IC6601 Advanced Control System Department of ICE 2018-19

Unit I
State feedback controller design
Pole placement approach
In the conventional approach to the design of a single input, single-output control system,
we design a controller (compensator) such that the dominant closed-loop poles have a desired
damping ratio  and a desired undamped natural frequency n. In this approach, the order of the
system may be raised by 1 or 2 unless pole–zero cancellation takes place. Note that in this approach
we assume the effects on the responses of nondominant closed-loop poles to be negligible. Different
from specifying only dominant closed-loop poles (the conventional design approach), the present
pole-placement approach specifies all closed-loop poles.

The pole-placement method is somewhat similar to the root-locus method in that we place
closed-loop poles at desired locations. The basic difference is that in the root-locus design we place
only the dominant closed-loop poles at the desired locations, while in the pole-placement design
we place all closed-loop poles at desired locations.

There is a requirement on the part of the system for the closed-loop poles to be placed at
arbitrarily chosen locations. The requirement is
1. The system must be completely state controllable.
2. All the state variables are measurable and available for feedback.

Consider a control system

𝑥̇ = 𝐴𝑥 + 𝐵𝑢
𝑦 = 𝐶𝑥 + 𝐷𝑢 (1)

Where, x = state vector (n-vector)

y = output vector (scalar)
u = control signal (scalar)
A = n × n constant matrix
B = n × 1 constant matrix
C = 1 × n constant matrix
D = constant (scalar)

The control signal to be,

𝑢 = −𝐾𝑥

This means that the control signal u is determined by an instantaneous state. Such a scheme
is called state feedback. The 1×n matrix K is called the state feedback gain matrix. We assume that
all state variables are available for feedback. A block diagram for this system is shown in figure.

[Link]’s College of Engineering 1 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
This closed-loop system has no input. Its objective is to maintain the zero output. Because
of the disturbances that may be present, the output will deviate from zero. The nonzero output will
be returned to the zero reference input because of the state feedback scheme of the system. Such a
system where the reference input is always zero is called a regulator system.

Substituting Equation (1) into Equation (2) gives

𝑥̇ = (𝐴 − 𝐵𝐾) 𝑥(𝑡)
The solution of this equation is given by
𝑥(𝑡) = 𝑒 (𝐴−𝐵𝐾)𝑡 𝑥(0)

where, 𝑥(0) is the initial state caused by external disturbances. The stability and transient response
characteristics are determined by the eigenvalues of matrix(𝐴 − 𝐵𝐾). If matrix K is chosen
properly, the matrix (𝐴 − 𝐵𝐾) can be made an asymptotically stable matrix, and for all 𝑥(0) ≠ 0x,
it is possible to make 𝑥(𝑡) approach 0 as t approaches infinity. The eigenvalues of matrix (𝐴 − 𝐵𝐾)
are called the regulator poles. If these regulator poles are placed in the left-half s plane, then 𝑥(𝑡)
approaches 0 as t approaches infinity. The problem of placing the regulator poles (closed-loop
poles) at the desired location is called a pole-placement problem.

Ackermanns’ formula - Derivation

The system is defined by

𝑥̇ = 𝐴𝑥 + 𝐵𝑢

and the control signal is given by

𝑢 = −𝐾𝑥

The system is completely state controllable and the desired closed loop poles are at 𝑠 = 𝜇1 , 𝑠 =
𝜇2 , 𝑠 = 𝜇3 , … … , 𝑠 = 𝜇𝑛 . Use of the state feedback control
𝑢 = −𝐾𝑥
modifies the system equation to
𝑥̇ = (𝐴 − 𝐵𝐾)𝑥 (1)
Let us define 𝐴̃ = 𝐴 − 𝐵𝐾. The desired characteristic equation is
|𝑠𝐼 − 𝐴 + 𝐵𝐾| = |𝑠𝐼 − 𝐴̃| = (𝑠 − 𝜇1 ) (𝑠 − 𝜇2 ) (𝑠 − 𝜇3 ) . .. (𝑠 − 𝜇𝑛 )
= 𝑠 𝑛 + 𝛼1 𝑠 𝑛−1 + 𝛼2 𝑠 𝑛−2 + ⋯ . . +𝛼𝑛−1 𝑠 + 𝛼𝑛 = 0
Since Cayley-Hamilton theorem states that 𝐴̃ satisfies its own characteristic equation.

∅(𝐴̃) = 𝐴̃𝑛 + 𝛼1 𝐴̃𝑛−1 + 𝛼2 𝐴̃𝑛−2 + ⋯ … … . +𝛼𝑛−1 𝐴 + 𝛼𝑛 𝐼 = 0 (2)

We shall utilize Equation (2) to derive Ackermann’s formula. To simplify the derivation, we
consider the case where n=3. (For any other positive integer n, the following derivation can be
easily extended.)

Consider the following identities:

𝐼=𝐼
𝐴̃ = 𝐴 − 𝐵𝐾
𝐴̃2 = (𝐴 − 𝐵𝐾)2 = 𝐴2 − 𝐴𝐵𝐾 − 𝐵𝐾𝐴̃
𝐴̃3 = (𝐴 − 𝐵𝐾)3 = 𝐴3 − 𝐴2 𝐵𝐾 − 𝐴𝐵𝐾𝐴̃ − 𝐵𝐾𝐴̃2

Multiplying the preceding equations in order by 𝛼3 , 𝛼2 , 𝛼1 , and 𝛼0 respectively, and adding the
results, we obtain,
𝛼3 𝐼 + 𝛼2 𝐴̃ + 𝛼1 𝐴̃2 + 𝐴̃3
= 𝛼3 𝐼 + 𝛼2 (𝐴 − 𝐵𝐾) + 𝛼1 (𝐴2 − 𝐴𝐵𝐾 − 𝐵𝐾𝐴̃) + 𝐴3 − 𝐴2 𝐵𝐾 − 𝐴𝐵𝐾𝐴̃ − 𝐵𝐾𝐴̃2
[Link]’s College of Engineering 2 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
= 𝛼3 𝐼 + 𝛼2 𝐴 + 𝛼1 𝐴2 + 𝐴3 − 𝛼2 𝐵𝐾 − 𝛼1 𝐴𝐵𝐾 − 𝛼1 𝐵𝐾𝐴̃ − 𝐴2 𝐵𝐾 − 𝐴𝐵𝐾𝐴̃ − 𝐵𝐾𝐴̃2 (3)

Referring to equation (2), we have,

𝛼3 𝐼 + 𝛼2 𝐴̃ + 𝛼1 𝐴̃2 + 𝐴̃3 = ∅(𝐴̃) = 0

Also, we have
𝛼3 𝐼 + 𝛼2 𝐴 + 𝛼1 𝐴2 + 𝐴3 = ∅(𝐴) ≠ 0

Substituting the last two equations into equation (3), we have

∅(𝐴̃) = ∅(𝐴) − 𝛼2 𝐵𝐾 − 𝛼1 𝐵𝐾𝐴̃ − 𝛼1 𝐴𝐵𝐾 − 𝐴𝐵𝐾𝐴̃ − 𝐴2 𝐵𝐾

Since ∅(𝐴̃) = 0, we obtain

∅(𝐴) = 𝐵(𝛼2 𝐾 + 𝛼1 𝐾𝐴̃ + 𝐾𝐴̃2 ) + 𝐴𝐵(𝛼1 𝐾 + 𝐾𝐴̃) + 𝐴2 𝐵𝐾

𝛼2 𝐾 + 𝛼1 𝐾𝐴̃ + 𝐾𝐴̃2
[𝐵 𝐴𝐵 𝐴 𝐵 ] [ 2
𝛼1 𝐾 + 𝐾𝐴̃ ] (4)
𝐾
Since the system is completely state controllable, the inverse of the controllability matrix
[𝐵 𝐴𝐵 𝐴2 𝐵 ]
exists. Premultiplying both sides of equation (4) by the inverse of the controllability matrix, we
obtain

𝛼2 𝐾 + 𝛼1 𝐾𝐴̃ + 𝐾𝐴̃2
[𝐵 𝐴𝐵 𝐴2 𝐵 ∅(𝐴) = [ ]−1 𝛼1 𝐾 + 𝐾𝐴̃ ]
𝐾
Premultiplying both sides of this last equation by [0 0 1] , we obtain

𝛼2 𝐾 + 𝛼1 𝐾𝐴̃ + 𝐾𝐴̃2
[0 0 1][𝐵 𝐴𝐵 𝐴2 𝐵 ]−1 ∅(𝐴) = [0 0 1] [ 𝛼1 𝐾 + 𝐾𝐴̃ ]=𝐾
𝐾
Which can be written as
𝐾 = [0 0. . . 0 1 ] [𝐵 𝐴𝐵 … … … … . . 𝐴𝑛−1 𝐵 ]−1 ∅(𝐴) (5)

This equation gives the required state feedback gain matrix K. this is known as Ackermann’s
formula for the determination of the state feedback gain matrix K

Procedure for calculating State feedback controller gain

Method I : Using Transformation Matrix T

Step 1: Check the controllability of the given system. If it is controllable, then go to step2

Controllability matrix |𝐶𝑀| = |𝐵 𝐴𝐵 𝐴2 𝐵 ,……. 𝐴𝑛−1 𝐵 | ≠ 0

Step 2: Find the characteristic polynomial

|𝑆𝐼 − 𝐴| = 𝑆 𝑛 + 𝑎1 𝑆 𝑛−1 + 𝑎2 𝑆 𝑛−2 +, … … … … + 𝑎𝑛−1 𝑆 + 𝑎𝑛

and hence determine the values of 𝑎1 , 𝑎2 , 𝑎3 , … … . 𝑎𝑛

Step 3: Determine the transformation matrix ′𝑇′ which transforms system state equation into
controllable canonical form.
[Link]’s College of Engineering 3 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
𝑎𝑛−1 𝑎𝑛−2 ,…… 𝑎1 1
𝑎𝑛−2 𝑎𝑛−3 ,…… 1 0
𝑇 = 𝑀 × 𝑊 = [𝐵 𝐴𝐵 𝐴2 𝐵 𝑛−1 ]
,……….. 𝐴 𝐵 ; ; ,……. ; ;
𝑎1 1 ,……. 0 0
[ 1 0 ,……. 0 0]

Note : If the given system is already in controllable canonical form, then 𝑇 = 𝐼

Step 4: Using desired closed loop poles, write the desired polynomial. i.e.

(𝑆 − 𝜇1 )(𝑆 − 𝜇2 )(𝑆 − 𝜇3 ), … … … … … … (𝑆 − 𝜇𝑛 )

= 𝑆 𝑛 + 𝛼1 𝑆 𝑛−1 + 𝛼2 𝑆 𝑛−2 +, … … … … + 𝛼𝑛−1 𝑆 + 𝛼𝑛

and determine the values of 𝛼1 , 𝛼2 , 𝛼3 , … … … … . . 𝛼𝑛

Step 5: The required state feedback gain matrix ′𝐾′ is given by

𝐾 = [𝛼𝑛 − 𝑎𝑛 𝛼𝑛−1 − 𝑎𝑛−1 , . , , … . 𝛼2 − 𝑎2 𝛼1 − 𝑎1 ] 𝑇 −1

Method II : Using Direct Substitution method

Step 1: Check the controllability of the given system. If it is controllable, then go to step2

Controllability matrix |𝐶𝑀| = |𝐵 𝐴𝐵 𝐴2 𝐵 ,……. 𝐴𝑛−1 𝐵 | ≠ 0

Note: If the system is of low order (𝑛 ≤ 3), direct substitution of matrix K into the desired
characteristics polynomial may be simpler.

Step 2: If 𝑛 = 3, then 𝐾 = [𝐾1 𝐾2 𝐾3 ]

Step 3: Substitute this 𝐾 matrix into the desired characteristics polynomial |𝑆𝐼 − 𝐴 + 𝐵𝐾| and
equate it to (𝑠 − 𝜇1 )(𝑠 − 𝜇2 )(𝑠 − 𝜇3 ).

|𝑆𝐼 − 𝐴 + 𝐵𝐾| = (𝑠 − 𝜇1 )(𝑠 − 𝜇2 )(𝑠 − 𝜇3 )

Since both sides of this characteristic equation are polynomials in ′𝑆′.

Step 4: By equating the coefficients of like power of ′𝑆′ on both sides, it is possible to determine
the values of 𝐾1 , 𝐾2 , 𝐾3

Method III : Using Ackermann’s formula

Step 1: Check the controllability of the given system. If it is controllable, then go to step2

Controllability matrix |𝐶𝑀| = |𝐵 𝐴𝐵 𝐴2 𝐵 ,……. 𝐴𝑛−1 𝐵 | ≠ 0

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝜇1 , 𝜇2 , 𝜇3 , . . , 𝜇𝑛
i.e., (𝑠 − 𝜇1 ) (𝑠 − 𝜇2 ) (𝑠 − 𝜇3 ). . . . . . . . (𝑠 − 𝜇𝑛 ) = 𝑠 𝑛 + 𝛼1 𝑠 𝑛−1 +, … … . . , +𝛼𝑛−1 𝑠1 + 𝛼𝑛
and determine the values of 𝛼1 , 𝛼2 , … … . , 𝛼𝑛 .

Step 3: Determine the matrix ∅(𝐴) using the coefficient of desired characteristics polynomial.

∅(𝐴) = 𝐴𝑛 + 𝛼1 𝐴𝑛−1 + 𝛼2 𝐴𝑛−2 +, … … . . , +𝛼𝑛−1 𝐴1 + 𝛼𝑛 𝐼

[Link]’s College of Engineering 4 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
Step 4: Calculate the state feedback gain matrix, ′𝐾′ using the Ackermann’s formula

𝐾 = [0 0 0 , … … . 1] [𝐵 𝐴𝐵 𝐴2 𝐵 ,…….. 𝐴𝑛−1 𝐵 ]−1 [∅(𝐴)]

Design of Type 1 Servo System when the Plant has an Integrator.

Consider a plant, defined by

𝑥̇ = 𝐴𝑥 + 𝐵𝑢
𝑦 = 𝐶𝑥 (1)

Where, x = state vector (n-vector)

y = output vector (scalar)
u = control signal (scalar)
A = n × n constant matrix
B = n × 1 constant matrix
C = 1 × n constant matrix
The control signal 𝑢 and the output signal 𝑦 are scalars. By proper choice of a set of state
variables, it is possible to choose the output to be equal to one of the state variables. Figure
shows a general configuration of the type 1 servo system when the plant has an integrator. Here
we assumed that 𝑦 = 𝑥1 . In the present analysis assume that the reference input 𝑟 is a step
function.

In this system, the following state feedback control system is used

𝑥1
𝑥
𝑢 = −[0 𝑘1 𝑘2 , … … … 𝑘𝑛 ] [ ;2 ] + 𝑘1 (𝑟 − 𝑥1 )
𝑥𝑛
= −𝐾𝑥 + 𝑘1 𝑟 (2)

Where, 𝐾 = [𝑘1 𝑘2 , , , … . . 𝑘𝑛 ]
Assume that the reference input is applied at t= 0. Then, for 𝑡 > 0, the system dynamics can
be described by equation (1) and (2), or
𝑥̇ = 𝐴𝑥 + 𝐵𝑢 = (𝐴 − 𝐵𝐾)𝑥 + 𝐵𝑘1 𝑟 (3)
Now we can design a type 1 servo system such that the closed loop poles are located at desired
positions. The designed system will be an asymptotically stable system, 𝑦(∞) will approach
the constant value 𝑟, and 𝑢(∞) will approach 0. At steadystate,

𝑥̇ (∞) = (𝐴 − 𝐵𝐾)𝑥(∞) + 𝐵𝑘1 𝑟(∞) (4)

[Link]’s College of Engineering 5 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
Where, 𝑟(𝑡) is a step input, we have 𝑟(∞) = 𝑟(𝑡) = 𝑟 (constant) for 𝑡 > 0. By subtracting
equation (4) from equation (3), we obtain

𝑥̇ (𝑡) − 𝑥̇ (∞) = (𝐴 − 𝐵𝐾)[𝑥(𝑡) − 𝑥(∞)] (5)

Define, 𝑥(𝑡) + 𝑥(∞) = 𝑒(𝑡), then equation (5) becomes

𝑒̇ = (𝐴 − 𝐵𝐾)𝑒 (6)

Equation (6) describes error dynamics. The design of the type 1 servo system here is converted
to the design of an asymptotically stable regulator system such that 𝑒(𝑡) approaches zero, given
any initial conditions 𝑒(0). If the system defined by equation (1) is completely state
controllable, then, by specifying the desired eigenvalues 𝜇1 , 𝜇2 , , … . . 𝜇𝑛 for the matrix 𝐴 − 𝐵𝐾,
matrix 𝐾 can be determined by the pole placement technique. The steadystate values of 𝑥(𝑡)
and 𝑢(𝑡) can be found as follows. At steady state 𝑡 = ∞, from equation (2)

𝑥̇ (∞) = 0 = (𝐴 − 𝐵𝐾)𝑥(∞) + 𝐵𝑘1 𝑟 (7)

Since the desired eigenvalues of 𝐴 − 𝐵𝐾 are all in the left half s-plane, the inverse of matrix
𝐴 − 𝐵𝐾 exists. Consequently, 𝑥(∞) can be determined as

𝑥(∞) = −(𝐴 − 𝐵𝐾)−1 𝐵𝑘1 𝑟 (8)

Also 𝑢(∞) can be obtained as

𝑢(∞) = −𝐾𝑥(∞) + 𝑘1 𝑟 = 0

Design of Type 1 Servo System when the Plant Has No Integrator

If the plant has no integrator (type 0 plant), the basic principle of the design of a type 1 servo
system is to insert an integrator in the feedforward path between the error comparator and the plant,
as shown in Figure 1. (The block diagram of Figure 1 is a basic form of the type 1 servo system
where the plant has no integrator.)

From the diagram, we obtain

𝑥̇ = 𝐴𝑥 + 𝐵𝑢 (1)
𝑦 = 𝐶𝑥 (2)
𝑢 = −𝐾𝑥 + 𝑘1  (3)
 = 𝑟 − 𝑦 = 𝑟 − 𝐶𝑥 (4)

Where, x = state vector (n-vector)

y = output vector (scalar)
u = control signal (scalar)
 = output of the integrator(scalar)
[Link]’s College of Engineering 6 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
A = n × n constant matrix
B = n × 1 constant matrix
C = 1 × n constant matrix
Assume that the plant given by Equation (1) is completely state controllable. The transfer function
of the plant can be given by
𝐺𝑝 (𝑠) = 𝐶(𝑆𝐼 − 𝐴)−1 𝐵
To avoid the possibility of the inserted integrator being cancelled by the zero at the origin of the
plant, we assume that 𝐺𝑝 (𝑠) has no zero at the origin. Assume that the reference input (step
function) is applied at 𝑡 = 0. Then, for 𝑡 > 0, the system dynamics can be described by an equation
that is a combination of Equations (1) and (4)
𝑥̇ (𝑡) 𝐴 0 𝑥(𝑡) 𝐵 0
[̇ ]=[ ][ ] + [ ] 𝑢(𝑡) + [ ] 𝑟(𝑡) (5)
 (t) −𝐶 0 (𝑡) 0 1
We can design an asymptotically stable system such that 𝑥(∞), (∞) and 𝑢(∞) approach constant
values, respectively. Then, at steady state, and we get (𝑡) = 0, and 𝑦(∞) = 𝑟. At steady state we
have
𝑥̇ (∞) 𝐴 0 𝑥(𝑡) 𝐵 0
[̇ ]=[ ][ ] + [ ] 𝑢(∞) + [ ] 𝑟(∞) (6)
 (∞) −𝐶 0 (𝑡) 0 1
Noting that 𝑟(𝑡) is a step input, we have 𝑟(∞) = 𝑟(𝑡) = 𝑟 (constant) for 𝑡 > 0. By subtracting
Equation (6) from Equation (5), we obtain

𝑥̇ (𝑡) − 𝑥̇ (∞) 𝐴 0 𝑥(𝑡) − 𝑥(∞) 𝐵

[̇ ]=[ ][ ] + [ ] [𝑢(𝑡) − 𝑢(∞)] (7)
 (t) − ̇ (∞) −𝐶 0 (𝑡) − (∞) 0

Define
𝑥(𝑡) − 𝑥(∞) = 𝑥𝑒 (𝑡)
(𝑡) − (∞) = 𝑒 (𝑡)
𝑢(𝑡) − 𝑢(∞) = 𝑢𝑒 (𝑡)
Then, equation (7) can be written as
𝑥𝑒̇ (𝑡) 𝐴 0 𝑥𝑒 (𝑡) 𝐵
[̇ ]=[ ][ ] + [ ] 𝑢𝑒 (𝑡) (8)
𝑒 (t) −𝐶 0 𝑒 (𝑡) 0
Where, 𝑢𝑒 (𝑡) = −𝐾𝑥𝑒 (𝑡) + 𝑘𝑟 𝑒 (𝑡) (9)
Define a new (𝑛 + 1) th order error vector 𝑒(𝑡) by
𝑥𝑒 (𝑡)
𝑒(𝑡) = [ (𝑡)] = (𝑛 + 1) vector
𝑒

Then equation (8) becomes,

𝑒̇ = 𝐴̂𝑒 + 𝐵̂ 𝑢𝑒 (10)
Where
𝐴 0 ̂ 𝐵
𝐴̂ = [ ],𝐵[ ]
−𝐶 0 0
Equation (9) becomes
𝑢𝑒 = −𝐾 ̂𝑒 (11)
Where
𝐾̂ = [𝐾 ⋮ −𝑘1 ]
The state error equation can be obtained by substituting equation (11) into equation (10)
𝑒̇ = (𝐴̂ − 𝐵̂ 𝐾
̂ )𝑒 (12)
If the desired eigenvalues of matrix (𝐴̂ − 𝐵̂ 𝐾 ̂ ) are specified as 𝜇1 , 𝜇2 , , … . . 𝜇𝑛 then the state
feedback gain matrix 𝐾 and the integral gain constant 𝑘1 can be determined by the pole placement
technique, provided that the system defined by equation (10) is completely state controllable. If the
matrix
𝐴 0
[ ]
−𝐶 0
[Link]’s College of Engineering 7 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
n\has rank 𝑛 + 1 then the system defined by equation (10) is completely controllable.

State Observer Design

In the pole-placement approach to the design of control systems, we assumed that all state
variables are available for feedback. In practice, however, not all state variables are available for
feedback. Then we need to estimate unavailable state variables. Estimation of unmeasurable state
variables is commonly called observation. A device (or a computer program) that estimates or
observes the state variables is called a state observer, or simply an observer. If the state observer
observes all state variables of the system, regardless of whether some state variables are available
for direct measurement, it is called a full-order state observer.

There are times when this will not be necessary, when we will need observation of only the
unmeasurable state variables, but not of those that are directly measurable as well. For example,
since the output variables are observable and they are linearly related to the state variables, we need
not observe all state variables, but observe only 𝑛 − 𝑚 state variables, where 𝑛 is the dimension of
the state vector and 𝑚 is the dimension of the output vector.

An observer that estimates fewer than 𝑛 state variables, where 𝑛 is the dimension of the
state vector, is called a reduced-order state observer or simply, a reduced-order observer. If the
order of the reduced-order state observer is the minimum possible, the observer is called a
minimum-order state observer or minimum-order observer.

State Observer Design: A state observer estimates the state variables based on the measurements
of the output and control variables. Here the concept of observability plays an important role. State
observers can be designed if and only if the observability condition is satisfied.

In many practical cases, the observed state vector is used in the state feedback to generate
the desired control vector. Consider the plant defined by

𝑥̇ = 𝐴𝑥 + 𝐵𝑢 (1)
𝑦 = 𝐶𝑥 (2)

The observer is a subsystem to reconstruct the state vector of the plant. The mathematical
model of the observer is basically the same as that of the plant, except that we include an additional
term that includes the estimation error to compensate for inaccuracies in matrices A and B and the
lack of the initial error. The estimation error or observation error is the difference between the
measured output and the estimated output. The initial error is the difference between the initial state
and the initial estimated state. Thus, we define the mathematical model of the observer to be

𝑥̃ = 𝐴𝑥̃ + 𝐵𝑢 + 𝐾𝑒 (𝑦 − 𝐶𝑥̃)
= (𝐴 − 𝐾𝑒 𝐶)𝑥̃ + 𝐵𝑢 + 𝐾𝑒 𝑦 (3)

where 𝑥̃ is the estimated state and 𝐶 𝑥̃ is the estimated output. The inputs to the observer
are the output y and the control input u. 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 𝐶 𝑥̃. This term continuously corrects the model output and improves the
performance of the observer. Figure shows the block diagram of the system and the full-order state
observer.

[Link]’s College of Engineering 8 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Full-Order State Observer. The order of the state observer that will be discussed here is the same
as that of the plant. Assume that the plant is defined by Equations

𝑥̇ = 𝐴𝑥 + 𝐵𝑢 (1)
𝑦 = 𝐶𝑥 (2)

and the observer model is defined by Equation

𝑥̃ = 𝐴𝑥̃ + 𝐵𝑢 + 𝐾𝑒 (𝑦 − 𝐶𝑥̃)
= (𝐴 − 𝐾𝑒 𝐶)𝑥̃ + 𝐵𝑢 + 𝐾𝑒 𝑦 (3)

To obtain the observer error equation, let us subtract Equation (3) from Equation (1)

(4)
Define the difference between x and 𝑥̃ as the error vector e, or
𝑒 = 𝑥 − 𝑥̃
Then Equation (4) becomes
𝑒̇ = (𝐴 − 𝐾𝑒 𝐶)𝑒 (5)

From Equation (5), the dynamic behavior of the error vector is determined by the
eigenvalues of matrix 𝐴 − 𝐾𝑒 𝐶. If matrix 𝐴 − 𝐾𝑒 𝐶 is a stable matrix, the error vector will converge
to zero for any initial error vector e(0). That is, will converge to x(t) regardless of the values of x(0)
and If the eigenvalues of matrix 𝐴 − 𝐾𝑒 𝐶 are chosen in such a way that the dynamic behavior of
the error vector is asymptotically stable and is adequately fast, then any error vector will tend to
zero (the origin) with an adequate speed.

If the plant is completely observable, then it can be proved that it is possible to choose
matrix 𝐾𝑒 such that 𝐴 − 𝐾𝑒 𝐶 has arbitrarily desired eigenvalues. That is, the observer gain matrix
𝐾𝑒 can be determined to yield the desired matrix 𝐴 − 𝐾𝑒 𝐶.

Duality Property

The design of the full-order observer becomes that of determining an appropriate 𝐾𝑒 such
that 𝐴 − 𝐾𝑒 𝐶 has desired eigenvalues. Thus, the problem here becomes the same as the pole-
placement problem. In fact, the two problems are mathematically the same. This property is called
duality.

Consider the system defined by

[Link]’s College of Engineering 9 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
𝑥̇ = 𝐴𝑥 + 𝐵𝑢 (1)
𝑦 = 𝐶𝑥

In designing the full-order state observer, we may solve the dual problem, that is, solve the
pole-placement problem for the dual system.

𝑧̇ = 𝐴∗ 𝑧 + 𝐶 ∗ 𝑣
𝑛 = 𝐵∗𝑧

Assuming the control signal v to be 𝑣 = −𝐾𝑧

If the dual system is completely state controllable, then the state feedback gain matrix 𝐾
can be determined such that matrix 𝐴∗ − 𝐶 ∗ 𝐾 will yield a set of the desired eigenvalues. If
𝜇1 , 𝜇2 , 𝜇3 … . . 𝜇𝑛 are the desired eigenvalues of the state observer matrix, then by taking the same
𝜇𝑖′ 𝑠 as the desired eigenvalues of the state-feedback gain matrix of the dual system, we obtain

|𝑠𝐼 − (𝐴∗ − 𝐶 ∗ 𝐾)| = (𝑠 − 𝜇1 ) (𝑠 − 𝜇2 ) (𝑠 − 𝜇3 ) . .. (𝑠 − 𝜇𝑛 )

Noting that the eigenvalues of 𝐴∗ − 𝐶 ∗ 𝐾 and those of 𝐴 − 𝐾 ∗ 𝐶 are the same, we have,

|𝑠𝐼 − (𝐴∗ − 𝐶 ∗ 𝐾)| = |𝑠𝐼 − (𝐴 − 𝐾 ∗ 𝐶)|

Comparing the characteristic polynomial |𝑠𝐼 − (𝐴 − 𝐾 ∗ 𝐶)| and the characteristic polynomial for
the observer system |𝑠𝐼 − (𝐴 − 𝐾𝑒 𝐶)|, we find that 𝐾𝑒 and 𝐾 ∗ are related by

|𝑠𝐼 − (𝐴 − 𝐾 ∗ 𝐶)| = |𝑠𝐼 − (𝐴 − 𝐾𝑒 𝐶)|

Thus, using the matrix 𝐾 determined by the pole-placement approach in the dual system, the
observer gain matrix 𝐾𝑒 for the original system can be determined by using the relationship
𝐾𝑒 = 𝐾 ∗
Procedure for calculating Observer gain
Method I : Using Transformation Matrix T

Step 1: Check the observability of the given system. If it is observable, then go to step2

𝐶
𝐶𝐴
| 𝐶𝐴2 |
Observability matrix |𝑶𝑴| = ≠𝟎
| ; |
𝐶𝐴𝑛−2
𝐶𝐴𝑛−1
Step 2: Find the characteristic polynomial

|𝑺𝑰 − 𝑨| = 𝑺𝒏 + 𝒂𝟏 𝑺𝒏−𝟏 + 𝒂𝟐 𝑺𝒏−𝟐 +, … … … … + 𝒂𝒏−𝟏 𝑺 + 𝒂𝒏

and hence determine the values of 𝒂𝟏 , 𝒂𝟐 , 𝒂𝟑 , … … . 𝒂𝒏

Step 3: Determine the transformation matrix ′𝑸′

𝑄 = (𝑊𝑁 ∗ )−1

and

[Link]’s College of Engineering 10 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

𝑎𝑛−1 𝑎𝑛−2 ,…… 𝑎1 1

𝑎𝑛−2 𝑎𝑛−3 ,…… 1 0
𝑊= ; ; ,……. ; ;
𝑎𝑛−1 1 ,……. 0 0
[ 1 0 ,……. 0 0]

If the given system is in observable canonical form then, transformation matrix 𝑄 =

(𝑊𝑁 ∗ )−1 is 𝐼

Step 4: Using desired closed loop poles, write the desired polynomial. i.e.

(𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 ), … … … … … … (𝒔 − 𝝁𝒏 )

= 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 + 𝜶𝟐 𝒔𝒏−𝟐 +, … … … … + 𝜶𝒏−𝟏 𝒔 + 𝜶𝒏

and determine the values of 𝜶𝟏 , 𝜶𝟐 , 𝜶𝟑 , … … … … . . 𝜶𝒏

Step 5: The required state observer gain matrix ′𝑲𝒆 ′ is given by

𝜶𝒏 − 𝒂𝒏 𝜶𝒏 − 𝒂𝒏
𝜶𝒏−𝟏 − 𝒂𝒏−𝟏 𝜶𝒏−𝟏 − 𝒂𝒏−𝟏
𝑲𝒆 = 𝑸 ; = (𝑾𝑵∗ )−𝟏 ;
; ;
[ 𝜶𝟏 − 𝒂𝟏 ] [ 𝜶𝟏 − 𝒂𝟏 ]

Method 2: Direct comparison method

Step 1: Check the observability of the given system. If it is observable, then go to step2

𝐶
𝐶𝐴
| 𝐶𝐴2 |
Observability matrix |𝑶𝑴| = ≠𝟎
| ; |
𝐶𝐴𝑛−2
𝐶𝐴𝑛−1
𝒌𝒆𝟏
Step 2: 𝒏 = 𝟐, then 𝑲𝒆 = [ ]
𝒌𝒆𝟐

Step 3: Substitute this 𝑲𝒆 matrix into the desired characteristics polynomial |𝒔𝑰 − (𝑨 − 𝑲𝒆 𝑪)|
and equate it to (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 ).

|𝑺𝑰 − 𝑨 + 𝑩𝑲| = (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 )

Step 4: By equating the coefficients of like power of ′𝑺′ on both sides, it is possible to determine
the values of 𝑲𝒆𝟏 , 𝑲𝒆𝟐

Method III : Using Ackermann’s formula

Step 1: Check the observability of the given system. If it is observable, then go to step2

[Link]’s College of Engineering 11 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
𝐶
𝐶𝐴
| 𝐶𝐴2 |
Observability matrix |𝑶𝑴| = ≠𝟎
| ; |
𝐶𝐴𝑛−2
𝐶𝐴𝑛−1
Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

Step 4: Calculate the state feedback gain matrix, ′𝑲𝒆 ′ using the Ackermann’s formula
𝐶 −𝟏 𝟎
𝐶𝐴 𝟎
𝑲𝒆 = [∅(𝑨)] 𝐶𝐴2 ;
; ;
[𝐶𝐴 ]𝑛 [ 𝟏]

Minimum order Observer

In certain systems the state variables may not be available for measurement feedback. In such
situation we need to estimate the unmeasurable state variables from the knowledge of input and
output. Hence a state observer is employed which estimates the state variables from the input and
output of the systems. The estimated state variable can be used for feedback to design the system
by pole placement.
Full-order Observer: If the state observer observes all the 𝑛 number of state variable of the system
then it is called full-order observer.
Reduced-order Observer: If the state observer observes 𝑚 number of state variables, where 𝑚
less than 𝑛 then the observer is is called reduced-order observer.
Minimum-order Observer: If the order of the reduced-order observer is the minimum possible,
then the observer is called a minimum-order observer.

