State Space and

State Variables
1.1 INTRODUCTION
a can
A physical system which is to be controlled is called a plant. The dynamics of plant
t as independent variable. It can
be represented by ordinary differential equations with time
also be represented.by many first-order differential equations, and the number of first-order
differential equations depends on the order of the differential equation that describes the
system. A differential equation is said to be linear if each term of the equation contains at
can be
most only the first power of the dependent variable or its derivatives. A system that
described by a linear differential equation is said to be a linear system.
or a linear time-varying
A linear system can be either a linear time-invariant system
system. A linear time-invariant system has its parameters that do not change with
a
time. A differential equation that represents such system will have the
coefficient of
are time varying. For a linear time
each term of it as constant. Most physical systems
will be a function of
varying system, many terms of its differential equation description
time.
powers of the dependent variable or its
A
nonlinear differential equation contains higher
are classified as nonlinear
derivatives. Systems described by nonlinear differential equations
systems.

1.2 CLASSICAL VERSUS MODERN CONTROL THEORY

The analysis and designof linear controlsystems
are carried out using root locus plots, Bode
Nyquist plots. These plots are drawn based on the transfer functions of linear sys
plots and
tems. A transfer function establishes only the relationship between the input and the output.
Transfer functions are used to analyse a system by determining the output
response and error
on transfer
of the system for a set of given input test signals. This control system theory based
function is called classical or conventional control theory., The classical control theory is based
frequency domain approach.
2 Modern Control Theory

to all types of control sys.

This transfer function approach is not generally applicable
A few limitations are listed
tems and hence these methods suffer from certain limitations.
below: dif.
(1) Conventional control theory is based on transfer functions derived from linear
the analy.
ferential equations with constant coefficients. Hence, it is well suited for
sis and design of linear time-invariant single-input single-output control systems
output
and is not applicable to nonlinear, time-varying and multiple-input-multiple
systems.
are set to
(2) While deriving the transfer function, the initial conditions of the system
zero. Hence, in the methods based on transfer functions, initial conditions of a systen
cannot be included in the analysis and design.
(3) The transfer function approach does not give
any information about the internal varí
ables of the system.
(4) Frequency response methods such as Bode plots and Nyquist plots
cannot be applied
for the analysis and design of nonlinear and time varying systems.
may not yield
(5) The system design is carried out on trial-and-error procedures and
optimal control systems.
step, unit
(6) The error analysis is based on certain test signals like unit impulse, unit
ramp and unit parabolic inputs.
design
The state space technique is a time-domain approach applicable to the analysis and
on certain chosen intérnal variables
of both linear and nonlinear control systems. It is based
in a system. The control theory based on state space technique is also called modern
contrò!
theory. Following are major advantages of the state space methods:
(1) It is applicable to both linear and nonlinear systems and time-invariant and time
varying systems.
(2) Multiple-input multiple-output systems can be represented and analysed in
state
space.
(3) Initialconditions can be incorporated in the analysis and design.
a system,
(4) The state space approach is generally based on certain internalvariables in
and these variables, at any time, can be known.
can
(5) From the state model of a system, a discrete-time model can be obtained which
be analysed using a digital computer.

1.3 CONCEPT OF STATE AND STATE VARIABLES

Consider a linear time-invariant continuous time system with an input u() and output
response y() as shown in Figure 1.1. The output response y() is given by the convolution
integral,

yl) =| h( - 9)u(0) de nts e (11)

the
 State Space and State Variables 3

where h() is the impulse response of the system, that is, the response of the system for an
input of unit impulse signal.

Linear Time Invariant y()

Continuous Time System
Input Signal Output Response

Figure 1.1 Input-Output relationship in a linear time-invariant (LTI) system

Consider that the input is applied at timet= 0.Then equation (1.1) can be written as

-00

The first term on the right hand side is the response of the system at t=0, and this is the
an
initialcondition of the system. The second term represents the response of the system for
input u(). Hence, the output depends not only
on the input but also on the initialcondition.
Then vt) can be written as

= y(0)+[u-9)u() de (1.2)
)
a given input u(), if the initial condi
The output y() can be determined completely for
a state of the system at t=0. The state of
tion y(0) is known. In this equation, v(0) is
a system is the summary of the complete status of the
system at a particular instant of
a
a system can be described by the values of set
time. At any particular time, the state of are called
of variables of the system. These variables that represent the state of a system
state variables.
are required to represent the complete state of
There may be many state variables that
are the minimum set of variables that are
a system. The state variables of a dynamic system
necessary to specify the state of a system. Given the initial
states at t= to and the inputs for
at least
is possible to determine the system response completely for any timeth.If
2,it output of a system, then these n vari
n
variables are required to completely determine the
ables are called the state variables. State variables need
not be physical quantities. They need
not even be measurable.

1.4 THE STATEVECTOR

are necessary tocompletely determine the
Ifn state variables
x
(),x,(),x}(),----,) can be represented as a vector X(), and
state of a system response, then these n state ariables
this vector is called a state vecto.
4 Modern Control Theory

x,()
X(1) =
x()

1.5 THE STATE SPACE

The n-dimensional space whose co-ordinate axes are the .x-axis, X - axis, ---,
x, - axis
is called a state space. The variables x, (), x, (),x;((), -----,,
() are state variables. A state
a
of system can a
be represented by point in the state space.

1.6 STATE MODEL FOR LINEAR SYSTEMS

A state model for a dynamic system consists of state equations and output equations. The state
equations are first-order differential equations. In each state equation, the derivative of a state
variable is expressed as a linear combination of the state variables chosen for the system and the
system inputs. There will be as many first-order differential equations as there are state variables.
The number of state equations is equal to the order of the system. The output equation is a set of
equations for output variables which are linear combinations of the state variables and the inputs.
Consider a third-order linear time-invariant system, which is described by a third-order
differential equation. The number of state variables required is three, which is the order of the
system. Let the state variables be x(),x,() and x(t). Let there be two inputs () and r(0)
.
and two outputs y() and y, (1) A set of three first-order differential equations can be writ
ten in the form
dx, (1) S
(1.3)
dt
dx, (0) =
dt
a21 ()+ agg *, (t)+ dz3 x,(0) + b K) +byn h(4) (1.4)

dx, (1)
ay
=
()+ 442 X2()+ ag3 X,(1) + b n() +bg 5() e
(.5)
di

These equations are linearly independent. The state equations (1.3), (1.4) and (1.5) can be
written in the matrix form as
[dx, (0)
d
dx,(0)
x4(0) +| b ba (1.6)
d
by
3t
dx,(1)
dt
 State Space and State Variables 5

The output equations are

i)=41 ()+2 4(0) +43 x4()+d )+dg h(1) (1.7)

2(0)=21 x()+o2 x,()+% ()+ dy H()+ dg hl0) (1.8)

The output equations (1.7) and (1.8) can be expressed in the matrix form as

()1
|C21 C22 C23
x,() |+| (1.9)

The two equations (1.6) and (1.9) together are said to be the state model of the system.
For an n-th order system, the state equations are a set of n first-order differential equa
tions. For an n-th order system with p inputs and g outputs, the state equation in the matrix
form is
ds,
()
d
dx,() a2 a13...in
|4(0)]
dt x,(1)
dx,(1) a23..aan () (1.10)
d

dx, (0)
dt

The output can be expressed in the matrix form as,

dyg
---dip )
x() (1)
2() C21 C2 --2n |
+ (1.11)

In general, these are written as,

X()=A X(0) +B R() - state equation

Y() =C X()+D R() – output equation

6 Modern Control Theory

where X(:) represents derivative state variables,

of X() is the state vector and R() is the input
vector representing multiple inputs; Y() is the output vector.

n)
R():
() al)
and Y()=

A is called the system matrix, Bis the input matrix, Cis the output matrix and is called
D
the input-output coupling matrix or feed forward matrix. The coupling matrix represents direct
coupling between the input and the output. For an n-th order system with p
number of inputs
and q number of outputs, A is an n xn matrix, B is an nxp matrix, Cis a qxn
matrix and
Dis a qXp matrix.
A state equation and the output egquation together make the state
model for a system.
A state model is not unique for a given system. State model for a system
depends on the choice
of state variables. As such, there can be many state models for the same system.
The advantage
of first-order equations is that, in addition to the input-output characteristics, the
internal
characteristics of the system are represented by these equations.
For time-varying systems, a state equation will be a function of time,
the independent
variable. They are represented in terms of n state variables and pinputs as

x=S(,X--s,---p).
The general form of a state equation is

dt
dx,() 41() ay2() a13
()...a,()
dt
z1(!) ay2() ays()....a2, (0)
dr,(1) b,i(4) byg(t) bzs()---b,() ()
dt

dx, (0)
dt
 State Space and State Variables 7

X= A()X()+ B()
R().

The general form of an output equation is

0] i () 2
(0)---G() fd,() dyp()---d,(0) 0]
2(0) C1 (0) Cz2
() --- (0)|x() dai
() dygl)---d, ())

Y() = C(() X(1)+D(0) R().

1.7 THE STATE DIAGRAM

A pictorial representation of the state equation and the output equation is called the state
diagram and can be drawn from the state model:
=
X) A
X()+B R(t)

Y()=C X()+D R()

The state diagram is shown in Figure 1.2.

X(0) Y()
B
R(C)
|+
Integrators

Figure 1.2 The State Diagram

1.8 STATE MODEL FROM DIFFERENTIAL EQUATIONS

(A) Differential Equations without any Derivative of Input

The method of obtaining state modelfrom differential equations is explained by means

of some examples.
 Modern Control Theory

a system is
Example 1.1 The differential equation that describes linear time-invariant

d'y)y()48y{)=r().
di' dt
Obtain a state model for the system and draw the state diagram.

Solution: Choosea set of two state variables

and x, =x =y

Sothat =y: (A)

From the given differential equation,

di?
=-8y()-6r)
dt

y=x =-8x -6:x, +r(). (B)

Equations (A) and (B) are state equations, and in matrix form, they can be written as

The output equation is

The state diagram is shown in Figure E1.1.

Figure E 1.1 State diagram for Example 1.1

 State Space and State Variables 9

Example 1.2 A third-order LTIsystem is represented by the differential equation

di
+6d04110+6y() =3r().
d? dt
Represent the system in a state space model and draw the state diagram.

Solution: Choose a set of three state variables

So that (A)

and (B)

From the given differential equation,

d'y) =-6y)-11906yO.+3r(t)
di' d di2

y =X =-6x-1lx,-6x, +3r()
Equations (A), (B) and (C) are state equations, and in matrix form, they can be written as

X
|-6 -11-6||

The output equation is

y=[ 0

The state diagram is shown in Figure E 1.2.

10 Modern Control Theory

Figure E 1.2 State diagram for Example 1.2

(B) Differential Equations Involving Derivatives of Input

A derivative of input cannot be a part of the state equation. A method to avoid a term with
a derivative of input in the state equation is to choose the state variables a
in different form.
Considera second-order system represented by the differential equation

y+4 +a0) =b, r+b, r+br.t s (1.12)

Choosea set of two state variables as

=y-kor (1.13)

=X-kr. (1.14)

From equation (1.14),

r.
=+k (A)
From equation (1.13),

y=x+k,r. (1.15)
Differentiating equation (1.15),

(1.16)
Combining equations (A) and (1.16).

y=(4+khr)+kyt. (1.17)

Differentiating equation (1.17).

(1.18)
 State Space and State Variables 11

Substituting equations (1.15), (1.17) and (1.18) in the given differential equation,

(g+ k r+ ky r)+a (ig +k, r+kg ) + 4, ( +ko r) =by r+ b, r+by r

xy+ a, xy + a, x = (by-ko)r+(b-k, -ak )r+ (b -ayk, -akydr. (1.19)

For the state equation to be independent of derivatives of input,

(b, -ko)=0 or ko =b (1.20)

(b,-k-ako)=0 or k, =b-ako (1.21)

k, = (bo -ak -a,k,) (1.22)

From equation (1.19),

(B)

Equations (A) and (B) are state equations and they can be kept in the matrix form as

The output equation is

y= +kr=[1

Example 1.3
Represent the system in the state space mnodel, the differential equation description of which is
y+sy+4y=3r(0) +2r() .Choose the state variables as x = y-kyr and x, =,-kr.
Solution:

(2)
12 Modern Control Theory

From equation (2).

(A)

From equation (1),

y= +kor.
X,
)
Differentiating this equation,

y=x+kyt. (4)

Combining equations (A) and (4),

y=(y +khr) +kyt. 6)

Differentiating this equation,

y=n+hr+yt. (6)
Substituting equations (3), (5) and (6) ín the given differentíal equation,

y+hr+k r+5(%, +khr+ky )+ 45, +ky) =3r+2r

=4x4-5x -ky r+(2-kh-5k4)r+(3-5k -4ky)r. (7)

For the state equation to be independent of derivatives

of input,
ko =0.

