Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
90 views

Nonlinear Lect3

EG4321/EG7040 covers nonlinear control and system analysis. Dr. Matt Turner introduces nonlinear stability concepts including phase portraits and linearization, which only provide local information. A.M. Lyapunov developed sufficient conditions for nonlinear stability using energy functions. Lyapunov showed that if the derivative of a positive, scalar energy function V(x) is negative, the system is stable. This allows proving stability without solving differential equations.

Uploaded by

win alfalah
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
90 views

Nonlinear Lect3

EG4321/EG7040 covers nonlinear control and system analysis. Dr. Matt Turner introduces nonlinear stability concepts including phase portraits and linearization, which only provide local information. A.M. Lyapunov developed sufficient conditions for nonlinear stability using energy functions. Lyapunov showed that if the derivative of a positive, scalar energy function V(x) is negative, the system is stable. This allows proving stability without solving differential equations.

Uploaded by

win alfalah
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

EG4321/EG7040

Nonlinear Control

Dr. Matt Turner


EG4321/EG7040

[An introduction to]


Nonlinear Control

Dr. Matt Turner


EG4321/EG7040

[An introduction to]


Nonlinear [System Analysis]
and Control

Dr. Matt Turner


Summary of where we are
Nonlinear systems have complex stability characteristics
Nonlinear system stability is (in general) a function of system initial
conditions and parameters
Phase portraits are an intuitive way of looking at stability BUT
Only useful for 2nd order systems
Construction effectively requires (computer) solution of differential
equations
(using a sufficiently dense grid of initial conditions)
Linearisation is a useful way of looking at stability BUT
Only gives local information
Not always applicable/reliable
Is there a 3rd way?
A.M. Lyapunov
Russian mathematician born in the 1850s.
Many mathematical contributions, but
especially known for contributions to
dynamical systems theory
Ideas form cornerstone of nonlinear control
theory
Died tragically at 61
(shot himself after wife died from tuberculosis)

Lyapunovs main contribution:

Sufficient conditions for stability/asymptotic stability of nonlinear systems


Non-local stability properties about general nonlinear systems
Not reliant on linearisation
Stability Definitions I

Theorem (Stability) x2
Consider the system

x = f (x) x Rn f (.) : Rn 7 Rn

then if there exists a () for each


such that x(0)
x1

kx(0)k kx(t)k t 0

then x = 0 is stable; otherwise it is


unstable

Initial conditions starting near x = 0 dont stray too far


x(t) remains bounded for all time.
Stability Definitions II

Theorem (Asymptotic x2
Stability)
Consider the system
x(0)
x = f (x) x Rn f (.) : Rn 7 Rn

then if x1
lim x(t) = 0 x(0) B
t

then x = 0 is asymptotically stable


with region of attraction B.

All states starting in B eventually converge to x = 0


x(t) decays over time.
Global stability
Preceeding definitions can be made global, if they apply to all initial
conditions x(0) Rn

Global stability:

kx(t)k bounded for all t 0 and all x(0) Rn

Global asymptotic stability:

limt x(t) = 0 for all x(0) Rn

Natural Questions
1. Given a system x = f (x) how can we guarantee stability/asymptotic
stability (of the origin x = 0)?
2. Can we do the above without solving the differential equations?
Rough idea of Lyapunov
1. Select an energy function
2. Show that as time progresses, the energy decreases
3. If energy decreases to zero, origin is asymptotically stable

V (t) Notes
STABLE BUT NOT
VALID ENERGY FUNCTION
Sufficient condition for
stability
GUARANTEED
Origin may be stable if energy
STABLE
function does not
monotonically decrease
t

Critical point: Choice of Energy Function is of paramount


importance in Lyapunovs 2nd method
Adding a little maths
Lyapunovs idea requires that V (t) be decreasing
This implies for t2 > t1

V (t2 ) < V (t1 ) t2 > t1

Because t2 t1 > 0 this is equivalent to

V (t2 ) V (t1 ) V (t1 + ) V (t1 )


<0
t2 t1
if t2 = t1 +
Taking the limit as 0, this implies

V (t) < 0
Interim example: pendulum with friction

State-space equations
l
=
h g b
= sin 2
l ml
mg b, m, l, g > 0

Potential energy = mgh = mg (h0 l cos )


Rotational Kinetic Energy = 12 m 2
Energy function:

V (, ) = P.E . + K .E .
1
= mg (h0 l cos ) + m 2
2
Interim example: pendulum with friction
Stability requires V < 0
 
d 1
V = mg (h0 l cos ) + m 2
dt 2
d d d 1 d
= (mgl cos ) + ( m 2 )
d dt d 2 dt
= (mgl sin ) + m
g b
= mgl sin + m( sin 2 )
l ml
mg b
= mgl sin sin 2 2
l l
Choosing = 1/l 2 then implies

b
V = 2 2 < 0
l
[Actually there is a small issue with this proof - see later]
Lyapunov interpretation on the phase plane
Kinetic Energy type energy function: V (x) = x 2
Balls of energy
B = x : V (x) = x 2


x2
v(x)=x 2

Main idea
1. If state trajectory
traverses nested sets....
2. .....energy decreases as x1
time increases

System is stable
Lyapunov interpretation on the phase plane
Kinetic Energy type energy function: V (x) = x 2
Balls of energy
B = x : V (x) = x 2


x2
v(x)=x 2
V(t 1) V(t 2)<V(t 1)
Main idea
1. If state trajectory
traverses nested sets.... V(t 2)
2. .....energy decreases as x1
time increases

System is stable
Lyapunov interpretation on the phase plane
Kinetic Energy type energy function: V (x) = x 2
Balls of energy
B = x : V (x) = x 2


x(t) x2
v(x)=x 2
Main idea
1. If state trajectory
traverses nested sets....
2. .....energy decreases as
time increases
x1

System is stable
More general Lyapunov functions
Main idea of Lyapunov was to use abstract energy functions
Energy functions did not necessarily correlate with physical energy
in systems
Distillation of idea:
1. Choose positive (scalar) function of system state: V (x) > 0
2. If derivative negative V (x) < 0 System stable
Advantages: huge choice of Lyapunov function candidates V (x)
Disadvantages: Sufficiency!
If, for a certain V (x) > 0, we have that V (x) < 0 - great!
If, for a certain V (x) > 0, V (x) 0....
.... system still may be stable - just have not proved it!

You might also like