EACT631 - A DAPTIVE C ONTROL S YSTEM

C HAPTER 2 :: A DVANCED TOOLS FOR STABILITY OF NON - AUTONOMOUS

NONLINEAR SYSTEMS

Dr. Saravanakumar Gurusamy

Associate Professor,
Department of Electrical and Electronics Technology,
FDRE Technical and Vocational Training Institute,
Addis Ababa, Ethiopia

December 22, 2023

 C ONTENTS IN THE C HAPTER - I

Contents
▶ Definitions,
▶ Adaptive backstepping,
▶ Converse Lyapunov theorems,
▶ LaSalle/Yoshizawa theorem,
▶ Passivity theory,
▶ Zero-state detectability,
▶ Positive real and strictly positive real transfer functions,
▶ Kalman-Yakubovich-Popov lemma

D R .S ARAVANAKUMAR G URUSAMY EACT631 1 / 91

 C HAPTER 3 - A DVANCED TOOLS FOR STABILITY OF
NON - AUTONOMOUS NONLINEAR SYSTEMS

1 Lyapunov Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Lyapunov Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3 Lyapunov’s Linearization Method . . . . . . . . . . . . . . . . . . . 20
4 Sylvester’s Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5 Lyapunov’s direct Method . . . . . . . . . . . . . . . . . . . . . . . . 39
6 Lyapunov Stability of Linear System . . . . . . . . . . . . . . . . . . 52

7 La Salle’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
8 Variable Gradient Method for Lyapunov Functions . . . . . . . . . . 65

9 Krasovski’s method . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
10 Passivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

D R .S ARAVANAKUMAR G URUSAMY EACT631 2 / 91

 LYAPUNOV S TABILITY I
I NTRODUCTION

▶ Stability theory is categorized into three parts:

1. Stability of equilibrium points
2. Stability of periodic orbits
3. Input/output stability
▶ Importance of Stability
• Stability is a fundamental property in control systems.
• A system is deemed stable if it maintains its desired operating
point despite perturbations.
• Unstable systems can be useless or even dangerous
▶ Understanding Stability
• Stability is qualitatively described as a system’s ability to remain
around an equilibrium point after being disturbed.
• Motion around equilibrium points (e.g., vertical up and down)
illustrate stable and unstable behaviors of dynamic systems.
▶ Historical Context

D R .S ARAVANAKUMAR G URUSAMY EACT631 3 / 91

 LYAPUNOV S TABILITY II
I NTRODUCTION

• Lyapunov’s work initially received limited attention outside

Russia.
• The publication of Lur’e and La Salle-Lefschetz brought
Lyapunov’s work to the control engineering community’s
attention.
• Refinements of Lyapunov’s methods have since evolved.
▶ Lyapunov’s work ”The General Problem of Motion Stability published
in 1892 includes two methods:
• Linearization Method: studies nonlinear local stability around an
Equ. point from stability properties of its linear approximation
• Direct Method: not restricted to local motion. Stability of nonlinear
system is studied by proposing a scalar energy-likefunction for the
system and examining its time variation
▶ Lyapunov 1st Method
• Consists of all procedure in which the explicit form of the solutions
of the differential equations are used for the analysis
D R .S ARAVANAKUMAR G URUSAMY EACT631 4 / 91
 LYAPUNOV S TABILITY III
I NTRODUCTION

▶ Lyapunov 2nd Method

• Does not require solutions of the differential equations.
• Determine the stability of the system without solving the state
equation.
• Bcoz, Solving non-linear and /or time varying state equation is very
difficult.

D R .S ARAVANAKUMAR G URUSAMY EACT631 5 / 91

 I SOLATED E QUILIBRIUM

▶ An equilibrium point xe in Rn is an isolated equilibrium point if there is

an r > 0 such that the r−neighborhood of xe contains no equilibrium
points other than xe
▶ Linear System
▶ Nonlinear System

D R .S ARAVANAKUMAR G URUSAMY EACT631 6 / 91

 N EIGHBORHOOD OF A POINT xe

The r-neighborhood of a point xe in Rn refers to a set of points around xe that

satisfy a specific condition regarding their distance from xe .
Mathematically, the r-neighborhood of xe can be defined as follows:

Br (xe ) = {x ∈ Rn | ||x − xe || ≤ r}

where:
▶ - Br (xe ) denotes the r-neighborhood of xe .
▶ - x represents a point in Rn .
▶ - || · || signifies any norm in Rn , for instance, the p-norm.
The distance ||x − xe || can be calculated using different norms such as the
p-norm, where p can take various values (Euclidean norm, Manhattan
norm, etc.). The norm chosen defines how the distance between x and xe is
measured.

D R .S ARAVANAKUMAR G URUSAMY EACT631 7 / 91

 S TABILITY IN THE SENSE OF LYAPUNOV I

▶ Consider a system defined by the state equation

ẋ = f (x, t)

Assume this system have unique solution at the given initial condition.
▶ Let us consider this solution as F(t : x0 , t0 ) Where x = x0 at t = t0 and
t → Observed time.
▶ Let xe one of the state for system with equation ẋ = f (x, t) In such a way
that, f (xe , t) = 0 for all t. Then this xe is called equilibrium state.
▶ For linear & Time invariant system If State matrix A is Non-Singular -¿
only one equilibrium state. If State matrix A is Singular -¿ more than
one equilibrium state.
▶ For Non-linear system, There are more than one equilibrium states.
These equilibrium states can be obtained from the solution of equation

f (xe , t) = 0

D R .S ARAVANAKUMAR G URUSAMY EACT631 8 / 91

 S TABILITY IN THE SENSE OF LYAPUNOV II
▶ Now we Consider the stability analysis of equilibrium states
at the origin. Consider a spherical region of radius R about
an equilibrium state xe such that

∥x − xe ∥ ≤ R

Where ∥x − xe ∥ is called Euclidean norm

∥x − xe ∥ = (x1 − x1e )2 + (x2 − x2e )2 . . . (x1 − xne )2

▶ S(δ) consists of all points such that ∥x − xe ∥ ≤ δ for all t ≥ t0 .

D R .S ARAVANAKUMAR G URUSAMY EACT631 9 / 91

 A SYMPTOTIC S TABILITY
▶ An equilibrium state xe is said to be a asymptotically stable if it is stable
in the sense of Lyapunov and every solution starting within s(δ)
converges without leaving s(ϵ) to xe as t increases indefinitely.
▶ It is a local stability.
▶ Establishing asymptotic stability does not indicate the proper operation
of the system.

D R .S ARAVANAKUMAR G URUSAMY EACT631 10 / 91

 A SYMPTOTIC S TABILITY
▶ An equilibrium state xe is said to be a asymptotically stable if it is stable
in the sense of Lyapunov and every solution starting within s(δ)
converges without leaving s(ϵ) to xe as t increases indefinitely.
▶ It is a local stability.
▶ Establishing asymptotic stability does not indicate the proper operation
of the system.
Asymptotic Stability in Large
▶ If the asymptotic stability holds for all states (all points in state space)
from which trajectories originate, then the equilibrium state is said to
be asymptotically stable in the large.
▶ [ Or ]
▶ The equilibrium state xe of the system is said to be asymptotically stable
in the large if it is stable and if every solution converges to xe as t
increases indefinitely.

