Applied Nonlinear Control Nguyen Tan Tien - 2002.

3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

3. Fundamentals of Lyapunov Theory

The objective of this chapter is to present Lyapunov stability has a single equilibrium point (the origin 0) if A is nonsingular.
theorem and illustrate its use in the analysis and the design of If A is singular, it has an infinity of equilibrium points, which
nonlinear systems. contained in the null-space of the matrix A, i.e., the subspace
defined by Ax = 0. A nonlinear system can have several (or
3.1 Nonlinear Systems and Equilibrium Points infinitely many) isolated equilibrium points.

Nonlinear systems Example 3.1 The pendulum___________________________

A nonlinear dynamic system can usually be presented by the
set of nonlinear differential equations in the form

x& = f (x, t ) (3.1) R

where
n
f R : nonlinear vector function
x Rn : state vectors Fig. 3.1 Pendulum
n : order of the system
Consider the pendulum of Fig. 3.1, whose dynamics is given
The form (3.1) can represent both closed-loop dynamics of a by the following nonlinear autonomous equation
feedback control system and the dynamic systems where no
control signals are involved. MR 2&& + b& + MgR sin = 0 (3.5)

A special class of nonlinear systems is linear system. The where R is the pendulums length, M its mass, b the friction
dynamics of linear systems are of the from x& = A(t ) x with coefficient at the hinge, and g the gravity constant. Leting
x1 = , x2 = & , the corresponding state-space equation is
A R nn .
x&1 = x2 (3.6a)
Autonomous and non-autonomous systems
Linear systems are classified as either time-varying or time- g b
x& 2 = x sin x1
2 2
(3.6b)
invariant. For nonlinear systems, these adjectives are replaced MR R
by autonomous and non-autonomous.
Therefore the equilibrium points are given by x2 = 0,
Definition 3.1 The nonlinear system (3.1) is said to be
sin( x1 ) = 0, which leads to the points (0 [2 ], 0) and
autonomous if f does not depend explicitly on time, i.e., if the
systems state equation can be written ( [2 ], 0) . Physically, these points correspond to the
pendulum resting exactly at the vertical up and down points.
x& = f (x) (3.2) __________________________________________________________________________________________

In linear system analysis and design, for notational and

Otherwise, the system is called non-autonomous. analytical simplicity, we often transform the linear system
equations in such a way that the equilibrium point is the origin
Equilibrium points of the state-space.
It is possible for a system trajectory to only a single point.
Such a point is called an equilibrium point. As we shall see Nominal motion
later, many stability problems are naturally formulated with
Let x* (t ) be the solution of x& = f (x) , i.e., the nominal motion
respect to equilibrium points.
trajectory, corresponding to initial condition x* (0) = x 0 . Let
*
Definition 3.2 A state x is an equilibrium state (or us now perturb the initial condition to be x(0) = x 0 + x 0 , and
equilibrium points) of the system if once x(t ) is equal to x * , it study the associated variation of the motion error
remains equal to x * for all future time. e(t ) = x(t ) x* (t ) as illustrated in Fig. 3.2.
x2
Mathematically, this means that the constant vector x *
satisfies
x(t ) e(t )
*
0 = f (x ) (3.3) x* (t )
x1
Equilibrium points can be found using (3.3).
A linear time-invariant system xn

x& = A x Fig. 3.2 Nominal and perturbed motions (3.4)

___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 7
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Since both x * (t ) and x(t ) are solutions of (3.2): x& = f (x) , we R > 0, r > 0, x(0) < r t 0, x(t ) < R
have or, equivalently
x& * = f (x * ) x * (0) = x 0
R > 0, r > 0, x(0) B r t 0, x(t ) B r
x& = f (x) x ( 0) = x 0 + x 0
Essentially, stability (also called stability in the sense of
then e(t ) satisfies the following non-autonomous differential Lyapunov, or Lyapunov stability) means that the system
equation trajectory can be kept arbitrarily close to the origin by starting
sufficiently close to it. More formally, the definition states that
e& = f (x * + e, t ) f (x * , t ) = g (e, t ) (3.8) the origin is stable, if, given that we do not want the state
trajectory x(t ) to get out of a ball of arbitrarily specified radius
with initial condition e(0) = x(0) . Since g (0, t ) = 0 , the new B R . The geometrical implication of stability is indicated in
dynamic system, with e as state and g in place of f, has an Fig. 33.
equilibrium point at the origin of the state space. Therefore, 3 curve 1 - asymptotically stable
instead of studying the deviation of x(t ) from x * (t ) for the 1 curve 2 - marginally stable
original system, we may simply study the stability of the 2 curve 3 - unstable
perturbation dynamics (3.8) with respect to the equilibrium
point 0. However, the perturbation dynamics non-autonomous 0
system, due to the presence of the nominal trajectory x * (t ) on x(0) S r
the right hand side.

Example 3.2________________________________________
SR
Consider the autonomous mass-spring system Fig. 3.3 Concepts of stability

Asymptotic stability and exponential stability

m &x& + k1 x& + k 2 x 3 = 0
In many engineering applications, Lyapunov stability is not
enough. For example, when a satellites attitude is disturbed
which contains a nonlinear term reflecting the hardening effect
from its nominal position, we not only want the satellite to
of the spring. Let us study the stability of the motion
maintain its attitude in a range determined by the magnitude of
x* (t ) which starts from initial point x0 . Assume that we the disturbance, i.e., Lyapunov stability, but also required that
slightly perturb the initial position to be x(0) = x0 + x0 . The the attitude gradually go back to its original value. This type
of engineering requirement is captured by the concept of
resulting system trajectory is denoted as x(t ) . Proceeding as asymptotic stability.
before, the equivalent differential equation governing the
motion error e is Definition 3.4 An equilibrium points 0 is asymptotically stable
if it is stable, and if in addition there exist some r > 0 such
m &e& + k1e + k 2 [e 3 + 3e 2 x * (t ) + 3e x *2 (t )] = 0 that x(0) r implies that x(t ) 0 as t .

