ECSL

3 Observer-based control design via

Lyapunov stability approach
• Let us consider a continuous-time linear system of the following
general form:

ẋ(t) = Ax(t) + Bu(t)
(13)
y(t) = Cx(t)

where x(t) ∈ Rn represents the state, u(t) ∈ Rm represents the

control input, and y(t) ∈ Rp represents the output.
 ECSL

• Based on (10) and (11), the observer-based control is described as

 ¡ ¢
x̃(t)
˙ = Ax̃(t) + Bu(t) − L y(t) − C x̃(t)
(14)
u(t) = K x̃(t)

where x̃(t) ∈ Rn represents the estimated state; K ∈ Rm×n and

L ∈ Rn×p are the control gain and the observer gain, respectively, to
be designed later.
 ECSL

• As a result, based on
 
x(t)
ζ(t) =   ∈ R2n×2n
e(t)

the closed-loop system with (13) and (14) is described as follows:

ζ̇(t) = Āζ(t) (15)

where
 
A + BK −BK
Ā =  .
0 A + LC
 ECSL

Theorem 2. For a prescribed value µ, close-loop control system (15)

is stable at origin if there exist matrices P̄1 = P̄1T ∈ Rn×n ,
P2 = P2T ∈ Rn×n , K̄ ∈ Rm×n , and L̄ ∈ Rn×p such that

P̄1 > 0 (16)

P2 > 0 (17)
 © ª 
He AP̄1 + B K̄ −B K̄ 0
 © ª 
 ? −He µP̄1 µI  < 0. (18)
 
© ª
? ? He P2 A + L̄C

Furthermore, the control and observer gains are constructed as

follows:

K = K̄ P̄1−1 , L = P2−1 L̄.

 ECSL

Proof: Let us choose a Lyapunov function of the following form:

V (t) = xT (t)P1 x(t) + eT (t)P2 e(t) = ζ T (t)P ζ(t)

where P1 = P1T > 0, P2 = P2T > 0, and

 
P1 0
P =P =T   > 0.
0 P2

The time derivative of V (t) along with (15) is given by

T
© ª
V̇ (t) = ζ (t)He P Ā ζ(t).
 ECSL

Thus, the observer-based control design condition becomes

© ª
He P Ā < 0 (19)

that is,
 
 P (A + BK) −P1 BK 
1
He   < 0. (20)
 0 P2 (A + LC) 
 ECSL

Furthermore, using
 
P1−1 0
M = 
0 P1−1

the congruent transformation of (20) is given from Lemma 1 as

follows:
 
 P (A + BK) −P BK 
1 1
M T · He   ·M <0
 0 P2 (A + LC) 

that is,
 
 (A + BK)P −1 −BKP1−1 
1
He   < 0. (21)
 0 P1−1 (P2 A + P2 LC)P1−1 
 ECSL

By replacing

P̄1 = P1−1 , K̄ = KP1−1 , L̄ = P2 L

condition (21) can be expressed as

 © ª 
He AP̄1 + B K̄ −B K̄
 © ª  < 0. (22)
? P̄1 He P2 A + L̄C P̄1
 ECSL

Under the following condition:

© ª
He P2 A + L̄C < 0 (23)

Lemma 3 allows
© ª © ª ³ © ª´−1
2
P̄1 He P2 A + L̄C P̄1 ≤ −He µP̄1 − µ He P2 A + L̄C .

Thus, condition (22) is ensured by (23) and

 © 
He AP̄1 + B K̄} −B K̄
 © ª ³ © ª ´−1  < 0.
? −He µP̄1 − µ2 He P2 A + L̄C
(24)
 ECSL

Finally, by Lemma 4, condition (24) is converted into

 © ª 
He AP̄1 + B K̄ −B K̄ 0
 © ª 
0>  ? −He µP̄1 µI .
 (25)
© ª
? ? He P2 A + L̄C

Here, it is worth noting that condition (25) guarantees (23). Thus,

there is no need to additionally include (23) in the observer-based
control design condition.
 ECSL

