Sign Definiteness of Scalar Functions

• Consider V ( x1 , x 2 ,  , x n ) be a scalar function of

state variables x1 , x 2 ,  , x n , Then following
definition are useful for the Lyapunov’s method:

• Definition 1: Positive definiteness: A scalar function

V ( x ) is said to be positive definite if V ( x ) is such
that V ( x ) is positive at all points in the state space
except at the origin, x  0 where V ( x ) is equal to
zero.
• Definition 2: Positive semi-definiteness:
A scalar functionV ( x ) is said to be positive semi-
definite if V ( x ) is such that V ( x ) is positive at all
points in the state space except at one or more
points in the state space including the origin
where it is equal to zero.
 Positive definite function will be zero only at the
origin.
 The positive semi-definite function can be zero at
several points in the state space with function
remaining positive at all other places.
• Definition 3: Negative definiteness:
A scalar function V ( x ) is said to negative
definite if V ( x ) is such that V ( x ) is negative at all
points in the state space except at the origin, x  0
where V ( x ) is equal to zero.
• Definition 4: Negative semi-definiteness:
A scalar functionV ( x ) is said to be negative semi-
definite if V ( x ) is such that V ( x ) is negative at all
points in the state space except at one or more
points in the state space including the origin
where it is equal to zero.
• Definition 5:Indefiniteness:
A scalar function V ( x ) is said to indefinite if
V ( x ) is such that V ( x ) is positive, negative or zero
anywhere in the state space.
 Example:
Positive V(0,0)=0 only at origin
V ( x )  x12  x 22 definite and V(x) > 0 at all
other points.
Positive Positive definite in
definite in region where
V ( x )  x1  x 2 particular
2 2

 x 3 3  x 3 
2 2 region x3  3
 It can be zero
Positive whenever
V ( x )   x1  x 2 
2
semi x1   x 2
x 2 definite
3
and x 3  0
Negative definite
Negative in two dimension
V ( x )    x1  x 2  definite.
2 2 and negative
semi-definite in
three dimensions
V(x) can be poitive
Sign negative or zero at
 x1 x2  x2 x1 
2 Indefinite anywhere in state
V ( x)    space.
  5x 2 x 
 1 2 
 Sign Definiteness of Quadratic form
• For a given any scalar function V(x) there is no
general method by which one can find whether it is
positive, negative or semi-definite
• However, if the function is in quadratic forms, one
can use Sylvester Theorem.
• V(x) is in the quadratic form if V(x) can be in
following format:

  k xi x j 
n n
V ( x)  ij
i 1 j 1

where k ij are real constants

• Quadratic forms of V(x) can also written in matrix
form as:
V ( x)  xT Q x
• Where Q is given as:

 q11 q12  q1 n 
q q 22  q2n 
Q   21

    
 
 q n1 qn2  q nn 
• Where elements of Q matrix is given by:
q ii  k ij whenever i  j

  k ij  k  q
1
q ij ji ji whenever i  j
2
• Sylvester theorem :
V ( x)  x Q x
A quadratic function T is
positive definite if only if all the principal
determinants of the matrix Q are positive.
 Lyapunov Second Method of Stability
Concepts:
 If an autonomous system has its total energy
decreasing indefinitely, it should ultimately move to
an equilibrium point.
 If the total energy in a system continuously
increases, some time or other, such a system will
have to be separated from other parts of system,
 As no physical device can feed infinite amount
energy to system.
• Consider a mechanical system with a unit mass,
friction and spring coefficients as B and K
respectively. . The state equation is given as:
x1  x2
.
x 2   Kx 1  Bx 2
• At any instant, total energy in the system consists of
kinetic energy in the mass and the potential energy
stored in the spring.
1 2 1
V ( x1 , x 2 )  x 2  Kx 12
2 2
• The V(x) > 0 for all x except when x = 0 when V(0) =0
• Total energy is positive unless the system is at rest
at equilibrium point xe = 0 where energy is zero.
• Rate of change of energy :
d  V dx 1  V dx 2
V ( x1 , x 2 )  
dt  x1 dt  x 2 dt
1 . 1 .
  2 Kx 1  x1   2 x 2  x 2
2 2
. .
  Kx 1  x1   x 2  x 2
x2

 Kx 1 x 2  x 2   Kx 1  Bx 2 
  Bx 2
2

• Thus, rate of change of energy is negative at all

points except where x  0 at which d
2 V ( x)  0
dt
• Under the positive damping ( B > 0 ), the energy of
the system cannot increase.
• From
.
the state equations, it can seen that:
x 2   K x1 at the points where x 2  0
• Thus, system cannot remain in the non-equilibrium
state for which x 2  0 .
• Therefore, energy cannot remain constant except at
the equilibrium point where it will be zero.
• The example is illustrated in the figure below:
• V-function plots into a different ellipses forming a
cup-shape in x1-x2-V space.
• These ellipses can be considered as loci of a
combination of points (x1, x2) such they will have
same total energy, given by V(x1,x2) = k.
• Constant V-loci are on surface of x1-x2 plane.
• If one draw the energy trajectory on the cup as
shown in figure, representative point x(t) crosses
the constant V constant and moves towards the
lowest point of the cup which is equilibrium point,
 V ( x1 , x 2 )
k 3  k 2  k1
V  k3
t 3  t 2  t1
V  k2

V  k1
x1

x2

t3
t2 x1
0 0
t1 (x , x )
1 2
• If V(x) is a positive definite function, then state x
which satisfy V(x) = k (Where k is a positive
constant), lie on a closed hyper surface in the n-
dimensional state space at least in the
neighborhood of origin.