Clearly, this is a non-autonomous system. Asymptotic stability means that the equilibrium is stable, and
__________________________________________________________________________________________
in addition, states start close to 0 actually converge to 0 as
3.2 Concepts of Stability time goes to infinity. Fig. 3.3 shows that the system
trajectories starting form within the ball B r converge to the
Notation origin. The ball B r is called a domain of attraction of the
B R : spherical region (or ball) defined by x R equilibrium point.
S R : spherical itself defined by x = R
In many engineering applications, it is still not sufficient to
: for any know that a system will converge to the equilibrium point
: there exist after infinite time. There is a need to estimate how fast the
: in the set system trajectory approaches 0. The concept of exponential
: implies that stability can be used for this purpose.
: equivalent
Definition 3.5 An equilibrium points 0 is exponential stable if
Stability and instability there exist two strictly positive number and such that
Definition 3.3 The equilibrium state x = 0 is said to be stable
if, for any R > 0 , there exist r > 0 , such that if x(0) r then t > 0, x(t ) x(0) e t (3.9)
x(t ) R for all t 0 . Otherwise, the equilibrium point is
unstable. in some ball B r around the origin.

Using the above symbols, Definition 3.3 can be written in the (3.9) means that the state vector of an exponentially stable
form system converges to the origin faster than an exponential
___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 8
 Applied Nonlinear Control
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
function. The positive number is called the rate of
exponential convergence.
For example, the system x& = (1 + sin 2 x) x is exponentially
convergent to x = 0 with the rate = 1 . Indeed, its solution is
0t [1+sin 2 x ( )] d
x(t ) = x(0) e
Note that exponential stability implies asymptotic stability.
But asymptotic stability does not implies guarantee
exponential stability, as can be seen from the system
x& = x 2 , x(0) = 1
whose solution is x = 1 /(1 + t ) , a function slower than any
exponential function e t .
Local and global stability
Definition 3.6 If asymptotic (or exponential) stability holds
for any initial states, the equilibrium point is said to be
asymptotically (or exponentially) stable in the large. It is also
called globally asymptotically (or exponentially) stable.
3.3 Linearization and Local Stability
Lyapunovs linearization method is concerned with the local
stability of a nonlinear system. It is a formalization of the
intuition that a nonlinear system should behave similarly to its
linearized approximation for small range motions.
Consider the autonomous system in (3.2), and assumed that
f(x) is continuously differentiable. Then the system dynamics
can be written as
f
x& = x + f h.o.t . (x)
x x =0
where f h.o.t . stands for higher-order terms in x. Let us use the
constant matrix A denote the Jacobian matrix of f with respect
f thus letting =
to x at x = 0: A = . Then, the system
x x =0
x& = A x (3.12)
is called the linearization (or linear approximation) of the
original system at the equilibrium point 0.
In practice, finding a systems linearization is often most
easily done simply neglecting any term of order higher than 1
in the dynamics, as we now illustrate.
Example 3.4________________________________________
Consider the nonlinear system
x&1 = x22 + x1 cos x 2
x& 2 = x 2 + ( x1 + 1) x1 + x1 sin x 2
Its linearized approximation about x = 0 is
x&1 = 0 + x1.1
x& 2 = x 2 + 0 + x1 + x1 x2 x2 + x1
1 0
The linearized system can thus be written x& = x .
1 1 __________________________________________________________________________________________
___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory
Nguyen Tan Tien - 2002.3
A similar procedure can be applied for a controlled system.
Consider the system &x& + 4 x& 5 + ( x 2 + 1) u = 0 . The system can
be linearly approximated about x = 0 as &x& + 0 + (0 + 1) u = 0 or
&x& = u . Assume that the control law for the original nonlinear
system has been selected to be u = sin x + x 3 + x cos 2 x , then
, and therefore x(t ) x(0) e t .
the linearized closed-loop dynamics is &x& + x& + x = 0 .
__________________________________________________________________________________________
The following result makes precise the relationship between
the stability of the linear system (3.2) and that of the original
nonlinear system (3.2).
(3.10)
Theorem 3.1 (Lyapunovs linearization method)
If the linearized system is strictly stable (i.e., if all
eigenvalues of A are strictly in the left-half complex plane),
then the equilibrium point is asymptotically stable (for the
actual nonlinear system).
If the linearizad system is un stable (i.e., if at least one
eigenvalue of A is strictly in the right-half complex plane),
then the equilibrium point is unstablle (for the nonlinear
system).
If the linearized system is marginally stable (i.e., if all
eigenvalues of A are in the left-half complex plane but at
least one of them is on the j axis), then one cannot
conclude anything from the linear approximation (the
equilibrium point may be stable, asymptotically stable, or
unstable for the nonlinear system).
Example 3.5________________________________________
Consider the equilibrium point ( = ,& = 0) of the pendulum
(3.11) in the example 3.1. Since the neighborhood of = , we can
write
sin = sin + cos ( ) + h.o.t. = + h.o.t.
~
, the systems linearization about the
equilibrium point ( = ,& = 0) is
~
&& ~& g ~
b
+ =0
2
MR R
Hence its linear approximation is unstable, and therefore so is
the nonlinear system at this equilibrium point.
__________________________________________________________________________________________
Example 3.5________________________________________
Consider the first-order system x& = a x + b x 5 . The origin 0 is
one of the two equilibrium of this system. The linearization of
this system around the origin is x& = a x . The application of
Lyapunovs linearization method indicate the following
stability properties of the nonlinear system
a < 0 : asymptotically stable
a > 0 : unstable
a = 0 : cannot tell from the linearization
In the third case, the nonlinear system is x& = b x 5 . The
linearization method fails while, as we shall see, the direct
method to be described can easily solve this problem.
9
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

