Lecture 6: Stability of Dynamic Systems Lyapunov's Direct Method 6.1 Notions of Stability
Lecture 6: Stability of Dynamic Systems Lyapunov's Direct Method 6.1 Notions of Stability
Lecture 6: Stability of Dynamic Systems Lyapunov's Direct Method 6.1 Notions of Stability
1
Example 1 (Unstable Equilibrium Point That Attracts All Trajectories)
Consider the second-order system with state variables x1 and x2 whose dynamics
are most easily described in polar coordinates via the equations
r_ = r(1 ; r)
_ = sin2 (=2) (6.3)
q
where the radius r is given by r = x21 + x22 and the angle by 0 =
arctan (x2 =x1 ) < 2. (You might try obtaining a state-space description directly
involving x1 and x2 .) It is easy to see that there are precisely two equilibrium
points: one at the origin, and the other at r = 1, = 0. We leave you to verify
with rough calculations (or computer simulation from various initial conditions)
that the trajectories of the system have the form shown in the gure below.
Evidently all trajectories (except the trivial one that starts and stays at the
origin) end up at r = 1, = 0. However, this equilibrium point is not stable
i.s.L., because these trajectories cannot be conned to an arbitrarily small ball
around the equilibrium point when they are released from arbitrary points with
any ball (no matter how small) around this equilibrium.
3
in this ball. Such a V (x) may be thought of as an \energy" function. Let V_ (x) denote the
time derivative of V (x) along any trajectory of the system, i.e. its rate of change as x(t)
varies according to (6.8). If this derivative is negative throughout the region (except at
the origin), then this implies that the energy is strictly decreasing over time. In this case,
because the energy is lower bounded by 0, the energy must go to 0, which implies that all
trajectories converge to the zero state. We will formalize this idea in the following sections.
6.4.2 Lyapunov Functions
Denition 6.2 Let V be a continuous map from R n to R . We call V (x) a locally positive
denite (lpd) function around x = 0 if
1. V (0) = 0,
2. V (x) > 0; 0 < kxk < r for some r.
Similarly, the function is called locally positive semidenite (lpsd) if the strict inequality on
the function in the second condition is replaced by V (x) 0. The function V (x) is locally
negative denite (lnd) if ;V (x) is lpd, and locally negative semidenite (lnsd) if ;V (x) is
lpsd. What may be useful in forming a mental picture of an lpd function V (x) is to think
of it as having \contours" of constant V that form (at least in a small region around the
origin) a nested set of closed surfaces surrounding the origin. The situation for n = 2 is
illustrated in Figure 6.1.
V(x)=c 2
V(x)=c 3
V(x)=c 1
Figure 6.1: Level lines for a Lyapunov function, where c1 < c2 < c3 .
Throughout our treatment of the CT case, we shall restrict ourselves to V (x) that have
continuous rst partial derivatives. (Dierentiability will not be needed in the DT case |
continuity will suce there.) We shall denote the derivative of such a V with respect to
time along a trajectory of the system (6.8) by V_ (x(t)). This derivative is given by
V_ (x(t)) = dV (x) x_ = dV (x) f (x)
dx dx
where dVdx(x) is a row vector | the gradient vector or Jacobian of V with respect to x |
containing the component-wise partial derivatives @x @V .
i
Denition 6.3 Let V be an lpd function (a \candidate Lyapunov function"), and let V_ be
its derivative along trajectories of system (6.8). If V_ is lnsd, then V is called a Lyapunov
function of the system (6.8).
4
6.4.3 Lyapunov Theorem for Local Stability
Theorem 6.1 If there exists a Lyapunov function of system (6.8), then x = 0 is a stable
equilibrium point in the sense of Lyapunov. If in addition V_ (x) < 0, 0 < kxk < r for some
1
r1, i.e. if V_ is lnd, then x = 0 is an asymptotically stable equalibrium point.
Proof: First, we prove stability in the sense of Lyapunov. Suppose > 0 is given. We need
to nd a > 0 such that for all kx(0)k < , it follows that kx(t)k < ; for all t > 0. The
Figure 6.2 illustrates the constructions of the proof for the case n = 2. Let 1 = min(; r).
