State Space Solutions and Realizations

EE-601: Linear System Theory

Prof. Bidyadhar Subudhi

School of Electrical Sciences
Indian Institute of Technology Goa
 Outline

• Solution of LTI State Equation

• Equivalent State Equations
• Realizations
• Solution of LTV Equations
• Equivalent Time-Varying Equations
• Time-Varying Realizations

2
 Introduction
 Linear System can be described by convolution and SS (if lumped)
 Convolution t
y (t ) 

 g (t , )u ( )d
t 0

k
 Discrete version y (k)   g (k, m)u(m)
m  k0

 LTI systems Integration Step size

 Transfer function, involves Laplace transform, poles and zeros,
inverse Laplace transform
 Inconvenient and inaccurate

3
 Non-homogenous Scalar Differential Equation
𝑥 𝑡 = 𝑎𝑥 𝑡 + 𝑏𝑢(𝑡) 𝑥 𝑡 − 𝑎𝑥 𝑡 = 𝑏𝑢(𝑡)
Multiply both sides by

Integrate between 0 to t

Contribution due to IC
Zero-input response Contribution due input
Zero-state response

Note: 𝑥 𝑡 𝑖𝑠 𝑎 𝑙𝑖𝑛𝑒𝑎𝑟 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑥 0 𝑎𝑛𝑑 𝑢 𝜏 𝑓𝑜𝑟 𝜏 ∈ [0, 𝑡]

4
Recollect the zero input state equation: 𝑥 𝑡 = 𝑎𝑥 𝑡 + 𝑏𝑢(𝑡)
Taking Laplace Transform:
𝑠𝑋 𝑠 − 𝑥 0 = 𝑎𝑋 𝑠 + 𝑏𝑈(𝑠)
⇒ 𝑠 − 𝑎 𝑋 𝑠 = 𝑏𝑈 𝑠 + 𝑥(0)

𝑥 0 𝑏𝑈(𝑠)
⇒𝑋 𝑠 = +
𝑠−𝑎 (𝑠 − 𝑎)

Taking inverse Laplace Transform:

𝑥 𝑡 = 𝑒 𝑎𝑡 𝑥 0 + 𝐶𝑜𝑛𝑣𝑜𝑙𝑢𝑡𝑖𝑜𝑛[𝑏𝑢(𝑡)𝑒 𝑎𝑡 ]

𝑡
𝑥 𝑡 = 𝑒 𝑎𝑡 𝑥 0 + 𝑒𝑎 𝑡−𝜏 𝑏𝑢 𝜏 𝑑𝜏
0

5
 Solution of State Equation

The state equation: 𝑥 𝑡 = 𝐴𝑥 𝑡 + 𝐵𝑢 𝑡

The state equation:

𝑥 𝑡 − 𝐴𝑥 𝑡 = 𝐵𝑢 𝑡

Multiplying both sides of the state equation with 𝑒 −𝐴𝑡 :

𝑒 −𝐴𝑡 𝑥 𝑡 − 𝑒 −𝐴𝑡 𝐴𝑥 𝑡 = 𝑒 −𝐴𝑡 𝐵𝑢 𝑡

Since 𝑒 −𝐴𝑡 and 𝐴 are commutative, i.e. 𝑒 −𝐴𝑡 𝐴 = 𝐴𝑒 −𝐴𝑡 :

𝑒 −𝐴𝑡 𝑥 𝑡 − 𝐴𝑒 −𝐴𝑡 𝑥 𝑡 = 𝑒 −𝐴𝑡 𝐵𝑢 𝑡

𝑑 𝑒 −𝐴𝑡 𝑥 𝑡
⇒ = 𝑒 −𝐴𝑡 𝐵𝑢 𝑡
𝑑𝑡
⇒ 𝑑 𝑒 −𝐴𝑡 𝑥 𝑡 = 𝑒 −𝐴𝑡 𝐵𝑢 𝑡 𝑑𝑡

6
 Solution of State Equation (contd…)
Integrating both sides for evaluation of states from initial-time 𝑡 = 0 to arbitrary-time 𝑡:
𝑡 𝑡
𝑑 𝑒 −𝐴𝜏 𝑥 𝜏 = 𝑒 −𝐴𝜏 𝐵𝑢 𝜏 𝑑𝜏
0 0

