ l − θ̂l 
4. d 1 4. d 2  gl  Table 1. Design Parameters for Observer-Based
Controller
~  1 &   1 &  ~  1 & 
+ θl  τ l − θ̂l  − υ2 . z 2 g u  τ u − θ̂u  + θu T  τ u − θ̂u 
 gl   g u   g u  Description value
3 T 3 T Lyapunov gain c3 = c4= 9
≤ − c1 z1 2 − c2 z 2 2 − ε .ε − ε .ε Observer gain k1 = 30 ; k2 = 300
4. d 1 4. d 2
Damping gain d1 = d2 =1
( ~ 
gl 
)
1 &  ~
+ − λ2 . z 2 g l + θl . τ l − θ̂l  + − υ2 . z 2 g u + θu T ( ). τ u−
1 & 
θ̂u 
g u  Spline parameter adaptation gain gl=700
 
(30) Load parameter adaptation gain gu=11
20 0.1
where:
0

( ) 10

Commande u
τ l = c1∗ z 2 + z1 .λ2

z z1
-0.1

Control input u

1
error
= (c + z ).ν
(31) 0 -0.2

erreur
∗ T
τu

Tracking
1 z2 1 2 -10
-0.3
-0.4
-20 -0.5
0 0.5 1 0 0.5 1
Restating the definition of the control, and defining the temps(s)
Time (seconds) temps(s)
Time (seconds)
update laws: 1
1
--- q

[
0
2

α2 1 __ q.est
Tracking error z

u= = − c2 . z 2 − d 2 .(c1∗ )2 . z 2
2

-1
erreur z

θ̂u θ̂u 0.5

− {z − k .(ζ + λ θ̂ + υ .θ̂ )+ c .(ζ )


-2

1 2 1 1 l 1 u 1 2 + λ2 θ̂l + υ2 .θ̂u -3

}]
0

− c1 . y& r + k 2 y − &y&r + ϕ (y).θ̂l + λ2 . g l . τ l + υ2 . g u . τ u -4
0 0.5 1 0 0.5 1
temps(s) temps(s)
(32) Time (seconds) Time (seconds)

&
(
θ̂l = g l τ l = g l . c1∗ z 2 + z1 . λ2 ) (33)
( )
4 1.2
& --- yr
θ̂u = g u τ u = g u . c1∗ z 2 + z1 .ν 2 T
--- V
__ V.est __ y
2 1

0 0.8
The final derivative of Lyapunov function is given by: -2 0.6

-4 0.4
2 2 0 0.5 1 0 0.5 1
3 T
∑ ∑
temps(s)
Time (seconds) temps(s)
V&2 ≤ − c j z 2j − ε .ε (34) Time (seconds)

j =1 i =1
4.d i -9 11

10.5
θu.estiméeu

-9.5
Estimate θ
θl.estimée
θl

Based on the Lyapunov function V = V2 , we have shown


Estimate

10
& < 0 , ∀ (z, ε ) ≠ 0 , implying asymptotic stability of the
that V -10
9.5
system and the observer. -10.5 9
0 0.5 1 0 0.5 1
temps(s) temps(s)
Time (seconds) Time (seconds)

Fig. 5. Results of the observer adaptive backstepping


control for a simple pendulum “constant input”

89

Downloaded 28 Jun 2013 to 10.0.50.139. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://proceedings.aip.org/about/rights_permissions
It is clearly noticed that the error tends towards zero and the 5. CONCLUSIONS
state x1 follows perfectly the trajectory of reference y r . This In this paper, observer-based adaptive backstepping is
means that this technique of control gives a good response by synthesized, using the simple pendulum as an example.
It is choice
a good clearly of
noticed that the error
the constants tends
c3 and c4.towards zero and the 5. CONCLUSIONS
state x1 follows perfectly the trajectory of reference y r . This The adaptive backstepping control is studied by introducing
In this paper, observer-based adaptive backstepping is
means that this technique of control gives a good response by the observer approach. The application of this technique for
synthesized, using the simple pendulum as an example.
a good choice of the constants c3 and c4. the nonlinear systems gives good results. The convergence
Theand
adaptive backstepping
the total stabilitycontrol
of theiscontrolled
studied by system
introducing
well express
4.2 Pursuit : yr = sin(2π.t) the the
observer approach. Thestrategy
application of this technique for
advantage of this of control.
the nonlinear systems gives good results. The convergence
In this case, we will maintain the same parameters and "the and The backstepping
the total stability ofcontrol with ansystem
the controlled observer
well was also studied
express
4.2 Pursuit : yr = sin(2π.t)
true" values and estimated values. the advantage
to make of thethis strategy of
classical control.
backstepping control more robust by
In this case, we will maintain the same parameters and "the Thesolving the problem
backstepping of measurement
control with states.
an observer was also studied
true" values and estimated values. to make the classical backstepping control more robust by
In this case, the technique of control presented rests on the
solving the problem of measurement states.
10
0.6 use of an observer by supposing that the states of the system
In this
are case,
not the
all technique of control
To presented
carry outrests theonobjectives
the
Tracking error z1