ε1
r δ
Since V (x) is continuous, the above m is well dened and positive. Choose satisfying 0 <
< 1 such that for all kxk < , V (x) < m. Such a choice is always possible, again because
of the continuity of V (x). Now, consider any x(0) such that kx(0)k < , V (x(0)) < m,
and let x(t) be the resulting trajectory. V (x(t)) is non-increasing (i.e. V_ (x(t)) 0) which
results in V (x(t)) < m. We will show that this implies that kx(t)k < 1 . Suppose there
exists t1 such that kx(t1 )k > 1 , then by continuity we must have that at an earlier time
t2, kx(t2 )k = 1 , and minkxk=1 kV (x)k = m > V (x(t2 )), which is a contradiction. Thus
stability in the sense of Lyapunov holds.
Using the rst part of the Theorem, choose > 0 such that for all initial conditions x(0)
that satisfy kx(0)k < , we have that kx(t)k < min(; r; r1 ). To prove asymptotic stability
when V_ is lnd, we need to show that as t ! 1, V (x(t)) ! 0; then, by continuity of V ,
kx(t)k ! 0. Suppose x(0) satises kx(0)k < , and x(t) is the resulting trajectory. Since
V (x(t)) is strictly decreasing, and V (x(t)) 0 we know that V (x(t)) ! c, with c 0. We
want to show that c is in fact zero. We can argue by contradiction and suppose that c > 0.
Let the set S be dened as
S = fx 2 R njV (x) cg ;
and let B be a ball inside S of radius ,
B = fx 2 R n jkxk < g :
We know that V (x(t)) is decreasing monotonically to c and V (x(t)) > c for all t, therefore,
x(t) 2= B; recall that B S which is dened as all the elements in R n for which V (x) c.
5
We can dene the largest derivative of V (x) as
;
= k
maxxk
V_ (x):
-1 1 x
6
point is globally asymptotically stable (or asymptotically stable \in the large") if its basin
of attraction is the entire state space.
If a function V (x) is positive denite on the entire state space, and has the additional
property that jV (x)j % 1 as kxk % 1, and if its derivative V_ is negative denite on the
entire state space, then the equilibrium point at the origin is globally asymptotically stable.
We omit the proof of this result. Other versions of such results can be stated, but are also
omitted.
Example 3
Consider the nth-order system
x_ = ;C (x)
with the property that C (0) = 0 and x0 C (x) > 0 if x 6= 0. Convince yourself that
the unique equilibrium point of the system is at 0. Now consider the candidate
Lyapunov function
V (x) = x0x
which satises all the desired properties, including jV (x)j % 1 as kxk % 1.
Evaluating its derivative along trajectories, we get
V_ (x) = 2x0 x_ = ;2x0 C (x) < 0 for x 6= 0:
Hence, the system is globally asymptotically stable.
Example 4 Consider the following dynamical system
x_ 1 = ;x1 + 4x2
x_ 2 = ;x1 ; x32 :
The only equlibrium point for this system is the origin x = 0. To investigate the
stability of the origin let us propose a quadratic Lyapunov function V = x21 +ax22,
where a is a positive constant to be determined. It is clear that V is positive
denite on the entire state space R 2 . In addition, V is radially unbounded, that
is it satises jV (x)j % 1 as kxk % 1. The derivative of V along the trajectories
of the system is given by
h i " ;x + 4x #
_V = 2x 2ax 1 2
1 2
;x ; x 1
3
2
If we choose a = 4 then we can eliminate the cross term x1 x2 , and the derivative
of V becomes
V_ = ;2x21 ; 8x42 ;
which is clearly a negative denite function on the entire state space. Therefore
we conclude that x = 0 is a globally asymptotically stable equilibrium point.
Example 5 A highly studied example in the area of dynamical systems and
chaos is the famous Lorenz system, which is comprised of the following set of
dierential equations
x_ = (y ; x)
y_ = rx ; y ; xz
z_ = xy ; bz;
7
where , r and b are positive constants. This system of equations provides
an approximate model of a horizontal
uid layer that is heated from below.
The warmer
uid from the bottom rises and thus causes convection currents.