(2-k-5ky)=0 or k =2
ky =
(3-5k -4ky) =-1.

Equation (7) becomes

y=4x,-5x-Tr. (B)
Equations (A)and (B) are state equations
and they can be kept in the matrix forn as
 State Space and State Variables 13

The output equation is

y'=+k r= =|1

Example 1.4 Represent the system in the state space model, the differential equation descrip
tion of vwhich is

y+6y+5y+4y = 3r()+2 r(t) +

r.
Choose the state variables as

j=y-kor xy=x-k
-||564
r and S
x =X-k,r.
Solution:
=y-kr (1)

(2)

and x = X-khr (3)

From equation (2),

x=x +k r (A)

This is a state equation.

From equation (3),

=Xy tk, r (B)

This is also a state equation.

From equation (1),
y=x +k r. (4)

Differentiating this equation,

y=x+k, r=x +kh r+ko

.
Differentiating this equation,

y= tkr+k, r+kyt. (6)

14 Modern Control Thcory

Differentiating this cquation again,

(7)

Substituting cquations (4),(5), (6) and (7) in the given differential equation,

r+ ko r)+5(x, +ky r+kyr)+44x +kyr)

(tk, r+k, r+ko r)+6(x, +k, r+k,
=3r()+2(0)+r.
- 6x,
y
=-4-5x, +(3-4kok, -6k,)r +(2-Sk, - 6ky-k,)rho
+(1-6ky -k,) r-ko (8)

For the state equation to be independent of the derivatives of

ínput r, the coefficients of r,
and r in equation (8) are to be zero.

Hence ka =0 (1-6k, -k,) =0 or k=1.

2-5k-6k -k, =0 or ky
=4.
ky =3-4ko -5k -6k,=22.

From equation (8),

xy .
=-4x,-5x, -6x, +22r (C)

Keeping equations A, B
and C in matrix form,

0
-5

The output equation is

y=x +kor= X
=[1 0
 State Space and State Variables 15

49 STATE MODEL FROM A STATE DIAGRAM

For a system, state cquationscan bc written down from a given state diagram.The method is
to choose the output of every integrator as a state variable. The input toan integrator is the
derivative of a state variable, and its equationcan be written down from the state diagram.

Example 1.5 A state diagram for a dynamicsystem is shown in Figure E 1.5. Write down the
state equations and the output equation.

Figure E 1.5 for Example 1.5

Solution:There are two integrators and hence two state variables. The output of each integra
tor is taken as a state variable. The state variables and their derivatives are marked in the state
diagram given in Figure E 1.5(a).

Figure E 1.5 (a) for Example E 1.5

From the state diagram,

(A)

and x =-2.x, +r(). B)

Keeping equations (A)and (B) in matrix form,

The output equation is

16 Modern Control Theory

Example 1.6 A
state diagram for a dynamic systemis shown in Figure E 1.6. Write doOWn
state equations and the outputequation. the

n)

Figure E 1.6 for Example 1.6

Solution: The output of each integrator is taken to be a state

variable. The state variables and
their derivatives are marked in the state diagram given in Figure 1.6(a).
E

r)

Figure E 1.6 (a) for Example 1.6

32agih ka so
From the state diagram,

=-2x +x, +r() (A)

and x, =-+r(). (B)

Equations (A) and (B) are state equations and they can
be kept in the matrix form,

The output equation is

Example 1.7 A state diagram of a dynamic system is shown in

Figure E 1.7. Write down the
state equations and the output equation.

y()

Figure E 1.7 for Example E 1.7

 State Space and State Variables 17

Solution: The outputof each integrator is taken tobe a state variable. The state variables and
their derivatives are marked in the state diagram given in Figure E 1.7(a).

Figure E 1.7 (a) for Example E 1.7

From the state diagram,

(1)

xy
=-2x + 2x, + x (2)

Xy =-2x, +r(). (3)

Combining equations (2) and(3).so alot a b

d ts esetSes
x, =-2x,+ 2.r -2x, +r()=-2x, +r(). (4)

Combining equations (1) and (4),

X=-
+
-2x, tr()=--+r(). (5)

Equations (3), (4) and (5) are state equations and can be represented in matrix form,a

-1 -1 0|
0 -2
0
0 -2|x3

The output equation is

o) =, =|!
 18 Modern Control Theory

MODEL FOR NONLINEAR SYSTEMS

1.10 STATE
systems generally contain higher powers of derivative
the output equation cr.
Differentialequations for nonlinear
For such a system, the state equation and
products of derivatives. a time-invariant system describedt
an example, consider nonlinear
be kept in matrix form. As
2

+a
d'y d di
*y=0.
di di' di
This can be written as
y+a)6)+by+c(y +y=0.
state equations
Let x =J, x = x = y
and x =,= y
bethree state variables. Then the two
are

and x, 3.
as
The thirdstate equation can be obtained from the giver differential equation

-ady dy
di d dt

Or
Itcan be observed that these equations cannot be kept in matrix form. The general form of
the state and output equations are
X=f(X0),R(0)

and Y(¢) s(K(0), R().

For a nonlinear time-varying system, the state and the output equations are also functions of
tíme. The general form is

X=f(X(), R(),)

and Y) =g(X(0), R().1).

Example 1.8 A dynamic system is represented by the differential equation

d,,(+6y)y
di? dt
=0
Represent the system in a state space model.

Solution: Let x
=y and x, =y=x be the two state variables.

(A)
From the given differentialequation

d'y -(1+8'y')y
di? dt

y=-ay-(|+b°y')y =-ay-y -by

This can be written as

(B)

Equations (A) and (B) are state equations but they cannot be kept in matrix form.

1.11 LINEARIZATION
A nonlinear system can be linearized about its operating state. In order to obtain a linear
mathematical model for a nonlinear system, it must be assumed that the variables deviate only
slightly froma given state. The method of linearization is based on Taylor's series. Consider a
function y= f(r).If the operating point is(,Jo), using Taylor's seriest
f()= f(%)+f (%) (x-)+) (x-)+---
When x is very close to x, or in other words if the variation (x -X) is very small, higher order
termscan be neglected so that as an approximation

f()= f()+f (*%)

(*- x)

or y=yo +f(0) (x-)

This shows that (y-yo) isproportional to (x-x,).This equation gives a linear mathematical
model for a nonlinear system.

Linearization with Two Variables

Let z=f (x, y) be a nonlinear equation where x and y are two variables. Let (x0Jo) be an
operatíng point. Using Taylor's series expansion and neglecting higher degree terms
 20 Modern Control Theory

(a-k) (r-) (y-)

(2-) (-)+k,(y- )
k,

Let
Hence

(1.23)

where

Equation (1.23) represents a lincar model for the

-
nonlincar system:=f(, ), about the point
(oJo).The variables in this linearized model do not represcnt the actualvalues but the de.
viations from the values at the steady state
operating point.

Linearization of State Equations

Consider the state equations for a time-invariant
second-order system.

=(y) and x =/i,4)

Let (x (0), x,(0)} be an operating point.

sd n
(1.24)

Gx, =
where N-x (0)
Ox
=N -x(0) &x,
=-(0).
Equation (1.24) is a set of lincarized state equations
in terms of small variations in state
ables x, and x,. vari

Jty0).4,0)) = ka
(1.25)

ON2
J0), v2(0))
 State Space and State Variables 21

Using equation (1.25), the Eigen values of the linearized state model can be determined by
solving the equation
A-k -k| =0

Example 1.9 Lincarize the cquation y+ 2y' -2 =0 about the point y= 0.5.

Solution: Replace the nonlinear term bya variable z. Let z = 2y3

The linearized model for the system is

(z-z)= (y-%) (1)

d' ly=0.5
= = 2(0.5) = 0.25
Zo
2(o )

=1.5.
dy ly=0.5
Substituting these values in equation (1),

(2-0.25) =1.5(y-0.5) or z =1.5y-0.5.

The given nonlinear equation can be linearized about the point y
=0.5 as

y+1.5y-0.5-2 =0 or y+1.5y -2.5 =0.

Example 1.10 Linearize the differential equation x+2x

x+3(x) +x=0 about the operat
ing point (x,x) = (0.5,1).

Solution: x+2' x+3) +x=0. (1)

Let y=x.The given differential equation is

y+2(*)'y+3(y) +x=0
The operating state is(%0,o) (0.5,1)
Let the nonlinear term be written as

z=2xx+ 3(.
= 2(x)° y + 3(9)° Zy = 2(0.5)'l+3(1) =3.5.

=4xyo5) =2 - 2(1)°
Oxo5,1) +6|,l(0.s,1) =6.5.
 22 Modern Control Theory

Z-ž0 = 2(.r-)+6.5(- o)

(2-3.5) = 2(r- 0.5) + 6.5() -1).

Or z=2r+6.5y -4. (4)

Substituting equation (2) in equation (1),

x+[2x+6y-4]+x =0

+6X+3x-4=0.
Example 1.11 A nonlinear equation is represented by the differential equation

x+2r+ 3(x) +x=0.

Obtain a state model for the system about the operating point (x,x)= (0.5,1).

Solution: x+2xx+3(x)+x=0
Let x, =x.
x=* and
The given differential equation now becomes

x,+ 2(%,)x, +3(,)

+=0,,U sti.

The operating state is

The state equations are

=,=(,)
(2)

Ox
o.5,1)
=(4x, -1)051 =-3
l(o.s, 1)

x(0) = (,20)=
=see
N20
t
(0) = f(xjosX9) = -2(0.5)²1-3(0)²-0.5 = -4.
 State Space and State Variables 23

The lineatized state cquations are

-(0)
|-(0)

The Eigen valucs of the linearized model can be determined as

-1
1+6.5
2'+6.51+3=0.
3

The Eigcn valucs arc l=-0.5 and h= -6.

1.12STATE VARIABLES FOR PHYSICAL SYSTEMS

Electrical System
The Capacitor

ic) +

Figure 1.3 State variable for a capacitor

Consider a capacitor as shown in Figure 1.3. The current through the capacitor is

ic) =C dv)
di
The voltage across the capacitor is

The first term on the right hand side is the initial voltage across
the capacitor at =0.It indi
cates the state of the capacitor at
(= 0.
 24 Modern ControlTheory

v)=ve(0)+jeod. (1.26)

Comparing equations (1.2) and (1.26), it can be concluded that the capacitor
voltage can
taken as a state variable. Since the charge on a capacitor is given by q(1) = be
Cve(),thecharge
on the capacitor g() can also be taken as a state variable.

The Inductor

iz ()

Figure 1.4 State variable for an inductor

Consider an inductor as shown in Figure 1.4.The voltage across the inductor is

(")
V,
() =L di,dt
The curTent through the inductor is

-00

-00

The first term on the right hand side is the initial current through the inductor at =0.
indicates the state of the inductor at
t=0.
(1.27)
i0-i,(0+,9d. 0

Comparing equations (1.2) and (1.27), it can be concluded that the inductor current c
linkage
taken as state variable. Since the electromagnetic flux linkage is v() = Li,(),tthe flux
can also be taken as a state variable.
are
Note that inductors variables
and capacitors are energy-storing elements. State
associated with energy-storing, elements. A as heat. It
resistor absorbs energy and dissipates it
Cannot store energy. Hencc, nostate variable is associated with a resistor
 Central
NIT, LIbrary
SIlchar
State Space and State Variables 25

Mechanical Translational System

Just as a resistor, a capacitor and an inductor form the elements of n basic clectrical circuit,
mass, lincar spring and a damper form the clements of a basic mechanical translational system.

The Mass

M -fu)

Figurc 1.5 State variable for a mass M

Consider a mass M moving on a plane, as shown in Figure 1.5. The cxternally

applied force
that causes the motion is f(). The reactive force is (), which is the force due to accel
eration of the mass. If () is the displacement of mass and
u() is the velocity of mass, the
dynamic equation of motion is

d² dt
Thevelocity of mass can be found out as

M -00

The first term on the right

hand side is the initial velocity of mass at t =0, It indicates the state
of the mass at t =0.

u(1) = u(0)+
(1.28)
0
Comparing equations (1.2) and (1.28), it can
be concluded that the velocity of nmass can
taken as a state variable. be
Since u(t)=), d
the displacement of mass can also be taken asa state
dt variable.
The Spring

Figure 1.6 State variable

for a linear spring
26 Modern Control Theory

Considera mechanical spring, which is considered to be linear, as shown in Figure 1.6 T

externally applied force is f((). The force
of reaction of the spring is proportional to th
displacement and the constant of proportionality is
the spring constant K. If Sk)is
the
reactive force,

The first termon the right hand side is the initial

value of the spring force at t =0. indicates
the state of the spring at =0. It

Se)=fa0)+Ku()d.
0
Comparing equations (1.2) and (1.29), can
as state variable. Since it be concluded that the spring force can
be taken
the spring force is proportional to
spring can also be taken as a state variable. the displacement, displacement of
Amass
in motion stores kineticenergy,
potentialenergy. Thus, state variables are anda spring, stretched or compressed, stores
associated with a mass and a spring.
damper or a dash pot absorbs energy A mechanical
no state variable is associated and dissipates it as heat. It cannot store energy. Hence,
with a dashpot.

