Anda di halaman 1dari 26

Nonlinear Systems: an Introduction

to Lyapunov Stability Theory

A. Pascoal , April 2013 (draft under revision)

Linear versus Nonlinear


Control
Nonlinear
Plant

Linear based control laws

+ Good engineering intuition


for linear designs (local
stability and performance)
- Lack of global stability
and performance results
- Poor physical intuition

Nonlinear control laws

+ Powerful robust stability


analysis tools
+ Possible deep physical
insight
- Need for stronger
theoretical background
- Limited tools for
performance analysis

AUV speed control

Dynamics

Nonlinear Control: Key


Ingredients
T

v
2

Nonlinear
Plant

Objective: generate T(t) so that v (t ) tracks the reference speed vr (t )


Tracking error

Error Dynamics

e vr v

Nonlinear Control: Key


Ingredients
Error Dynamics

mT

de(t )
dv
mT r T v fv 2
dt
dt

Nonlinear Control Law


dvr
T mT (
Ke) v fv 2
dt
de
dvr
dvr

mT
mT
mT (
ke) v fv 2 v fv 2
dt
dt
dt

de
e( t ) e(0) exp( kt ); t 0
ke 0
dt

Nonlinear Control: Key


Ingredients
T mT (

dvr
Ke) v fv 2
dt

Tracking error tends to


zero exponentially fast.

e ( t ) e (0) exp( kt ); t 0

Simple and elegant!


Catch: the nonlinear dynamics are known EXACTLY.

Key idea: i) use simple concepts, ii) deal with


robustness against parameter uncertainty.
New tools are needed: LYAPUNOV theory

SIMPLE EXAMPLE

Lyapunov theory of stability: a soft


Intro

(free mass, subjected to a simple motion resisting force)

dv
m
fv 0
dt

fv
v

dv
f
v
dt
m
v(t ) e

f
( t t0 )
m

m/f

v(t0 )

v=0 is an equilibrium point; dv/dt=0 when v=0!

v=0 is attractive
(trajectories
converge to 0)

fv

Lyapunov theory of stability: a soft


Intro

How can one prove that the trajectories go to the equilibrium point
WITHOUT SOLVING the differential equation?

1
V (v) mv 2
2

V (v) 0, v 0;
V 0, v 0

(energy function)

dV (v(t )) V
dv (t )

dt
v |v (t ) dt
dV
dv
mv
fv 2 0
dt
dt

V positive and bounded below by zero;


dV/dt negative implies convergence
of V to 0!

fv

Lyapunov theory of stability: a soft


Intro

What are the BENEFITS of this seemingly strange approach to investigate


convergence of the trajectories to an equilibrium point?

f(v)

dv
m
f (v ) 0
dt

1
V (v) mv 2
2
dV
dv
mv
vf (v) 0
dt
dt

f a general dissipative force


e.g. v|v|

Q-I

V positive and bounded below by zero;

0
Q-III

dV/dt negative implies convergence of V to 0!


Very general form of nonlinear equation!

2-D case

Lyapunov theory of stability: a soft


Intro

dx1
x2 k1 ( x1 );
dt
dx2
x1 k 2 ( x2 )
dt

x1
x
x2
State
vector

1 2
2
V ( x) ( x1 x2 )
2
V ( x)

1 T
x Qx; Q I 0
2

Q-positive definite

dx
f (x)
dt

V (v) 0, v 0;
V 0, v 0

Lyapunov theory of stability: a soft


Intro

2-D case

dx1
x2 k1 ( x1 );
dt
dx2
x1 k 2 ( x2 )
dt

1 2
2
V ( x) ( x1 x2 )
2

V ( x(t )) x (t ) t
V ( x(t )) R x(t ) R 2 t R
dV ( x) V

dt
x
1x1

1x2

dx
dt
2x1

x1

x
2

dx
f (x)
dt

x2 k1 ( x1 )
dV ( x )
x1 , x2

k
(
x
)
dt
1
2
2

x2 k1 ( x1 )
dV ( x )
x1 , x2

k
(
x
)
dt
1
2
2

dV ( x)
x1 x2 x1k1 ( x1 ) x2 x1 x2 k 2 ( x2 ) 0
dt

V positive and bounded below by zero;


dV/dt negative implies convergence of V to 0!

x tends do 0!

Lyapunov theory of stability: a soft


Intro

Shifting

Is the origin always the TRUE origin?


y-measured from spring at rest
dy
f( )
dt

k ( y)

dy
d2y
k ( y ) f ( ) mg m 2
dt
dt
Equilibrium point xeq: dx/dt=0

y
mg

y
yeq
x dy ; xeq

dt

k ( y eq ) mg

dx
F ( x); F ( xeq ) 0
dt

Examine if yeq is attractive!

x xeq
dx dxeq d
d

dt
dt
dt
dt

d
dx

F ( xeq ) G ( )
dt
dt
Examine the
G (0) F ( xeq 0) 0
ZERO eq. Point!

