Figure 8 Observer state feedback control system with a minimum order observer

It is important to note, however, that if the measurement of output variables involves

significant noises and is relatively inaccurate, then the use of the full-order observer may
result in a better system performance.
To present the basic idea of the minimum-order observer, without undue mathematical
complications, we shall present the case where the output is a scalar (that is, m = 1) and derive
the state equation for the minimum-order observer. Consider the system:
𝒙̇ = 𝑨𝒙 + 𝒃𝒙
𝒚 = 𝒄𝒙
where the state vector 𝒙 can be partitioned into two parts 𝑥𝑎 (a scalar) and 𝒙𝒃 [an (n - 1)-
vector]. Here the state variable𝑥𝑎 is equal to the output y and thus can be directly measured,
and 𝒙𝒃 is the unmeasurable portion of the state vector. Then the partitioned state and output
equations become:
𝑥̇ 𝑎 𝐴 𝑨𝒂𝒃 𝑥𝑎 𝐵
[ ] = [ 𝑎𝑎 ] [𝒙 ] + [ 𝑎 ] 𝑢
𝒙̇ 𝒃 𝑨𝒃𝒂 𝑨𝒃𝒃 𝒃 𝑩𝒃
𝑥𝑎
𝑦 = [1 𝟎] [𝒙 ]
𝒃

where: 𝐴𝑎𝑎 is scalar;

𝑨𝒂𝒃 is 1 × (𝑛 − 1) matrix ;
𝑨𝒃𝒂 is (𝑛 − 1) × 1 matrix ;
𝑨𝒃𝒃 is (𝑛 − 1) × (𝑛 − 1) matrix;

28
 𝐵𝑎 is scalar;
and 𝑩𝒃 is (𝑛 − 1) × 1 matrix.
Figure 9 shows the block diagram of the observed state feedback control system with the
minimum order observer.

Figure 9 block diagram of the observed state feedback control system with the minimum order observer

From Figure 9 the state space equation of the minimum order observer is given by:
̃
𝒅𝜼
̂𝜼
=𝑨 ̂𝒚 + 𝑭
̂+𝑩 ̂𝑢 (1.30)
𝒅𝒕
̂ = 𝑨𝒃𝒃 − 𝒌𝒆 𝑨𝒂𝒃 ;
where: 𝑨
̂=𝑨
𝑩 ̂ 𝒌𝒆 + 𝑨𝒃𝒂 − 𝒌𝒆 𝐴𝑎𝑎
̂ = 𝑩𝒃 − 𝒌𝒆 𝐵𝑎
𝑭
𝒙 ̂𝜼
̃=𝑪 ̂𝑦
̃+𝑫 (1.31)

̂=[ 𝟎 ̂ =[1]
where: 𝑪 ]; and 𝑫
𝑰𝒏−𝟏 𝒌𝒆
̃
𝑢 = −𝒌𝒙 (1.32)
The characteristic equation for the minimum-order observer is obtained from Equation (1.30)
as follows:
|𝑠𝐼 − 𝐴̃| = |𝑠𝐼 − 𝐴𝑏𝑏 + 𝑘𝑒 𝐴𝑎𝑏 | = (𝑠 − 𝜇1 )(𝑠 − 𝜇2 ) … (𝑠 − 𝜇𝑛−1 )
= 𝑠 𝑛−1 + 𝛼̃1 𝑠 𝑛−2 + 𝛼̃2 𝑠 𝑛−3 + ⋯ + 𝛼̃𝑛−2 𝑠 + 𝛼̃𝑛−1 = 0 (1.33)

29
where 𝜇1 , 𝜇2 , … , 𝜇𝑛−1 are desired eigenvalues for the minimum-order observer. The observer
gain matrix 𝒌𝒆 can be determined by first choosing the desired eigenvalues for the minimum-
order observer [that is, by placing the roots of the characteristic equation, Equation (1.33), at
the desired locations] and then using the procedure developed for the full-order observer with
appropriate modifications.
The observer gain matrix 𝒌𝒆 can be obtained using Ackermann’s Formula as shown in the
following equation:
𝑨𝒂𝒃 −𝟏
𝑨𝒂𝒃 𝑨𝒃𝒃 0
𝑨 𝑨 𝟐 0
𝒌𝒆 = 𝝓(𝑨𝒃𝒃 ) 𝒂𝒃: 𝒃𝒃 : (1.34)
0
𝑨𝒂𝒃 𝑨𝒏−𝟑
𝒃𝒃 [1]
𝒏−𝟐
[𝑨𝒂𝒃 𝑨𝒃𝒃 ]
Where: 𝝓(𝑨𝒃𝒃 ) = 𝑨𝒏−𝟏 ̃1 𝑨𝒏−𝟐
𝒃𝒃 + 𝛼 ̃2 𝑨𝒏−𝟑
𝒃𝒃 + 𝛼 𝒃𝒃 + ⋯ + 𝛼
̃𝑛−2 𝑨𝒃𝒃 + 𝛼̃𝑛−1 𝑰

