Stability
EE-601
Linear System Theory
Prof. Bidyadhar Subudhi
School of Electrical Sciences
Indian Institute of Technology Goa
Contents
• Introduction
• Input-output stability of LTI systems
• Internal stability
• Lyapunov theorem
• Stability of LTV systems
1
Stability
• Stability of a system can be thought as a
continuity in its dynamic behaviour. If a small
perturbation arises in the system inputs or
initial conditions, a stable system will present
small modifications in its perturbed response
• Stability is a basic requirement of dynamic
systems that perform operations or process
signals; the first objective in control design
In an unstable system, any perturbation, no matter how small, will make
states or outputs grow unbounded or until the system disintegrates or
saturates.
2
Input-Output Stability of LTI Systems
• Response of a LTI system: due to initial conditions +
response to inputs
• Concept of Input-Output Stability refers to stability of
the response to inputs only, assuming zero initial
conditions
• BIBO Stability: A system is BIBO (bounded-input
bounded-output) stable if every bounded input
produces a bounded output
3
Input-Output Stability
• Consider a SISO LTI system
𝑡 𝑡
y(t) = 0
g(t )u(τ)d = 0
g(τ)u(t )d
system is causal and relaxed at 0
• An input u(t) is said to be bounded if there exist constant um such
that, u(t) ≤ 𝑢𝑚 < , ∀t ≥ 0
• A system is said to be BIBO stable if every bounded input excites
a bounded output (for zero initial state).
4
Theorem(BIBO Stability and Impulse Response)
• A SISO system is BIBO stable if and only if g(t) is absolutely
integrable in [0, ), or
0
g(t) dt ≤ M < for some constant M≥ 0
Proof If g(t) is absolutely integrable, every bounded input excites a
bounded output
• Let u(t) be an arbitrary input: u(t) ≤ 𝑢𝑚 < , ∀t ≥ 0
𝑡 𝑡
y(t) = 0
g(τ)u(t )d ≤ 0
g(τ) u(t ) d
≤ 𝑢𝑚 g(τ) d ≤ 𝑢𝑚 M
0
Thus output is bounded.
5
• Let g(t) is not absolutely integrable
𝑡1
g(τ) d = ∞
0
1, 𝑖𝑓 g(τ) ≥ 0
𝑢 𝑡1 −τ = 𝑓 𝑥 =
−1, 𝑖𝑓 g(τ) < 0
𝑡1 𝑡1
𝑦 𝑡1 = g(𝑡1 −τ)𝑢 𝑡1 −τ = 𝑔(𝑡1 − 𝜏) =
0 0
Which is not bounded
6
Theorem(BIBO Stability and Steady-State Response)
If a system with impulse response g(t) is BIBO stable, then, as
t:
[Link] output excited by u(t) = a, for t 0, approaches g 0 . 𝑎
2. The output excited by 𝑢(𝑡) = sin0 𝑡, for 𝑡 ≥ 0,
approaches g (j0) sin(0t g ( j0 ))
g (s) :Laplace transform of g(t)
g (s) = 0
g(τ)𝑒 −𝑠𝑡 d
This is a basic result: specifies the response of a BIBO system to constant and
sinusoidal signals once the transients have extinguished. Filtering of signals is
based essentially on this theorem.
7
Input-Output Stability of LTI Systems
𝑠+1
Consider transfer function𝐺 𝑠 = 𝑆 2 + 15𝑠 + 26
Poles at 𝑠 = −13 𝑎𝑛𝑑 𝑠 = −2 so it is BIBO stable
8
Theorem (BIBO Stability)
A SISO system with proper rational transfer g (s) is BIBO stable if
and only if every pole of g (s) has a negative real part or,
equivalently, lies inside the left-half s-plane.
• If g (s) has pole 𝑝𝑖 with multiplicity 𝑚𝑖 , then its partial fraction
1 1 1
expansion contains the factors ,
𝑠−𝑝𝑖 (𝑠−𝑝𝑖 )2,…, 𝑠−𝑝𝑖 𝑚𝑖 .
• Inverse Laplace transform of g (s) or the impulse contains the factors
𝑒 𝑝𝑖𝑡 , 𝑡𝑒 −𝑝𝑖 𝑡 ,…, 𝑡 𝑚𝑖−1 𝑒 𝑝𝑖𝑡
• Every such term is absolutely integrable if and only if 𝑝𝑖 has a
negative real part.
9
MIMO System
Theorem: A MIMO LTI system with impulse response matrix
𝐺(𝑡) = [𝑔𝑖𝑗(𝑡)] is BIBO stable if and only if every 𝑔𝑖𝑗(𝑡) is
absolutely integrable in 0, ∞ for all i, j.