[Link]’s College of Engineering 12 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
If the measurement of output variables involves significant noises and is relatively inaccurate, then
use of full order observer may result in a better system performance
Consider the system

𝑥̇ = 𝐴𝑥 + 𝐵𝑢
𝑦 = 𝐶𝑥 (1)

Where, x = state vector (n-vector)

y = output vector (scalar)
u = control signal (scalar)
A = n × n constant matrix
B = n × 1 constant matrix
C = 1 × n constant matrix
The state 𝑥 can be partitioned into two parts 𝑥𝑎 and 𝑥𝑏 . Here the state variable 𝑥𝑎 is equal to the
output 𝑦 and thus can be directly measured and 𝑥𝑏 is the unmeasured portion of the state vector.
The portioned state and output equations become.

(2)
Where, 𝐴𝑎𝑎 = scalar
𝐴𝑎𝑏 = 1 × (𝑛 − 1) matrix
𝐴𝑏𝑎 = (𝑛 − 1) × 1 matrix
𝐴𝑏𝑏 = (𝑛 − 1) × (𝑛 − 1) matrix
𝐵𝑎 = scalar
𝐵𝑏 = (𝑛 − 1) × 1 matrix

From (2) the equation for the measure portion of the state becomes
[Link]’s College of Engineering 13 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
𝑥̇ 𝑎 = 𝐴𝑎𝑎 𝑥𝑎 + 𝐴𝑎𝑏 𝑋𝑏 + 𝐵𝑎 𝑢

𝑥̇ 𝑎 − 𝐴𝑎𝑎 𝑥𝑎 − 𝐵𝑎 𝑢 = 𝐴𝑎𝑏 𝑋𝑏 (3)

The term on the left side of (3) can be measured. It act as output equation. In designing minimum
order observer, the left side of (3) to be known quantities. Thus it relates the measurable quantites
and unmeasurable quantities of the state. From (1) the equation for the unmeasured portion of the
state becomes

𝑥̇ 𝑏 = 𝐴𝑏𝑎 𝑥𝑎 + 𝐴𝑏𝑏 𝑥𝑏 + 𝐵𝑏 𝑢 (4)

𝐴𝑏𝑎 𝑥𝑎 and 𝐵𝑏 𝑢 are known quantities and it describes the dynamics of unmeasured portion of the
state.
Design of Minimum order observer
The state equation for the full order observer is
𝑥̇ = 𝐴𝑥 + 𝐵𝑢
The state equation for the minimum order observer is
𝑥̇ 𝑏 = 𝐴𝑏𝑎 𝑥𝑎 + 𝐴𝑏𝑏 𝑥𝑏 + 𝐵𝑏 𝑢
The output equation for the full-order observer is
𝑦 = 𝐶𝑥
The output equation for the minimum-order observer is
𝑥̇ 𝑎 − 𝐴𝑎𝑎 𝑥𝑎 − 𝐵𝑎 𝑢 = 𝐴𝑎𝑏 𝑋𝑏

The observer equation for the full-order observer is

.
𝑥 = (𝐴 − 𝐾𝑒 𝐶)𝑥̂ + 𝐵𝑢 + 𝐾𝑒 𝑦
⏞ (5)

Then making the substitutions of table into equation (5),

.
𝑥 𝑏 = (𝐴𝑏𝑏 − 𝐾𝑒 𝐴𝑎𝑏 )𝑥̂𝑏 + 𝐴𝑏𝑎 𝑥𝑎 + 𝐵𝑏 𝑢 + 𝐾𝑒 (𝑥̇ 𝑎 − 𝐴𝑎𝑎 𝑥𝑎 − 𝐵𝑎 𝑢)
⏞ (6)

Where the state observer gain matrix 𝐾𝑒 is an (𝑛 − 1) × 1 matrix. To estimate 𝑥̂𝑏 we need the
derivative of 𝑥𝑎 . This presents a difficulty, because differentiation amplifies noise, if 𝑥𝑎 (= 𝑦)is
noisy, the use of 𝑥̇ 𝑎 is unacceptable.

Table 1. List of necessary substitutions for writing the observer equation for the minimum-order
observer

To avoid this difficulty 𝑥̇ 𝑎 should be eliminated

[Link]’s College of Engineering 14 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Equation (10) and (8) together defines the minimum-order observer.

Where 0 is a row vector consisting of (𝑛 − 1) zeros, if we define

By subtracting equation (2) from equation (4)

Define

Then Equation (3) becomes

This is the error equation for the minimum-order observer. 𝑒 is an (𝑛 − 1) vector.

Solved problem
[Link]’s College of Engineering 15 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
1. Consider the regulator system shown in following figure. The plant is given by
𝒙̇ 𝟏 𝟎 𝟏 𝟎 𝒙𝟏 𝟎
[𝒙̇ 𝟐 ] = [ 𝟎 𝟎 𝟏 ] [𝒙𝟐 ] + [𝟎] 𝒖(𝒕);
𝒙̇ 𝟑 −𝟏 −𝟓 −𝟔 𝒙𝟑 𝟏

The system uses the state feedback control 𝒖 = −𝑲𝒙. The desired eigenvalues are 𝝁𝟏 = −𝟐 +
𝒋𝟒, 𝝁𝟏 = −𝟐 − 𝒋𝟒, 𝝁𝟑 = −𝟏𝟎. Determine the state feedback gain matrix K

Solution

0 1 0 0
Given 𝐴 = [ 0 0 1 ]; 𝐵 = [0]; 𝜇1 = −2 + 𝑗4, 𝜇2 = −2 − 𝑗4, 𝜇3 = −10; 𝑂𝑟𝑑𝑒𝑟, 𝑛 = 3
−1 −5 −6 1
Method I : Using Transformation Matrix T

Step 1: Check the controllability of the given system. If it is controllable, then go to step2

Controllability matrix |𝑪𝑴| = |𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . 𝑨𝒏−𝟏 𝑩| ≠ 𝟎

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵 | ≠ 0
0
𝐵 = [0] ;
1
0 1 0 0 0
𝐴×𝐵 = [ 0 0 1 ] × [0] = [ 1 ]
−1 −5 −6 1 −6
0 1 0 0 1
𝐴2 𝐵 = 𝐴 × 𝐴𝐵 = [ 0 0 1 ] × [ 1 ] = [−6]
−1 −5 −6 −6 31
0 0 1
1 −6 0 −6 0 1
|𝐶𝑀| = |𝐵 𝐴𝐵 𝐴 𝐵 | = [0 1 −6] = 0 |
2 | − 0| | + 1| | = −1 ≠ 0
−6 31 1 31 1 −6
1 −6 31
Hence the system is completely state controllable.

Step 2: Find the characteristic polynomial

|𝑺𝑰 − 𝑨| = 𝑺𝒏 + 𝒂𝟏 𝑺𝒏−𝟏 + 𝒂𝟐 𝑺𝒏−𝟐 +, … … … … + 𝒂𝒏−𝟏 𝑺 + 𝒂𝒏

and hence determine the values of 𝒂𝟏 , 𝒂𝟐 , 𝒂𝟑 , … … . 𝒂𝒏

𝑠 0 0 0 1 0 𝑠 −1 0
|𝑆𝐼 − 𝐴| = |[0 𝑠 0] − [ 0 0 1 ]| = |0 𝑠 −1 |
0 0 𝑠 −1 −5 −6 1 5 𝑠+6

[Link]’s College of Engineering 16 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

𝑠 −1 0 −1 0 𝑠
= 𝑠| | − (−1) | | + 0| | = 𝑠(𝑠 2 + 6𝑠 + 5) + 1
5 𝑠+6 1 𝑠+6 1 5
= 𝑠 3 + 6𝑠 2 + 5𝑠 + 1 = 0

On comparing with 𝑠 3 + 𝑎1 𝑠 2 + 𝑎2 𝑠 + 𝑎3 we get, 𝑎1 = 6; 𝑎2 = 5; 𝑎3 = 1

Step 3: Determine the transformation matrix ′𝑻′

The given system is already in controllable canonical form, then 𝑇 = 𝐼

Step 4: Using desired closed loop poles, write the desired polynomial. i.e.

(𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 ), … … … … … … (𝒔 − 𝝁𝒏 )

= 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 + 𝜶𝟐 𝒔𝒏−𝟐 +, … … … … + 𝜶𝒏−𝟏 𝒔 + 𝜶𝒏

and determine the values of 𝜶𝟏 , 𝜶𝟐 , 𝜶𝟑 , … … … … . . 𝜶𝒏

The desired characteristic equation

(𝑠 + 2 + 𝑗4)(𝑠 + 2 − 𝑗4)(𝑠 + 10) = 𝑠 3 + 14𝑠 2 + 60𝑠 + 200

On comparing with 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 , we get,

𝛼1 = 14; 𝛼2 = 60; 𝛼3 = 200

Step 5: The required state feedback gain matrix ′𝑲′ is given by

𝑲 = [𝜶𝒏 − 𝒂𝒏 𝜶𝒏−𝟏 − 𝒂𝒏−𝟏 , . , , … . 𝜶𝟐 − 𝒂𝟐 𝜶𝟏 − 𝒂𝟏 ] 𝑻−𝟏

𝐾 = [𝛼3 − 𝑎3 𝛼2 − 𝑎 2 𝛼1 − 𝑎1 ]𝑇 −1

1 0 0
Since the system is in controllable canonical form 𝑇 = 𝐼 = [0 1 0]
0 0 1
1 0 0
𝐾 = [200 − 1 ]
60 − 5 14 − 6 [0 1 0]
0 0 1
Hence, 𝐾 = [199 55 8]
𝑥1
𝑢 = −𝐾𝑥 = −[𝐾1 𝐾2 𝐾3 ] [𝑥2 ]
𝑥3
𝑥1
= −[199 55 8] [𝑥2 ]
𝑥3
𝑢 = −199𝑥1 − 55𝑥2 − 8𝑥3

Method II : Using Direct Substitution method

Step 1: Check the controllability of the given system. If it is controllable, then go to step2

Controllability matrix |𝑪𝑴| = |𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . 𝑨𝒏−𝟏 𝑩| ≠ 𝟎

[Link]’s College of Engineering 17 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
Note: If the system is of low order (𝒏 ≤ 𝟑), direct substitution of matrix K into the
desired characteristics polynomial may be simpler.

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵 | ≠ 0
0
𝐵 = [0] ;
1
0 1 0 0 0
𝐴×𝐵 = [ 0 0 1 ] × [0] = [ 1 ]
−1 −5 −6 1 −6
0 1 0 0 1
𝐴2 𝐵 = 𝐴 × 𝐴𝐵 = [ 0 0 1 ] × [ 1 ] = [ −6]
−1 −5 −6 −6 31
0 0 1
1 −6 0 −6 0 1
|𝐶𝑀| = |𝐵 𝐴𝐵 𝐴2 𝐵 | = [0 1 −6] = 0 | | − 0| | + 1| | = −1 ≠ 0
−6 31 1 31 1 −6
1 −6 31
Hence the system is completely state controllable.

Step 2: If 𝒏 = 𝟑, then 𝑲 = [𝑲𝟏 𝑲𝟐 𝑲𝟑 ]

Here 𝑛 = 3, hence, 𝐾 = [𝐾1 𝐾2 𝐾3 ]

Step 3: Substitute this 𝑲 matrix into the desired characteristics polynomial |𝒔𝑰 − 𝑨 + 𝑩𝑲|
and equate it to (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 ).

|𝒔𝑰 − 𝑨 + 𝑩𝑲| = (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 )

Since both sides of this characteristic equation are polynomials in ′𝒔′.

𝑠 0 0 0 1 0 0
[𝑠𝐼 − 𝐴 + 𝐵𝐾] = [0 𝑠 0] − [ 0 0 1 ] + [0] [𝑘1 𝑘2 𝑘3 ]
0 0 𝑠 −1 −5 −6 1
𝑠 −1 0 0 0 0 𝑠 −1 0
= [0 𝑠 −1 ] + [ 0 0 0]=[ 0 𝑠 −1 ]
1 5 𝑠+6 𝑘1 𝑘2 𝑘3 1 + 𝑘1 5 + 𝑘2 𝑠 + 6 + 𝑘2

𝑠 −1 0
| 0 𝑠 −1 | = 𝑠 3 + (6 + 𝑘3 )𝑠 2 + (5 + 𝑘2 )𝑠 + 1 + 𝑘1
1 + 𝑘1 5 + 𝑘2 𝑠 + 6 + 𝑘2

(𝑠 − 𝜇1 )(𝑠 − 𝜇2 )(𝑠 − 𝜇3 ) = (𝑠 + 2 + 𝑗4)(𝑠 + 2 − 𝑗4)(𝑠 + 10)

= 𝑠 3 + 14𝑠 2 + 60𝑠 + 200

Step 4: By equating the coefficients of like power of ′𝒔′ on both sides, it is possible to determine
the values of 𝑲𝟏 , 𝑲𝟐 , 𝑲𝟑

|𝒔𝑰 − 𝑨 + 𝑩𝑲| = (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 )

𝑠 3 + (6 + 𝑘3 )𝑠 2 + (5 + 𝑘2 )𝑠 + 1 + 𝑘1 = 𝑠 3 + 14𝑠 2 + 60𝑠 + 200

6 + 𝑘3 = 14; 𝑘3 = 14 − 6 = 8
[Link]’s College of Engineering 18 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

5 + 𝑘2 = 60; 𝑘2 = 60 − 5 = 55

1 + 𝑘1 = 200; 𝑘1 = 200 − 1 = 199

Hence 𝐾 = [𝑘1 𝑘2
𝑘3 ] = [199 55 8]
𝑥1
𝑢 = −𝐾𝑥 = −[𝐾1 𝐾2 𝐾3 ] [𝑥2 ]
𝑥3
𝑥1
= −[199 55 8] [𝑥2 ]
𝑥3
𝑢 = −199𝑥1 − 55𝑥2 − 8𝑥3

Method III : Using Ackermann’s formula

Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵 | ≠ 0
0
𝐵 = [0] ;
1
0 1 0 0 0
𝐴×𝐵 = [ 0 0 1 ] × [0] = [ 1 ]
−1 −5 −6 1 −6
0 1 0 0 1
𝐴2 𝐵 = 𝐴 × 𝐴𝐵 = [ 0 0 1 ] × [ 1 ] = [ −6]
−1 −5 −6 −6 31
0 0 1
1 −6 0 −6 0 1
|𝐶𝑀| = |𝐵 𝐴𝐵 𝐴2 𝐵 | = [0 1 −6] = 0 | | − 0| | + 1| | = −1 ≠ 0
−6 31 1 31 1 −6
1 −6 31
Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏

i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏

and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .

The desired characteristic equation is

(𝑠 + 2 − 𝑗4) (𝑠 + 2 + 𝑗4) (𝑠 + 10) = (𝑠 2 + 4𝑠 + 4 + 16) (𝑠 + 10) = 0

= 𝑠 3 + 14𝑠 2 + 60𝑠 + 200 = 0
= 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 = 0

Hence, 𝛼1 = 14; 𝛼2 = 60; 𝛼3 = 200

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.

∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

[Link]’s College of Engineering 19 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

0 1 0
𝐴=[ 0 0 1]
−1 −5 −6
0 1 0 0 1 0 0 0 1
𝐴2 = 𝐴 × 𝐴 = [ 0 0 1 ] × [ 0 0 1 ] = [ −1 −5 −6]
−1 −5 −6 −1 −5 −6 6 29 31
0 0 1 0 1 0 −1 −5 −6
𝐴3 = 𝐴2 × 𝐴 = [−1 −5 −6] × [ 0 0 1 ]=[ 6 29 31 ]
6 29 31 −1 −5 −6 −31 −149 −157

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

0 1 0 3 0 1 0 2 0 1 0 1 0 0
=[ 0 0 1 ] + 14 [ 0 0 1 ] + 60 [ 0 0 1 ] + 200 [0 1 0]
−1 −5 −6 −1 −5 −6 −1 −5 −6 0 0 1
−1 −5 −6 0 0 1 0 1 0 1 0 0
=[ 6 29 31 ] + 14 [ −1 −5 −6] + 60 [ 0 0 1 ] + 200 [ 0 1 0]
−31 −149 −157 6 29 31 −1 −5 −6 0 0 1
199 55 8
= [ −8 159 7 ]
−7 −43 117
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝑲 = [𝟎 𝟎 𝟏] [𝑪𝑴]−𝟏 [∅(𝑨)]

0 0 1 −1 𝐴𝑑𝑗(𝐶𝑀)
𝐶𝑀−1 = [0 1 −6] =
|𝐶𝑀|
1 −6 31
0 0 1
1 −6 0 −6 0 1
|𝐶𝑀| = |𝐵 𝐴𝐵 2 |
𝐴 𝐵 = |0 1 −6| = 0 | | − 0| | +1| | = −1
−6 31 1 31 1 −6
1 −6 31

1 −6 0 −6 0 1 𝑇
+| | −| | +| |
−6 31 1 31 1 −6 −5 −6 −1 𝑇 −5 −6 −1
0 1 0 1 0 0
𝐴𝑑𝑗(𝐶𝑀) = − | | +| | −| | = [−6 −1 0 ] = [−6 −1 0 ]
−6 31 1 31 1 −6
0 1 0 1 0 0 −1 0 0 −1 0 0
[ + |1 −6
| −|
0 −6
| +|
0 1
|]

𝐴𝑑𝑗(𝐶𝑀) −5 −6 −1 5 6 1
𝐶𝑀−1 = = −1 [−6 −1 0 ] = [6 1 0]
|𝐶𝑀|
−1 0 0 1 0 0
5 6 1 199 55 8
𝐾 = [0 0 1] × [ 6 1 0] × [ −8 159 7 ]
1 0 0 −7 −43 117
199 55 8
𝐾 = [1 0 0] × [ −8 159 7 ] = [199 55 8]
−7 −43 117
[Link]’s College of Engineering 20 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

𝐾 = [𝑘1 𝑘2 𝑘3 ] = [199 55 8]
𝑥1
= −𝐾𝑥 = −[𝐾1 𝐾2 𝐾3 ] [𝑥2 ]
𝑥3
𝑥1
= −[199 55 8] [𝑥2 ]
𝑥3
𝑢 = −199𝑥1 − 55𝑥2 − 8𝑥3

2. A regulator system has the plant 𝒙̇ = 𝑨𝒙 + 𝑩𝒖; 𝒚 = 𝑪𝒙

𝟎 𝟎 −𝟔 𝟏
With 𝑨 = [𝟏 𝟎 −𝟏𝟏] ; 𝑩 = [𝟎] ; 𝑪 = [𝟎 𝟎 𝟏]
𝟎 𝟏 −𝟔 𝟎
Compute 𝑲 so that the control law 𝒖 = −𝑲𝒙, places the control loop poles at −𝟐 ±
𝐣 𝟑. 𝟒𝟔𝟒, −𝟑. Give the state variable model of the closed loop system
Solution

0 0 −6 1
Given 𝐴 = [1 0 −11]; 𝐵 = [0]; 𝜇1 = −2 + 𝑗3.464, 𝜇2 = −2 − 𝑗3.464, 𝜇3 = −3; 𝑛 = 3
0 1 −6 0
Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵 | ≠ 0
1
𝐵 = [0] ;
0
0 0 −6 1 0
𝐴 × 𝐵 = [1 0 −11] × [0] = [1]
0 1 −6 0 0
0 0 −6 0 0
2
𝐴 𝐵 = 𝐴 × 𝐴𝐵 = [1 0 −11] × [1] = [0]
0 1 −6 0 1
1 0 0
1 0 0 0 0 1
|𝐶𝑀| = |𝐵 𝐴𝐵 𝐴 𝐵2| = |0 1 0| = 1 | | − 0| |+ 0| |=1≠0
0 1 0 1 0 0
0 0 1
Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .

The desired characteristic equation is

(𝑠 + 2 − 𝑗3.464) (𝑠 + 2 + 𝑗3.464) (𝑠 + 3) = (𝑠 2 + 4𝑠 + 16) (𝑠 + 3) = 0

= 𝑠 3 + 7𝑠 2 + 28𝑠 + 48 = 0
= 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 = 0

[Link]’s College of Engineering 21 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
Hence, 𝛼1 = 7; 𝛼2 = 28; 𝛼3 = 48

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

0 0 −6
𝐴 = [1 0 −11]
0 1 −6
0 0 −6 0 0 −6 0 −6 36
𝐴2 = 𝐴 × 𝐴 = [1 0 −11] × [1 0 −11] = [0 −11 60]
0 1 −6 0 1 −6 1 −6 25
0 −6 36 0 0 −6 −6 36 −150
3 2
𝐴 = 𝐴 × 𝐴 = [0 −11 60] × [1 0 −11] = [−11 60 −239]
1 −6 25 0 1 −6 −6 25 −90

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

0 0 −6 3 0 0 −6 2 0 0 −6 1 0 0
=[1 0 −11] + 7 [1 0 −11] + 28 [1 0 −11] + 48 [0 1 0]
0 1 −6 0 1 −6 0 1 −6 0 0 1
−6 36 −150 0 −6 36 0 0 −6 1 0 0
= [−11 60 −239] + 7 [0 −11 60] + 28 [1 0 −11] + 48 [0 1 0]
−6 25 −90 1 −6 25 0 1 −6 0 0 1
42 −6 −66
= [17 31 −127]
1 11 −35
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝑲 = [𝟎 𝟎 𝟏] [𝑪𝑴]−𝟏 [∅(𝑨)]

1 0 0 −1 𝐴𝑑𝑗(𝐶𝑀)
𝐶𝑀−1 = [0 1 0] =
|𝐶𝑀|
0 0 1
1 0 0
1 0 0 0 0 1
|𝐶𝑀| = |𝐵 𝐴𝐵 2
𝐴 𝐵 | = |0 1 0| = 1 | | − 0| |+ 0| |=1
0 1 0 1 0 0
0 0 1

1 0 0 0 0 1 𝑇
+| | −| | +| |
0 1 0 1 0 0 1 0 0𝑇 1 0 0
0 0 1 0 1 0
𝐴𝑑𝑗(𝐶𝑀) = − | | +| | −| | = [0 1 0] = [0 1 0]
0 1 0 1 0 0
0 0 1 0 1 0 0 0 1 0 0 1
[+ |1 0
| −|
0 0
| +|
0 1
|]

𝐴𝑑𝑗(𝐶𝑀) 1 0 0 1 0 0
𝐶𝑀−1 = = 1 [0 1 0 ] = [ 0 1 0]
|𝐶𝑀|
0 0 1 0 0 1
1 0 0 42 −6 −66
𝐾 = [0 0 1] × [0 1 0] × [17 31 −127]
0 0 1 1 11 −35
[Link]’s College of Engineering 22 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

42 −6 −66
𝐾 = [0 0 1] × [17 31 −127] = [1 11 −35]
1 11 −35
𝐾 = [𝑘1 𝑘2 𝑘3 ] = [1 11 −35]
𝑥1
𝑢 = −𝐾𝑥 = −[𝐾1 𝐾2 𝐾3 ] [𝑥2 ]
𝑥3
𝑥1
= −[1 11 −35] [𝑥2 ]
𝑥3
𝑢 = −1𝑥1 − 11𝑥2 + 35𝑥3

3. Consider the regulator system shown in following figure.

The plant is given by

𝐱̇ 𝟏 𝟏 𝟐 𝟏 𝐱𝟏 𝟎
[𝐱̇ 𝟐 ] = [𝟎 𝟏 𝟑] [𝐱 𝟐 ] + [𝟎] 𝐮(𝐭);
𝐱̇ 𝟑 𝟏 𝟏 𝟏 𝐱𝟑 𝟏

Determine the state feedback gain for each state variable to place the poles at −𝟏 + 𝐣, −𝟏 −
𝐣, −𝟑
Solution

1 2 1 0
Given 𝐴 = [0 1 3]; 𝐵 = [0]; 𝜇1 = −1 + 𝑗, 𝜇2 = −1 − 𝑗, 𝜇3 = −3; 𝑜𝑟𝑑𝑒𝑟, 𝑛 = 3
1 1 1 1
Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵 | ≠ 0
0
𝐵 = [0] ;
1
1 2 1 0 1
𝐴 × 𝐵 = [0 1 3] × [0] = [3]
1 1 1 1 1
1 2 1 1 8
𝐴2 𝐵 = 𝐴 × 𝐴𝐵 = [0 1 3] × [ 3] = [ 6]
1 1 1 1 5
0 1 8
3 6 0 6 0 3
|𝐶𝑀| = |𝐵 𝐴𝐵 𝐴2 𝐵 | = |0 3 6| = 0 | | − 1| |+ 8| | = −18 ≠ 0
1 5 1 5 1 1
1 1 5
[Link]’s College of Engineering 23 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .

The desired characteristic equation is

(𝑠 + 1 − 𝑗) (𝑠 + 1 + 𝑗) (𝑠 + 3) = (𝑠 2 + 2𝑠 + 2) (𝑠 + 3) = 0
= 𝑠 3 + 5𝑠 2 + 8𝑠 + 6 = 0
= 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 = 0

Hence, 𝛼1 = 5; 𝛼2 = 8; 𝛼3 = 6

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

1 2 1
𝐴 = [0 1 3]
1 1 1
1 2 1 1 2 1 2 5 8
2
𝐴 = 𝐴 × 𝐴 = [0 1 3 ] × [0 1 3] = [3 4 6]
1 1 1 1 1 1 2 4 5
2 5 8 1 2 1 10 17 25
𝐴3 = 𝐴2 × 𝐴 = [3 4 6] × [0 1 3] = [ 9 16 21]
2 4 5 1 1 1 7 13 19

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

1 2 13 1 2 12 1 2 1 1 0 0
=[0 1 3] + 5 [0 1 3] + 8 [0 1 3] + 6 [0 1 0]
1 1 1 1 1 1 1 1 1 0 0 1
10 17 25 2 5 8 1 2 1 1 0 0 34 58 73
=[9 16 21] + 5 [3 4 6] + 8 [0 1 3] + 6 [0 1 0] = [24 50 75]
7 13 19 2 4 5 1 1 1 0 0 1 25 41 58
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝑲 = [𝟎 𝟎 𝟏] [𝑪𝑴]−𝟏 [∅(𝑨)]

0 1 8 −1 𝐴𝑑𝑗(𝐶𝑀)
𝐶𝑀−1 = [0 3 6] =
|𝐶𝑀|
1 1 5
0 1 8
3 6 0 6 0 3
|𝐶𝑀| = |𝐵 𝐴𝐵 2
𝐴 𝐵 | = [ 0 3 6] = 0 | | − 1| | + 8| | = −18
1 5 1 5 1 1
1 1 5

[Link]’s College of Engineering 24 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
3 6 0 6 0 3 𝑇
+| | −| | +| |
1 5 1 5 1 1 9 6 −3 𝑇 9 3 −18
1 8 0 8 0 1
𝐴𝑑𝑗(𝐶𝑀) = − | | +| | −| | =[ 3 −8 1 ] = [ 6 −8 0 ]
1 5 1 5 1 1
1 8 0 8 0 1 −18 0 0 −3 1 0
[+ |3 6
| −|
0 6
| +|
0 3
|]

−1
𝐴𝑑𝑗(𝐶𝑀) 1 9 3 −18 −0.5 −0.1667 1
𝐶𝑀 == = − [ 6 −8 0 ] = [−0.3333 0.4444 0]
|𝐶𝑀| 18
−3 2 0 0.1667 −0.0556 0
−0.5 −0.1667 1 34 58 73
𝐾 = [0 0 1] × [−0.3333 0.4444 0] × [24 50 75]
0.1667 −0.0556 0 25 41 58
34 58 73
𝐾 = [0.1667 −0.0556 0] × [24 50 75] = [4.3334 6.8886 7.9991]
25 41 58
𝐾 = [𝑘1 𝑘2 𝑘3 ] = [4.3334 6.8886 7.9991]

4. A single input system is described by the following state equation

−𝟎. 𝟏𝟒 𝟎. 𝟑𝟑 −𝟎. 𝟑𝟑 𝟎
𝑿̇ = [ 𝟎. 𝟏 −𝟎. 𝟐𝟖 𝟎 ] 𝑿 + [ 𝟎 ] 𝑼; 𝒀 = [𝟐 𝟎 𝟎] 𝑿
𝟎 𝟏. 𝟕 −𝟎. 𝟕𝟕 −𝟎. 𝟎𝟐𝟓

Design a state feedback controller which will give closed loop poles at −𝟎. 𝟔𝟕, −𝟎. 𝟔𝟕 & −
𝟎. 𝟔𝟕

Solution

−0.14 0.33 −0.33 0

Given, 𝐴 = [ 0.1 −0.28 0 ] ; 𝐵 = [ 0 ] ; 𝜇1 = −0.67, 𝜇2 = −0.67, 𝜇3 =
0 1.7 −0.77 −0.025
−0.67; 𝑛 = 3

Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵 | ≠ 0
0
𝐵=[ 0 ]
−0.025
−0.14 0.33 −0.33 0 0.0083
𝐴 × 𝐵 = [ 0.1 −0.28 0 ] ×[ 0 ] =[ 0 ]
0 1.7 −0.77 −0.025 0.0193
−0.14 0.33 −0.33 0.0083 −0.0075
𝐴2 𝐵 = 𝐴 × 𝐴𝐵 = [ 0.1 −0.28 0 ] × [ 0 ] = [ 0.0008 ]
0 1.7 −0.77 0.0193 −0.0148
0 0.0083 −0.0075
|𝐶𝑀| = |𝐵 𝐴𝐵 𝐴 𝐵| = | 0
2 0 0.0008 |
−0.025 0.0193 −0.0148

[Link]’s College of Engineering 25 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

0 0.0008 0 0.0008 0 0
= 0| | − 0.0083 | | + −0.0075 | |
0.0193 −0.0148 −0.025 −0.0148 −0.025 0.0193
= −1.66 × 10−7 ≠ 0

Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .
The desired characteristic equation is
(𝑠 + 0.67) (𝑠 + 0.67) (𝑠 + 0.67) = 𝑠 3 + 2.01𝑠 2 + 1.3467𝑠 + 0.3008 = 0
= 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 = 0

Hence, 𝛼1 = 2.01; 𝛼2 = 1.3467; 𝛼3 = 0.3008

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

−0.14 0.33 −0.33

𝐴 = [ 0.1 −0.28 0 ]
0 1.7 −0.77
−0.14 0.33 −0.33 −0.14 0.33 −0.33
2
𝐴 = 𝐴 × 𝐴 = [ 0.1 −0.28 0 ] × [ 0.1 −0.28 0 ]
0 1.7 −0.77 0 1.7 −0.77
0.0526 −0.6996 0.3003
= [−0.042 0.1114 −0.033]
0.17 −1.785 0.5929
0.0526 −0.6996 0.3003 −0.14 0.33 −0.33
𝐴3 = 𝐴2 × 𝐴 = [−0.042 0.1114 −0.033] × [ 0.1 −0.28 0 ]
0.17 −1.785 0.5929 0 1.7 −0.77
−0.0773 0.7238 −0.2486
= [ 0.017 −0.1012 0.0393 ]
−0.2023 1.5638 −0.5126

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

−0.14 0.33 −0.33 3 −0.14 0.33 −0.33 2

= [ 0.1 −0.28 0 ] + 2.01 [ 0.1 −0.28 0 ] +
0 1.7 −0.77 0 1.7 −0.77
−0.14 0.33 −0.33 1 0 0
1.3467 [ 0.1 −0.28 0 ] + 0.3008 [0 1 0]
0 1.7 −0.77 0 0 1

−0.0773 0.7238 −0.2486 0.0526 −0.6996 0.3003

=[ 0.017 −0.1012 0.0393 ] + 2.01 [−0.042 0.1114 −0.033] +
−0.2023 1.5638 −0.5126 0.17 −1.785 0.5929
−0.14 0.33 −0.33 1 0 0
1.3467 [ 0.1 −0.28 0 ] + 0.3008 [ 0 1 0]
0 1.7 −0.77 0 0 1
[Link]’s College of Engineering 26 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

0.1407 −0.238 −0.0894

= [0.0673 0.0465 −0.0271]
0.1394 0.2654 −0.0571
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝑲 = [𝟎 𝟎 𝟏] [𝑪𝑴]−𝟏 [∅(𝑨)]

0 0.0083 −0.0075 −1 𝐴𝑑𝑗(𝐶𝑀)

𝐶𝑀−1 =[ 0 0 0.0008 ] =
|𝐶𝑀|
−0.025 0.0193 −0.0148
0 0.0083 −0.0075
|𝐶𝑀| = | 0 0 0.0008 |
−0.025 0.0193 −0.0148
0 0.0008 0 0.0008 0 0
= 0| | − 0.0083 | | + −0.0075 | |
0.0193 −0.0148 −0.025 −0.0148 −0.025 0.0193
|𝐶𝑀| = −1.66 × 10−7
𝑇
0 0.0008 0 0.0008 0 0
+| | −| | +| |
0.0193 −0.0148 −0.025 −0.0148 −0.025 0.0193
0.0083 −0.0075 0 −0.0075 0 0.0083
𝐴𝑑𝑗(𝐶𝑀) = − | | +| | −| |
0.0193 −0.0148 −0.025 −0.0148 −0.025 0.0193
0.0083 −0.0075 0 −0.0075 0 0.0083
[+ | 0 0.0008
| −|
0 0.0008
| +|
−0.025
|
0.0193 ]
𝑇
−0.0000159 −0.0000206 0
= [−0.00002222 −0.0001877 −0.0002063]
0.0000068 0 0
−0.0000159 −0.0000222 0.0000068
= [−0.0000206 −0.0001877 0 ]
0 −0.0002063 0

