Scribd
Upload
101 views

Local Invariant Set Theorem

This document summarizes techniques for analyzing convergence to invariant sets for nonlinear systems, including: 1. Using Barbalat's lemma to show that the time derivative of a Lyapunov function approaches zero, implying convergence. 2. Defining invariant sets as sets that solutions remain in once entered. 3. Applying the global invariant set theorem - if a Lyapunov function satisfies certain conditions, solutions converge to the largest invariant set where its time derivative is zero. 4. The local invariant set theorem guarantees convergence to invariant sets within a bounded region defined by a Lyapunov function.

Uploaded by

Ahmad Rahan
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
101 views

Local Invariant Set Theorem

This document summarizes techniques for analyzing convergence to invariant sets for nonlinear systems, including: 1. Using Barbalat's lemma to show that the time derivative of a Lyapunov function approaches zero, implying convergence. 2. Defining invariant sets as sets that solutions remain in once entered. 3. Applying the global invariant set theorem - if a Lyapunov function satisfies certain conditions, solutions converge to the largest invariant set where its time derivative is zero. 4. The local invariant set theorem guarantees convergence to invariant sets within a bounded region defined by a Lyapunov function.

Uploaded by

Ahmad Rahan
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

C21 Nonlinear Systems

Mark Cannon
4 lectures
webpage: www.eng.ox.ac.uk/conmrc/nlc

Michaelmas Term 2013

0-1

Lecture 3

Convergence and invariant sets

3-1

Convergence and invariant sets

Review of Lyapunovs direct method


Convergence analysis using Barbalats Lemma
Invariant sets
Global and local invariant set theorem
Example

3-2

Review of Lyapunovs direct method


Positive definite functions
If
V (0) = 0
V (x) > 0 for all x 6= 0
then V (x) is positive definite
If S is a set containing x = 0 and
V (0) = 0
V (x) > 0 for all x 6= 0, x S
then V (x) is locally positive definite (within S)
e.g.

3-3

V (x) = xT x

positive definite

V (x) = xT x(1 xT x)

locally positive definite


within S = {x : xT x < 1}

Review of Lyapunovs direct method


System: x = f (x),

f (0) = 0

Storage function: V (x)


V dx
Time-derivative of V : V (x) =
= V (x)T x = V (x)T f (x)
x dt
If

(i). V (x) is positive definite
for all x S
(ii). V (x) 0
then the equilibrium x = 0 is stable
If
for all x S
(iii). V (x) is negative definite
then the equilibrium x = 0 is asymptotically stable
If
(iv). S = entire state space
(v). V (x) as kxk
then the equilibrium x = 0 is globally asymptotically stable

3-4

Convergence analysis
What can be said about convergence of x(t) to 0
if V (x) 0 but V (x) is not negative definite?
Revisit m-s-d example:
c(y)

6
k(y)

0
m

k(y)

6
c(y)

Equation of motion: m
y + c(y)
+ k(y) = 0
Storage function:

V = K.E. + P.E. =
V = c(y)
y

3-5

- y

1
my 2
2

Z
+

k(y) dy
0

-y

Convergence analysis

V is p.d. and V 0 so:

(y, y)
= (0, 0) is stable
and V (y, y)
tends to a finite limit as t

but does (y, y)


converge to (0, 0)?
l equivalent to
can V (y, y)
get stuck at V = V0 6= 0 as t ?

need to consider motion at points (y, y)


for which V = 0

3-6

Example
Equation of motion: m
y + c(y)
+ k(y) = 0
k(y)

c(y)

k(y) = 5 tan1 (y/5)

c(y)
= 0.1y(0.5

+ |y|)(2

e0.1|y|
)

15

V = 100

10

V = 50

dy/dt

V = 0 when y = 0

15

3-7

V = c(y)
y 0

10

Storage function:
Z y
1 2
V = 2 y +
5 tan1 (y/5) dy

BM
x(t) B
20

10

0
y

10

20

but k(y) 6= 0 = y 6= 0 = V =
6 0

V continues to decrease until y = y = 0

Convergence analysis
Summary of method:
1. show that V (x) 0 as t

2. determine the set R of points x for which V (x) = 0

3. identify the subset M of R for which V (x) = 0 at all future times

then x(t) has to converge to M as t

This approach is the basis of the invariant set theorems

3-8

Barbalats Lemma
Barbalats lemma: For any function (t), if
t

( ) d converges to a finite limit as t

(i).
0

(ii). (t)
is finite for all t
then limt (t) = 0
Obvious for the case that (t) 0 for all t
Condition (ii) is needed to ensure that (t) remains continuous for all t

Z
Can construct discontinuous (t) for which

( ) d converges
0

but (t) 6 0 as t
3-9

Barbalats Lemma
Example: pulse train (t) =

k=0

e4

(tk)2

1.5
1
(t)

(t) :

0.5
0
0

10

10

( ) d :

(t) dt

2
1
0
0

From the plots itZ is clear that


t
(s) ds tends to a finite limit
0

but

(t) 6 0 as t

because (t)
as t

3 - 10

Barbalats Lemma contd.