Lyapunov theory of stability: a soft


Intro

Shifting

Is the origin always the TRUE origin?


dxref (t )
dx(t )
F ( x(t ));
F ( xref (t ))
dt
dt
xref(t) is a solution
Examine if xref(t) is attractive!

x xref
dx dxref
d

dt
dt
dt
d
F ( xref )
dt

d
dx

F ( xref ) F ( xref ) F ( xref ) G ( , t )


dt
dt
G (0, t ) F ( xref (t ) 0) F ( xref (t )) 0

Examine the
ZERO eq. Point!

Lyapunov theory of stability: a soft


Intro
dx
f ( x, u ); y g ( x); g (0) 0
dt

Control Action
u

Nonlinear

plant
Static control
law

u h( y ); h(0) 0

dx
f ( x, h( g ( x))); f (0) 0
dt
dx
F ( x); F (0) 0
dt

Investigate if 0
is attractive!

dx
f ( x) ; f (0) 0
dt

Lyapunov Theory

Stability of the zero solution


The zero solution is STABLE if

x-space

0, ( ) 0 : x(to ) B (0) x(t ) B (0); t t0

dx
f ( x) ; f (0) 0
dt

Lyapunov Theory

Attractiveness of the zero solution


The zero solution is locally ATTRACTIVE if

x-space

0 : x(to ) B (0) lim x(t ) t 0

dx
f ( x) ; f (0) 0
dt

Lyapunov Theory

The zero solution is locally


ASYMPTOTICALLY STABLE if
it is STABLE and ATTRACTIVE
One may have attractiveness but NOT
Stability!

(the two conditions are required for


Asymptotic Stability!)

Key (a
Ingredients
for Nonlinear
Lyapunov Theory
formal approach)
Control
dx
f (x)
dt

(1)

Lyapunov Theory

(the two conditions are required for

Asymptotic Stability!)

Lyapunov Theory
There are at least three ways of assessing the stability (of
an equilibrium point of a) system:
Solve the differential equation (brute-force)
Linearize the dynamics and examine the behaviour
of the resulting linear system (local results for hyperbolic
eq. points only)
Use Lypaunovs direct method (elegant and powerful,
may yield global results)

Lyapunov Theory

Lyapunov Theory

If

dV ( x)
0;V ( x) as x
dt
then the origin is
globally asymptotically
stable

Lyapunov Theory
What happens when

dV ( x)
0?
dt

Is the situation hopeless?

No!

dV ( x)
Let
0; define
dt
dV ( x)

: x :
0
dt

Suppose the only trajectory of the system


entirely contained in is the null trajectory.
Then, the origin is asymptotically stable

Krazovskii-LaSalle

(Let M be the largest invariant set


contained in . Then all solutions
converge to M. If M is the origin, the
results follows)

Krazovskii-La Salle
f(

dy
)
dt

Lyapunov Theory
2

dy
d y
k ( y) f ( ) m 2
dt
dt

x dy ; xeq

dt

k ( y)

dx1
x2

dt

1
1
k(x )
f ( x2 )
dx2
1
m

m
dt

dx
F ( x); F (0) 0
dt

V ( x) Kinetic Energy Potential Energy


x

1
1
2
V ( x) mx2 k ( )d
2
0

Krazovskii-La Salle
dy
f( )
dt

Lyapunov Theory
f(.), k(.) 1st and 3rd quadrants

k ( y)

f(0)=k(0)=0

dx1
x2

dt

1
1
k(x )
f ( x2 )
dx2
1

m
m
dt

1
1
2
V ( x) mx2 k ( )d
2
0

V(x)>0!

dV ( x)
dx2
dx
mx
k ( x1 ) 1
dt
dt
2 dt
mx 2 (

1
1
k ( x1 )
f ( x2 )) k ( x1 ) x2 f ( x2 ) x2 0!
m
m

Krazovskii-La Salle
f(

dy
)
dt

Lyapunov Theory
dx1
x2

dt


1
1
k(x )
f ( x2 )
dx2
1
m
m

dt

k ( y)

dV ( x)
f ( x2 ) x2 0!
dt
dV
0 for x2 0!
dt

x2

x1

dx1
0

dt

1
k ( x1 )
dx2
m

dt

Examine dynamics here!

Trajectory
leaves
unless x1=0!

M is the origin.
The origin is
asymptotically
stable!

Nonlinear Systems: an Introduction


to Lyapunov Stability Theory

Anda mungkin juga menyukai