D R .S ARAVANAKUMAR G URUSAMY EACT631 10 / 91

 A SYMPTOTIC S TABILITY
▶ An equilibrium state xe is said to be a asymptotically stable if it is stable
in the sense of Lyapunov and every solution starting within s(δ)
converges without leaving s(ϵ) to xe as t increases indefinitely.
▶ It is a local stability.
▶ Establishing asymptotic stability does not indicate the proper operation
of the system.
Asymptotic Stability in Large
▶ If the asymptotic stability holds for all states (all points in state space)
from which trajectories originate, then the equilibrium state is said to
be asymptotically stable in the large.
▶ [ Or ]
▶ The equilibrium state xe of the system is said to be asymptotically stable
in the large if it is stable and if every solution converges to xe as t
increases indefinitely.
Instability
▶ An equilibrium state xe is said to be unstable if for some real number
ε > 0 and any real number δ > 0, irrespective of how small it is, there is
always a state x0 in S(δ) such that the trajectory starting at this state
leaves S(ε).
D R .S ARAVANAKUMAR G URUSAMY EACT631 10 / 91
 A SYMPTOTIC S TABILITY OF L INEAR AND
N ONLIENAR S YSTEM

▶ An LTI system is asymptotically stable, meaning, the equilibrium state

at the origin is asymptotically stable, if and only if the eigenvalues of A
have negative real parts
▶ For LTI systems asymptotic stability is equivalent with convergence
(stability condition automatically satisfied)
▶ For LTI systems asymptotic stability is equivalent with convergence
(stability condition automatically satisfied)

D R .S ARAVANAKUMAR G URUSAMY EACT631 11 / 91

 A SYMPTOTIC S TABILITY OF L INEAR AND
N ONLIENAR S YSTEM

▶ An LTI system is asymptotically stable, meaning, the equilibrium state

at the origin is asymptotically stable, if and only if the eigenvalues of A
have negative real parts
▶ For LTI systems asymptotic stability is equivalent with convergence
(stability condition automatically satisfied)
▶ For LTI systems asymptotic stability is equivalent with convergence
(stability condition automatically satisfied)
▶ For nonlinear systems the state may initially tend away from the
equilibrium state of interest but subsequently may return to it

D R .S ARAVANAKUMAR G URUSAMY EACT631 11 / 91

 E XPONENTIAL S TABILITY

Definition -Exponential Stability An equilibrium point 0 is exponentially

stable if there exist two strictly positive numbers α and λ such that

∀t > 0, ∥x(t)∥ ≤ α∥x(0)∥e−λt

in some ball Br around the origin.

D R .S ARAVANAKUMAR G URUSAMY EACT631 12 / 91

 G RAPHICAL REPRESENTATION OF S TABILITY

D R .S ARAVANAKUMAR G URUSAMY EACT631 13 / 91

 E XAMPLE OF A N ONLINEAR S YSTEM WITH
M ULTIPLE E QUILIBRIUM P OINTS I

Given nonlinear system has the following form:

(
ẋ1 = −x21 + x22
ẋ2 = 2x1 x2 − x1

This system possesses two equilibrium points (x1 , x2 ) where the derivatives
are both zero:
1. First equilibrium point:
(
−x21 + x22 = 0
2x1 x2 − x1 = 0

2. Second equilibrium point: x1 = 0; x2 = 0 To find these equilibrium points,

solve the system of equations for the first equilibrium: From the first
equation: −x21 + x22 = 0, we get x22 = x21 or x2 = ±x1 . Substitute x2 = ±x1 into

D R .S ARAVANAKUMAR G URUSAMY EACT631 14 / 91

 E XAMPLE OF A N ONLINEAR S YSTEM WITH
M ULTIPLE E QUILIBRIUM P OINTS II

the second equation 2x1 x2 − x1 = 0: When

1
x2 = x1 :2x21 − x1 = 0; x1 (2x1 − 1) = 0 → x1 = 0 or x1 = 2 When x2 = −x1 :

−2x21 − x1 = 0; x1 (−2x1 − 1) = 0

1
x1 = 0 or x1 = −
2
Therefore, the equilibrium points are:
▶ 1. (x1 , x2 ) = (0, 0)
▶ 2. (x1 , x2 ) = 12 , 12

▶ 3. (x1 , x2 ) = − 12 , − 12

D R .S ARAVANAKUMAR G URUSAMY EACT631 15 / 91

 F ORMAL D EFINITIONS - LYAPUNOV M ETHODS I

▶ Consider the autonomous system:

ẋ = f (x)

where f : D → Rn is a locally Lipschitz function on a domain D ⊂ Rn .

• - Let x̄ ∈ D be an equilibrium point, denoted as f (x̄) = 0.
• - Objective: Characterize the stability of x̄.
• - Without loss of generality (WLOG), let x̄ = 0.
• - If x̄ ̸= 0, introduce a coordinate transformation: y = x − x̄, then

ẏ = ẋ = f (y + x̄) = g(y) with g(0) = 0

▶ Stability Definitions:
• Stable Equilibrium: - For each ϵ > 0, there exists δ = δ(ϵ) > 0 such
that:
∥x(0)∥ < δ ⇒ ∥x(t)∥ < ϵ ∀t ≥ 0
• Unstable Equilibrium: - If it is not stable.

D R .S ARAVANAKUMAR G URUSAMY EACT631 16 / 91

 F ORMAL D EFINITIONS - LYAPUNOV M ETHODS II

• Asymptotically Stable Equilibrium: - It is stable, and δ can be

chosen such that:

∥x(0)∥ < δ ⇒ lim x(t) = 0

t→∞

▶ Therefore, Lyapunov stability ensures the system trajectory can stay

arbitrarily close to the origin by starting sufficiently close to it.
▶ An equilibrium point that is Lyapunov stable but not asymptotically
stable is termed ”Marginally stable.”

D R .S ARAVANAKUMAR G URUSAMY EACT631 17 / 91

 STABILITY AND INSTABILITY
M ATHEMATICAL R EPRESENTATION

Definition 2.1
The equilibrium state x = 0 is said to be stable if, for any R > 0, there exists r > 0
such that if ∥x(0)∥ < r then ∥x(t)∥ < R for all t ≥ 0. Otherwise, the equilibrium
point is unstable.
Essentially, stability (also called stability in the sense of Lyapunov, or
Lyapunov equilibrium point) means that some trajectory can be kept
arbitrarily close to the origin by an arbitrarily specified radius BR , a value
r(R) can be found such that starting the state from within the ball Br at time
t = 0 guarantees that the state will stay within ball BR thereafter.
▶ ∀ to mean ”for any”
▶ ∃ for ”there exists”
▶ ∈ for ”in the set”
▶ =⇒ for ’for implies’
Using these symbols, the definition can be written as:

∀R > 0, ∃r > 0, ∥x(0)∥ < r =⇒ ∀t ≥ 0, ∥x(t)∥ < R

D R .S ARAVANAKUMAR G URUSAMY EACT631 18 / 91
 S TABILITY C ONCEPT IN 1D

D R .S ARAVANAKUMAR G URUSAMY EACT631 19 / 91

 LYAPUNOV ’ S L INEARIZATION M ETHOD I

Lyapunov’s linearization metho aims to analyze local stability for nonlinear

systems by approximating their behavior using linear approximations.
Lyapunov’s Linearization Method:
▶ 1. Objective: Analyze local stability of a nonlinear system.
▶ 2. Linear Approximation: For an autonomous system with f (x)
continuously differentiable:
• - The system dynamics ẋ = f (x) are written as ẋ = Ax + fh (x),
• - where fh (x) represents higher-order terms in x.
• - A denotes the Jacobian matrix of f with respect to x at the
equilibrium x = 0.
• - The linearization ẋ = Ax is considered the linear approximation
of the original nonlinear system at the equilibrium x = 0.
▶ 3. Extension to Non-Autonomous Systems with Control Input:
• - For a non-autonomous system with a control input u, the
dynamics ẋ = f (x, u) can be expressed as ẋ = Ax + Bu + fh (x, u).

D R .S ARAVANAKUMAR G URUSAMY EACT631 20 / 91

 LYAPUNOV ’ S L INEARIZATION M ETHOD II

• - A represents the Jacobian matrix of f with respect to x at

(x = 0, u = 0), and B denotes the Jacobian matrix of f with respect
to u at the same point.
▶ 4. Linear Approximation of Non-Autonomous Systems with Control:
• - The linear approximation is ẋ = Ax + Bu, where G reflects the
linearization of the function f with respect to x at its equilibrium
point x = 0.
•
▶ 5. Practical Approach to Linearization: - Often, obtaining a system’s
linearization is simplified by neglecting terms higher than the first
order in the system dynamics.
Key Implications:
▶ - Lyapunov’s linearization method justifies using linear control
techniques for stability, especially in analyzing local behavior around
equilibrium points.

D R .S ARAVANAKUMAR G URUSAMY EACT631 21 / 91

 LYAPUNOV ’ S L INEARIZATION M ETHOD III

▶ - Linear approximations aid in understanding the behavior of nonlinear

systems by providing simpler models that capture essential dynamics
for small range motions.
▶ This methodological approach to linearize nonlinear systems provides
a crucial foundation for stability analysis, enabling the use of linear
control strategies even in inherently nonlinear physical systems.

