95
Emergent complexity Chaos and fractals

Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Embed Size (px)

Citation preview

Page 1: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Emergent complexity

Chaos and fractals

Page 2: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Uncertain Dynamical Systems

c-plane

Page 3: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z

0c

planec

bounded open and connectedkc Mset z

bounded is "Cantor dust"kc Mset z

Page 4: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z c i

boundedkc Mset z

Page 5: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

planez

1 1k k kz cz z

Julia sets

Page 6: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z 0c

bounded open and connectedkc Mset z

Undecidable No algorithm to

determine has bounded run timec Mset

Page 7: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Overcoming computational

complexity

Page 8: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1 ,k k kz cz z z c c

boundedkc Mset z

What does this have to do with complex systems?

• This classic computational problem illustrates an important idea, but in an easily visualized way.

• Most computational problems involve uncertain dynamical systems, from protein folding to complex network analysis. Not easily visualized.

• Natural questions are typically computationally intractable, and conventional methods provide little encouragement that this can be systematically overcome.

Page 9: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Main idea

1

"Fragile" means

membership changes when

the map is perturbed:

1k k kz c z z

e.g. the boundary moves.

Page 10: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Main idea

Points near the boundary are “fragile.”

Merely stating the obvious in this case.

But illustrates general principle that can be exploited by the right

algorithms.

Page 11: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

5

10

15

20

25

30

# iterations

Points not in M.

Page 12: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

5

10

15

20

25

30

# iterations

Color indicates number of

iterations of simulation to show point is not in M.

Page 13: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

But simulation cannot show that points are in M.

1 1 k k kz c z z

boundedkc Mset z

Page 14: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

planec

1 1k k kz cz z

But simulation is fundamentally limited• Gridding is not scalable• Finite simulation inconclusive

0

1

2z

Page 15: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

It’s easy to prove that this disk is in M.

Other points in M are fragile to the definition of the map.

1 1 k k kz c z z

Merely stating the obvious.

Main idea

Page 16: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z

planec

2

1

221 0

1 1

k k

k k k

k

V z z

V z V z

z cz z

c z

2 decreases

1 1

V z z

c z

Sufficient condition

Page 17: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z

2 decreases

1 1

V z z

c z

21

2 11 1 0 1z

2c

1c Mset

Page 18: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Trivial to prove that these points are in Mandelbrot set.

1c 2 1c

Page 19: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Main idea

The longer the proof, the more fragile the remaining regions.

The proof of this region is a bit longer.

Page 20: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Main idea

And so on…

Proof even longer.

Page 21: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane
Page 22: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Easy to prove these points are in Mset.

Page 23: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Easy to prove these points are not in Mset.

Page 24: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Proofs get harder.(But all still “easy.”)

What’s left gets more fragile.

Page 25: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Complexity»Chaos

Fractals

Emergent complexity.

Page 26: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Complexityimplies fragility

What matters to organized

complexity.

Page 27: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Emergent complexity

Page 28: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

cHow might this help

with organized complexity and

“robust yet fragile”?

• Long proofs indicate a fragility.• Either a true fragility (a useful answer) or an

artifact of the model (which must then be rectified)• Potentially fundamentally changes computational

complexity for organized complexity• Brings back together two research areas that have

been separated for decades:• Numerical analysis and ill-conditioning• Computational complexity (P, NP/coNP,

undecidable)

Page 29: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Proof?

New proof methods that is scalable and systematic (can be

automated).

Page 30: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Breaking hard problems

• SOSTOOLS proof theory and software• Nested family of (dual) proof algorithms• Each family is polynomial time• Recovers most “gold standard” algorithms

as special cases, and immediately improves• No a priori polynomial bound on depth

(otherwise P=NP=coNP)• Conjecture: Complexity implies fragility

Page 31: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Safety Verification and Reachability Analysis

• Safety critical applications.• Exhaustive simulation is not exact.• Set propagation is computationally expensive.

Find a barrier certificate B(x)

( ) 0

( ) 0

( ) 0

B x x UnsafeSet

B x x InitialSet

Bf x x StateSpace

x

Initial set

Unsafe set

B(x) = 0

Scalable computation using SOS machinery.

Page 32: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

• Parametric• Memoryless• Dynamic (IQC)

Hybrid, Uncertain, Stochastic

Hybrid systems can be handled easily,even for systems with uncertainty:

• Use supermartingales as certificates.• Get guaranteed bound on reach probability.