Mechanical Rotational System

Rotating Shaft

T() 0()
Figure 1.7 State Variable for
the inertia of shaft
Consider a rotating shaft with moment
torque to the shaft and of inertia J, shown in Figure 1.7.
(1) is the angular displacement, then the torque If T() is the input
of the shaft T, (t) is due to acceleratioM

T) =,(0)=J0)
di2
do()
dt
where a) 401) is the angular speed
dt of rotation,
 State Space and State Variables 27

sO)-4jroajroga.
The term on the right hand side is the initial value of the angular spcedat (=), It indicates
the state of theshaft at ( 0,

= (1.30)
i)

Comparingequation (1.2) and(1.30), it can be understood that the angular speed of the rotat
ing shaft can be taken as a state variable. Sincc oX) =the angular displacement of shaft
can also be taken as a state variable.

Torsion Spring
Consider a torsion spring shown in Figure 1.8. The externally applied torque is T(). The re
action torque T() of the spring is equal to and opposite of the applied torque.i;ie

Figure 1.8 State variable for a torsion spríng

T() =Tz(0) = K0:) = K o)

The first term on the right hand side is the initial value of the spring reaction torque at = 0.
It indicates the state of the spring at = 0.

T(0-T,(0)+Kfa) (1.31)

Comparing equation (1.2) and (1.31), it can be concluded that for a torsion spring, the spring
torque can be taken as a state variable. Since the spríng torque is proportional to the angular
displacement, the angular displacement can also be taken as astate variable.
 28 Modern Control Theoryte

A rotating mass stores kinetic energy, and torsion spring stores potentialenergy. A damn
or a dash pot in a rotating system absorbs energy and dissipates it as heat. It cannot ste
energy. Hence, no state variable is associated witha damper.
The variables for electrical and mechanicalsystems are listed below in Table 1.1.

System State Variables

Electrical Inductor current, magnetic flux linkage

Capacitor voltage, capacitor charge

Mechanical (translational) Displacement of mass, velocity of mass

Displacement of spring, spring force

Mechanical (rotational) Angular displacement, angular velocity

Reaction torque of torsion spring

Table 1.1 State Variables for Different Systems

1.13 CHOICE OF STATE VARIABLES

The number of state variables required for a system is equal to
the order of the system, that is
the order of the differential equation that describes the system.
The choice of state variables
for a system is not unique. Many sets of state variables may
be chosen, each containing the
minimum number required, and each set results in a different state
model.
(1) For an electrical network, inductor currents and
capacitor voltages can be chosen as
state variables.
(2) For mechanical translational system,
displacement and velocity of mass, displacement
of spring and spring force qualify to be state variables.
(3) For mechanical rotational system,
angular displacement and angular velocity of
rotating shaft and the angular displacement and the spring torque can
be taken as
state variables.
(4) Asa general guidance, after writing
simple derivative equation for each energy stor
ing element, each differentiated variable can
be chosen as a state variable.
(5) The state variables chosen must
be linearly independent.

1.14 PHYSICAL VARIABLE MODEL AND

PHASE VARIABLE MODEL
Inductor current, capacitor voltage, mechanical
displacement, velocity and spring force are
physical variables. If some of these are
selected as state variables, then the resulting state
model is called a physical variable model.
For a physical system, consider that a
particular quantity is chosen as a state variable
labelled as x,. If the successive derivatives
this variable are also taken as state variables,ad
state model thus obtained is said to be of tne
in phase variable form. For example,
order system. Three state variables are necessary to consider a thir
a
obtain state model, For this systet,
 State Space and State Variables 29

=
consider that the variable x is chosen as a state variable. Then x, = and x,are also
dt dt
taken as state variables required to represent the system. If r,() andr,(t) are the two inputsto
the system, then the threc state cquations can be written in the form,

dxy
dt
dxy
dt
This set of equations are said to be in phase variable form. These equations are denoted in
matrix form as

dt 0 1 0
dx,() 0
1
X() +|0
dt
dx()
dt

Example 1.12 An electrical network is shown in Figure E 1.12. Select a set of proper state
variables and write down a state equation, in physical variable form, to represent the system.

WmR
s Cm vo

Figure E 1.12

Vs(t)
Wmi R
L
+

Figure E 1.12(a)

Solution There are two energy-storing elements, the inductor and the capacitor. The order
of the system is two and two state variables are required to get a state model. The current
through the inductor and the voltage across the capacitor are taken as state variables. Writing
the KVL equation, with respect to Figure E 1.12(a),
 30 Modern Control Theory

s) =Ri,()+z20
dt
+ ve()

Or
dOi0-e (0+t0). (A)

The capacitor current is the same as the inductor current and is given by,

d
=i() (2)

dvc
Or ()1
dt (B)

Equations (A)and (B) are state equationsand can be represented in the matrix form as.

di,(C) R
dt
= State equation
dvc (0)
dt

The output equation is

Example 1.13 For the network of Example 1.12,

obtain a state model in phase variable form.
Select the capacitor voltage as a state variable.

Solution Thenetwork is shown

in Figure E 1.13.

L
Vs()
Vo

Figure E 1.13 Example 1.13

The capacitor voltage x =v.(0) and x,

=C dt
are taken as the two state variables

(A)

Writing the KVL equation,

 State Space and State Variables 31

+
s() =Riy () L0+
dt
e(). (1)

The inductor current is the same as the capacitor current.

=Cx) (2)
dt

di()Cx
dt
(3)

From equations (1). (2) and (3). vg(0) = R

C+ LCx+ 5
1
R 1

L
X
LC 's(). (B)
Keeping equations (A) and (B) in the matrix form, the state model is

0
- Vs().

Taking capacitor voltage as output,

vo (()
=Vc(.)=1

Example 1.14 For the network of Example 1.12, obtain a state model in phase variable form,
selecting the inductor current as a state variable.

Solution: The network is shown in Figure E 1.14.

R
Vs(t)

Figure E1.14 Example 1.14

Let x=,()di and the inductor current x, =i (0) be two state variables. Then

x =i,() =X (A)
32 Modern Control Theory

Writing the KVL equation,

= +
sl) Ri, L,d
dt

dt (2

From equation (1),

"s()= Rx, +
Lt L

Or R
(B)

Expressing equations (A) and (B) in the matrix form,

the state equation is
1

If the capacitor voltage is the output, the output equation

is

or

Example 1.15 An electrical network is

shown in Figure El.15. Select a set proper state var
ables and write down state equations, of
in physical variable form, to represent
the system.

R1
WWm L
+ R Wm R
Vs(t)
Vo()
+ RiL
Vs(t)
|+
o()

Figure E1.15 Example

1.15a Figure E1.15(a) Example 1.15

Solution: The current through the inductor

i, (0) and the voltage cross capacitor ve()21
taken as state variables.
Writing the KVL equation, with respect
to FigureE 1.15(a)
 State Space and State Variables 33

Ri, () + 44()
= L
Vs (0) + V().
di

Or di,0_,0)-c0,).
d L
1

(A)

The capacitor current is given by,

dt
=i,()-c)
R
dv (() 1

Or (B)
dt R,C
Equations (A) and (B) are state equations and can be represented in the matrix form as,

di, () | -15645 1

dt
dv, (") 1

R,C
dt
The output equation is

vo() =v¢0=[0

Example 1.16 For the network of Example 1.15, obtain a state model in phase variable form,
selecting the capacitor voltage as a state variable.

are
Solution: The network is shown in Figure E1.16.The state variables selected

x=Vc() and x, Ve().

=

Hence x=X.
+ R
R
vs(t) volt)

Figure E 1.16 for Example 1.16 tsag

Writing the KVL equation, with respect to Figure E 1.16,

"s)= Ri () + L
dt
+
V) (1

L()=cWcQ,+ ()
dt R, (2

Differentiating equation (2),

(3)
dt di' R dt
Substituting equations (2) and (3) in equation (1),

"c). (4)

Rearranging equation (4),

(R + Ry
1

R,LC)" LG's(). (B)

Keeping equations (A)and (B) in the

matrix form, the state equation is
0 1
Vc()
(R+ R,) (RR,C+L)
Vs().
R,LC RLC LLC
The output equation is

e)]
Example 1.17 Represent
the network shown in
the state model. Figure E 1,17in the physical
variable tormno

Ve2
R,
Rg +
vgt) C R
Vo(t) i R
vs(0)C
Figure E 1.17 for
Example 1.17
Figure E 1.17(a) for Example 1.17
 State Space and State Variables 35

Solution: Select the inductor current i, and the capacitor voltages v,y and v ,
shown in
Figure E 1.17(a), as state variables. Writing the KVL equation,

"y() = R i, + L (1)
dt

From equation
(1),i-ts).
dt L
(A)

The inductor current = V1-Vc2,

i, CVt: (2)
R

From equation (2), Vca

C i-RG'atG:
R,C R,C
(B)

The current through the capacitor C, is

e
(3)
R
1 1
From equation (3), Yc2 (C)

Equations (A), (B) and (C) are state equations and keeping them in the matrix form,toct

0
L
1 1 1
L
VCi
C R,C R,C
Vc+os().
1 1 'ca
R,C,
If the capacitor voltage v, is taken as the output

= [0 0
Vo) =Vez()

Example 1.18 An electrical network is shown in Figure El1.18. Represent it

in a state model
in physical variable form.
 36 Modern Control Theory

+ L
L + i 4+
R, C R, v) vct) R2
W

Figure E 1.18 for Example 1.18 Figure E 1.18 (a) for Example 1.18

Solution: There are two energy-storing elements, and hence two state variables are to b
selected. Choose the current through the inductor
i, and the voltage across the capacitor y.
shown in Figure E 1.18 (a). Writing down the KCL equation,

-+i,. (1)
R dt

From equation (1),

dt (A)

The capacitor voltage is v = L diL+

dt
RiL: (3)

di
From equation (2),
dt
Equations (A) and (B) are state
L i+c
equations and can be written
in the matrix form asip
1

di RC
di, (0) 1

dt L
If the voltage across the
resistor R, is taken as
output, then the output
equation is
vo(/)
From equation (3), =i, R (3)

Example 1.19
For the network
form. Select of
the current through Example 1.18, obtain a state
Solution: The the inductor as a model in its phase variable
network is re-drawn state variable.
as shown in
igure E 1.19.
 State Space and State Variables 37

W + +
R
Civc) R2 Vol)

Figure E 1.19 Example 1.19

Select .x =i, and .x =x = d. as two state variables.

d
(A)
Hence x =X,.

Writing KCL equation,

(1)
dt

The capacitor voltage v = L+R,i,

dt
=Lx +R, x. (2)

Differentiating equation (2),

V =Lx+R x
=
Lxy+R, x. (3)

Substituting equations (2) and (3) in equation (1),

1
4)+C(Lxy+R, xy)+i,.
isl)=Lg+Ry
o
(4)

Re-arranging equation (4),

(R
RLC
+R)._(L+Rgyt).
-x
RLC
1

L
(B)

can be expressed in the matrix form as

Equations (A)and (B) are state equations and
1

The output equation is v,(/) =, , =x R, =|R,

38 Modern Control Theory

for the clectrical network slhown

Example 1.20 Write down the necessary equations as the state variables, A
E
1.20 and represent it
in a state model.
Use
the pysical variables
draw the corresponding state diagram.

WR L W
v(t) i(t) L

Figure E 1.20 for Example 1.20

Solution: The network is drawn in Figure El.20 (a) showing the inductor currents and th

capacitor voltage. Let the state variables be.x =ipl Xy = V: and x, =):
L

W R1
W
Vs(t) i(t) vo()

Figure E 1.20 (a) for Example 1.20

Writing the KVL equation,

vs() = Ri) +L4

dt
From equation (1), 4 1 1

=--t;(). (A

Writing the KCL equation,

i +ig() =Cv¢+iL2•
Or
+is()=Cx,+*.
From equation (2), , 1 1

s). (B

Thecapacitor voltage v: = di2

R,L2
tL dt

Or
(3
 State Space and State Variables 39

R
From equation(3), X}
=) -X. (C)

Eauations (A). (B) and (C) arcstatc cquationsand kecping them in

the matrix form,

L
If the capacitor voltagc is takcn asoutput, then the output
equation is

Vo(1) = v(0) =
=|0 1

The state diagram is shown in Figure 1.20 (b).

l)

Figure E1.20 (b) State diagram for Example E 1.20

Example 1.21 Realize a state model for the
mechanical system shown in Figure El.21.

