You are on page 1of 36

Outline

Modern Control systems


Lecture-3. Solution of State Equations

V. Sankaranarayanan

V. Sankaranarayanan Modern Control systems


Outline

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

V. Sankaranarayanan Modern Control systems


Outline

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

2 State Transition Matrix


Properties of State Transition Matrix

V. Sankaranarayanan Modern Control systems


Outline

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

2 State Transition Matrix


Properties of State Transition Matrix

3 Computational Methods of Matrix Exponential


Laplace Transformation Approach
Diagonal Transformation
Cayley-Hamilton Theorem Approach

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

2 State Transition Matrix


Properties of State Transition Matrix

3 Computational Methods of Matrix Exponential


Laplace Transformation Approach
Diagonal Transformation
Cayley-Hamilton Theorem Approach

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution to Homogeneous State Equations

Solution to Scalar D.E.s

Let us consider the scalar differential equation,


ẋ = ax, where
x(t) = b0 + b1 t + b2 t2 + · · · + bk tk + · · ·
On substituting x(t) in our scalar differential Equation, we get..

b1 + 2b2 t + 3b3 t2 + · · · + kbk tk−1 + · · · =


a(b0 + b1 t + b2 t2 + · · · + bk tk + · · · )

Equating the coefficients of equal powers of ’t’

b1 = ab0
1 1
b2 = ab1 = a2 b0
2 2
1 1 3
b3 = ab2 = a b0
3 2∗3
..
.
1 k
bk = a b0
k!
V. Sankaranarayanan Modern Control systems
Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Continuation of Solution of scalar D.E.s...


The value of b0 can be determined by substituting t = 0 in
x(t) = b0 + b1 t + b2 t2 + · · · + bk tk + · · ·
x(0) = b0

Hence the solution x(t) can be written as,


1 2 2 1 3 3 1 k k
x(t) = (1 + at + 2! a t + 3! a t + · · · + k! a t + · · · )b0
1 2 2 1 3 3 1 k k
= (1 + at + 2!
a t + 3!
a t + ··· + k!
a t + · · · )x(0)

= eat x(0)

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

2 State Transition Matrix


Properties of State Transition Matrix

3 Computational Methods of Matrix Exponential


Laplace Transformation Approach
Diagonal Transformation
Cayley-Hamilton Theorem Approach

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Solution of Vector-matrix D.E.s


Let us consider the vector differential equation,
ẋ = Ax,

where, x ∈ Rn → n-vector
A ∈ Rn∗n → n ∗ n constant matrix

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Solution of Vector-matrix D.E.s


Let,

ẋ1 (t) = a11 x1 (t) + a12 x2 (t) + · · · + a1n xn (t) (1)


ẋ2 (t) = a21 x1 (t) + a22 x2 (t) + · · · + a2n xn (t)
..
.
..
.
ẋn (t) = an1 x1 (t) + an2 x2 (t) + · · · + ann xn (t)

This can be written in Matrix form as,

ẋ1 (t) a11 ··· an1 x1 (t)


    
 .   . .. ..   .. 
 ..  =  .. . .  . 
ẋn (t) an1 ··· ann xn (t)

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Solution of Vector D.E.s


Let,

x1 (t) = a0 + a1 t + a2 t2 + · · · + an tn (2)
x2 (t) = b0 + b1 t + b2 t2 + · · · + bn tn (3)
x3 (t) = c0 + c1 t + c2 t2 + · · · + cn tn (4)
..
.

Consider the equation,


x1 (t) = a0 + a1 t + a2 t2 + · · · + an tn
Differentiating with respect to t, we get,
x˙1 (t) = a1 + 2a2 t + 3a3 t2 + · · · + nan tn−1

Substituting this in equation (1) , we get,


a1 + 2a2 t + 3a3 t2 + · · · + nan tn−1 =
a11 (a0 + a1 t + a2 t2 + · · · + an tn ) + a12 (b0 + b1 t + b2 t2 + · · · + bn tn )+
a13 (c0 + c1 t + c2 t2 + · · · + cn tn ) + ....

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Solution of Vector-matrix D.E.s


Equating coefficients of equal powers of ’t’,

a1 = a11 a0 + a12 b0 + a13 c0 + ...


b1 = a21 a0 + a22 b0 + a23 c0 + ...
c1 = a31 a0 + a32 b0 + a33 c0 + ...
..
.