𝐴𝑑𝑗(𝐶𝑀) 1 −0.0000159 −0.0000222 0.0000068

𝐶𝑀−1 = =− [−0.0000206 −0.0001877 0 ]
|𝐶𝑀| 1.66 × 10−7
0 −0.0002063 0
93.3 130.7 −40
𝐶𝑀−1 = [121.2 1103 0 ]
0 1212.1 0
93.3 130.7 −40 0.1407 −0.238 −0.0894
𝐾 = [0 0 1] × [ 121.2 1103 0 ] × [ 0.0673 0.0465 −0.0271]
0 1212.1 0 0.1394 0.2654 −0.0571
0.1407 −0.238 −0.0894
𝐾 = [0 1212.1 0] × [0.0673 0.0465 −0.0271] = [81.5394 56.3018 −32.8]
0.1394 0.2654 −0.0571
𝐾 = [𝑘1 𝑘2 𝑘3 ] = [81.5394 56.3018 −32.8]

[Link]’s College of Engineering 27 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
𝟏 𝟏 −𝟐 𝟏
5. Find the state feedback gain K for the state equation 𝑿̇ = [𝟎 𝟏 𝟏 ] 𝑿 + [𝟎] 𝑼, so that
𝟎 𝟎 𝟏 𝟏
the resulting system has eigenvalues −𝟐, −𝟏 ± 𝒋

Solution

1 1 −2 1
Given, 𝐴 = [0 1 1 ] ; 𝐵 = [0] ; 𝜇1 = −2, 𝜇2 = −1 + 𝑗, 𝜇3 = −1 − 𝑗; 𝑛 = 3
0 0 1 1
Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵 | ≠ 0
1
𝐵 = [0]
1
1 1 −2 1 −1
𝐴 × 𝐵 = [0 1 1 ] × [0 ] = [ 1 ]
0 0 1 1 1
1 1 −2 −1 −2
2
𝐴 𝐵 = 𝐴 × 𝐴𝐵 = [0 1 1 ] ×[ 1 ]=[ 2 ]
0 0 1 1 1
1 −1 −2
|𝐶𝑀| = |𝐵 | 1 2 0 2 0 1
𝐴𝐵 2
𝐴 𝐵 = |0 1 2 | = 1| | − (−1) | | − 2| | = −1 ≠ 0
1 1 1 1 1 1
1 1 1
Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .
The desired characteristic equation is
(𝑠 + 2) (𝑠 + 1 + 𝑗) (𝑠 + 1 − 𝑗) = 0
= 𝑠 3 + 4𝑠 2 + 6𝑠 + 4 = 0
= 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 = 0

Hence, 𝛼1 = 4; 𝛼2 = 6; 𝛼3 = 4

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

1 1 −2
𝐴 = [0 1 1]
0 0 1
1 1 −2 1 1 −2 1 2 −3
𝐴2 = 𝐴 × 𝐴 = [0 1 1 ] × [0 1 1 ] = [0 1 2 ]
0 0 1 0 0 1 0 0 1

[Link]’s College of Engineering 28 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

1 2 −3 1 1 −2 1 3 −3
𝐴3 = 𝐴2 × 𝐴 = [0 1 2 ] × [0 1 1 ] = [0 1 3]
0 0 1 0 0 1 0 0 1

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

1 1 −2 3 1 1 −2 2 1 1 −2 1 0 0
= [0 1 1 ] + 4 [0 1 1 ] + 6 [0 1 1 ] + 4 [0 1 0]
0 0 1 0 0 1 0 0 1 0 0 1

1 3 −3 1 2 −3 1 1 −2 1 0 0 15 17 −27
=[0 1 3 ] + 4 [0 1 2 ] + 6 [0 1 1 ] + 4 [0 1 0] = [ 0 15 17 ]
0 0 1 0 0 1 0 0 1 0 0 1 0 0 15
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝑲 = [𝟎 𝟎 𝟏] [𝑪𝑴]−𝟏 [∅(𝑨)]

1 −1 −2 −1 𝐴𝑑𝑗(𝐶𝑀)
𝐶𝑀−1 = [0 1 2] =
|𝐶𝑀|
1 1 1
1 −1 −2
|𝐶𝑀| = |0 1 1 2 0 2 0 1
2 | = 1| | − (−1) | | − 2| | = −1
1 1 1 1 1 1
1 1 1

1 2 0 2 0 1 𝑇
+| | −| | +| |
1 1 1 1 1 1 −1 2 −1 𝑇 −1 −1 0
−1 −2 1 −2 1 −1
𝐴𝑑𝑗(𝐶𝑀) = − | | +| | −| | = [−1 3 −2] = [ 2 3 −2]
1 1 1 1 1 1
−1 −2 1 −2 1 −1 0 −2 1 −1 −2 1
[+ | 1 2
| −|
0 2
| +|
0 1
|]

𝐴𝑑𝑗(𝐶𝑀) −1 −1 0 1 1 0
𝐶𝑀−1 == = −1 [ 2 3 −2 ] = [ −2 −3 2 ]
|𝐶𝑀|
−1 −2 1 1 2 −1
1 1 0 15 17 −27
𝐾 = [0 0 1] × [−2 −3 2 ] × [ 0 15 17 ]
1 2 −1 0 0 15
15 17 −27
𝐾 = [1 2 ]
−1 × [ 0 15 17 ] = [12 47 −8]
0 0 15
𝐾 = [𝑘1 𝑘2 𝑘3 ] = [12 47 −8]

6. A single input system is described by the following state equation

−𝟏 𝟎 𝟎 𝟎
̇𝑿 = [ 𝟎 −𝟐 𝟎 ] 𝑿 + [𝟎] 𝑼; 𝒀 = [𝟏 𝟎 𝟎]𝑿
𝟎 𝟎 −𝟑 𝟏

[Link]’s College of Engineering 29 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
By using state feedback control 𝒖 = −𝑲𝒙 it is desired to have the closed loop poles at
−𝟔, −𝟏 ± 𝟐𝒋. Determine the state feedback matrix K

Solution

−1 0 0 0
Given, 𝐴 = [ 0 −2 0 ] ; 𝐵 = [0] ; 𝜇1 = −6, 𝜇2 = −1 + 2𝑗, 𝜇3 = −1 − 2𝑗; 𝑛 = 3
0 0 −3 1
Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵 | ≠ 0
0
𝐵 = [0]
1
−1 0 0 0 0
𝐴 × 𝐵 = [ 0 −2 0 ] × [0] = [ 0 ]
0 0 −3 1 −3
−1 0 0 0 0
2
𝐴 𝐵 = 𝐴 × 𝐴𝐵 = [ 0 −2 0 ] × [ 0 ] = [0]
0 0 −3 −3 9
0 0 0
0 0 0 0 0 0
|𝐶𝑀| = |𝐵 𝐴𝐵 2
𝐴 𝐵 | = |0 0 0| = 0 | | − 0| | + 0| |=0
−3 9 1 9 1 −3
1 −3 9
Hence the system is NOT state controllable. Design of state feedback controller is not possible.

𝐱̇ 𝟐 𝟏 𝐱𝟏 𝟏 𝐱𝟏
7. Given [ 𝟏 ] = [ ] [𝐱 ] + [ ] 𝐮; 𝐲 = [𝟏 𝟏] [𝐱 ]. Find the state feedback gain K, so
𝐱̇ 𝟐 −𝟏 𝟏 𝟐 𝟐 𝟐
that the state feedback system has -1 and -2 as its eigenvalues.

Solution

2 1 1
Given, 𝐴 = [ ] ; 𝐵 = [ ] ; 𝜇1 = −1, 𝜇2 = −2
−1 1 2
Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 | ≠ 0

1
𝐵=[ ]
2
2 1 1 4
𝐴×𝐵 = [ ] ×[ ]= [ ]
−1 1 2 1

|𝐶𝑀| = |𝐵 1 4
𝐴𝐵 | = [ ] = −7 ≠ 0
2 1
Hence the system is completely state controllable.
[Link]’s College of Engineering 30 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .
The desired characteristic equation is
(𝑠 + 1) (𝑠 + 2) = 0
= 𝑠 2 + 3𝑠 + 2 = 0
= 𝑠 2 + 𝛼1 𝑠 + 𝛼2 = 0

Hence, 𝛼1 = 3; 𝛼2 = 2

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

2 1
𝐴=[ ]
−1 1
2 1 2 1 3 3
𝐴2 = 𝐴 × 𝐴 = [ ]×[ ]=[ ]
−1 1 −1 1 −3 0

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

2 12 2 1 1 0
=[ ] + 3[ ] + 2[ ]
−1 1 −1 1 0 1
3 3 2 1 1 0 11 6
=[ ] + 3[ ]+ 2[ ]=[ ]
−3 0 −1 1 0 1 −6 5
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝑲 = [𝟎 𝟏] [𝑪𝑴]−𝟏 [∅(𝑨)]

1 4 −1 𝐴𝑑𝑗(𝐶𝑀)
𝐶𝑀−1 = [ ] =
2 1 |𝐶𝑀|

|𝐶𝑀| = |𝐵 1 4
𝐴𝐵 | = | | = −7
2 1
1 4 1 −4
𝐴𝑑𝑗(𝐶𝑀) = 𝑎𝑑𝑗 [ ]=[ ]
2 1 −2 1
𝐴𝑑𝑗(𝐶𝑀) 1 1 −4 −0.1429 0.5714
𝐶𝑀−1 = =− [ ]=[ ]
|𝐶𝑀| 7 −2 1 0.2857 −0.1429

−0.1429 0.5714 11 6
𝐾 = [0 1] × [ ]×[ ]
0.2857 −0.1429 −6 5
11 6
𝐾 = [0.2857 −0.1429] × [ ] = [4 1]
−6 5
𝐾 = [𝑘1 𝑘2 ] = [4 1]

[Link]’s College of Engineering 31 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
𝟎 𝟏 𝟏
8. Consider the system defined by 𝐱̇ = 𝐀𝐱 + 𝐁𝐮 where, 𝐀 = [ ] ; 𝐁 = [ ]. The
𝟐𝟎. 𝟔 𝟎 𝟎
characteristic equation for the system is

|𝑺𝑰 − 𝑨| = | 𝟎 𝟏
| = 𝑺𝟐 − 𝟐𝟎. 𝟔 = 𝟎
𝟐𝟎. 𝟔 𝒔
Since the characteristic roots are 𝑺 = ±𝟒. 𝟓𝟑𝟗, the system is unstable. By using the state
feedback control 𝒖 = −𝑲𝒙, it is desired to have closed loop poles at 𝑺 = −𝟏. 𝟖 ± 𝒋𝟐. 𝟒 ( that is
the eigenvalues of 𝑨 − 𝑩𝑲 to be 𝝁𝟏 = −𝟏. 𝟖 + 𝒋𝟐. 𝟒 and 𝝁𝟐 = −𝟏. 𝟖 − 𝒋𝟐. 𝟒. determine the
state feedback gain matrix K

Solution

𝟎 𝟏 𝟏
Given, 𝐴 = [ ] ; 𝐵 = [ ] ; 𝝁𝟏 = −𝟏. 𝟖 + 𝒋𝟐. 𝟒 ; 𝝁𝟐 = −𝟏. 𝟖 − 𝒋𝟐. 𝟒
𝟐𝟎. 𝟔 𝟎 𝟎
Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 | ≠ 0

1
𝐵=[ ]
0
0 1 1 0
𝐴×𝐵 = [ ] ×[ ]=[ ]
20.6 0 0 20.6
1 0
|𝐶𝑀| = |𝐵 𝐴𝐵 | = [ ] = 20.6 ≠ 0
0 20.6
Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .
The desired characteristic equation is
(𝑠 + 1.8 + 𝑗2.4) (𝑠 + 1.8 − 𝑗2.4) = 0
= 𝑠 2 + 3.6𝑠 + 9 = 0
= 𝑠 2 + 𝛼1 𝑠 + 𝛼2 = 0

Hence, 𝛼1 = 3.6; 𝛼2 = 9

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

0 1
𝐴=[ ]
20.6 0
0 1 0 1 20.6 0
𝐴2 = 𝐴 × 𝐴 = [ ]×[ ]=[ ]
20.6 0 20.6 0 0 20.6

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

0 12 0 1 1 0
=[ ] + 3.6 [ ] + 9[ ]
20.6 0 20.6 0 0 1
[Link]’s College of Engineering 32 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
20.6 0 0 1 1 0 29.6 3.6
=[ ] + 3.6 [ ] + 9[ ]=[ ]
0 20.6 20.6 0 0 1 74.16 29.6
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝑲 = [𝟎 𝟏] [𝑪𝑴]−𝟏 [∅(𝑨)]

1 0 −1 𝐴𝑑𝑗(𝐶𝑀)
𝐶𝑀−1 = [ ] =
0 20.6 |𝐶𝑀|

1 0
|𝐶𝑀| = |𝐵 𝐴𝐵 | = | | = 20.6
0 20.6
1 0 20.6 0
𝐴𝑑𝑗(𝐶𝑀) = 𝑎𝑑𝑗 [ ]=[ ]
0 20.6 0 1
𝐴𝑑𝑗(𝐶𝑀) 1 20.6 0 1 0
𝐶𝑀−1 = = [ ]=[ ]
|𝐶𝑀| 20.6 0 1 0 0.0485

1 0 29.6 3.6
𝐾 = [0 1] × [ ]×[ ]
0 0.0485 74.16 29.6
29.6 3.6
𝐾 = [0 0.0485] × [ ] = [3.6 1.4369]
74.16 29.6
𝐾 = [𝐾1 𝐾2 ] = [3.6 1.4369]
𝑥1
𝑢 = −𝐾𝑥 = −[𝐾1 𝐾2 ] [𝑥 ]
2
𝑥1
= −[3.6 1.4369] [𝑥 ]
2
𝑢 = −3.6𝑥1 − 1.4369𝑥2

9. A single input system is described by the following state equation

𝒙̇ 𝟏 −𝟏 𝟎 𝟎 𝒙𝟏 𝟏𝟎
[𝒙̇ 𝟐 ] = = [ 𝟏 −𝟐 𝟎 ] [𝒙𝟐 ] + [ 𝟏 ] 𝑼
𝒙̇ 𝟑 𝟐 𝟏 −𝟑 𝒙𝟑 𝟎
Design a state feedback controller which will give closed loop poles at −𝟏 + 𝒋𝟐, −𝟏 − 𝒋𝟐, 𝟐 (May
2017)

Step 1: Check the controllability of the given system. If it is controllable, then go to step2

Controllability matrix |𝐶𝑀| = |𝐵 𝐴𝐵 𝐴2 𝐵 , … … . 𝐴𝑛−1 𝐵| ≠ 0

Note: If the system is of low order (𝑛 ≤ 3), direct substitution of matrix K into the desired
characteristics polynomial may be simpler.

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵| ≠ 0

10
𝐵 = [ 1 ];
0
−1 0 0 10 −10
𝐴 × 𝐵 = [ 1 −2 0 ] × [ 1 ] = [ 8 ]
2 1 −3 1 21

[Link]’s College of Engineering 33 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

−1 0 0 −10 10
𝐴2 𝐵 = 𝐴 × 𝐴𝐵 = [ 1 −2 0 ] × [ 8 ] = [−26]
2 1 −3 21 −75
10 −10 10
|𝐶𝑀| = |𝐵 𝐴𝐵 2
𝐴 𝐵 | = [ 1 8 −26]
0 21 −75
8 −26 1 −26 1 8
= 10 | | − (−10) | | + 10 | | = −1080 ≠ 0
21 −75 0 −75 0 21

Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles, 𝜇1 , 𝜇2 , 𝜇3 , . . , 𝜇𝑛

i.e., (𝑠 − 𝜇1 ) (𝑠 − 𝜇2 ) (𝑠 − 𝜇3 ). . . . . . . . (𝑠 − 𝜇𝑛 ) = 𝑠 𝑛 + 𝛼1 𝑠 𝑛−1 +, … … . . , +𝛼𝑛−1 𝑠1 + 𝛼𝑛

and determine the values of 𝛼1 , 𝛼2 , … … . , 𝛼𝑛 .

The desired characteristic equation is

(𝑠 − 𝜇1 )(𝑠 − 𝜇2 )(𝑠 − 𝜇3 ) = (𝑠 − (−1 + j2))(𝑠 − (−1 − j2))(𝑠 − 2) = 𝑠 3 + 𝑠 − 10

= 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 = 0

Hence, 𝛼1 = 0; 𝛼2 = 1; 𝛼3 = −10

Step 3: Determine the matrix ∅(𝐴) using the coefficient of desired characteristics polynomial.

∅(𝐴) = 𝐴𝑛 + 𝛼1 𝐴𝑛−1 + 𝛼2 𝐴𝑛−2 +, … … . . , +𝛼𝑛−1 𝐴1 + 𝛼𝑛 𝐼

−1 0 0
𝐴 = [ 1 −2 0 ]
2 1 −3
−1 0 0 −1 0 0 1 0 0
𝐴2 = 𝐴 × 𝐴 = [ 1 −2 0 ] × [ 1 −2 0 ] = [−3 4 0]
2 1 −3 2 1 −3 −7 −5 9
1 0 0 −1 0 0 −1 0 0
𝐴3 = 𝐴2 × 𝐴 = [−3 4 0] × [ 1 −2 0 ] = [ 7 −8 0 ]
−7 −5 9 2 1 −3 20 −19 −27

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

−1 0 0 3 −1 0 0 2 −1 0 0 1 0 0
=[ 1 −2 0 ] + 0 [ 1 −2 0 ] + 1 [ 1 −2 0 ] − 10 [0 1 0]
2 1 −3 2 1 −3 2 1 −3 0 0 1
−1 0 0 1 0 0 −1 0 0 1 0 0
=[ 7 −8 0 ] + 0 [−3 4 0] + 1 [ 1 −2 0 ] − 10 [0 1 0]
20 −19 −27 −7 −5 9 2 1 −3 0 0 1
−12 0 0
=[ 8 −20 0 ]
20 19 −27

Step 4: Calculate the state feedback gain matrix, ′𝐾′ using the Ackermann’s formula
𝐾 = [0 0 0 , … … . 1] [𝐵 𝐴𝐵 𝐴2 𝐵 , … … . . 𝐴𝑛−1 𝐵]−1 [∅(𝐴)]

𝐾 = [0 0 1] [𝐶𝑀]−1 [∅(𝐴)]
[Link]’s College of Engineering 34 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

10 −10 10 −1 𝐴𝑑𝑗(𝐶𝑀)
−1
𝐶𝑀 =[1 8 −26] =
|𝐶𝑀|
0 21 −75
8 −26 1 −26 1 8
|𝐶𝑀| = |𝐵 𝐴2 𝐵| == 10 |21 −75| − (−10) |0 −75| + 10 |0 21| = −1080
𝐴𝐵
𝑇
8 −26 1 −26 1 8
+| | −| | +| |
21 −75 0 −75 0 21 −54 75 21 𝑇
−10 10 10 10 10 −10
𝐴𝑑𝑗(𝐶𝑀) = − | | +| | −| | = [−540 −750 −210]
21 −75 0 −75 0 21
−10 10 10 10 10 −10 180 270 90
[+ | 8 −26
| −|
1 −26
| +|
0 21
|]
−54 −540 180
= [ 75 −750 270]
21 −210 90
−54 −540 180 0.05 0.5 −0.1667
𝐴𝑑𝑗(𝐶𝑀) 1
-𝐶𝑀−1 = |𝐶𝑀|
= − 1080 [ 75 −750 270] = [−0.0694 0.6944 −0.25 ]
21 −210 90 −0.0194 0.1944 −0.0833
0.05 0.5 −0.1667 −12 0 0
𝐾 = [0 0 1] × [−0.0694 0.6944 −0.25 ] × [ 8 −20 0 ]
−0.0194 0.1944 −0.0833 20 19 −27
−12 0 0
𝐾 = [−0.0194 0.1944 −0.0833] × [ 8 −20 0 ] = [−0.0444 −5.5556 3.3333]
20 19 −27

𝐾 = [𝑘1 𝑘2
𝑘3 ] = [−0.0444 −5.5556 3.3333]
𝑥1 𝑥1
𝑥 𝑥
𝑢 = −𝐾𝑥 = −[𝐾1 𝐾2 𝐾3 ] [ 2 ] = −[−0.0444 −5.5556 3.3333] [ 2 ]
𝑥3 𝑥3
𝑢 = 0.0444𝑥1 + 5.5556𝑥2 − 3.3333𝑥3

10. Consider a system with the following state equations,

𝟎 𝟏 𝟎 𝟏
𝑿̇ = [ ] 𝑿 + [ ] 𝒖𝟏 + [ ] 𝒖 𝟐
−𝟔 −𝟓 𝟏 𝟏
Design a state feedback controller to place the closed loop poles at −𝟏𝟎 and − 𝟏𝟏 for the three
following cases. (i) Using input one only (ii) Using input two only Case (i): Using input one
only.

Solution

0 1 0
Given, 𝐴 = [ ] ; 𝐵 = [ ] ; 𝜇1 = −1.8 + 𝑗2.4 ; 𝜇2 = −1.8 − 𝑗2.4
−6 −5 1
Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 | ≠ 0

0
𝐵=[ ]
1
0 1 0 1
𝐴×𝐵 = [ ] ×[ ]=[ ]
−6 −5 1 −5
0 1
|𝐶𝑀| = |𝐵 𝐴𝐵 | = [ ] = −1 ≠ 0
1 −5
[Link]’s College of Engineering 35 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .
The desired characteristic equation is
(𝑠 + 10) (𝑠 + 11) = 0
= 𝑠 2 + 21𝑠 + 110 = 0
= 𝑠 2 + 𝛼1 𝑠 + 𝛼2 = 0

Hence, 𝛼1 = 21; 𝛼2 = 110

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

0 1
𝐴=[ ]
−6 −5
0 1 0 1 −6 −5
𝐴2 = 𝐴 × 𝐴 = [ ]×[ ]=[ ]
−6 −5 −6 −5 30 19

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

0 1 2 0 1 1 0
=[ ] + 21 [ ] + 110 [ ]
−6 −5 −6 −5 0 1
−6 −5 0 1 1 0 104 16
=[ ] + 21 [ ] + 110 [ ]=[ ]
30 19 −6 −5 0 1 −96 24
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝑲 = [𝟎 𝟏] [𝑪𝑴]−𝟏 [∅(𝑨)]

0 1 −1 𝐴𝑑𝑗(𝐶𝑀)
𝐶𝑀−1 = [ ] =
1 −5 |𝐶𝑀|

0 1
|𝐶𝑀| = |𝐵 𝐴𝐵 | = [ ] = −1
1 −5
0 1 −5 −1
𝐴𝑑𝑗(𝐶𝑀) = 𝑎𝑑𝑗 [ ]=[ ]
1 −5 −1 0
𝐴𝑑𝑗(𝐶𝑀) −5 −1 5 1
𝐶𝑀−1 = = −1 [ ]=[ ]
|𝐶𝑀| −1 0 1 0

5 1 104 16
𝐾 = [0 1] × [ ]×[ ]
1 0 −96 24
104 16
𝐾 = [1 0] × [ ] = [104 16]
−96 24
𝐾 = [𝑘1 𝑘2 ] = [104 16]

[Link]’s College of Engineering 36 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
Case (ii): Using input Two only

Solution

0 1 1
Given, 𝐴 = [ ] ; 𝐵 = [ ] ; 𝜇1 = −1.8 + 𝑗2.4 ; 𝜇2 = −1.8 − 𝑗2.4
−6 −5 1
Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 | ≠ 0

1
𝐵=[ ]
1
0 1 1 1
𝐴×𝐵 = [ ] ×[ ]=[ ]
−6 −5 1 −11

|𝐶𝑀| = |𝐵 1 1
𝐴𝐵 | = [ ] = −12 ≠ 0
1 −11
Hence the system is completely state controllable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .
The desired characteristic equation is
(𝑠 + 10) (𝑠 + 11) = 0
= 𝑠 2 + 21𝑠 + 110 = 0
= 𝑠 2 + 𝛼1 𝑠 + 𝛼2 = 0

Hence, 𝛼1 = 21; 𝛼2 = 110

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

0 1
𝐴=[ ]
−6 −5
0 1 0 1 −6 −5
𝐴2 = 𝐴 × 𝐴 = [ ]×[ ]=[ ]
−6 −5 −6 −5 30 19

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

0 1 2 0 1 1 0
=[ ] + 21 [ ] + 110 [ ]
−6 −5 −6 −5 0 1
−6 −5 0 1 1 0 104 16
=[ ] + 21 [ ] + 110 [ ]=[ ]
30 19 −6 −5 0 1 −96 24
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝑲 = [𝟎 𝟎 𝟎 , … … . 𝟏] [𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . . 𝑨𝒏−𝟏 𝑩]−𝟏 [∅(𝑨)]

𝐾 = [0 1] [𝐶𝑀]−1 [∅(𝐴)]

[Link]’s College of Engineering 37 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

1 1 −1 𝐴𝑑𝑗(𝐶𝑀)
𝐶𝑀−1 = [ ] =
1 −11 |𝐶𝑀|

|𝐶𝑀| = |𝐵 1 1
𝐴𝐵 | = [ ] = −12
1 −11
1 1 −11 −1
𝐴𝑑𝑗(𝐶𝑀) = 𝑎𝑑𝑗 [ ]=[ ]
1 −11 −1 1
𝐴𝑑𝑗(𝐶𝑀) 1 −11 −1 0.9167 0.0833
𝐶𝑀−1 = =− [ ]=[ ]
|𝐶𝑀| 12 −1 1 0.0833 −0.0833

0.9167 0.0833 104 16

𝐾 = [0 1] × [ ]×[ ]
0.0833 −0.0833 −96 24
104 16
𝐾 = [0.0833 −0.0833] × [ ] = [16.66 −0.6667]
−96 24
𝐾 = [𝑘1 𝑘2 ] = [16.66 −0.6667]

Case (iii): Using both inputs u1 and u2

Solution

0 1 0 1
𝑋̇ = [ ] 𝑋 + [ ] 𝑢1 + [ ] 𝑢2
−6 −5 1 1

𝑋̇ 0 1 𝑋1 0 1
[ 1] = [ ] [ ] + [ ] 𝑢1 + [ ] 𝑢2
𝑋̇2 −6 −5 𝑋2 1 1

Ẋ1 = 0 × X1 + 1 × X 2 + 0 × u1 + 1 × u2

Ẋ 2 = −6 × X1 − 5 × X 2 + 1 × u1 + 1 × u2

Combining the above equations

𝑋̇ 0 1 𝑋1 0 1 𝑢1
[ 1] = [ ][ ] + [ ][ ]
̇
𝑋2 −6 −5 𝑋2 1 1 𝑢2

𝟎 𝟏 𝟎 𝟏
Given, 𝐴 = [ ]; 𝑩 = [ ] ; 𝝁𝟏 = −𝟏𝟎 ; 𝝁𝟐 = −𝟏𝟏
−𝟔 −𝟓 𝟏 𝟏

Use Direct substitution method for multi input systems (Ackerman’s formula cannot be used

Step 1: Check the controllability of the given system. If it is controllable, then go to step 2

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵| ≠ 0

0 1
𝐵=[ ]
1 1
0 1 0 1 1 1
𝐴×𝐵 =[ ] ×[ ]=[ ]
−6 −5 1 1 −5 −11
0 1 1 1
|𝐶𝑀| = |𝐵 𝐴𝐵| = | |≠0
1 1 −5 −11

Hence the system is completely state controllable.

[Link]’s College of Engineering 38 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏.
The desired characteristic equation is
(𝑠 + 10) (𝑠 + 11) = 0
= 𝑠 2 + 21𝑠 + 110 = 0
= 𝑠 2 + 𝛼1 𝑠 + 𝛼2 = 0

Hence, 𝛼1 = 21; 𝛼2 = 110

Step 2: order 𝒏 = 𝟐 𝒂𝒏𝒅 𝒇𝒐𝒓 𝟐 𝒊𝒏𝒑𝒖𝒕 𝒔𝒚𝒔𝒕𝒆𝒎𝒔,

𝐾 𝐾2
𝑲=[ 1 ]
𝐾3 𝐾4

Step 3: Substitute this 𝑲 matrix into the desired characteristics polynomial |𝒔𝑰 − 𝑨 + 𝑩𝑲| and equate
it to (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 ).

|𝒔𝑰 − 𝑨 + 𝑩𝑲| = (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )

Since both sides of this characteristic equation are polynomials in ′𝒔′.

[𝑠𝐼 − 𝐴 + 𝐵𝐾] = [ 𝑠 0 0 1 0 1 𝐾1 𝐾2
]−[ ]+[ ][ ]
0 𝑠 −6 −5 1 1 𝐾3 𝐾4

𝑠 −1 𝐾3 𝐾4 𝑠 + 𝐾3 −1 + 𝐾4
=[ ]+[ ]=[ ]
6 𝑠+5 𝐾1 + 𝐾3 𝐾2 + 𝐾4 6 + 𝐾1 + 𝐾3 𝑠 + 5 + 𝐾2 + 𝐾4

𝑠 + 𝐾3 −1 + 𝐾4
| | = (𝑠 + 𝐾3 ) (𝑠 + 5 + 𝐾2 + 𝐾4 ) − (6 + 𝐾1 + 𝐾3 )(−1 + 𝐾4 )
6 + 𝐾1 + 𝐾3 𝑠 + 5 + 𝐾2 + 𝐾4

= 𝑠 2 + 𝑠(5 + 𝐾2 + 𝐾4 + 𝐾3 ) + 5𝐾3 + 𝐾2 𝐾3 + 𝐾3 𝐾4 − (−6 + 6𝐾4 − 𝐾1 + 𝐾1 𝐾4 − 𝐾3 + 𝐾3 𝐾4 )

= 𝑠 2 + 𝑠(5 + 𝐾2 + 𝐾3 + 𝐾4 ) + 5𝐾3 + 𝐾2 𝐾3 + 𝐾3 𝐾4 + 6 − 6𝐾4 + 𝐾1 − 𝐾1 𝐾4 + 𝐾3 − 𝐾3 𝐾4

= 𝑠 2 + 𝑠(5 + 𝐾2 + 𝐾3 + 𝐾4 ) + 6𝐾3 + 𝐾2 𝐾3 + 6 − 6𝐾4 + 𝐾1 − 𝐾1 𝐾4

(𝑠 − 𝜇1 )(𝑠 − 𝜇2 ) = (𝑠 + 10) (𝑠 + 11) = 𝑠 2 + 21𝑠 + 110 = 0

Step 4: By equating the coefficients of like power of ′𝒔′ on both sides, it is possible to determine the
values of 𝑲𝟏 , 𝑲𝟐 , 𝑲𝟑 , 𝑲𝟒 , 𝑲𝟓 , 𝑲𝟔

|𝒔𝑰 − 𝑨 + 𝑩𝑲| = (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )

𝑠 2 + 𝑠(5 + 𝐾2 + 𝐾3 + 𝐾4 ) + 6𝐾3 + 𝐾2 𝐾3 + 6 − 6𝐾4 + 𝐾1 − 𝐾1 𝐾4 = 𝑠2 + 21𝑠 + 110

On comparing like power of ‘s’

5 + 𝐾2 + 𝐾3 + 𝐾4 = 21 (1)

6𝐾3 + 𝐾2 𝐾3 + 6 − 6𝐾4 + 𝐾1 − 𝐾1 𝐾4 = 110 (2)

Solving the above equations,

K K2 −16 5
K=[ 1 ]=[ ]
K3 K4 10 1

[Link]’s College of Engineering 39 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
𝟏 𝟎 𝟎 𝟏 𝟎
11. Given the system 𝒙̇ = 𝑨𝒙 + 𝑩𝒖 where 𝑨 = [𝟎 𝟐 𝟎] ; 𝑩 = [𝟎 𝟏]. Design a linear state
𝟎 𝟎 𝟑 𝟏 𝟏
variable feedback such that the closed-loop poles of the system are located at −𝟏, −𝟐 and
−𝟑 (May 2016)

Solution
1 0 0 1 0
Given 𝐴 = [0 2 0 ] ; 𝐵 = [ 0 1] ; 𝜇1 = −1; 𝜇2 = −2; 𝜇3 = −3
0 0 3 1 1

Method II : Using Direct Substitution method

Step 1: Check the controllability of the given system. If it is controllable, then go to step2

Controllability matrix |𝑪𝑴| = |𝑩 𝑨𝑩 𝑨𝟐 𝑩 , … … . 𝑨𝒏−𝟏 𝑩| ≠ 𝟎

Controllability matrix 𝐶𝑀 = |𝐵 𝐴𝐵 𝐴2 𝐵| ≠ 0

1 0
𝐵 = [ 0 1] ;
1 1
1 0 0 1 0 1 0
𝐴 × 𝐵 = [0 2 0 ] × [ 0 1] = [ 0 2]
0 0 3 1 1 3 3
1 0 0 1 0 1 0
𝐴2 𝐵 = 𝐴 × 𝐴𝐵 = [0 2 0] × [0 2] = [0 4]
0 0 3 3 3 9 9
1 0 1 0 1 0
|𝐶𝑀| = |𝐵 𝐴𝐵 2
𝐴 𝐵 | = |0 1 0 2 0 4| ≠ 0
1 1 3 3 9 9
Hence the system is completely state controllable.

Step 2: 𝒏(𝑵𝒐. 𝒐𝒇 𝒔𝒕𝒂𝒕𝒆 𝒗𝒂𝒓𝒊𝒂𝒃𝒍𝒆𝒔) = 𝟑, & 𝒖(𝒏𝒐. 𝒐𝒇 𝑰𝒏𝒑𝒖𝒕𝒔) = 𝟐, Hence

𝐾1 𝐾2 𝐾3
𝑲=[ ]
𝐾4 𝐾5 𝐾6

Step 3: Substitute this 𝑲 matrix into the desired characteristics polynomial |𝒔𝑰 − 𝑨 + 𝑩𝑲| and equate
it to (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 ).

|𝒔𝑰 − 𝑨 + 𝑩𝑲| = (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 )

Since both sides of this characteristic equation are polynomials in ′𝒔′.