𝑡 𝑡 𝑡
⇒ 𝑒 −𝐴𝜏 𝑥 𝜏 = 𝑒 −𝐴𝜏 𝐵𝑢 𝜏 𝑑𝜏 ⇒ 𝑒 −𝐴𝑡 𝑥 𝑡 −𝑥 0 = 𝑒 −𝐴𝜏 𝐵𝑢 𝜏 𝑑𝜏
0 0 0

𝑡
⇒ 𝑥 𝑡 = 𝑒 𝐴𝑡 𝑥 0 + 𝑒 𝐴𝑡 𝑒 −𝐴𝜏 𝐵𝑢 𝜏 𝑑𝜏
0

𝑡
⇒ 𝑥 𝑡 = 𝑒 𝐴𝑡 𝑥 0 + 𝑒𝐴 𝑡−𝜏 𝐵𝑢 𝜏 𝑑𝜏
0

Zero-input
response Zero-state response

We will refer summation of Zero-input and Zero-state responses as Forced State Response.

7
 Laplace Transform Approach

𝑥 𝑡 = 𝐴𝑥 𝑡 + 𝐵𝑢 𝑡
Taking Laplace Transform:
𝑠𝑋 𝑠 − 𝑥 0 = 𝐴𝑋 𝑠 + 𝐵𝑈 𝑠
⇒ 𝑠𝐼 − 𝐴 𝑋 𝑠 = 𝑥 0 + 𝐵𝑈(𝑠)
⇒ 𝑋 𝑠 = 𝑠𝐼 − 𝐴 −1 𝑥 0 + 𝑠𝐼 − 𝐴 −1 𝐵𝑈 𝑠
Taking inverse Laplace Transform:
𝑥 𝑡 = ℒ −1 𝑠𝐼 − 𝐴 −1
𝑥 0 + ℒ −1 𝑠𝐼 − 𝐴 −1
𝐵𝑈(𝑠)

𝑡
𝑥 𝑡 = 𝑒 𝐴𝑡 𝑥 0 + 𝑒𝐴 𝑡−𝜏 𝐵𝑢 𝜏 𝑑𝜏
0
Comparing the above with zero-input response 𝑥 𝑡 = Φ 𝑡 𝑥 0 = 𝑒 𝐴𝑡 𝑥(0):

Φ 𝑡 = 𝑒 𝐴𝑡 = ℒ −1 𝑠𝐼 − 𝐴 −1 8
 The Linearity

𝐴𝑡 𝑡 𝐴 𝑡−𝜏
Solution of state equation: 𝑥 𝑡 = 𝑒 𝑥 0 + 0
𝑒 𝐵𝑢 𝜏 𝑑𝜏
𝑡 𝐴 𝑡−𝜏
Consider the zero-state response: 𝑥 𝑡 = 0
𝑒 𝐵𝑢 𝜏 𝑑𝜏

For linearity, it must satisfy superposition properties.

𝑡 𝐴 𝑡−𝜏
For additive: 𝑥 𝑡 = 0
𝑒 𝐵 𝑢1 𝜏 + 𝑢2 𝜏 𝑑𝜏

𝑡 𝑡
= 𝑒𝐴 𝑡−𝜏
𝐵𝑢1 𝜏 𝑑𝜏 + 𝑒𝐴 𝑡−𝜏
𝐵𝑢2 𝜏 𝑑𝜏
0 0

𝑡 𝐴 𝑡−𝜏 𝑡 𝐴 𝑡−𝜏
For homogeneity: 𝑥 𝑡 = 0
𝑒 𝐵 𝛼𝑢 𝜏 𝑑𝜏 = 𝛼 0
𝑒 𝐵𝑢 𝜏 𝑑𝜏

 Similar property can be shown with respect to the initial state 𝑥(0) for
the zero-input response.
9
 State Transition Matrix
𝑡 𝐴 𝑡−𝜏
Solution of state equation: 𝑥 𝑡 = 𝑒 𝐴𝑡 𝑥 0 + 0
𝑒 𝐵𝑢 𝜏 𝑑𝜏