Similarly,
a2 = a11 a1 + a12 b1 + a13 c1 + ..

= a11 (a11 a0 + a12 b0 + a13 c0 + ...) + a12 (a21 a0 + a22 b0 + a23 c0 + ...) + · · ·

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Solution of Vector-matrix D.E.s


Substituting t = 0 in equations (2), (3), · · · we get

x1 (0) = a0
x2 (0) = b0
x3 (0) = c0
..
.

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Solution of Vector-matrix D.E.s


Summing up all the results obtained so far, we get

x1  a0  a0
     
1 ··· 0 a11 ··· a1n
 
x2  b  b 
  . .. ..   0  +  .. .. ..   0  t +
x3  =  .. . c
.  0  . . c 
.  0
 
.. .. ..
 
0 ··· 1 an1 ··· ann
. . .
2 a0
 
a11 ··· a1n

b 
 . .. ..   0  t2 + · · ·
 .. . c 
 0
.  
an1 ··· ann ..
.

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Solution of Vector-matrix D.E.s


Replacing a0 , b0 , · · · withx1 (0), x2 (0), · · · we get,

x1  x1 (0)  x1 (0)
     
1 ··· 0 a11 ··· a1n
 
 x2  x (0) x (0)
  . .. ..   2  +  .. .. ..   2  t +
x3  =  .. . x
.  3  (0) . x (0)
. .  3 
   
.. .. ..
 
0 ··· 1 an1 ··· ann
. . .
2 x1 (0)
 
a11 ··· a1n

x2 (0)
 . .. ..  
 .. x (0) t2 + · · ·

.  3 
.  
an1 ··· ann .
..

If ,

x1 x1 (0)
   
a11 ··· a1n 1 ··· 0
   
 x2  x2 (0)
 . .. .. . .. .. 
x(t) = x3  , x(0) = x3 (0) A =  . I = .
    
. . .  . . .
.. ..
   
an1 ··· ann 0 ··· 1
. .

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Homogeneous State Equations

Continuation of Solution of Vector-matrix D.E.s..


the solution x(t) can be written as,
1 1 1
x(t) = (I + At + (A)2 t2 + (A)3 t3 + · · · + (A)k tk + · · · ) x(0)
2! 3! k!
| {z }

The expression in the under brace on the R.H.S of the last equation is an
n ∗ n matrix
It is similar to the infinite power series for a scalar exponential.It is called
matrix exponential and can be written as:
1 1 1
eAt = I + At + 2!
(A)2 t2 + 3!
(A)3 t3 + ··· + k!
(A)k tk + ···

Thus, the solution can be written as


x(t) = eAt x(0)
= φ(t)x(0)

where,φ(t) is called the ’State Transition Matrix’

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Non Homogeneous State equations

Scalar Case
Consider a scalar state equation,
ẋ = ax + bu
ẋ − ax = bu,
Multiplying this equation by e−at on both sides and integration between 0
and t gives,
x(t) = eat x(0) + 0t (ea(t−τ ) u(τ )dτ )
R

Vector Case
Consider the non homogeneous state equation described by
ẋ = Ax + Bu
where, x ∈ Rn → n-vector
u ∈ Rm →m-vector
A∈ Rn∗n → n∗n-constant matrix,
B∈ Rn∗m → n∗m-constant matrix,

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
Solution of Scalar D.E.s
State Transition Matrix
Solution of Vector D.E.s
Computational Methods of Matrix Exponential

Solution of Non Homogeneous State Equations

Continuation of Vector Case...