3.4 Lyapunovs Direct Method 3.4.1. Positive definite functions and Lyapunov functions
Definition 3.7 A scalar continuous function V (x) is said to be
The basic philosophy of Lyapunovs direct method is the locally positive definite if V (0) = 0 and, in a ball B R0
mathematical extension of a fundamental physical
observation: if the total energy of a mechanical (or electrical)
system is continuous dissipated, then the system, whether x0 V ( x) > 0
linear or nonlinear, must eventually settle down to an
equilibrium point. Thus, we may conclude the stability of a If V (0) = 0 and the above property holds over the whole state
system by examining the variation of a single scalar function. space, then V (x) is said to be globally positive definite.

Let consider the nonlinear mass-damper-spring system in Fig. 1

For instance, the function V (x) = MR 2 x22 + MR(1 cos x1 )
3.6, whose dynamic equation is 2
which is the mechanical energy of the pendulum in Example
m &x& + b x& x& + k 0 x + k1 x 3 = 0 (3.13) 3.1, is locally positive definite.

Let us describe the geometrical meaning of locally positive

with definite functions. Consider a positive definite function
b x& x& : nonlinear dissipation or damping V (x) of two state variables x1 and x2 . In 3-dimensional space,
k 0 x + k1 x 3 : nonlinear spring term V (x) typically corresponds to a surface looking like an
upward cup as shown in Fig. 3.7. The lowest point of the cup
is located at the origin.
nonlinear spring
and damper m V = V3
V

V = V2
Fig. 3.6 A nonlinear mass-damper-spring system
V = V1
Total mechanical energy = kinetic energy + potential energy x2

1 2 x 1 1 1 0
(k x + k x )dx = 2 mx& x1
3 2
V ( x) = mx& + k 0 x 2 + k1 x 4
0 1 + V3 > V2 > V1
0 2 2 4
(3.14)
Comparing the definitions of stability and mechanical energy, Fig. 3.7 Typical shape of a positive definite function V ( x1 , x 2 )
we can see some relations between the mechanical energy and
the concepts described earlier: The 2-dimesional geometrical representation can be made as
follows. Taking x1 and x2 as Cartesian coordinates, the level
zero energy corresponds to the equilibrium point
(x = 0, x& = 0) curves V ( x1 , x 2 ) = V typically present a set of ovals
assymptotic stability implies the convergence of surrounding the origin, with each oval corresponding to a
mechanical energy to zero positive value of V .These ovals often called contour curves
instability is related to the growth of mechanical energy may be thought as the section of the cup by horizontal planes,
projected on the ( x1 , x 2 ) plane as shown in Fig. 3.8.
The relations indicate that the value of a scalar quantity, the
mechanical energy, indirectly reflects the magnitude of the
V = V2 x2 V = V1
state vector, and furthermore, that the stability properties of
the system can be characterized by the variation of the
mechanical energy of the system. x1
0
The rate of energy variation during the systems motion is
obtained by differentiating the first equality in (3.14) and
using (3.13) V = V3 V3 > V2 > V1
3
V& (x) = m x& &x& + (k 0 x + k1 x 3 ) x& = x& (b x& x& ) = b x& (3.15) Fig. 3.8 Interpreting positive definite functions using contour
curves
(3.15) implies that the energy of the system, starting from
some initial value, is continuously dissipated by the damper Definition 3.8 If, in a ball B R0 , the function V (x) is positive
until the mass is settled down, i.e., x& = 0 . definite and has continuous partial derivatives, and if its time
derivative along any state trajectory of system (3.2) is
The direct method of Lyapunov is based on generalization of
negative semi-definite, i.e., V& (x) 0 then, V (x) is said to be a
the concepts in the above mass-spring-damper system to more
complex systems. Lyapunov function for the system (3.2).

___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 10
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

A Lyapunov function can be given simple geometrical Obviously, this function is locally positive definite. As a mater
interpretations. In Fig. 3.9, the point denoting the value of of fact, this function represents the total energy of the
V ( x1 , x2 ) is seen always point down an inverted cup. In Fig. pendulum, composed of the sum of the potential energy and
3.10, the state point is seen to move across contour curves the kinetic energy. Its time derivative yields
corresponding to lower and lower value of V .
V& (x) = & sin + &&& = & 2 0

V Therefore, by involving the above theorem, we can conclude

that the origin is a stable equilibrium point. In fact, using
physical meaning, we can see the reason why V& (x) 0 ,
namely that the damping term absorbs energy. Actually,
V
V& (x) is precisely the power dissipated in the pendulum.
However, with this Lyapunov function, we cannot draw
0 x2 conclusion on the asymptotic stability of the system,
because V& (x) is only negative semi-definite.
x1 __________________________________________________________________________________________
x(t )
Example 3.8 Asymptotic stability_______________________
Fig. 3.9 Illustrating Definition 3.8 for n=2
Let us study the stability of the nonlinear system defined by

V = V2 x2 V = V1 x&1 = x1 ( x12 + x 22 2) 4 x1 x 22
x& 2 = 4 x12 x2 + x 2 ( x12 + x22 2)
x1
0 around its equilibrium point at the origin.