M y()
y()

B,
3K2

Figure E
1.21 for Example
1.21Figure E 1.21 (a) for Example 1.21
 40 Modem Control Theory

Solution: For the econvenience of analysis, an equivalent mechanical diagramn

shown in Figure E l.21(a).Fron this diagram, the cquation of motion is
is drawn

f)= M y) +(B,+ B,)j(t0) +(K,+ K,)y).

The variablesyand y are, respeetively, the displacement and the velocity of mass, and y.

are selected asthe two state variables.

N=)' and x,
= y

so
that =

Equation () can be written as

=
f) Mx,+ (B +B,)x, +(K,+K,)x. ()

Rearranging equation (2),

(Ky+K (B+B,) 1

M M * ). (B

Equations (A) and (B) are state equations and keeping them in matrix form,

(B +B,)
M M
If the velocity of mass is taken as the output, then the output
equation is

v,() = y = x, =0

Example 1.22 Obtain a state model for an armature-controlled

DC motor and draw the state
diagram.

Solution: The schematic diagram of a DC motor

driving a shaft is shown in Figure E 1.22.

+
W
R
Constant

Field
v)
)shaft

T(), 0
Figure E 1.22 for Example E
1.22x l I3
 State Space and State Variables 41

Solution For an clcctromcchanical systcm, a set of three cquations are to be written: one for
the clcctrical circuit, another for the mcchanicalsystem and the third equation relating these
(wo.
R - Resistance of armaturc circuit, Q
L-Inductance armaturc circuit, H
of

i()- Instantancous armaturc currcnt, A

v.()-Applicd voltage across the armature, V
c, - Back emf of the moto, V
1,-Constant ficld current, A
J- Momcnt of incrtia of rotating parts, Kg-m²
B-Cocficicnt of viscous friction, N-sec/radian
0-Angular displaccmcnt of the shaft, radians
T, - Externalload torque on the motor, N.
Thecquation for the clectrical circuit (armature circuit) is

v,() = R, i,()+L, di,() -te,(). (1)

dt

The back emf is proportional to the angular speed and is given by

e, () = k,: (2)
dt

where k, is a constant
of proportionality.
Combining equations (1)and (2),

v„() = R, ,()+L, di,() +ky de (3)

dt dt

The angular velocity is

de (4)
rad/sec.
dt

Hence, equation (3) can be written as

di,).+k, 0. (5)
dt
The torque developed by the motor has to overcome the torque due to angular acceleration,
the torque lost to overcome the viscous friction and the externally applied load torque. Hence,
the equation for the mechanical system is

T() =J+-+B+TL
dt
=J+Bo+T,.
dt
(6)
 to the armature current when the field f..
The torque developed bythe motor is proportional
is constant. Hence,

T() = k, i, (1), (7)

where k, is a constant.
Combining equations (6) and (7),

k, i,(1) = do
J+Bo+T.
dt
(8)

Select the armature current i,(0) and the angular speed o as state variables. From
equation (5),

di,()K,()-
di L
-wt().
Lu L,
(A)

From equation (8),

dw
(B)
dt JJotoa
Equations (A) and (B) are state equations and can be expresed in the matrix form as

di, ()1R kh
dt
do
dt

If the angular speed is taken as output

o
() = 0=[o
The state diagram is shown in Figure E 1.22(a).

T
dik) do
dt i)
dt
J

kb
L
Figure E 1.22 (a) State model for an
armature-controlled DC motor
 State Space and State Variables 43

Example 1.23 Represent a licld-controlled DC motor in a state space model.

Solutio: The sclhcmatic dingram of a ficld-controllcd DC motor is shown in Figure E 1.23.

Ficld Constant I,
WWm
L
Rr
.shaf?

T), 0

Figure E 1.23 for Example E 1.23

Equation for the ficld circuit is

di (")
R, i,(0+L -= y(). (1)
dt
The clectromagnetic torque developed by the motor is proportional tothe field current when
the armature current is constant, Hence,

T() =kri;(). (2)

The torque developed by the motor has to overcome the torque due to acceleration, the torque
required to overcome the frictional torque and to counterbalance the externally applied load
torque on the motor shaft. Thus,

-BO -+T,
T() =J+
d' dt
=J+Bo+T,, dt
(3)

do
where isthe angular speed of rotation.
w

di
Combining equations (2) and (3),

J+dt Bo+T, =k,i, 0). (4)

Select the field current , () and the angular speed of the shaft as two state variables.
From equation (1),

(A)
dt
 44 Modern Control Theory

From equation (4).

da k
dt

Equations (A)and (B) are state equations and they can be expressed in the matrix formm ae

di,()] R
dt
do
dt
If the angular speed of the motor is taken as output, then the outputequation is

= @=[0
)
The state diagram is shown in Figure El.23(a).

TL
di) do
v) dt dt

R
Figure E1.23 (a) State model for a field-controlled DCmotor gt

1.15 NON-UNIQUENESS OF A SET OF STATE VARIABLES

For a given system, there can be a number of sets of state variables., many
In of the earit
examples, it was shown that a system can be represented a state
in model either in terms
physical variables or in terms of phase variables. There can be
different sets of physical vat
ables andso also different sets of phase variables. These
variables need not have physie
existence. For example, in an electricalsystem,
if i, is an inductor current and v is a capae
voltage, then x=i,, + v can be a state variable. Clearly, this variable
has no physical entity,
general, if X1X2,X3,---,Xn are state variables,
then z,z,Z3,--,Zn given byY
 State Space and State Variables 45

can also be selected as state variables to represent the same system; but they must be lincarly
independent. The selection of a set of state variables affects the analytical and computational
effort and complexity. A set of state variables even if they are not mcasurable or not physical
variables may greatly reduce the computational effort.

SUMMARY
The conventional approach to the analysis and design are by using root locus, Bode plots
are not
and Nyquist plots. They are applicable only to linear time-invariant systems and
applicable to nonlinear and time-varying systems.
The conventional method is based on open-1loop and closed-loop transfer functions and
does not throw any light on the internal variables of the system.
Transfer function approach is applicable only to single-input single-output systems.
a method.
State space method of analysis and design of a control system is time-domain
systems and mul
State space method is also applicable to nonlinear systems, time-varying
tiple-inputmultipleoutput systems.
a a
A set of state variables of a dynamic system is set of minimum
number of variables
with which the system can be modelled and analysed completely.
a to the order of the
The number of state variables required to represent system is equal
system.
in terms of the state
In a state equation, the first derivative of state variable is expressed
a
variables selected and the inputs to the system.
The state equations are linearly independent.
state variable selected and
Output equation expresses the output variable in terms of the
the inputs to the system.
to state variables.
Variables associated with energy-storing elements qualify be
are taken as state
For an electrical network, inductor current and capacitor voltage
variables.
velocity of mass, displacement
For a mechanical translational system, displacement and
can be taken as state variables.
of spring and spring force
angular velocity of a shaft,
For a mechanical rotational system, angular displacement and
can be taken as state variables.
displacement of torsion spring and spring torque
A physical variable model is a state model in
which the state variables are physical
variables.
are successive derivatives of the first state
In phase variable model, the state variables
a

variable selected.
variables,
State variables need 'not be physical
state equation and the output
State diagram is a pictorial representation of the
equation. a system can be linearized using Taylor's
A nonlinear differential equation representing
series expansion.
 2 State Models from
Transfer Functions
2.1 TRANSFER MATRIX
Consider an n-th order system with p inputs and g outputs. The state model is
X= X+ BR:state equation
A

Y=CX+DR:output equation
where X= state vector with n state variables

R=input vector with p inputs

Y= output vector with q outputs
A=system matrix, n X n
B= input matrix, n Xp
C= output matrix, qX n
D= coupling matrix, q Xp.
Taking Laplace transform of the state equation,

SX($)-X(0) = A X()+B R(s)

[sI-AjX(S) = X(0)+B R(S)
(2.1)

where I is the identity matrix.

Assume that the initial state X(0) is zero. Then from equation (2.1),

[sl-A]X(s) = B R(9)

X(s) =[s1-AJ BR(S) (2.2)

Taking Laplace transform of the output equation

Y(s)= CX(s) + D
R(s)
52 Modern Control Theory

Using equations (2.2) and (2. 1)

Y(s) = C[sl-AJ'B R(s)+ D R(s)
-[clsl-Ar' B+D]R)
Or
Y(s) = Tr (s) R(s)
The transfer matrix is
Tu (s) = C[sI- AJ'B+D
This is the transfer matrix.
If D=0, the Transfer matrix is that matrix which pre-multiplying the input vector yield
output vector, when the initial states are all zeros.
Example 2.1 A system is represented by the state equation as

Derive the transfer matrix.

Solution: The transfer matrix, from equation (2.3)isoyw 89)

Tu (5) =C[sl -
A|'B+D

D=0

s+1 1 [s+2
(s+)(s +2)0 s+1
s+2]

Tu(s)=C[sl - AT B+D= 0s+2

s+2
(s+l)(s+2)s+2
 State Models from Transfer Functions 53

fs+2 -(s+1) ]
(s+l)(s+ 2)|s+2 - 2(s+ 1)

S+1 S+2
2
s+2

This is the transfer matrix and

1

S+2
Y(9) = |R(s).
2
Ls+1 s+2J

STATE EQUATION
2.2 TRANSFERFUNCTION FROM THE
a single-input/single
Transfer function is a special
case of transfer matrix when the system is
the Laplace transform of the output to the Laplace transform
output system. It the
is ratio of
zero.
of the input, with all initial states being -1|5645
single-input/single-output system shown in Figure
Consider a linear time-invariant
2.1 described by a state model,

X=AX+Br :state equation

y=CX+Dr : output equation
r single input, v() is the single output and X(t) is the state vector.
where () the
is

x(0)
X()=X()

,(0]

Linear Time Invariant

Continuous Time System
Output Response
Input Signal

Figure 2.1 Linear time-invariant system

Taking Laplace transform of the state equation, with zero initial states
sX(s) = A X(S) + B R(s)
 [sl-AJX(s) = B R(s)
X(s)-(sl-Ar' BR(s)
where Iis the identity matrix.
Taking Laplace transform of the output cquation
Y(s) CX(s)+ DR(S)

Combining cquations (2.4) and (2.5)

Y(s) = C[s l- Aj' BR(S)
+
DR(s)
The transfer function is
Y(S) =
C[sI-AJ-'B+D (2.6)
R(s)
Since this is a single-input-single-output system, this is a ratio
a denominator polynomial in terms
of a numerator polynomial and
ofs.
Y(s)
R(s)
= Adilsl-AB+D.sit
Det[s I- A]
WithD = 0, the transfer function js"

Y() Ad|lsl-Alg
R(s) Det [s A]
I- (23

Eigen Values
The equation Det [s
I- A]=0is called the characteristic equation
this characteristic equation are
the poles of the transfer function,
and the roots at
these roots are known as Eigen
values. These are the critical
In matrix terminologi
The Eigen values decide values of the complex frequency
the stability of a system. There a
every Eigen value. is system response corresponng
Eigen values decide the nature
of the natural response of a system.
Example 2.2 For a
lincar time-invariant system,
the system matrix is

Determine the characteristic

equation and the Eigen
values.
Solution: The
characteristic cquation is

|s-A|=0
 State Models from Transfer Functions 55

This simplifies to s+3s +2=0.

This is the characteristic equation. Solving this equation, we get
the Eigen values are s =-1 and s=-2.
s=-l and s= -2 and

2.3 STATE TRANSITION MATRIX

The matrix [sI-AJ is called the resolvent matrix or the state transition matrix in the
s-domain and is denoted as D(s). The state transition matrix in the time domain is obtained
by taking the inverse Laplace transform of (s) and can be expressed as

d)=1{s1-Ar"}
It is known that
A A2
[sI-AJ+I,
Taking inverse Laplace transforms
A,3
L'{(sI-Ar}=+A+ A'2
2! 3!

This is called the matrix exponential or the state transition matrix. The importance of state
transition matrix willbe discussed in Chapter 3.

d -eA! =AeA! and evaluate

Example 2.3 Show that (0).
dt

Solution: The state transition matrix is defined as

eAl=I+Al+ A;4 A
t..
2! 3! 4! k!

Differentiating each term on the right hand side,

A' A' A
A* +
2! 3! 4! k!

=A+A't+ A2 A
3! * (1)
2!

2!
A'A 4!
3! +...+
 as
Equation (1)can also be written
A'
A
A'4
2! 3! 4!

Hence
A=AeA=eAlA
dt

dt
=
O0) A
I=A.
Example 2.4 Evaluate e^de.

Solution:
A²e-+Ae -+: Ae4! k!
2! 3!

Integrating term by term on the right hand side,

A' A3,4
=I4Ar, t...
2! 3!

A[eAi0=At A'A',
2! 3!
A

() 4!

I+A|e^ d =l+At+AAA#
2! 3! 4!