D R .S ARAVANAKUMAR G URUSAMY EACT631 22 / 91

 E XAMPLE I
The system equations are:

ẋ1 = x1 + x1 · cos(x2 )
ẋ2 = (x1 + 1) · x1 + x1 · sin(x2 )

Calculating the Jacobian matrix at x = 0:

1. Partial Derivatives:

∂ ẋ1
= 1 + cos(x2 )
∂x1
∂ ẋ1
= −x1 · sin(x2 )
∂x2
∂ ẋ2
= 2 · x1 + sin(x2 )
∂x1
∂ ẋ2
= x1 · cos(x2 )
∂x2
2. Evaluate at x = 0:
D R .S ARAVANAKUMAR G URUSAMY EACT631 23 / 91
 E XAMPLE II

∂ ẋ1
= 1 + cos(0) = 2
∂x1
∂ ẋ1
= −0 · sin(0) = 0
∂x2
∂ ẋ2
= 2 · 0 + sin(0) = 0
∂x1
∂ ẋ2
= 0 · cos(0) = 0
∂x2
Jacobian Matrix at x = 0: The Jacobian matrix A is formed from these partial
derivatives: " # 
∂ ẋ1 ∂ ẋ1 
∂x1 ∂x2 2 0
A = ∂ ẋ2 ∂ ẋ2 =
∂x ∂x
0 0
1 2

D R .S ARAVANAKUMAR G URUSAMY EACT631 24 / 91

 E XAMPLE III

Linearized Approximation: The linearized system around x = 0 is

represented by:
    
2 0 x1 2x1
ẋ = Ax = =
0 0 x2 0
This linear approximation describes the behavior of the system near the
equilibrium point x = 0. It predicts that the ẋ2 component remains at zero,
while ẋ1 doubles the value of x1 itself.

D R .S ARAVANAKUMAR G URUSAMY EACT631 25 / 91

 LYAPUNOV ’ S L INEARIZATION M ETHOD
Theorem (Lyapunov’s 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.
▶ - Strict Stability: All eigenvalues of matrix A are strictly in the left-half
complex plane.
▶ If the linearized system is unstable (i.e., if at least one eigenvalue of A is
strictly in the right-half complex plane), then the equilibrium point is
unstable for the nonlinear system.
▶ - Unstable Case: At least one eigenvalue of matrix A is strictly in the
right-half complex plane.
▶ If the linearized system is marginally stable (i.e., all eigenvalues of A
are in the left-half complex plane, but at least one of them is on the
imaginary axis), then one cannot conclusively determine the stability of
the equilibrium point from the linear approximation.
▶ - Marginal Stability: All eigenvalues of A are in the left-half complex
plane, but at least one of them is on the imaginary axis. In this case, the
equilibrium point may be stable, asymptotically stable, or unstable for
the nonlinear system, and no clear conclusion can be drawn from the
linear
D R .S ARAVANAKUMAR
G approximation.
URUSAMY EACT631 26 / 91
 D EFINITENESS OF S CALAR F UNCTIONS
Positive Definiteness: F(x) is positive and greater than zero.
▶ A scalar function F(x) is positive definite in a region including the
origin if F(x) > 0 for all non-zero states x in that region and F(0) = 0.
▶ Example: F(x) = x21 + 2x22 .
Negative Definiteness:
▶ A scalar function F(x) is negative definite if −F(x) is positive definite.
▶ Example: F(x) = −x21 − (3x1 + 2x2 )2 .
Positive Semi-definiteness: F(x) is positive or zero.
▶ A scalar function F(x) is positive semi-definite if it is positive except at
the origin and certain other states where it is zero.
▶ Example: (x1 + x2 )2 .
Negative Semi-definiteness:
▶ A scalar function F(x) is negative semi-definite if −F(x) is positive
semi-definite.
Indefiniteness:
▶ A scalar function F(x) is indefinite in a region if it assumes positive and
negative values regardless of the region’s size.
▶ Example: F(x) = x1 x2 + x22 .
D R .S ARAVANAKUMAR G URUSAMY EACT631 27 / 91
 Q UADRATIC FORM

- Quadratic Form:

P11 P12 · · · P1n

  
x1
 P12 P22 · · · P2n   x2 
e.g. F(x) = xT Px = [x1 x2 . . . xn ] 
 ...
 . 
  .. 
P1n · · · · · · Pnn xnn

P is real symmetric matrix and x is a real vector. - Hermitian Form If x is a

complex n vector and P is a Hermitian matrix then the complex quadratic
form is called Hermitian form.
P11 P12 · · · P1n
  
x1
 P̄12 P22 · · · P2n   x2 
e.g. F(x) = x∗ Px = [x1 x2 · · · xn ] 
 ...
 . 
  .. 
P̄1n P̄2n · · · Pnn xn

D R .S ARAVANAKUMAR G URUSAMY EACT631 28 / 91

 GLOBALLY POSITIVE DEFINITE
D EFINITION - GLOBALLY POSITIVE DEFINITE : A SCALAR
CONTINUOUS FUNCTION V(x) IS SAID TO BE LOCALLY
POSITIVE DEFINITE IF V(0) = 0 AND , IN A BALL BR , V(x) > 0
FOR x ̸= 0. I F V(0) = 0 AND THE ABOVE PROPERTY HOLDS
OVER THE WHOLE STATE SPACE , THEN V(x) IS SAID TO BE
GLOBALLY POSITIVE DEFINITE .

D R .S ARAVANAKUMAR G URUSAMY EACT631 29 / 91

 QUADRATIC FORM AND S YMMETRIZING

▶ A quadratic form is a function Q : V → F defined as Q(x) = xT Ax,

where x is a vector in V and A is a symmetric matrix.
▶ To symmetrize a quadratic form, given a matrix A that is not symmetric,
you can compute its symmetric counterpart B using the formula:

1
B = (A + AT )
2
▶ Consider the polynomial f (x, y) = 2x2 + 3xy − 4y2 . To express this
polynomial in terms of a quadratic form f (x, y) = xT Ax, we construct
the symmetric matrix A based on the coefficients of the quadratic terms:
a = 2, b = 3, and c = −4.
▶ The corresponding matrix A is given by:

2 32
 
A= 3
2 −4

D R .S ARAVANAKUMAR G URUSAMY EACT631 30 / 91

 S YMMETRIC Q UADRATIC F ORM E XAMPLE

▶ Given a quadratic form Q(x) = xT Ax with:

 
2 −1
A=
−1 3

This matrix A is symmetric.

 
1
▶ Compute Q(x) for x = using matrix A:
2
  
T
  2 −1 1
Q(x) = x Ax = 1 2 = 13
−1 3 2

D R .S ARAVANAKUMAR G URUSAMY EACT631 31 / 91

 D ERIVATION OF LYAPUNOV M ATRIX E QUATION I
The Lyapunov matrix equation is derived from considering the stability of a
linear time-invariant system described by the differential equation ẋ = Ax.
The Lyapunov equation is based on the search for a symmetric positive
definite matrix P that characterizes the system’s stability.
To derive the Lyapunov matrix equation step by step:
Consider the linear time-invariant system:

ẋ = Ax

To find a symmetric positive definite matrix P such that the system ẋ = Ax

is stable.
1. Lyapunov Stability Theory: The system ẋ = Ax is stable if there exists a
symmetric positive definite matrix P such that the following condition
holds for all x ̸= 0:
xT Px > 0
This condition ensures that the quadratic form xT Px is positive definite.
2. Lyapunov Function: Let’s define a candidate Lyapunov function as
V(x) = xT Px. This function represents the energy or potential of the system.
D R .S ARAVANAKUMAR G URUSAMY EACT631 32 / 91
 D ERIVATION OF LYAPUNOV M ATRIX E QUATION II
3. Derivative of the Lyapunov Function: Compute the derivative of V(x)
with respect to time:
d T
V̇(x) = (x Px)
dt
Using the chain rule, we get:

V̇(x) = ẋT Px + xT Pẋ

Given ẋ = Ax:

V̇(x) = (Ax)T Px + xT P(Ax)

V̇(x) = xT AT Px + xT PAx
V̇(x) = xT (AT P + PA)x
4. Stability Condition: For the system to be stable, the derivative V̇(x) must
be negative definite (for asymptotic stability) or negative semidefinite (for
stability):
D R .S ARAVANAKUMAR G URUSAMY EACT631 33 / 91
 D ERIVATION OF LYAPUNOV M ATRIX E QUATION III

xT (AT P + PA)x < 0

or
xT (AT P + PA)x ≤ 0
5. Lyapunov Matrix Equation: To find P for stability, equate AT P + PA to a
negative definite (or negative semidefinite) matrix Q:

AT P + PA = −Q
where Q is a symmetric negative definite (or negative semidefinite) matrix.
This process derives the Lyapunov matrix equation, which plays a crucial
role in stability analysis and control system design. Solving this equation
helps in determining the appropriate P to ensure stability or to design
control strategies for stable behavior in linear systems.

