Examples of Dynamical Systems

Example Dynamical Systems
Reading for this lecture:
NDAC, Sec. 5.0-5.2, 6.0-6.4, 7.0-7.3, & 9.0-9.4
Lecture 3: Natural Computation & Self-Organization, Physics 256A (Winter 2014); Jim Crutchfield
Tuesday, January 13, 15 1
Example Dynamical Systems ...
1D Flows: Fixed Points

model of static equilibrium
1D Flow: x R
x = F (x)
Fixed Points:
x R such that
x |x = 0
or
F (x ) = 0
1D Flows: Fixed Points ...
Stability: What is linearized system at x ?

Investigate evolution of perturbations: x0 = x + x
dF
Local Flow: x = x
dx x(t)
Local Linear System: x = x
Solution: x(t) / e t
x(0)
1D Flows ...
Stability Classification of Fixed Points:
Slope of F (x) at x :
x
1. Stable: <0
Attractor Attractor
2. Unstable: >0 x
Repellor
3. Neutral: =0
2D Flows: Fixed Points

model of static equilibrium
2D Flow: !
x R2
x = f (x, y)
x = F (x) or
x = (x, y) y = g(x, y)
F = (f, g)
Fixed Points:
(x , y ) such that
~x |(x ,y ) = (0, 0)
or
0 = f (x , y )
0 = g(x , y )
Stability: What is linearized system at x ?

Investigate evolution of perturbations x : ~x0 = ~x + ~x
x = F (x)
!
"
F ! !

Local Flow: "x = "x
"x !
!
x(t)
!
Initial conditions: x(0) x(0)
Local Linear System: x = A x

! "
f f
F"
Jacobian: A = = x
g
y
g
"x x y
Solution:
At
~x(t) / e ~x(0)
2D Flows: Fixed Points (an aside) ...
Solve linear ODEs: Find ~x(t) given

~x(0)
~x = A~x
Eigenvalues and eigenvectors: j and ~vj :
A~vj = j~
vj , j = 1, 2
Solution:
2
X
~x(t) = j e jt
~vj where calculate j :
j=1
~x(0) = 1~v1 + 2~v2
2D Flows ...
Eigenvalues of Jacobian A at x : 1 & 2 C

(Review: NDAC, Chapter 5)
Stable fixed point (aka sink, attractor):
!(1 ), !(2 ) < 0
2D Flows ...
Stability Classification of Fixed Points ...
Eigenvalues of Jacobian A at x : 1 & 2 C
Unstable fixed point (aka source, repellor):
!(1 ), !(2 ) > 0
2D Flows ...
Eigenvalues of Jacobian at x : 1 & 2 C
Saddle fixed point (mixed stability):
!(1 ) > 0 & !(2 ) < 0
2D Flows ...
Eigenvalues of Jacobian at x : 1 & 2 C
Center:
!(1 ) = !(2 ) = 0
2D Flows ...
Magnitude of (in)stability: Det(A) = 1 2
Det(A) < 0 : 1 , 2 R, 1 > 0 2 < 0 Saddles
Det(A) > 0 :
Stable: Tr(A) < 0 Tr(A) = 1 + 2
Unstable: Tr(A) > 0
Marginal: Tr(A) = 0
2D Flows ...
Tr(A)
Tr2 (A) 4Det(A) = 0
Unstable
Saddles
Unstable Spirals
Det(A)
Stable Spirals
Saddles
Stable
2D Flows ...
Tr(A)
Tr2 (A) 4Det(A) = 0
Unstable
Saddles
Unstable Spirals
Det(A)
Stable Spirals
Saddles
Stable
2D Flows ...
Tr(A)
Tr2 (A) 4Det(A) = 0
Unstable
Saddles
Unstable Spirals
Det(A)
Stable Spirals
Saddles
Stable
2D Flows ...
Tr(A)
Tr2 (A) 4Det(A) = 0
Unstable
Saddles
Unstable Spirals
Det(A)
Stable Spirals
Saddles
Stable
2D Flows ...
Tr(A)
Tr2 (A) 4Det(A) = 0
Unstable
Saddles
Unstable Spirals
Det(A)
Stable Spirals
Saddles
Stable
2D Flows ...
Tr(A)
Tr2 (A) 4Det(A) = 0
Unstable
Saddles
Unstable Spirals
Det(A)
Stable Spirals
Saddles
Stable
2D Flows ...
Tr(A)
Tr2 (A) 4Det(A) = 0
Unstable
Saddles
Unstable Spirals
Det(A)
Stable Spirals
Saddles
Stable
2D Flows ...
s u
Hyperbolic intersection of W and W :
Robust, if !(i ) "= 0, i