𝑠 0 0 1 0 0 1 0 𝐾 𝐾2 𝐾3
[𝑠𝐼 − 𝐴 + 𝐵𝐾] = [0 𝑠 0] − [0 2 0] + [0 1] [𝐾1 𝐾5 𝐾6
]
4
0 0 𝑠 0 0 3 1 1

𝑠−1 0 0 𝑘1 𝑘2 𝑘3
=[ 0 𝑠−2 0 ] + [ 𝑘4 𝑘5 𝑘6 ]
0 0 𝑠−3 𝑘1 + 𝑘4 𝑘2 + 𝑘5 𝑘3 + 𝑘6

[Link]’s College of Engineering 40 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
(𝑠 − 1) + 𝑘1 𝑘2 𝑘3
=| 𝑘4 (𝑠 − 2) + 𝑘5 𝑘6 |
𝑘1 + 𝑘4 𝑘2 + 𝑘5 (𝑠 − 3) + 𝑘3 + 𝑘6

Step 4: By equating the coefficients of like power of ′𝒔′ on both sides, it is possible to determine the
values of 𝑲𝟏 , 𝑲𝟐 , 𝑲𝟑

|𝒔𝑰 − 𝑨 + 𝑩𝑲| = (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 )

(𝑠 − 1) + 𝑘1 𝑘2 𝑘3
| 𝑘4 (𝑠 − 2) + 𝑘5 𝑘6 | = (𝑠 + 1)(𝑠 + 2)(𝑠 + 3)
𝑘1 + 𝑘4 𝑘2 + 𝑘5 (𝑠 − 3) + 𝑘3 + 𝑘6

Solving the above equations we get

𝐾1 𝐾2 𝐾3 −3.8237 −9.3055 11.7215

𝐾=[ ]=[ ]
𝐾4 𝐾5 𝐾6 −0.3389 3.564 0.5382

𝐱̇ 𝟏 𝟎 𝟐𝟎. 𝟔 𝐱 𝟏 𝟎 𝐱𝟏
12. Consider the system [ ]=[ ] [𝐱 ] + [ ] 𝐮(𝐭); 𝐲 = [𝟎 𝟏] [𝐱 ]. Use observer
𝐱̇ 𝟐 𝟏 𝟎 𝟐 𝟏 𝟐
based approach to design state feedback control such that 𝐮 = −𝐊𝐱̌. Design a full order
state observer assume that the desired eigenvalues of the observer matrix are 𝛃𝟏 =
−𝟏𝟎 𝐚𝐧𝐝 𝛃𝟐 = −𝟏𝟎

Method I : Using Transformation Matrix T

Step 1: Check the observability of the given system. If it is observable, then go to step2

𝐶
𝐶𝐴
| 𝐶𝐴2 |
Observability matrix |𝑶𝑴| = ≠𝟎
| ; |
𝐶𝐴𝑛−2
𝐶𝐴𝑛−1
𝐶
Observability matrix |𝑂𝑀| = | | ≠ 0
𝐶𝐴
𝐶 = [0 1];
0 20.6
𝐶 × 𝐴 = [0 1] × [ ] = [1 0]
1 0

|𝑂𝑀| = | 𝐶 | = [0 1] = −1 ≠ 0
𝐶𝐴 1 0
Hence the system is completely state observable.

Step 2: Find the characteristic polynomial

|𝑺𝑰 − 𝑨| = 𝑺𝒏 + 𝒂𝟏 𝑺𝒏−𝟏 + 𝒂𝟐 𝑺𝒏−𝟐 +, … … … … + 𝒂𝒏−𝟏 𝑺 + 𝒂𝒏

and hence determine the values of 𝒂𝟏 , 𝒂𝟐 , 𝒂𝟑 , … … . 𝒂𝒏

|𝑆𝐼 − 𝐴| = |[ 𝑠 0] − [0 20.6]| = | 𝑠 −20.6

| = 𝑠 2 − 20.6
0 𝑠 1 0 −1 𝑠

[Link]’s College of Engineering 41 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
On comparing with 𝑠 2 + 𝑎1 𝑠 + 𝑎2 we get,

𝑎1 = 0; 𝑎2 = −20.6

Step 3: Determine the transformation matrix ′𝑻′

The given system is already in observable canonical form. Hence, the transformation
matrix 𝑄 = (𝑊𝑁 ∗ )−1 is 𝐼

Step 4: Using desired closed loop poles, write the desired polynomial. i.e.

(𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 ), … … … … … … (𝒔 − 𝝁𝒏 )

= 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 + 𝜶𝟐 𝒔𝒏−𝟐 +, … … … … + 𝜶𝒏−𝟏 𝒔 + 𝜶𝒏

and determine the values of 𝜶𝟏 , 𝜶𝟐 , 𝜶𝟑 , … … … … . . 𝜶𝒏

The desired characteristic equation

(𝑠 + 10)(𝑠 + 10) = 𝑠 2 + 20𝑠 + 100

On comparing with 𝑠 2 + 𝛼1 𝑠 + 𝛼2 , we get,

𝛼1 = 20; 𝛼2 = 100

Step 5: The required state feedback gain matrix ′𝑲′ is given by

𝜶𝒏 − 𝒂𝒏 𝜶𝒏 − 𝒂𝒏
𝜶𝒏−𝟏 − 𝒂𝒏−𝟏 𝜶𝒏−𝟏 − 𝒂𝒏−𝟏
𝑲𝒆 = 𝑸 ; = (𝑾𝑵∗ )−𝟏 ;
; ;
[ 𝜶𝟏 − 𝒂𝟏 ] [ 𝜶𝟏 − 𝒂𝟏 ]
𝛼2 − 𝑎2
𝐾𝑒 = (𝑊𝑁 ∗ )−1 [𝛼 − 𝑎 ]
1 1

1 0
Since the system is in observable canonical form (𝑊𝑁 ∗ )−1 = 𝐼 = [ ]
0 1
1 0 100 − (−20) 120.6
𝐾𝑒 = [ ][ ]=[ ]
0 1 20 − 0 20
𝑘𝑒1 120.6
Hence, 𝐾𝑒 = [ ]=[ ]
𝑘𝑒2 20

Method 2: Direct comparison method

Step 1: Check the observability of the given system. If it is observable, then go to step2

𝐶
𝐶𝐴
| 𝐶𝐴2 |
Observability matrix |𝑶𝑴| = ≠𝟎
| ; |
𝐶𝐴𝑛−2
𝐶𝐴𝑛−1
𝐶
Observability matrix |𝑂𝑀| = | | ≠ 0
𝐶𝐴
[Link]’s College of Engineering 42 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

𝐶 = [0 1];
0 20.6
𝐶 × 𝐴 = [0 1] × [ ] = [1 0]
1 0

|𝑂𝑀| = | 𝐶 | = [0 1] = −1 ≠ 0
𝐶𝐴 1 0
Hence the system is completely state observable.

𝒌𝒆𝟏
Step 2: 𝒏 = 𝟐, then 𝑲𝒆 = [ ]
𝒌𝒆𝟐

Step 3: Substitute this 𝑲𝒆 matrix into the desired characteristics polynomial |𝒔𝑰 − (𝑨 − 𝑲𝒆 𝑪)|
and equate it to (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 ).

|𝑺𝑰 − 𝑨 + 𝑩𝑲| = (𝒔 − 𝝁𝟏 )(𝒔 − 𝝁𝟐 )(𝒔 − 𝝁𝟑 )

The characteristic equation

𝑘 𝑠 −20.6 + 𝑘𝑒1
|𝑠𝐼 − (𝐴 − 𝐾𝑒 𝐶)| = |[ 𝑠 0] − [0 20.6] + [ 𝑒1 ] [0 1]| = [ ]
0 𝑠 1 0 𝑘𝑒2 −1 𝑠 + 𝑘𝑒2

= 𝑠 2 + 𝑘𝑒2 𝑠 − 20.6 + 𝑘𝑒1

The desired characteristic equation

(𝑠 − 𝜇1 ) (𝑠 − 𝜇2 ) (𝑠 − 𝜇3 ) = (𝑠 + 10)(𝑠 + 10) = 𝑠 2 + 20𝑠 + 100

Step 4: By equating the coefficients of like power of ′𝑺′ on both sides, it is possible to determine
the values of 𝑲𝒆𝟏 , 𝑲𝒆𝟐

𝑘𝑒2 = 20;

−20.6 + 𝑘𝑒1 = 100

𝑘𝑒1 = 120.6

𝑘𝑒1 120.6
∴ 𝑘𝑒 = [ ]=[ ]
𝑘𝑒2 20

Method III : Using Ackermann’s formula

Step 1: Check the observability of the given system. If it is observable, then go to step2

𝐶
𝐶𝐴
| 𝐶𝐴2 |
Observability matrix |𝑶𝑴| = ≠𝟎
| ; |
𝐶𝐴𝑛−2
𝐶𝐴𝑛−1
𝐶
Observability matrix |𝑂𝑀| = | | ≠ 0
𝐶𝐴
𝐶 = [0 1];
[Link]’s College of Engineering 43 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

0 20.6
𝐶 × 𝐴 = [0 1] × [ ] = [1 0]
1 0

|𝑂𝑀| = | 𝐶 | = [0 1] = −1 ≠ 0
𝐶𝐴 1 0
Hence the system is completely state observable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 +
𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .
(𝑠 − 𝜇1 ) (𝑠 − 𝜇2 ) (𝑠 − 𝜇3 ) = (𝑠 + 10)(𝑠 + 10) = 𝑠 2 + 20𝑠 + 100

Hence, 𝛼1 = 20; 𝛼2 = 100

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

0 20.6
𝐴=[ ]
1 0
0 20.6 0 20.6 20.6 0
𝐴2 = 𝐴 × 𝐴 = [ ]×[ ]=[ ]
1 0 1 0 0 20.6
20.6 0 0 20.6 0 424.36
𝐴3 = 𝐴2 × 𝐴 = [ ]×[ ]=[ ]
0 20.6 1 0 20.6 0

∅(𝑨) = 𝑨𝟐 + 𝜶𝟏 𝑨 + 𝜶𝟐 𝑰

0 20.6 2 20.6 0 1 0
=[ ] + 20 [ ] + 100 [ ]
1 0 0 20.6 0 1
0 424.36 20.6 0 1 0
=[ ] + 20 [ ] + 100 [ ]
20.6 0 0 20.6 0 1
120.6 412
=[ ]
20 120.6
Step 4: Calculate the state feedback gain matrix, ′𝑲𝒆 ′ using the Ackermann’s formula
𝐶 −𝟏 𝟎
𝐶𝐴 𝟎
𝑲𝒆 = [∅(𝑨)] 𝐶𝐴2 ;
; ;
[𝐶𝐴𝑛 ] [𝟏]
𝟎
𝑲𝒆 = [∅(𝑨)] [𝑶𝑴]−𝟏 [ ]
𝟏
𝑘𝑒1 120.6 412 0 1 0 120.6
𝑲𝒆 = [ ]=[ ]×[ ]×[ ] =[ ]
𝑘𝑒2 20 120.6 1 0 1 20

13. Consider a linear system described by the equation

[Link]’s College of Engineering 44 VI Semester ICE

 IC6601 Advanced Control System Department of ICE 2018-19

𝟏 𝟐 𝟎 𝟐
𝒙̇ = [𝟑 −𝟏 𝟏] 𝒙 + [𝟐] 𝒖; 𝒚 = [𝟎 𝟎 𝟏] 𝒙
𝟎 𝟐 𝟎 𝟏
Design a state observer so that eigen values of state matrix (𝑨 − 𝑮𝑪) are −𝟒, −𝟑 ± 𝑱𝟏

Solution

1 2 0 2
Given 𝐴 = [3 −1 1] ; 𝐵 = [2] ; 𝐶 = [0 0 1]; 𝜇1 = −4, 𝜇2 = −3 − 𝑗1, 𝜇3 = −3 −
0 2 0 1
𝑗1; 𝑛 = 3

Step 1: Check the Observability of the given system. If it is observable, then go to step 2

𝐶
Observability matrix |𝑂𝑀| = | 𝐶𝐴 | ≠ 0
𝐶𝐴2
𝐶 = [0 0 1]
1 2 0
𝐶𝐴 = 𝐶 × 𝐴 = [0 0 1] × [3 −1 1] = [0 2 0]
0 2 0
1 2 0
𝐶𝐴2 = 𝐶𝐴 × 𝐴 = [0 2 0] × [3 −1 1] = [6 −2 2]
0 2 0
𝐶 0 0 1
|𝑂𝑀| = | 𝐶𝐴 | = |0 2 0| = 0 | 2 0| − 0 |0 0 0 2
| + 1| | = −12 ≠ 0
−2 2 6 2 6 −2
𝐶𝐴2 6 −2 2
Hence the system is completely state Observable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .

The desired characteristic equation is

(𝑠 + 4) (𝑠 + 3 + 𝑗1) (𝑠 + 3 − 𝑗1) = (𝑠 + 4)(𝑠 2 + 6𝑠 + 10) = 0
= 𝑠 3 + 10𝑠 2 + 34𝑠 + 40 = 0
= 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 = 0

Hence, 𝛼1 = 10; 𝛼2 = 34; 𝛼3 = 40

Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.

∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

1 2 0
𝐴 = [3 −1 1]
0 2 0

[Link]’s College of Engineering 45 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

1 2 0 1 2 0 7 0 2
𝐴2 = 𝐴 × 𝐴 = [3 −1 1] × [3 −1 1] = [0 9 −1]
0 2 0 0 2 0 6 −2 2
7 0 2 1 2 0 7 18 0
𝐴3 = 𝐴2 × 𝐴 = [0 9 −1] × [3 −1 1] = [27 −11 9]
6 −2 2 0 2 0 0 18 −2

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

1 2 0 3 1 2 02 1 2 0 1 0 0
=[3 −1 1] + 10 [3 −1 1] + 34 [3 −1 1] + 40 [0 1 0]
0 2 0 0 2 0 0 2 0 0 0 1
7 18 0 7 0 2 1 2 0 1 0 0
= [27 −11 9 ] + 10 [0 9 −1] + 34 [3 −1 1] + 40 [0 1 0]
0 18 −2 6 −2 2 0 2 0 0 0 1
151 86 20
= [129 85 33]
60 66 58

Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝐶 −𝟏 𝟎
𝐶𝐴 𝟎
𝑲= [∅(𝑨)] 𝐶𝐴2 ;
; ;
[𝐶𝐴𝑛 ] [𝟏]
𝟎
[∅(𝑨)] [𝑶𝑴] −𝟏
𝑲= [𝟎 ]
𝟏
0 0 1 −1 𝐴𝑑𝑗(𝑂𝑀)
[𝑂𝑀]−1 = [0 2 0] =
|𝑂𝑀|
6 −2 2
𝐶 0 0 1
|𝑂𝑀| = | 𝐶𝐴 | = |0 2 0| = 0 | 2 0| − 0 |0 0| + 1 |0 2 | = −12
−2 2 6 2 6 −2
𝐶𝐴2 6 −2 2

2 0 0 0 0 2 𝑇
+| | −| | +| |
−2 2 6 2 6 −2 4 0 −12 𝑇 4 2 −2
0 1 0 1 0 0
𝐴𝑑𝑗(𝑂𝑀) = − | | +| | −| | = [ 2 −6 0 ] =[ 0 −6 0 ]
−2 2 6 2 6 −2
0 1 0 1 0 0 −2 0 0 −12 0 0
[ + |2 0
| −|
0 0
| +|
0 2
|]

𝐴𝑑𝑗(𝑂𝑀) 1 4 2 −2 −0.3333 0.1667 0.1667

𝑂𝑀−1 = =− [ 0 −6 0 ] = [ 0 0.5 0 ]
|𝑂𝑀| 12
−12 0 0 1 0 0
151 86 20 −0.3333 0.1667 0.1667 0
𝐾 = [129 85 33] × [ 0 0.5 0 ] × [0]
60 66 58 1 0 0 1
[Link]’s College of Engineering 46 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

−18.8333 −3.5 8.5 0 8.5

𝐾=[ 64.5 42.5 16.5] × [0] = [16.5]
151 86 20 1 20
𝑘1 8.5
𝐾 = [𝑘2 ] = [16.5]
𝑘3 20

14. A single input system is described by the following state equation

𝟎 𝟏 𝟎 𝟎
̇𝑿 = [ 𝟎 𝟎 𝟏 ] 𝑿 + [𝟎] 𝑼; 𝒀 = [𝟏 𝟎 𝟎] 𝑿
−𝟔 −𝟏𝟏 −𝟔 𝟏
Design a full order observer with eigenvalues −𝟓, −𝟐 ± 𝟑. 𝟒𝟔𝟒𝒋

Solution

𝟎 𝟏 𝟎 𝟎
Given 𝐴 = [ 𝟎 𝟎 𝟏 ] ; 𝐵 = [ 𝟎 ] ; 𝐶 = [𝟏 𝟎 𝟎]; 𝜇1 = −5, 𝜇2 = −2 − 3.464𝑗, 𝜇3 =
−𝟔 −𝟏𝟏 −𝟔 𝟏
−2 − 3.464𝑗; 𝑛 = 3

Step 1: Check the Observability of the given system. If it is observable, then go to step 2

𝐶
Observability matrix |𝑂𝑀| = | 𝐶𝐴 | ≠ 0
𝐶𝐴2
𝐶 = [𝟏 𝟎 𝟎]
𝟎 𝟏 𝟎
𝐶𝐴 = 𝐶 × 𝐴 = [𝟏 𝟎 𝟎] × [ 𝟎 𝟎 𝟏 ] = [0 1 0]
−𝟔 −𝟏𝟏 −𝟔
𝟎 𝟏 𝟎
2
𝐶𝐴 = 𝐶𝐴 × 𝐴 = [0 1 0] × [ 𝟎 𝟎 𝟏 ] = [0 0 1]
−𝟔 −𝟏𝟏 −𝟔
𝐶 1 0 0
1 0
|𝑂𝑀| = | 𝐶𝐴 | = |0 1 0| = 1 | |=1≠0
0 1
𝐶𝐴2 0 0 1
Hence the system is completely state Observable.

Step 2: Determine the desired characteristic equation from the specified closed loop poles,
𝝁𝟏 , 𝝁𝟐 , 𝝁 𝟑 , . . , 𝝁𝒏
i.e., (𝒔 − 𝝁𝟏 ) (𝒔 − 𝝁𝟐 ) (𝒔 − 𝝁𝟑 ). . . . . . . . (𝒔 − 𝝁𝒏 ) = 𝒔𝒏 + 𝜶𝟏 𝒔𝒏−𝟏 +, … … . . , +𝜶𝒏−𝟏 𝒔𝟏 + 𝜶𝒏
and determine the values of 𝜶𝟏 , 𝜶𝟐 , … … . , 𝜶𝒏 .

The desired characteristic equation is

(𝑠 + 5) (𝑠 + 2 + 𝑗3.464) (𝑠 + 2 − 𝑗3.464) = (𝑠 + 5)(𝑠 2 + 4𝑠 + 16) = 0
= 𝑠 3 + 9𝑠 2 + 36𝑠 + 80 = 0
= 𝑠 3 + 𝛼1 𝑠 2 + 𝛼2 𝑠 + 𝛼3 = 0

Hence, 𝛼1 = 9; 𝛼2 = 36; 𝛼3 = 80
[Link]’s College of Engineering 47 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
Step 3: Determine the matrix ∅(𝑨) using the coefficient of desired characteristics polynomial.
∅(𝑨) = 𝑨𝒏 + 𝜶𝟏 𝑨𝒏−𝟏 + 𝜶𝟐 𝑨𝒏−𝟐 +, … … . . , +𝜶𝒏−𝟏 𝑨𝟏 + 𝜶𝒏 𝑰

𝟎 𝟏 𝟎
𝐴=[ 𝟎 𝟎 𝟏]
−𝟔 −𝟏𝟏 −𝟔
𝟎 𝟏 𝟎 𝟎 𝟏 𝟎 0 0 1
2
𝐴 =𝐴×𝐴=[ 𝟎 𝟎 𝟏 ]×[ 𝟎 𝟎 𝟏 ] = [−6 −11 −6]
−𝟔 −𝟏𝟏 −𝟔 −𝟔 −𝟏𝟏 −𝟔 36 60 25
0 0 1 𝟎 𝟏 𝟎 −6 −11 −6
𝐴3 = 𝐴2 × 𝐴 = [−6 −11 −6] × [ 𝟎 𝟎 𝟏 ] = [ 36 60 25 ]
36 60 25 −𝟔 −𝟏𝟏 −𝟔 −150 −239 −90

∅(𝑨) = 𝑨𝟑 + 𝜶𝟏 𝑨𝟐 + 𝜶𝟐 𝑨 + 𝜶𝟑 𝑰

𝟎 𝟏 𝟎 3 𝟎 𝟏 𝟎 2 𝟎 𝟏 𝟎 1 0 0
=[ 𝟎 𝟎 𝟏 ] +9[ 𝟎 𝟎 𝟏 ] + 36 [ 𝟎 𝟎 𝟏 ] + 80 [0 1 0]
−𝟔 −𝟏𝟏 −𝟔 −𝟔 −𝟏𝟏 −𝟔 −𝟔 −𝟏𝟏 −𝟔 0 0 1
−6 −11 −6 0 0 1 𝟎 𝟏 𝟎 1 0 0
= [ 36 60 25 ] + 9 [−6 −11 −6] + 36 [ 𝟎 𝟎 𝟏 ] + 80 [ 0 1 0]
−150 −239 −90 36 60 25 −𝟔 −𝟏𝟏 −𝟔 0 0 1
74 25 3
= [−18 41 7]
−42 −95 −1
Step 4: Calculate the state feedback gain matrix, ′𝑲′ using the Ackermann’s formula
𝐶 −𝟏 𝟎
𝐶𝐴 𝟎
𝑲 = [∅(𝑨)] 𝐶𝐴2 ;
; ;
[𝐶𝐴 ]𝑛 [ 𝟏]
𝟎
𝑲 = [∅(𝑨)] [𝑶𝑴]−𝟏 [𝟎]
𝟏
1 0 0 −1 𝐴𝑑𝑗(𝑂𝑀)
[𝑂𝑀]−1 = [0 1 0] =
|𝑂𝑀|
0 0 1
𝐶 1 0 0
1 0
|𝑂𝑀| = | 𝐶𝐴 | = |0 1 0| = 1 | |=1
2 0 1
𝐶𝐴 0 0 1
1 0 0
𝐴𝑑𝑗(𝑂𝑀) = [0 1 0]
0 0 1

𝐴𝑑𝑗(𝑂𝑀) 1 1 0 0 1 0 0
𝑂𝑀−1 = = [0 1 0 ] = [0 1 0 ]
|𝑂𝑀| 1
0 0 1 0 0 1
74 25 3 1 0 0 0
𝐾𝑒 = [−18 41 7 ] × [0 1 0 ] × [0 ]
−42 −95 −1 0 0 1 1

[Link]’s College of Engineering 48 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

74 25 3 0 3
𝐾𝑒 = [−18 41 7 ] × [0] = [ 7 ]
−42 −95 −1 1 −1
𝑘𝑒1 3
𝐾𝑒 = [𝑘𝑒2 ] = [ 7 ]
𝑘𝑒3 −1

UNIT II PHASE PLANE ANALYSIS

Nonlinear Systems
Nonlinear systems are systems which does not obey the principle of superposition. Principle of
superposition implies that if a system has responses 𝑦1 (𝑡) and 𝑦2 (𝑡) to any two inputs 𝑥1 (𝑡) and
𝑥2 (𝑡) respectively then the system response to the linear combination of these inputs 𝛼1 (𝑡)𝑥1 (𝑡) +
𝛼2 (𝑡)𝑥2 (𝑡) is given by the linear combination of the individual outputs. i.e., 𝛼1 (𝑡)𝑦1 (𝑡) +
𝛼2 (𝑡)𝑦2 (𝑡) where 𝛼1 and 𝛼2 constants.

To satisfy the principle of superposition, 𝑦3 = 𝑦1 + 𝑦2 = 𝛼1 (𝑡)𝑦1 (𝑡) + 𝛼2 (𝑡)𝑦2 (𝑡)

Features of Nonlinear systems

1. The response of nonlinear system to a particular test signal is no guide to their behaviour to
their inputs, since the principle of superposition does not hold good for nonlinear systems.

2. Nonlinear system response may be highly sensitive to input amplitude. The stability study
of nonlinear systems requires the information about the type and amplitude of the
anticipated inputs, initial conditions, etc., in addition to the usual requirement of the
mathematical model.

[Link]’s College of Engineering 49 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
3. Nonlinear systems may exhibit limit cycles which are self-sustained oscillations of fixed
frequency and amplitude.

4. Nonlinear systems may have jump resonance in the frequency response.

5. The output of a nonlinear system will have harmonics and sub-harmonics when excited by
sinusoidal signals.

6. Nonlinear systems will exhibit phenomena like frequency entrainment and asynchronous
quenching.

Frequency amplitude dependence

The frequency-amplitude dependence is one of the most fundamental characteristics of the

oscillations of nonlinear systems. The frequency-amplitude dependence can be best studied by
considering the mechanical system shown in figure in which the spring is nonlinear.

The differential equation governing the dynamics of the system may be written as

𝑀𝑥̈ + 𝐵𝑥̇ + 𝐾𝑥 + 𝐾 ′ 𝑥 3 = 0

Where 𝐾𝑥 + 𝐾 ′ 𝑥 3 = opposing force due to nonlinear spring

The parameters M, B and K are positive constants. The parameter K' may be positive or
negative. If K' is positive, the spring is called hard spring and if K' is negative the spring is called
soft spring. The equation is nonlinear differential equation and it is called Duffing’s equation.
When the system has non zero initial conditions, the free response is damped oscillatory. The
frequency of free oscillations depends on the amplitude of oscillations. When K' < 0 (soft spring)
the frequency decreases with decreasing amplitude. When K' > 0 (hard spring) the frequency
increases with decreasing amplitude. When K' = 0 (corresponding to linear system) the frequency
remains unchanged as the amplitude of free oscillation decreases. The frequency-amplitude
dependence characteristic of nonlinear mechanical system is shown in figure.

Jump resonance

In the frequency response of nonlinear systems, the amplitude of the response (output) may
jump from one point to another for increasing or decreasing values of frequency, . This
phenomenon is called jump resonance and it can be observed in the frequency response of the
system shown in figure when it is subjected to sinusoidal input. Let the mechanical system shown
in figure 1 be subjected to an input of type 𝐴 𝐶𝑜𝑠 𝜔𝑡. Now the differential equation governing
the mechanical system is

𝑀𝑥̈ + 𝐵𝑥̇ + 𝐾𝑥 + 𝐾 ′ 𝑥 3 = 𝐴 𝐶𝑜𝑠 𝜔𝑡

[Link]’s College of Engineering 50 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

Let X be the amplitude of the response or output of the system. In frequency response studies,
the amplitude, A of the input is held constant, while its  is varied and the amplitude X of the
output is observed. The frequency response curve is plotted between X and . The frequency
response curves of the mechanical system of figure 1are shown in figure 2.
In the frequency response curve shown in figure, as the frequency  is increased, the
amplitude X increases, until point-2 is reached. A further increase in frequency will cause a jump
from point-2 to point-3. This phenomenon is called jump resonance. As the frequency is
increased further, the amplitude X follows the curve from point-3 towards point-4. When the
frequency is reduced starting from a high value corresponding to point-4, the amplitude X slowly
increases through point-3, until point-5 is reached. A further decrease in  will cause another
jump from point-5 to point-6. This phenomenon is called jump resonance. After this jump, the
amplitude X decreases with  and follows the curve from point-6 towards point-1. For jump
resonance to take place, it is necessary that the damping term be small and the amplitude of the
forcing function be large enough to drive the system into a region of appreciably nonlinear
operation.

Subharmonic oscillations

When a nonlinear system is excited by a sinusoidal signal, the response or output will have
steadystate oscillation whose frequency is an integral submultiple of the forcing frequency.
These oscillations are called subharmonic oscillations. The generation of subharmonic
oscillations depends on the system parameters and initial conditions. It also depends on
amplitude and frequency of the forcing functions.

Limit Cycles

The response of nonlinear system may exhibit oscillations with fixed amplitude and
frequency. These oscillations are called limit cycles. Consider a mechanical system with a
nonlinear damping and described by the equation.

𝑀𝑥̈ + 𝐵(1 − 𝑥 2 )𝑥̇ + 𝐾𝑥 = 0

where M, B and K are positive constants. The equation is called the van der pol equation. For small
values of x the damping will be negative which implies the stored energy in the damper is fed to
the system. For large values of x the damping is positive which implies that it absorbs energy from
the system. Thus, it can be expected that such, a system may exhibit a sustained oscillation. Since
the system explained above is not a forced system, this oscillation is called a self-excited oscillation
or zero input limit cycle.

[Link]’s College of Engineering 51 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
Frequency entrainment

The phenomenon of frequency entrainment is observed in the frequency response of

nonlinear systems that exhibit limit cycles. Consider a system capable of exhibiting a limit cycle
of frequency, . If a periodic input of frequency  is applied to this system then the phenomenon
of beats is observed. (The beat is the oscillation whose frequency is the difference between l
and . This frequency is also called beat frequency). In linear systems, the beat frequency
decreases indefinitely as  approaches l. But in nonlinear systems, the frequency , of the limit
cycle falls in synchronistical with or is entrained by the forcing frequency, or within a certain
band of frequencies. This phenomenon is called frequency entrainment. The band of frequency
in which entrainment occurs is called the zone of frequency entrainment. In this zone, the
frequencies  and l coalesce and only one frequency, or exists. The relationship between
|𝜔 − 𝜔𝑙 | and  is shown in figure.

Asynchronous quenching

In a nonlinear system that exhibits a limit cycle of frequency l it is possible to quench (stop
or eliminate) the limit cycle oscillation by forcing the system of a frequency q where q and ,
are not related to each other. The phenomenon is called signal stabilization or asynchronous
quenching

Common Physical Non-linearities

The nonlinearities can be classified as incidental and intentional. The incidental

nonlinearities are those which are inherently present in the system. Common examples of
incidental nonlinearities are saturation, dead-zone, coulomb friction, stiction, backlash, etc.

The intentional nonlinearities are those which are deliberately inserted in the system to
modify system characteristics. The most common example of this type of nonlinearity is a relay.

Saturation: In this type of nonlinearity the output is proportional to input for a limited range
of input signals. When the input exceeds this range, the output tends to become nearly constant
as shown in figure.

[Link]’s College of Engineering 52 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

All devices when driven by sufficiently large signals, exhibit the phenomenon of saturation due
to limitations of their physical capabilities. Saturation in the output of electronic, rotating and
flow (hydraulic and pneumatic) amplifiers, speed and torque saturation in electric and hydraulic
motors, saturation in the output of sensors for measuring position, velocity, temperature, etc., are
the well known examples.

Dead Zone : The dead zone is the region in which the output is zero for a given input. Many
physical devices do not respond to small signals, i.e., if the input amplitude is less than some small
value, there will be no output. The region in which the output is zero is called dead zone. When the
input is increased beyond this dead zone value, the output will be linear. Figure shows the Dead zone
nonlinearity and combination of dead zone a nd saturation nonlinearity.

Friction: Friction exists in any system when there is relative motion between contacting surfaces.
The different types of friction are viscous friction, coulomb friction and stiction. The viscous
friction is linear in nature and the frictional force is directly proportional to relative velocity of the
sliding surfaces. The coulomb friction and stiction are nonlinear frictions. The coulomb friction
offers a constant retarding force only when the motion is initiated. Due to interlocking of surface
irregularities, more force is required to move an object from rest than to maintain it in motion.
Hence the force of stiction is always greater than that of coulomb friction.

In actual practice, the stiction force gradually decreases with velocity and changes over to coulomb
friction at reasonably low velocities as shown in figure. The composite characteristics of various
frictions are shown in figure.