This approximates what happens in the atmosphere. Under intense heating
this model exhibits complex dynamical behaviour. However, in this example we
would like to analyze the stability of the origin under the condition r < 1, which
is known not to lead to complex behaviour. Le us dene V = 1 x2 + 2 y2 + 3 z 2 ,
where 1 , 2 , and 3 are positive constants to be determined. It is clear that V
is positive denite on R 3 and is radially unbounded. The derivative of V along
the trajectories of the system is given by
2 3
h i6 (y ; x) 7
_V = 21 x 22 y 23 z rx ; y ; xz 5
4
xy ; bz
= ;21 x2 ; 22 y2 ; 23 bz 2
+xy(21 + 2r2 ) + (23 ; 22 )xyz:
If we choose 2 = 3 = 1 and 1 = 1 then the V_ becomes
V_ = ;2 x2 + y2 + 2bz 2 ; (1 + r)xy
" 2 #
1 1 + r
= ;2 x ; 2 (1 + r)y + 1 ; ( 2 ) y2 + bz 2 :
2
Since 0 < r < 1 it follows that 0 < 1+2 r < 1 and therefore V_ is negative
denite on the entire state space R 3 . This implies that the origin is globally
asymptotically stable.
Example 6 (Pendulum)
The dynamic equation of a pendulum comprising a mass M at the end of a rigid
but massless rod of length R is
MR + Mg sin = 0
where is the angle made with the downward direction, and g is the acceleration
due to gravity. To put the system in state-space form, let x1 = , and x2 = _;
then
x_1 = x2
x_2 = ; Rg sin x1
Take as a candidate Lyapunov function the total energy in the system. Then
V (x) = 12 MR2x22 + MgR(1 ; cos x1) = kinetic + potential
" #
_V = dV f (x) = [MgR sin x1 MR2 x2 ] x2
dx ; g sin x R 1
= 0
Hence, V is a Lyapunov function and the system is stable i.s.L. We cannot
conclude asymptotic stability with this analysis.
8
Consider now adding a damping torque proportional to the velocity, so that the
state-space description becomes
x_1 = x2
x_2 = ;Dx2 ; Rg sin x1
With this change, but the same V as before, we nd
V_ = ;DMR2x22 0:
From this we can conclude stability i.s.L. We still cannot directly conclude
asymptotic stability. Notice however that V_ = 0 ) _ = 0. Under this con-
dition, = ;(g=R) sin : Hence, 6= 0 if 6= k for integer k, i.e. if the
pendulum is not vertically down or vertically up. This implies that, unless we
are at the bottom or top with zero velocity, we shall have 6= 0 when V_ = 0, so
_ will not remain at 0, and hence the Lyapunov function will begin to decrease
again. The only place the system can end up, therefore, is with zero velocity,
hanging vertically down or standing vertically up, i.e. at one of the two equilib-
ria. The formal proof of this result in the general case (\LaSalle's invariant set
theorem") is beyond the scope of this course.
The conclusion of local asymptotic stability can also be obtained directly through
an alternative choice of Lyapunov function. Consider the Lyapunov function
candidate
V (x) = 21 x22 + 21 (x1 + x2 )2 + 2(1 ; cos x1 ):
It follows that
V_ = ;(x22 + x1 sin x1 ) = ; ; (_2 + sin ) 0:
Also, _ 2 + sin = 0 ) _2 = 0; sin = 0 ) = 0; _ = 0: Hence, V_ is strictly
negative in a small neighborhood around 0. This proves asymptotic stability.
6.4.5 Discrete-Time Systems
Essentially identical results hold for the system
x(k + 1) = f (x(k)) (6.9)
provided we interpret V_ as
4
V_ (x) = V (f (x)) ; V (x) ;
i.e. as
V (next state) ; V (present state)
Example 7 (DT System)
Consider the system
x1 (k + 1) = 1 +x2x(k2 ()k)
2
x 1 (k )
x2 (k + 1) = 1 + x2 (k)
2
9
which has its only equilibrium at the origin. If we choose the quadratic Lyapunov
function
V (x) = x21 + x22
we nd
_V (x(k)) = V (x(k)) 1
[1 + x2 (k)]2 ; 1 0
2
from which we can conclude that the equilibrium point is stable i.s.L. In fact,
examining the above relations more carefully (in the same style as we did for the
pendulum with damping), it is possible to conclude that the equilibrium point
is actually globally asymptotically stable.
10