V ( x1 , x 2 ) = x12 + x22
V = V3 V3 > V2 > V1

Fig. 3.10 Illustrating Definition 3.8 for n=2 using contour its derivative V& along any system trajectory is
curves
V& = 2( x12 + x 22 )( x12 + x 22 2)
3.4.2 Equilibrium point theorems
Lyapunovs theorem for local stability Thus, is locally negative definite in the 2-dimensional ball B 2 ,
Theorem 3.2 (Local stability) If, in a ball B R0 , there exists a i.e., in the region defined by ( x12 + x22 ) < 2 . Therefore, the
scalar function V (x) with continuous first partial derivatives above theorem indicates that the origin is asymptotically
such that stable.
__________________________________________________________________________________________

V (x) is positive definite (locally in B R0 )

Lyapunov theorem for global stability
V& (x) is negative semi-definite (locally in B R0 )
Theorem 3.3 (Global Stability) Assume that there exists a
then the equilibrium point 0 is stable. If, actually, the scalar function V of the state x, with continuous first order
derivative V& (x) is locally negative definite in B R0 , then the derivatives such that
stability is asymptotic. V (x) is positive definite
V& (x) is negative definite
In applying the above theorem for analysis of a nonlinear V (x) as x
system, we must go through two steps: choosing a positive
Lyapunov function, and then determining its derivative along then the equilibrium at the origin is globally asymptotically
the path of the nonlinear systems. stable.

Example 3.7 Local stability___________________________ Example 3.9 A class of first-order systems_______________

A simple pendulum with viscous damping is described as Consider the nonlinear system

x& + c( x) = 0
&& + & + sin = 0

Consider the following scalar function where c is any continuous function of the same sign as its
scalar argument x , i.e., such as x c( x) > 0 x 0 . Intuitively,
1 this condition indicates that c(x ) pushes the system back
V (x) = (1 cos ) + & 2
2 towards its rest position x = 0 , but is otherwise arbitrary.
___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 11
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Since c is continuous, it also implies that c(0) = 0 (Fig. 3.13). Local invariant set theorem
Consider as the Lyapunov function candidate the square of The invariant set theorem reflect the intuition that the decrease
distance to the origin V = x 2 . The function V is radially of a Lyapunov function V has to graduate vanish (i.e., ) V&
has to converge to zero) because V is lower bounded. A
unbounded, since it tends to infinity as x . Its derivative
precise statement of this result is as follows.
is V& = 2 x x& = 2 x c ( x) . Thus V& < 0 as long as x 0 , so
that x = 0 is a globally asymptotically stable equilibrium point. Theorem 3.4 (Local Invariant Set Theorem) Consider an
autonomous system of the form (3.2), with f continuous, and
let V (x) be a scalar function with continuous first partial
c (x )
derivatives. Assume that
for some l > 0 , the region l defined by V (x) < l is
bounded
0 x V& (x) 0 for all x in l
Let R be the set of all points within where V& (x) = 0 , and l
M be the largest invariant set in R. Then, every solution x(t )
originating in l tends to M as t .
Fig. 3.13 The function c(x )
Note that:
For instance, the system x& = sin 2 x x is globally convergent - M is the union of all invariant sets (e.g., equilibrium
to x = 0 , since for x 0 , sin 2 x sin x x . Similarly, the points or limit cycles) within R
- In particular, if the set R is itself invariant (i.e., if once
system x& = x 3 is globally asymptotically convergent to x = 0 .
V& = 0 , then 0 for all future time), then M=R
Notice that while this systems linear approximation ( x& 0) is
inconclusive, even about local stability, the actual nonlinear The geometrical meaning of the theorem is illustrated in Fig.
system enjoys a strong stability property (global asymptotic 3.14, where a trajectory starting from within the bounded
stability). region l is seen to converge to the largest invariant set M.
__________________________________________________________________________________________
Note that the set R is not necessarily connected, nor is the set
Example 3 .10______________________________________ M.
The asymptotic stability result in the local Lyapunov theorem
Consider the nonlinear system can be viewed a special case of the above invariant set
theorem, where the set M consists only of the origin.
x&1 = x2 x1 ( x12 + x22 )
x& 2 = x1 x 2 ( x12 + x22 ) V =l
V
The origin of the state-space is an equilibrium point for this
l
system. Let V be the positive definite function V = x12 + x22 .
Its derivative along any system trajectory is V& = 2( x12 + x 22 ) 2 R
which is negative definite. Therefore, the origin is a globally
asymptotically stable equilibrium point. Note that the
globalness of this stability result also implies that the origin is M
the only equilibrium point of the system. x0 x2
__________________________________________________________________________________________
x1
Note that:
Fig. 3.14 Convergence to the largest invariant set M
- Many Lyapunov function may exist for the same system.
- For a given system, specific choices of Lyapunov
Let us illustrate applications of the invariant set theorem using
functions may yield more precise results than others.
some examples.
- Along the same line, the theorems in Lyapunov analysis
are all sufficiency theorems. If for a particular choice of
Example 3 .11______________________________________
Lyapunov function candidate V , the condition on V& are
not met, we cannot draw any conclusions on the stability Asymptotic stability of the mass-damper-spring system
or instability of the system the only conclusion we
For the system (3.13), we can only draw conclusion of
should draw is that a different Lyapunov function
marginal stability using the energy function (3.14) in the local
candidate should be tried.
equilibrium point theorem, because V& is only negative semi-
3.4.3 Invariant set theorem definite according to (3.15). Using the invariant set theorem,
Definition 3.9 A set G is an invariant set for a dynamic system however, we can show that the system is actually
if every system trajectory which starts from a point in G asymptotically stable. TO do this, we only have to show that
remains in G for all future time. the set M contains only one point.