D R .S ARAVANAKUMAR G URUSAMY EACT631 34 / 91

 S YLVESTER ’ S C RITERION

▶ The positive definiteness of Quadratic form or Hermitian form F(x) can

be determined by Sylvester’s criterion.
▶ Necessary and sufficient conditions for Quadratic or Hermitian form to
be positive definite is

P11 · · · · · · P1n
 

P P  P̄12 · · · · · · P2n 
P11 > 0 11 12 > 0 . . . · · · 
 ...
>0
P̄12 P22 
P̄1n · · · · · · Pnn

F(x) = x∗ Px is positive semidefinite if P is singular and all the principle

minors are non-negative. F(x) is negative definite if −F(x) is positive
definite. Similarly F(x) is negative semidefinite if −F(x) is positive
semidefinite.

D R .S ARAVANAKUMAR G URUSAMY EACT631 35 / 91

 S YLVESTER ’ S CRITERION I

Sylvester’s criterion says that actually, this characterizes positive definite

matrices:

Theorem 1.1 (Sylvester’s criterion). Let A be an n × n symmetric matrix.

Then:
▶ A > 0 if and only if ∆1 > 0, ∆2 > 0, . . . , ∆n > 0.
▶ A < 0 if and only if (−1)1 ∆1 > 0, (−1)2 ∆2 > 0, . . . , (−1)n ∆n > 0.
▶ A is indefinite if the first ∆k that breaks both patterns is the wrong sign.
▶ Sylvester’s criterion is inconclusive (A can be positive or negative
semidefinite, or indefinite) if the first ∆k that breaks both patterns is 0.

D R .S ARAVANAKUMAR G URUSAMY EACT631 36 / 91

 S YLVESTER ’ S CRITERION II

Example Calculation To determine the definiteness of the matrix A using

Sylvester’s criterion.
Given matrix A:
 
4 −1 2
A = −1 3 1
2 1 5
Step 1: Determinants of Principal Minors 1. |A1 | - Principal Minor of Order 1
(1x1 Matrix):
|A1 | = 4
2. A2 - Principal Minor of Order 2 (2x2 Matrix):

4 −1
A2 =
−1 3

|A2 | = (4 × 3) − (−1 × −1) = 12 − 1 = 11

D R .S ARAVANAKUMAR G URUSAMY EACT631 37 / 91

 S YLVESTER ’ S CRITERION III

3. A3 - Principal Minor of Order 3 (3x3 Matrix):

4 −1 2
A3 = −1 3 1
2 1 5

|A3 | = 4 × (3 × 5 − 1 × 1) − (−1) × (2 × 5 − 1 × 2) + 2 × (−1 × 1 − 3 × 2)

|A3 | = 4 × (15 − 1) − (−1) × (10 − 2) + 2 × (−1 − 6)
|A3 | = 4 × 14 + 8 + (−14) = 56 + 8 − 14 = 50
Step 2: Sylvester’s Criterion The determinants of principal minors are: -
|A1 | = 4 - |A2 | = 11 - |A3 | = 50
As all determinants (A1 , A2 , and A3 ) are positive, Sylvester’s criterion
confirms that all eigenvalues of matrix A are positive. Therefore, based on
Sylvester’s criterion, matrix A is positive definite.

D R .S ARAVANAKUMAR G URUSAMY EACT631 38 / 91

 LYAPUNOV ’ S DIRECT M ETHOD I
▶ The philosophy behind Lyapunov’s direct method is rooted in the
observation that in physical systems, if the total energy continuously
dissipates, the system—regardless of being linear or nonlinear—will
eventually settle down to an equilibrium point.
▶ The stability of such a system can be inferred by examining the
behavior of a single scalar function representing the system’s energy.
Consider a nonlinear mass-damper-spring
system described by the dynamic equation:

mẍ + bẋ + k0 x + k1 x3 = 0
where
▶ m represents the mass
▶ b is the nonlinear dissipation or damping
term
▶ k0 x represents linear spring behavior
▶ k1 x3 represents a nonlinear spring term
▶ Initial displacement of mass from spring’s natural length, then released.
D R .S ARAVANAKUMAR G URUSAMY EACT631 39 / 91
 LYAPUNOV ’ S DIRECT M ETHOD II

▶ Inquiry: Will resulting motion be stable?

▶ —Challenges:
▶ Unavailability of a general solution for the nonlinear equation impedes
traditional stability analysis.
▶ Inapplicability of linearization methods due to motion starting beyond
the system’s linear range.
▶ Expected marginal stability if a linear approximation were feasible.
▶ — Focus on Energy of the system
▶ Examining energy variations offers insights despite nonlinear equation
complexity.
▶ Provides understanding of motion patterns and stability tendencies.
▶ Avoids reliance solely on explicit mathematical solutions or linear
approximations.

D R .S ARAVANAKUMAR G URUSAMY EACT631 40 / 91

 LYAPUNOV ’ S DIRECT M ETHOD III

Instead of relying on explicit solutions or intricate stability definitions, this

method focuses on constructing a scalar function, often termed an
”energy-like” function, for the dynamic system.
The fundamental procedure of Lyapunov’s direct method involves these
steps:
1. Generation of Scalar Function: The primary step is to create a scalar
function that behaves akin to an energy function for the system. This
function, typically denoted as the Lyapunov function, characterizes the
system’s behavior and captures its dynamics.
2. Examination of Time Variation: Once the Lyapunov function is
established, the method involves analyzing the rate of change or time
variation of this scalar function. By observing how this function evolves
over time, conclusions can be drawn regarding the system’s stability.

D R .S ARAVANAKUMAR G URUSAMY EACT631 41 / 91

 LYAPUNOV ’ S DIRECT M ETHOD IV

3. Drawing Stability Conclusions: The behavior and time variation of the

scalar function provide insights into the stability of the set of
differential equations. If the function’s variation satisfies certain
criteria, stability conclusions about the system can be deduced without
necessitating explicit knowledge of the solutions or resorting to
complex stability definitions.

D R .S ARAVANAKUMAR G URUSAMY EACT631 42 / 91

 LYAPUNOV D IRECT M ETHOD (2nd M ETHOD )

▶ A system, which is vibrating, is stable if its total energy is continuously

decreasing.
▶ This is indicates that the time derivative of the total energy must be
negative.
▶ Total energy is a positive definite function.
▶ If the system has an asymptotically stable at equilibrium state then the
stored energy decays with increase in time till it attains minimum value
at the equilibrium state.
▶ There is no simple way for defining an energy function.
▶ This difficulty was overcome as Lyapunov function method which is
fabricated energy function.
▶ In this method sign behaviour of F(x, t) and its time derivative
dF(x, t)/dt gives us information about stability of an equilibrium state
without solving the equation.

