Professional Documents
Culture Documents
3. Lyapunov Stability
In this chapter we will study stability of equilibrium points of autonomous
system
x f ( x )
including linear systems. For a linear time-invariant system
x Ax
the stability of equilibrium point x 0 can be completely defined by
location of eigenvalues of matrix A.
Stability of equilibrium points is usually characterized in the sense of
Lyapunov (1892). Alexander Mikhilovich Lyapunov, Russian
mathematician introduced in his work The general Problem of Motion
Stability the basic definition of stability that are in use today and proved
many of fundamental theorems.
Lyapunov stability is concerned with behavior of the trajectories of
a system when its initial state is near an equilibrium.
- An equilibrium point is stable if all solutions starting at nearby
points stay nearby, otherwise, it is unstable.
- It is asymptotically stable if all solutions starting at nearby points
not only stay nearby, but tend to the equilibrium points with t .
Lyapunov stability theorems give sufficient conditions for stability,
asymptotic stability and exponential stability. They do not say whether the
given conditions are also necessary.
Some Lyapunov stability theorems establish sufficient and
necessary conditions. Such theorems are usually called converge
theorems. We will consider two converge theorems. We use this theorems
for exponential stability to show that an equilibrium point of a nonlinear
system is exponentially stable if and only if the linearization of the system
about this point has an exponentially stable equilibrium at the origin.
g
x 1 x 2 , x 2 sin x12
l
we have seen that trajectories in the neighborhood of the first equilibrium
x1 0, x 2 0 are closed orbit. Choosing 0 , we have a ball B with radius
. Then choose 0 such that the ball B contents the ball B . So, all
trajectories starting in B will stay within B . Therefore, the
requirement for stability is satisfied. This equilibrium point however, is not
asymptotically stable since the trajectories starting off the equilibrium
point do not tend to it. They remain in their closed orbits.
When friction is taken in consideration ( k 0 ) the equilibrium point
x1 0, x 2 0 becomes a stable focus. Inspection of the phase portrait of the
stable focus shows that the requirement for stability is satisfied. In
addition, trajectories starting close to the equilibrium point tend to it as t
tends to .
The second equilibrium point at x1 , x 2 0 is a saddle point. Clearly
the requirement cannot be satisfied since for any 0 , there is always
a trajectory will leave the ball B x : x even x(0) is arbitrarily close to
the equilibrium point.
Remark. Having definition of stability, we determined the stability of
pendulum equilibrium points using the phase portrait of the pendulum
equation. But in general case this may be difficult or even impossible.
Our task now to find ways to determine stability
forms a closed contour around x 0 for small c, we can again arrive at the
conclusion that x 0 is stable equilibrium point.
Case 2. With friction ( k 0 ), we have the energy dissipate during the
motion of the system, that is,
dE 1 g k g k
2 x2 sin x1 x2 x2 sin x1 x2 x22 0
dt 2 l m l m
along the trajectories of the system. Due to friction energy E cannot
remain constant along the motion. Hence, E is decreasing until it reaches
zero. Hence, the trajectory tends to x 0 as t tends to .
V ( x ) 0 V ( x(t )) V ( x(0)) , t 0
Since is a compact set (it is closed by definition and bounded
since it is contained in Br ), and x(t ) t 0 , then the system (1) by
Theorem 3 has a unique solution defined for all t 0
Since V (x) is continuous in x and V ( 0) 0 , there exist 0
x V (x )
Then
B Br
and
x( 0) B x( 0) x( t ) x( t ) Br
Therefore,
x ( 0) x (t ) r , t 0
which exists because the continuous function has maximum over the
compact set d x r . By condition (4), 0 . It follows that
t
V ( x( t )) V ( x( 0)) V ( x( ))d V ( x( 0)) t
0
Since the right-hand side will become negative, the inequality contradicts
the assumption that c>0. Therefore, V ( x( t )) 0 as t , and by definition
V ( x(t )) 0 if and only if x(t ) 0 . Hence,
x( t ) 0 as t