Apply Barbalats Lemma to V x(t) = (t) 0:
Integrate:
Z t


(s) ds = V x(t) V x(0)

finite limit as t

Differentiate:

2V
f
T

(t) = V x(t) = f (x) 2 (x)f (x) + V (x) (x)f (x)


x
x
= finite for all t if f (x) continuous and V (x) continuously differentiable

V (x) 0 as t
The above arguments rely on kx(t)k remaining finite for all t,
which is implied by:
V (x) positive definite
V (x) 0
V (x) as kxk
3 - 11

Convergence analysis
Summary of method:
1. show that V (x) 0 as t
true whenever V 0 & V, f are smooth & kx(t)k is bounded
[by Barbalats Lemma]
2. determine the set R of points x for which V (x) = 0
algebra!
3. identify the subset M of R for which V (x) = 0 at all future times
M must be invariant
then x(t) has to converge to M as t

This approach is the basis of the invariant set theorems

3 - 12

Invariant sets
A set of points M in state space is invariant if
x(t0 ) M

x(t) M

for all t > t0

Examples:
? Equilibrium points
? Limit cycles
? Level sets of V (x)
provided V (x) 0

i.e. {x : V (x) V0 } for constant V0

If V (x) 0 as t , then
x(t) must converge to an invariant set M contained
within the set of points on which V (x) = 0
as t
3 - 13

Global invariant set theorem


If there exists a continuously differentiable function V (x) such that
V (x) is positive definite
V (x) 0
V (x) as kxk
then: (i). V (x) 0 as t
(ii). x(t) M = the largest invariant set contained in R
where R = {x : V (x) = 0}
M



0
x(t)
R

V (x) negative definite = M = 0

(c.f. Lyapunovs direct method)

Determine M by considering system dynamics within R


3 - 14

Global invariant set theorem


Revisit m-s-d example (for the last time)
V (x) is positive definite, V (x) as kxk , and
V (y, y)
= c(y)
y 0
therefore V 0, implying y 0 as t
i.e. R = {(y, y)
: y = 0}
but y = 0 implies y = k(y)/m
therefore y 6= 0 unless y = 0, so y(t)

= 0 for all t only if y(t) = 0


i.e. M = {(y, y)
: (y, y)
= (0, 0)}

(y, y)
= (0, 0) is a globally asymptotically stable equilibrium!

3 - 15

Local invariant set theorem


If there exists a continuously differentiable function V (x) such that
the level set = {x : V (x) V0 } is bounded for some V0
and V (x) 0 whenever x
then:
(i). is an invariant set
(ii). x(0)
=
V (x) 0 as t
(iii). x(t) M = largest invariant set contained in R
where R = {x : V (x) = 0}




0
x(t)
R

3 - 16

Local invariant set theorem

V (x) doesnt have to be positive definite or radially unbounded

Result is based on Barbalats Lemma applied to V

applies here because finite implies kx(t)k finite for all t


since x(0) and V 0

is a region of attraction for M

3 - 17

Example: local invariant set theorem


Second order system:

x 1 = x2
x 2 = (x1 1)2 x32 x1 + sin(x1 /2)

Equilibrium points: (x1 , x2 ) = (0, 0), (1, 0), (1, 0)


Trial storage function:
Z x1

1 2
V (x) = 2 x2 +
y sin(y/2) dy
0

V (x1 , 0)

V is not positive definite


but V (x) if x1 or x2

level sets of V are finite


1

x1

x1 sin(x1 /2)
3 - 18

Example: local invariant set theorem contd.


Differentiate: V (x) = (x1 1)2 x42 0
V (x) = 0

x R = {x : x1 = 1 or x2 = 0}

From the system model, x R implies:


x1 = 1
=
(x 1 , x 2 ) = (x2 , 0)
and
x2 = 0
=
(x 1 , x 2 ) = (0, sin(x1 /2) x1 )

x(t) remains on line x1 = 1 only if x2 = 0
therefore
x(t) remains on line x2 = 0 only if x1 = 0, 1 or 1
= M = {(0, 0), (1, 0), (1, 0)}
Apply local invariant set theorem to any level set = {x : V (x) V0 }:

is finite
=
x(t) M = {(0, 0), (1, 0), (1, 0)} as t
V 0

3 - 19

Example: local invariant set theorem contd.


From any initial condition, x(t) converges asymptotically to (0, 0), (1, 0)
or (1, 0)
but x = (0, 0) is unstable
p
(linearized system at (0, 0) has poles 2 1 so is unstable)
Contours of V (x):

2
1.5
1

V (x) = 0

x2

0.5
0
0.5
1
1.5
2
2

1.5

0.5

0
x

0.5

1.5

Use local invariant set theorem on level sets = {x : V (x) V0 } for V0 < 0

x = (1, 0), x = (1, 0) are stable equilibrium points


3 - 20

Summary

Convergence analysis using Barbalats lemma


Invariant sets
Invariant set methods for convergence: local invariant set theorem
global invariant set theorem

3 - 21

You might also like