Also stochastic hybrid systems:

(Prajna, Jadbabaie – HSCC ’04)

(Prajna, Jadbabaie, Pappas – CDC ’04)

Page 33: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Feedback control

Variable supply/demand

Physical network

Components

Functionalrequirements

Hardware constraints

“Horizontal” Decompositions

“Ver

tica

l”

laye

ring

Unifying role of dual proofs and decomp-ositions

Page 34: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Main idea

Think of this as a robustness problem.

Page 35: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1

0k k kz cz z

0c

planec

How robust is stability to perturbations in c?

Globally stable.

Page 36: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z

0c

planec

Region of convergence.

How robust is stability to perturbations in c?

Page 37: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z planec planez

Page 38: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

planec 1 1k k kz cz z planez

Simulation is fundamentally limited

Page 39: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

planec 1 1k k kz cz z

planez

Simulation is fundamentally limited

0

1

2z

Page 40: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

5

10

15

20

25

30

# iterations

Page 41: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

10

20

30

40

50

60

iterations

Page 42: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

10

20

30

40

50

60

iterations

-2

Page 43: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

10

20

30

40

50

60

iterations

-2

Page 44: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

10

20

30

40

50

60

iterations

Page 45: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

10

20

30

40

50

60

iterations

Page 46: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

70

120

iterations

Page 47: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

120

iterations

60

1 1k k kz cz z

Page 48: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kx cx x

realc

realx

Page 49: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kx cx x

realc

kx kx

0c 0c

1kx 1kx

Page 50: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

realc

Page 51: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

realc

1 1k k kx cx x

realx

Page 52: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3-2

-1

0

1

2

3

4

1 1k k kx cx x

realc

1

10,1

x cx x

xc

realx

Fixed points

StableUnstable

Stable

Page 53: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3-2

-1

0

1

2

3

4

1x cx x realc

realx

Unstable

Stable

Stable

Fixed points

Page 54: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3-2

-1

0

1

2

3

4

1x cx x realc

realx

Unstable

Stable

Equilibria

Page 55: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3-2

-1

0

1

2

3

4

1 1k k kx cx x

realc

realx

2

1

221 0

1 1

k k

k k k

k

V x x

V x V x

x cx x

c x

2 decreases

1 1

V z z

c z

Page 56: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z

planec

2

1

221 0

1 1

k k

k k k

k

V z z

V z V z

z cz z

c z

2 decreases

1 1

V z z

c z

Sufficient condition

Page 57: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1k k kz cz z

2 decreases

1 1

V z z

c z

21

2 11 1 0 1z

2c

1c Mset

Page 58: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Trivial to prove that these points are in Mandelbrot set.

1c 2 1c

Page 59: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3-2

-1

0

1

2

3

4

1x cx x realc

realx

Unstable

Equilibria

Page 60: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

realc

Page 61: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Bifurcations for 1 1k k kx cx x

c

“last 200 x”

Page 62: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

“last 200 x”

c

Zoom-in

Page 63: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

“last 200 x”

c

Zoom-in

Page 64: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3

-2

-1

0

1 1k k kx cx x 0 realc

Bounded for -2< 0 realc

stable

2 2 2

2 2 221 1 1 22 1 1 2 1 2

2 2x x c x c

c c

Bifurcations to chaos

Page 65: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-2

-1

0

-3 -2 -1 0 1 2 3-2

-1

0

1

2

1

221 0

1 1

k k

k k k

k

V x x

V x V x

x cx x

c x

2 decreases

1 1

V z z

c z

Page 66: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-2

-1

0

-3 -2 -1 0 1 2 3-2

-1

0

1

2

1

221 0

1 1

k k

k k k

k

V x x

V x V x

x cx x

c x

1 1

1 1 1 ?

k

k k

c x

c cx x

Page 67: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3-2

-1

0

1

2

3

4

Page 68: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3-2

-1

0

1

2

3

4

1 1k k kx cx x

realc

realx

2

1

221 0

1 1

k k

k k k

k

V x x

V x V x

x cx x

c x

2 decreases

1 1

V z z

c z

Page 69: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

2 2 22 2 221 1 1 2

2 1 1 2 1 22 2

2 0 2 0

x x c x cc c

c c c

-3 -2 -1 0 1 2 3

-2

-1

0

1 1k k kx cx x

Bounded for - 2 0 real?c

stable

Bifurcations to chaos

Page 70: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

2 22 2 1 2

2 0

c x c

c c

1 1k k kx cx x Invariant set?

2 22 2 1 1 2c cx x c

Invariance

2 22

2 222 1 22 1 1 2

2 0

c x cc cx x c

c c

2 22

2 222 1 2, 2 1 1 2 ?

2 0

c x cc cx x c

c c

Page 71: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3

-2

-1

0

1 1k k kx cx x

Bifurcations to chaos

Page 72: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Special case of SOS

Page 73: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Special case of SOS

Page 74: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Special case of SOS

Special case of SOS

Contradiction!