Consider the zero-input response:

𝑥 𝑡 = 𝑒 𝐴𝑡 𝑥 0
=Φ 𝑡 𝑥 0

It represents how the initial-state 𝑥(0) is transited to 𝑥(𝑡).

State Transition Matrix (STM): Φ 𝑡 = 𝑒 𝐴𝑡

𝐴2 𝑡 2 𝐴3 𝑡 3
In general, Φ 𝑡 = 𝑒 𝐴𝑡 = 𝐼 + 𝐴𝑡 + + +⋯
2! 3!

10
 Properties of STM Φ(𝑡)
𝐴2 𝑡 2 𝐴3 𝑡 3
STM:Φ 𝑡 = 𝑒 𝐴𝑡 = 𝐼 + 𝐴𝑡 + + +⋯
2! 3!

Property 1: Φ 0 = 𝐼

Property 2: Φ−1 𝑡 = Φ −𝑡

Property 3: Φ 𝑡1 Φ 𝑡2 = Φ(𝑡1 + 𝑡2 )

Property 4: If 𝐴 is diagonal then Φ(𝑡) is diagonal as well.

11
 Solution of STM Φ(𝑡) Method 1: Power Series Method

−1 1
Consider 𝐴 = .
0 −2

𝐴2 𝑡 2 𝐴3 𝑡 3
Φ 𝑡 = 𝑒 𝐴𝑡 = 𝐼 + 𝐴𝑡 + + +⋯
2! 3!
2 3
1 0 −1 1 1 −3 𝑡 −1 7 𝑡
= + 𝑡+ + +⋯
0 1 0 −2 0 4 2! 0 −8 3!

𝑡2 𝑡3 𝑡2 𝑡3
1−𝑡+ − +⋯ 𝑡−3 +7 −⋯
= 2! 3! 2! 3!
𝑡2 𝑡3
0 1 − 2𝑡 + 4 − 8 + ⋯
2! 3!
−𝑡
= 𝑒 𝑒 −𝑡 − 𝑒 −2𝑡
0 𝑒 −2𝑡
 Con: Recognition of series terms is difficult 12
 Solution of STM Φ(𝑡) Method 2: Using Inverse Laplace Transform
Recollect the zero input state equation: 𝑥 𝑡 = 𝐴𝑥 𝑡

Taking Laplace Transform:

𝑠𝑋 𝑠 − 𝑥 0 = 𝐴𝑋 𝑠
⇒ 𝑠𝐼 − 𝐴 𝑋 𝑠 = 𝑥(0)
⇒ 𝑋 𝑠 = 𝑠𝐼 − 𝐴 −1 𝑥(0)

Taking inverse Laplace Transform:

𝑥 𝑡 = ℒ −1 𝑠𝐼 − 𝐴 −1 𝑥(0)

Comparing the above with zero-input response 𝑥 𝑡 = Φ 𝑡 𝑥 0 = 𝑒 𝐴𝑡 𝑥(0):

Φ 𝑡 = 𝑒 𝐴𝑡 = ℒ −1 𝑠𝐼 − 𝐴 −1

13
 Solution of STM Φ(𝑡) Method 2: Using Inverse Laplace Transform
−1 1
Example: Consider the same 𝐴 = .
0 −2
To compute: Φ 𝑡 = 𝑒 𝐴𝑡 = ℒ −1 𝑠𝐼 − 𝐴 −1