METHOD OF LINEARISATION

Local linearization has been extensively used while analyzing nonlinear phenomena in many
engineering situations. Sometimes, this may also be called perturbation analysis. Consider a time
invariant nonlinear system
𝑥̇ = 𝑓(𝑥, 𝑢) (1)

[Link]’s College of Engineering 53 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
Where the function 𝑓(𝑥, 𝑢) is assumed to be continuously differentiable with respect to each
component of 𝑥 and 𝑢. When the control function 𝑢 is either zero or a constant, the equilibrium
states are those where 0 = 𝑓(𝑥𝑒 , 𝑢0 ) is satisfied.
Let
𝑥(𝑡) = 𝑥𝑒 (𝑡) + 𝛿𝑥(𝑡); u(𝑡) = 𝑢0 + 𝛿𝑢(𝑡) so that
𝑑
𝑥̇ (𝑡) = 𝑥̇ 𝑒 (𝑡) + 𝛿𝑥̇ (𝑡) where 𝛿𝑥̇ (𝑡) = 𝑑𝑡 𝛿𝑥(𝑡) (2)
Here 𝛿𝑥(𝑡) and 𝛿𝑢(𝑡) are perturbations on 𝑥𝑐 and 𝑢0 , and 𝛿𝑥̇ (𝑡) is the time derivative of 𝛿𝑥(𝑡).
Substituting equation (2) into the original system equations;
𝑥̇ (𝑡) = 𝑥̇ 𝑒 (𝑡) + 𝛿𝑥̇ (𝑡) = 𝑓(𝑥𝑒 (𝑡) + 𝛿𝑥(𝑡), 𝑢0 + 𝛿𝑢(𝑡))
𝜕𝑓 𝜕𝑓
𝑓(𝑥𝑒 (𝑡), 𝑢0 ) + 𝜕𝑥 (𝑥𝑒 , 𝑢0 )𝛿𝑥 + 𝜕𝑢 (𝑥𝑒 , 𝑢0 )𝛿𝑢 + ℎ(𝑥𝑒 , 𝑢0 , 𝛿𝑥, 𝛿𝑢)
(3)
𝜕𝑓 𝜕𝑓
Where and are the following Jacobian matrices;
𝜕𝑥 𝜕𝑢

𝜕𝑓1 𝜕𝑓1 𝜕𝑓1 𝜕𝑓1 𝜕𝑓1 𝜕𝑓1

,… ,…
𝜕𝑥1 𝜕𝑥2 𝜕𝑥𝑛 𝜕𝑢1 𝜕𝑢2 𝜕𝑢𝑛
𝜕𝑓2 𝜕𝑓2 𝜕𝑓2 𝜕𝑓2 𝜕𝑓2 𝜕𝑓2
𝜕𝑓 ,… 𝜕𝑓 ,…
= 𝜕𝑥1 𝜕𝑥2 𝜕𝑥𝑛 and 𝜕𝑢 = 𝜕𝑢1 𝜕𝑢2 𝜕𝑢𝑛 (4)
𝜕𝑥
; ; ; ; ; ; ; ;
𝜕𝑓𝑛 𝜕𝑓𝑛 𝜕𝑓𝑛 𝜕𝑓𝑛 𝜕𝑓𝑛 𝜕𝑓𝑛
[𝜕𝑥1 𝜕𝑥2
; 𝜕𝑥𝑛 ] [𝜕𝑢1 𝜕𝑢2
; ] 𝑒𝑣𝑎𝑙𝑢𝑎𝑡𝑒𝑑 𝑎𝑡 𝑥𝑒,𝑢0
𝑒𝑣𝑎𝑙𝑢𝑎𝑡𝑒𝑑 𝑎𝑡 𝑥𝑒 ,𝑢0 𝜕𝑢

And ℎ(𝑥𝑒 , 𝑢0 , 𝛿𝑥, 𝛿𝑢) contains the remainder higher order terms. Assume that as 𝛿𝑥 and 𝛿𝑢 go to
zero, the term ℎ(𝑥𝑒 , 𝑢0 , 𝛿𝑥, 𝛿𝑢) tends to zero at a faster rate that the first order terms on the right
hand side of equation (4). Specifically, we assume that;

Then, in a sufficiently small region about 𝑥𝑒 and for sufficiently small values of |𝛿𝑢|, the behaviour
of the perturbed system can be approximated by the locally linearized equation;

𝜕𝑓 𝜕𝑓
𝛿𝑥̇ (𝑡) = 𝜕𝑥 (𝑥𝑒 , 𝑢0 )𝛿𝑥 + 𝜕𝑢 (𝑥𝑒 , 𝑢0 )𝛿𝑢 (6)

This being a set of linear and time invariant system of equations makes it much simpler to study
from an analytical point of view. All the methods available for linear systems like analysis based
on eigen value or roots of the characteristic equation are applicable to determine the stability of this
linearized system.

Concept of phase portraits, Singular points and Limit cycles

The phase plane method of analysis is a graphical method for the analysis of linear and
nonlinear systems. The analysis is carried by constructing phase trajectories. It gives an idea about
the transient behaviour and stability of the system. The phase plane analysis is usually restricted
to second order systems excited by step or ramp inputs. This analysis technique can be extended
to a higher order system if it is approximated as a second order system. The dynamics of control
systems can be represented by differential equations. A second order linear system can be
represented by the differential equation
𝑑2 𝑥 𝑑𝑥
+ 2𝜔𝑛 𝑑𝑡 + 𝜔𝑛2 𝑥 = 0 (1)
𝑑𝑡 2
[Link]’s College of Engineering 54 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
where,
𝑥 = One of the system variable (e.g, displacement in mechanical system, current in
electrical system, etc.,)
 = Damping ratio
n = Natural frequency of oscillation.
The state of the second order system represented by equ(1) can be described by choosing
two state variables. In state space modelling using phase variables we choose one of the system
variable and its derivatives as state variables. Let 𝑥1 and 𝑥2 be the state variables of the second
order system.
𝑑𝑥
Here 𝑥1 = 𝑥 and 𝑥2 = (2)
𝑑𝑡
On substituting the state variables in equation (1)
𝑥̇ 2 + 2𝜔𝑛 𝑥2 + 𝜔𝑛2 𝑥1 = 0 (3)
The state equations of the system are obtained from equation (2) and (3). The state equations
are,
𝑥̇ 1 = 𝑥2 (4)
𝑥̇ 2 = −𝜔𝑛 𝑥1 − 2𝜔𝑛 𝑥2
2
(5)
For linear systems the state equations are a set of first order linear differential equations
and solutions of state equations can be easily obtained by integration. But for nonlinear systems,
the state equations are a set of first-order nonlinear differential equations and solving the nonlinear
differential equations will not be an easy task. Hence, for nonlinear systems the phase plane
method of analysis will be a useful tool.

Phase plane: The coordinate plane with state variables 𝑥1 and 𝑥2 as two axis is called the phase
plane.

Phase Trajectory: The curve describing the state point (𝑥1 , 𝑥2 ) in the phase plane with time as
running parameter is called phase trajectory.

Phase Portrait: A trajectory can be constructed in the phase plane for each set of initial conditions.
Hence a family of trajectories can be constructed for a system in a phase plane and such a family
of trajectories is called the ‘Phase Portrait’.

Singular point: A point in phase plane at which the derivatives of all state variables are zero is
called a singular point. It is also called equilibrium point. At the singular point, the slope of the
trajectory is indeterminant. If the system is placed at a singular or equilibrium point, it will
continue to lie there if left undisturbed (i.e., the derivatives of all the phase variables are zero and
so the system state remains unchanged).
The singular points are classified as Nodal point, Saddle point, Focus point and Centre
or Vortex point depending on the eigen values of the system matrix.
Table 1. Phase portrait of systems with various types of singular points.

[Link] Eigen values of system Type of singular Phase portrait of the system
matrix point with singular point at origin

[Link]’s College of Engineering 55 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Distinct, real and

the two eigen values are
1 Stable node
negative

Distinct, real and

the two eigen values are
2 Unstable node
positive

Distinct, real, one eigen

value is positive and
3 Saddle point
the other is negative

Complex-conjugate
4 with negative real part Stable focus

Complex conjugate
5 with positive real part Unstable focus

Purely imaginary and

Centre or Vortex
6 conjugate
point

Stability analysis of nonlinear systems using Phase trajectories

For linear time invariant systems, the concept of stability can be defined as follows,
1. When the input is zero, the system is stable for arbitrary initial conditions if the resulting
trajectory tends towards the equilibrium state.
2. When the system is excited by a bounded input, the system is stable if the system output is
bounded.

In nonlinear systems the concept of stability is not clear-cut. There are many types of
stability definitions in the literature. The linear autonomous system has only one equilibrium state.
[Link]’s College of Engineering 56 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
The behaviour of linear system about the equilibrium state completely determines the qualitative
behaviour in the entire state-plane.

In nonlinear systems there may be multiple equilibrium state. The behaviour of nonlinear
system about the equilibrium point may be different for small deviations and large deviations about
the equilibrium point. In nonlinear systems with multiple equilibrium states, the system trajectories
may move away from one equilibrium point and tend to other as time progresses. Hence in
nonlinear systems, stability is discussed relative to the equilibrium state and the general stability of
system cannot be defined.

Consider an autonomous system described by the state equation, x = f(x). Let us assume
that the system has one equilibrium point and the origin of phase plane is the equilibrium point. For
this system, the following definitions of stability are proposed.

1. The autonomous system defined by equation 𝑥 = 𝑓(𝑥) is stable at the origin, if for every
initial state 𝑥(𝑡0 ) which is sufficiently close to origin, 𝑥(t) remains near the origin for all t.
2. The autonomous system defined by equation 𝑥 = 𝑓(𝑥) is asymptotically stable if 𝑥(t)
approaches the origin as 𝑡 → .
3. The autonomous system defined by equation 𝑥 = 𝑓(𝑥) is asymptotically stable in the large
if it is asymptotically stable for every initial state regardless of how near or for it is from the
origin.

Limit cycles in Phase Portrait

The limit cycles are oscillations of fixed amplitude and period. The existence of limit cycle
in nonlinear systems can be predicted from closed trajectories in the phase portrait. In linear
systems, when oscillations occur, the resulting trajectories will be closed curves as shown in figure
2. The amplitude of the oscillations is not fixed. It changes with the size of the initial conditions.
Slight changes in system parameters will destroy the oscillations

The limit cycles are oscillations of the response of nonlinear systems with fixed amplitude and
frequency. If these oscillations or limit cycles exists when there is no input then they are called
zero input limit cycles.

In nonlinear systems, there can be limit cycles (oscillations) that are independent of the size
of initial conditions as shown in figure 3. These limit cycles are usually less sensitive to system
parameter variations. Limit cycles of fixed amplitude and period can be sustained over a finite range
of system parameters. The limit cycle is stable if the paths in its neighbourhood converge towards
the limit cycle as shown in figure 3. The limit cycle is unstable if the paths in the neighbourhood
of a limit cycle diverge away from it as shown in figure 4.

[Link]’s College of Engineering 57 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Figure 1. Phase portrait showing limit Figure 2. Unstable limit Figure 3. Stable limit cycle
cycle behavior in linear system cycle

Construction of Phase trajectories

The state equations of a second order autonomous system are,

𝑥̇ 1 = 𝑓1 (𝑥1 , 𝑥2 ) (1)
𝑥̇ 2 = 𝑓2 (𝑥1 , 𝑥2 ) (2)
Where 𝑥1 and 𝑥2 are the state variables of the system. On dividing (2) by (1),
𝑥̇ 2 𝑓 (𝑥 ,𝑥 )
= 𝑓2 (𝑥1 ,𝑥2 ) (3)
𝑥̇ 1 1 1 2
𝑑𝑥2 𝑑𝑥1
Since, 𝑥̇ 2 = and 𝑥̇ 1 = , equation (3) can be written as
𝑑𝑡 𝑑𝑡
𝑑𝑥2 𝑓2 (𝑥1 ,𝑥2 )
= 𝑓 (𝑥 (4)
𝑑𝑥1 1 1 ,𝑥2 )

The equation (4) defines the slope of phase trajectory at every point in the phase- plane,
except at singular points. At singular points, the slope of the phase trajectory is indeterminate. The
phase trajectory can be constructed using the slope equation (4) either analytically or graphically.
In analytical method of construction, the equation (4) is integrated and the resulting equation is
used to construct phase trajectories for a given set of initial conditions. In many cases, it is not
possible to perform integration of the slope equation (4). Hence a number-of graphical methods
have been developed for construction of phase trajectories and the two popular methods are isocline
method and delta method.

Construction of Phase trajectories by analytical method

The analytical method is used if the differential equations describing the system can be
approximated by piecewise linear differential equations. (i.e., the equations are linearized for small
regions). The slope equation 𝑑𝑥2 ⁄𝑑𝑥1 is formed from the state equations. Then the slope equation
is splitted into sections of linear equations. Each section of linear slope equation is directly
integrated to get the solution of state equations. Each section of solution equation is used to
construct a section of phase trajectories for various sets of initial conditions. The solution of slope
equation will be a function of 𝑥1 and 𝑥2 . For a given set of initial conditions, assume different
values of 𝑥1 , and calculate 𝑥2 for each value of 𝑥1 , using the solution of slope equation. The
values of 𝑥1 and 𝑥2 are tabulated. In an ordinary graph sheet, take 𝑥1 and 𝑥2 is, choose appropriate
scales and mark the phase points (𝑥1 and 𝑥2 ). Join all the points by a smooth curve. For each set
of initial conditions one phase trajectory can be constructed using the procedure described above:
Eventhough, time domain solutions are obtained by direct integration, the construction of
phase-trajectories (or phase plane analysis) will help in investigating system behaviour and the
design of system parameters to achieve a desired response. Also the existence of limit cycles is
sharply brought into focus by the phase-portrait.
[Link]’s College of Engineering 58 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
Example 1
Consider a system with an ideal relay as shown in figure. Determine the singular point.
Construct phase trajectories, corresponding to initial conditions, (𝒊) 𝒄(𝟎) = 𝟐, 𝒄̇ (𝟎) =
𝟏 𝒂𝒏𝒅 (𝒊𝒊) 𝒄(𝟎) = 𝟐, 𝒄̇ (𝟎) = 𝟏. 𝟓. Take 𝒓 = 𝟐 volts amd M = 𝟏. 𝟐 volts.

SOLUTION

[Link]’s College of Engineering 59 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 60 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 61 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Construction of phase trajectory by Isocline method

Let, S = slope at any point in the phase plane

𝑑𝑥 𝑓 (𝑥 ,𝑥 )
Slope 𝑆 = 𝑑𝑥2 = 𝑓2 (𝑥1 ,𝑥2) (1)
1 1 1 2

Let, S1= slope at a point on phase trajectory-1

𝑓2 (𝑥1 , 𝑥2 ) = 𝑆1 × 𝑓1 (𝑥1 , 𝑥2 ) (2)

Equation (2) defines the locus of all such points in phase plane at which the slope of the
phase trajectory is S1. A locus passing through the points of same slope in phase plane is called
isocline. The slope of a phase trajectory at the crossing point of an isocline will be the slope of
corresponding isocline. A typical plot of isoclines for various values of slope, S is shown in figure.
Using these isoclines the phase trajectories can be constructed as explained below.

[Link]’s College of Engineering 62 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

The phase trajectory starts at a point corresponding to initial conditions. (For each set of
initial conditions one phase trajectory can be constructed).

Let, S1, S2, S3 etc, be the slopes associated with isoclines 1,2,3, etc.,

Let, 𝛼1 = 𝑡𝑎𝑛−1 (𝑆1 ); 𝛼2 = 𝑡𝑎𝑛−1 (𝑆2 ); 𝛼3 = 𝑡𝑎𝑛−1 (𝑆3 ) etc.,

Note: If a straight line is drawn at an angle  from a point, then the slope of the line at that point is
tan 𝛼,

In figure, let point A on isocline-1be the point corresponding to a set of initial conditions.
The phase-trajectory will leave the point A at a slope S1. When the trajectory reaches the isocline-
2, the slope changes to S2. Draw two lines from point A one at a slope of S 1 (i.e., at angle of 𝛼1 =
𝑡𝑎𝑛−1 (𝑆1 ) and the other at a slope of S2, (i.e., at angle of 𝛼1 = 𝑡𝑎𝑛−1 (𝑆1 )). Let these two lines
meet the isocline- 2 at p and q. Now we can say that the trajectory would cross the iscoline-2 at a
point midway between p and q. Mark the point B on the isocline- 2 approximately midway between
p and q. The constructional procedure is now repeated at B to find the crossing point C on the
isocline-3. By similar procedures the crossing points on the isolines are determined.

A smooth curve drawn through the crossing points gives the phase-trajectory starting at
point A. The accuracy of the trajectory is closely related to the spacing of the isoclines. The phase
trajectory will be more accurate if large numbers of isoclines are used, which are very close to each
other. It should be noted that using a set of isoclines, any number of trajectories can be constructed.

Example : A linear second order servo is described by the equation

𝒆̈ + 𝟐𝝎𝒏 𝒆̇ + 𝝎𝟐𝒏 𝒆 = 𝟎

𝒓𝒂𝒅
Where  = 𝟎. 𝟏𝟓, 𝝎𝒏 = 𝟏 𝒔𝒆𝒄 , 𝒆(𝟎) = 𝟏. 𝟓 and 𝒆̇ (𝟎) = 𝟎. Determine the singular point.
Construct the phase trajectory using the method of isoclines.
Solution
Let 𝑥1 and 𝑥2 be the state variables of the system and they are related to the system variables, 𝑒
𝑥1 = 𝑒 (1)
𝑥2 = 𝑒̇ (2)
On differentiating (1)
𝑥̇ 1 = 𝑒̇ = 𝑥2 (3)
On differentiating (2)
𝑥̇ 2 = 𝑒̈ (4)
Given that
𝑒̈ + 2 𝜔𝑛 𝑒̇ + 𝜔2𝑛 𝑒 = 0 (5)
[Link]’s College of Engineering 63 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
Substitute equation (1), (2) and (4) in (5)
𝑥̇ 2 + 2 𝜔𝑛 𝑥2 + 𝜔2𝑛 𝑥1 = 0
𝑥̇ 2 = −2 𝜔𝑛 𝑥2 − 𝜔2𝑛 𝑥1 (6)
The state equation of the system are given by equation (3) and (6)
𝑥̇ 1 = 𝑥2

𝑥̇ 2 = −2 𝜔𝑛 𝑥2 − 𝜔2𝑛 𝑥1

Singular point is obtained from state equation by putting 𝑥̇ 1 = 0 and 𝑥̇ 2 = 0

Let the coordinates of singular point in phase plane = (𝑥10 , 𝑥20 )

On substituting 𝑥̇ 1 = 0 and 𝑥2 = 𝑥02 in equation (3) we get, 𝑥02 = 0.

On substituting 𝑥̇ 2 = 0 , 𝑥1 = 𝑥01 and 𝑥2 = 𝑥02 in equation (6) we get,

0 = −2 𝜔𝑛 𝑥2 − 𝜔𝑛2 𝑥1

But 𝑥02 = 0, ∴ 0 = −𝜔2𝑛 𝑥1 or 𝑥01 = 0.

Therefore the coordinates of singular point are (0,0) and so the origin is the singular point.

The slope of the phase trajectory is given by

𝑑𝑥 ⁄𝑑𝑡 𝑥̇
𝑆 = 𝑑𝑥2⁄𝑑𝑡 = 𝑥̇ 2 (7)
1 1

On substituting for 𝑥1̇ and 𝑥̇ 2 from equations (3) and (6) in equation (7), we get

From equation (9) we can conclude that the isoclines are straight lines. For each value of S
we can draw one isocline. Using equation (9), the coordinates (𝑥1 , 𝑥2 ) in the phase plane for various
slopes can be calculated. Since there are three variables. Let us assume two variables and calculate
the third variable.

Let us choose S as −2, −1.0, −0.5, 0, 0.5, 1.0, 2.0

For each value of S, choose two values of 𝑥1 and calculate 𝑥2 using equation (9_. The value
of 𝑆, 𝑥1 and 𝑥2 are tabulated. The slope angle, 𝛼 is calculated for each value of S, using the
expression, 𝛼 = tan−1(𝑆), and tabulated in table.
S -2.0 -1.0 -0.5 0 0.5 1.0 2.0
𝛼 -63 -45 -27 0 27 45 63
𝑥1 𝑥2 𝑥1 𝑥2 𝑥1 𝑥2 𝑥1 𝑥2 𝑥1 𝑥2 𝑥1 𝑥2 𝑥1 𝑥2
1.0 0.6 1.0 1.4 0.25 1.25 0.25 -0.8 1.0 -1.25 1.0 -0.77 1.0 -0.43
2.0 1.2 1.5 2.1 0.5 2.5 0.75 -2.5 2.0 -2.5 2.0 -1.54 2.0 -0.86
Result
1. The singular point lies at origin
2. From figure it is observed that the phase trajectory spiral towards the origin, hence the type
of singular point is stable focus.
[Link]’s College of Engineering 64 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19

Unit III
DESCRIBING FUNCTION FOR COMMON NON-LINEARITIES

CHARACTERISTICS OF NON-LINEAR SYSTEMS

1. Non-linear equations, unlike linear ones, cannot, in general be solved analytically.
2. Powerful mathematical tools like Laplace and Fourier transforms do not apply to non-linear
systems.
3. Non-linear systems do not obey principle of superposition. Thus, they respond quite differently
to external inputs and initial conditions. In other words, the response of a non-linear system to
a particular test signal is no guide to their behavior to other inputs.
4. The stability of non-linear system may depend on initial conditions.
5. Non-linear systems frequently have more than one equilibrium points (where as linear system
has unique equilibrium point).
6. Non-linear system may exhibit limit cycles which are self-sustained oscillations of fixed
amplitude and fixed period (without external excitation). These oscillations are called limit
cycles.
7. Non-linear systems with a periodic input may exhibit a periodic output whose frequency is
either a sub harmonic or a harmonic of the input frequency. For example, an input of frequency
of 10Hz may result in an output of 5Hz for sub-harmonic case or 30Hz for a harmonic case.
8. A non-linear system can display jump resonance, a form of hysteresis, in its frequency
response.
[Link]’s College of Engineering 65 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
TYPES OF NON-LNEARITIES

SATURATION

 It is the most commonly encountered nonlinearity in control system. It is often associated with
amplifiers and actuators.
 In transistor amplifiers, the output varies linearly with the input,
only for a small amplitude limits.
 When the input amplitude gets out of the linear range of the
amplifier, the output changes very little and stays close to its
maximum value.
 Most of the actuators display saturation characteristics.
 For eg., the output torque of a servo motor cannot increase
infinitely and tends to saturate due to the properties of magnetic
material. Similarly, valve controlled hydraulic actuators are saturated by the maximum flow
rate.

DEAD ZONE
 The dead-zone is the region in which the output is zero for a
given input. When the input is increased beyond this dead-
zone value, the output will be linear. The region where the
applied voltage has no effect is called the deadband e.g.
sensors, friction in all components
 Dead zone nonlinearity may occur in sensors, amplifiers and
actuators.
 In a de motor, we assume that any voltage applied to armature winding will cause armature to
rotate if the field current is maintained constant.
 In reality, due to static friction at the motor shaft, rotation will occur only if the torque provided
by the motor is sufficiently large. This corresponds to dead zone.
 This phenomenon occurs also in valve controlled pneumatic and hydraulic actuators.

BACKLASH

 Backlash nonlinearity commonly occurs in mechanical

components of control systems. In gear trains, small gap exist
between a pair of mating gears.
 As a result, when the driving gear a smaller angle than the gap
H, the driven gear does not move at all, which corresponds to
the dead zone (OA).
 After the contact is established between the two gears, the
[Link]’s College of Engineering 66 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
driven gear follows the rotation of the driving gear in a linear fashion (AB segment).
 When the driving gear rotates in the reverse direction, by a distance of 2H, the driven gear again
does not move, corresponding to the segment BC.
 After the contact between the two gears is reestablished, the driven gear linearly follows the
rotation of driving gear in the reverse direction (CD segment).
 Thus, if the driving gear is in periodic motion, the driven gear will move in the fashion
represented by closed path EBCD

FRICTION
 Friction exists in a system, when there is relative motion between contacting surfaces.
 The different types of friction are viscous friction, coulomb friction and stiction. The viscous
fiction is linear in nature and the frictional force is directly proportional to relative velocity of
sliding surfaces.
 The coulomb friction is a drag force which opposes motion (For eg: drag due to rubbing contact
between brushes and the commutator).
 Due to interlocking of surface irregularities, more force is required to move an object from rest
than to maintain it in motion. Hence, force of stiction is always greater than that of Coulomb
friction.

[Link]’s College of Engineering 67 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

DESCRIBING FUNCTION FUNDAMENTALS

Describing Function (DF) method provides a linear approximation to the nonlinear element based
on the assumption that the input to nonlinear element is a sinusoidal input of known constant amplitude.
The fundamental harmonic element's output is compared with input sinusoidal to determine steady state
amplitude and phase relation. This relation is the describing function for the nonlinear element. Thus, the
method can also be viewed as 'harmonic linearization' of a non- linear element. DF method is based on
'Fourier series. DF method predicts whether limit cycle oscillation exist or not and gives numerical
estimates of amplitude and frequency of oscillation, when limit cycles are predicted.
To discuss the basic concept of DF analysis, consider the block diagram of a nonlinear system
shown in figure below, where blocks 𝐺1 (𝑠)) and 𝐺2 (𝑠) represent linear elements while the block N
represents nonlinear element.

Let 𝑥 = 𝑋 sin 𝜔𝑡 be a sinusoidal input given to the nonlinear system. The output ‘y' of the nonlinear
element will be a non-sinusoidal periodic function, which may be expressed in terms of Fourier series.

𝑦 = 𝐴0 + 𝐴1 sin 𝜔𝑡 + 𝐵1 cos 𝜔𝑡 + 𝐴2 sin 𝜔𝑡 + 𝐵2 cos 𝜔𝑡 + 𝐴3 sin 𝜔𝑡 + 𝐵3 cos 𝜔𝑡+, , …

If the nonlinearity is symmetrical the average value of y is zero and hence the output y is given by

𝑦 = 𝐴1 sin 𝜔𝑡 + 𝐵1 cos 𝜔𝑡 + 𝐴2 sin 𝜔𝑡 + 𝐵2 cos 𝜔𝑡 + 𝐴3 sin 𝜔𝑡 + 𝐵3 cos 𝜔𝑡+, , …

In the absence of external input (i.e., when 𝑟 = 0) the output y of the nonlinearity N is feedback to
input through 𝐺1 (𝑠) and 𝐺2 (𝑠) in tandem. If 𝐺1 (𝑠) 𝐺2 (𝑠) has low pass characteristics, then all the
harmonics of y are filtered, so that the input 𝑥 to the nonlinear element N is mainly contributed by
fundamental component of 𝑦 and hence 𝑥 remains sinusoidal. Under such conditions the harmonics of the
output are neglected and the fundamental component of 𝑦 alone considered for the purpose of analysis.
𝑦 = 𝑦1 = 𝐴1 sin 𝜔𝑡 + 𝐵1 cos 𝜔𝑡 = 𝑌1  ∅1 = 𝑌1 sin(𝜔𝑡 + ∅)
𝐵
Where, Y1 = √A21 + 𝐵12 and ∅1 = tan−1 𝐴1
1

𝑌1 = Amplitude of the fundamental harmonics component of the output.

∅1 =Phase shift of the fundamental harmonic component of the output with respect to the input.

The coefficients 𝐴1 and 𝐵1 of the Fourier series are given by

2 2𝜋
𝐴1 = 2𝜋 ∫0 𝑦 sin 𝜔𝑡 𝑑(𝜔𝑡)
2𝜋
2
𝐵1 = ∫ 𝑦 cos 𝜔𝑡 𝑑(𝜔𝑡)
2𝜋
0

When the input 𝑥 to the nonlinearity is sinusoidal, the describing function of the nonlinearity is
defined as,

𝑌1
𝐾𝑁 (𝑋, 𝜔) =  ∅1
𝑋

The nonlinear element N in the system can be replaced by the following describing function.
[Link]’s College of Engineering 68 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
If the nonlinearity is replaced by the describing function then all linear theory frequency domain
techniques can be used for the analysis of the system. The describing functions are used only for stability
analysis and it is not directly applied to the optimization of the system design. The describing function is a
frequency domain approach and no general correlation is possible between time and frequency response.

Saturation Nonlinearity
The input-output relation is linear for 𝑥 = 0 to 𝑆. When the input 𝑥 > 𝑆, the output reaches a saturated value of 𝐾𝑆.
The input is a sinusoidal,
𝑥 = 𝑋 sin 𝜔𝑡 (1)
Where 𝑋 is the maximum value of input.
When 𝜔𝑡 = 𝛽, 𝑥 = 𝑆
Hence equation (1) can be written as,
𝑆 = 𝑋 sin 𝛽 (2)
𝑆 −1 𝑆
sin 𝛽 = 𝑜𝑟 𝛽 = sin ( ) (3)
𝑋 𝑋
The output 𝑦 of the nonlinearity can be divided into three regions in a period of 𝜋. The output equation for the three
regions are
𝐾𝑥 ; 0 ≤ 𝜔𝑡 ≤ 𝛽
𝑦 = { 𝐾𝑆 ; 𝛽 ≤ 𝜔𝑡 ≤ (𝜋 − 𝛽) (4)
𝐾𝑥 ; (𝜋 − 𝛽) ≤ 𝜔𝑡 ≤ 𝜋

Let 𝑌1 = Amplitude of the fundamental harmonic component of the output

∅1 = Phase shift of the fundamental harmonic component of the output with respect to the Input.
𝑌
The describing function is given by, 𝐾𝑁 (𝑋, 𝜔) = 𝑋1  ∅1
𝐵
Where 𝑌1 = √𝐴12 + 𝐵12 and ∅1 = tan−1 (𝐴1 )
1
The output y has half wave and quarter wave symmetries
𝜋
2
∴ 𝐵1 = 0 and 𝐴1 = 𝜋 ∫02 𝑦 sin 𝜔𝑡 𝑑(𝜔𝑡) (5)
⁄ 2
The output y is given by two different expressions in the period 0 to pi/2. Hence, equation (5) can be written as
4 𝛽 4 𝜋⁄
𝐴1 = 𝜋 ∫0 𝑦 sin 𝜔𝑡 𝑑(𝜔𝑡) + 𝜋 ∫𝛽 2 𝑦 sin 𝜔𝑡 𝑑(𝜔𝑡) (6)
On substituting the values of y from equ.(5) in equ.(6)
4 𝛽 4 𝜋⁄
𝐴1 = 𝜋 ∫0 𝐾𝑥 sin 𝜔𝑡 𝑑(𝜔𝑡) + 𝜋 ∫𝛽 2 𝐾𝑆 sin 𝜔𝑡 𝑑(𝜔𝑡)
On substituting 𝑥 = 𝑋 sin 𝜔𝑡,
4𝐾 𝛽 4𝐾𝑆 𝜋⁄2
𝐴1 = 𝜋 0
∫ 𝑋 𝑠𝑖𝑛 𝜔𝑡 × sin 𝜔𝑡 𝑑(𝜔𝑡) + ∫ sin 𝜔𝑡
𝜋 𝛽
𝑑(𝜔𝑡)
4𝐾𝑋 𝛽 4𝐾𝑆 𝜋⁄
= 𝜋 0
∫ 𝑠𝑖𝑛2 𝜔𝑡 𝑑(𝜔𝑡) + 𝜋 ∫𝛽 2 sin 𝜔𝑡 𝑑(𝜔𝑡)
4𝐾𝑋 𝛽 1−cos 2𝜔𝑡 4𝐾𝑆 𝜋⁄2
= 𝜋
∫0 2
𝑑(𝜔𝑡) + ∫ sin 𝜔𝑡 𝑑(𝜔𝑡)
𝜋 𝛽

[Link]’s College of Engineering 69 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
2𝐾𝑋 sin 2𝜔𝑡 𝛽 4𝐾𝑆 𝜋⁄
= [𝜔𝑡 − ] + [− cos 𝜔𝑡]𝛽 2
𝜋 2 0 𝜋
2𝐾𝑋 sin 2𝛽 4𝐾𝑆 𝜋
= 𝜋
[𝛽 − 2 ] + 𝜋 [− cos 2 + cos 𝛽]
2𝐾𝑋 sin 2𝛽 4𝐾𝑆
= 𝜋
[𝛽 − 2 ] + 𝜋 cos 𝛽 (7)
On substituting for 𝑆, (𝑖. 𝑒, 𝑆 = 𝑋 sin 𝛽 from eq(2) in eq(7), we get,
2𝐾𝑋 sin 2𝛽 4𝐾
𝐴1 = 𝜋 [𝛽 − 2 ] + 𝜋 𝑋 sin 𝛽 cos 𝛽
2𝐾𝑋 2 sin 𝛽 cos 𝛽 4𝐾𝑋
= [𝛽 − ] + 𝜋 sin 𝛽 cos 𝛽
𝜋 2
2𝐾𝑋
= 𝜋 [𝛽 − sin 𝛽 cos 𝛽] + 2 sin 𝛽 cos 𝛽
2𝐾𝑋
= [𝛽 + sin 𝛽 cos 𝛽] (8)
𝜋
2𝐾𝑋
𝑌1 = √𝐴21 + 𝐵21 = √𝐴21 + 0 = 𝐴1 = [𝛽 + sin 𝛽 cos 𝛽] (9)
𝜋
−1 𝐵1 −1
∅1 = tan = tan 0=0 (10)
𝐴1
𝑌
The describing function 𝐾𝑁 (𝑋, 𝜔) = 𝑋1 ∅1 (11)
Using equations (9) and (10), the describing function if eq(11) can be writtene as
𝑌 2𝐾𝑋
𝐾𝑁 (𝑋, 𝜔) = 𝑋1 ∅1 = 𝜋 [𝛽 + sin 𝛽 cos 𝛽] 00 (12)
Depending on the maximum value of input 𝑋, the describing function can be written as,
𝜋
If 𝑋 < 𝑆, then 𝛽 = 2 , 𝐾𝑁 (𝑋, 𝜔) = 𝐾 (13)
2𝐾
If 𝑋 > 𝑆, then 𝐾𝑁 (𝑋, 𝜔) = 𝜋 [𝛽 + sin 𝛽 cos 𝛽] (14)
Equation (14) can be expressed in another form as follows,
𝑆
From equation (2) 𝑆 = 𝑋 sin 𝛽 , ∴ sin 𝛽 = t (15)
𝑋
On constructing right angle triangle with unity hypotenuse as shown in figure, cos 𝛽 can be evaluated
𝑆 2 𝑎𝑑𝑗 𝑆 2
𝑎𝑑𝑗 = √1 − ( ) ∴ cos 𝛽 = = √1 − ( ) (16)
𝑋 ℎ𝑦𝑝 𝑋
In describing function of eq(14), substitute for 𝛽, sin 𝛽 and cos 𝛽 from equations (3), (15) and (16), we get

2𝐾 2
−1 𝑆 𝑆
√1 − ( 𝑆 )
∴ 𝐾𝑁 (𝑋, 𝜔) = 𝜋
[sin (𝑋
) + (𝑋
)] 𝑋
𝑓𝑜𝑟 𝑋 > 𝑆 (17)

Dead Zone nonlinearity

 A dead zone nonlinearity may occur in sensors, amplifiers and actuators.
 Ina dc motor, we assume that any voltage applied to armature winding will cause armature to
rotate if the field current is maintained constant.
 In reality, due to static friction at the motor shaft, rotation will occur only if the torque provided
by the motor is sufficiently large.
 This corresponds to dead zone.
 This phenomenon occurs also in valve controlled pneumatic and hydraulic actuators.

