SISTEMAS

LINEARES

Prof. Dr. João Edgar Chaves Filho

Matrix polynomials and functions of matrices

• Matrix polynomials and functions of matrices have extensive applications in different

engineering problems such as modern control system design, vibration analysis, queueing
theory and communication systems

• Matrix Polynomials

• Consider the polynomial f(x) of degree m given by:

• If the scalar variable x is replaced by the n n matrix A, then the corresponding matrix polynomial
f(A) is defined by

• where Am= A*A*A*………*A, m times and I is the nxn identity matrix

• The polynomial f(x) can be written in factor form as:

Where 𝛼1, 𝛼2,……..,𝛼m are the roots of the polynomial f(x). Similarly, the matrix polynomial
f(A) can be factored as:

• Note that m, the degree of the polynomial f(x), can be greater than, equal to or less than n,
the size of matrix A.
• The MATLAB® command to factor polynomial f(x) is p =roots(a), where a =[a0 a1 … am] is
the vector of polynomial coefficients and p =[a0 a1 … am] is the vector of the roots of the
polynomial f(x).
Infinite Series of Matrices

• An infinite series of matrix A is defined as:

• where ak is a scalar fn k =0, 1, … and A is a real or complex nxn square

matrix.
Convergence of an Infinite Matrix Series
• It can be shown that the infinite matrix series S(A) converges if and only if
the scalar infinite series S(li) converges for all values of i, where li is the
ith eigenvalue of matrix A. For example, the geometric matrix series:

• Is a convergent series and converges to

• if and only if the scalar geometric s

:

converges when l is replaced by any eigenvalue of matrix

A
• The exponential matrix polynomial defined by:
Also converge ( 5.10)
Cayley–Hamilton Theorem
• Theorem 5.1: Any n n square matrix satisfies its own characteristic polynomial, that is:
P(A)=0
• Where P(l)=|lI-A| is the characteristic polynomial of the matrix A.

• Show that the following 2x2 matrix A satisfies its characteristic polynomial.

• The characteristic polynomial of A is:

Functions of Matrices

• There are several techniques available to find a function of a matrix.

1. Sylvester’s Expansion
• Theorem: Sylvester’s theorem.
• Sylvester’s theorem states that if f (A) is a matrix polynomial in A and if the eigenvalues of A are
distinct, and then f (A) can be decomposed as:

• where {li}i=1 n are the eigenvalues of the nxn matrix A and z(li) is an nxn matrix given by
Example

• Let A be the 2 2 matrix

• Using Sylvester‘s theorem, find

• a. f (A) =eAt if t>0

• b. f (A) =Ak

• The eigenvalues of A are l1 =0 and l2 = -1. Therefore:

• 2. Modal Matrix Technique
• Assume that the n n matrix A has n distinct eigenvalues and let M be the modal matrix of
A, then:
• (7)
• Let the Taylor series expansion of the function f(A) be:

• (8)
• Now substitute Equation 8 into 7
Special Matrix Functions
• Certain functions of matrices have applications in linear system theory, control and
communication. Among them are two functions frequently used in modern control system
design. These are eAt and Ak.
Matrix Exponential Function eAt
• The matrix exponential function eAt is a very important function with applications in analysis and
design of continuous time control systems. It is defined by the infinite series:

• The derivative of this function with respect to t is

• The integral of the exponential function eAt is computed by:

• The above closed form solution is valid if and only if matrix A is nonsingular

• a) First we find eAt using the modal matrix technique. The eigenvalues of matrix A are:
Example.

t  A k k
t
e At = 
 dτ = ?
Aτ
e
k =0 k!
0

t
A k t k +1

  k +1 k +1

0 e dτ = 
A t
= A 
-1
  ~ Assuming
Aτ

k = 0 (k + 1)!
 
k = 0 (k + 1)!
  that A-1 exists

 
( ) (eAt - I)A-1
 k k
A t
= A  
-1
+ I - I  = A -1 e At - I =
 k =1 k! 
t
 t

0 
  -1
= = -
Aτ Aτ At
e Bdτ e dτ  B (e I)A B
0 
➢ This will be used to compute the output response
under constant inputs.
21
• Matrix Function Ak
• Matrix function Ak is also useful in the analysis and design of discrete-time control systems. It is
defined as:

•
The State Space Modeling of Linear Continuous-time Systems

• Modern control theory, is based on the state space formulation of the system. One advantage of
modern control theory over the conventional approach is in its applicability to both single-input
single-output (SISO) and multiple-input multiple-output (MIMO) systems. Another advantage
of the state space formulation is that it is a unified approach which can be applied to both linear
as well as nonlinear dynamic systems.

• Concept of States
Definition: The state of a continuous-time dynamic system is the minimum number of variables
called states variables such that the knowledge of these variables at time t =t0, together with
input for t ≥ t0, uniquely determines the behavior of the system for t ≥ t0. If n variables are
needed, then these n variables are considered components of a n-dimensional vector x called
the state vector. The n-dimensional space whose coordinates are the states of the system is
called the state space.
State Equations of Continuous Time Systems

• The state space equations describing a multiple-input-multiple-output

(MIMO) continuous-time dynamic system have a general form given by:
• The state equations given are applicable to both linear and nonlinear, time
invariant (TI) and time varying systems. If the system is linear and time
invariant, then the state and output equations can be written as:
Solution of Continuous-time State Space Equations
• Solution of Homogenous State Equations and State Transition Matrix
The homogenous solution represents the solution of the state equations to an
arbitrary initial condition with zero input. The homogenous state equation is given
by

• Assuming that the initial state is x(0), then:

• where eAt is the matrix exponential function

Problem
• Descreva o sistema da figura abaixo em espaço de estados
The matrix exponential eAt is called the state transition matrix and is denoted
by f(t). The state transition matrix f(t) is an n n matrix and is the solution to
the homogenous equation:

The state transition matrix is given by:

Computing State
Transition Matrix
• There are several methods to
compute the state transition
matrix eAt as described before.
Here we discuss a new
approach based on the Laplace
transform.
• Find the state-transition matrix of the following system using the Laplace
transform approach:

• Using the Laplace transform, we have:

• Or
• Using the partial fraction expansion, the above matrix can be transformed as
follows:

• Hence
Complete Solution of State Equations

Consider the system:

•
x = Ax + Bu; y = Cx + Du

• Given x(0) and u(t) for t ≥ 0. The solution is

t
x(t) = e x(0) +  e A(t - τ) Bu( )d ;
At
0
t
y(t) = Ce x(0) +  Ce A(t - τ) Bu( )d + Du(t)
At
0
State Space Representation of Discrete-time Systems

• The general expressions for the state equation of a MIMO discrete-time

system are given by:
Solution of Discrete-time Equations
The DT system: x[k + 1] = Ax[k] + Bu[k]
y[k] = Cx[k] + Du[k]
Given x[0] and u(k), k 0, the solution is:
k -1
x[k] = A x[0] +  A k - m -1Bu[m]
k

m =0
k -1
y[k] = CA k x[0] +  CA k - m -1Bu[m] + Du[k]
m =0

The main problem involved is to compute Ak.

The system is internally stable iff all the eigenvalues
of A are within the unit disk: |li(A)| < 1 for all i
 x[k] → 0 if u[k]=0
43
Properties of State Transition Matrix
Computing the State Transition Matrix
• Here we discuss an approach based on the Z-transform.
Solution:
Next we find