If Aexists,

-I].
Example 2.5 For
the system matrix of Example 2.2,
determinethe state transition matrix and
a
show that (0)= A
 State Models from Transfer Functions 57

Solution:

a) =(sl-Ar' -
-1 T'
s+3 so+5)+2-2
s+3 1

(s+1)(s + 2) (s+1)(s + 2)
-2 S

(s+1)(s+2) (s+1)(s+2)
2 1 1

s+1 s+2 s+1 s+2

-2 2 -1 2
TAls+2 s+ls+2]
Taking inverse Laplace transforms, the state transition matrix is

| 2e--e e-e
-2e +2e-21 -e+2e-2
|-2e- +2e-21 -e+2e2
2e~4p-21 e-4e21

Example 2.6 The state equation for a system is

Find the resolvent matrix and the state transition matrix.

Solution: The resolvent matrix is

0
s+1
20)-1-Ar'-6
0
S+2.
Taking inverse Laplace transforms, the state transition matrix is
 58 Modern Control Theory

()

,-2/

Example 2.7 The state equation for a system

is

Find the state transition matrix.

*-.
Solution: The resolvent matrix is

o0=s1-4'
1
s+2 1] S
+2 (s+2)²
(s+2) 0 s+2| 1

S+2
Taking inverse Laplace transforms, the state transition matrix is

)eAr_Je

Example 2.8 For the given system matrix,

determine the state transition matrix.

-2 1

=0
A

-2 0
0 -3
Solution: The resolvent
matrix is

=
|S+2 -1 0 7-!
D(s) [sl-AJ =0 S+2
0
s+3

+(s+ 2)\(s+3)
-{-(s+3)} +0
(s+ 2'(s+3) -0 +(s+2)(s +3) -0
+0 -0 +(s+2)°
 State Models from Transfer Functions 59

1 1

(s+2) (s+ 2)?

0
(s+2)
1

0
(s+3) |
Taking inverse Laplace transforms,

le
P)=Ar =|0 e2
e

2.4 PROPERTIES OF STATE TRANSITION MATRIX

The state transition matrix is

A) =e^=sI-AT
1.
Itcan be expressed as an infinite series as
eA=J+At+ t.
2! 3!. k! k=0 k!
This converges absolutely for all values of t.
2. 0) =I is the identity matrix
3. 0) =A isthe system matrix.
4 The state transition matrix is non-singular, and its inverse exists for all : 1

5. )A-)=el-A! =I
The state transitíon matrix for a system can
be determined using properties (2) and (3). This
is illustrated in Example 2.12.

Example 2.9 For the state variable

description of a system given by

y) =[1
]x.
 60 Modern Control Theory

Draw a state diagram and determine the Ligen values, state transitionmntrix and the trane
function.
Solution: The state equations are

=- +r() (A)

(B)
The output equation is

Using these equations, the state diagram can be drawn as shown in Figurc E2.9.

Figure E 2.9 The state diagram for Example

2.9
The Eigen values are obtained by solving
the equation
0
=0.
The characteristic equation is

(s+ 1)(s +2)= 0.

The Eigen values are s
=-1 and s =-2.
The state transition matrix in the
s-domain is

s+2) (s+1)(s+
S+1|
1

0
S+1
1
1
1

L(s+1)(s+2) S+2 L(s+1)

The state transition in (s+2) S+2
the time domain is

L(s+I) (s+2)
S+2
Lee2
 MIT,
Bhcher
fitate Mdels fron Transfer FunctiE

The tanfer functionis

)-cqu-4'#-!
Y()
R(*)

ExAmple 2,10 A
statemodel for an 1TIsystem is givenhy

Determine the (a) charactcristic cquation, (b) Pigen values, (c) state transition
matriz and
(d) transfer function.

Solutlon: The characteristic equation is

--0or s+2s+5 m0.

The Bigen values aress-1 J2.
The state transition matrix in the s-domain or the resolvent matrix is

-1
s+2)
s+2 1

s+2s +5 +2s +5

+2s +5 +2s +
5.
s+1 1 2
(s+1)° +222(+1) +2 2 (s+1)² +22
5 2 S+1 1

2 (s+1) +22 (s+1)° +22 2(s+1 4+22

Taking inverse Laplace transforms,

thestate transition matrix in thetime domain is
cos21
+e'sin 21 1

ze sin 2:

sin 2! 21
cos20-esin
The transfer function is
Y(s)
=C[sl- 4|'B- CO()B
R(s)

(s+2s+5)'
s+8
1
1

|6s-5
7s +3
(s+2s +5)

Example 2.11 Draw the state diagram and find the state transition matrix and the transfer
function of a system represented by the state model
2 1]
X=| 0 -2
2 x+o
-1 -1 -4
o
y) =[1 o]x.

Solution: The state equations are

+ x
=2x, xy
=-2x +2x3 =- -y
*y –4.x, +2r.
The output equation is y= x .
Using these equations, the state diagram can
be drawn, keeping the output an integra
as
tor a state variable. The state diagram of
is shown in Figure E2.11.

4
+
+
-24
+

Figure E 2.11 The state

diagram for Example 2.11
 State Models from Transfer Functions 63

The state transition matrix is

-2
S+2 -2
s+4
Dei
sl- a]=s(s + 2)(s +
4)+2]+4+s+2 =g +6s' +1 1s+6
+(s +6s+8+2) -(-2s-8+1) +(4+s+2) |
1
D(s)= <2) +(s +4s +1) -(-2s)
s+6s+lls +6
#-(s+2)) -(s+ 2) +(s+2s)

1
(s +6s +10) 2s+7 (s+6)
-2 (s + 4s+1) 2s
s+6s+1ls +6 (s²+23)|
-(s+ 2) -(s+2)

The transfer function is

Y() =CO(s) B
R(s)

s+6s+10) 2s+7 (s+6) 0

1
0
o]-2 (s²+4s+1) 2s
+6s+1ls+6
(s+2) -(s+ 2)t (s+2s) 2

2(s+6)
+6s +1ls+6

Example 2.12 A system matrix for a linear time-invariant system is given as

Determine the state transition matrix using its properties.

Solution: The Eigen values are obtained by solving the equation

-1|
s1-A|=0 =0.
s+3
Or

(s+1)(s +3)= 0.
 state transition matrix wit
values are s =-land s= -3. Bnch clenment of thc
The 'igen
polynomialin terns of
eund e, llence

Using the property 0)=1,

=
Using the property 0) A,
-a-3a) o
-3 -d-34,
From cquations (1) and (2),
a +
Solving these equations a, =1 a, =0
3a, =l
=
h +b, =0 4 +36, =-1 Solving these equations b,
2
q +3e, = 0 Solving these equations c =0
d + 3d, =3 d =0
+
d d, =1 Solving these equations d, =1
Hence, the state transition matrix is

let ,1-3t
p-3t

2.5 REALIZATION OF STATE MODELS

FROM THE TRANSFER
FUNCTION
Asmentioned earlier, a state model obtained
for a systenm is not unigue and depends on
choice of state variables. The state
variables can be physical variables or
other set of variables which even may phase variables or
not have physicalentity. Hence,
state models can be derived
representing the same system, fora given system,
system,many state models can Given the transter function
be realized using different
function. methods of processing the tra
For an n-th order system,
system matrix
and g outputs, the input matrix the
inputs
is of size n X n,
If there are P
The input-output coupling
B is of size n
xp and the output matrix C is of siz 9
matrix is of size q X p. Hence,
the number of elements e
A
ILC. And D matrieCs s cqual to n' np m qp,T his unter is verylary a4 cAn
ilicumtly teducol by doriving state models fron transter fuicticis sóng stzin tatiat
d
aeecntcd methods. Such state nodels withtmatries having miniun toutes h AHWA)
aleuehts nrc cnlled canonieal fornms, State models in canonical fortns ate very UmYIM
the analysis and desipn of plysicnl systems. Canonical state noels hæe
tain tra1aHA
averother state models. In certain canonleal forns, the state transiticn matríz can te
wfts
down by inspection,
The rcalization of the following four canonical forms will be diuw here.
1.
Controllable canonical form
2. Obscrvable canonical form
3. Diagonal canonical form
4
Jordan canonical form.

2.6 CONTROLLABLE CANONICAL FORM

The transfer function of a physical system is the ratio of polynomials in terns cf the variatile s.
Usually the degrec of thc numerator polynomial ís less than the degree of the denominztor
polynomial and such a transfer function is said to be strlctly proper.
Consider the transfer function of a thírd-order system given as

Y(3) b,s +bh,s +bo

This can be written as

Y() bys
+b,s+b, W)e

and can be represented by a block diagram shown ín Figure 2. 2.

R(S) 1 W(s) Y(s)

bzs + bs + bo

Figure 2.2 Realization into controllable canonical form

Ws) is an intermediate variable. From the block diagram,

W(s)
R(s) s' +4,s + 4s+ ay

(s +4,s + ajs + ay) W(s) = R()

 Taking inverse Laplace transforms,

.
W+a, W+ a 1w+ a W=r() (2.9)

Choose }=W, -X =" and ; =X, = w as the three state variables.

(2.10)

(2.1)
From equation (2.9),

W=-dWp - a w-ay W+r()

Or

(2.1)
Equations (2.10), (2.11) and (2.12) are state equations and can be expressed in the matrir
form as

0 1

0 0

The output equation, from the block diagram, is

Y(9)= (b,s +hs+b,)W()

Taking inverse Laplace transforms,

=
y() b,w+b w+b, w = b, x
+b xy +b, xy =|b (2.13)
b

Note: It can be observed that in the system

matrix A, the elements of all the rows excep
last are either 0 or1. The last row contains
the coefficients of the denominator polynomlal
the transfer function,with negative
sign: The elements of the input matrix B are O e
the last row, which is 1. The elements of the output all
numerator polynomial matrix C are the coefficients ot
of the transfer function.
The statediagram can be
drawn as shown in figure 2.3.
 State Models from Transfer Functions 67

Figure 2.3 State diagram - controllable canonical form

Example 2.13 The transfer function of a dynamic system is
Y(s) 2s +5
R(s) s+7s+10
Obtain the state model in the controllable canonical form.

Solution: The transfer function can be written as

W(s)
Y(s) 2s+5 oitot Tul
tdT szsfs
R(s) s+7s+10 W(s)

(s +7s+10)W() = R(s).
Taking inverse Laplace transforms,

w+7w+10w =r(). (1)

Choose x = w
and x, =x, =
W
as the two state variables.

(A)

From equation (1),

w=-7w-10w +r()

Or

x =-10x -7x, +
r(0). (B)

Keeping equations (A)and (B) in thematrix form,

 The output equation is
Y(s) = (2s +5)W(s).

Taking inverse Laplace transforms,

= 2w+5w =
() 5x +2.x, =|5

Example 2.14 The transfer function of a dynamicsystem is

Y(s) 10(s +4s +8)

R(s) s +7s +10s
Derive the state model in controllable canonical form.

Solution: The transfer function can be

written as
Y(s) 10(s+4s+8) W(s)
R(s) s'+7s +10s W(s)
(s +7s +
10s)W(s) = R(s).
Taking inverse Laplace transforms,

w+7w+10w =r(t) (0)

Choose X
x =,
= W,
Xy =X = w and = W as the
three state variables.

(A)

(B)

From equation (1),

W =-7w-101w+r(t)
Or

Xy
=-10x, -7x,+r().
 State Models from Transfcr Functions 69

Keeping equations (A), (B) and (C) in the matrix form,

(0 1

-)
|0 -10

The output equation

is
Y(s) = (10s + 40s +80) W(s).

Taking inverse Laplace transforms,

() =10w+ 40 w+80 w = 10.x, +40.x, + 80x = (80 40 10]||xz

Example 2.15 Derive the state model in its controllable canonical form for a system with
transfer function

Y(s) 5s+14s+8
R(s) S+9s+ 23s +15

Solution: The transfer function can be written as

Y(s) 5s+14s+8 W(s)
s
R(s) +9s+23s+15 W(s)
(s+9s²+ 23s +15)W(s) = R(s). (1)

Taking inverse Laplace transforms of equation (1), etsinel ya

w+9w+ 23 w+15w =r(t). (2)

ioose x
= W,
X =X =W and x =X = W
as the three state variables.

(A)

x = (B)

From equation
(2),

w=-9w- 23 w-15v +r()

Or

x, (C)
=-15x-23x, -9.x, +r().
 70 Modern Control Theory

Keeping equations (A), (B) and (C) in the matrix forn,

|-15 - 23

The output equation is

Y(9) =
(5s +14s+8)W(s).