___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 12
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

The set R defined by x& = 0 , i.e., the collection of states with x2

zero velocity, or the whole horizontal axis in the phase
plane ( x, x& ) . Let us show that the largest invariant set M in limit cycle
this set R contains only the origin. Assume that M contains a
point with a non-zero position x1 , then, the acceleration at that
x1
point is &x& = (k 0 / m) x (k1 / m) x 3 0 . This implies that the 0
trajectory will immediately move out of the set R and thus
also out of the set M, a contradiction to the definition.
__________________________________________________________________________________________

Example 3 .12 Domain of attraction____________________

Consider again the system in Example 3.8. For l = 1 , the Fig. 3.15 Convergence to a limit circle
region l , defined by V ( x1 , x 2 ) = x12 + x22 < 1 , is bounded.
Moreover, the equilibrium point at the origin can actually be
The set R is simply the origin 0, which is an invariant set shown to be unstable. Any state trajectory starting from the
(since it is an equilibrium point). All the conditions of the region within the limit cycle, excluding the origin, actually
local invariant set theorem are satisfied and, therefore, any converges to the limit cycle.
trajectory starting within the circle converges to the origin. __________________________________________________________________________________________
Thus, a domain of attraction is explicitly determined by the
invariant set theorem. Example 3.11 actually represents a very common application
__________________________________________________________________________________________
of the invariant set theorem: conclude asymptotic stability of
Example 3 .13 Attractive limit cycle_____________________ an equilibrium point for systems with negative semi-definite V& .
The following corollary of the invariant set theorem is more
Consider again the system specifically tailored to such applications.

x&1 = x2 x17 ( x14 + 2 x 22 10) Corollary: Consider the autonomous system (3.2), with f
continuous, and let V (x) be a scalar function with continuous
x& 2 = x13 3 x 25 ( x14 + 2 x 22 10)
partial derivatives. Assume that in a certain neighborhood
of the origin
Note that the set defined by x14 + 2 x22 = 10 is invariant, since is locally positive definite
V& (x) is negative semi-definite
d 4
( x1 + 2 x 22 10) = (4 x110 + 12 x 26 )( x14 + 2 x22 10) the set R defined by V& (x) = 0 contains no trajectories of
dt (3.2) other than the trivial trajectory x 0
Then, the equilibrium point 0 is asymptotically stable.
which is zero on the set. The motion on this invariant set is Furthermore, the largest connected region of the form
described (equivalently) by either of the equations (defined by V (x) < l ) within is a domain of attraction of the
equilibrium point.
x&1 = x 2
x& 2 = x13 Indeed, the largest invariant set M in R then contains only the
equilibrium point 0.
Therefore, we see that the invariant set actually represents a
Note that:
limit circle, along which the state vector moves clockwise. Is
- The above corollary replaces the negative definiteness
this limit circle actually attractive ? Let us define a Luapunov
condition on V& in Lyapunovs local asymptotic stability
function candidate V = ( x14 + 2 x22 10) 2 which represents a
theorem by a negative semi-definiteness condition on V& ,
measure of the distance to the limit circle. For any arbitrary
combined with a third condition on the trajectories within
positive number l , the region l , which surrounds the limit
R.
circle, is bounded. Its derivative - The largest connected region of the form l within is a
domain of attraction of the equilibrium point, but not
V& = 8( x110 + 3x 26 )( x14 + 2 x22 10) 2 necessarily the whole domain of attraction, because the
function V is not unique.
Thus V& is strictly negative, except if x14 + 2 x 22 = 10 or - The set itself is not necessarily a domain of attraction.
Actually, the above theorem does not guarantee that is
x110 + 3 x 26 = 0 , in which cases V& = 0 . The first equation is invariant: some trajectories starting in but outside of
simply that defining the limit cycle, while the second equation the largest l may actually end up outside .
is verified only at the origin. Since both the limit circle and the
origin are invariant sets, the set M simply consists of their Global invariant set theorem
union. The above invariant set theorem and its corollary can be
simply extended to a global result, by enlarging the involved
Thus, all system trajectories starting in l converge either to region to be the whole space and requiring the radial
the limit cycle or the origin (Fig. 3.15) unboundedness of the scalar function V .
___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 13
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Theorem 3.5 (Global Invariant Set Theorem) Consider an as long as x 0 . Thus the system cannot get stuck at an
autonomous system of the form (3.2), with f continuous, and equilibrium value other than x = 0 ; in other words, with R
let V (x) be a scalar function with continuous first partial being the set defined by x& = 0 , the largest invariant set M in R
derivatives. Assume that contains only one point, namely [ x = 0, x& = 0] . Use of the
V& (x) 0 over the whole state space local invariant set theorem indicates that the origin is a locally
asymptotically stable point.
V (x) as x
Let R be the set of all points where V& (x) = 0 , and M be the b(x& ) c (x )
largest invariant set in R. Then all solutions globally
asymptotically converge to M as t

For instance, the above theorem shows that the limit cycle
0 x& 0 x
convergence in Example 3.13 is actually global: all system
trajectories converge to the limit cycle (unless they start
exactly at the origin, which is an unstable equilibrium point).
Fig. 3.17 The functions b(x& ) and c(x )
Because of the importance of this theorem, let us present an
additional (and very useful) example.
x