The solution of x(t) can be written as
x(t) = eAt x(0) + 0t (eA(t−τ ) u(τ )dτ )
R

This equation can also be written


R as
x(t) = φ(t)x(0) + 0t (φ(t − τ )u(τ )dτ )
where, φ(t)=eAt , is the State Transition Matrix’

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
State Transition Matrix Properties of State Transition Matrix
Computational Methods of Matrix Exponential

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

2 State Transition Matrix


Properties of State Transition Matrix

3 Computational Methods of Matrix Exponential


Laplace Transformation Approach
Diagonal Transformation
Cayley-Hamilton Theorem Approach

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
State Transition Matrix Properties of State Transition Matrix
Computational Methods of Matrix Exponential

State Transition Matrix- ’φ’

Properties of State Transition Matrix- ’φ’


For the time invariant system, ẋ = Ax
φ(t) = eAt

The properties of the State Transition Matrix are:

φ(0) = I
φ−1 (t) = φ(−t)
φ(t1 + t2 ) = φ(t1 )φ(t2 )
[φ(t)]n = φ(nt)
φ(t2 − t1 )φ(t1 − t0 ) = φ(t2 − t0 )

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation
State Transition Matrix Properties of State Transition Matrix
Computational Methods of Matrix Exponential

State Transition Matrix- ’φ’

Example

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Methods to compute eAt


Matrix exponential can be computed by
Numerical Methods
Analytic Methods

Numerical Methods
If matrix A is given with all elements in numerical values, MATLAB provides
a simple way to compute eAT , where T is a constant.

Analytic Methods
Some of the Analytic methods to be discussed are given below
Laplace Transformation Approach
Diagonal Transformation
Cayley-Hamilton Theorem

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

2 State Transition Matrix


Properties of State Transition Matrix

3 Computational Methods of Matrix Exponential


Laplace Transformation Approach
Diagonal Transformation
Cayley-Hamilton Theorem Approach

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Laplace Transformation Approach


We know that,
ẋ(t) = Ax(t)
Applying Laplace Transformation on both sides, we get,
sX(s) − X(0) = AX(s)
⇒ (sI-A)X(s) = X(0)
Pre-multiplying with (sI-A)−1 on both sides and taking Inverse Laplace
Transform on both sides, we get

x(t) = L−1 ((sI-A)−1 )x(0)

We know that,

x(t) = eAt x(0)

Comparing equations, (1) and (2) we can write


eAt = L−1 ((sI-A)−1 ))

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Example of Laplace Transformation approach


 
−1 0
Consider the following matrix A =
−3 −2
   
s 0 −1 −0
sI-A = −
0 s −3 −2
 −1
s+1 0
⇒ (sI-A)−1 =
3 s+2
" #
1
0
⇒ (sI-A)−1 = s+1
−3 1
(s+1)(s+2) s+2

Applying Inverse Laplace Transformation on both sides we get,

e−t
 
0
eAt = L−1 ((sI-A)−1 ) =
−3e−t+ 3e−2t e−2t

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

2 State Transition Matrix


Properties of State Transition Matrix

3 Computational Methods of Matrix Exponential


Laplace Transformation Approach
Diagonal Transformation
Cayley-Hamilton Theorem Approach

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Diagonal Transformation
Matrix A is diagonalized using a diagonalizing matrix P
The resultant matrix is given by 
eλ1 t ··· 0
−1
 . .. .. 
At
e = Pe P Λt = P .  P−1
.

. . 
0 ··· eλn t
If matrix A can be transformed into Jordan Canonical form, then eAt can be
given by
eAt = SeJt S−1
where, S is a transformation matrix that transforms matrix A into
Jordan canonical form J

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Example of Diagonal Transformation


 
0 1
Let, matrix A =
−2 −3
characteristic equation of this matrix is given by λ2 − 3λ + 2
Eigenvalues of matrix A are λ1 = −1, λ2 = −2  
1
The corresponding eigenvectors of the eigenvalues λ1 and λ2 are and
−1
 
1
respectively
2
 
1 1
The modal matrix formed by these eigenvectors is given by P =
−1 2
the diagonal matrix Λ is obtained
 by the transformation P−1 AP
−1 0
Λ=
0 −2
 −t 
e 0
∴ eJt = eΛt = −2t
0 e

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Continuation of Example of Diagonal Transformation


Having transformed Matrix A into Jordan Canonical form, Matrix
exponential eAt can be obtained by the transformation PeJt P−1

1 e−t
   
1 0 2 1
eAt = −2t
−1 2 0 e −1 −1
 −t
2e − e−2t e−t − e−2t

= −t −2t −t −2t
−2e 2e −e + 2e

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Outline

1 Solution of Differential Equation


Solution of Scalar D.E.s
Solution of Vector D.E.s

2 State Transition Matrix


Properties of State Transition Matrix

3 Computational Methods of Matrix Exponential


Laplace Transformation Approach
Diagonal Transformation
Cayley-Hamilton Theorem Approach

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Cayley-Hamilton Theorem Approach


For large systems, this method is far more convenient computationally as
compared to the other two methods discussed earlier.