𝑠 + 1 −1
𝑠𝐼 − 𝐴 =
0 𝑠+2

1 1
−1 1
−1 𝑠 + 1 −1 𝑠+2 1 𝑠+1 𝑠+1 𝑠+2
⇒ 𝑠𝐼 − 𝐴 = = =
0 𝑠+2 𝑠+1 𝑠+2 0 𝑠+1 1
0
𝑠+2
Taking inverse Laplace Transform:

1 1
𝑠+1 𝑠+1 𝑠+2 𝑒 −𝑡
𝑒 −𝑡 − 𝑒 −2𝑡
ℒ −1 𝑠𝐼 − 𝐴 −1
= ℒ −1 =
1 0 𝑒 −2𝑡
0
𝑠+2
14
 Solution of STM Φ(𝑡) Method 3: Cayley-Hamilton Method

Cayley-Hamilton Theorem: A square matrix 𝐴 satisfies its own characteristic

equation. If we have for an 𝑛 × 𝑛 matrix 𝐴, the characteristics equation

∆  = 𝑛 + 𝛼1 𝑛−1 + ⋯ + 𝛼𝑛−1  + 𝛼𝑛
Then according to this theorem
∆ 𝐴 = 𝐴𝑛 + 𝛼1 𝐴𝑛−1 + ⋯ + 𝛼𝑛−1 𝐴 + 𝐼
The degree of matrix polynomial can be reduced .
𝑛−1 + ⋯ + 𝛼𝑛−1  + 𝛼𝑛

15
 Theorem: Cayley-Hamilton
Δ A = An + α1 An−1 + ⋯ + αn−1 A + αn I = 0 (11)

Cayley-Hamilton theorem implies that An can be written as linear combination of

{I, A,…., An−1 }.
Multiplying (11) by A yields

An+1 + α1 An + ⋯ + αn−1 A2 + αn A = 0

which implies that An+1 can be written as a linear combination of {A, A2 ,…, An }
In conclusion, for any polynomial 𝑓 A , no matter how large its degree is, 𝑓 A can
be expressed as 𝑓 A = β0 𝐼 + β1 𝐴 + ⋯ + β𝑛−1 An−1 12

One way to compute (12) is to use long division to express 𝑓 λ as

𝑓 λ =𝑞 λ Δ λ +ℎ λ (13)
𝑞 λ :quotient; ℎ λ is the remainder polynomial with degree < n

Then 𝑓 A = 𝑞 A Δ A + ℎ A = 𝑞 A 0 + ℎ A = ℎ A

30
• If the degree is large, the long division method is not convenient to carryout.
• In this case, h(λ) can be solved directly from (13).
• Let
𝑓 λ = β0 + β1 λ + ⋯ + β𝑛−1 λn−1
• n unknowns β𝑖 are to be solved.
If all n eigen values of A are distinct, these β𝑖 can be solved from the n
equations.
𝑓 𝜆𝑖 = 𝑞 𝜆𝑖 Δ 𝜆𝑖 + ℎ 𝜆𝑖 = ℎ 𝜆𝑖

31
 If A has repeated eigenvalues (𝜆𝑖 of multiplicity m) , then
𝑓 𝜆𝑖 = 𝑞 𝜆𝑖 Δ 𝜆𝑖 + ℎ 𝜆𝑖 = ℎ 𝜆𝑖 (1)
must be differentiated to yield additional equations.

• 𝑓 𝜆𝑖 = ℎ 𝜆𝑖
• The remaining (m-1) linear equations must be obtained to
solve for 𝛼𝑖 by differentiating both sides of (1)
𝑑𝑙 𝑓 𝜆
• 𝑓𝑙 𝜆𝑖 ≔ for 𝑙 = 0,1, … , 𝑚 − 1 and 𝑖 = 1,2, … , 𝑛-m
𝑑𝜆𝑙 𝜆=𝜆
𝑖

𝑓 A = ℎ A and ℎ 𝜆 is said to be equal 𝑓 𝜆 on the spectrum of A

 Example
0 1
Compute 𝐴100 with A= ?
−1 −2
Solution: 𝑓 𝐴 = 𝐴100
The characteristic polynomial of A is
Δ λ = (λ + 1)2 = λ2 + 2λ + 1.