Taking inverse Laplace transforms,

|
14
v() =Sw+14w+ 8v=8x +14x, +5x =|8 5|

2.7 OBSERVABLE CANONICAL FORM

Consider a third-order system with transfer function
Y(s) b,s +b,s+bn
R(s) s + a,s + ahS+ a
After cross multiplication, this can be written as

(s+ays +
as+a,) Y() = (b,s +hs+b,)R(s).
=
Taking inverse Laplace transforms and using the symbol D

(D'+a,D' +aD+ a,)y= (6,D +hD+b,)r

D'y = D'(b,r-a,y) + D(Gr-ay)+(byr - a))
Or
(2.14)
y= D(b,r-d,y) + D"(6r-ay) + D
(byr-aa).
Let the three state variables be x
=), x, and x;.
Hence, from equation (2.14),
x =D(b,r- d,x)+ Dbr-a:4)+ D'(byr-ay).
 Iransrer runctions 71

Differentiating this equation,

=
(b,-a,)+D"(br-ax)+ D*(byr- ag). (2.15)
Let
x, =
D(br-ax)+ D
(bar-agx). (2.16)
Hence, from equation (2.15)
and (2.16),

(2.17)
Differentiating equation (2.16),

=(hr-a4)+D"(bor- ag;). (2.18)

Let
Xy = D(6yr-4%;).
(2.19)
Hence, from equation (2.18)
and (2.19),

(2.20)
Differentiating equation (2.19),

Keeping equations (2.17), (2.20)

and (2.21) in the matrix form,

1
by
=|-a
00

The output equation

is

0 x
y=x=|1 0|

Note: It can be observed

that in the system matrix, the elements
the coefficients of the denominator
polynomial of the transfer function
of the first column are
theelements of the other columns are with negative sign and
either 0 or 1, Theelements
the coefficients of the numerator of the column matrix B are
polynomial of the transfer function,
Output row matrix The first element of the
Cis 1' and all other elementsare 0.
 The state diagram is shown in Figure 2.4.
r(t)
bo

Figure 2.4 State diagram

- observable canonical form
form for a system with trane
Example 2.16 Derive the state model in its observable canonical
fer function
Y() 2s+5
R(s) s47s+10

Solution: Rearranging the transfer function,

(s +7s+10)Y(S) = (2s+ 5)R(s).
Taking inverse Laplace transforms,

y+7y+10y= 2r+5r.
This can be written in terms of the differential operator as
(D +7D+10)y = (2D +5)r
D'y= D(2r-7y) +(5r -10y)

y=D(2r-7y) + D
(5r-10y).
Let x = y and x, be the two state variables.
Equation (1) can be written as
x = D"(2r-7x;)+ D(Sr-10x).
Differentiating equation (2),

=(2r-7x)+ D"(Sr-10x,).
Let
xy = D"(5r -10:x,).
 Slate Models (rom Transler Functions 73

Combining equations (3) and (4),

(A)

Diflerentiating cquation (4),

(B)

Keping equations (A) and (B) in the matrix form, the state equntion is

The output equation is

a system with
Example 2.17 Derive the state model in its observable canonical form for
transfer function
Y(s) Ss+14s+8
R(s) s +9s + 23s +15

Solution: The transfer function can be written as

(s' +9s+23s +15)Y (s) = (5s+14s+8)R(s).

Taking inverse Laplace transforms,

y+9y+23 y+15y
= S r+14r+&r.
can be written as
In terms of the differential operator D, this
+
D'y+9D'y+ 23 Dy + 15y' = 5D'r +14Dr &r.
From this equation,
D
(8r-15y), (1)
y= D'(5r-9y)+ D(14r-23y)+
Let x = J, ,
and x, be the three state variables.
From equation (1),
x= D(5r -9x) + D"(14r- 23x)+
D
(8r-1Sy) (2)

Differentiating equation (2),

D
(8r-15x). (3)
j = (Sr-9x)+D"(14r- 23x)+
 Let
x, = D(14r-23x,)+
D (8r-15x).

Combiningequations (3) and (4),

Differentiating equation (4),

x = (14r- 23.x )
+ D(8r-15x).

Let
X =
D(8r-15x).
(6),
Combining equations (5) and

Differentiating equation (6),

form
Equations (A), (B) and (C) are state equations and keeping them in the matrix

1 0
-9 |
=-23 o
00 8
-15

The output equation is

y==| 0

2.8 DIAGONAL CANONICAL FORM

This canonical form of
the state model is obtained usingthe partial fraction expansion oft
given transfer function,
provided the Eigen values are
Considera third-order system distinct.
with transfer function

Y()
R(S)
b,s +bs+ bo
+4,s +as+40
 State Models frornTransfer Functions 75

Let the three Eigen values be distinct and let them be

Hence the transfer function can be written as
s-h s- and s = -h.

Y(s) b,s +hs +ho

R(s) (s +4, )(s+, )(s + 4)
Let the partial fraction expansion be

Y(O)
k k, k
R(s) (s+4)' (s+4) (s+ h)
Rearranging this transfer function,

Y(s)= k (9)+ k, R(s) + k (2.22)

(s+4, (s+4)
-R(s)
(s+ 4)
Or
Y9) =k,X,() + k, X,(s)+k,X,(9). (2.23)
These equations are represented by means of a block diagram as shown
in Figure 2.5. Since
the blocks are in parallel, this method of realization is also
called parallel realization.

1 |X(s)
s+
R(S) 1 |X,(s) Y'S)

1 |Xs(s)

s+
Figure 2.5 Block diagram - Diagonal formite
From equations (2.22) and (2.23),

X,(9) =. R(S) X(4)=R)

s+h s+
(s+4)X,(s) = R(s).
Taking inverse Laplace transforms,

*j+4x r() x=-4x+r(/) (2.24)

(s+ ,)X,(s) = R(s).
Taking inverse Laplace
transforms,
 76 Modern Control Theory

x,+ dyty = r()

(Q2
and (s+ ,)X,(s) = R(s).
Taking inverse Laplace transforms,

j+4.X; = r() 2.26)

Equations (2.24). (2.25) and (2.26) are the state equations and keeping them in the matrix for

Taking inverse Laplace transform of equation (2.23), the output equation is

=k +
v) +ky kyy =[k, k, k
k, ||

Note:
1. The system matrix is diagonal, with the Eigen values as
the diagonal elements and al
the off-diagonal elements being 0.
2. The elements of input
matrix B are all 1.
3. The elements of the output
matrix C are the residues of the transfer
function at its pols
Theadvantage of this method realization
of is that each state equation is a first-order
ential equation in only one variable. difir
Thus these equations can be solved independently.
these equations are said to be decoupled Henc
and the state model is said tobe in decoupled form
The state diagram is shown in Figure
2.6.

Figure 2.6 State

diagram- Diagonal canonical
form
 State Models (rom Transler Purnctlons 77

Example 2, 18 Derive the state model in its diagonal canonicnl form for asystem with transfer
function
Y(*) 2s45
R(s) +7s+10
Solution: The denominator polynominlcan be factorized as

s+7s+10 (s+2)(s +5).

Hence, the Eigen values are and s-S.
s-2
Expanding the transfer function in partial fractions,

Y(s) 2.s +5 ,W3)(5/3)

R(s) s
+7s+10 s+2 S+5

Y(o) = U3) R()S

S+2 S+5
R--x0)+x,). (1)

where
R(s)
X(9) = R(s) and X,(s) =
s+2 S+5
(s+ 2)X, (s) = R(s).

Taking inverse Laplace transforms,

x+2.x = r() or xj =-2.x +r(0) (A)

(s+5)X,(s) = R(s).

Taking inverse Laplace transforms,

xy+ Sx, = r() (B)

or
them in the matrix form,
Equations (A)and (B) are the state cquationsand kecping

From equation (1),

),
vu-00).
 78 Modern ControlTheory

Taking inverse Laplace transforms,

Example 2.19 Derive the state model in its diagonal canonical form and draw the tate
.
gram for a system with transfer function

Y(s) 5s+14s+8
R(s) +
952+ 23s+ 15

Solution:
The denominator polynomial can be factorized as

+9s+23s + 15 = (s+1)(s+3)(s +5).

Hence, the Eigen values are s =-1, s=-3and s=5.
Expanding the transfer function in partial fractions
Y(s) 5s+14s +8 -(1/8) (11/4), (63/8)
R(s) s +9s+23s +15 S+1 S+3 S+5
1 R(S) 11 R(s), 63 R(s)
Y() = +
8 s+1 4 s+3 8 s+5
Let

Y)---) 8
From equations (1)and (2),

X})=R6) X, (s) = :R(s) R(s)

S+1 X,(s) =
S+3 S+5
(s+1)X,(s) =
R(s).
Taking inverse Laplace transforms,

x+x =r() or x=-+r(t)

(s+3)X, (s) = R(s).
Taking inverse Laplace transforms,

g+3:x, =r) or x =3.x, +

r(/)
(s+5)X,($) = R(s).
 79
State Models from Transfer Functions

Taking inverse Laplace

transforms,
(C)
1y+5x, =r() Or x =-5x, +r().
form,
Keeping equations (A), (B) and (C) in the matrix

0 -3
0
-5

From equation (2), the output equation is

Y)--x9-x)+9.
Taking inverse Laplace transforms,

The state diagram is shown in Figure E 2.19.

1/8

11/4
+

63/8

Figure E 2.19 State diagram for Example 2.19

2.9 JORDAN CANONICAL FORM

The Jordan canonical form of a state model is also obtained by the partial fraction expansion
Of the given transfer function. But in thiscase some of the Eigen values are repeated. The
diagonal
canonical form is a special case of Jordan canonical form.
Consider athírd-urÓez systern
with transíer function

Y() b, +hs h,

Lathe three Eizen values be sm

-h, sm -, and s =

tranfer function can be written -h.

Eizen value. Henc the Note that
as s=-hisa
Y) thhs+by
Ra) (s4 4Ý(s+4)
La the partialfration expanion be
Ya)
R(:) kuk2(s+4)
(3+
(s+)
Y()= k R(3+k2(st RG)
(s+4, (s+4) 22
Or

Y)= khy X,)+kyg X(s) + ky

X,(s). (22
From equations (2.27) and (2.28),

X(s)=R)
(s+4
X,()= R) (23
(s+4)
X,(s) = R() (23
(s+4)
From equations (2.29) and (2.30),

X,()= X,6) (23

(s+4)"
Thee equatíons are represented by means
of a block diagram as shown in Figure 2.1.

|X(3)
R(s) +
Y()
X(6)

Figure 2.7 Block

diagram - Jordan canonical stA
form
 State Models from Transfer Functions 81

From equation (2.32),

or (s+ 4)X, (s) = X,(s).

Taking inverse Laplace transforms,

Or (2.33)

From cquation (2.30),

R(s)
X, (s) = or (s+4,)Xz(s) = R(S).

Taking inverse Laplace transforms,

or xy
=-h xy +r(). (2.34)

From equation (2.31)

R(S)
X, (s) = or (s+)X, (s)= R(s)b
Taking inverse Laplace transforms,

(2.35)

Keeping equations (2.33), (2.34) and (2.,35) in the matrix form, the state equation is

laking inverse Laplace transforms of cquation (2.28), the output equation is

= +
=[kii ki2
yl0)
k +ki2x) ky,

The state diagram is shown in Figure 2.8.

 Figure 2.8 State diagram - Jordan canonical form
Note:
1. The sub-matrix in the system matrix is called a Jordan block.
0
2. The diagonal elements of the system matrix are the Eigen values.
3. All the elements below the diagonal are 0.
4. The element in the previous row and just above the repeated
Eigen value is land all
other elements above the diagonal are 0.
5. The element of the output
matrix B corresponding to the last row of the Jordan
block is 1 and all the previous elements corresponding
to each row of the Jordan
block are 0.
6. The element of the output
matrix B corresponding to the non-repeated
in the system matrix is 1. Eigen value

Example 2.20 The transfer function of a system is e.)hos (sE.S)8E53 noitsyo argpt
Y() 2s+5
R(s)
(s+1'
Obtain a state model in the Jordan
canonical form and draw the state
diagram.
Solution: The Eigen values are
repeated and are equal to s -1. Expanding in
partial fraction,
Y(s)2s+53 20lutai sosiq.i ves
R(s) (s+1 (s+1 (s+1)
Or

Y(o) = 3
-R(s)+: 2 R(S). ()
(s+1)? (s+1)
Let

Y()=3X(s)+ 2X,(s). (2)

 State Models from Transíer Functions 83

From equations (1) and (2),

X, (s) = R(s) (3)

(s+1)?

X, (s) = R(s) (4)

s+1
From equations (3) and (4),

(5)
S+1
From equation (5),
(s+1)X{(s) = X,(9).

Taking inverse Laplace transforms,

(A)
From equation (4),

(s+1)X, (s) = R(s).

Taking inverse Laplace transforms,

Xy+xy =r() or xy =-y tr(). (B)

Keeping equations (A) and (B) in the matrix form,

Taking inverse Laplace transform of equation (2), the output equation is

y() = 3x, +2x, =|3 3

The state diagram is shown in Figure E2.20.

