Linear System Control

Stability

Unggul Wasiwitono

Mechanical Engineering Department

Faculty of Industrial Technology and Systems Engineering
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Introduction

U. Wasiwitono [Linear System Control] - Stability

2 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Introduction

1 Systems are designed to perform some tasks or to process signals.

2 Stability is a basic requirement for all systems.
3 Stability is one of the most important properties of dynamic systems.
4 There are several possible definitions of stability.
1 Most of them involve the notion of an equilibrium point.
5 The first is an internal notion of stability and involves the qualitative behavior of the zero-input
state response
1 the response of a related homogeneous state equation that depends solely on the initial state.
2 explicit stability criteria that involve the eigenvalues of the system dynamics matrix A.
3 energy-based stability analysis (Lyapunov stability theory)
6 The second type of stability focuses on external, or input-output, behavior.

U. Wasiwitono [Linear System Control] - Stability

3 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

U. Wasiwitono [Linear System Control] - Stability

4 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

Consider the nonlinear state equation

ẋ (t) = f [x (t)] x (0) = x0 (1)

for which equilibrium states are constant n × 1 vectors x̃ that satisfy f (x̃) = 0.
1 Stability of an equilibrium state refers to the qualitative behavior of trajectories that start in the
vicinity of the equilibrium state.
2 A nonlinear state equation can have multiple isolated equilibrium states each with different
stability properties.

U. Wasiwitono [Linear System Control] - Stability

5 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

U. Wasiwitono [Linear System Control] - Stability

6 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

Definition 1
The equilibrium state x̃ = 0 of Equation (1) is
Stable if, given any  > 0 there corresponds a δ > 0 such that kkx0 kk < δ implies that
kkx (t)kk <  for all t ≥ 0.
Unstable if it is not stable

U. Wasiwitono [Linear System Control] - Stability

7 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

U. Wasiwitono [Linear System Control] - Stability

8 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

Definition 2
The equilibrium state x̃ = 0 of Equation (1) is
Asymptotically stable if it is stable and it is possible to choose δ > 0 such that kx0 k < δ implies
that limt→∞ kx (t)k = 0. Specifically, given any  > 0, there exist T > 0 for which the
corresponding trajectory satisfies kx (t)k ≤  for all t ≥ T .
Globally asymptotically stable if it is stable and limt→∞ kx (t)k = 0 for any initial state.
Specifically, given any M > 0 and  > 0, there exists T > 0 such that kx0 k < M implies that the
corresponding trajectory satisfies kx (t)k ≤  for all t ≥ T .

U. Wasiwitono [Linear System Control] - Stability

9 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

U. Wasiwitono [Linear System Control] - Stability

10 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

Definition 3
The equilibrium state x̃ = 0 of Equation (1) is
Exponentially stable if there exist positive constants δ, k, and λ such that kx0 k < δ implies that
kx (t)k < ke−λt kx0 k for all t ≥ 0.
Globally exponentially stable if there exist positive constants k and λ such that
kx (t)k < ke−λt kx0 k for all t ≥ 0 for all initial states.

U. Wasiwitono [Linear System Control] - Stability

11 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

Definition 3
The equilibrium state x̃ = 0 of Equation (1) is
Exponentially stable if there exist positive constants δ, k, and λ such that kx0 k < δ implies that
kx (t)k < ke−λt kx0 k for all t ≥ 0.
Globally exponentially stable if there exist positive constants k and λ such that
kx (t)k < ke−λt kx0 k for all t ≥ 0 for all initial states.

An equilibrium state is stable if the state trajectory can be made to remain as close as
desired to the equilibrium state for all time by restricting the initial state to be sufficiently
close to the equilibrium state.

U. Wasiwitono [Linear System Control] - Stability

11 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability
Our ultimate focus is on the homogeneous linear time-invariant state equation

ẋ (t) = Ax (t) x (0) = x0 (2)

for which x̃ = 0 ∈ Rn is seen easily to be an equilibrium state.

Definition 4
The equilibrium state x̃ = 0 of Equation (2) is
Stable if there exists a finite positive constant γ such that for any initial state x0 the
corresponding trajectory satisfies kx (t)k ≤ γ kx0 k for all t ≥ 0.
Unstable if it is not stable.
(Globally) asymptotically stable if given anyµ > 0 there exists T > 0 such that for any initial
state x0 the corresponding trajectory satisfies kx (t)k ≤ µ kx0 k for all t ≥ T .
(Globally) exponentially stable if there exist positive constants k and λ such that that for any
initial state x0 the corresponding trajectory satisfies kx (t)k < ke−λt kx0 k for all t ≥ 0.