D R .S ARAVANAKUMAR G URUSAMY EACT631 43 / 91

 P OSITIVE D EFINITENESS AND LYAPUNOV
F UNCTION
▶ -Local Positive Definiteness:
▶ A scalar continuous function V(x) is considered locally positive definite
if it satisfies two conditions:
• 1. V(0) = 0: The function evaluates to zero at the origin.
• 2. In a ball BR : Within a neighborhood BR around the origin,
V(x) > 0 for all points x other than the origin.
▶ -Global Positive Definiteness:
• If V(0) = 0 and the property of V(x) > 0 holds true over the entire
state space (not just within a specific neighborhood), then the
function V(x) is termed globally positive definite.
▶ Definition: Lyapunov Function If, within a ball BR , the function V(x) is
positive definite and possesses continuous partial derivatives, and if its
time derivative along any state trajectory of the system is negative
semi-definite, i.e.,

V̇(x) ≤ 0 for all x ̸= 0

then V(x) is said to be a Lyapunov function for the system.
D R .S ARAVANAKUMAR G URUSAMY EACT631 44 / 91
 LYAPUNOV T HEOREM I
Consider the system
ẋ = f(x)
Equilibrium state x0 (f(0) = 0) A function V(x) is called a Lyapunov
function if
1. V(x) > 0, ∀x ̸= 0
2. V(0) = 0 for x = 0
3. dV(x) dV(x)
dt = dx < 0
Then equilibrium state of the system (ẋ = f(x)) is stable.
Moreover, if the Lyapunov function satisfies

dV(x) dV(x)
< 0, ∀x ̸= 0 and =0⇔x=0
dt dt

▶ Then equilibrium state of the system (ẋ = f(x)) is asymptotically stable.

▶ The derivative of the Lyapunov function along the trajectory is
negative.
D R .S ARAVANAKUMAR G URUSAMY EACT631 45 / 91
 LYAPUNOV T HEOREM II

▶ The Lyapunov function may be considered as an energy function of the

system.
▶ It is only sufficient for stability.
▶ In case of nonlinear system, we take a candidate V(x) ≥ 0 and check If
along the trajectories of ẋ = f (x), V(x) ≤ 0
▶ In case V(x) ≤ 0 then either,The equilibrium point is not asymptotically
stable or We did not choose our candidate v(x) carefully. We have to
look for another V ( Lyapunov function).

D R .S ARAVANAKUMAR G URUSAMY EACT631 46 / 91

 LYAPUNOV T HEOREM FOR S TABILITY

Lyapunov Theorem for Local Stability

▶ If, within a ball BR , there exists a scalar function V(x) with continuous
first partial derivatives such that:
• V(x) is positive definite (locally in BR )
• V(x) is negative semi-definite (locally in BR )
then the equilibrium point x = 0 is stable. If, in fact, the derivative V̇(x)
is locally negative definite in BR , then the stability is asymptotic.
Lyapunov Theorem for Global Stability
Assume that there exists a scalar function V of the state x, with continuous
first-order derivatives, such that:
▶ V(x) is positive definite
▶ V(x) is negative definite
▶ V̇(x) < 0 for all x ̸= 0
then the equilibrium at the origin is globally asymptotically stable.

D R .S ARAVANAKUMAR G URUSAMY EACT631 47 / 91

 LYAPUNOV T HEOREM - E XAMPLE I
EX 1: Consider the system described by the differential equation:
dx
= −x3
dt
To analyze the stability of the equilibrium point at x = 0, Lyapunov’s
method is used. The first step is to find a Lyapunov function V(x) that
satisfies the following conditions:
1. V(x) is positive definite.
2. V ′ (x) is negative definite or negative semi-definite.
▶ Choose a Lyapunov function as V(x) = 21 x2 .
▶ Now, calculate V ′ (x), which represents the derivative of V(x) along the
trajectories of the system:
dV dx
V ′ (x) = =x· = x · (−x3 ) = −x4
dx dt
Since V ′ (x) = −x4 , it’s always negative for x ̸= 0 (except at x = 0). Thus,
V ′ (x) is negative definite.
D R .S ARAVANAKUMAR G URUSAMY EACT631 48 / 91
 LYAPUNOV T HEOREM - E XAMPLE II

▶ Now, shown that V ′ (x) is negative definite for x ̸= 0, and V(x) is

positive definite (as it’s 21 x2 ). Therefore, according to Lyapunov’s
method, the equilibrium point at x = 0 is stable.
▶ Additionally, to determine asymptotic stability, we need to check if
V ′ (x) is negative definite (not just negative semi-definite). In this case,
since V ′ (x) = −x4 , which is negative definite (except at x = 0), the
equilibrium point x = 0 is asymptotically stable.
Ex 2: Consider the following system described by the differential equation:
ẋ = −x + x3
To determine the stability and asymptotic stability of the equilibrium point
at x = 0, we’ll use Lyapunov’s method.
▶ Step 1: Choose a Lyapunov Function Let’s choose a Lyapunov function
candidate, V(x), which should be positive definite, i.e., V(x) > 0 for all
x ̸= 0 and V(0) = 0.
A common choice for this system is V(x) = 12 x2 .

D R .S ARAVANAKUMAR G URUSAMY EACT631 49 / 91

 LYAPUNOV T HEOREM - E XAMPLE III
▶ Step 2: Calculate V ′ (x) Now, compute the derivative of V(x) along the
trajectories of the system, denoted as V ′ (x) or dV
dt .
 
dV d 1
V ′ (x) = = x2 = xẋ
dt dt 2
Substitute ẋ = −x + x3 into the equation:

V ′ (x) = x(−x + x3 ) = −x2 + x4

▶ Step 3: Analyze V ′ (x) - For stability, V ′ (x) should be negative definite or
negative semi-definite. - For asymptotic stability, V ′ (x) should be
negative definite.
Now, evaluate V ′ (x) at the equilibrium point x = 0:

V ′ (0) = −02 + 04 = 0
The equilibrium point’s stability can’t be determined directly from
V ′ (x) at x = 0.
▶ Step 4: Analyze V ′ (x) for All x ̸= 0
D R .S ARAVANAKUMAR G URUSAMY EACT631 50 / 91
 LYAPUNOV T HEOREM - E XAMPLE IV

Consider the sign of V ′ (x) for x ̸= 0:

•
- When x is small (near 0), x4 term will dominate.
•
- For x > 0.1, x4 becomes larger than x2 .
•
- For x < −0.1, x4 also dominates.
•
So, V ′ (x) = −x2 + x4 is negative definite for x ̸= 0 but V ′ (0) = 0.
•
▶ Conclusion: - The Lyapunov function V(x) = 21 x2 confirms that the
equilibrium point at x = 0 is stable but not asymptotically stable.
▶ Additional Problems for Practice:
1. 1. Analyze the stability of the equilibrium point x = 0 for the
system ẋ = −2x.
2. 2. Determine the stability of the equilibrium point x = 0 for the
system ẋ = x3 − x.
3. 3. Investigate the stability of the equilibrium point x = 0 for the
system ẋ = −x3 .

D R .S ARAVANAKUMAR G URUSAMY EACT631 51 / 91

 LYAPUNOV S TABILITY OF L INEAR S YSTEM
Assume that the dynamic system of the Linear form
ẋ = Ax
Let P > 0 be a symmetric positive definite matrix, then we define
V(x) = xT Px. We have:
V̇(x) = ẋT Px + xT Pẋ
V̇(x) = xT AT Px + xT PAx
V̇(x) = xT (AT P + PA)x
Theorem: An autonomous linear system (x) ˙ = Ax is stable if and only if for
any symmetric, positive definite matrix Q there exists a corresponding
symmetric positive definite matrix P such that
AT P + PA = −Q
Is this negative definite?
V̇(x) = xT (AT P + PA)x
V̇(x) = xT (−Q)x
As Q is positive definite, V̇(x) is negative definite.
D R .S ARAVANAKUMAR G URUSAMY EACT631 52 / 91
 LYAPUNOV S TABILITY OF L INEAR S YSTEM I
E XAMPLE

 
−2 1
Determine the stability of the linear system ẋ = Ax with A = ,
−1 −2
Calculate the P matrix using the Lyapunov equation

AT P + PA = −Q

.
Given A, Assume a specific Q matrix as an example:
 
1 0
Q=
0 1
Now, solve the Lyapunov equation AT P + PA = −Q to find the P matrix.