• 𝑓 𝜆 𝑖 = ℎ 𝜆𝑖 𝑑𝑙 𝑓 𝜆
𝑓 𝑙 𝜆𝑖 ≔
𝑑𝜆𝑙 𝜆=𝜆𝑖

• Let h(λ)= β0 + β1 λ. spectrum of A we have

𝑓 −1 = ℎ −1 ; (−1)100 = β0 − β1
𝑓 ′ −1 = ℎ′ −1 ; 100. (−1)99 = β1
• Thus 𝑑𝑓 𝑑100
=
𝑑 𝑑
• β1 =-100, β0 = 1 + β1 = −99, ℎ λ = −99 − 100 λ and
𝐴100 = β0 𝐼 + β1 A=−99I−100A.

1 0 0 1 −199 −100
=−99 −100 =
0 1 −1 −2 100 101

33
 Solution of STM Φ(𝑡) Method 3: Cayley-Hamilton Method

Cayley-Hamilton Theorem: A square matrix 𝐴 satisfies its own characteristic

equation.
∆ 𝜆 = 𝜆𝐼 − 𝐴 = 𝛽0 + 𝛽1 𝜆 + ⋯ + 𝛽𝑛−1 𝜆𝑛−1 =0
Characteristic equation of a matrix: 𝜆𝐼 − 𝐴 = 0.
⇒ 𝜆𝑛 = 𝛽0 + 𝛽1 𝜆 + ⋯ + 𝛽𝑛−1 𝜆𝑛−1
where 𝛽0 , … , 𝛽𝑛−1 are coefficients of characteristic polynomial.
Following Cayley-Hamilton theorem:
𝐴𝑛 = 𝛽0 + 𝛽1 𝐴 + ⋯ + 𝛽𝑛−1 𝐴𝑛−1

Therefore: 𝑒 𝐴𝑡 = 𝛼0 𝐼 + 𝛼1 𝐴 + ⋯ + 𝛼𝑛 𝐴𝑛−1
Where 𝛼0 , … , 𝛼𝑛 need to be find out.

On the other hand, the above is satisfied for all the eigenvalues 𝜆𝑖 of 𝐴, i.e.
𝑒 𝜆𝑖𝑡 = 𝛼0 + 𝛼1 𝜆𝑖 + ⋯ + 𝛼𝑛 𝜆𝑖 𝑛−1 , 𝑖 = 1, … , 𝑛

20
 Solution of STM Φ(𝑡) Method 3: Cayley-Hamilton Method (contd…)

Computing steps:

Step 1: Find the 𝑛 eigenvalues (say 𝜆1 , 𝜆2 , … , 𝜆𝑛 ) of the 𝐴 matrix

Step 2: Constitute the 𝑛 number of equations satisfying the characteristic equation,

i.e.
𝑒 𝜆𝑖𝑡 = 𝛼0 + 𝛼1 𝜆𝑖 + ⋯ + 𝛼𝑛 𝜆𝑖 𝑛−1 , 𝑖 = 1, … , 𝑛

Step 3: Solve for the coefficients 𝛼𝑖 , 𝑖 = 0,1, … , 𝑛 − 1 from the 𝑛 simultaneous

scalar equations

Step 4: Finally compute 𝑒 𝐴𝑡 = 𝛼0 𝐼 + 𝛼1 𝐴 + ⋯ + 𝛼𝑛 𝐴𝑛−1

Con: Applicable to non-singular (invertible) 𝐴 matrix.