[Link]’s College of Engineering 70 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 71 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 72 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Relay with Dead zone and Hysteresis

[Link]’s College of Engineering 73 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Relay with Dead zone and Hysteresis

[Link]’s College of Engineering 74 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 75 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 76 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Backlash Nonlinearity

[Link]’s College of Engineering 77 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 78 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 79 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 80 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Dead Zone and Saturation

[Link]’s College of Engineering 81 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 82 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 83 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Ideal Relay(On-OFF) with Hysteresis

The input-output characteristics of an ideal relay with hysteresis for a sinusoidal input are shown in
figure. When the input exceed, a the input is positive and constant at ‘N’. when the input is gradually
reduced the output becomes’-N’ when the inout is ‘-a’ and remains at ‘-N’. the sinusoidal input to the
nonlinear element is
𝑚(𝑡) = 𝑀 sin 𝜔𝑡; 𝑀 > 𝑎 (1)
The output 𝑦 of the nonlinearity can be divided into three regions. The output equation for the three regions are

[Link]’s College of Engineering 84 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
−𝑁 ; 0 ≤ 𝜔𝑡 ≤ 𝛽
𝑦= { 𝑁 ; 𝛽 ≤ 𝜔𝑡 ≤ (𝜋 + 𝛽) (2)
−𝑁 ; (𝜋 + 𝛽) ≤ 𝜔𝑡 ≤ 2𝜋 + 𝛽

The output equation has no symmetry so that its Fourier series contains both sine and cosine terms. The average
value of the output is zero and hence 𝐴0 = 0
2 𝜋
𝐴1 = 𝜋 ∫0 𝑦 cos 𝜔𝑡 𝑑(𝜔𝑡)
2 𝜋+𝛽 −4𝑁
𝐴1 = 𝜋 ∫𝛽 𝑁 cos 𝜔𝑡 𝑑(𝜔𝑡) = 𝜋
sin 𝛽 (3)
2 𝜋
𝐵1 = 𝜋 ∫0 𝑦 sin 𝜔𝑡 𝑑(𝜔𝑡)
2 𝜋+𝛽
= 𝜋 ∫𝛽 𝑁 sin 𝜃 𝑑𝜃 (4)
Let 𝑌1 = Amplitude of the fundamental harmonic component of the output
∅1 = Phase shift of the fundamental harmonic component of the output with respect to the I input.
𝑌
The describing function is given by, 𝐾𝑁 (𝑋, 𝜔) = 𝑋1  ∅1
𝐵
Where 𝑌1 = √𝐴12 + 𝐵12 and ∅1 = tan−1 (𝐴1 )
1
The output y has half wave and quarter wave symmetries
4𝑁 4𝑁 −𝑗𝛽
(cos 𝛽 − 𝑗 𝑠𝑖𝑛𝛽) = 𝑒
𝜋𝑀 𝜋𝑀
4𝑁
𝜋𝑀
 −𝛽 ; 𝑀 ≥𝑎 (5)

[Link]’s College of Engineering 85 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Describing Function analysis of nonlinear systems

Describing functions of nonlinear elements can be used for stability analysis of nonlinear control
systems. Also it is used to predict the sustained oscillations or limit cycles in the output of the system.
Consider a unity feedback system shown, in which the nonlinearity is represented by its describing function
𝐾𝑁 (𝑋, 𝜔) or 𝐾𝑁 . Let 𝐶(𝑗𝜔)/𝑅(𝑗𝜔) be the closed loop sinusoidal transfer function of the system
𝐶(𝑗𝜔) 𝐾𝑁 G(𝑗𝜔)
=
𝑅(𝑗𝜔) 1 + 𝐾𝑁 (𝑗𝜔)
The characteristic equation of the system is obtained by equating the denominator to zero.
1 + 𝐾𝑁 𝐺(𝑗𝜔) = 0
The Nyquist stability criteria can also be extended to the stability analysis of nonlinear systems. According
to the Nyquist stability criterion the system will exhibit systained oscillations or limit cycles when
𝐾𝑁 𝐺(𝑗𝜔) = −1
This implies that the sustained oscillations or limit cycles will occur if 𝐾𝑁 𝐺(𝑗𝜔) locus pass through the
critical point, −1 + 𝑗0 in the complex plane. Above equation can be written as
1
𝐺(𝑗𝜔) = −
𝐾𝑁
This implies that the critical point, −1 + 𝑗0 becomes the critical locus which is the locus of −1/𝐾𝑁 . Hence
the intersection point of 𝐺(𝑗𝜔) locus and −1/𝐾𝑁 locus will give the amplitude and frequency of limit cycle.

To determine the stability of the system due to nonlinearity sketch the −1/𝐾𝑁 locus and 𝐺(𝑗𝜔) locus (polar
plot) in complex plane. From the sketches the following conclusions can be obtained.

1. If the −1⁄𝐾 locus is not enclosed by the 𝐺(𝑗𝜔) locus then the system is stable or there is no limit
𝑁
cycle at steady state.
2. If the −1⁄𝐾 locus is enclosed by the 𝐺(𝑗𝜔) locus then the system is unstable.
𝑁
3. If the −1⁄𝐾 locus and the 𝐺(𝑗𝜔) locus intersect, then the system output may exhibit a sustained
𝑁
oscillation or a limit cycle. The amplitude of the limit cycle is, given by the value of −1⁄𝐾 locus
𝑁
at the intersection point. The frequency of the limit cycle is given by the frequency of G(jω)
corresponding to the intersection point.

CONCEPT OF ENCLOSURE
In a complex plane the −1/𝐾𝑁 locus is said to be enclosed by𝐺(𝑗𝜔) locus if it lies in the region to the
right of an observer travelling through 𝐺(𝑗𝜔) locus in the direction of increasing .
[Link]’s College of Engineering 86 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
In a complex plane −1/𝐾𝑁 locus is not enclosed by 𝐺(𝑗𝜔) if it lies in the region to the left of an
observer travelling through 𝐺(𝑗𝜔) locus in the direction of increasing  .
If the −1/𝐾𝑁 l ocus and 𝐺(𝑗𝜔) locus intersect as shown in figure, then for an observer
travelling through 𝐺(𝑗𝜔) locus in the direction of increasing  , the region on the Right is unstable
region and the region on the left is stable region .

STABLE AND UNSTABLE LIMIT CYCLES

The −1/𝐾𝑁 locus may intersect 𝐺(𝑗𝜔) locus at one or more points. There exists a limit cycle
at every intersecting point. These limit cycles can be either stable or unstable limit cycles, as shown
in figure.
If −1/𝐾𝑁 locus travels in unstable region and it intersect 𝐺(𝑗𝜔) locus to-enter stable region then
the limit cycle corresponding to that intersection point is stable limit cycle.
If −1/𝐾𝑁 locus travels in stable region and it intersect 𝐺(𝑗𝜔) locus to enter unstable region
then the limit cycle corresponding to that intersection point is unstable limit cycle.

Problem : A servo system used for positioning a load has backlash characteristics as shown in figure
1. The block diagram of the system is shown in figure 2. The magnitude and phase of the describing
function of backlash nonlinearity for various values of b/X are listed in table-1, where X = Maximum
value of input sinusoidal signal to the nonlinearity.

[Link]’s College of Engineering 87 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Show that the system is stable if K=1. Also show that limit cycle exists when K=2. Investigate the
stability of these limit cycles and determine their frequency and b/X.

Solution

The magnitude of 𝐺(𝑗𝜔) in 𝑑𝑏 and phase of 𝐺(𝑗𝜔) are calculated for various values of 𝜔 and
tabulated. The Nichols pot of 𝐺(𝑗𝜔) is sketched in an ordinary graph sheet.

Nichols plot of G(jω) when K=2

When K=2 the magnitude of G(jω) increases by an amount 20 log2=6db. The phase of G(jω) is not
altered. The increase in magnitude is independent of frequency. Hence, G(jω) locus when K=2 is obtained
by shifting the locus of G(jω) when K=1 , by 6db upwards.

The magnitude and phase of the describing function of backlash, KN is listed in the problem in table for
various values of b/X. Using these values of KN and  KN the values of 1/KN in db and  1/KN are
calculated for various values of b/X are listed in table. Using these values the locus of -1/Kn is sketched.

[Link]’s College of Engineering 88 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

STABILITY ANALYSIS
Case (i) when K = 1

From the Nichols plots it is observed that when K = 1, 𝐺(𝑗𝜔) locus does not enclose −1/𝐾𝑛 locus. Hence
the system is stable.

Case (ii) when K = 2

From the Nichols plots it is observed that when K = 2; 𝐺(𝑗𝜔) locus intersects −1/𝐾𝑛 locus at two points.
At one intersection point unstable limit cycle exists and at another intersection point stable limit cycle
exists.

The coordinates corresponding to unstable limit cycle

= (14.2 db, -120) = 1014.2/20 -120=5.1-120

The coordinates corresponding to stable limit cycle

= (1.1 db, -165) = 101.1/20 -165=1.14-165

RESULT

• The unstable limit cycle exists when 𝑏⁄𝑋 = 0.316 and the frequency of oscillation is 0.36 rad/sec.

• The stable limit cycle exists when 𝑏⁄𝑋 = 0.464 and the frequency of oscillation is 1.07 rad/sec.

[Link]’s College of Engineering 89 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 90 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 91 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 92 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

[Link]’s College of Engineering 93 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

For the nonlinear system given below, investigate the possibility of a limit cycle. If limit cycle exists, find
its magnitude and frequency. (May 2017)

Solution
10
Given 𝑁 = 1, 𝐺(𝑠) = 𝑠(𝑠+1)(𝑠+2)
4𝑁
The describing function of the nonlinear element is 𝐾𝑁 = 𝜋𝑀
1 𝜋𝑀 𝜋𝑀 𝜋𝑀
− =− =− =−
𝐾𝑁 4𝑁 4×1 4
𝟏
Locus of −
𝑲𝑵
1
When 𝑀 → 0, − 𝐾 → 0
𝑁
1
When M → ∞, − K → ∞
N

[Link]’s College of Engineering 94 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
1
The locus of −𝐾 is shown in figure.
𝑁
Locus of 𝑮(𝒋𝝎)
10
𝐺(𝑠) = 𝑠(𝑠+1)(𝑠+2)
Convert pole zero form into time constant form and replace 𝑠 by 𝑗𝜔
5
G(jω) = jω(1+jω)(1+0.5jω)
5  0°
=
𝜔 90° √1+𝜔2  tan−1 𝜔 √1+0.25𝜔2  tan−1 0.5𝜔
5
|G(jω)| =
ω √1+𝜔2 √1+0.25𝜔2

G(jω) = −90° − tan−1 𝜔 − tan−1 0.5𝜔

When 𝜔 → 0, 𝐺(𝑗𝜔) → ∞
When 𝜔 → ∞, G(j𝜔) → 0
The locus of 𝐺(𝑗𝜔) is shown in figure.

The intersection point P can be determined by solving the equation. At the intersection point,
∴ |G(jω)| = 1 and G(jω) = −180°
Let, 𝜔𝑙1 = frequency when G(jω) = −1
At 𝜔 = 𝜔𝑙1 ,
G(jω) = −90° − tan−1 𝜔𝑙1 − tan−1 0.5𝜔𝑙1 = −180°
− tan−1 𝜔𝑙1 − tan−1 0.5𝜔𝑙1 = −180° + 90°
tan−1 𝜔𝑙1 + tan−1 0.5𝜔𝑙1 = 90°
On taking 𝑡𝑎𝑛 on either side,
𝑡𝑎𝑛(tan−1 𝜔𝑙1 + tan−1 0.5𝜔𝑙1 ) = tan 90°
tan(tan−1 𝜔𝑙1 ) + tan(tan−1 0.5𝜔𝑙1 )
= tan 90°
1 − tan(tan−1 𝜔𝑙1 ) × tan(tan−1 0.5𝜔𝑙1 )
𝜔𝑙1 + 0.5𝜔𝑙1 1
=∞=
1 − 𝜔𝑙1 × 0.5𝜔𝑙1 0
2 2 2
1
1 − 𝜔𝑙1 × 0.5𝜔𝑙1 = 0; 1 − 0.5𝜔𝑙1 = 0; 0.5𝜔𝑙1 = 1; 𝜔𝑙1 = ; 𝜔𝑙1 = √2 𝑟𝑎𝑑/𝑠𝑒𝑐
0.5
The frequency of limit cycle = √2 𝑟𝑎𝑑/𝑠𝑒𝑐
At 𝜔 = 𝜔𝑙1 , |G(jω)| = 1
5
|G(jω)| = 2
ω √1+𝜔 √1+0.25𝜔2
5
|G(jω)| =
ω √1+𝜔2 √1+0.25𝜔2
5 5 5
= 2 ×√3×1.5
= 3 = 1.66
2 2 √
√2 √1+(√2) √1+0.25(√2)

1
The 𝐺(𝑗𝜔) locus intersects the − K at (−1.66,0). Limit cycle exists at the intersection point
N
𝜋𝑀
Amplitude of limit cycle= − 4
= −1.66
Hence 𝑀 = 2.1146. the limit cycle is −2.1146 sin √2 𝑡

[Link]’s College of Engineering 95 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Unit IV
Introduction to Optimal Control
INTRODUCTION
The sources available on the planet earth are being consumed at an alarming rate and in order to
preserve whatever is left for the future generations, it is essential that consumption has to be reduced to
the minimum possible. The situation applies to almost all resources, energy, fuel, food and so on. Any
system that is designed in these years should confirm to stringent conditions on energy, efficiency and
quality to avoid unnecessary wastage of resources. For this, it is essential that design is made according
to strict constraints rather than trial and error approach as followed in the early development of control
systems. The requirements on the system formulated as specifications should be satisfied and all the
physical and mathematical constraints met simultaneously. This requires formulation of the design problem
as maximization or minimization of a performance measure subject to the set of constraints. Such a design
is said to be an optimal design.

Any optimization problem is characterised by a performance measure which translates the

requirements on the system into a mathematical function and a set of constraints which have to be met
while the performance measure is maximized or minimized. The success and effectiveness of the
optimization problem lies in how well the performance measure reflects the requirements on the system
and how well the constraints describe the physical realities. Optimization in classical control theory did
not receive much attention due to the limited computing facility available at that time. But we start by
reviewing the attempts at optimization in the classical control systems before proceeding to the modern
approach using state space models.

CLASSICAL CONTROL AND OPTIMIZATION

In the classical control using transfer function and frequency domain approach some attempts were
made to arrive at an optimal design. The performance measures were based on the integral of the error
and its adaptations. Some of the performance measures used in classical control are the Integral of Error
(IE), Integral of the Absolute Error (IAE), Integral of the Squared Error (ISE) and Int egral of Time
Absolute Error (ITAE) criteria are discussed below.

1. Integral of the Error (IE)

∞
J1 = ∫ e(t)dt
0

[Link]’s College of Engineering 96 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
This will ensure that even if the error itself is small, if it persists for a long time, its integral will be
significant and sufficient penalty will be given.

2. Integral of Absolute Error (IAE)

∞
J1 = ∫ |e(t)|dt
0

Positive and negative errors are given equal weightage in this measure.
3. Integral of the Squared Error (ISE)

∞
J1 = ∫ e(t)2 dt
0

Penalty for small errors is significantly reduced as t he square of small quantities will be still smaller in
relation to the square of large errors which becomes relatively larger when squared.
4. Integral of the Time Absolute Error (ITAE)

∞
J1 = ∫ t e(t)dt
0

Absolute value of the error integrated after multiplying with time.

5. Integral of Time Squared Error (ITSE)

∞
J1 = ∫ t e(t)2 dt
0

Among these the 1SE criterion has been reasonably well studied due to this being a quadratic function and
the solution is mathematically elegant and tractable.

15. b. For the system of the following figure, compute the value of 𝐾 that minimizes 𝐼𝑆𝐸 for
the unit step input. (May 2013)

Problem 1: Consider the feedback system shown. Compute the value of K that minimizes ISE for the
unit step input.

Solution

[Link]’s College of Engineering 97 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
For the system under consideration,
𝐸(𝑠) 1 1 𝑠
= = =
𝑅(𝑠) 1 + 𝐺(𝑠)𝐻(𝑠) 1 + 𝐾 𝑠 + 𝐾
𝑠
For unit step input,
1 1
𝑅(𝑠) = ∴ 𝐸(𝑠) =
𝑠 𝑠+𝐾
On taking inverse Laplace Transform,
𝑒(𝑡) = 𝑒 −𝐾𝑡
∞
1
𝐼𝑆𝐸 = ∫ 𝑒 2 (𝑡)𝑑𝑡 =
0 2𝐾
Obviously, the minimum value of ISE is obtained as 𝐾 → ∞. This is an impractical solution resulting in
excessive strain on the physical components of the system. We must include the ‘cost’ of the control effort.
The quadratic performance index,
∞
𝐽 = ∫ [𝑒 2 (𝑡) + 𝑢2 (𝑡)]𝑑𝑡
0

may serve the objective. From figure, 𝑢(𝑡) = 𝐾𝑒(𝑡) = 𝐾 𝑒 −𝐾𝑡

Therefore,
∞ ∞
𝑒 −3𝐾𝑡 𝐾 2 𝑒 −3𝐾𝑡 1 𝐾
𝐽=∫ [𝑒 −2𝐾𝑡 2 −2𝐾𝑡 ]𝑑𝑡
+𝐾 𝑒 = [ + ] =0−[ + ]
0 −2𝐾 −2𝐾 0 −2𝐾 −2
1 𝐾
∴𝐽= +
2𝐾 2
The minimum value of J is obtained when
𝜕𝐽 1 1
= − 2 + = 0 𝑜𝑟 𝐾 = 1
𝜕𝐾 2𝐾 2
Note that,
2 𝐽 1
= 3 >0
𝐾 2 𝐾
The minimum value of J is 1. This solution weighs error and control effort equally, seems to be acceptable.
We assign larger weight to error minimization.
∞
𝐽 = ∫ [𝑒 2 (𝑡) +  𝑢2 (𝑡)]𝑑𝑡 ;  = 0.5
0

For the system under consideration,

1 𝐾
+ 𝐽=
2𝐾 2
𝜕𝐽 1 1 1 1
=− 2+ =0=− 2+ = 0 𝑜𝑟 𝐾 = √2 𝑎𝑛𝑑 𝐽 = 0.707
𝜕𝐾 2𝐾 2 2𝐾 2×2

[Link]’s College of Engineering 98 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
When  is greater than unity, it means that more importance is given to the constraints on amplitude of
𝑢(𝑡) compared to the performance of the system. A suitable value of  is chosen so that relative importance
of the system performance is contrasted with the importance of the limit on control effort.

FORMULATION OF THE OPTIMAL CONTROL PROBLEM

For a continuous time system, the state equation in general can be written as
ẋ = f(x(t), u(t), t), t ∈ [t 0 , t i ]
and for a discrete time system
x(k + 1) = f(x(k), u(k), k), k ∈ [k 0 , k j ]
The problem is to control the plant so as to achieve certain specified objectives while satisfying the
operational constraints. The steps involved in the solution of the optimal control problem are:

1. Compute the performance index 𝐽 as a function of the free parameters 𝑘1 , 𝑘2 , … .. , 𝑘𝑛 of the system
with fixed configuration:
𝐽 = 𝐽(𝑘1 , 𝑘2 , . … .. , 𝑘𝑛 )
2. Determine the solution set 𝑘𝑖 of the equations
𝜕𝐽
= 0; 𝑖 = 1, 2, … … … , 𝑛
𝜕𝑘𝑖
This gives the necessary conditions for J to be minimum. From the solution set of these equations, find
the subset that satisfies the sufficient conditions which require that the Hessian matrix is positive
definite.
𝜕 2𝐽 𝜕 2𝐽 𝜕 2𝐽
,,,,,,
𝜕𝑘12 𝜕𝑘1 𝜕𝑘2 𝜕𝑘1 𝜕𝑘𝑛
𝜕 2𝐽 𝜕 2𝐽 𝜕 2𝐽
𝐻 = 𝜕𝑘2 𝜕𝑘1 ,,,,,,
𝜕𝑘22 𝜕𝑘2 𝜕𝑘𝑛
; ; ;
,,,,,,
𝜕 2𝐽 𝜕 2𝐽 𝜕 2𝐽
,,,,,,
[𝜕𝑘𝑛 𝜕𝑘1 𝜕𝑘𝑛 𝜕𝑘2 𝜕𝑘𝑛2 ]
𝜕2 𝐽 𝜕2 𝐽
Since = 𝜕𝑘
𝜕𝑘𝑖 𝜕𝑘𝑗 𝑗 𝜕𝑘𝑖

the matrix H is always symmetric.

The design of an optimal controller is based on the following factors relating to the plant and to
the nature of its connection with the controller.
[Link]’s College of Engineering 99 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
1. Characteristics of the plant.
2. The requirements to be satisfied by the plant.
3. The nature of information supplied to the controller about the plant.

1. Characteristics of the Plant

It will be normally necessary to put constraints on the control variables

u1 (t), u2 (t), u3 (t), … . , up (t) due to the realistic demands of the situation. For example, an automobile
can take only a certain maximum fuel based on its cylinder capacity, combustion mechanism and other
limitations. The state variables of the system, preferably, physical variables cannot take unconstrained
values as they will have to be kept within specified limits. A control Signal which satisfies all the
constraints during the entire control interval t ∈ [t 0 , t i ], of interest is called an admissible control.

Let U be the set of admissible control with 𝑢(𝑡) ∈ 𝑈 f o r a l l 𝑡 ∈ [𝑡0 , 𝑡1 ]

Constraints can be laid on the states 𝑥(𝑡) also during the interval of operation of the optimal control.
A state trajectory that satisfies the state variable constraints during the time interval 𝑡 ∈ [𝑡0 , 𝑡1 ] will be
called the admissible trajectory. The set of admissible state trajectory is denoted by 𝑋. Then 𝑥(𝑡) is
admissible if 𝑥(𝑡) ∈ 𝑋 for all 𝑡 ∈ [𝑡0 , 𝑡1 ].
2. Requirements on the Plant

The designer translates the requirements or specifications to be satisfied by the plant into a
mathematical performance criterion or index and then tries to find a solution that optimises the performance
measure. In certain cases, the performance measure can be chosen without difficulty, but in some other
cases it may have to be selected with difficulty with lot of subjectivity involved. In some cases several
performance measures will have to be attempted to find a satisfactory and suitable solution.

Typical optimal control performance measures

14 a (i) Explain briefly about the minimum-time control of linear time-invariant systems. (8) May 2016

2.1.1 Minimum Time Control Problem

The time required for the solution of the problem is to be minimized similar to reaching particular
destination from a starting point. The starting point is specified as initial state 𝑥(𝑡0 ) = 𝑥0 . The final state
is required to be on as specific region 𝑆 of the state space. 𝑆 may be called a target set. If the final state
is fixed, the target is a straight line. The objective is to transfer the system from the initial stale 𝑥(𝑡0 ) =
𝑥0 to the specified target in the minimum possible time. The performance measure thus becomes:
𝑡1
𝐽 = 𝑡1 − 𝑡0 = ∫ 𝑑𝑡
𝑡0
[Link]’s College of Engineering 100 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
where 𝑡0 is the first instant of time when 𝑥(𝑡) is in 𝑆 of interest.

The interception of an attacking aircraft by a missile is a typical example. The attacking aircraft
has to be destroyed before it can do any damage to human beings or machinery. This type of problem is
always associated with inequality constraints on the magnitude of the input-output vector elements. In the
absence of any constraints, the change can be made in zero time which is impossible as no physical
device can give infinite energy to a device.
2.1.2 Minimum Energy Problem

The objective is to transfer the system from the initial state 𝑥(𝑡0 ) = 𝑥0 to the specified target with
minimum use of energy during the process. The performance measure thus becomes:
𝑡1
𝐽 = ∫ 𝑢2 (𝑡)𝑑𝑡
𝑡0

If there are several inputs, i.e. 𝑢(𝑡) is a vector, then the performance measure is modified as:
𝑡1
𝐽 = ∫ [𝑢𝑇 (𝑡) 𝑢(𝑡)]𝑑𝑡
𝑡0

If it is required to give different weightage or penalty for different components of the input, we may
introduce a weighting matrix 𝑅 into the performance measure as:
𝑡1
𝐽 = ∫ [𝑢𝑇 (𝑡) 𝑅 𝑢(𝑡)]𝑑𝑡
𝑡0

Here 𝑅 is a diagonal matrix with suitable values on the diagonal to give proper penalty for the different
components of 𝑢.
2.1.3 Minimum Fuel Problem

The quantity of fuel consumed in rockets and propulsion of space crafts is to be optimised to the
maximum extent as the mass of fuel is a significant part of the whole vehicle and any reduction of the fuel
consumed can result in a better utilization of the payload to carry additional equipment for scientific
investigations. The rate of consumption of fuel in jet engines can be taken as directly proportional to the
thrust (input to the space vehicle):
𝑡1
𝐽 = ∫ |𝑢(𝑡)|𝑑𝑡
𝑡0

Allowing the various jets present to different consumption ratio, it can be modified to:
𝑡1
𝐽 = ∫ [𝑘1 |𝑢1 (𝑡)| + 𝑘2 |𝑢2 (𝑡)| + 𝑘3 |𝑢3 (𝑡)|+, … … . . 𝑘𝑝 |𝑢𝑝 (𝑡)|]𝑑𝑡
𝑡0

Where 𝑘𝑖 ’s are nonnegative weighting functions.

2.1.4 Regulator Problem

Here the objective of the optimal control problem is to transfer a system from an initial state 𝑥(𝑡0 ) =
𝑥 0 to the desired state 𝑥(𝑡1 ) = 𝑥1 (may be the equilibrium point) with minimal integral squared error.
[Link]’s College of Engineering 101 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
1 1)
Relative to the state 𝑥 the quantity (𝑥(𝑡) − 𝑥 can be viewed as instantaneous error. If we transform the
system co-ordinates such that 𝑥1 can be the origin, 𝑥(𝑡) itself will be the error.
𝑡1 𝑛 𝑡1
𝐽 = ∫ [∑(𝑥𝑖 (𝑡)) ] 𝑑𝑡 = ∫ [𝑥 𝑇 (𝑡)𝑥(𝑡)] 𝑑𝑡
2
𝑡0 𝑖=1 𝑡0

or in general with different weights given to different errors:

𝑡1
𝐽 = ∫ [𝑥 𝑇 (𝑡) 𝑄(𝑥) 𝑥(𝑡)] 𝑑𝑡
𝑡0

where 𝑄 is a real symmetric positive definite matrix, a diagonal matrix in the simplest case. Further, if it
is also desired to minimize the deviation of the final state, the performance index can be modified as:
𝑡1
𝐽 = 𝑥 𝑇 (𝑡1 ) 𝐻 𝑥(𝑡1 ) + ∫ [𝑥 𝑇 (𝑡) 𝑄(𝑥) 𝑥(𝑡)] 𝑑𝑡
𝑡0

Where the first term is introduced to minimize the deviation from the final state and 𝐻 is again a positive
definite or positive semi definite matrix. A more realistic performance measure may also put some penalty
on the energy consumed as well.
𝑡1
𝐽 = 𝑥 𝑇 (𝑡1 ) 𝐻 𝑥(𝑡1 ) + ∫ [𝑥 𝑇 (𝑡) 𝑄(𝑥) 𝑥(𝑡) + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
𝑡0

2.1.5 Output Regulator Problem

In output regulator, the concern is to make the output 𝑦(𝑡) to be as close as possible to the desired
output. The performance measure can be modified as:
1 𝑡1
𝐽 = 𝑦 𝑇 (𝑡1) 𝐻 𝑦(𝑡1 ) + ∫ [𝑦 𝑇 (𝑡) 𝑄(𝑥) 𝑦(𝑡) + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 𝑡0
2.1.6 Tracking Problem

Here the objective is to maintain the system state 𝑥(𝑡) as close as possible to a desired state or
reference state 𝑥(𝑡) in t he interval [𝑡0 , 𝑡1 ]. i.e.
1 𝑇 1 𝑡1
𝐽= 𝑒 (𝑡1 ) 𝐻 𝑒(𝑡1 ) + ∫ [𝑒(𝑡) 𝑄(𝑥) 𝑒(𝑡) + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 2 𝑡0
Where 𝑒(𝑡) = 𝑥(𝑡) − 𝑟(𝑡)
2.2 Choice of Performance Measure

The choice of the performance index or measure is dictated by the physical requirements which
have to be translated to a mathematical function. In some cases very complex functions will be required
to meet all the specifications.

3. Nature of Information about the Plant Supplied to the Controller

If the optimal control is determined as a function of time for a specified initial state,
𝑢∗ (𝑡) = 𝑓(𝑥(𝑡0 ), 𝑡)
[Link]’s College of Engineering 102 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
then the optimal control is said to be of the open loop form. This requires no additional information and in
the absence of meaningful disturbances and errors, open loop control can be successful. If a functional
relationship of the form:
𝑢∗ (𝑡) = 𝑓(𝑥(𝑡), 𝑡)
can be found for the optimal control at time 𝑡, then the optimal control is said to be closed loop and
𝑓() is called an optimal control law. Here the controller gets information on the actual state 𝑥(𝑡) of the
process and will be more effective. The effect of disturbances and errors can be significantly reduced in
closed loop control.

Definition l: The optimal control problem can thus be formally defined as: Find an admissible control 𝑢∗ (𝑡)
which causes the system 𝑥̇ (𝑡) = 𝑓(𝑥(𝑡), 𝑢(𝑡), 𝑡) 𝑑𝑡 to follow an admissible trajectory 𝑥 ∗ (𝑡) and minimizes
the performance measure:
𝑡1
𝐽 = ℎ(𝑥(𝑡1 ), 𝑡1 ) + ∫ 𝑔(𝑥(𝑡), 𝑢(𝑡), 𝑡) 𝑑𝑡
𝑡0

Several approaches are available for the solution of the optimal control problem based on the nature of
the performance measure and constraints. Some of the commonly used methods are:
1. Method of calculus of variations.
2. Maximum principle due to Pontryagin.
3. Dynamic Programming approach of Bellman,
4. Lyapunov's method.
5. Linear Programming approach
6. Integer programming approach etc.

STATE REGULATOR PROBLEM

State Regulator Design through Matrix Riccati equation(LQR)

15. b. Derive the Ricatti’s equation as a solution of steady state optimal control problem. (Nov
2015)
Quadratic performance measures have been extensively used in the design of optimal control
systems because of its mathematical tractability and the ability to accommodate realistic functions such
as energy function and squared errors or deviations. When squared errors are used as performance
measures, large errors are given a much higher penalty than small errors as the square of large quantities
are very large and squares of small quantities are smaller.
Consider the system described by
𝑥̇ = 𝐴𝑥 + 𝐵𝑢 (1)
where x is the n × 1 state vector, 𝑢 is the 𝑝 × 1 input vector, A and B are respectively, 𝑛 × 𝑛 and 𝑛 × 𝑝
real constant matrices. Find the control law,
𝑢 = −𝐾𝑥 (2)
[Link]’s College of Engineering 103 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
Where 𝐾 is 𝑝 × 𝑛 real constant unconstrained gain matrix, that minimizes the following performance index
subject to the initial conditions 𝑥(0) ≜ 𝑥 0 :
1 ∞
𝐽 = 2 ∫0 [𝑥 𝑇 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡 (3)

where, 𝑄 is a positive definite, real, symmetric, constant matrix and 𝑅 is the 𝑝 × 𝑝 positive definite, real,
symmetric, constant matrix.
There are several ways to solve this optimal control problem and here we use Laypunov approach.
Substitute equation (2) in (1)
𝑥̇ = 𝐴𝑥 − 𝐵𝐾𝑥 = (𝐴 − 𝐵𝐾) 𝑥 (4)
Since, (𝐴, 𝐵) pair is completely controllable, there exists a feedback matrix 𝐾 such that (𝐴 − 𝐵𝐾) is a
stable matrix. The controllability of the given plant, thus, ensures the existence of a 𝐾 that minimizes 𝐽.
The time derivative of Laypunov function is
1
𝑉̇ (𝑥) = − 2 𝑥 𝑇 (𝑄 + 𝐾 𝑇 𝑅𝐾)𝑥 (5)

The laypunov function

1
𝑉(𝑥) = 2 𝑥 𝑇 𝑃 𝑥 (6)

Where 𝑃 is a positive definite, real, symmetric, constant matrix, and

(𝐴 − 𝐵𝐾)𝑇 𝑃 + 𝑃(𝐴 − 𝐵𝐾) + 𝐾 𝑇 𝑅𝐾 + 𝑄 = 0 (7)
The performance index
1
𝐽 = 2 𝑥 𝑇 (0) 𝑃 𝑥(0) (8)

Since feedback matrix 𝐾 is unconstrained, the optimum value of 𝐽 is independent of initial conditions. The
optimal 𝐾𝑖𝑗 ’s are obtained from the equations,
𝑃
= 0, 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖, 𝑗 (9)
 𝑘𝑖𝑗

Since 𝑅 has been assumed to be a positive definite matrix, we can write,

𝑅 = T𝑇 T
Where  is a nonsingular matrix. Then equation (7) can be written as
(𝐴𝑇 − 𝐾 𝑇 𝐵 𝑇 )𝑃 + 𝑃(𝐴 − 𝐵𝐾) + 𝑄 + 𝐾 𝑇 𝑇 T 𝑇𝐾 = 0
𝐴𝑇 𝑃 + 𝑃𝐴 − 𝐾 𝑇 𝐵 𝑇 𝑃 − 𝑃𝐵𝐾 + 𝑄 + 𝐾 𝑇 T T T𝐾 = 0
𝐴𝑇 𝑃 + 𝑃𝐴 + [T𝐾 − (𝑇 𝑇 )−1 𝐵𝑇 𝑃]𝑇 [T𝐾 − (T 𝑇 )−1 𝐵 𝑇 𝑃] − 𝑃𝐵𝑅 −1 𝐵 𝑇 𝑃 + 𝑄 = 0 (10)
The condition (9) for unconstrained minimization of J leads to the following equations:

[T𝐾 − (T 𝑇 )−1 𝐵𝑇 𝑃]𝑇 [T𝐾 − (T 𝑇 )−1 𝐵𝑇 𝑃] = 0
 𝑘𝑖𝑗
Since the matrix within the brackets is non-negative definite, the minimum occurs when it is zero, or when
𝑇𝐾 = (𝑇 𝑇 )−1 𝐵 𝑇 𝑃
Hence, 𝐾 = 𝑇 −1 (T 𝑇 )−1 𝐵 𝑇 𝑃 = 𝑅 −1 𝐵 𝑇 𝑃 (11)
Equation (11) gives the optimal gain matrix 𝐾. Thus the optimal control law is given by,

[Link]’s College of Engineering 104 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
−1 𝑇
𝑢 = −𝐾𝑥 = 𝑅 𝐵 𝑃 (12)
The matrix 𝑃 in equation (12) must satisfy equation (10), or the following reduced equation:
𝐴𝑇 𝑃 + 𝑃𝐴 − 𝑃𝐵𝑅 −1 𝐵𝑇 𝑃 + 𝑄 = 0 (13)
Equation (13) is called Matrix Riccati equation. This optimal control is called
the Linear Quadratic Regulator (LQR)

Steps to be followed in designing Optimal Control La w using Riccatti equation

• Plant model 𝑥̇ = 𝐴𝑥 + 𝐵𝑢;

1 ∞
• Performance index 𝐽 = 2 ∫0 (𝑥 𝑇 𝑄𝑥 + 𝑢𝑇 𝑅𝑢)𝑑𝑡

• Feedback control law 𝑢 = −𝐾𝑥

Step 1: Solve the matrix Riccatti equation

𝑨𝑻 𝑷 + 𝑷𝑨 − 𝑷𝑩𝑹−𝟏 𝑩𝑻 𝑷 + 𝑸 = 𝟎 𝒇𝒐𝒓 𝑷
Step 2: Substitute 𝑃 into 𝑲 = 𝑹−𝟏 𝑩𝑻 𝑷 to find the optimal gain matrix.
Step 3: If the matrix [𝐴 − 𝐵𝐾] is stable, this method always gives the correct result.
(i.e. poles should lie on LH of S plane). |𝑰 − (𝑨 − 𝑩𝑲)| = 𝟎

14. a (ii) Explain the properties and use of LQR. (8) (May 2016)

Properties and Use of the LQR

Static Gain: The LQR generates a static gain matrix K, which is not a dynamical system. Hence, the order
of the closed-loop system is the same as that of the plant.