1.4.3.1 Observed-State Feedback Control System with Minimum-

Order Observer:
For the case of the observed-state feedback control system with full-order state observer, we
have shown that the closed-loop poles of the observed-state feedback control system consist
of the poles due to the pole-placement design alone, plus the poles due to the observer design
alone. Hence, the pole-placement design and the full-order observer design are independent of
each other.
For the observed-state feedback control system with minimum-order observer, the same
conclusion applies. The system characteristic equation can be derived as:
|𝑠𝑰 − 𝑨 + 𝒃𝒌||𝑠𝑰 − 𝑨𝒃𝒃 + 𝒌𝒆 𝑨𝒂𝒃 | = 0 (1.35)
Example 1.15: Consider the system:
𝒙̇ = 𝑨𝒙 + 𝒃𝒙
𝒚 = 𝒄𝒙
0 1 0 0
Where: 𝐴 = [ 0 0 1 ]; 𝑏 = [0 ] ; 𝑐 = [1 0 0]
−6 −11 −6 1
Let us assume that we want to place the closed loop poles at:
𝑠1,2 = −2 ∓ 𝑗2√3; 𝑠3 = −6
And assume that we choose the desired observer poles to be at:
𝑠1,2 = −𝜇1,2 = −10
Design minimum-order observer.

30
Solution:
Using Ackermann’s Formula, the state feedback gain matrix can be obtained as follows:
𝑘 = [90 29 4].
The characteristic equation for the minimum-order observer is:
|𝑠𝐼 − 𝐴𝑏𝑏 + 𝑘𝑒 𝐴𝑎𝑏 | = (𝑠 − 𝜇1 )(𝑠 − 𝜇2 ) = (𝑠 + 10)(𝑠 + 10) = 𝑠 2 + 20𝑠 + 100 = 0
We shall use Ackermann’s formula:
𝑨𝒂𝒃 −𝟏 0
𝒌𝒆 = 𝝓(𝑨𝒃𝒃 ) [ ] [ ]
𝑨𝒂𝒃 𝑨𝒃𝒃 1
Where; 𝝓(𝑨𝒃𝒃 ) = 𝑨𝒃𝒃 2 + 20𝑨𝒃𝒃 + 100𝐼
Since:

0 1 0 0
𝐴=[ 0 0 1 ]; 𝑏 = [0];
−6 −11 −6 1
0 0 1 0
∴ 𝐴𝑎𝑎 = 0; 𝐴𝑎𝑏 = [1 0]; 𝐴𝑏𝑎 = [ ] ; 𝐴𝑏𝑏 = [ ] ; 𝐵𝑎 = 0 𝑎𝑛𝑑 𝐵𝑏 = [ ]
−6 −11 −6 1
0 1 2 0 1 1 0 1 0 −1 0 14
∴ 𝑘𝑒 = {[ ] + 20 [ ] + 100 [ ]} [ ] [ ]=[ ]
−11 −6 −11 −6 0 1 0 1 1 5
Since;
𝒅𝜼̃
=𝑨 ̂𝜼
̂+𝑩 ̂𝒚 + 𝑭
̂𝑢
𝒅𝒕
̂ = 𝑨𝒃𝒃 − 𝒌𝒆 𝑨𝒂𝒃 = [ 0
𝑨
1 14
] − [ ] [1 0] = [
−14 1
];
−11 −6 5 −16 −6
𝑩̂=𝑨 ̂ 𝒌𝒆 + 𝑨𝒃𝒂 − 𝒌𝒆 𝐴𝑎𝑎 = [−14 1 ] [14] + [ 0 ] − [14] ∗ 0 = [−191]
−16 −6 5 −6 5 −260
̂ = 𝑩𝒃 − 𝒌𝒆 𝐵𝑎 = [0] − [14] ∗ 0 = [0]
𝑭
1 5 1
𝑑𝜂̃2
−14 1 𝜂̃2 −191 0
∴ [ 𝑑𝑡 ] = [ ][ ] + [ ]𝑦 + [ ]𝑢
𝑑𝜂̃3 −16 −6 𝜂̃3 −260 1
𝑑𝑡
𝜂̃ 𝑥̃
Where: [ 2 ] = [ 2 ] − 𝑘𝑒 𝑦, or
𝜂̃3 𝑥̃3
𝑥̃ 𝜂̃
[ 2 ] = [ 2 ] + 𝑘𝑒 𝑥1
𝑥̃3 𝜂̃3
If the observed-state feedback is used, then the control signal 𝑢 becomes:
𝑥1
𝑢 = −𝒌𝒙 ̃ = −𝑘 [ ̃2 ]
𝑥
𝑥̃3
Figure 10 is a block diagram showing the configuration of the system with observed-state
feedback, where the observer is the minimum-order observer
31
1.4.3.2 Transfer Function of Minimum-Order Observer Based
Controller:
From Equations (1.30, 1.31 and 1.32) the transfer function of minimum-order observer is
given by:
𝑈(𝑠) 𝑛𝑢𝑚
= = − [𝑪 ̃ )−𝟏 𝑩
̃ (𝒔𝑰 − 𝑨 ̃ +𝐷
̃] (1.36)
−𝑌(𝑠) 𝑑𝑒𝑛
Where:
̃=𝑨
𝑨 ̂−𝑭
̂ 𝑲𝒃
̃=𝑩
𝑩 ̂ −𝑭
̂ (𝐾𝑎 + 𝑲𝒃 𝒌𝒆 )
̃ = −𝑲𝒃
𝑪
̃ = −(𝐾𝑎 + 𝑲𝒃 𝒌𝒆 )
𝐷
We defined the state feedback matrix as:
𝒌 = [𝐾𝑎 𝐾𝑏 ]
𝐾𝑎 is a scalar; and 𝑲𝒃 is 1 × (𝑛 − 1) row vector.

32