21
 Solution of STM Φ(𝑡) Method 3: Cayley-Hamilton Method (contd…)
−1 1
Example: Consider the same 𝐴 = .
0 −2
Step 1: Eigenvalues are -1, -2

Step 2: The scalar equations are:

𝑒 −𝑡 = 𝛼0 − 𝛼1 and 𝑒 −2𝑡 = 𝛼0 − 2𝛼1

Step 3: Solving 𝛼0 = 2𝑒 −𝑡 − 𝑒 −2𝑡 and 𝛼1 = 𝑒 −𝑡 − 𝑒 −2𝑡

Step 4: Finally, 𝑒 𝐴𝑡 = 𝛼0 𝐼 + 𝛼1 𝐴

1 0 −1 1
= 2𝑒 −𝑡 − 𝑒 −2𝑡 + 𝑒 −𝑡 − 𝑒 −2𝑡
0 1 0 −2
−𝑡
= 𝑒 𝑒 −𝑡 − 𝑒 −2𝑡
0 𝑒 −2𝑡

22
 Solution of LTI Equations
1 0  1 
We use methods 1 and 2 to compute ,(sI  A) where A   
Method 1: We use  1  2 
1
s 1  1  s  2  1
(sI  A) 1    
 1 s  2 s 2  2 s  1  1 s 
s  2 1 
 
 
2 2
( s 1) ( s 1 )
 1 s 
 ( s  1) 2
( s  1) 
2

Method 2: The eigenvalues of A are -1, -1. Let h(λ) = β0+β1λ. If h(λ) equals
f(λ): = (s-λ)-1 on the spectrum of A, then
f (1)  h(1) : s  11   0  1
f  (1)  h(1) : s  12  1
Thus we have
 
h( )  (s  1) 1  (s  1) 2  (s  2) 2 
and sI  A1  h( A)  s  11  s  12 I  ( s  1) 2 A
s  2  1 

 s  12
s  1 
2

 1 s 
 s  12 s  12  23

 Similarity Transformation
𝐴𝑥 = 𝑦 (1)
(A maps 𝑥 in 𝑅𝑛 into y in 𝑅𝑛 )
Wrt the basis 𝑞1, 𝑞2, … , 𝑞𝑛 (1) becomes 𝐴𝑥 = 𝑦 (2)
where 𝑥, 𝑦 representations of 𝑥, 𝑦 wrt basis 𝑞1, 𝑞2, … , 𝑞𝑛 .
α1
α2
𝑥=𝑄 . 𝑥 (3)
.
αn
𝑇
= α1 α2 … α𝑛 is the representation of x wrt basis of {q1, q2,…,qn}

𝑥 = 𝑄x (4) Substitute (4) in (1)

𝐴𝑄x = 𝑄𝑦 or 𝑄−1 𝐴𝑄x = 𝑦
𝑦 = 𝑄y
Comparing this with (2)

𝑄 = [𝑞1, 𝑞2, … , 𝑞𝑛 ] 𝐴 = 𝑄−1 𝐴𝑄 or 𝐴 = 𝑄𝐴𝑄−1 Similarity transformation

A and 𝑨 are said to be similar

 Zero-Input Response
Suppose we have a matrix A whose Jordan form is

Q :a non-singular matrix that makes the change of coordinates that

brings A to 𝐴.

scalars 1 and 2 :eigenvalues of 𝐴, which are also those of A. The

matrix exponential of a matrix in its Jordan form is easy to
compute. For the above example we have

25
 Jordan form
• A matrix with repeated eigenvalues and a deficient number of
associated eigenvectors cannot be diagonalised. However, it can
always be taken to a block-diagonal and triangular form called the
Jordan form. For example,
This matrix has two distinct
eigenvalues, 1 and 2 ; 1 is
repeated five times, while 2
appears only once.

There are two Jordan blocks

associated with 1 ; one of order
3 and one of order 2.