Example 3.1. By using Theorem 2 for µ = 100, obtain both control

and observer gains for
   
1 −2 1 h i
A= , B =   , C = 0.1 0 .
1 4 0.1

And see if the state trajectory

 on the x1 -x2 plane 
converges to the
1.0 0.0
origin, where x(0) =   and x̃(0) =  . Finally,
−1.0 0.0
compare it with the state trajectory of Example 2.1.
 ECSL

4 Robust observer-based control

• Let us consider a continuous-time uncertain system of the following
general form:

ẋ(t) = (A + ∆A)x(t) + (B + ∆B)u(t)
(26)
y(t) = Cx(t)

where ∆A ∈ Rn×n and ∆B ∈ Rn×m denote the parameter

uncertainties expressed as

∆A = EΥH1 , ∆B = EΥH2 . (27)

 ECSL

• As in (14), the observer-based control is configured by using the

nominal matrices A, B, and C:
 ¡ ¢
x̃(t)
˙ = Ax̃(t) + Bu(t) − L y(t) − C x̃(t)
u(t) = K x̃(t)

where x̃(t) ∈ Rn represents the estimated state; K ∈ Rm×n and

L ∈ Rn×p are the control gain and the observer gain, respectively, to
be designed later.
 ECSL

• Hence, based on
 
x(t)
ζ(t) =   ∈ R2n×2n
e(t)

the closed-loop system with (26) and (14) is described as follows:

ζ̇(t) = (Ā + ∆Ā)ζ (28)

where
   
A + BK −BK ∆A + ∆BK −∆BK
Ā =   , ∆Ā =  .
0 A + LC ∆A + ∆BK −∆BK
 ECSL

• Furthermore, the closed-loop system can be transformed into

ζ̇(t) = (Ā + Ē ῩH̄)ζ(t)

where
   
A + BK −BK E 0
Ā =   , Ē =  
0 A + LC 0 E
   
Υ 0 H1 + H2 K −H2 K
Ῡ =   , H̄ =  .
0 Υ H1 + H2 K −H2 K
 ECSL

Theorem 3. For prescribed values µ and ² > 0, close-loop control

system (15) is robustly stable at origin if there exist matrices
P̄1 = P̄1 , P2 = P2T , K̄, and L̄ such that

P̄1 > 0 (29)

P2 > 0 (30)
 
−²I 0 Ψc −H2 K̄ 0 0
 
 0 −²I Ψc −H2 K̄ 0 0 
 
 
 ? ? Ωc + ²EE T −B K̄ 0 0 
 
 © ª <0 (31)
 ? ? ? −He µP̄1 µI 0 
 
 
 ? ? ? ? Ωo ²P2 E 
 
? ? ? ? ? −²I
 ECSL

where
© ª © ª
Ωc = He AP̄1 + B K̄ , Ωo = He P2 A + L̄C
Ψc = H1 P̄1 + H2 K̄.

Moreover, the control and observer gains are constructed as follows:

K = K̄ P̄1−1 , L = P2−1 L̄.

 ECSL

Proof: let us choose a Lyapunov function of the following form:

V (t) = ζ T (t)P ζ(t)

where
 
P1 0
P =P T
=  > 0.
0 P2

The time derivative of V (t) is given by

T
© ª
V̇ (t) = ζ (t)He P (Ā + Ē ῩH̄) ζ(t).

Thus, the observer-based control design condition becomes

© ª
He P (Ā + Ē ῩH̄) < 0. (32)
 ECSL

To be specific, condition (32) is represented as

 
 P (A + BK) −P1 BK
1
He  
 0 P2 (A + LC)
   
P1 E 0 H1 + H2 K −H2 K 
+  Ῡ   < 0. (33)
0 P2 E H1 + H2 K −H2 K 
 ECSL

Furthermore, using
 
P1−1 0
M = 
0 P1−1

the congruent transformation of (33) is given from Lemma 1 as

follows:
 