-
-
Figure E2.20 State diagram for Example 2.20
 order six, the six Eigen values are repeated thrice,
s =
Example 2.21 For a a system of s =,
Write down the state equation in its Jordan canonical forman
repcated twicc and s=h.
comment on the structurc of the state cquation.
The six Eigen values are s thrice, s= repeated twice ands = .
Solution: =h repeated
can be written down as
The state equation in the Jordan canonical form

0 0
- 1 0 0
0 - 0 0 0 1
0
1
0
0 0
0 0 -4]

There are two Jordan blocks as expected. In the output matrix, the element that corresponds
to the last rowof each Jordan block and the element that corresponds to the non-repeated
Eigen values are 1 and allother elements are 0. This is indicated by the arrows.

Example 2.22 Obtain a state model in Jordan canonical for a system with transfer function

Y() 2(s +2s +3)

R($) (s+1)' (s+3)
Also draw the state diagram.

Solution: Expanding the given transfer function in partial fractions

Y(s) 2(s +2s +3) 2 1

3
R(s) +
(s+1' (s +3) (s+1) (s+1) (s +3)
2R(s) R(s),3R(s) (0)
Y(s) =
(s+1)' (s+1) (s+3)
Let

Y()= 2X} (s)- X,()+3X}(s).

From equations (1) and (2),

X(s)= KS)
 State Models fronm Transfer Functions 85

X,(s) = R(s) (4)

(s+1)

X}(s) = R(s) (5)

(s+3)
From equations (3) and (4),

X} (s) = X,(s) or (s+1)X, (s) = X,(5).

(s+1)
Taking inverse Laplace transforms,

x+x =X or x=-x +x. (A)

From equation (4),
(s+1)X,(s) = R().
Taking inverse Laplace transforms,

Xyty =r() =- +r). (B)

From equation (5),

R(s)
X,(s) = or (s+3)X,(s) = R(S).hst ciBsl
(s+3)
Taking inverse Laplace transforms,

Xy+3x, = r() or x =-3x, +r(). ()

Keeping equation (A), (B) and (C) in the matrix form

-1 1
0
x
0 -1 +|1|r().
0 -3

laking inverse Laplace transforms of equation (2), the output

equation is

y() = 2x, - +3x, =[2 -1 3]|

86 Modern Control Theory

The state diagram is shown in Figure E2.22.

n)

Figure E2.22 State diagram for Example 2.22

Example 2.23 Obtain a state model in Jordan canonical for a system with transfer function
Y(s) 16
R(S) (s+i(s+3)(s+ 5)
Also draw the state diagram.

Solution: Expanding the given transfer function in partial fractions

Y(s) 16 (312), 2
(1/2)
R(S) (s+1)' (s+3)($+5) (s+1) (s+1) (s+3) (s +5)/

Y(s)=2-R(s) 3 R(s) R(s) 1

R(s) rh
s+1 -+2
2 (s +1) (s+3) 2 (s+5)8: s4) 4otscs goR
Let

Y)-2x,0)-x,)+2x,0)-0.
From equations (1) and (2),

R(s)
X{(s) =

R(s)
X,(s) =
(s+1)
R(s)
(s+3)
X4 (s)= R(s)
(s+5)
 CIB
irom Transfer 'unctions
87
From equations (3) and (4),

(s+1) (7)
From equation (7),

(s+1)X{(s)= X,(6).
Taking inverse Laplace transforms,

or x -X +
From equation (4), X. (A)

(s+1)X,(s)= R(9).
Taking inverse Laplace transforms,

Xy+xy =r()
(B)
From equation (5),

(s+3)X,(s) = R(s).
Taking inverse Laplace transforms,

x+3x, =r() or x =-3x, +r(). (C)

From equation (6),

(s+5)X4(s)=R($).
Taking inverse Laplace transforms, MOFTAS.ABA BOA32A3 0r.S

Keeping equations (A), (B), (C) and (D) in the matrix form, the state equation is

-1 1
001|
0-1 ti 0
+| r().
0 -3 0
|
0 0 -5
 88 Modern Control Theory

Taking inverse Laplace transforms of equation (2), the output equation

is

3
=
) 2x -+2r3 2

The state diagram is shown in Figure E2.23.

B2
r)

X4

Figure E2.23 State diagram for Example

2.23

2.10 CASCADE REALIZATION

This method of realization is achieved by
writing the given transfer function as a proa=
of simple ratios of polynomials. The degree the numerator
of polynomial or the den
nator polynomial is at most two. Generally,
the denominator polynomial is of in
second degree and the numerator is either a
first-degree polynomial ora constant
These factors can be represented by a
series of blocks in a block diagram.
of each block is taken as a state variable, A Ine
differential equation can be obtaine
each block and, hence, a state equation.
The method is illustrated by means o
examples.

Example 2.24 The transfer

function of a second-order system given by
Y(S)
k(s + b)
R(s) (s + a)(s+ a,)
Obtain a state model by cascade
realization.
 State Models fronm Transfer Functions 89

Solution:
Y(s) k(s+b)
R(s) (s +4,)(s+ a,)
Splitting the right hand side into two ratios,

Y(s) k (s+b)
R(s) (s +a) (s +
a,)
These factors can be represented by block diagram as shown in Figure E2.24.

R(S) Vs) Y(S)h

sta1 s t a2

Figure E2.24 Cascade realization for Example 2.24

From the block diagram

V(s) k or (s+a)V(s) = k R(s).

R(S) s+ a
Taking inverse Laplace transforms,

V+a v= kr() or v=- v+kr(). (1)

Also

Y()_s+b or (s +a,)Y(s) = (s+ b)V(s).

V) s+ag
Taking inverse Laplace transforms,

y+ a, y =y+by or y=-a, y+bv+v. (2)

Select the output of the blocks y and v as state variables. Hence x = y and x, = v are cho
Sen as state variables.
From equation (2),

(3)
From equation (1),

(A)
Substituting equation (A) in equation (3),

4=-4yy +bx-4Xy +kr() =-a, +(6-4, )x +kr(). Xj

(B)
 Keeping equations (A)and (B) in the matrix form,

The output equation is

v(0) =x =[1

The corresponding state diagram is shown in Figure E2,24 (a).

(b-a)

Figure E2.24 (a) State diagram of the cascade

realization for Example 2,24
Example 2.25 For the example 2.24,
of obtain an alternate state model by cascade realization.
Solution:

Y() k(s +b)

R(S) (s+4)ls+ a,)

(s+a) (s +ay)
These factors can be represented
by a block diagram as shown
in Figure E2.25.
R($) V(s)
k Ys)
s+
sta
Figure E2.25 Cascade realization
for Example 2.25
V() k
R(s) s+ a Or (s+4)V(s) =k R(s).
Taking inverse Laplace
transforms,

V+ a,v= k r(o)
or v=-d v+kr(i).
 State Models from Transfer Functions 91

Choose x =v as a state variable. Hence, cquation (1) can be written as

X =-ax +kr. (A)

Y(9) s+h -=1+(6-a4)

V(s) S+a, S+a
Y() =V(o)40-ay).
S+a)
Let
Y(s) =V(s) +W(9). (2)
where

Wo)= (b-alvs) (s+a,)W(s) = (b-a,)v(s).

S+a,
Taking inverse Laplace transforms,

w+a, = (b-a)v
1W
or w=-a, 1w+(6-a,)v.s (3)

Select x, =W as the second state variable. Hence, equation (3) can be written as

(B)

Keeping equations (A) and (B) in the matrix form

Taking inverse Laplace transforms of equation (2) the output equation is

y=V+w= x+x) =||

The corresponding state

diagram is shown in Figure E2.25(a).

-- a

Figure E2.25 (a) Alternate form

(b-a)

a
of state diagram – cascade realization
 92 Modern Control Theory

Example 2.26 For the example 2.24, obiain another state model by cascade realization,

Solution:
Y(s)_ k(s+b)
R(s) (s +a,)(s + a,)

Rearranging the factor terms of the transfer function,

Y() (s+b)
R(s) (s+ a,) (s + a)
These factors can be represented by a block diagram as shown in Figure E2.26.

R($) VS) Y(s)

s+ a
sta
Figure E2.26 Rearrangement of blocks- cascade realization

From the block diagram

V(s) S+b -=l4(-g)

R(s) s+ ay (s+a,)

V(s)= R(s) +- (b-4 R(9)=R(s)+ W(s).

W()= R(S) or
(s+a,)W(s)=(b-a,)R(S).
(s+a)
Taking inverse Laplace transforms,

W+a, w= (b-a,)r()
or w:
W= -4, W+
(b-a,)r().
Choose x =W asa state variable. Hence,
equation (2) can be written as

Also

Y()_ k
V(s) s+a (o or (st4)Y(s) = k V().
Taking inverse Laplace
transforms,

y+ a,
y=kvor y=-d y+kv.
2832
 Slate Models from Transfer Functions 93

Taking inerse aplace transtormsof cquntlon(1),

(4)
"=tw.
Combining equations(3) and (4),

(5)
J'm-4)+ k (r+w).
Select ay ny as another statevariable. Thus, from equation (5),

(B)

Keeping cquations (A) and (B) in the matrix form,

The output equation is

}
y= =[0

The state diagram is shown in Figure E2.26 (a).

(a)

Figure E2.26 (a) State diagram - cascade realization

The three state models obtained for thesame system of Example 2.24 are different. It can.
be seen that the cascade realization
does not give rise to a unique state model. Hence, the
state model obtained by the cascade realization
does not belong to the class of canonical
forms.

Example 2.27 The transfer function

of a system is
Y(s) 5(s +3)
R(s) s(s+1)(s +5)
Obtain a state model by cascade realization.
 04 Modern Control Theory

Nolutlon The plvon tnalor hwmetloD enn bg wrltten In a fnctorod form 8

'(*)
R(a) ( 5) (&+)*

This is vepvesented lby a block dlugrmm as alown In lgure B2.27.

V*) Y(*)
5

gure E2.27 Block dlagram for Bxamplo 2.27

Conslder I()
R() s+5
From this factor,

(s+S)V(s)=5 R(1).
Taking iversc laplace transforms,

P+Sv= 5r() or y-5v+5r(0).

Now

VG) (s+3) or (s+1)WV (s) =

(s +3)V().
Taking inversc Laplace transforms,

Also

Y(G)

IV()
1tlo or sY(s) W).
Taking inverse Lapluco
transforms,utoakstsS es 2th

Sclcct x y, X w and x V as stato varlables.

Henco, from cquation (3),

From cquntion
(2),
From equation (1),

+
}=-5x, Sr(1). (B)
Substituting equation (B) in equation (4),

x=-) +{-5x, +5 r()}+3 x, =-X-2x +5r(). (C)

Equations (A), (3) and (C)are the state equations. Keeping them in the matrix form,

1
07
0
0 -1 -2 +|5|().
0 0

The output equation is

y=x =1 0 x
0]|

The state diagram is shown in Figure E 2.27(a).

Figure E 2.27(a) The state diagram for Example 2.27

Example 2.28 A state diagram of a linear time-invariant system is shown in Figure E 2.28.
Realize a state model for the systemn and hence find the transfer function,

2
Figure E 2.28 The state diagram for Example 2.28
 96 Modern Control Theory

taken as state variable. The given state

a
Solution: The output of each integrator
gram is redrawn as shown in Figure E dia.
2.28 (a), selecting the output of each integratoT
as
state variables.

r)
+

Figure E 2.28(a) The redrawn state diagram for

Example 2.28
From the state diagram, the state equations can
be written as

X=-2x +7r()
()

These are the state equations

x =- +3x +r().
and keeping them in the matrix
form,

The output equation is

y=x+x=1
The transfer function
of the system is given
by

Cisl - A B =[1

|S+1

(s+1

8s+30
(s+1)(s + 2)
 State Models from Transfer Functions 97

SUMMARY
The transfer matrix is that matrix which pre-multiplying the input vector yields the output
vector.
Thetransfer function is a spccial casc of transfer matrix, It is defincd with respect to
(SISO) system. It is defined as the ratio of the Laplace
a single-input/single-output

transform of the output to the Laplacc transform of the input, with all zero initial
states.
Eigen valucs are the system poles, They are the roots of the characteristic equation
-
|sl A=0 where A is the system matrix.
State transition matrix in the s-domain is called the resolvent matrix: (sl-AT
The inverse Laplace transform of this is the state transition matrix in the time
domain.
A
state model with the matrices A, B, Cand D which have the minimum number of non
zero elements is said to be in canonical form.
The diferent canonical forms are controllable canonical form, observable canonical form,
diagonal canonical form and Jordan canonical form.
Diagonal form of a state model is associated with a system with distinct Eigen values.
In the diagonal canonical form of the state model, the system matrix is diagonal
with the diagonal elements as the Eigen values and all the off-diagonal elements as
Zeros.
Diagonal canonical form is also called parallel form or decoupled form.
In the diagonal canonical form, each state equation is a first-order differential equation
in only one state variable and the input. The equations are said to be decoupled. The state
equations can be solved independently.
Systems with repeated Eigen values can be expressed in Jordan canonical forms.
The diagonal and Jordan canonical forms are obtained by the method of partial fraction
expansion of the transfer function.
The state model obtained using cascade realization is not unique. More than one state
model can be realized by this method from the same transfer function.
For obtaining a state model from a given state diagram, the variable at the output of each
integrator is taken as a state variable and the equation for the variable at the input of each
integrator is written down.

