Design of Control Systems in

State Space
State Observers

By
H. M. Yusuf
 State Observers
• In the pole placement design of state feedback, we assumed
that all state variables are available for feedback
• In practice, however, not all state variables are available for
feedback i.e. not all states are measurable.
• Therefore we need to estimate unmeasurable state variables by
a method called observation using a replica of the system to
provide an “estimate” of the system states based on the
measured output of the system.
• A device that estimates or observes unmeasurable state
variables is called a state observer or simply an observer
• If the observer observes all states of the system, it is called a
full-order state observer
• If the observer observes less than the total number of states of
the system, it is called reduced-order observer
 Full Order State Observer
A full-order observer estimates all the state variables of a system based on the measurement
of the output and control variables
 Full Order State Observer
Consider the system
𝑥 = 𝐴𝑥 + 𝐵𝑢 (1)
𝑦 = 𝐶𝑥

Let the observed state be 𝑥

The observer dynamics will be
𝑥 = A𝑥 + Bu + 𝐾𝑒 𝑦 − 𝑦 (2)
𝑦 = 𝐶𝑥 (3)

∴ 𝑥 = A𝑥 + Bu + 𝐾𝑒 𝑦 − 𝐶𝑥 (4)

Matrix 𝐾𝑒 which is called the observer gain matrix, is a weighting

matrix to the correction term involving the difference between
the measured output 𝑦 and the estimated output 𝐶𝑥
 Full Order State Observer
To obtain the observer error equation, let us subtract equation (1)
from equation (2)
𝑥 − 𝑥 = 𝐴𝑥 − 𝐴𝑥 − 𝐾𝑒 𝐶𝑥 − 𝐶𝑥
𝑥 − 𝑥 = (𝐴 − 𝐾𝑒 𝐶)(𝑥 − 𝑥 ) (5)
Define the difference between 𝑥 and 𝑥 as the error vector
𝑒 =𝑥−𝑥
Equation (5) becomes
𝑒 = 𝐴 − 𝐾𝑒 𝐶 𝑒 (6)
It can be seen from Equation (6) that the dynamic behaviour of the
error vector is determined by the eigenvalues of matrix 𝐴 − 𝐾𝑒 𝐶
given by the characteristic equation;
𝐴 − 𝐾𝑒 𝐶 = 0. (7)

If the matrix 𝐴 − 𝐾𝑒 𝐶 is a stable matrix, the error vector will

converge to zero for any initial error vector.
 Observer Design Methods
The observer design involves the determination of the
observer gain matrix 𝐾𝑒 that will yield the arbitrarily
desired eigenvalues of 𝐴 − 𝐾𝑒 𝐶 under the condition that
the system is Observable.

We will discuss three methods used for observer design

(determination of 𝐾𝑒 )
– Direct substitution method
– Transformation matrix method
– Ackermann’s formula
 Method 1: Direct Substitution Method
If the desired eigenvalues of the observer matrix are given by
s  1 , s  2 ,..., s  n (8)

Then, the characteristic equation of the observer system given by (7) should be equal
to the characteristic equation of the desired eigenvalues of the observer matrix

sI  ( A  K eC )  ( s  1 )( s  2 )...( s   n ) (9)

Since both sides of this characteristic equation are polynomials in s, by equating the
coefficients of the like powers of s on both sides, it is possible to determine the
value of 𝐾𝑒

Solving for 𝐾𝑒 in equation (9) will give the elements of the observer gain matrix

NB: This method is applicable to systems having 𝒏 ≤ 𝟑

 Method 1: Direct Substitution Method
Example 1:
Consider the system
𝑥 = 𝐴𝑥 + 𝐵𝑢
𝑦 = 𝐶𝑥

0 20.6  0 
Where A  , B    , C   0 1
1 0  1 
Design a full-order state observer, assuming that the desired eigenvalues of the observer matrix
are
𝜇1 = −10 and 𝜇2 = −10
Solution:
Step1: Check Observability;
0 1
N  C T AT C T   
1 0 
Since det(N)≠0, The system is observable
Step 2: Define 𝐾𝑒

K 
K e   e1 
 Ke2 
Method 1:Direct Substitution Method
Step 3: Form the closed loop characteristic equation

s 0 0 20.6   ke1 
sI  A  K eC         0 1
0 s  1 0   ke 2 
=s 2  ke 2 s  ( 20.6  ke1 )

Step 4: Form the characteristic equation of the desired closed loop poles
( s  1 )( s   2 )  ( s  10)( s  10)
= s 2  20 s  100