 (A + BK)P −1 −BKP −1
1 1
He  
 0 P1−1 (P2 A + P2 LC)P1−1
   
E 0 (H1 + H2 K)P1−1 −H2 KP1−1 
+  Ῡ   < 0.
0 P1−1 P2 E (H1 + H2 K)P1−1 −H2 KP1−1 
(34)
 ECSL

By replacing

P̄1 = P1−1 , K̄ = KP1−1 , L̄ = P2 L

condition (34) can be expressed as

 
Ω −B K̄ © ª
 c  + He ĒῩH̄ < 0 (35)
? P̄1 Ωo P̄1

where

Ωc = He{AP̄1 + B K̄}, Ωo = He{P2 A + L̄C}

   
E 0 Ψc −H2 K̄
Ē =   , H̄ =   , Ψc = H1 P̄1 + H2 K̄.
0 P̄1 P2 E Ψc −H2 K̄
 ECSL

From Lemma 2, it follows that

n o
He ĒΥH̄ ≤ ²ĒĒT + ²−1 H̄T ΥT ΥH̄ ≤ ²ĒĒT + ²−1 H̄T H̄.

Thus, by the Schur complement, condition (35) holds if

 
−²I H̄
   
 
 Ωc −B K̄ <0
 ?   + ²ĒĒT 
? P̄1 Ωo P̄1

that is,
 
−²I 0 Ψc −H2 K̄
 
 0 −²I Ψc −H2 K̄ 
 
  < 0. (36)
 ? ? Ωc + ²EE T −B K̄ 
 
? ? ? P̄1 (Ωo + ²P2 EE T P2 )P̄1
 ECSL

Under the following condition:

Ωo + ²P2 EE T P2 < 0 (37)

Lemma 3 allows
³ ´ © ª ³ ´−1
P̄1 Ωo + ²P2 EE T P2 P̄1 ≤ −He µP̄1 − µ2 Ωo + ²P2 EE T P2 .

Accordingly, condition (36) is ensured by (37) and

 
−²I 0 Ψc −H2 K̄
 
 0 −²I Ψc −H2 K̄ 
 
 
 ? ? Ωc + ²EE T
−B K̄ < 0. (38)
  
 © ª 
 −He µP̄1 
 ? ? ?  
−µ2 (Ωo +²P2 EE T P2 )−1
 ECSL

Using the Schur complement, condition (38) is transformed into

 
−²I 0 Ψc −H2 K̄ 0
 
 0 −²I Ψc −H2 K̄ 0 
 
 
 ? ? Ωc + ²EE T
−B K̄ 0 <0
 
 © ª 
 ? ? ? −He µ P̄ µI 
 1 
? ? ? ? Ωo + ²P2 EE T P2

and it is transformed once more into (31).

 ECSL

Finally, it is worth noting that (31) implies

 
Ω ²P2 E
 o <0
? −²I

which guarantees (37) according to Schur complement. Thus, there is

no need to additionally include (37) in the robust observer-based
control design condition.
 ECSL

5 H∞ observer-based control
• Let us consider the following linear system:


ẋ(t) = Ax(t) + Bu(t) + Ew(t)

y(t) = Cx(t) + Dw(t) (39)



z(t) = Gx(t) + Hu(t)

where x(t) ∈ Rn , u(t) ∈ Rm , w(t) ∈ Rd , y(t) ∈ Rp , and z(t) ∈ Rq .

 ECSL

• Thus, the closed-loop system is given by


ζ̇(t) = Āζ(t) + Ēw(t)
(40)
z(t) = Ḡζ(t)

where
   
A + BK −BK E
Ā =   , Ē =  
0 A + LC E + LD
h i
Ḡ = G + HK −HK .
 ECSL

Theorem 4. For a prescribed value µ > 0, close-loop control system

(40) is asymptotically stable at origin and has the H∞ performance
γ, if there exist matrices P̄1 = P̄1T , P2 = P2T , K̄, L̄, and a scalar
γ > 0, such that

P̄1 > 0 (41)

P2 > 0 (42)
 
−I Ψc −H K̄ 0 0 0
 
 ? Ω −B K̄ E 0 0 
 c 
 
 ? ? −2µP̄1 −N1 µI N1 
 
 <0 (43)
 ? ? ? −He{N2 } 0 N2 
 
 
 ? ? ? ? Ωo Ψo 
 
? ? ? ? ? −γ 2 I
 ECSL

where
© ª © ª
Ωc = He AP̄1 + B K̄ , Ωo = He P2 A + L̄C
Ψc = GP̄1 + H K̄, Ψo = P2 E + L̄D.