Cayley-Hamilton theorem
Statement:Every square matrix A satisfies its own characteristic equation.

If,q(λ) = λn + a1 λn−1 + · · · + an−1 λ + an = 0 is the characteristic equation


of A, then
q(A) = An + a1 An−1 + · · · + an−1 A + an = 0

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Cayley-Hamilton theorem
Let λ1 , λ2 , · · · , λn be the eigenvalues of the matrix A
Consider the scalar polynomial f (λ) = k0 + k1 λ + k2 λ2 + · · · + kn λn + · · · ,
where λ is the eigenvalue of the matrix.
The matrix polynomial f (A) = k0 I + k1 A + k2 A2 + · · · + kn An + · · · an be
computed by considering the scalar polynomial f (λ)
f (λ) R(λ)
Dividing f (λ) by q(λ) we get q(λ) = Q(λ) + q(λ)
∴ f (λ) = Q(λ)qλ + R(λ)
where, R(λ) = α0 + α1 λ + α2 λ2 + · · · + αn−1 λn−1 is the remainder
polynomial
For λ = λ1 , λ2 · · · , λn (for eigenvalues) q(λ) = 0
∴ f (λi ) = R(λi ); i = 1, 2, 3, · · ·

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Cayley-Hamilton theorem
The coefficients of the remainder polynomial α0 , α1 , · · · , αn−1 can be
obtained by substituting λ = λ1 , λ2 , · · · , λn in the relation f (λi ) = R(λi )
Replacing λ with matrix A we get,
f (A) = R(A)
= α0 I + α1 A + · · · + αn−1 An−1

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Cayley-Hamilton theorem
Procedure to compute eAt :
Step-1:
Find the eigenvalues of matrix A
Step-2:
Case-1: If all the eigenvalues are distinct, the coefficients α0 , α1 , · · · , αn−1
can be obtained by solving 0 n0 simultaneous equations given by f (A) = R(A)
Case-2: If A possess an eigenvalue λK of order 0 m0 then,
Only one independent equation can be obtained by substituting λk in the
equation f (A) = R(A)
The remaining m − 1 linear equations can be obtained by differentiating
f (λ) = R(λ) on both sides
dj f (λ) dj R(λ)
∴ = ;j = 0, 1, 2 · · · , m − 1
dλj λ=λk dλj λ=λk

Step-3:
The required result is obtained by substituting the values of α0 , α1 , · · · , αn−1
in f (A) = α0 I + α1 A + · · · + αn−1 An−1

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Example of Cayley-Hamilton theorem Approach


 
0 1
Consider the matrix, A=
−1 −2
Step-1:
Eigenvalues of the matrix are λ1 = λ2 = −1
Step-2:
Since the eigenvalues equal and of order 2
∴ 2 − 1 = 1 equation can be obtained by differentiating the coefficients of
f (A) = R(A)
Another equation is obtained by substituting λ = −1 directly in the equation
f (A) = R(A)

V. Sankaranarayanan Modern Control systems


Solution of Differential Equation Laplace Transformation Approach
State Transition Matrix Diagonal Transformation
Computational Methods of Matrix Exponential Cayley-Hamilton Theorem Approach

Computational Methods of Matrix exponential-eAt

Example of Cayley-Hamilton theorem Approach


Since A is of second-order, The polynomial R(λ) will be of the form α0 + α1 λ
The coefficients of α0 , α1 can be obtained as follows:

f (λ) = eλt = α0 + α1 λ
∴ f (−1) = e−t = α0 − α1 (4)
d d λt d
f (λ) = e = (α0 + α1 λ)
dλ λ=−1 dλ λ=−1 dλ

⇒ te−t = α1

Substituting the value of α1 in equation(4) we get α0 = (1 + t)e−t


Step-3:
The required result is f (A) = eAt = α
 0 I + α1−t
A
te−t

(1 + t)e
=
−te−t (1 − t)e−t

V. Sankaranarayanan Modern Control systems

You might also like