Scribd
Upload
36 views

Back Stepping and Non-CE Method

The document summarizes adaptive control techniques for double integrator systems and scalar first order systems with unknown parameters. For double integrators, backstepping control is used assuming known and unknown parameters. For unknown parameters, a non-certainty equivalence approach is taken using parameter filters. For scalar first order systems, the error dynamics are derived assuming unknown constants for the system. A non-certainty equivalence control law is designed using parameter filters, and it is shown that the parameter estimation error and tracking error are asymptotically stable through a Lyapunov analysis.

Uploaded by

Sagar Joshi
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
36 views

Back Stepping and Non-CE Method

The document summarizes adaptive control techniques for double integrator systems and scalar first order systems with unknown parameters. For double integrators, backstepping control is used assuming known and unknown parameters. For unknown parameters, a non-certainty equivalence approach is taken using parameter filters. For scalar first order systems, the error dynamics are derived assuming unknown constants for the system. A non-certainty equivalence control law is designed using parameter filters, and it is shown that the parameter estimation error and tracking error are asymptotically stable through a Lyapunov analysis.

Uploaded by

Sagar Joshi
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

SC 617 Adaptive Control Theory

Lecture 12 - 03/03/2015

Back Stepping and Non-CE method


Lecturer: Srikant Sukumar

Scribe: Srinath Dama

Double integral systems (Continued)

Consider the Double integrator system dynamics given as follows:


x 1 = x2 + f (x1 )
x 2 = u

(1)

1. Assuming is known
In order to design a controller for the above system we can use the classical backstepping design
approach. The control objective given as follows:
lim x1 0

lim x2 x2d

Where,
x2d = K1 x1 f (x1 )

(2)

1 2
x1
2
1
2
V2 = (x2 x2d )
2
V 2 = (x2 x2d )(u x 2d )

(3)

Consider V1 () and V2 () as follows:


V1 =

Consider the update law, u = x 2d K2 (x2 x2d ). Hence, the overall Lyapunov function is given as
follows:
V = V1 + V2
V = x1 x 1 K2 (x2 x2d )2
= x1 x2 + x1 f (x1 ) K2 (x2 x2d )

= x1 x2 + x1 (x2d K1 x1 ) K2 (x2 x2d )


1
1
2
(K1 )x1 2 (K2 )(x2 x2d )
2
2
V 0
2. Assuming is unknown
First control

(x1 )
x2d = K1 x1 f
if x2 = x2d
(x1 )
x 1 = K1 x1 + f
1

(4)

1 2
1 2
x1 +

2
2
1
V1 = x1 x 1

Let us take

V1 =

(x1 )) 1
= x1 (K1 x1 + f

1 f (x1 ) )
= K1 x1 2 + (x

= x1 f (x1 )
V1 K1 x1 2

First Update

Let us take

1
(x2 x2d )2
2
V 2 = (x2 x2d )(x 2 x 2d )

V2 =

u = x 2d K2 (x2 x2d )
x 2d = K1 x 1 x1 f 2 (x1 ) f(x1 )
= K1 (x2 + f (x1 )) x1 f 2 (x1 ) f(x1 )
f
)(x2 + f (x1 )) x1 f 2 (x1 )
= (K1 +
x1
f
)(x2 + f (x1 )) x1 f 2 (x1 )
x
2d = (K1 +
x1
f
= x 2d + f (x1 )(K1 +
)
x1
W here =

Let the overall Lyapunov function V = V1 + V2 , where, V1 () and V2 () are given as follows:
1 2
1 2
x1 +

2
2
1
1 2
V2 = (x2 x2d )2 +

2
2

V1 =

Therefore, V evaluated as follows:

second control

1
1
2d )
V = x1 x2 + x1 f (x1 ) x
1 f (x1 ) + (x2 x2d )(u x

u = x 2d K2 (x2 x2d )
f
= x 2d + f (x)(K1 +
)
x1
1 f (x1 ) K2 (x2 x2d )2
= x1 (x2 x2d ) + x1 x2d + x1 f (x1 ) x
f
1
+ (x2 x2d )f (x1 )(K1 +
)
x1 )

= K1 x1 2 K2 (x2 x2d )2 + x1 (x2 x2d )


0

Second update

f
= (x2 x2d )f (x1 )(K1 +
)
x1
2

Hence,by signal chasing we can prove x1 0, x2 0 and x2 x2d , x 2 x 2d

Creating an attractive set for parameter estimation error

Non-certainty Equivalence

(5)

x = ax + bu

(6)

Consider the scalar first order system

where, a, b are the unknown constants.


Control objective:
lim x(t) r(t) = 0

The error dynamics given as follows:

Let, 1 =

a
b

e = ax + bu r(t)

a
1

= b(u + x r(t))
b
b
and 2 = 1b . Reconsider the control law, u = 1 x 2 (Ke r)

1
a

e = Ke + b(u + x + (Ke r))


b
b
e = Ke + b([x, (Ke r)]
+ u)

Where, = [1 , 2 ]T and W = [x, (Ke r)].


The filters are given as follows:
e f = ef + e
f = Wf + W
W
u f = uf + u
e f = Kef + b(Wf + uf )
If non CE uf = Wf ( + )
Where - Dynamics and - static
e f = Kef + b(Wf Wf ( + ))
= Kef bWf ( + )
= Kef bWf Z

where Z = + , Z = + ,
= Wf T ef Sgn(b)

f T ef Sgn(b) + Wf T (Kef bWf Z)Sgn(b)


Z = + W
T

let = W f ef Sgn(b) + KWf T ef Sgn(b)


Z = | b | Wf T Wf ZSgn(b)

Consider the Lyapunov function V = Z2 +

e2f
2

V = Z Z + ef e f
= Z( | b | Wf T Wf Z) + ef (Kef bWf Z)
= | b | k Wf Z k2 Ke2f bef Wf Z
( | b |

b2
1
)kWf Zk2 (K )e2f
2
2

Likewise, signal chasing we can proove the following,


ef 0
Wf Z 0
e0

You might also like