Robustness: The LQR achieves infinite gain margin: 𝑘𝑔 = ∞, implying that the loci of (PC) (scalar case)
or (det(1 + 𝑃𝐶) − 1) (MIMO case) approach the origin along the imaginary axis. The LQR also
guarantees phase margin 𝛾 = 60 degrees. This is in good agreement with the practical guidelines for control
system design.

Output Variables: In many cases, it is not the states 𝑥 which are to be minimized, but the output variables
𝑦. In this case, we set the weighting matrix 𝑄 = 𝐶 𝑇 𝑄 ′ 𝐶, since 𝑦 = 𝐶𝑥, and the auxiliary matrix 𝑄 ′ weights
the plant output.

Behavior of Closed-Loop Poles: Expensive Control. When 𝑅 ≫ 𝐶 𝑇 𝑄 ′ 𝐶R the cost function is dominated
by the control effort 𝑢, and so the controller minimizes the control action itself. In the case of a completely
stable plant, the gain will indeed go to zero, so that the closed-loop poles approach the open-loop plant
poles in a manner consistent with the scalar root locus.
The optimal control must always stabilize the closed-loop system, however, so there should be some
account made for unstable plant poles. The expensive control solution puts stable closed-loop poles at the
mirror images of the unstable plant poles.

[Link]’s College of Engineering 105 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
𝑇 ′
Behavior of Closed-Loop Poles: Cheap Control. When 𝑅 ≪ 𝐶 𝑄 𝐶C, the cost function is dominated by
the output errors 𝑦, and there is no penalty for using large 𝑢. There are two groups of closed-loop poles.
First, poles are placed at stable plant zeros, and at the mirror images of the unstable plant zeros. This part
is akin to the high-gain limiting case of the root locus. The remaining poles assume a Butterworth pattern,
whose radius increases to infinity as 𝑅 becomes smaller and smaller.

The Butterworth pattern refers to an arc in the stable left-half plane. The angular separation of n closed-
loop poles on the arc is constant, and equal to 180°/𝑛. An angle 90°/𝑛 separates the most lightly-damped
poles from the imaginary axis.
𝒙̇ 𝟎 𝟏 𝒙𝟏 𝟎
Problem 2. Consider a double integrator plant having state model [ 𝟏 ] = [ ] [𝒙 ] + [ ] 𝒖; 𝑪 =
𝒙̇ 𝟐 𝟎 𝟎 𝟐 𝟏
[𝟏 𝟎] . It is desired to design an optimal state feedback controller of the form 𝒖(𝒕) = −𝑲𝒙(𝒕) such
∞ 𝟏 𝟎
that the performance measure 𝑱 = ∫𝟎 [𝒙𝑻 𝑸𝒙 + 𝒖𝑻 𝒖]𝒅𝒕 where 𝑸 = [ ] , 𝑹 = 𝟏 is minimized.
𝟎 𝟎
SOLUTION
0 1 0 1 0 𝑝1 𝑝2
Given 𝐴 = [ ];𝐵 = [ ]; 𝑄 = [ ] ; 𝑅 = 1; 𝑃 = [𝑝 𝑝3 ]
0 0 1 0 0 2

The reduced Riccatti equation is

𝐴𝑇 𝑃 + 𝑃𝐴 − 𝑃𝐵𝑅 −1 𝐵 𝑇 𝑃 + 𝑄 = 0
0 1 𝑇 𝑝1 𝑝2 𝑝1 𝑝2 0 1 𝑝1 𝑝2 0 𝑝1 𝑝2 1 0 0 0
[
0
] [
0 𝑝2 𝑝3 ] + [𝑝2 𝑝3 ] [0 0
] − [𝑝
2 𝑝3 ] [1] [1] [0 1] [𝑝
2 𝑝3 ] + [0 0] = [0 0]

0 0 𝑝1 𝑝2 𝑝1 𝑝2 0 1 𝑝1 𝑝2 0 𝑝1 𝑝2 1 0 0 0
[
1
][
0 𝑝2 𝑝3 ] + [𝑝2 𝑝3 ] [0 0
] − [𝑝
2 𝑝3 ] [1] [1] [0 1] [𝑝2 𝑝3 ] + [0 0] = [0 0
]

0 0 0 0 𝑝1 𝑝2 1 0 𝑝1 𝑝2 1 0 0 0
[ ]+[ ] − [𝑝 𝑝3 ] [0 0] [𝑝2 𝑝3 ] + [0 0] = [0 0]
𝑝1 𝑝2 0 𝑝2 2

0 0 𝑝 2 𝑝1 𝑝2 1 0 0 0
[ ]− [ 1 2 ] + [0 ]=[ ]
𝑝1 2𝑝2 𝑝1 𝑝2 𝑝2 0 0 0

Expanding
1 − 𝑝22 = 0 𝑝2 = 1
𝑝1 − 𝑝2 𝑝3 = 0
2𝑝2 − 𝑝32 = 0

Solving for 𝑝1 , 𝑝2 , 𝑝3 we have 𝑝1 = √2; 𝑝2 = 1; 𝑝3 = √2

𝑝1 𝑝2 √2 1
To check whether 𝑃 is positive definite, we get 𝑃 = [𝑝 𝑝3 ] = [ 1 √2] which indeed is positive
2

definite.

The feedback gain matrix is obtained as 𝐾 = 𝑅 −1 𝐵 𝑇 𝑃

𝐾 = [1] [0 √2 1 ] = [
1] [ 1 √2]
1 √2

[Link]’s College of Engineering 106 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

The optimal control signal 𝑢(𝑡) = −𝐾𝑥 = −𝑥1 − √2 𝑥2

Which is independent of any initial state under the given performance index. Under state feedback, the
system matrix becomes:
0 1
𝐴 − 𝐵𝐾 = [ ]
−1 −√2
The characteristic equation of the closed loop system is :
2 + √2 + 1 = 0
√2
Which has the roots at  = (−1 ± 𝑗) with damping ratio of 0.707 which gives a good performance.
2

The open loop transfer function of the optimally designed system is

∞
15 a. (ii) Derive the control law which minimizes the performance index 𝐽 = ∫0 (𝑥12 + 𝑢2 ) 𝑑𝑡
for the system as follows. (Nov 2014)
𝑥̇ 1 0 1 𝑥1 0
[ ]=[ ][ ] + [ ]𝑢
𝑥̇ 2 0 0 𝑥2 1

1
(𝑠 + )
𝐾∅(𝑠)𝐵 = 𝐾(𝑠𝐼 − 𝐴)−1 𝐵 = √2 √2
𝑠2
∞
Problem 3. Obtain the control law which minimize the performance index 𝑱 = ∫𝟎 (𝒙𝟐𝟏 + 𝒖𝟐 )
𝟎 𝟏 𝟎
for the system 𝑿̇ = [ ]𝑿 + [ ]𝒖
𝟎 𝟎 𝟏
SOLUTION
0 1 0
Given, 𝐴 = [ ] ;𝐵 = [ ]
0 0 1
Modify the equation to
1 ∞ 𝑇
𝐽= ∫ [𝑥 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 0
1 ∞ 2
𝐽 = ∫ [2𝑥1 + 2𝑢2 ] 𝑑𝑡
2 0
On comparing the above equations,
2 0
𝑄=[ ] ; 𝑅 = [2]
0 0
The reduced matrix Riccati equation is
𝐴𝑇 𝑃 + 𝑃𝐴 − 𝑃𝐵𝑅 −1 𝐵 𝑇 𝑃 + 𝑄 = 0
0 1 𝑇 𝑝11 𝑝12 𝑝11 𝑝12 0 1 𝑝11 𝑝12 0 1 𝑝11 𝑝12 2 0 0 0
[
0
] [
0 𝑝12 𝑝22 ] + [𝑝12 𝑝22 ] [0 0] − [𝑝12 𝑝22 ] [1] [2] [0 1] [𝑝
12 𝑝22 ] + [0 0
]=[
0 0
]

0 0 𝑝11 𝑝12 𝑝11 𝑝12 0 1 𝑝11 𝑝12 0 1 𝑝11 𝑝12 2 0 0 0

[ ][
1 0 𝑝12 𝑝22 ] + [𝑝12 𝑝22 ] [0 0
] − [𝑝
12 𝑝22 ] [1] [2] [0 1] [𝑝12 𝑝22 ] + [0 0] = [0 0
]

[Link]’s College of Engineering 107 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

0 0 0 𝑝11 𝑝12 𝑝11 𝑝12 2 0 0 0

[ ]+[ ] − [𝑝 ] [0 1⁄2] [𝑝 𝑝22 ] + [0 ]=[ ]
𝑝11 𝑝12 0 𝑝12 22 12 0 0 0

0 𝑝11 𝑝12 𝑝11 𝑝12 2 0 0 0

[ ] − [𝑝 ] [0 1⁄2] [𝑝 𝑝22 ] + [0 0] = [0 0]
𝑝11 2𝑝12 22 12

𝑝12
0
[
0 𝑝11
]−[ 2 ] [𝑝11 𝑝12 2 0 0 0
𝑝11 2𝑝12 𝑝22 𝑝12 𝑝22 ] + [0 0
]=[
0 0
]
0
2

2
𝑝12 𝑝12 𝑝22
0 𝑝11 0+ 0+
2 2 2 0 0 0
[ ]− 2 +[ ]=[ ]
𝑝11 2𝑝12 𝑝12 𝑝22 𝑝22 0 0 0 0
[ 0 + 0+
2 2 ]
Expanding the above,

2
𝑝12
0− +2=0 (1)
2

𝑝12 𝑝22
𝑝11 − +0=0 (2)
2

𝑝12 𝑝22
𝑝11 − +0=0 (3)
2

2
𝑝22
2𝑝12 − +0=0 (4)
2
From (1),
2
𝑝12 2
− + 2 = 0; 𝑝12 = 4; 𝑝12 = 2
2
From (4)
2
𝑝22
2𝑝12 − =0
2
2
𝑝22 2 2
= 2𝑝12 ; 𝑝22 = 4𝑝12 ; 𝑝22 = 8; 𝑝12 = 2√2
2
From (2)
𝑝12 𝑝22
𝑝11 − =0
2
𝑝12 𝑝22 2 × 2√2
𝑝11 = ; = 2√2
2 2
𝑝11 𝑝12
] = [2√2 2 ]
𝑃 = [𝑝
12 𝑝22 2 2√2

[Link]’s College of Engineering 108 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
∗ −1 𝑇
The optimal control law 𝑢 = −𝐾𝑥(𝑡) = −𝑅 𝐵 𝑃 𝑥(𝑡)

1 2√2 2 𝑥 (𝑡)
= − [2] [0 1] [ ][ 1 ]
2 2√2 𝑥2 (𝑡)

2√2 2 𝑥 (𝑡)
= − [0 1⁄2] [ ][ 1 ]
2 2√2 𝑥2 (𝑡)

𝑥 (𝑡)
=−[1 √2] [ 1 ]
𝑥2 (𝑡)
𝑢∗ = −𝑥1 (𝑡) − √2𝑥2 (𝑡)

∴ 𝐾 = [1 √2]

State Regulator Design through the Lyapunov Equation (Constrained Optimization)

Consider the system described by
𝑥̇ = 𝐴𝑥 + 𝐵𝑢 ; 𝑥(0) ≜ 𝑥 0 (1)
where x is the n × 1 state vector, 𝑢 is the 𝑝 × 1 input vector, A and B are respectively, 𝑛 × 𝑛 and 𝑛 × 𝑝
real constant matrices. Find the control law,
𝑢 = −𝐾𝑥 (2)
Where 𝐾 is 𝑝 × 𝑛 real constant constrained gain matrix, that minimizes the following performance index
subject to the initial conditions 𝑥(0) ≜ 𝑥 0 :
1 ∞
𝐽 = 2 ∫0 [𝑥 𝑇 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡 (3)

where, 𝑄 is a positive definite, real, symmetric, constant matrix and 𝑅 is the 𝑝 × 𝑝 positive definite, real,
symmetric, constant matrix.
With the control law, the closed loop system becomes
𝑥̇ = 𝐴𝑥 − 𝐵𝐾𝑥 = (𝐴 − 𝐵𝐾) 𝑥 (4)
Since, (𝐴, 𝐵) pair is completely controllable, there exists a feedback matrix 𝐾 such that (𝐴 − 𝐵𝐾) is a
stable matrix. The controllability of the given plant, thus, ensures the existence of a 𝐾 that minimizes 𝐽.
The optimization problem is to determine the values of free parameters of the matrix 𝐾 so as to minimize
the performance index.
Substituting the control vector 𝑢 from equation (2) in the performance index 𝐽 of equation (3)
1 ∞
𝐽 = 2 ∫0 [𝑥 𝑇 𝑄 𝑥 + 𝑥 𝑇 𝐾 𝑇 𝑅 𝐾𝑥] 𝑑𝑡
1 ∞
𝐽 = 2 ∫0 𝑥 𝑇 (𝑄 + 𝐾 𝑇 𝑅 𝐾)𝑥 𝑑𝑡 (5)

The value of the performance index for the system trajectory starting at 𝑥(0) is 𝑉(𝑥(0))
[Link]’s College of Engineering 109 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
The time derivative of Laypunov function is
1 ∞
𝑉̇ (𝑥) = 2 𝑥 𝑇 (𝑄 + 𝐾 𝑇 𝑅𝐾)𝑥|
𝑡

1 𝑇 1
= 𝑥 (∞)[𝑄 + 𝐾 𝑇 𝑅 𝐾] 𝑥(∞) − 𝑥 𝑇 (𝑡)[𝑄 + 𝐾 𝑇 𝑅 𝐾] 𝑥(𝑡)
2 2
The matrix (𝐴 − 𝐵𝐾) is stable. From equation (4)
𝑥(∞) → 0
1
Therefore, 𝑉̇ (𝑥) = − 2 𝑥 𝑇 (𝑄 + 𝐾 𝑇 𝑅𝐾)𝑥 (6)

Since 𝑉̇ (𝑥) is a quadratic in 𝑥 and the plant equation in linear, let us assume that 𝑉(𝑥) is also given by
quadratic form:
1
𝑉(𝑥) = 2 𝑥 𝑇 𝑃 𝑥 (7)

Where 𝑃 is a positive definite, real, symmetric, constant matrix, therefore,

1 𝑇
𝑉̇ (𝑥) = (𝑥 𝑃𝑥 + 𝑥 𝑇 𝑃 𝑥̇ )
2
Substituting for 𝑥̇ from equation (4)
1
𝑉̇ (𝑥) = 𝑥 𝑇 [(𝐴 − 𝐵𝐾)𝑇 𝑃 + 𝑃(𝐴 − 𝐵𝐾)]𝑥 (8)
2

Comparing (8) and (6)

1 𝑇 1
𝑥 [(𝐴 − 𝐵𝐾)𝑇 𝑃 + 𝑃(𝐴 − 𝐵𝐾)]𝑥 = − 𝑥 𝑇 (𝑄 + 𝐾 𝑇 𝑅𝐾)𝑥
2 2
Since the above equality holds for arbitrary 𝑥(𝑡),

(𝐴 − 𝐵𝐾)𝑇 𝑃 + 𝑃(𝐴 − 𝐵𝐾) + 𝐾 𝑇 𝑅𝐾 + 𝑄 = 0 (9)

This equation is of the form of Lyapunov equation. Equation (9) has 𝑛2 nonlinear algebraic equation. Since
𝑛(𝑛+1)
𝑛 × 𝑛 matrix 𝑃 is symmetric, we need to solve only equations for the elements 𝑃𝑖𝑗 as functions of
2

the feedback matrix 𝐾.

𝑉(𝑥(0)) is the value of performance index for the system trajectory starting at 𝑥(𝑡), from equation (7)
1
𝐽 = 2 𝑥 𝑇 (0) 𝑃 𝑥(0) (10)

A suboptimal control law may be obtained by minimizing 𝐽 with respect to the available elements 𝑘𝑖𝑗 of 𝐾,
 [𝑥 𝑇 (0)𝑃 𝑥(0)]
=0
 𝑘𝑖𝑗
If for the suboptimal solution thus obtained, the matrix (A − BK) is stable, then the minimization of J as
the procedure described above gives the correct result.

[Link]’s College of Engineering 110 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Steps to be followed in designing Optimal Control Law using Lyapunov

Function
• Plant model
𝒙̇ = 𝑨𝒙 + 𝑩𝒖; 𝐱(𝟎) ≜ 𝒙𝟎
• Performance index
𝟏 ∞ 𝑻
𝑱= ∫ (𝒙 𝑸𝒙 + 𝒖𝑻 𝑹𝒖)𝒅𝒕
𝟐 𝟎
• Feedback control law
𝒖 = −𝑲𝒙
Step 1: Identify A, B, Q and R matrices and solve the Lyapunov equation and determine the
P matrix.
[𝑨 − 𝑩𝑲]𝑻 𝑷 + 𝑷[𝑨 − 𝑩𝑲] + 𝑸 + 𝑲𝑻 𝑹𝑲=0
Step 2: Compute the performance index 𝐽. The performance index becomes a function of K.
𝟏 𝑻
𝑱=𝒙 (𝟎)𝑷𝒙(𝟎)
𝟐
Step 3: Determine the solution of the equation.
𝑱
=𝟎
 𝒌𝒊𝒋
From the solution set of these equations, find the subset that satisfies the sufficient
conditions for minimization. This subset is the optimum parameter set.
Step 4: Examine (𝐴 − 𝐵𝐾) for stability. |𝑆𝐼 − 𝐴 + 𝐵𝐾| = 0, Poles/roots should be negative

Additional Problems
Problem 4. Consider the second order system shown in figure. Determine the value of damping ratio
 > 𝟎 which minimizes the integral square error
∞
𝑱 = ∫ 𝒆𝟐 (𝒕)𝒅𝒕
𝟎
For the initial conditions 𝒚(𝟎) = 𝟏, 𝒚̇ (𝟎) = 𝟎. What is the minimum value of the performance
index?

Solution
The differential equation governing the system
1⁄
𝑌(𝑠) 𝐺(𝑠) 𝑠(𝑠 + 2) 1
= = = 2
𝑅(𝑠) 1 + 𝐺(𝑠)𝐻(𝑠) 1 + 1⁄ 𝑠 + 2𝑠 + 1
𝑠(𝑠 + 2)
[Link]’s College of Engineering 111 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
𝑅(𝑠) = 0 ; ∴ 𝑌(𝑠)[𝑠 + 2𝑠 + 1] = 0
2

Taking inverse Laplace Transform,

𝑦̈ + 2𝑦̇ + 𝑦 = 0 (1)
Defining the state variables as 𝑥1 = 𝑦; 𝑥2 = 𝑦̇ , equation (1) becomes,
𝑥̇ 2 + 2𝑥2 + 𝑥1 = 0
𝑥̇ 2 = −𝑥1 − 2𝑥2 and 𝑥̇ 1 = 𝑥2
𝑥̇ 0 1 𝑥1
𝑥̇ = [ 1 ] = [ ] [𝑥 ] ;
𝑥̇ 2 −1 −2 2
0 1 1
𝐴=[ ] 𝑎𝑛𝑑 𝑥(0) = [ ]
−1 −2 0
Modify the equation to
1 ∞ 𝑇
𝐽 = ∫ [𝑥 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 0
The performance index,
∞ 𝟏 ∞ 𝟏 ∞
𝑱 = ∫𝟎 𝒆𝟐 (𝒕)𝒅𝒕 = 𝟐 ∫𝟎 𝟐𝒙𝟐𝟏 (𝒕)𝒅𝒕 = 𝟐 ∫𝟎 𝟐𝑥 𝑇 𝑥𝒅𝒕
2 0
From the above equation, 𝑄 = [ ] ; 𝑅 = 0; 𝐾 = 0
0 0
The Lyapunov equation, [𝑨 − 𝑩𝑲]𝑻 𝑷 + 𝑷[𝑨 − 𝑩𝑲] + 𝑸 + 𝑲𝑻 𝑹𝑲
Since, 𝐾 = 0, the Lyapunov equation becomes,
∴ 𝑨𝑻 𝑷 + 𝑷𝑨 + 𝑸 = 𝟎

0 −1 𝑝11 𝑝12 𝑝11 𝑝12 0 1 2 0 0 0

[ ][
1 −2 𝑝12 𝑝22 ] + [𝑝12 𝑝22 ] [−1 −2] + [0 0
]=[
0 0
]

−𝑝12 −𝑝22 −𝑝12 𝑝11 − 2 𝑝12 2 0 0 0

[𝑝 ] + [ ]+[ ]=[ ]
11 − 2 𝑝12 𝑝12 − 2 𝑝22 −𝑝22 𝑝12 − 2 𝑝22 0 0 0 0
Expanding the above,

−𝑝12 − 𝑝12 + 2 = 0 (2)

−𝑝22 + 𝑝11 − 2 𝑝12 + 0 = 0 (3)

2𝑝12 − 4 𝑝22 = 0 (4)

From (2), −𝑝12 − 𝑝12 + 2 = 0
−2𝑝12 = −2; 𝑝12 = 1
From (4), 2𝑝12 − 4 𝑝22 = 0
2𝑝12 2 × 1 1
4 𝑝22 = 2𝑝12 ; 𝑝22 = = =
4 4 2
From (3), −𝑝22 + 𝑝11 − 2 𝑝12 = 0

[Link]’s College of Engineering 112 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
1
𝑝11 = 𝑝22 + 2 𝑝12 = + 2
2
1
2 + 1
𝑝 𝑝12 2
𝑃 = [𝑝11 𝑝22 ] =
12 1
1
[ 2]
𝟏 𝑻
𝑱= 𝒙 (𝟎)𝑷𝒙(𝟎)
𝟐
1
2 + 2 1
1 1
=2 [1 0] [ 1] [ ]
1 0
2

1 1 1 1
𝐽= [2 + 1] [ ] =  +
2 2 0 4
To find the minimum 𝐽, differentiate and w.r.t  and equate to 0
𝜕𝐽 1 𝜕 2𝐽
= 1 − 2 = 0;  = 0.5; 2 > 0; 𝐽𝑚𝑖𝑛 = 1
𝜕 4 𝜕

Problem 5. A unity feedback system has the forward path transfer function
𝑲
𝑮(𝒔) =
𝒔(𝒔 + 𝜶)
The reference input is a unit step function, and the system is at rest initially.
(i) Taking K as constant, determine the value of 𝜶 which minimizes the integral square error.
(ii) Taking 𝜶 as constant, determine the value of K which minimizes the integral square error.
Solution
The closed loop transfer function,
𝑌(𝑠) 𝐾
= 2
𝑅(𝑠) 𝑠 + 𝛼𝑠 + 𝐾
𝑌(𝑠)[𝑠 2 + 𝛼𝑠 + 𝐾] = 𝐾 𝑅(𝑠)
Take inverse Laplace Transform,
𝑦̈ + 𝛼𝑦̇ + 𝐾 𝑦 = 𝐾 𝑟 (1)
𝑒 = 𝑟 − 𝑦; 𝑒̇ = 𝑦̇ ; 𝑒̈ = 𝑦̈ (2)
Substitute (2) in (1), 𝑒̈ + 𝛼𝑒̇ + 𝐾 𝑒 = 0 (3)
Assume the state variable as, 𝑥1 = 𝑒; 𝑥2 = 𝑒̇
Equation (3) becomes, 𝑥̇ 2 + 𝛼𝑥2 + 𝐾 𝑥1 = 0
𝑥̇ 2 = −𝛼𝑥2 − 𝐾 𝑥1 and 𝑥̇ 1 = 𝑥2
The state equation,
𝑥̇ 0 1 𝑥1
𝑥̇ = [ 1 ] = [ ][ ];
𝑥̇ 2 −𝐾 −𝛼 𝑥2
0 1 1
𝐴=[ ] 𝑎𝑛𝑑 𝑥(0) = [ ]
−𝐾 −𝛼 0
Modify the equation to
[Link]’s College of Engineering 113 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
∞
1
𝐽= ∫ [𝑥 𝑇 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 0
The performance index,
∞ 𝟏 ∞
𝑱 = ∫𝟎 𝒆𝟐 (𝒕)𝒅𝒕 = 𝟐 ∫𝟎 𝟐𝒙𝟐𝟏 (𝒕)𝒅𝒕
2 0
From the above equation, 𝑄 = [ ] ; 𝑅 = 0; 𝐾 = 0
0 0
The Lyapunov equation, [𝑨 − 𝑩𝑲]𝑻 𝑷 + 𝑷[𝑨 − 𝑩𝑲] + 𝑸 + 𝑲𝑻 𝑹𝑲 = 𝟎
Since, 𝐾 = 0, the Lyapunov equation becomes,
∴ 𝑨𝑻 𝑷 + 𝑷𝑨 + 𝑸 = 𝟎

0 −𝐾 𝑝11 𝑝12 𝑝11 𝑝12 0 1 2 0 0 0

[ ][ 𝑝22 ] + [𝑝12 𝑝22 ] [−𝐾 ]+[ ]=[ ]
1 −𝛼 𝑝12 −𝛼 0 0 0 0

−𝐾𝑝12 −𝐾𝑝22 −𝐾𝑝12 𝑝11 − 𝛼 𝑝12 2 0 0 0

[ ]+[ ]+[ ]=[ ]
𝑝11 − 𝛼 𝑝12 𝑝12 − 𝛼 𝑝22 −𝐾𝑝22 𝑝12 − 𝛼 𝑝22 0 0 0 0
Expanding the above,

−𝐾𝑝12 − 𝐾𝑝12 + 2 = 0 (4)

2𝑝12 − 2𝛼 𝑝22 = 0 (5)

−𝐾𝑝22 + 𝑝11 − 𝛼 𝑝12 + 0 = 0 (6)

From (4), −𝐾𝑝12 − 𝐾𝑝12 + 2 = 0
−2 1
−2𝐾𝑝12 = −2; 𝑝12 = −2𝐾 = 𝐾

From (5), 2𝑝12 − 2𝛼 𝑝22 = 0

−2𝑝12 1
−2𝛼 𝑝22 = −2𝑝12 ; 𝑝22 = =
−2𝛼 𝐾𝛼
From (6), −𝐾𝑝22 + 𝑝11 − 𝛼 𝑝12 = 0
𝛼 𝐾 𝛼 1
𝑝11 = 𝛼𝑝12 + 𝐾 𝑝22 = + = +
𝐾 𝐾𝛼 𝐾 𝛼
𝛼 1 1
𝑝 𝑝12 +
∴ 𝑃 = [𝑝11 𝐾 𝛼 𝐾]
12
𝑝22 ] = [ 1 1
𝐾 𝐾𝛼
𝟏 𝑻
𝑱= 𝒙 (𝟎)𝑷𝒙(𝟎)
𝟐
𝛼 1 1
1
+𝛼 1
𝐾 𝐾
=2 [1 0] [ 1 1 ][ ]
0
𝐾 𝐾𝛼

1 𝛼 1 1 1 𝛼 1
𝐽= [ + ] [ ]= +
2 𝐾 𝛼 𝐾 0 2𝐾 2𝛼

[Link]’s College of Engineering 114 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
(i) Taking K as constant, determine the value of 𝜶 which minimizes the integral square error.
𝜕𝐽 1 1 1 1
= − 2 = 0; 2
= ; 𝛼 = √𝐾
𝜕𝛼 2𝐾 2𝛼 2𝛼 2𝐾

(ii) Taking 𝜶 as constant, determine the value of K which minimizes the integral square error.
𝜕𝐽 𝛼 1
= (− 2 ) = 0; 𝐾 → ∞
𝜕𝐾 2 𝐾

15. b .Consider a system described by the equations

𝟎 𝟏 𝟎
𝒙̇ = [ ]𝒙 + [ ]𝒖
𝟎 𝟎 𝟏
𝒙𝟏 (𝟎) = 𝒙𝟐 (𝟎) = 𝟏. Consider the feedback control law 𝒖 = −𝒙𝟏 − 𝑲𝒙𝟐
𝟏 ∞
(i) Find the value of K so that 𝑱 = 𝟐 ∫𝟎 (𝒙𝟐𝟏 + 𝒙𝟐𝟐 ) 𝒅𝒕 is minimized.
(ii) Find minimum value of 𝑱. (May 2014)

Problem 6. Consider a system described by the equations

𝟎 𝟏 𝟎
𝒙̇ = [ ]𝒙 + [ ]𝒖
𝟎 𝟎 𝟏
𝒙𝟏 (𝟎) = 𝒙𝟐 (𝟎) = 𝟏. Consider the feedback control law 𝒖 = −𝒙𝟏 − 𝑲𝒙𝟐
𝟏 ∞
(i) Find the value of K so that 𝑱 = 𝟐 ∫𝟎 (𝒙𝟐𝟏 + 𝒙𝟐𝟐 ) 𝒅𝒕 is minimized.
(ii) Find minimum value of 𝑱.