Theorem: A MIMO system with proper rational transfer matrix
Ĝ (𝑠) = [ĝ𝑖𝑗(𝑠)] is BIBO stable if and only if every pole of every
ĝ𝑖𝑗(𝑠) has a negative real part.
10
BIBO Stability (State Space Description)
• Consider 𝑥 = 𝐴𝑥 + 𝐵𝑢
𝑦 = 𝐶𝑥 + 𝐷𝑢
−1 1
• Ĝ 𝑠 = C(𝑠𝐼 − 𝐴) 𝐵 + 𝐷 = 𝐶 𝐴𝑑𝑗 𝑠𝐼 − 𝐴 𝐵 + 𝐷
det(𝑠𝐼−𝐴)
• Every pole of Ĝ (𝑠) is an eigen value of A, but eigenvalue of A
might not be a pole because of cancellation
11
Input Output Stability of LTI Systems
Note that not every eigenvalue of A is a pole of G(s), since there may
be pole-zero cancellations while computing G(s).
Thus, a state equation may be BIBO stable even when some
eigenvalues of A do not have negative real part.
Example
1 10 2 has one eigenvalue with
x(t ) = x(t ) u (t ) positive real part = 1, it
0 1 0 is BIBO stable, since its
y (t ) = 2 3x(t ) 2u (t ) transfer function
21 s
G ( s ) = C sI A B D =
1 has a single pole at s = -1.
s 1
12
Internal Stability of LTI Systems
Concept of Internal Stability: refers to stability of the system
response to initial conditions only, assuming zero inputs.
Stability of the response of the state equation
with x(t ) = Ax(t ), x ( 0) = x 0
Because the solution of this equation is
given by x(t ) = e At x 0
stability is determined by eigenvalues of A (state response to
initial conditions)
If the eigenvalues of A have all negative real part, the system
response will decay to 0 as t
13
Internal Stability of LTI Systems
Lyapunov Stability: the system 𝑥 𝑡 = Ax(t) is a Lyapunov
stable, or marginally stable, or simply stable, if every finite initial
state 𝑥0 excites a bounded response x t .
Asymptotic Stability: The system 𝑥 𝑡 = Ax(t) is asymptotically
stable if every finite initial state 𝑥0 excites a bounded response
x t that approaches 0 as t→ ∞.
Instability: The system 𝑥 𝑡 = Ax(t) is unstable if it is not stable.
14
Internal Stability of LTI Systems
Theorem (Internal Stability): The equation 𝑥 (𝑡) = 𝐴𝑥(𝑡) is
1. Lyapunov stable if and only if all the eigenvalues of A have zero
or negative real parts, and those with zero real part are associated
with a Jordan block of order 1.
2. Asymptotically stable if and only if all eigenvalues of A have
negative real parts.
0 0 0
x t = 0 0 0 x t
0 0 1
A has eigenvalues 0 = 0 with multiplicity 2, and
1 = −1 with multiplicity 1. eigenvalue 1 = −1 is
associated to Jordan blocks of order 1, so the equation
is Lyapunov stable.
15
Example 0 1 0
x t = 0 0 0 x t
0 0 1
A has the same eigenvalues and same multiplicities of the
previous example. However, the repeated eigenvalue𝜆1 = 0 is
associated to a Jordan block of order 2, so the equation is
unstable. 0 1 0 1 t 0 x1 (0)
t x(0) = 0 1 0 x (0)
solution of this equation: x (t ) = exp 0 0 0 2
0 0 -1 0 0 e-t x3 (0)
x1 (0) tx2 (0)
= x2 (0)
e-t x3 (0)
from which we see that 𝑥1 grows unbounded if 𝑥2 (0) ≠ 0
16
Internal Stability of LTI
Systems
A block diagram interpretation of the difference between the
two systems
0 0 0 1
x(t ) = x(t ) x(t ) = x(t )
0 0 0 0
the system represents two decoupled
the system represents two decoupled integrators
integrators
x1 (t ) x1 (t )
ʃ ʃ x2 (t) ʃ ʃ x2 (t)
x1 (0) x2(0) x1 (0) x2(0)
17
Internal Stability of LTI
Systems
−1
Every pole of the transfer matrix 𝐺 𝑠 = 𝑠𝐼 − 𝐴 𝐵+𝐷
is an eigenvalue of A.
Asymptotic stability⟹BIBO stability
However, not all eigenvalues of A are necessarily poles of
𝐺(𝑠),
However, not all eigenvalues of A are necessarily poles of 𝐺(𝑠),
Hence asymptotic stability ⇍ BIBO stability
Hence asymptotic stability ⇍ BIBO stability
18
Discrete Time Case(I/O Stability)
• Consider a discrete-time SISO system
𝑔[𝑘]: impulse response sequence or the output sequence
excited by an impulse sequence applied at 𝑘 = 0.
