Assignment EE-313

Tutorial
November 2023

Tutorial on Internal Stability

Problem:
Consider a real symmetric matrix A with eigenvalues λ1 , λ2 , . . . , λn . The
eigenvalues are ordered such that λ1 ≥ λ2 ≥ . . . ≥ λn .
Rayleigh’s inequality states that for any nonzero vector v, the following
inequality holds:

v T Av
λn ≤ ≤ λ1
vT v
Part A:
Let A be the following symmetric matrix:
 
4 −1 0
A = −1 4 −1
0 −1 4
Find the eigenvalues λ1 , λ2 , λ3 of matrix A.
Part B:
Use Rayleigh’s inequality to find lower and upper bounds on the eigenvalues
1
of matrix A using the vector v = 1.
1
Part C:
Verify your results in Part B by finding the actual eigenvalues of matrix A
and comparing them with the bounds obtained from Rayleigh’s inequality.
Solution:
Part A: Eigenvalues of Matrix A
The matrix A is given by:
 
4 −1 0
A = −1 4 −1
0 −1 4
To find the eigenvalues, we solve the characteristic equation |A − λI| = 0,
where I is the identity matrix:

1
 det(A − λI) = 0
Solving this equation yields the eigenvalues λ1 , λ2 , λ3 :

λ1 = 5, λ2 = 4, λ3 = 3
Part B: Rayleigh’s Inequality Bounds
Rayleigh’s inequality for a symmetric matrix A and a nonzero vector v is
given by:

v T Av
λn ≤ ≤ λ1
vT v
 
1
Using v = 1, we can calculate the bounds:
1
v T Av
λ3 ≤ T ≤ λ1
v v
Substitute the values and simplify to obtain the bounds:

v T Av
3≤ ≤5
vT v
Part C: Verification
Verify the obtained bounds by finding the actual eigenvalues of matrix A:

Eigenvalues of A : λ1 = 5, λ2 = 4, λ3 = 3
The actual eigenvalues lie within the bounds obtained from Rayleigh’s in-
equality, confirming the validity of the bounds.
Problem:
Consider a continuous-time linear time-invariant system described by the
state-space representation:

ẋ(t) = Ax(t),
n
where x(t) ∈ R is the state vector, and A is an n × n constant matrix.
Assume that the system is stable.
Part A: Show that if there exists a symmetric positive definite matrix P
satisfying the Lyapunov equation:

AT P + P A = −Q,
where Q is a symmetric positive definite matrix, then the system is stable.
Part B: Consider the following system with the matrix A:
 
−1 0
A=
0 −2

2
 Find a symmetric positive definite matrix P that satisfies the Lyapunov
equation AT P + P A = −Q, where:
 
1 0
Q=
0 2
Part C: Determine the stability of the system described by matrix A in
Part B using the Lyapunov stability criterion and the matrix P obtained.
Solution:
Part A: Stability via Lyapunov Equation
If there exists a symmetric positive definite matrix P satisfying the Lyapunov
equation AT P + P A = −Q, where Q is symmetric positive definite, then the
system is stable.
Part B: Finding Matrix P
For the system with A:
 
−1 0
A=
0 −2
and Q:
 
1 0
Q=
0 2
we need to find a symmetric positive definite matrix P that satisfies AT P +
P A = −Q. The solution to this Lyapunov equation is obtained by solving for
P.
Part C: Stability Analysis
Determine the stability of the system using the Lyapunov matrix P obtained
in Part B. Apply the Lyapunov stability criterion to check whether the matrix
A is stable.
Finding Matrix P
For the system with A:
 
−1 0
A=
0 −2
and Q:
 
1 0
Q=
0 2
we need to find a symmetric positive definite matrix P that satisfies the
Lyapunov equation AT P + P A = −Q. The solution to this Lyapunov equation
is obtained by solving for P .
The Lyapunov equation is:

 T       
−1 0 p11 p12 p p12 −1 0 1 0
T
A P + PA = + 11 =−
0 −2 p21 p22 p21 p22 0 −2 0 2

3
 Calculate the left-hand side:

       
−1 0 p11 p12 p p12 −1 0 −2p11 −2p12
T
A P +P A = + 11 =
0 −2 p21 p22 p21 p22 0 −2 −p21 −4p22

Set this equal to the negative of Q:

   
−2p11 −2p12 1 0
=−
−p21 −4p22 0 2
Solving for the elements of P :
1 1
p11 = , p12 = 0, p21 = 0, p22 =
2 2
So, the matrix P that satisfies the Lyapunov equation is:
1 
2 0
P =
0 12
This matrix P is symmetric positive definite and satisfies the Lyapunov
equation. Therefore, the system is stable.
Problem: Consider a practical example of a stability analysis using Lya-
punov theory for a linear perturbed system. Imagine a simple pendulum that is
subject to small perturbations. The governing equation for a simple pendulum
without perturbations is given by:
g
θ̈ + sin(θ) = 0
L
where: - θ̈ is the angular acceleration, - g is the acceleration due to gravity,
- L is the length of the pendulum, - θ is the angular displacement.
Now, introduce a small perturbation to this system by adding a disturbance
term d(t) to the equation:
g
θ̈ +sin(θ) = d(t)
L
This perturbation term could represent external forces or disturbances acting
on the pendulum.
Analyze the stability of the perturbed system using Lyapunov theory.

Solution:
Linearize the system around its equilibrium point, which is usually the down-
ward vertical position (θ = π), and use Lyapunov’s stability criterion.
The linearized system is:
g
θ̈ + (θ − π) = d(t)
L
Define a Lyapunov function V (θ, θ̇) as follows:

4
 1 2 g
θ̇ + (θ − π)2
V (θ, θ̇) =
2 L
The time derivative of V along the trajectories of the perturbed system is:
g
V̇ = − (θ − π)θ̇ + θ̇d(t)
L
Choose the disturbance term d(t) such that V̇ is negative definite:
g
V̇ = − (θ − π)θ̇ − k θ̇
L
where k is a positive constant. By Lyapunov’s stability criterion, the per-
turbed simple pendulum system is stable when the disturbance term is d(t) =
−k θ̇ with k > 0.
Problem: Consider the following system of nonlinear ordinary differential
equations:

ẋ1 = x2
ẋ2 = −4x1 − 2x2 + βx32 , β≥0
Analyze the stability by considering the linear and perturbed part as g(x) =
[0, βx32 ]. Estimate the region of attraction.
Solution: To analyze the stability, let’s first consider the linearized system
around the equilibrium point (0, 0). The linearized system is obtained by ne-
glecting the nonlinear terms:

ẋ1 = x2
ẋ2 = −4x1 − 2x2
The matrix form of this system is Ẋ = AX, where X = [x1 , x2 ]T and
 
0 1
A=
−4 −2
The eigenvalues of matrix A are complex with negative real parts, indicating
stability:
√
λ = −1 ± 7i
Next, consider the perturbed part g(x) = [0, βx32 ]. The Lyapunov function
chosen is V (x) = 21 x21 + 12 x22 . The time derivative of V along the trajectories of
the perturbed system is:

V̇ (x) = −3x1 x2 + βx42

For stability, we want −3x1 x2 + βx42 < 0. The stability depends on the
value of β. If β is sufficiently small, the system will remain stable, and the
perturbation will not destabilize the equilibrium point.

5
 In conclusion, the linearized system is stable, and the stability of the per-
turbed system depends on the value of β. For small values of β, the equilibrium
point is likely to remain stable. However, for larger values of β, further analysis
or numerical simulations may be needed to assess stability.