s
W Ws
Fixed point
persists under
u u
perturbation W W
2D Flows ...
s u
Non-hyperbolic intersection of W and W :
Fragile
s u
W W
Fixed point
changes structure
under perturbation
2D Flows: Limit Cycles
isolated, closed trajectory:
a periodic orbit: ~
x(t) = ~x(t + p), for all t
( p is the period)
model of stable oscillation
this is a new behavior type
not possible in 1D flows
Stable limit cycle
2D Flows: Limit Cycles ...
Unstable cycle
2D Flows: Limit Cycles ...
Saddle cycle
2D Flows ...
Limit Cycle Examples
Easy in polar coordinates:
r = r(1 r2 )
= 1
2D Flows ...
Limit Cycle Examples ... y
(a, ) = (0.1, 10.0)
2
Van der Pol Equations:
2
x + (x a)x + x = 0
x
or -2 2
x = y
y = x + y(a x2 )
Nonlinear damping changes sign:

Small oscillation: growth -2
Large oscillation: damped
2D Flows ...
Limit cycle existence
(requires real work to show!)
Systems that cant have stable oscillations:

1. Simple harmonic oscillator
2. Gradient systems: x = V (x)
3. Lyapunov systems
2D Flows ...
Limit cycle existence
(requires real work to show!)
How to find limit cycles?
Poincar-Bendixson Theorem:
(a) trajectory confined to trapping region
(b) no fixed points
R
then have limit cycle C C
somewhere inside R .
3D Flows:
Fixed points
Limit cycles
and ... ?
3D Flows: Quasiperiodicity
product of two limit cycles:
two irrational frequencies 1 = 2

1 2
3D Flows: Quasiperiodicity ...

a new kind of behavior not possible in 1D or 2D
Torus attractor
1 1 = 2
3D Flows: Chaos
recurrent instability
one way to do this:

Orbit reinjection near unstable fixed point
not possible in lower D flows
a new behavior type
3D Flows: Chaos ...
A topological construction:
saddle fixed point at origin: 0
1D unstable manifold: dim(W u (0)) = 1
2D stable manifold: dim(W (0)) = 2
s
two fixed points: C + & C
C+ C
0
Orbits Cannot Cross: Need 3D!
Does any ODE implement this flow design?

3D Flows: Chaos ...

Does any ODE implement this design?
Yes, the Lorenz equations:
x = (y x)
y = rx y xz
z = xy bz
Parameters: , r, b > 0
Exercise: Show fixed point at the origin can be a saddle, with 2 stable and 1 unstable directions
Exercise: Show there is a symmetry (x, y) (x, y)
3D Flows: Chaos ...

Lorenz ODE properties:
Trajectories stay in a bounded region near origin
No stable fixed points or stable limit cycles inside
Volume shrinks to zero (everywhere inside):
Z
V = dV r F~ (~x)
region
r F~ (~x) = Tr(A) = 1 b
V = ( + 1 + b)V
Region volume shrinks
V (t) = e (+1+b)t
exponentially fast!
What does the invariant set look like?

3D Flows: Chaos ...
Lorenz simulation demo:
fixed point:
limit cycle:
chaotic attractor:
(, r, b) = (10, 8/3, 28)
3D Flows: Chaos ...
Lorenz attractor structure
Branched manifold
From Continuous-Time Flows to Discrete-Time Maps:
50
Series of z-maxima: zb1 , zb2 , zb3 , . . .

z
What happens if you plot
zbn+1 versus zbn ?
0
-45 x+y 45
Max-z Return Map: zn+1 = f (zn )

Z Max Return Map
50
Zn+1
30
30 Zn 50
Time of Return Function: T (zn )
Return Time Map: Tn+1 = h(Tn )