Moreover, the control and observer gains are constructed as follows:

K = K̄ P̄1−1 , L = P2−1 L̄.

 ECSL

Proof: Let us choose a Lyapunov function of the following form:

V (t) = ζ T (t)P ζ(t)

where
 
P1 0
P =P T
=  > 0.
0 P2
 ECSL

Then it is obtained that

V̇ (t) + z T (t)z(t) − γ 2 wT (t)w(t)

³ ´ ³ ´T
= ζ T (t)P Āζ(t) + Ēw(t) + Āζ(t) + Ēw(t) P ζ(t)

+ ζ T (t)ḠT Ḡζ(t) − γ 2 wT (t)w(t)

 T   
ζ(t) He{P Ā} + ḠT Ḡ P Ē ζ(t)
=      .
w(t) ? −γ 2 I w(t)
 ECSL

Thus, the H∞ control design condition is given as follows:

 
He{P Ā} + ḠT Ḡ P Ē
0> . (44)
? −γ 2 I

Furthermore, by the Schur complement, condition (44) is transformed

into
 
−I Ḡ 0
 
0 >  ? He{P Ā} P Ē 

. (45)
? ? −γ 2 I
 ECSL

Specifically, condition (45) becomes

 
−I G + HK −HK 0
 
 ? He{P (A + BK)} −P1 BK P1 E 
 1 
0> .
 ? ? He{P2 (A + LC)} P2 (E + LD) 
 
? ? ? −γ 2 I
(46)
 ECSL

Furthermore, using
M = diag(I, P1−1 , P1−1 , I)
the congruent transformation of (46) is given from Lemma 1 as
follows:
 
−1
−I Ψc −HKP1 0
 
 ? Ω −BKP1 −1
E 
 c 
0> . (47)
 ? −1
? P1 Ωo P1 −1
P1 Ψo 
−1
 
? ? ? −γ 2 I
where
Ωc = He{(A + BK)P1−1 }, Ωo = He{P2 (A + LC)}
Ψc = (G + HK)P1−1 , Ψo = P2 (E + LD).
 ECSL

Letting

P̄1 = P1−1 , F̄ = F P̄1 , L̄ = P2 L

condition (47) is represented as

 
−I Ψc −H K̄ 0
 
 ? Ωc −B K̄ E 
 
0>  (48)
 ? ? P̄1 Ωo P̄1 P̄1 Ψo 
 
? ? ? −γ 2 I

where
n o n o
Ωc = He AP̄1 + B K̄ , Ωo = He P2 A + L̄C
Ψc = GP̄1 + H K̄, Ψo = P2 E + L̄D.
 ECSL

Under the following condition:

 
Ω Ψo
 o <0 (49)
? −γ 2 I

Lemma 3 allows
     
P̄ Ω P̄ P̄1 Ψo P̄1 0 Ωo Ψo P̄1 0
 1 o 1 =   
? −γ 2 I 0 I ? −γ 2 I 0 I
  
 µI N P̄1 0 
1
≤ −He   
 0 N2 0 I 
  −1  T
µI N1 Ωo Ψo µI N1
−     .
0 N2 ? −γ 2 I 0 N2
 ECSL

Thus, condition (48) holds by (49) and

 
−I Ψc −H K̄ 0
 
 ? Ω −B K̄ E 
 c 
0> 
 ? ? −2µP̄1 −N1 
 
? ? ? −He{N2 }
   T
0 0 0 0
  −1  
 0 0  Ω Ψ  0 0 
  o o
  
−    . (50)
 µI N1  ? −γ I 2  µI N1 
   
0 N2 0 N2
 ECSL

Finally, by the Schur complement, condition (50) can be transformed

into
 
−I Ψc −H K̄ 0 0 0
 
 ? Ω −B K̄ E 0 0 
 c 
 
 ? ? −2µP̄1 −N1 µI N1 
 
0> 
 ? ? ? −He{N2 } 0 N2 
 
 
 ? ? ? ? Ωo Ψo 
 
? ? ? ? ? −γ 2 I

which ensures (49). Thus, there is no need to additionally include

(49) in the H∞ observer-based control design condition.
 ECSL

Figure 1: State estimation through an observer

 ECSL

Figure 2: Observer-based output-feedback control