• An input sequence u[k] is said to be bounded if there exists
constant um such that, 𝑢[𝑘] ≤ 𝑢𝑚 < ∞, for 𝑘 = 0,1,2 …
Definition: A system
𝑘 𝑘
𝑦𝑘 = 𝑚=0 𝑔 𝑘−𝑚 𝑢 𝑚 = 𝑚=0 𝑔 𝑚 𝑢 𝑘−𝑚
is said to be BIBO stable if every bounded input excites a
bounded output sequence (under zero initial state).
19
Theorem (Discrete-time BIBO
stability and impulse response)
A discrete-time SISO system
𝑘 𝑘
𝑦𝑘 = 𝑔 −𝑚 𝑢 𝑚 = 𝑔 𝑚 𝑢 𝑘−𝑚
𝑚=0 𝑚=0
is BIBO stable if and only if g[k] is absolutely summable in [0, ∞) or
∞
𝑔[𝑘] ≤ 𝑀 for some constant 𝑀 ≥ 0
𝑘=0
Theorem :
If a discrete-time system with impulse response sequence 𝑔[𝑘] is BIBO
stable, then, as k :
1. The output excited by 𝑢[𝑘] = 𝑎, for k 0, approaches g 1 . a .
2. The output excited by u[k] = sin𝜔0 𝑘, for k 0, approaches
𝑔 𝑒 𝑗𝜔0 sin(𝜔0 𝑘 + ∠𝑔 𝑒 𝑗𝜔0 )where ĝ(z) is the z-transform of g[k].
20
Theorem:
• A discrete-time SISO system with proper rational transfer
function ĝ(𝑧) is BIBO stable if and only if every pole of ĝ(z)
has a magnitude less than 1, or lies inside the unit circle on
the z-plane.
• If g (z) has pole 𝑝𝑖 with multiplicity 𝑚𝑖 , then its partial
1
fraction expansion contains the factors ,
𝑧−𝑝𝑖
1 1
,…, .
(𝑧−𝑝𝑖 )2 𝑧−𝑝𝑖 𝑚𝑖
• Thus the inverse z - transform of 𝑔 𝑧 or the impulse
contains the factors 𝑝𝑖 𝑘 , 𝑘𝑝𝑖 𝑘 ,…, 𝑘 𝑚𝑖 −1 𝑝𝑖 𝑘
• Every such term is absolutely summable if and only if
𝑝𝑖 < 1
21
Theorem:
• A MIMO discrete-time system with impulse response
sequence matrix 𝐺[𝑘] = [𝑔𝑖𝑗[𝑘]] is BIBO stable if and only if
every 𝑔𝑖𝑗[𝑘] is absolutely summable.
Theorem:
• A MIMO discrete-time system with discrete proper rational
transfer matrix Ĝ (𝑧) = [ĝ𝑖𝑗(𝑧)] is BIBO stable if and only if
every pole of every ĝ𝑖𝑗(𝑧) has a magnitude less than 1.
22
State Space Description
𝑥 𝑘 + 1 = 𝐴𝑥 𝑘 + 𝐵𝑢[𝑘]
• Consider
𝑦 𝑘 = 𝐶𝑥 𝑘 + 𝐷𝑢[𝑘]
1
• Ĝ 𝑧 = C(𝑧𝐼 − 𝐴)−1 𝐵 +𝐷 = 𝐶 𝐴𝑑𝑗 𝑧𝐼 − 𝐴 𝐵 + 𝐷
det(𝑧𝐼−𝐴)
• Every pole of Ĝ (𝑧) is an eigen value of A, is contained inside the
unit circle for BIBO system to be stable.
23
Internal Stability(for zero-input response)
• The stability of zero-input response of 𝑥(𝑡) = 𝐴𝑥(𝑡) is
marginally stable or stable in the sense of Lyapunov if every finite
initial state x0 excites a bounded response.
• It is asymptotically stable if every finite initial state excites a
bounded response, which, in addition, approaches 0 as t and
the solution of 𝑥(𝑡) is 𝑒𝐴𝑡𝑥0
Theorem:
𝑥(𝑡) = 𝐴𝑥(𝑡) is marginally stable if and only if all eigenvalues of
A have zero or negative real parts.
𝑥(𝑡) = 𝐴𝑥(𝑡) is asymptotically stable if and only if all eigenvalues
of A have negative real parts.
24
Discrete-Time Case (Internal Stability)
• The internal stability of discrete-time system: the stability of
𝑥[𝑘 + 1] = 𝐴𝑥[𝑘].
• Its solution is 𝑥[𝑘] = 𝐴𝑘 𝑥0.
• x[k+1] = Ax[k] is marginally stable if and only if all eigenvalues
of A have magnitude less than or equal to 1.
• x[k+1] = Ax[k] is asymptotically stable if and only if all
eigenvalues of A have magnitude less than 1.
25