For any square matrix A, there is always a

non-singular matrix Q such that

26
 Forced State response

𝑡 𝐴 𝑡−𝜏
The forced solution: 𝑥 𝑡 = 𝑒 𝐴𝑡 𝑥 0 + 0
𝑒 𝐵𝑢 𝜏 𝑑𝜏
Problem: Find the response of the system given by:

𝑥1 𝑡 −1 1 𝑥1 𝑡 0
= + 𝑢 𝑡
𝑥2 𝑡 0 −2 𝑥2 𝑡 1

𝑥1 0 0
for = and 𝑢 𝑡 = 2, 𝑡 > 0.
𝑥2 0 1

𝑒 −𝑡 𝑒 −𝑡 − 𝑒 −2𝑡
Solution: Already computed STM 𝑒 𝐴𝑡 =
0 𝑒 −2𝑡
𝑡
𝑥1 𝑡
= 𝑒 𝐴𝑡 𝑥 0 + 𝑒𝐴 𝑡−𝜏
𝐵𝑢 𝜏 𝑑𝜏
𝑥2 𝑡 0

−𝑡
Zero-Input Response: 𝑒 𝐴𝑡 𝑥 0 = 𝑒 𝑒 −𝑡 − 𝑒 −2𝑡 0 = 𝑒 −𝑡 − 𝑒 −2𝑡
0 𝑒 −2𝑡 1 𝑒 −2𝑡 27
 Forced State response
𝑥1 𝑡 −1 1 𝑥1 𝑡 0
The system: = + 𝑢 𝑡
𝑥2 𝑡 0 −2 𝑥2 𝑡 1

𝑒 −𝑡 𝑒 −𝑡 − 𝑒 −2𝑡
𝐴𝑡
Solution (continued): Already computed STM 𝑒 =
0 𝑒 −2𝑡
𝑡 𝑡
𝑒𝐴 𝑡−𝜏 𝐵𝑢 𝜏 𝑑𝜏 = 𝑒 𝐴𝑡 𝑒 −𝐴𝜏 𝐵𝑢 𝜏 𝑑𝜏
0 0

𝑡
= 𝑒 −𝑡 𝑒 −𝑡− 𝑒 −2𝑡 𝑒𝜏 𝑒 𝜏 − 𝑒 2𝜏 0 2 𝑑𝜏
0 𝑒 −2𝑡 0 0 𝑒 2𝜏 1

𝑡
= 𝑒 −𝑡 𝑒 −𝑡 − 𝑒 −2𝑡 2 𝑒 𝜏 − 𝑒 2𝜏
𝑑𝜏
0 𝑒 −2𝑡 0 𝑒 2𝜏

−𝑡 𝑡
= 𝑒 𝑒 −𝑡 − 𝑒 −2𝑡 2𝑒 𝜏 − 𝑒 2𝜏 = 𝑒 −𝑡 𝑒 −𝑡 − 𝑒 −2𝑡 2𝑒 𝑡 − 𝑒 2𝑡 − 1
0 𝑒 −2𝑡 𝑒 2𝜏 0 0 𝑒 −2𝑡 𝑒 2𝑡 − 1 28
 Forced State response
𝑥1 𝑡 −1 1 𝑥1 𝑡 0
The system: = + 𝑢 𝑡
𝑥2 𝑡 0 −2 𝑥2 𝑡 1

𝑒 −𝑡 𝑒 −𝑡 − 𝑒 −2𝑡
𝐴𝑡
Solution (continued): Already computed STM 𝑒 =
0 𝑒 −2𝑡
Final Solution:

𝑥1 𝑡 𝑡 𝐴 𝑡−𝜏
𝑥 𝑡 = = 𝑒 𝐴𝑡 𝑥 0 + 0
𝑒 𝐵𝑢 𝜏 𝑑𝜏
𝑥2 𝑡
−𝑡 −2𝑡 −𝑡
= 𝑒 − 𝑒 + 𝑒 𝑒 −𝑡 − 𝑒 −2𝑡 2𝑒 𝑡 − 𝑒 2𝑡 − 1
𝑒 −2𝑡 0 𝑒 −2𝑡 𝑒 2𝑡 − 1
−𝑡
= −𝑒 +1
1

29