Example 3 .14 A class of second-order nonlinear systems___

Furthermore, if the integral c(r )dr is unbounded as x ,

0
then V is a radially unbounded function and the equilibrium
Consider a second-order system of the form point at the origin is globally asymptotically stable, according
to the global invariant set theorem.
&x& + b( x& ) + c( x) = 0 __________________________________________________________________________________________

where b and c are continuous functions verifying the sign Example 3 .15 Multimodal Lyapunov Function___________
conditions x& b( x& ) > 0 for x& 0 and x c( x& ) > 0 for x 0 . The Consider the system
dynamics of a mass-damper-spring system with nonlinear
damper and spring can be described by the equation of this x
&x& + x 2 1 x& 3 + x = sin
form, with the above sign conditions simply indicating that the 2
otherwise arbitrary function b and c actually present y 1 2 x
damping and spring effects. A nonlinear R-L-C (resistor-
inductor-capacitor) electrical circuit can also be represented by
Chose the Lyapunov function V =
2
dy .
2
x& +
y sin
0

the above dynamic equation (Fig. 3.16) This function has two minima, at x = 1, x& = 0 , and a local
maximum in x (a saddle point in the state-space) at
vC = c (x ) v L = &x& v R = b(x& ) x = 0, x& = 0 . Its derivative V& = x 2 1 x& 4 , i.e., the virtual

power dissipated by the system. Now V& = 0 x& = 0 or

x = 1 . Let us consider each of cases:
x
Fig. 3.16 A nonlinear R-L-C circuit x& = 0 &x& = sin x 0 except if x = 0 or x = 1
2
Note that if the function b and c are actually linear x = 1 &x& = 0
(b( x& ) = 1 x& , c( x) = x ) , the above sign conditions are simply
Thus the invariant set theorem indicates that the system
the necessary and sufficient conditions for the systems
converges globally to or ( x = 1, x& = 0) . The first two of these
stability (since they are equivalent to the conditions
1 > 0, 0 > 0 ). equilibrium points are stable, since they correspond to local
minima of V (note again that linearization is inconclusive
about their stability). By contrast, the equilibrium point
Together with the continuity assumptions, the sign conditions
( x = 0, x& = 0) is unstable, as can be shown from linearization
b and c are simply that b(0) = 0 and c = 0 (Fig. 3.17). A
positive definite function for this system is ( &x& = ( / 2 1) x) , or simply by noticing that because that point
1 2 x is a local maximum of V along the x axis, any small deviation
2 0
V = x& + c( y ) dy , which can be thought of as the sum of in the x direction will drive the trajectory away from it.
__________________________________________________________________________________________
the kinetic and potential energy of the system. Differentiating
V , we obtain Note that: Several Lyapunov function may exist for a given
system and therefore several associated invariant sets may be
V& = x& &x& + c( x) x& = x& b( x& ) x& c( x) + c( x) x& = x& b( x& ) 0 derived.

which can be thought of as representing the power dissipated 3.5 System Analysis Based on Lyapunovs Direct Method
in the system. Furthermore, by hypothesis, x& b( x& ) = 0 only if
How to find a Lyapunov function for a specific problem ?
x& = 0 . Now x& = 0 implies that &x& = c (x) , which is non-zero
There is no general way of finding Lyapunov function for
___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 14
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

nonlinear system. Faced with specific systems, we have to use Lyapunov functions for linear time-invariant systems
experience, intuition, and physical insights to search for an Given a linear system of the form x& = A x , let us consider a
appropriate Lyapunov function.
quadratic Lyapunov function candidate V& = xT P x , where P
In this section, we discuss a number of techniques which can is a given symmetric positive definite matrix. Its derivative
facilitate the otherwise blind of Lyapunov functions. yields

3.5.1 Lyapunov analysis of linear time-invariant systems V& = x& T P x + x T P x& = -xT Q x (3.18)
Symmetric, skew-symmetric, and positive definite matrices where
Definition 3.10 A square matrix M is symmetric if M=MT (in
other words, if i, j M ij = M ji ). A square matrix M is skew- A T P + P A = -Q (3.19)

symmetric if M = M T (i.e., i, j M ij = M ji ). (3.19) is so-called Lyapunov equation. Note that Q may be not
p.d. even for stable systems.
Note that:
Example 3 .17 ______________________________________
- Any square n n matrix can be represented as the sum of
a symmetric and a skew-symmetric matrix. This can be
Consider the second order linear system with A = 0 4 .
shown in the following decomposition 8 12
M + MT M - MT
M= + If we take P = I , then - Q = P A + A T P = 0 4 . The
142 243 142 243 4 24
symmetric skew symmetric matrix Q is not p.d.. Therefore, no conclusion can be draw
- The quadratic function associated with a skew-symmetric from the Lyapunov function on whether the system is stable or
matrix is always zero. Let M be a n n skew-symmetric not.
__________________________________________________________________________________________
matrix and x is an arbitrary n 1 vector. The definition of
skew-symmetric matrix implies that xT M x = xT M T x . A more useful way of studying a given linear system using
T T T quadratic functions is, instead, to derive a p.d. matrix P from a
Since x M x is a scalar, x M x = x M x which yields given p.d. matrix Q, i.e.,
x, x T M x = 0 (3.16)
choose a positive definite matrix Q
In the designing some tracking control systems for robot, solve for P from the Lyapunov equation
this fact is very useful because it can simplify the control check whether P id p.d.
law.
- (3.16) is a necessary and sufficient condition for a matrix
M to be skew-symmetric. If P is p.d., then (1 / 2)x T P x is a Lyapunov function for the
linear system. And the global asymptotical stability is
Definition 3.11 A square matrix M is positive definite (p.d.) if guaranteed.
x 0 xT M x > 0 .
Theorem 3.6 A necessary and sufficient condition for a LTI
Note that: system x& = A x to be strictly stable is that, for any symmetric
- A necessary condition for a square matrix M to be p.d. is p.d. matrix Q, the unique matrix P solution of the Lyapunov
that its diagonal elements be strictly positive. equation (3.19) be symmetric positive definite.
- A necessary and sufficient condition for a symmetric
matrix M to be p.d. is that all its eigenvalues be strictly Example 3 .18 ______________________________________
positive. Consider again the second order linear system in Example
- A p.d. matrix is invertible.
p p
- A .d. matrix M can always be decomposed as 3.18. Let us take Q = I and denote P by P = 11 12 ,
p 21 p 22
M = U T U (3.37)
where due to the symmetry of P, p 21 = p12 . Then the
where U T U = I , is a diagonal matrix containing the
Lyapunov equation is
eigenvalues of M
- There are some following facts
p11 p12 0 4 0 8 p11 p12 1 0
p 21 p 22 8 12 + 4 12 p 21 p 22 = 0 1
2 2
min (M ) x xT Mx max (M ) x
x T Mx = xT U T Ux = z T z where Ux = z
min (M ) I max (M ) I whose solution is p11 = 5 , p12 = p 22 = 1 . The corresponding