SOLUTION
Modify the equation to
1 ∞ 𝑇
𝐽 = ∫ [𝑥 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 0
1 ∞ T 1 ∞
𝐽= ∫ x x 𝑑𝑡 = ∫ (𝑥12 + 𝑥22 ) 𝑑𝑡
2 0 2 0
On comparing the above equations,
1 0
𝑄=[ ] ; 𝑅 = [0]
0 1
𝑥1
u(t) = −x1 − kx2 = −[1 𝐾 ] [𝑥 ] = −𝐾𝑥
2
0 1 0 1
Given 𝐴 = [ ] ; 𝐵 = [ ] ; 𝑥(0) = [ ] 𝐾 = [1 𝐾 ]
0 0 1 1
The Lyapunov equation, [𝑨 − 𝑩𝑲]𝑻 𝑷 + 𝑷[𝑨 − 𝑩𝑲] + 𝑸 + 𝑲𝑻 𝑹𝑲
Since, 𝑅 = 0, the Lyapunov equation becomes,
[𝑨 − 𝑩𝑲]𝑻 𝑷 + 𝑷[𝑨 − 𝑩𝑲] + 𝑸 = 𝟎

[𝐴 − 𝐵𝐾] = [0 1] − [0] [1 0 1 0 0 0 1
𝐾] = [ ]−[ ]=[ ]
0 0 1 0 0 1 𝐾 −1 −𝐾

0 −1 𝑝11 𝑝12 𝑝11 𝑝12 0 1 1 0 0 0

[ ][ 𝑝22 ] + [𝑝12 𝑝22 ] [−1 −𝐾 ] + [0 ]=[ ]
1 −𝐾 𝑝12 1 0 0

[Link]’s College of Engineering 115 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

−𝑝12 −𝑝22 −𝑝12 𝑝11 − 𝐾 𝑝12 1 0 0 0

[𝑝 𝑝12 − 𝐾 𝑝22 ] + [−𝑝22 ]+[ ]=[ ]
11 − 𝐾𝑝12 𝑝12 − 𝐾 𝑝22 0 1 0 0
Expanding the above,

−𝑝12 − 𝑝12 + 1 = 0 (1)

−𝑝22 + 𝑝11 − 𝐾 𝑝12 = 0 (2)

𝑝12 − 𝐾𝑝22 + 𝑝12 − 𝐾 𝑝22 + 1 = 0 (3)

Solving the above equations,
From (1), −𝑝12 − 𝑝12 + 1 = 0
1
2𝑝12 = 1; 𝑝12 =
2
From (3), 𝑝12 − 𝐾𝑝22 + 𝑝12 − 𝐾 𝑝22 + 1 = 0
2𝑝12 − 2𝐾 𝑝22 = −1
−2𝐾 𝑝22 = −1 − 2𝑝12
1
1 + 2𝑝12 1 + 2 2 2 1
𝑝22 = = = =
2𝐾 2𝐾 2𝐾 𝐾
From (2), −𝑝22 + 𝑝11 − 𝐾 𝑝12 = 0
1 𝐾 𝐾2 + 2
𝑝11 = 𝑝22 + 𝐾 𝑝12 = + =
𝐾 2 2𝐾

𝐾2 + 2 1
𝑝 𝑝12 2𝐾 2]
∴ 𝑃 = [𝑝11 𝑝22 ] = [ 1
12 1
2 𝐾
𝟏
(ii) 𝑱 = 𝟐 𝒙𝑻 (𝟎)𝑷𝒙(𝟎)
𝐾2 +2 1
1 2𝐾 2 1
=2 [1 1] [ 1 1
] [ ]
1
2 𝐾

1 𝐾2 + 2 1 1 1 1 𝐾 2 + 2𝐾 + 4
𝐽= [ + ] [1] =
2 2𝐾 + 2 2 𝐾 4𝐾
𝜕𝐽 𝜕 𝐾 2 + 2𝐾 + 4 3
= ( ) = 0; 𝐾 = 2; 𝐽𝑀𝑖𝑛 =
𝜕𝐾 𝜕𝐾 4𝐾 2
To check the stability: eigen values should be negative. i. e. |I − (A − BK)| = 0
Solving this equation results,
2 + 2 + 1 = 0; 1 = −1; 2 = −1
Hence the system is stable.
To find the sensitivity
For 𝐾 = 1.5, 𝐽 = 1.54
For 𝐾 = 2, 𝐽 = 1.5
[Link]’s College of Engineering 116 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
∆𝑱⁄ 𝟎.𝟎𝟒⁄
𝑶𝒑𝒕 𝑱 𝟏.𝟒
Sensitivity 𝑺𝑲 = ∆𝑲⁄ = 𝟎.𝟓⁄ = 𝟎. 𝟏𝟎𝟕
𝑲 𝟐

Problem 7. Consider a system

𝟎 𝟏 𝟎
𝐱̇ = [ ]𝐱 + [ ]𝐮
𝟎 𝟎 𝟏
Assume the control law is
𝐮(𝐭) = −𝐤 𝟏 𝐱 𝟏 − 𝐤 𝟐 𝐱 𝟐 .
Fix the value of 𝐤 𝟏 so that the closed loop system has undamped natural frequency of 𝟐 𝒓𝒂𝒅/𝒔𝒆𝒄.
Determine the free parameter 𝐤 𝟐 so that
𝟏 ∞
𝐉 = ∫ (𝒙𝟐𝟏 + 𝒙𝟐𝟐 )𝐝𝐭
𝟐 𝟎
is minimized for 𝐱(𝟎) = [𝟏 𝟎]𝐓 . Find the minimum value of the performance index.

SOLUTION
Modify the equation to
1 ∞ 𝑇
𝐽= ∫ [𝑥 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 0
1 ∞ 2
𝐽 = ∫ (𝑥1 + 𝑥22 ) 𝑑𝑡
2 0
On comparing the above equations,
1 0
𝑄=[ ] ; 𝑅 = [0]
0 1
𝑥1
u(t) = −𝑘1 x1 − k 2 x2 = −[𝑘1 𝑘2 ] [𝑥 ] = −𝐾𝑥
2
0 1 0 1
𝐴=[ ] ; 𝐵 = [ ] ; 𝑥(0) = [ ] ; 𝐾 = [𝑘1 𝑘2 ]
0 0 1 0

The closed loop system,

0 1 0 0 1 0 0 0 1
[𝐴 − 𝐵𝐾] = [ ] − [ ] [𝑘1 𝑘2 ] = [ ]−[ ]=[ ]
0 0 1 0 0 𝑘1 𝑘2 −𝑘1 −𝑘2
0 1
ẋ = [ ] x or 𝑥̇ 1 = 𝑥2 ; 𝑥̇ 2 = −𝑘1 𝑥1 − 𝑘2 𝑥2 ; 𝑥̇ 2 = 𝑥̈ 1
−𝑘1 −𝑘2

Eliminating 𝑥2 from this equation yields, 𝑥̈ 1 + 𝑘2 𝑥̇ 1 + 𝑘1 𝑥1 = 0

Since the undamped natural frequency is specified as 2 𝑟𝑎𝑑/𝑠𝑒𝑐, we obtain 𝑘1 = 4 (compare with
standard second order system. i.e. 𝑥̈ 1 + 2𝜔𝑛 𝑥̇ 1 + 𝜔𝑛2 𝑥1 = 0

0 1
Therefore, [𝐴 − 𝐵𝐾] = [ ]
−4 −𝑘2
The Lyapunov equation, [𝑨 − 𝑩𝑲]𝑻 𝑷 + 𝑷[𝑨 − 𝑩𝑲] + 𝑸 + 𝑲𝑻 𝑹𝑲
Since, 𝑅 = 0, the Lyapunov equation becomes,
[𝑨 − 𝑩𝑲]𝑻 𝑷 + 𝑷[𝑨 − 𝑩𝑲] + 𝑸 = 𝟎

[Link]’s College of Engineering 117 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

0 −4 𝑝11 𝑝12 𝑝11 𝑝12 0 1 1 0 0 0

[ ][ ] + [ 𝑝22 ] [−4 −𝑘2 ] + [0 1] = [0 0]
1 −𝑘2 𝑝12 𝑝22 𝑝12

−4𝑝12 −4𝑝22 −4𝑝12 𝑝11 − 𝑘2 𝑝12 1 0 0 0

[ ]+[ ]+[ ]=[ ]
𝑝11 − 𝑘2 𝑝12 𝑝12 − 𝑘2 𝑝22 −4𝑝22 𝑝12 − 𝑘2 𝑝22 0 1 0 0
Expanding the above,

−4𝑝12 − 4𝑝12 + 1 = 0 (1)

−4𝑝22 + 𝑝11 − 𝑘2 𝑝12 = 0 (2)

𝑝12 − 𝑘2 𝑝22 + 𝑝12 − 𝑘2 𝑝22 + 1 = 0 (3)

Solving the above equations,
From (1), −4𝑝12 − 4𝑝12 + 1 = 0
1
8𝑝12 = 1; 𝑝12 =
8
From (3), 𝑝12 − 𝑘2 𝑝22 + 𝑝12 − 𝑘2 𝑝22 + 1 = 0
2𝑝12 − 2𝑘2 𝑝22 = −1
−2𝑘2 𝑝22 = −1 − 2𝑝12
1 1
1 + 2𝑝12 1 + 2 8 1 + 4 5
𝑝22 = = = =
2𝑘2 2𝑘2 2𝑘2 8𝑘2
From (2), −4𝑝22 + 𝑝11 − 𝑘2 𝑝12 = 0
4 × 5 𝑘2 5 𝑘2
𝑝11 = 𝑝22 + 𝑘2 𝑝12 = + = +
8𝑘2 8 2𝑘2 8

5 𝑘2 1
𝑝 𝑝12
+
2𝑘2 8 8
∴ 𝑃 = [𝑝11 𝑝22 ] =
12 1 5
[ 8 8𝑘2 ]
𝟏 𝑻
𝑱= 𝒙 (𝟎)𝑷𝒙(𝟎)
𝟐
5 𝑘2 1
+
1 2𝑘2 8 8 1
=2 [1 0] [ 1 5 ][ ]
0
8 8𝑘2

1 5 𝑘2 1 1 1 5 𝑘2
𝐽= [ + ] [ ]= ( + )
2 2𝑘2 8 8 0 2 2𝑘2 8
𝜕𝐽 𝜕 1 5 𝑘2 𝜕2 𝐽
= 𝜕𝑘 (2 (2𝑘 + )) = 0; 𝐾 = √20; >0
𝜕𝑘2 2 2 8 𝜕𝑘22

√5
For this value of 𝑘2 , 𝐽𝑀𝑖𝑛 = 4

[Link]’s College of Engineering 118 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

15. a. Consider the system described by the state model. (May 2013)
𝟎 𝟏 𝟎
𝒙̇ = [ ] 𝒙 + [ ] 𝒖 ; 𝒚 = [𝟏 𝟎]𝒙
𝟎 −𝟐 𝟐𝟎
Find the optimal control law that minimizes
𝟏 ∞
𝑱 = 𝟐 ∫𝟎 (𝒚𝟐 + 𝒖𝟐 ) 𝒅𝒕

Problem 8. Consider the system described by the state model.

𝟎 𝟏 𝟎
𝒙̇ = [ ] 𝒙 + [ ] 𝒖 ; 𝒚 = [𝟏 𝟎]𝒙
𝟎 −𝟐 𝟐𝟎
𝟏 ∞
Find the optimal control law that minimizes 𝐉 = 𝟐 ∫𝟎 (𝒚𝟐 + 𝒖𝟐 )𝐝𝐭

SOLUTION
For output regulator problem the performance index is given by,
1 ∞
J = ∫ (𝑦 𝑇 𝑄 𝑦 + 𝑢𝑇 𝑅 𝑢)dt
2 0
Modify the equation to
1 ∞ 𝑇
𝐽 = ∫ [𝑥 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 0
Substituting 𝑦 = 𝐶𝑥, the performance index becomes,
1 ∞ 𝑇 𝑇
J = ∫ (𝑥 𝐶 𝑄 𝐶 𝑥 + 𝑢𝑇 𝑅 𝑢)dt
2 0
𝑄 is replace by 𝐶 𝑇 𝑄 𝐶
1 ∞ 𝑇 1
J = ∫ (𝑥 [ ] [1 0]𝑥 + 𝑢𝑇 𝑅 𝑢) dt
2 0 0
1 ∞ 1 0
J = ∫ (𝑥 𝑇 [ ] 𝑥 + 𝑢𝑇 𝑅 𝑢) dt
2 0 0 0
On comparing the above equations,
1 0
𝑄=[ ] ; 𝑅 = [1]
0 0
0 1 0
Given, 𝐴 = [ ] ;𝐵 = [ ]
0 −2 20
The reduced matrix Riccati equation is
𝐴𝑇 𝑃 + 𝑃𝐴 − 𝑃𝐵𝑅 −1 𝐵 𝑇 𝑃 + 𝑄 = 0

0 0 𝑝11 𝑝12 𝑝11 𝑝12 0 1 𝑝11 𝑝12 0 𝑝11 𝑝12 1 0 0 0

[
1
][
−2 𝑝12 𝑝22 ] + [𝑝12 𝑝22 ] [0 −2
] − [𝑝
12 𝑝22 ] [20] [1][0 20] [𝑝
12 𝑝22 ] + [0 0
]=[
0 0
]

0 0 0 𝑝11 − 2𝑝12 20𝑝12 𝑝11 𝑝12 1 0 0 0

[ ]+[ ]− [ ] [0 20] [𝑝 𝑝22 ] + [0 ]=[ ]
𝑝11 − 2𝑝12 𝑝12 − 2𝑝22 0 𝑝12 − 2𝑝22 20𝑝22 12 0 0 0

0 0 0 𝑝11 − 2𝑝12 0 400𝑝12 𝑝11 𝑝12 1 0 0 0

[ ]+[ ]− [ ][ 𝑝22 ] + [0 ]=[ ]
𝑝11 − 2𝑝12 𝑝12 − 2𝑝22 0 𝑝12 − 2𝑝22 0 400𝑝22 𝑝12 0 0 0

[Link]’s College of Engineering 119 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

2
0 0 0 𝑝11 − 2𝑝12 400𝑝12 400𝑝12 𝑝22 1 0 0 0
[ ]+[ ]− [ ]+[ ]=[ ]
𝑝11 − 2𝑝12 𝑝12 − 2𝑝22 0 𝑝12 − 2𝑝22 400𝑝12 𝑝22 2
400𝑝22 0 0 0 0
Expanding the above,

2
0 − 400𝑝12 +1=0 (1)

𝑝11 − 2𝑝12 − 400𝑝12 𝑝22 = 0 (2)

2
𝑝12 − 2𝑝22 + 𝑝12 − 2𝑝22 − 400𝑝22 =0 (3)
From (1),

2 2 2
1 1
0 − 400𝑝12 + 1 = 0; 400𝑝12 = 1; 𝑝12 = ; 𝑝12 = = 0.05
400 20
From (3)

2
𝑝12 − 2𝑝22 + 𝑝12 − 2𝑝22 − 400𝑝22 =0

2
2𝑝12 − 4𝑝22 − 400𝑝22 =0

2
400𝑝22 + 4𝑝22 − 2𝑝12 = 0

2
400𝑝22 + 4𝑝22 − 0.1 = 0

Solving this equation yields, 𝑝22 = −0.0216, 0.0116

Since p matrix is positive, take positive solution, 𝑝22 = 0.0116

From, 𝑝11 − 2𝑝12 − 400𝑝12 𝑝22

𝑝11 = 2𝑝12 + 400𝑝12 𝑝22

= 2 × 0.05 + 400 × −.05 × 0.0116

= 0.1 + 20 × 0.0116

= 0.1 + 0.23166 = 0.3317

𝑝11 𝑝12 0.3317 0.05
𝑃 = [𝑝 𝑝 ]= [ ]
12 22 0.05 0.0116

The optimal control law 𝑢∗ = −𝐾𝑥(𝑡) = −𝑅 −1 𝐵𝑇 𝑃𝑥(𝑡)

0.3317 0.05 𝑥 (𝑡)

=−[1] [0 20] [ ][ 1 ]
0.05 0.0116 𝑥2 (𝑡)

[Link]’s College of Engineering 120 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
𝑥1 (𝑡)
=− [1 0.232] [ ]
𝑥2 (𝑡)

𝑢∗ = −𝑥1 (𝑡) − 0.232𝑥2 (𝑡)

∴ 𝐾 = [1 0.232]

Problem 9. Find the optimal control law for the system

𝟎 𝟏 𝟎
𝐱̇ = [ ]𝐱 + [ ]𝐮
−𝟏 −𝟐 𝟏
∞
with the performance index 𝐉 = ∫𝟎 (𝐱 𝟏𝟐 + 𝐮𝟐𝟏 )𝐝𝐭.
SOLUTION
Modify the equation to
1 ∞ 𝑇
𝐽= ∫ [𝑥 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 0
1 ∞ 2
𝐽 = ∫ [2𝑥1 + 2𝑢12 ] 𝑑𝑡
2 0
On comparing the above equations,
2 0
𝑄=[ ] ; 𝑅 = [2]
0 0
0 1 0
Given 𝐴 = [ ] ;𝐵 = [ ] ;
−1 −2 1
The reduced matrix Riccati equation is
𝐴𝑇 𝑃 + 𝑃𝐴 − 𝑃𝐵𝑅 −1 𝐵 𝑇 𝑃 + 𝑄 = 0
0 −1 p11 p12 p11 p12 0 1 p11 p12 0 1 𝑝11 𝑝12 2 0 0 0
[
1
][
−2 p12 p22 ] + [p12 p22 ] [−1 −2] − [p12 p22 ] [1] [2] [0 1] [𝑝12 𝑝22 ] + [0 0] = [0 0
]

−𝑝12 −𝑝22 −𝑝12 𝑝11 − 2𝑝12 𝑝12 𝑝11 𝑝12 2 0 0 0

[𝑝 − 2𝑝 𝑝12 − 2𝑝22 ] + [−𝑝22 ] − [𝑝 ] [0 1⁄2] [ 𝑝 𝑝22 ] + [0 0] = [0 ]
11 12 𝑝12 − 2𝑝22 22 12 0

𝑝12
−𝑝12 −𝑝22 0
[𝑝
−𝑝12 𝑝11 − 2𝑝12 2 ] [𝑝11 𝑝12 2 0 0 0
11 − 2𝑝12 𝑝12 − 2𝑝22 ] + [−𝑝22 𝑝12 − 2𝑝22
]− [ 𝑝22 𝑝12 𝑝22 ] + [0 0
]=[
0 0
]
0
2

2
𝑝12 𝑝12 𝑝22
−𝑝12 −𝑝22 −𝑝12 𝑝11 − 2𝑝12 2 2 2 0 0 0
[𝑝 𝑝12 − 2𝑝22 ] + [−𝑝22 ]− +[ ]=[ ]
11 − 2𝑝12 𝑝12 − 2𝑝22 𝑝12 𝑝22 2
𝑝22 0 0 0 0
[ 2 2 ]
Expanding the above,

2
𝑝12
−𝑝12 − 𝑝12 − +2=0 (1)
2

[Link]’s College of Engineering 121 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
𝑝12 𝑝22
−𝑝22 + 𝑝11 − 2𝑝12 + +0=0 (2)
2

2
𝑝22
𝑝12 − 2𝑝22 + 𝑝12 − 2𝑝22 − +0=0 (3)
2
2
𝑝12
From (1), −𝑝12 − 𝑝12 − +2=0
2
2
𝑝12
+ 2𝑝12 − 2 = 0
2

Solving the above equation yields, 𝑝12 = −4.828 𝑎𝑛𝑑 0.8284. Since P is a positive matrix, choose
positive value for 𝑝12
𝑖. 𝑒., 𝑝12 = 0.8284

2
𝑝22
From (3), 𝑝12 − 2𝑝22 + 𝑝12 − 2𝑝22 − +0=0
2
2
𝑝22
2𝑝12 − 4𝑝22 − =0
2
2
𝑝22
+ 4𝑝22 − 1.6568 = 0
2
Solving the above equation yields, 𝑝22 = −8.3947 𝑎𝑛𝑑 0.3947. Since P is a positive matrix, choose
positive value for 𝑝22
𝑖. 𝑒., 𝑝22 = 0.3947
From (2)
𝑝12 𝑝22
−𝑝22 + 𝑝11 − 2𝑝12 +
+0=0
2
𝑝12 𝑝22 0.8284 × 0.3947
𝑝11 =− + 2𝑝12 + 𝑝22 ; − + (2 × 0.8284) + 0.3947 = 2.2151
2 2
𝑝11 𝑝12 2.2151 0.8284
𝑃 = [𝑝 𝑝 ]= [ ]
12 22 0.8284 0.3947

The optimal control law 𝑢∗ = −𝐾𝑥(𝑡) = −𝑅 −1 𝐵𝑇 𝑃𝑥(𝑡)

1 2.2151 0.8284 𝑥1 (𝑡)

=− [2] [0 1] [ ][ ]
0.8284 0.3947 𝑥2 (𝑡)
2.2151 0.8284 𝑥1 (𝑡)
=− [0 1⁄2] [ ][ ]
0.8284 0.3947 𝑥2 (𝑡)
𝑥1 (𝑡)
=−[0.4142 0.1974] [ ]
𝑥2 (𝑡)
𝑢∗ = −0.4142𝑥1 (𝑡) − 0.1974𝑥2 (𝑡)
∴ 𝐾 = [0.4142 0.1974]

[Link]’s College of Engineering 122 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

15.a. Determine the optimal control law for the system described by (May 2014)
𝟎 𝟏 𝟎
𝒙̇ (𝒕) = [ ] 𝒙(𝒕) + [ ] 𝒖(𝒕),
−𝟐 −𝟑 𝟏
such that the following performance index is minimized.
∞
𝑱 = ∫𝟎 (𝒙𝑻 𝒙 + 𝒖𝟐 ) 𝒅𝒕.

Problem 10. Determine the optimal control law for the system described by
𝟎 𝟏 𝟎
𝐱̇ (𝐭) = [ ] 𝐱(𝐭) + [ ] 𝐮(𝐭)
−𝟐 −𝟑 𝟏
such that the following performance index is minimized.
∞
𝐉 = ∫ (𝐱 𝐓 𝐱 + 𝒖𝟐 )𝐝𝐭
𝟎

SOLUTION
Modify the equation to
1 ∞ 𝑇
𝐽= ∫ [𝑥 𝑄 𝑥 + 𝑢𝑇 𝑅 𝑢] 𝑑𝑡
2 0
1 ∞ 𝑇 1 ∞ 2
𝐽 = ∫ [2𝑥 𝑥 + 2𝑢 ] 𝑑𝑡 = ∫ [2𝑥1 + 2𝑥22 + 2𝑢2 ] 𝑑𝑡
2
2 0 2 0
On comparing the above equations,
2 0
𝑄=[ ] ; 𝑅 = [2]
0 2
0 1 0
Given 𝐴 = [ ] ;𝐵 = [ ] ;
−2 −3 1
The reduced matrix Riccati equation is
𝐴𝑇 𝑃 + 𝑃𝐴 − 𝑃𝐵𝑅 −1 𝐵 𝑇 𝑃 + 𝑄 = 0
0 −2 p11 p12 p11 p12 0 1 p11 p12 0 1 𝑝11 𝑝12 2 0 0 0
[
1
][
−3 p12 p22 ] + [p12 p22 ] [−2 −3] − [p12 p22 ] [1] [2] [0 1] [𝑝12 𝑝22 ] + [0 2] = [0 0
]

−2𝑝12 −2𝑝22 −2𝑝12 𝑝11 − 3𝑝12 𝑝12 𝑝11 𝑝12 2 0 0 0

[ ]+[ ] − [𝑝 ] [0 1⁄2] [ 𝑝 𝑝22 ] + [0 2] = [0 0]
𝑝11 − 3𝑝12 𝑝12 − 3𝑝22 −2𝑝22 𝑝12 − 3𝑝22 22 12

𝑝12
0
[
−2𝑝12 −2𝑝22
]+[
−2𝑝12 𝑝11 − 3𝑝12
]− [ 2 ] [𝑝11 𝑝12 ] + [2 0] = [0 0]
𝑝11 − 3𝑝12 𝑝12 − 3𝑝22 −2𝑝22 𝑝12 − 3𝑝22 𝑝 𝑝12 𝑝22 0 2 0 0
0 22
2

𝑝212 𝑝12 𝑝22

−2𝑝12 −2𝑝22 −2𝑝12 𝑝11 − 3𝑝12
[ ]+[ ]− 2 2 + [2 0] = [0 0]
𝑝11 − 3𝑝12 𝑝12 − 3𝑝22 −2𝑝22 𝑝12 − 3𝑝22 𝑝12 𝑝22 2
𝑝22 0 2 0 0
[ 2 2 ]
Expanding the above,

2
𝑝12
−2𝑝12 − 2𝑝12 − +2=0 (1)
2

[Link]’s College of Engineering 123 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
𝑝12 𝑝22
−2𝑝22 + 𝑝11 − 3𝑝12 + +0=0 (2)
2

2
𝑝22
𝑝12 − 3𝑝22 + 𝑝12 − 3𝑝22 − +2=0 (3)
2
2
𝑝12
From (1), −2𝑝12 − 2𝑝12 − +2=0
2
2
𝑝12
+ 4𝑝12 − 2 = 0
2

Solving the above equation yields, 𝑝12 = −8.4721 𝑎𝑛𝑑 0.4721. Since P is a positive matrix, choose
positive value for 𝑝12
𝑖. 𝑒., 𝑝12 = 0.4721

2
𝑝22
From (3), 𝑝12 − 3𝑝22 + 𝑝12 − 3𝑝22 − +2=0
2
2
𝑝22
2𝑝12 − 6𝑝22 − +2=0
2
2
𝑝22
+ 6𝑝22 − (2 × 0.4721) + 2 = 0
2
2
𝑝22
+ 6𝑝22 − 2.9442 = 0
2

Solving the above equation yields, 𝑝22 = −12.4721 𝑎𝑛𝑑 0.4721. Since P is a positive matrix, choose
positive value for 𝑝22
𝑖. 𝑒., 𝑝22 = 0.4721
From (2)
𝑝12 𝑝22
−2𝑝22 + 𝑝11 − 3𝑝12 ++0=0
2
𝑝12 𝑝22 0.4721 × 0.4721
𝑝11 =− + 3𝑝12 + 2𝑝22 ; − + (3 × 0.4721) + (2 × 0.4721) = 2.4721
2 2
𝑝11 𝑝12 2.4721 0.4721
𝑃 = [𝑝 𝑝 ]= [ ]
12 22 0.4721 0.4721
The optimal control law 𝑢∗ = −𝐾𝑥(𝑡) = −𝑅 −1 𝐵𝑇 𝑃𝑥(𝑡)
1 2.4721 0.4721 𝑥1 (𝑡)
=− [2] [0 1] [ ][ ]
0.4721 0.4721 𝑥2 (𝑡)
2.2151 0.8284 𝑥1 (𝑡)
=− [0 1⁄2] [ ][ ]
0.8284 0.3947 𝑥2 (𝑡)
𝑥 (𝑡)
=− [0.2361 0.2361] [ 1 ]
𝑥2 (𝑡)
𝑢∗ = −0.2361𝑥1 (𝑡) − 0.2361𝑥2 (𝑡)

∴ 𝐾 = [0.2361 0.2361]

[Link]’s College of Engineering 124 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19

Unit V Optimal Estimation

Estimation deals with finding an approximate value (called an estimate) of a quantity from
observations or measurements which contain the information on the quantity to be estimated. All
measurements are prone to errors and therefore an estimate can be only as close as possible to the correct
value depending on the methods used for the estimation. Estimation is to be done generally for the purpose
of implementing a control of a physical system.

Figure shows the general problem of estimation. The physical system is subjected to two types
of inputs, a control input which can be easily manipulated and a disturbance input which accounts for the
presence of internal or external phenomena which cannot be easily determined. The disturbance may be
inherent in the system or may be due to the environment, such as noise in electronic circuits, signal
interference due to radiation, turbulence of aircraft etc., which occur in an unpredictable or random
manner. The system variables which are the outputs of the system are then measured by a measurement
process or system. The measuring system not being the perfect one introduces its own errors which may
again be unpredictable or random. Some of the errors in the measurement may be systematic and can be
corrected for, but many estimation problems deal with the measurement errors also as random.

The estimation problem here is to correctly calculate the true values of the system variables in the
presence of the disturbances and the errors introduced due to error in measurement. Often the time history
of the measurements is available and the estimates are to be obtained from these. Usually, a performance
measure is defined to assess the quality of the approximation or the estimate and the estimate is derived
that will maximize or minimize this performance measure. In such a case, the estimate is called an optimal
estimate. What is being done in an optimal estimation problem is to determine an algorithm by which the
estimate is obtained from the measurements based on the knowledge of the dynamics of the physical system
and whatever information one has on the disturbance process and the measurement process or-the errors
caused by these. In an associated control problem, the estimates obtained are used to control the system in
a desired manner.

A typical procedure for this can be summarized as follows.

1. Development of the models: This involves the specification of the models for (a) the physical

[Link]’s College of Engineering 125 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
system (b) the disturbance process (c) the measurement system and (d) the measurement error
process. The models developed should be sufficiently accurate or at least be adequate for the
purpose.
2. Specification of the performance measure: The aim or the purpose of the whole procedure is
defined here. Obviously the performance measure must be realistic in terms of the physical problem
being studied and mathematically solvable.
3. Problem formulation: Here the information available from the two steps above is combined along
with a set of constraints that are to be imposed to define the problem.
4. Development of estimation and control algorithm: This is the implementation stage of the
problem in the physical system to achieve the purpose for which the whole exercise is carried out.
The usefulness of the results also will have to be examined.

Application of Optimal Estimation

1. Communication systems: To extract a message from received signal. The transmitted or received
signal will contain the message to be extracted and will be contaminated by unwanted disturbance
and measurement errors in different stages of the Transmission and reception.
2. Navigation: In a typical navigation problem the position and velocity of a vehicle have to be
correctly estimated from the available measurement on the two. Aircraft, space crafts, surface ships
and submarines use these for their movement.
3. Post experimental data analysis: Here the recoded data from an experiment is analyzed in detail
to assess the success. For example tracking and telemetry data available from the launching of a
spacecraft provides valuable information for subsequent missions.
4. Process control: Successful operation of large chemical processes require regulation and control for the
purpose of maintaining the efficiency of operation, quality of products and attainment of other specified
goals. Specific examples are machine tool control, aircraft and space craft flight control and multistage
chemical processes.

Kalman Filter

Kalman filter is essentially a set of mathematical equations that implement a predictor-corrector

type estimator that is optimal in the sense that it minimizes the estimated error covariance—when some
presumed conditions are met.

Since the time of its introduction, the Kalman filter has been the subject of extensive research and
application, particularly in the area of autonomous or assisted navigation. This is likely due in large part to
advances in digital computing that made the use of the filter practical, but also to the relative simplicity
and robust nature of the filter itself. Rarely do the conditions necessary for optimality actually exist, and
yet the filter apparently works well for many applications in spite of this situation.

Applications of Kalman filter.

The applications of a Kalman filter are numerous;
 Tracking objects
 Fitting Bezier patches to (noisy, moving,..) point data
 Economics
 Navigation
 Many computer vision applications
 Stabilizing depth measurements
 Feature tracking
[Link]’s College of Engineering 126 VI Semester ICE
IC6601 Advanced Control System Department of ICE 2018-19
 Cluster tracking
 Fusing data from radar, laser scanner and stero cameras for depth and velocity measurement
 Many more
Advantages
1. Progressive method - No large matrices has to be inverted
2. Proper dealing with system noise
3. Track finding and track fitting
4. Detection of outliers
5. Merging track from different segments
Assumptions
1. Linear system
 System parameters are linear function of parameters at some previous time
 Measurements are linear function of parameters
2. White Gaussian noise
 White: uncorrelated in time
 Gaussian: noise amplitude

Linear Quadratic Gaussian (LQG)

Linear Quadratic Gaussian (LQG) control is a modern state space technique for designing optimal
dynamic regulators. It enables you to trade off regulation performance and control effort, and to take into
account process and measurement noise. Like pole placement, LQG design requires a state-space model of
the plant. The LQG controller design methodology is based on the Kalman filter who in 1960 published
his famous paper describing a recursive solution to the discrete-data linear filtering problem.

The LQG controller is simply the combination of a Kalman filter i.e. a linear-quadratic estimator
(LQE) with a linear-quadratic regulator (LQR). When we use the combination of an optimal estimator
(Kalman filter) and an optimal regulator (LQR) to design the controller, the compensator is called Linear
Quadratic Gaussian (LQG).

This regulator has state space equations

𝑑
𝑥̂ = [𝐴 − 𝐿𝐶 − (𝐵 − 𝐿𝐷)𝐾]𝑥̂ + 𝐿𝑦𝑣
𝑑𝑡
𝑢 = −𝐾𝑥 (1)

[Link]’s College of Engineering 127 VI Semester ICE

IC6601 Advanced Control System Department of ICE 2018-19
The goal is to regulate the output 𝑦 around zero. The plant is subject to disturbances and is driven
by controls. The regulator relies on the noisy measurements 𝑦𝑣 = 𝑦 + 𝑣 to generate these controls. The
plant state and measurement equations are of the form,
𝑥̇ = 𝐴𝑥 + 𝐵𝑢 + 𝐺𝑤
𝑦𝑣 = 𝐶𝑥 + 𝐷𝑢 + 𝐻𝑤 + 𝑣 (2)
and both 𝑤 and 𝑣 are modeled as white noise. The LQG regulator consists of an optimal state feedback
gain and a Kalman state estimator. These two components can be designed independently.

1. Optimal State-Feedback Gain

In LQG control, the regulation performance is measured by a quadratic performance criterion of the form,
∞
𝐽(𝑢) = ∫0 {𝑥 𝑇 𝑄 𝑥 + 2𝑥 𝑇 𝑁 𝑥 + 𝑢𝑇 𝑅 𝑢} (3)
The weighting matrices Q, N and R are user specified and define the trade-off between regulation
performance (how fast goes to zero) and control effort. The first design step seeks a state feedback law that
minimizes the cost function. This gain is called the LQ-optimal gain.

2. Kalman State Estimator

As for pole placement, the LQ-optimal state feedback u = −Kx is not implementable without full state
measurement. However, we can derive a state estimate x̂ such that u = −Kx̂ remains optimal for the output-
feedback problem. This state estimate is generated by the Kalman filter.
𝑑
𝑥̂ = [𝐴𝑥̂ + 𝐵𝑢 + 𝐿(𝑦𝑣 − 𝐶𝑥̂ − 𝐷𝑢)] (4)
𝑑𝑡
with inputs u (controls) and yv (measurements). The noise covariance data,
𝐸(𝑤𝑤 𝑇 ) = 𝑄𝑛 , 𝐸(𝑣𝑣 𝑇 ) = 𝑅𝑛 , 𝐸(𝑤𝑣 𝑇 ) = 𝑁𝑛 (5)
Determines the Kalman gain L, through an algebraic Riccati equation. The Kalman filter is an
optimal estimator when dealing with Gaussian white noise. Specifically, it minimizes the asymptotic
covariance of the estimation error x − x̂.
lim 𝐸((𝑥 − 𝑥̂)(𝑥 − 𝑥̂)𝑇 ) (6)
𝑡→∞
The goal is to regulate the plant output y around zero. The input disturbance d is low frequency
with power spectral density (PSD) concentrated below 10 rad/sec. For LQG design purposes, it is modeled
as white noise driving a low-pass filter with a cutoff at 10 rad/sec.
There is some measurement noise n, with noise intensity given by,
𝐸(𝑛2 ) = 0.01 (7)
Use the cost function,
∞
𝐽(𝑢) = ∫0 (10𝑦 2 + 𝑢2 )𝑑𝑡 (8)
to specify the trade-off between regulation performance and cost of control.

Drawbacks of LQG
LQG robustness is not guaranteed. It is thus it is important to look at the stability radius and spectral
values sets of the closed loop design. This is related to the interaction of the transients of the true state, the
controller action and the observer.

*********************ALL THE BEST ******************************

Prepared By,

[Link], Asso. Professor, Department of ICE, [Link]’s College of

Engineering, Chennai 119
[Link]’s College of Engineering 128 VI Semester ICE