U. Wasiwitono [Linear System Control] - Stability

12 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Internal Stability

Theorem 5
The equilibrium state x̃ = 0 of Equation (2) is:
Stable if and only if all eigenvalues of A have a nonpositive real part and if eigenvalues on
the imaginary axis are simple (nonrepeated)..
(Globally) asymptotically stable if and only if every eigenvalue of A has strictly negative real
part.

U. Wasiwitono [Linear System Control] - Stability

13 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Example
Consider a continuous time 4th order system with three different eigenvalue locations
  
 −1.5  0  0
a. −2 b. −2 c. 0
−0.3 ± j1.5 −0.3 ± j1.5 −0.3 ± j1.5
  

The matrices are

   
0 1 0 0 0
 0 0 1 0   0   
A= , B =  , C = 1 0 0 0 , D = [0]
 0 0 0 1   0 
−a0 −a1 −a2 −a3 1

The characteristic polynomial of A is

P (A) = λ4 + a3 λ3 + a2 λ2 + a1 λ + a0

U. Wasiwitono [Linear System Control] - Stability

14 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Example

a3 a2 a1 a0
System a 4.1 7.44 9.99 7.02
System b 2.6 3.54 4.68 0
System c 0.6 2.34 0 0

1 System a has all it’s eigenvalues strictly in the left half plane and it is therefore asymptotically
stable.
2 System b is stable i.s.L. because it has an eigenvalue on the imaginary axis.
3 System c has a double eigenvalue on the imaginary axis and hence it is neither
asymptotically stable nor stable i.s.L. It is unstable.

U. Wasiwitono [Linear System Control] - Stability

15 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Example

Figure 1: System a
U. Wasiwitono [Linear System Control] - Stability
16 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Example

Figure 2: System b
U. Wasiwitono [Linear System Control] - Stability
17 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Example

Figure 3: System c
U. Wasiwitono [Linear System Control] - Stability
18 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Energy-Based Analysis

U. Wasiwitono [Linear System Control] - Stability

19 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Energy-Based Analysis
Consider the second-order linear translational mechanical system, which for zero
external applied force is governed by

mÿ (t) + cẏ (t) + ky (t) = 0

in which y(t) represents the displacement of both the mass and spring from rest.
Chosing the state variables x1 (t) = y (t) and x2 (t) = ẏ (t), yielding the homogeneous
state equation     
ẋ1 0 1 x1
= k c
ẋ2 −m −m x2
These state variables are related to energy stored in this system.
The spring displacement characterizes the potential energy stored in the spring, and
The mass velocity characterizes the kinetic energy stored in the mass.

U. Wasiwitono [Linear System Control] - Stability

20 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Energy-Based Analysis

Therefore the total energy stored in the system can expressed by the function

1 2 1
E (x1 , x2 ) = kx + mx22
2 1 2

The system energy is positive whenever

T T
[x1 , x2 ] 6= [0, 0]

Attains the minimum value of zero at the equilibrium state

T T
[x̃1 , x̃2 ] = [0, 0]

U. Wasiwitono [Linear System Control] - Stability

21 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Energy-Based Analysis
Evaluating the energy function along a system trajectory by computing the time
derivative
 
d d 1 2 1
E [x1 (t) , x2 (t)] = kx1 (t) + mx22 (t) = kx1 (t) ẋ1 (t) + mx2 (t) ẋ2 (t)
dt dt 2 2
 
k c
= kx1 (t) [x2 (t)] + mx2 (t) − x1 (t) − x2 (t) = −cx22 t
m m

dE
1 For zero damping (c = 0) we have dt
≡ 0, so the total system energy is constant along any
trajectory.
dE
2 For positive damping (c > 0), we have dt
< 0 along any trajectory for which the mass velocity
is not identically zero.