measurable. of
Control input u
Commande u

0 0.4
use continuation
of an observer and
by supposing that the
regulation, states ofofthean
a choice system
observer was
erreur z 1

0.2 are not all measurable. To carry out the objectives of


used. But, it is valid only for the systems which can be
-10 continuation and regulation, a choice of an observer was
0 represented
used. in triangular
But, it is valid only for theform. This
systems observer
which can be makes it
-20 -0.2 possible in
represented to carry out the
triangular continuation
form. as well
This observer as theitregulation
makes
0 1 2 3 4 0 1 2 3 4
temps(s)
Time (seconds) temps(s)
Time (seconds)
in a perfect
possible way.
to carry out the continuation as well as the regulation
0.6 --- q
in a perfect way.
1
Tracking error z 2

0.4 __ q.est

0.2 0.5
6. REFERENCES
erreur z 2

0 0
6. REFERENCES
-0.2 -0.5
-0.4
-1
0 1 2 3 4 0 1 2 3 4 [1] [1]
R. R. Milman,
Milman, Adaptive
Adaptive Backstepping
Backstepping Control Control
of the of the
temps(s)
Time (seconds) temps(s)
Time (seconds) Variable
Variable Reluctance
Reluctance Motor, Motor , Department
thesis, thesis, Department
of of
--- V --- yr Electrical
Electrical and Computer
and Computer Engineering,
Engineering, UniversityUniversity
of of
1
__ V.est 5 __ y
Toronto,
Toronto, 1997.1997.
0.5

0 0 B.
[2] [2] Yaon, Adaptive
B. Yaon, RobustRobust
Adaptive ControlControl
ofNonlinear Systems Systems
ofNonlinear
with with
Application to Control
Application to of Mechanical
Control of Systems, Systems,
Mechanical
-0.5 thesis, University of California, New York, 1996.
-5
thesis, University of California, New York, 1996.
-1
0 1 2 3 4 0 1 2 3 4 [3] M. Kristić, I. Kanellakopoulos, and P.V. Kokotović,
temps(s)
Time (seconds) temps(s)
Time (seconds) [3] M. Kristić,
Nonlinear I. Kanellakopoulos,
and Adaptive Control Design, and
John P.V.
WilleyKokotović,
-9 11
Nonlinear
and Sons, andYork,
Inc., New Adaptive Control
NY, 1995 . Design, John Willey
10.5 and Sons, Inc., New York, NY, 1995.
.estiméeθu
Estimate θl

-9.5 [4] A. Benaskeur and A. Desbiens, “Application of


θl.estimée

θuEstimate

10 backstepping to the stabilization of the Inverted


[4] A. Benaskeur
Pendulum,” andComputer
Electrical and A. Desbiens,
Engineering“Application
IEEE, of
-10
9.5 backstepping
Canadian Conference,tovol.the stabilization
1, May 24-28, 1998, of the Inverted
pp. 113-
116.Pendulum,”
, 1998. Electrical and Computer Engineering IEEE,
-10.5 9 Canadian Conference, vol. 1, May 24-28, 1998, pp. 113-
0 5 10 0 5 10
temps(s)
Time (seconds) temps(s)
Time (seconds)
116. , 1998.

Fig. 6. Results of the observer adaptive backstepping


control for a simple pendulum “variable input”
Fig. 6. Results of the observer adaptive backstepping
control for a simple pendulum “variable input”

90

90

View publication stats Downloaded 28 Jun 2013 to 10.0.50.139. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://proceedings.aip.org/about/rights_permissions

You might also like