Step 5: Apply equation (9)

s 2  ke 2 s  (20.6  ke1 )  s 2  20 s  100

Equating the coefficients,
ke 2  20 120.6 
Therefore, K e   
20.6  ke1  100  20 
 ke1  120.6
 Method 2: Transformation Matrix Method
Step 1: Check the controllability condition for the system. If the system is completely state controllable
Step 2: Form the characteristic polynomial for matrix A, that is,
𝒔𝑰 − 𝑨 = 𝒔𝒏 + 𝒂𝟏 𝒔𝒏−𝟏 + 𝒂𝟐 𝒔𝒏−𝟐 …𝒂𝒏−𝟏 𝒔 +𝒂𝒏
determine the values of 𝑎1 , 𝑎2 … 𝑎𝑛
Step 3: Determine the transformation matrix T that transforms the system state equation
into the controllable canonical form. Q  WN T
where, N is the observability matrix, N  C T AT C T ( AT ) 2 C T ... ( AT ) n 1 C T 
 
and,  an 1 an  2 ... a1 1 
a a ... 1 0 
 n  2 n  3 
W  
 
 a 1 1 0 0 
 1 0 0 0 

Note: If the given system equation is already in the observable canonical form, then Q= 𝑰

Step 4: Using the desired eigenvalues (desired closed-loop poles), write the desired
characteristic polynomial: 𝒔 − 𝝁𝟏 … 𝒔 − 𝝁𝒏 = 𝒔𝒏 +𝜶𝟏 𝒔𝒏−𝟏 + 𝜶𝟐 𝒔𝒏−𝟐 …𝜶𝒏−𝟏 𝒔+𝜶𝒏
And determine the values of 𝛼1 , ∝2 … ∝𝑛
Step 5: The required state feedback gain matrix K can be determined from

K  Q 1  n  an  n 1  an 1 ...  2  a2 1  a1 
T
 Method 1: Transformation Matrix Method
Example 1:
Consider the system
𝑥 = 𝐴𝑥 + 𝐵𝑢
𝑦 = 𝐶𝑥

0 20.6  0 
A  , B    , C   0 1
Where 1 0  1 

Design a full-order state observer, assuming that the desired eigenvalues of the observer matrix
are
𝜇1 = −10 and 𝜇2 = −10
Solution:
Step1: Check Observability;0 1
 
N  C T AT C T   
1 0 

Since det(N)≠0, The system is observable

Step 2: Form characteristic polynomial for matrix A
s 20.6  a1  0
sI  A   s 2  20.6
1 s a2  20.6
= s 2  a1s  a2 s
 Method 2: Transformation Matrix Method
Step 3: Find Q
a 1  0 1 
W  1   1 0 
 1 0   
0 1 
N  
1 0 
 0 1   0 1  1 0
Q  WN T    1 0    0
 1 0    1


Step 4: The desired characteristic equation is

( s  1 )( s   2 )  ( s  10)( s  10)
= s 2  20 s  100
1  20 and  2  100
Step 5: The required observer gain matrix 𝐾𝑒
 2  a2 
1
Ke  Q  
 
 1 1 a
1
1 0  100  ( 20.6)  120.6 
=     
0 1   20  0   20 
 Method 3: Ackermann’s formula
Step 1: Check the observability condition for the system.
Step 2: Form a matrix polynomial, ∅(𝐴)using the coefficients of the desired
characteristics polynomial
 ( s )  ( s  1 )( s   2 )...( s   n )
=s n  1s n 1  ...   n 1s   n  0
According to Cayley-Hamilton theorem, a matrix satisfies its own characteristic equation.
Therefore,
 ( A)=A  1 A
n n 1
 ...   n 1 A   n I  0
Step 3: Calculate the observer gain matrix using the Ackerman’s formula
0 
0 
 
K e   ( A) N  
1

 
0 
1 
Where N is the Observability matrix given by

N  C T AT C T ( AT ) 2 C T ... ( AT ) n 1 C T 
 Method 3: Ackermann’s formula
Example 1:
Consider the system
𝑥 = 𝐴𝑥 + 𝐵𝑢
𝑦 = 𝐶𝑥

0 20.6  0 
A  , B    , C   0 1
Where 1 0  1 

Design a full-order state observer, assuming that the desired eigenvalues of the observer matrix
are 𝜇1 = −10 and 𝜇2 = −10
Solution:
Step1: Check Observability; 0 1 
N  C T AT C T   
1 0 

Since det(N)≠0, The system is observable

Step 2: Form characteristic polynomial for desired eigenvalues of observer matrix
( s  1 )( s   2 )  ( s  10)( s  10)
= s 2  20s  100
  ( A)  A2  20 A  100 I
 Method 3: Ackermann’s Formula
2
 0 20.6  0 20.6  1 0
 ( A)     20 1  100 
 1 0   0  0 1 
120.6 412 
= 
 20 120.6 

0 1
N  T
AT C T  
 1
C
 0


Step 3: Determine the K matrix using the Ackerman’s formula

0 
K e   ( A) N  
1

1 
120.6 412  0 1  0  120.6 
Ke         
 20 120.6  1 0   
1 20 