AT P + PA = −Q
     
−2 −1 −2 1 −1 0
P+P =
1 −2 −1 −2 0 −1

D R .S ARAVANAKUMAR G URUSAMY EACT631 53 / 91

 LYAPUNOV S TABILITY OF L INEAR S YSTEM II
E XAMPLE


p11 p12
Let P = .
p21 p22
Solving the matrix equation, we get:
   
−2p11 − p21 + p11 −p12 − 2p22 + p12 −1 0
=
p21 − p11 − 2p22 p22 − p12 − 2p21 0 −1
Solving the corresponding equations:
▶ 1. −2p11 − p21 + p11 = −1 ⇒ −p21 − p11 = −1
▶ 2. p21 − p11 − 2p22 = 0 ⇒ p21 − p11 − 2p22 = 0
▶ 3. −p12 − 2p22 + p12 = 0 ⇒ −2p22 = 0 ⇒ p22 = 0

D R .S ARAVANAKUMAR G URUSAMY EACT631 54 / 91

 LYAPUNOV S TABILITY OF L INEAR S YSTEM III
E XAMPLE

From equation 3, we find that p22 = 0. Then, using equation 1:

−p21 − p11 = −1 ⇒ p21 = −1 − p11

Using this in equation 2:
1
−1 − p11 − p11 − 2 × 0 = 0 ⇒ p11 = −
2
Thus, found that p11 = − 21 and p21 = −1 − (− 21 ) = − 12 .
Therefore, the symmetric positive definite matrix P satisfying
AT P + PA = −Q is:
 1
− 2 − 12

P=
− 12 0
This computed P matrix can be used to evaluate the stability characteristics
of the given linear system.

D R .S ARAVANAKUMAR G URUSAMY EACT631 55 / 91

 I NVARIANT S ET T HEOREM I

A set G is considered an invariant set for a dynamic system if every system

trajectory starting from a point in G remains in G for all future time.
For example:
▶ Any equilibrium point is an invariant set.
▶ The domain of attraction of an equilibrium point is also an invariant set.
▶ The null state-space is a trivial invariant set.
▶ For an autonomous system, any trajectory in the state-space is an
invariant set.
▶ Limit cycles, as special cases of system trajectories (closed curves in the
phase plane), are also considered invariant sets.

D R .S ARAVANAKUMAR G URUSAMY EACT631 56 / 91

 I NVARIANT S ET T HEOREM II
Invariant set theorems, apart from often deducing asymptotic stability
when the derivative of the Lyapunov function candidate (V̇) is only
negative semi-definite, extend the concept of Lyapunov functions to
describe convergence to dynamic behaviors beyond equilibrium, such as
convergence to a limit cycle.
Theorem: (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 derivatives. Assume that:
▶ for some l > 0, the region Ω1 defined by V(x) < l is bounded,
▶ V(x) ≤ 0 for all x in Ω1 .
Let R be the set of all points within Ω1 where V(x) = 0, and M be the largest
invariant set in R. Then, every solution x(t) originating in Ω1 tends to M as
t → ∞.
Note that in the above theorem, the word ”largest” is understood in the
sense of set theory, i.e., M is the union of all invariant sets (e.g., equilibrium
points or limit cycles) within R. In particular, if the set R itself is invariant
(i.e., if once V̇ = 0, then V̇ = 0 for all future time), then M = R.
D R .S ARAVANAKUMAR G URUSAMY EACT631 57 / 91
 I NVARIANT S ET T HEOREM III

Theorem (Global Invariant Set Theorem) Consider the autonomous

system, with f continuous, and let V(x) be a scalar function with continuous
first partial derivatives. Assume that:
▶ V(x) ≤ 0 over the whole state space,
▶ V(x) → ∞ as ∥x∥ → ∞.
Let R be the set of all points where V(x) = 0, and M be the largest invariant
set in R. Then all solutions globally asymptotically converge to M as t → ∞

D R .S ARAVANAKUMAR G URUSAMY EACT631 58 / 91

 D EFINITION : I NVARIANT S ET FOR A D YNAMIC
S YSTEM

Definition: A set G is an invariant set for a dynamical system if every

trajectory within the system, initiated from a point in G, remains within G
for all future time.
Mathematically:

For all x0 ∈ G, ϕ(t, x0 ) ∈ G, for all t ≥ 0

where ϕ(t, x0 ) represents the trajectory of the system starting at x0 at time t.

D R .S ARAVANAKUMAR G URUSAMY EACT631 59 / 91

 T HEOREM : L OCAL I NVARIANT S ET T HEOREM

Theorem (Local Invariant Set Theorem): Consider an autonomous system

of the form ẋ = f (x), with f continuous, and let V(x) be a scalar function
with continuous first partial derivatives. Assume that for some ϵ > 0, the
region Qϵ , defined by V(x) < ϵ, is bounded, and p(x) ≤ 0 for all x in Qϵ .
Let R be the set of all points within Qϵ where p(x) = 0, and Ω be the largest
invariant set in R. Then, every solution x(t) originating in Qϵ tends to Ω as
t → ∞.
Mathematically: For all solutions x(t) with initial conditions in Qϵ ,

lim x(t) = M
t→∞

D R .S ARAVANAKUMAR G URUSAMY EACT631 60 / 91

 T HEOREM : G LOBAL I NVARIANT S ET T HEOREM
Theorem (Global Invariant Set Theorem): Consider the autonomous
system ẋ = f (x), with f continuous, and let V(x) be a scalar function with
continuous first partial derivatives. Assume that V(x) < 0 over the entire
state space.
Let R be the set of all points where dV
dx = 0, and M be the largest invariant set
in R. Then, all solutions globally asymptotically converge to M as t → ∞.
Mathematically: For all solutions x(t), regardless of initial conditions and
over the entire state space,
lim x(t) = M
t→∞

D R .S ARAVANAKUMAR G URUSAMY EACT631 61 / 91

 I NVARIANCE P RINCIPLE

Summary Positively Invariant Set (M):

▶ A set M is positively invariant with respect to ẋ = f (x) if
x(0) ∈ M ⇒ x(t) ∈ M, ∀t ≥ 0.
Invariance of Equilibrium Points and Limit Cycles:
▶ Equilibrium points and limit cycles are examples of invariant sets.
Positively Invariant Set Defined by Lyapunov Function (Ω):
▶ The set Ω = {x ∈ Rn | V(x) ≤ c} with V̇ ≤ 0 for all x ∈ Ω is a positively
invariant set.

D R .S ARAVANAKUMAR G URUSAMY EACT631 62 / 91

 L A S ALLE ’ S T HEOREM

Conditions for Lasalle’s Theorem:

▶ Consider a compact set Ω where solutions of ẋ = f (x) starting in Ω
remain in Ω for all future times.
▶ Let V : Ω → R be continuously differentiable with V̇(x) ≤ 0 in Ω.
▶ Define E as the set where V̇(x) = 0.
▶ M is the largest invariant set in E.
Conclusion of La Salle’s Theorem:
▶ Every solution starting in Ω approaches M as t → ∞.
▶ Unlike the Lyapunov theorem, Lasalle’s theorem doesn’t require V(x)
to be positive definite; it requires Ω to be bounded.
▶ Ω is bounded for small c if V is positive definite. If V is radially
unbounded, Ω remains bounded for all c, regardless of V being positive
definite.
▶ To show approach to the origin, prove that the largest invariant set in E
is the origin.

D R .S ARAVANAKUMAR G URUSAMY EACT631 63 / 91

 L A S ALLE ’ S T HEOREM

Example
▶ - Given V(x) = x2
▶ compute V̇(x):

dV d(x2 )
V̇(x) = = = 2xẋ = 2x(−x) = −2x2
dt dt
▶ - In this case, Ω = {x ∈ R | V(x) ≤ c} is the interval where V(x) ≤ c. If
V(x) = x2 , then Ω will be bounded for c > 0 as it represents an interval
on the real line.
▶ - Applying Lasalle’s Theorem, since V̇(x) = −2x2 ≤ 0 for x ∈ Ω, and
E = {0} (where V̇(x) = 0), M is the largest invariant set in E.
▶ Therefore, according to Lasalle’s Theorem, any solution starting in Ω
approaches M (in this case, M = {0}) as time goes to infinity.

D R .S ARAVANAKUMAR G URUSAMY EACT631 64 / 91

 VARIABLE G RADIENT M ETHOD FOR LYAPUNOV
F UNCTIONS I

Scalar Function and Gradient Relation:

▶ Scalar function V(x) related to its gradient Vv by:
Z
V(x) = Vv · dx

- Recovery of unique V(x) requires specific conditions on Vv .