zT z = x
2
matrix P = 5 1 is p.d., and therefore the linear system is
1 1
globally asymptotically stable.
The concepts of positive semi-definite, negative definite, and __________________________________________________________________________________________
negative semi-definite can be defined similarly. For instance, a
square n n matrix M is said to be positive semi-definite 3.5.2 Krasovskiis method
(p.s.d.) if x, x T M x 0 . A time-varying matrix M(t) is Krasovskiis method suggests a simplest form of Lyapunov
uniformly positive definite if > 0, t 0, M (t ) I . function candidate for autonomous nonlinear systems of the
___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 15
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

form (3.2), namely, V = f T f . The basic idea of the method is function for this system. If the region is the whole state
simply to check whether this particular choice indeed leads to space, and if in addition, V (x) as x , then the
a Lyapunov function. system is globally asymptotically stable.

Theorem 3.7 (Krasovkii) Consider the autonomous system 3.5.3 The Variable Gradient Method
defined by (3.2), with the equilibrium point of interest being The variable gradient method is a formal approach to
the origin. Let A(x) denote the Jacobian matrix of the system, constructing Lyapunov functions.
i.e.,
f To start with, let us note that a scalar function V (x) is related
A(x ) =
x to its gradient V by the integral relation
If the matrix F = A + A T is negative definite in a
x
neighborhood , then the equilibrium point at the origin is
asymptotically stable. A Lyapunov function for this system is
V ( x ) = V dx
0

V (x ) = f T ( x) f ( x) where V = {V / x1 ,K, V / x n }T . In order to recover a

unique scalar function V from the gradient V , the gradient
If is the entire state space and, in addition, V (x) as function has to satisfy the so-called curl conditions
x , then the equilibrium point is globally
asymptotically stable. Vi V j
= (i, j = 1,2,K, n)
x j xi
Example 3 .19 ______________________________________
Consider the nonlinear system Note that the ith component Vi is simply the directional
derivative V / xi . For instance, in the case n = 2 , the above
x&1 = 6 x1 + 2 x 2 simply means that
x& 2 = 2 x1 6 x2 2 x 23
V1 V2
=
We have x2 x1

f 6 2 12 4 The principle of the variable gradient method is to assume a

A= F = A + AT =
x 2 6 6 x 2 4 12 12 x 2
2 2 specific form for the gradient V , instead of assuming a
specific form for a Lyapunov function V itself. A simple way
The matrix F is easily shown to be negative definite. Therefore, is to assume that the gradient function is of the form
the origin is asymptotically stable. According to the theorem, a
n
Lyapunov function candidate is
Vi = a x
j =1
ij j (3.21)
2
V (x) = (6 x1 + 2 x 2 ) + (2 x1 6 x 2 2 x 23 ) 2

where the aij s are coefficients to be determined. This leads

Since V (x) as x , the equilibrium state at the
to the following procedure for seeking a Lyapunov function V
origin is globally asymptotically stable.
__________________________________________________________________________________________
assume that V is given by (3.21) (or another form)
The applicability of the above theorem is limited in practice, solve for the coefficients aij so as to sastify the curl
because the Jcobians of many systems do not satisfy the
equations
negative definiteness requirement. In addition, for systems of
higher order, it is difficult to check the negative definiteness of assume restrict the coefficients in (3.21) so that V& is
the matrix F for all x. negative semi-definite (at least locally)
compute V from V by integration
Theorem 3.7 (Generalized Krasovkii Theorem) Consider check whether V is positive definite
the autonomous system defined by (3.2), with the equilibrium
point of interest being the origin, and let A(x) denote the Since satisfaction of the curl conditions implies that the above
Jacobian matrix of the system. Then a sufficient condition for integration result is independent of the integration path, it is
the origin to be asymptotically stable is that there exist two usually convenient to obtain V by integrating along a path
symmetric positive definite matrices P and Q, such that which is parallel to each axis in turn, i.e.,
x 0 , the matrix
x1 x2
F (x) = A T P + PA + Q V ( x) =
0
V1 ( x1 ,0,K,0) dx1 +
0
V2 ( x1 ,0,K,0) dx2 + K +
xn
is negative semi-definite in some neighborhood of the 0
Vn ( x1 ,0,K,0) dx n
origin. The function V (x) = f T (x) f (x) is then a Lyapunov
___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 16
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Example 3 .20 ______________________________________ Estimating convergence rates for linear system