Return Time Map
1.5
Time of Return
1.5
Tn+1
T(Z)
0.5
30 Z 50
0.5
0.5 Tn 1.5
3D Flows ...
Lorenz reduces to a cusp 1D map:
normalize to zn [0, 1]
b
zn+1 = a(1 |1 2zn | )
1
Parameters:
height: a > 0
peak sharpness: 0 < b < 1 zn+1
0
0 zn 1
3D Flows ...
Rssler equations
x = y z
y = x + ay
z = b + z(x c)
Parameters: a, b, c > 0
3D Flows ...
Rssler chaotic attractor
Parameters: (a, b, c) = (0.2, 0.2, 5.7)
3D Flows ...
Rssler branched manifold
3D Flows ...
Rssler maximum-x return map: xn+1 = f (xn )
14
xn+1
0
0 xn 14
3D Flows ...
When normalized to xn [0, 1]
get the Logistic Map:
1
xn+1 = rxn (1 xn )
xn+1
Parameter (height): r [0, 4]

0
0 xn 1
Classification of Possible Behaviors
Dimension Attractor
1 Fixed point
2 Fixed point, Limit cycle
3 Fixed Point, Limit Cycle,

Torus, Chaotic
4 Above + Hyperchaos
5 Above + ?
Lorenz: x = (y x) , r, b > 0
y = rx y xz
z = xy bz
Rssler: x = y z
y = x + ay
z = b + z(x c)
Cusp Map: zn [0, 1] a > 0, 0 < b < 1
b
zn+1 = a(1 |1 2zn | )
Logistic map:
xn+1 = rxn (1 xn ) xn [0, 1] r [0, 4]
Play with these!

The Big Picture
Global view of the state space structures:

The attractor-basin portrait
The Learning Channel
You Are Here

B 1 0 1 0
...001011101000... 1 1
1 0
0
0 1
1
1 A C 0
0
System Instrument Process Modeller
Reading for next lecture:
NDAC, Chapter 3.

Examples of Dynamical Systems

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Examples of Dynamical Systems

Загружено:

Авторское право:

Доступные форматы

Example Dynamical Systems

Reading for this lecture:

NDAC, Sec. 5.0-5.2, 6.0-6.4, 7.0-7.3, & 9.0-9.4

1D Flows: Fixed Points

1D Flows: Fixed Points ...

Stability: What is linearized system at x ?

Local Linear System: x = x

2D Flows: Fixed Points

2D Flows: Fixed Points ...

Stability: What is linearized system at x ?

Initial conditions: x(0) x(0)

2D Flows: Fixed Points ...

Local Linear System: x = A x

2D Flows: Fixed Points (an aside) ...

Solve linear ODEs: Find ~x(t) given

Eigenvalues of Jacobian A at x : 1 & 2 C

Stable fixed point (aka sink, attractor):

!(1 ), !(2 ) < 0

Eigenvalues of Jacobian A at x : 1 & 2 C

Unstable fixed point (aka source, repellor):

!(1 ), !(2 ) > 0

Eigenvalues of Jacobian at x : 1 & 2 C

Saddle fixed point (mixed stability):

!(1 ) > 0 & !(2 ) < 0

Eigenvalues of Jacobian at x : 1 & 2 C

Det(A) < 0 : 1 , 2 R, 1 > 0 2 < 0 Saddles

Unstable: Tr(A) > 0

Robust, if !(i ) "= 0, i

Stable limit cycle

Easy in polar coordinates:

Van der Pol Equations:

Nonlinear damping changes sign:

Systems that cant have stable oscillations:

3D Flows: Quasiperiodicity ...

one way to do this:

not possible in lower D flows

a new behavior type

two fixed points: C + & C

Does any ODE implement this flow design?

3D Flows: Chaos ...

3D Flows: Chaos ...

What does the invariant set look like?

From Continuous-Time Flows to Discrete-Time Maps:

Series of z-maxima: zb1 , zb2 , zb3 , . . .

From Continuous-Time Flows to Discrete-Time Maps:

Max-z Return Map: zn+1 = f (zn )

From Continuous-Time Flows to Discrete-Time Maps:

Time of Return Function: T (zn )

Return Time Map: Tn+1 = h(Tn )

Parameters: (a, b, c) = (0.2, 0.2, 5.7)

Parameter (height): r [0, 4]

Classification of Possible Behaviors

2 Fixed point, Limit cycle

3 Fixed Point, Limit Cycle,

Play with these!

Global view of the state space structures:

You Are Here

System Instrument Process Modeller

Reading for next lecture:

Вам также может понравиться