Conditions for Recovering V(x):
▶ Curl equations in n = 2 dimensions:

∂V2 ∂V1
=
∂x1 ∂x2
- Directional derivatives in Vv relate through these conditions.
Principle of Variable Gradient Method:

D R .S ARAVANAKUMAR G URUSAMY EACT631 65 / 91

 VARIABLE G RADIENT M ETHOD FOR LYAPUNOV
F UNCTIONS II

▶ Instead of specifying V directly, assume a form for Vv :

   ∂V 
a11 ∂x1
a21   ∂V 
 ∂x2 
Vv =  ..  ·  . 
 
.  .. 
an1 ∂V
∂x n

- aij coefficients are to be determined.

Procedure for Seeking V:
▶ Assume Vv form, solve for coefficients to satisfy curl equations.
▶ Restrict coefficients in Vv for local negative semi-definiteness.
▶ Integrate Vv to derive Lyapunov function V.
▶ Verify positivity of V for validity.

D R .S ARAVANAKUMAR G URUSAMY EACT631 66 / 91

 E XAMPLE : VARIABLE G RADIENT M ETHOD FOR
LYAPUNOV F UNCTIONS I

Given Differential Equation:

ẋ = −2x

Steps for Finding Lyapunov Function:

1. Relationship between V(x) and Vv :
Z
V(x) = Vv · dx

2. Assumed Form for Vv :

Vv = a · x
3. Derive Gradient Vv Components:

∂V
Vv 1 = =a
∂x
D R .S ARAVANAKUMAR G URUSAMY EACT631 67 / 91
 E XAMPLE : VARIABLE G RADIENT M ETHOD FOR
LYAPUNOV F UNCTIONS II

4. Check Conditions for Recovering V(x): - Curl condition not applicable

in 1 dimension.
5. Construct Lyapunov Function V(x):
Z Z
V(x) = Vv · dx = a · dx = ax + C

(Where C is an integration constant)

6. Verify Positivity of V(x): - Choose a > 0 for positivity.
This demonstrates the process of using the Variable Gradient Method to
find a Lyapunov function for a given dynamical system.

D R .S ARAVANAKUMAR G URUSAMY EACT631 68 / 91

 LYAPUNOV S TABILITY OF N ON -L INEAR S YSTEM -
C ONSTRUCTION OF LYAPUNOV FUNCTION BY
K RASOVSKI ’ S METHOD
▶ For a given Lyapunov function V(x), there is no general method which
allows us to confirm that the function is positive definite.
▶ If V(x) is in the quadratic form in xi ’s then Sylvester’s theorem can be
applied to test the positive definiteness of the function.
▶ If it is not possible to obtain the Lyapunov’s function of required type
then it does not indicate that the system is unstable.
▶ Krasovskii’s method gives a sufficient condition for the equilibrium
state to be asymptotically stable. Consider the system described by
following equation. ẋ = f (x); f (0) = 0. The origin is a singular point.
▶ Let us consider Liapunov’s function as,
V = f T Pf (1)
Here P is a symmetric positive definite matrix. Now,
V̇ = f T Pḟ + f T Pḟ (2)
∂f ∂x
ḟ = · = Jf (3)
D R .S ARAVANAKUMAR G URUSAMY
∂x ∂t EACT631 69 / 91
 C ONSTRUCTION OF LYAPUNOV FUNCTION BY
K RASOVSKI ’ S METHOD
J is called Jacobian Matrix given by
 
∂f1 ∂f2 ∂f1
∂x1 ∂x2 ··· ∂xn
 ∂f2 ∂f2 ∂f2 

∂x1 ∂x2 ··· ∂xn

J=
 ... .. .. .. 

 . . . 
∂fn ∂fn ∂fn
∂x1 ∂x2 ··· ∂xn

Substituting ḟ in expression for V̇ we have,

V̇ = f T JT pf + f T PJf
= f T JT P + PJ f

Let
Q = JT P + PJ
As V is positive definite, for the system to be asymptotically stable, Q
should be negative definite. If in addition V(x) is tending to infinity as norm
of x tends to infinity (∥x∥ → ∞) the system is asymptotically stable in the
large.
D R .S ARAVANAKUMAR G URUSAMY EACT631 70 / 91
 LYAPUNOV S TABILITY OF N ON -L INEAR S YSTEM -
C ONSTRUCTION OF LYAPUNOV FUNCTION BY
K RASOVSKI ’ S METHOD
Introduction:
▶ There’s no general method to confirm positive definiteness of a given
Lyapunov function V(x).
▶ Sylvester’s theorem is applicable for V(x) in quadratic form to test
positive definiteness.
▶ Lack of obtaining the required Lyapunov function doesn’t indicate
system instability.
▶ Krasovskii’s method provides a sufficient condition for asymptotic
stability of the equilibrium state.
Krasovskii’s Method:
▶ Consider the system ẋ = f (x) with f (0) = 0 (origin is a singular point).
▶ Define the Lyapunov function as V = f T Pf with P being a symmetric
positive definite matrix.
▶ Express V̇ using J (Jacobian matrix) and Q = JT P + PJ.
▶ For asymptotic stability: Q should be negative definite if V(x) tends to
infinity as ||x|| → ∞.
D R .S ARAVANAKUMAR G URUSAMY EACT631 71 / 91
 E XAMPLE : C ONSTRUCTION OF LYAPUNOV
FUNCTION BY K RASOVSKI ’ S METHOD
Consider the non-linear system:

−x31
 
ẋ =
−x32
Jacobian Matrix:
−3x21
 
0
J=
0 −3x22
Lyapunov Function:
  
 1 0 x1
= x21 + x22

V = x1 x2
0 1 x2
Derivation:
2
  
−3x 0 x1
V̇ = −x1 −x32 = −3x51 − 3x52
 3  1
2
0 −3x2 x2
Conclusion: The system is asymptotically stable around the origin as V is
positive definite and V̇ tends to negative infinity as ||x|| → ∞.
D R .S ARAVANAKUMAR G URUSAMY EACT631 72 / 91
 C OMPUTING V̇ USING S YSTEM D YNAMICS AND
LYAPUNOV F UNCTION I

−x31

Example 1: ẋ = with V = x21 + x22
−x32 
−x31

System Dynamics: ẋ =
−x32
Lyapunov Function: V = x21 + x22
Derivation of V̇:
1. Compute V̇ using the chain rule: V̇ = ∇V · ẋ, where ∇V is the gradient
of V and ẋ represents the system dynamics.
 ∂V   
2x1
2. ∇V = ∂x ∂V = 2x
1

2
 ∂x 2

−x31

3. ẋ =
−x32
   3
2x1 −x1
4. V̇ = · = −2x41 − 2x42
2x2 −x32

D R .S ARAVANAKUMAR G URUSAMY EACT631 73 / 91

 C OMPUTING V̇ USING S YSTEM D YNAMICS AND
LYAPUNOV F UNCTION II
 
−2x1 − x2
Example 2: ẋ = with V = x21 + x22
−x1 − 2x
2 
−2x1 − x2
System Dynamics: ẋ =
−x1 − 2x2
Lyapunov Function: V = x21 + x22
Derivation of V̇:
1. Compute V̇ using the chain rule: V̇ = ∇V · ẋ, where ∇V is the gradient
of V and ẋ represents the system dynamics.
 ∂V   
2x1
2. ∇V = ∂x ∂V = 2x
1

2
 ∂x2 
−2x1 − x2
3. ẋ =
−x1 − 2x2
   
2x1 −2x1 − x2
4. V̇ = · = −6x21 − 6x22
2x2 −x1 − 2x2

D R .S ARAVANAKUMAR G URUSAMY EACT631 74 / 91

 PASSIVITY I

▶ Concept of Passivity:
▶ Passivity in control systems refers to a property where a system
dissipates or conserves energy when subjected to external inputs. It is
an essential concept in stability analysis and controller design.
▶ Key Elements:
▶ 1. Energy Storage and Dissipation: In control systems, energy storage
elements represent capacitance and inductance, while energy
dissipation elements are analogous to resistance in electrical circuits.
▶ 2. Passivity Criterion: For a dynamical system ẋ = f (x, u), it’s passive if
the total stored energy doesn’t increase more than the input energy.