The total energy in the system is strictly decreasing along all other trajectories and
converges to zero as time tends to infinity.
U. Wasiwitono [Linear System Control] - Stability
22 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Lyapunov Stability Analysis

U. Wasiwitono [Linear System Control] - Stability

23 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Lyapunov Stability Analysis

1 The Russian mathematician A. M. Lyapunov (1857–1918) observed that conclusions regarding

stability of an equilibrium state can be drawn from a more general class of energy-like
functions.
2 Lyapunov methods: very general methods to prove (or disprove) stability of nonlinear systems
3 Lyapunov’s stability theory is the single most powerful method in stability analysis of nonlinear
systems.

U. Wasiwitono [Linear System Control] - Stability

24 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Lyapunov Stability Analysis

The Lyapunov method is based on the notion of positive definite functions of the state
vector x.
The continuous scalar function W (x) is positive definite if and only if
1 W (0) = 0,
2 W (x) > 0 for all non zero x and
3 W (x) → ∞ for kxk → ∞

The scalar time dependent functions V (x, t) is positive definite if W (x) is positive definite
and if and only if V (x, t) ≥ W (x) for all x and t ≥ t0

U. Wasiwitono [Linear System Control] - Stability

25 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Lyapunov Stability Analysis

Consider nonlinear system
ẋ (t) = f (x (t) , t) (3)

Theorem 6
The system (3) is stable if V (x, t) is a positive definite function and if and only if

d
V (x, t) ≤ 0
dt

U. Wasiwitono [Linear System Control] - Stability

26 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Lyapunov Stability Analysis

Consider nonlinear system
ẋ (t) = f (x (t) , t) (3)

Theorem 6
The system (3) is stable if V (x, t) is a positive definite function and if and only if

d
V (x, t) ≤ 0
dt

Theorem 7
The system (3) is asymptotically stable if V (x, t) is a positive definite function and if and
only if
d
V (x, t) < 0
dt

U. Wasiwitono [Linear System Control] - Stability

26 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Lyapunov Stability Analysis

1 For the homogeneous LTI system,
ẋ (t) = Ax (t) (4)
a quadratic form can be formed by choosing a symmetric, constant matrix P ,

V (x) = xT P x

If P is positive definite, V (x) is a Lyapunov function candidate.

2 For the continuous time cases respectively,
d
V (x) = ẋT P x + xT P ẋ = xT AT P x + xT P Ax
dt  
= xT AT P + P A x

Applying theorem (7) it can now be seen that if it is possible to find another positive definite
matrix Q such that
AT P + P A = −Q
then the systems (4) are asymptotically stable.
U. Wasiwitono [Linear System Control] - Stability
27 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Matlab for Stability Analysis

The following MATLAB function is useful for Lyapunov stability analysis:

lyap(A’,Q) Solve A’P+PA=-Q for matrix P, given a positive-definite matrix Q.

U. Wasiwitono [Linear System Control] - Stability

28 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Input–Output Stability

U. Wasiwitono [Linear System Control] - Stability

29 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Input–Output Stability

1 Consider a SISO linear time-invariant (LTI) system described by

Z t Z t
y (t) = g (t − τ ) u (τ ) dτ = g (τ ) u (t − τ ) dτ (5)
0 0

where g (t) is the impulse response or the output excited by an impulse input applied at t = 0.
2 An input u(t) is said to be bounded if u(t) does not grow to positive or negative infinity or,
equivalently, there exists a constant um such that

|u (t)| ≤ um < ∞ for all t ≥ 0

A system is said to be BIBO stable (bounded-input bounded-output stable) if every bounded

input excites a bounded output.
3 This stability is defined for the zero-state response and is applicable only if the system is initially
relaxed.

U. Wasiwitono [Linear System Control] - Stability

30 / 33
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

Stability of Linear System

1 The state space model is an internal model and the stability investigation based on the state
model will reveal whether or not the system is internally stable.
2 For LTI systems this property is determined by the eigenvalues of the system matrix.
3 If the model is represented by a transfer function matrix then one is dealing with an external
model and the placement of the poles gives information on the system’s external stability.
4 If an unstable eigenvalue/ pole is cancelled by a zero, the system may be externally stable
although it is internally unstable.
5 Conversely, an internally stable LTI-system will also be externally stable.

U. Wasiwitono [Linear System Control] - Stability

31 / 33
Question ?
 Introduction Internal Stability Energy-Based Analysis Lyapunov Stability Analysis Input–Output Stability

U. Wasiwitono [Linear System Control] - Stability

33 / 33