Page 75: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane
Page 76: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-3 -2 -1 0 1 2 3

-2

-1

0

1 1k k kx cx x

2 22

2 222 1 2, 2 1 1 2 ?

2 0

c x cc cx x c

c c

What is the shortest proof possible?

Can prove the whole yellow region using SOSTOOLS!

Page 77: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1c 2 1c

Lyapunov argument

Page 78: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

How to prove membership?2-period lobes

Page 79: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

12

3

3

4

4

4Proof

lengths

Page 80: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Prove membership of 2-period lobe:• Using a stability argument of the 2-period map.• Using an invariance argument.

Page 81: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-1

Formulate the invarianceProblem as the emptinessof a semialgebraic set.

Then use SOSTOOLS to construct the certificate.

Page 82: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

-6 -4 -2 0 2 4 6-6

-4

-2

0

2

4

6

x1

x 2

1 1

1 1 2

2 32 1 20.1 2 0.1

x x x

x x x x x

Discrete → Continuous

Page 83: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Let 0 be an equilibrium of

( ), .

Let be a region containing 0 and let

: be a continuously differentiable

function such that

( ) 0 in

nx f x x

D

V D

V x D

V

( ) ( ) 0 in

Then 0 is asymptotically stable.

Vx f x D

x

Lyapunov’s theorem

Page 84: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

( ) 0 in

( ) ( ) 0 in

V x D

VV x f x D

x

-6 -4 -2 0 2 4 6-6

-4

-2

0

2

4

6

x1

x 2

Page 85: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Can we test these conditions

algorithmically?

Use the Sum of Squares decomposition!

( ) 0 in

( ) ( ) 0 in

V x D

VV x f x D

x

Page 86: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

| ( ) 0, 1, ,niD x g x i n

( ), SOS, SOS, ( ) 0, ( ) 0i iV x p q x x

1

( ) ( )N

i ii

V x x p g

1

( ) ( )N

i ii

V x x q g

is SOS

is SOS

Find

such that

SOSTOOLS

( ) 0 in

( ) ( ) 0 in

V x D

VV x f x D

x

Then equilibrium is asymptotically stable.

Page 87: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

( , ), ,(0, ) 0

( , ) 0 for ,

( , ) 0 for ,

x f x p x D p Pf p

V x p x D p P

V x p x D p P

Robust Stability?

Describe both and as semilalgebraic sets.D P

Use SOSTOOLS to construct V(x,p).

www.cds.caltech.edu/sostools

Page 88: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Chemical oscillator? ,

2 3constant

X A A BX Y X

Y B

2

2

x a x x y

y b x y

Nondimensional state equations

Page 89: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

2

2

x a x x y

y b x y

3

Limit cycle for

b a b a

0 0.2 0.4 0.6 0.8 1 1.2 1.40

0.2

0.4

0.6

0.8

1

a

b

Can be computed analytically, which is not scalable.

Page 90: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

0 0.2 0.4 0.6 0.8 1 1.2 1.40

0.2

0.4

0.6

0.8

1

a

b

0 0.1 0.2 0.3 0.4 0.5 0.6

1

1.5

2

2.5

3

a = 0.1, b = 0.13

2

2

x a x x y

y b x y

Numerical simulation.

Page 91: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

1 1.5 2 2.50

0.2

0.4

0.6

0.8

1

x

y

2.2 2.6 3 3.40

0.2

0.4

0.6

0.8

1

a = 1, b = 2

0 0.2 0.4 0.6 0.8 1 1.2 1.40

0.2

0.4

0.6

0.8

1

a

b

a = 0.6, b = 1.1

(1.1, 0.6) (2, 1)

2

2

x a x x y

y b x y

Page 92: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

0 0.2 0.4 0.6 0.8 1 1.2 1.40

0.2

0.4

0.6

0.8

1

a

b

a a

b b

,b a

Page 93: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

2

2

x a x x y

y b x y

2

2

0

0

a x x y

b x y

equilibrium

2

2

x a x x x x y y

y b x x y y

2 2

2 2

x x

y y

a a

b b

Search for ( , ) satisfying

the Lyapunov conditions using

SOSTOOLS.

V x y

Page 94: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

0 0.2 0.4 0.6 0.8 1 1.2 1.40

0.2

0.4

0.6

0.8

1

a

b

,b a

4 order ( , ) th V x y

Page 95: Emergent complexity Chaos and fractals. Uncertain Dynamical Systems c-plane

Modeling Analysis

Set of possible system behaviors

Set of bad system behaviors

Proof of robustness

More on Saturday