D R .S ARAVANAKUMAR G URUSAMY EACT631 75 / 91

 PASSIVITY II
Illustrative Example:
Consider a nonlinear system described by:

x˙1 = x2 , x˙2 = −x1 − x2 + u

where u(t) is the input and x1 , x2 are state variables.
Passivity Verification:
1. Storage Function Selection: Define a storage function V(x) = 21 (x21 + x22 ).
2. Derivative of the Storage Function: Compute V̇(x), the derivative of V(x)
along the system trajectories:
V̇(x) = ∇V(x) · f (x, u)
3. Calculate V̇(x): Substitute the system dynamics into V̇(x) to check if it
satisfies the passivity criterion.
Calculation:
Given the system dynamics:

V̇(x) = x1 x˙1 + x2 x˙2

Substituting the system equations into V̇(x):
D R .S ARAVANAKUMAR G URUSAMY EACT631 76 / 91
 PASSIVITY III

V̇(x) = x1 · x2 + x2 · (−x1 − x2 + u)
Simplify the expression:

V̇(x) = −x22 + x1 u
Conclusion and Analysis:
To fulfill the passivity criterion, V̇(x) ≤ 0 for all x and u. In this case, the
term x1 u indicates that the system’s passivity may depend on the nature of
the input u and initial state x.

D R .S ARAVANAKUMAR G URUSAMY EACT631 77 / 91

 LYAPUNOV S TABILITY OF N ON -L INEAR S YSTEM -
C ONSTRUCTION OF LYAPUNOV FUNCTION BY
K RASOVSKI ’ S METHOD
▶ For a given Lyapunov function V(x), there is no general method which
allows us to confirm that the function is positive definite.
▶ If V(x) is in the quadratic form in xi ’s then Sylvester’s theorem can be
applied to test the positive definiteness of the function.
▶ If it is not possible to obtain the Lyapunov’s function of required type
then it does not indicate that the system is unstable.
▶ Krasovskii’s method gives a sufficient condition for the equilibrium
state to be asymptotically stable. Consider the system described by
following equation. ẋ = f (x); f (0) = 0. The origin is a singular point.
▶ Let us consider Liapunov’s function as,
V = f T Pf (4)
Here P is a symmetric positive definite matrix. Now,
V̇ = f T Pḟ + f T Pḟ (5)
∂f ∂x
ḟ = · = Jf (6)
D R .S ARAVANAKUMAR G URUSAMY
∂x ∂t EACT631 78 / 91
 C ONSTRUCTION OF LYAPUNOV FUNCTION BY
K RASOVSKI ’ S METHOD
J is called Jacobian Matrix given by
 
∂f1 ∂f2 ∂f1
∂x1 ∂x2 ··· ∂xn
 ∂f2 ∂f2 ∂f2 

∂x1 ∂x2 ··· ∂xn

J=
 ... .. .. .. 

 . . . 
∂fn ∂fn ∂fn
∂x1 ∂x2 ··· ∂xn

Substituting ḟ in expression for V̇ we have,

V̇ = f T JT pf + f T PJf
= f T JT P + PJ f

Let
Q = JT P + PJ
As V is positive definite, for the system to be asymptotically stable, Q
should be negative definite. If in addition V(x) is tending to infinity as norm
of x tends to infinity (∥x∥ → ∞) the system is asymptotically stable in the
large.
D R .S ARAVANAKUMAR G URUSAMY EACT631 79 / 91
 LYAPUNOV T HEOREM
Consider the system
ẋ = f(x)
Equilibrium state x0 (f(0) = 0)
A function V(x) is called a Lyapunov function if
1. V(x) > 0, ∀x ̸= 0
2. V(0) = 0 for x = 0
3. dV(x) dV(x)
dt = dx < 0
Then equilibrium state of the system (ẋ = f(x)) is stable.
Moreover, if the Lyapunov function satisfies

dV(x) dV(x)
< 0, ∀x ̸= 0 and =0⇔x=0
dt dt
Then equilibrium state of the system (ẋ = f(x)) is asymptotically stable.
The derivative of the Lyapunov function along the trajectory is negative.
The Lyapunov function may be considered as an energy function of the
system.
D R .S ARAVANAKUMAR G URUSAMY EACT631 80 / 91
 LYAPUNOV T HEOREM

▶ It is only sufficient for stability.

▶ In case of nonlinear system, we take a candidate V(x) ≥ 0 and check If
along the trajectories of ẋ = f (x), V(x) ≤ 0
▶ In case V(x) ≤ 0 then either,
▶ The equilibrium point is not asymptotically stable or
▶ We did not choose our candidate v(x) carefully. We have to look for
another V ( Lyapunov function).

D R .S ARAVANAKUMAR G URUSAMY EACT631 81 / 91

 LYAPUNOV C ONVERSE THEOREM

D R .S ARAVANAKUMAR G URUSAMY EACT631 82 / 91

 L A S ALLE ’ S T HEOREM - L A S ALLE ’ S I NVARIANCE
P RINCIPLE

D R .S ARAVANAKUMAR G URUSAMY EACT631 83 / 91

 PASSIVITY

D R .S ARAVANAKUMAR G URUSAMY EACT631 84 / 91

 K ALMAN –YAKUBOVICH –P OPOV L EMMA

D R .S ARAVANAKUMAR G URUSAMY EACT631 85 / 91

 P ROBLEM 1

Check the definiteness of the function: f (x) = x2 + 4xy + 2y2

Solution
 
1 2
1. Symmetry Check: Matrix is symmetric:
2 2
2. Eigenvalue Analysis: Solve |A − λI| = 0 to find eigenvalues: λ1 = 3,
λ2 = 0
3. Definiteness: All eigenvalues are positive ⇒ Positive definite.

D R .S ARAVANAKUMAR G URUSAMY EACT631 86 / 91

 P ROBLEM 2

Determine the definiteness of the function: g(x) = x2 − 6xy + 9y2

Solution
 
1 −3
1. Symmetry Check: Matrix is symmetric:
−3 9
2. Eigenvalue Analysis: Solve |A − λI| = 0 to find eigenvalues: λ1 = 10,
λ2 = 0
3. Definiteness: All eigenvalues are positive ⇒ Positive definite.

D R .S ARAVANAKUMAR G URUSAMY EACT631 87 / 91

 P ROBLEM 3

Examine the definiteness of the function: h(x) = x2 + 2xy + y2 − 4z2

Solution
 
1 1
1. Symmetry Check: Matrix is symmetric:
1 1
2. Eigenvalue Analysis: Solve |A − λI| = 0 to find eigenvalues: λ1 = 0,
λ2 = 2
3. Definiteness: One positive and one zero eigenvalue ⇒ Indefinite.

D R .S ARAVANAKUMAR G URUSAMY EACT631 88 / 91

 P ROBLEM 4

Investigate the definiteness of the function: k(x) = x2 − 4xy + 4y2 − z2

Solution
 
1 −2
1. Symmetry Check: Matrix is symmetric:
−2 4
2. Eigenvalue Analysis: Solve |A − λI| = 0 to find eigenvalues: λ1 = 0,
λ2 = 5
3. Definiteness: One positive and one zero eigenvalue ⇒ Indefinite.

D R .S ARAVANAKUMAR G URUSAMY EACT631 89 / 91

 P ROBLEM 5

Check the definiteness of the function: m(x) = −x2 − 6xy − 9y2

Solution
 
−1 −3
1. Symmetry Check: Matrix is symmetric:
−3 −9
2. Eigenvalue Analysis: Solve |A − λI| = 0 to find eigenvalues: λ1 = −10,
λ2 = 0
3. Definiteness: All eigenvalues are negative ⇒ Negative definite.

D R .S ARAVANAKUMAR G URUSAMY EACT631 90 / 91

 LYAPUNOV I NDIRECT M ETHOD IN S TABILITY
A NALYSIS OF N ONLINEAR S YSTEMS

▶ There are two conventional approaches in the stability analysis of

nonlinear systems:
• Lyapunov direct method
• Lyapunov indirect method or linearization approach
▶ The direct method analyzes stability of the system (equilibrium point)
using the nonlinear equations of the system
▶ The indirect method analyzes the system stability using the linearized
equations about the equilibrium point.

D R .S ARAVANAKUMAR G URUSAMY EACT631 91 / 91

 End of Chapter 2

D R .S ARAVANAKUMAR G URUSAMY EACT631 91 / 91