Let denote the largest eigenvalue of the matrix P by max (P ) ,
Let us use the variable gradient method top find a Lyapunov
function for the nonlinear system the smallest eigenvalue of the matrix Q by min (Q) , and their
ratio max (P) / min (Q) by . The p.d. of P and Q implies
x&1 = 2x1 that these scalars are all strictly positive. Since matrix theory
x& 2 = 2 x 2 + 2 x1 x22 shows that P max (P ) I and min (Q) I Q , we have
min (Q) T
xT Q x x [ max (P ) I ] x V
We assume that the gradient of the undetermined Lyapunov max (P )
function has the following form
This and (3.18) implies that V& V .This, according to
V1 = a11 x1 + a12 x 2 lemma, means that x T Q x V (0) e t . This together with the
V2 = a 21 x1 + a 22 x2 2
fact xT P x min (P) x(t ) , implies that the state x
The curl equation is converges to the origin with a rate of at least / 2 .

V1 V2 a12 a The convergence rate estimate is largest for Q = I . Indeed, let

= a12 + x 2 = a 21 + x1 21
x2 x1 x 2 x1 P0 be the solution of the Lyapunov equation corresponding to
Q = I is
If the coefficients are chosen to be a11 = a 22 = 1, a12 = a 21 = 0
A T P0 + P0 A = I
which leads to V = x , V = x then V& can be computed
1 1 2 2 and let P the solution corresponding to some other choice of
as Q
x1 x2 x12 + x 22 A T P + PA = Q1
V ( x) =
0
x1 dx1 +
0
x2 dx 2 =
2
(3.22)
Without loss of generality, we can assume that min (Q1 ) = 1
since rescaling Q1 will rescale P by the same factor, and
This is indeed p.d., and therefore, the asymptotic stability is
guaranteed. therefore will not affect the value of the corresponding .
Subtract the above two equations yields
If the coefficients are chosen to be a11 = 1, a12 = x 22 , A T (P - P0 ) + (P - P0 ) A = (Q1 - I )
a 21 = 3 x22 , a 22 = 3 , we obtain the p.d. function Now since min (Q1 ) = 1 = max (I ) , the matrix (Q1 - I) is
positive semi-definite, and hence the above equation implies
x12 3 2 that (P - P0 ) is positive semi-definite. Therefore
V ( x) = + x2 + x1 x 23 (3.23)
2 2
max (P ) max (P0 )
Since min (Q1 ) = 1 = min (I ) , the convergence rate estimate
whose derivative is V& = 2 x12 6 x 22 2 x 22 ( x1 x 2 3 x12 x22 ) .
= min (Q) / max (P )
corresponding to Q = I the larger than (or equal to) that
We can verify that V& is a locally negative definite function
(noting that the quadratic terms are dominant near the origin), corresponding to Q = Q1 .
and therefore, (3.23) represents another Lyapunov function for
the system. Estimating convergence rates for nonlinear systems
__________________________________________________________________________________________
The estimation convergence rate for nonlinear systems also
3.5.4 Physically motivated Lyapunov functions involves manipulating the expression of V& so as to obtain an
explicit estimate of V . The difference lies in that, for
3.5.5 Performance analysis nonlinear systems, V and V& are not necessarily quadratic
function of the states.
Lyapunov analysis can be used to determine the convergence
rates of linear and nonlinear systems. Example 3 .22 ______________________________________

A simple convergence lemma Consider again the system in Example 3.8

Lemma: If a real function W (t ) satisfies the inequality
x&1 = x1 ( x12 + x 22 2) 4 x1 x 22
W& (t ) + W (t ) 0 (3.26) x& 2 = 4 x12 x2 + x 2 ( x12 + x22 2)

where is a real number. Then W (t ) W (0) e t Choose the Lyapunov function candidate V = x
2
, its
dV
The above Lemma implies that, if W is a non-negative derivative is V& = 2V (V 1) . That is = 2dt . The
function, the satisfaction of (3.26) guarantees the exponential V (1 V )
convergence of W to zero. solution of this equation is easily found to be

___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 17
 Applied Nonlinear Control Nguyen Tan Tien - 2002.3
_________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

e 2dt V ( 0)
V (x) = , where = .
1 + e 2dt 1 V (0)

2
If x(0) = V (0) < 1 , i.e., if the trajectory starts inside the
unit circle, then > 0 , and V (t ) < e 2t . This implies that
the norm x(t ) of the state vector converges to zero
exponentially, with a rate of 1.
However, if the trajectory starts outside the unit circle, i.e., if
V (0) > 1 , then < 0 , so that V (t ) and therefore x tend to
infinity in a finite time (the system is said to exhibit finite
escape time, or explosion).
__________________________________________________________________________________________

3.6 Control Design Based on Lyapunovs Direct Method

There are basically two ways of using Lyapunovs direct

method for control design, and both have a trial and error
flavor:
Hypothesize one form of control law and then finding a
Lyapunov function to justify the choice
Hypothesize a Lyapunov function candidate and then
finding a control law to make this candidate a real
Lyapunov function

Example 3 .23 Regulator design_______________________

Consider the problem of stabilizing the system &x& x& 3 + x 2 = u .

__________________________________________________________________________________________

___________________________________________________________________________________________________________
Chapter 3 Fundamentals of Lyapunov Theory 18