V ( x )   as x  
 If
then such as closed surfaces extend over the entire
state space.
V ( x )  k1
 Hyper-surface V ( x ) lies if k 2 inside
k 2 entirely k 1 the hyper-
surface
• for a given system, If there exist a positive function
V(x) can be found such that its time derivative taken
along the energy trajectory is negative,
• Then as time increase, V(x) takes smaller values of “ k
” and finally the V(x) shrinks to zero.
• Therefore, the operating points x also shrinks to origin.
• This implies the asymptotic stability of the origin of the
state space.
• Lyapunov second method provides a sufficient
condition for asymptotic stability.
• It is a generalization of the concept of the total energy
decreasing to zero for a stable system.
• For some system, it is easy to define an energy
function V(x) in terms of state variables.
• but it is not case with many system as the total
energy may not be easily be expressed as function
the state variables.
• Basic idea of second method of Lyapunov is to
develop a function which has all the properties of
energy functions
• Total energy is a positive definite function.
• Concept of second method of Lyapunov is extended
to systems with non-energy based dynamics.
 Lyapunov’s Stability Theorems
Lyapunov’s First theorem: .
• For a system described by : x  f ( x , t ),
where f (0, t )  0 for all t  t 0 , if there exits a scalar
function V(x, t) having continuous first partial derivatives
and satisfying the following conditions:
1) V(x, t) is positive definite.
2) Derivative of V(x, t) with respect to time is negative
definite.
Then, the equilibrium state at origin is uniformly
asymptotically stable.
V ( x )   as x  
and if , them equilibrium state
at origin is uniformly asymptotically stable in the large
Lyapunov’s Second theorem: .
• For a system described by : x  f ( x , t ) ,
where f (0, t )  0 for all t  t 0 , if there exits a
scalar function V(x, t) having continuous first partial
derivatives and satisfying the following conditions:
1) V(x, t) is positive definite.
2) Derivative of V(x, t) with respect to time is negative
semi-definite.
3) Derivative oft  t 0 t) with respect
V(x, t0 to time vanish
identically in for any , along the state
trajectory.
Then, the equilibrium state at origin is uniformly
asymptotically stable in the large.
• Example:
• Check the stability of the equilibrium state of system
described by:.

x 1  x 2  x1 x  x
2
1
2
2 
 
.
x 2   x1  x 2 x  x2
1
2
2

• Solution: Equilibrium point can be obtained by

equating the derivatives to zero.
 0
.
x 1  x 2  x1 x  x
2
1
2
2

x  x   0
.
x 2   x1  x 2 2
1
2
2

• This gives x 2  x1  0 , origin is the only equilibrium

point.
• Taking a positive definite scalar function :
V ( x)  x  x
2
1
2
2
• The time derivative of V(x) can be derived as
 
.
V ( x )   2 x12  x 22
• which is negative definite.
• therefore as per Lyapunov’s first theorem, the
equilibrium point is asymptotically stable.
• as V ( x )   as x , equilibrium
 state at origin
is asymptotically stable in the large
• V(x) is a measure of distance of the state x from the
. state space.
origin of the
V ( x)  0
• with , this distance continuously decreases,
as t increase, then x(t) will tend to origin.
• In the second theorem of Lyapunov, if the
derivative of V(x, t) is only negative semi-definite ,
then it should be possible to prove that it vanish
identically along the state trajectory.
• That is , one has to proof that even though
derivative of V(x, t) is negative semi-definite; as
per the system equations, it can become zero only
at origin of the state space.
Example-2:
• Check the stability of the equilibrium state of
.
system described by: x  x
1 2
.
x 2   x1  3 x 2
• Taking a positive definite scalar function :
V ( x)  x  x
2
1
2
2
• The time derivative of V(x) can be derived as
d V . V .
V ( x)  x1  x2  6 x2
2

dt  x1 x2
which is only negative semi-definite.
• From state. equations, one see that when x 2 , 0
x1  0
.
x 2   x1  x1  0
• Time derivative of V(x,t) vanishes only when both x1
and x2 are zero, that is only at the origin.
• Therefore equilibrium state is asymptotically stable as
per Lyapunov second theorem
Lyapunov’s Third theorem: (Instability
.
Theorem)
• For a system described by : x  f ( x , t ) ,
where f (0, t )  0 for all t  t 0 , if there exits a
scalar function V(x, t) having continuous first partial
derivatives and satisfying the following conditions:
1) V(x, t) is positive definite in some region about
origin.
2) Derivative of V(x, t) with respect to time is positive-
definite in the same region

Then, the equilibrium state at origin is unstable.

Procedure For Finding Stability by Second Method
of Lyapunov

1) Assume a positive definite function, V(x, t).

2) Determine the derivative of V( x, t) as:
d V  dx1  V  dx2  V  dxn 
V ( x)         
dt x1  dt  x2  dt  xn  dt 
3) Evaluate the derivative along the trajectory by
. . .
substituting the x , x ,  , x from the state
1 2 n
equation.
4) Check whether the derivative is negative definite in
the region where V(x) is positive definite.
5) If not, check whether the derivative is at least
negative semi-definite.

6) if the derivative is negative semi-definite, verify

whither the derivative vanishes identically only at
the origin by substituting the condition for the
derivative to vanish in the state equations.

7) Then if the derivative vanishes only at the origin or

the derivative is negative definite conclude about
the stability using with second theorem in first
case or first theorem in the second case.