PRACTICE PROBLEMS
PP 2.1 Determine the transfer matrix of a multiple-input/multiple-output system represented
by the state model
 Similarity
4 Transformations
4.1 SIMILARITY TRANSFORMATION
As mentioned in Chapter 2, canonical models have certain advantages over other state mod
cls. Often, it is convenicnt to represent the system in canonical forms for the simplicity of
design and analysis. A non-canonical form of state model can be converted into different
canonical forms by linear transformation. These systems are called similar systems and the
method of obtaining these is called similarity transformation. The method is based on select
ing a transformation matrix that transforms the given system matrix into a canonical form.

4.2 METHOD OF SIMILARITY TRANSFORMATION

Two (nX n) matrices A., and A, are said to be similar if a non-singular matrix P exists such that
Ay = [P"AyP).

The matrix A, is said to be obtained from A,, by a similarity transformation and P

is
called the transformation matrix. It was discussed in Chapter 2 that different state models
can
be obtained from the same transfer function representing a system. Hence, a state model for
a system is not unique. Consider the matrix equation
X= PZ.

The matrix P transforms a vector Z into a vector X. This transformation is valid only
When the inverse of P exists. Let the state model for a SISO system be

X= AX+Br (4.1)

and

y= CX (4.2)
146 Modern Control Theory

Consider that this state model is to be converted into another state model and let
the
is the transforma-
is the new state vector and
P

transformation be X(t)
=
PZ(), where Z(t)

tion matrix.

X= PŽ (4.3)

From equation (4.1) and (4.3)

=
PZ APZ+ Br

2 =[P'APJZ +[P'B]r= 4,Z+Byt (4.4)

A, -[P'AP) is the new system matrix and B, =PB]is the new input matrix.

y= (CP]Z (4.5)

Cz =(CP] is the new output matrix.

Equations (4.4) and (4.5) show the non-uniqueness of a state model. The set of equations
(4.1) and (4.2) and the set (4.4) and (4.5) represent the same system. A system can have dfer.
ent state space representations, but these models have the same transfer function and Ejgen
values.

4.3 EIGEN VECTORS

Consider a square matrix A and let represents its Eigen values. A column vector P that
satisfies the relation AP = 2Pis called the Eigen vector.
For every Eigen value, there is an as
sociated Eigen vector. Eigen vectors can be determined by any one
of the following methods
Method 1: By solving the equation AP
For an Eigen value ,, =APtke he
ict ihar ( t)

AP = , P
This gives

[41-A], =(9]
Consider a (3 × 3) system matrix with ,,
distinct Eigen values and
.
Let the EI3*
vector associated with these
Eigen values be

R=| 21 , R=P2
and =P23
LP31 LP2 LP33
 Similarity Transformations 147

Solving the equation

P21 =|o.
P31 0

can be obtained.
P21

Similarly =| P22 can be obtained by solving the equation

LP2

P2
[a1-A] Pz2
0.
P32

In a similar way, can be obtained by solving the equationi0

=| P23
LP33

| P3
[21-Al-|Pzs

fP Pi2
P=(R 2.
P22 P23

LP31 P32 P33.

clod 2; Select any column of fadjoint [,I-A}. av eih9

same as the cofactors of any one row of [, I-A], where h, represents the
Ihis is the
values. Then for each Eigen value, the corresponding Eigen vector
can be found out. In
8n can be removed.
*n vector obtained, the common factor, if any, in the column elements,
Hse methods are illustrated in the following examples.
Example
4.1
2
For a square
vectors. matrix A determine the Eigen values and the associated Bigen
 148 Modern Control Theory

Solutlon: The Eigen values are obtained by solving the equation

|a1-A|-0 or
2+51+6=().
The Eigen values are A, m-2and , -3
The Eigen vectors aredetermincd from

[a1-A]P =(9]

Let the Eigen vector associated

with the Eigen value
h=-2 be P
Then

d+1
These cquations are

2P1 -2p =0 and

These cquations cannot
P1-2 =0,
be uniquely solved to
10 assign any
value for p,,or p,, get the values p,
and get the other value. of and p,,. The methodis
Choose P1 so
=l that p
=1.Hence
Let the Eigen vector
associated with
the Eigen value , -3 be P, =
LP22
-2]
These equations are J=-3LP2 J

These equations P2-2P =0 and

value for P,, or
cannot
be uniquely solved Pn-2p» =0.
p,, the other value to get the values of p,,
can be obtained.
Choose and p,: Assigning any
p,,=2so that p,,= 1.
Hence P,
Example
4,2 For a square
sociated Eigen matrix
vctors.
Solution: determine the Eigen
The Eigen values and the
values are
obtained
by solving
the equation
 SImllarity Transformations
149

|a1-A|= 0
+32+2
, ).
Eigen values are =-l and
The

The Eigen vectors are determined from

,=-2.

[a1-A]P =[0] or
2
the Eigen vector associated with the Bigen
=.
Let
value 4, =- be P
-1] Ai|
LP21

|2
These equations are

-P-P) =0 and 2p +2pP1=0

p,,
Choose
=lso thatp,, =-1. Hence P
Let
the Eigen vector associated with the Eigen
value, =-2 be

-11P|=(0]
These equations
are

-2p2 -P =0 and 2p +P =0 5

Choose
P212=1 so that p,, =-2. Hence P:
Lxample 01 0]
4.3 For a
2
,determine the Eigen values and the
Square matrix A
=3 0

scatedEigen -12 -7-6eta oit

vectors.
Soution:.
The Eigen
values are obtained by solving the equation

|21-A|=0 or-31 -2=0

12 7 A+6

aa? +61+14)+(-31-18+ 24) = 0

150 Modern Control Theory

A
+62 +l1d+6= 0
(a+1)(2+ 2)(4+ 3) =0.
= =
The Eigen values are d -1, A -2 and A=-3.
The Eigen vectors can be determined using the cofactors of any one row of| ,1
-1
(a1-A]=-3 -2
12 a+6
The cofactors of the first row of [I-] are

+(2+62+14)| 2+61+14|
,=-34 +6)
31-6
(-21-124) |-122–21
9

For =-1, P =-9, this can be reduced to

R=
|-9

For
i=2 P,=-12, this can be reduced to P,:
3

For 2=-3, P =-15|, this can 1

be reduced to P=
15

4.4 GENERALIZED
EIGEN VECTORS
When the Eigen
values of a square
of Eigen vectors. matrix are distinct,
When the matrix is said
Eigen vectors. the Eigen values are to have a
Consider a repeated,
repeated Eigen (3x3) square matrix the matrix will not have a fullrul
set of
value. Corresponding with the Eigen
mined using a to = h and values A=
the repeated
method explained
earlier. There d= h,two Eigen h,,, win
Eigen value is an vectors can be deter-
Let P, = Cofactors and is called a generalized additional Eigen vector
of any one row Eigen vector. correspou
of / -A|
P=(Pla-,
 Similarity Transformations
151

generalized Eigen vector corresponding to the repeated

The Eigen value =
can be
hainedas

If there
are repeated Eigen values, the Eigen vector and
m

the remaining generalized

vectors can be determined as given below:

P= Cofactors of any one row of [2/-A

qm-l
(m-1)! dzm-Tlil
P, P, P, .. P, are generalized Eigen vectors.
Generalized Eigen vectors can also be found using the following
relations:

ngle
C4Eigen4A Fora square
bhton.
vectors. matrix A
0=,-!
2]
determine the Eigen values and the associ

The Eigen
values are obtained
by solving the equation
 152 Modern Control

.
2+1 -2 or (2+1)(2+1) = 0
Jal-A|=0

are =-land , =-1, with repeated Eigen

values.
The Eigen values

one Eigen value

Let the Eigen vector associated with
, =-1 be P, -
Cofactors of the second row of [/-A] are

The generalized Eigen vector corresponding to the repeated Eigen value A, =- is

Example 4.5 For a square matrix |-3

ated Eigen vectors.

A =determine
-2
the Eigen values and the associ
1

Solution: The Eigen values are

obtained by solving the equation

|a1-A|=0 |2+3 -2
or 2+21+1 = 0.

The Eigen values are

One Eigen vector

h=-land h, =-1, with repeated
Eigen
corresponding
to =-1 is obtained as the values.
cofactors of any one row
-2
Cofactors of
the first row 2-1|

The generalized
Eigen vector
corresponding
,
to =-1 is
 Similarity 'Iransformations 153

Famplo 4,.6 the lpe values and the ascinted

vvtora
ligen
Sohtlon: The igon values nrO obtalned by solving tho cquntion

or +42+4 -0 or (4+ 2 m0.

The ligen values are A

2nd A, m2, with repcatcd igen valuces,

(a1-A] [a+5 -1
9

Cofnctors of the first row of [l - A] are

The Eigen vector associnted with Eigen valueh =2 is

The generalized
equation
Eigen vector corresponding to =2 can be obtained by solving the
Ta+5 -1
2

fa+5 -1 1

-JLP2!

3p2 - P22 =3 and 9P2-3Py =9.

Let
P,=1sothat p,
=0 and |

[-2 4 01
Example
4.7 For a square matrix 0, determine the Eigen values and the
A =
0 -2
ASSociated
Eigen vectors. L-4 4 -2
154 Modern Control Theory

are obtaincd by solving the cquation

Solution: The Eigen values
|a+2 -4 ()
or (a+2)
Ja1-A|=0 or A+2 0 ()
-0
|4 -4 a+2

The Eigen values are , -2,, =-2 and ay = -2, with repcated Eigen values

To find the Eigen vector corresponding to , =-2,

H(2+2)°) (2+2)?
Cofactors of the first row of (2/-A is P =-{0} 0
+{-4((4
+2)}-4(2+2)|
(2+2)?]
-o
-4(2+2)|
An Eigen vector with
Ji-2
all-zero elements is not
acceptable.Ne
Cofactors -4(4+2)} [4(2+2)
ofthe second row of
(4I-A|is =|+{(2+2)) (a+2)?
-4(4+2) +16;
4(2+2)-16||
[4(2+2)
R= (2+2)
|4(2+2)-16|
A generalized Jim-2
Eigen vector
corresponding
to d, =-2 is
f4(2+2)
4
(a+2)2
da 2(2 +2) |
4(2+2)-16
4
Another generalized
Eigen vector

4(d 2) +
corresponding
to , Jdm-2

-2 is
L4

21 da? (a+2) 07
L:
 Similarity Transformations 155

Example 4.8 Repeat Examplee 4.7 using the definition

of Bigen vectors.
Solution:
|-2 4 0

=0
A

-2
|-4 4 -2
The Eigen values are obtained by solving the equation

|+2 -4 0
|a1-A=0 orl 0 1+2 =) or (2+2)' =0
4 -4 d+2
The Eigen values are A =-2,h=-2 and y =-2, with repeated Eigen values.

To find the Eigen vector R=|

Pu
P21 corresponding to , =2,
LP31

[a+2 -4
0 A+2 0 |Pai-[0].
4 -4 A+2LPJ
which gives the following equations:

(0)P1 -4P1 +(0)P =0 (1)

+
(0)P1 +(0)Pz1 (0)P1 =0 (2)

4P1 -4p21
+
(0)P =0 iLjeseeofl t adt etu t
(3)
These equations cannot be solved uniquely.)
From equation (1), P,,
=0. Pz P p
Substituting the value of in equation (3), =0.Let =4.
0
Hence
P =|0
4
Eigen vector corresponding to , =-2 is obtained from solving the equation

2+2 -4
0 A+2 0 [B]=-A]. s bs

4 -4 1+2Jjm-2
 156 Modern Control Theor

fi+2 -4
i+2
4 -4 i+2].lPu
following equations:
whih givs the
(0)r:-4P2 +(0)P: =0
(4)

(0n: +(0)P +(0)p2 =0

(5)

4:-4p2 +(0)P: =4 (6)

From equation (4). p,, =0.

In equation (6), let p,,=-l and
P
=4.

=
Heoce P 0

Another Eigen vector corresponding to =-2 is obtained from solving the equation

[2+2 -4 0
2+2 [R]--3]
4 -4 2+22-2
[2+2 -4 0
|A3
0 2+2 0
|P23
4 -4 2+2,2LPss
which gives the following
equations

(0)A3 -4P3 +(0)P3

=1etaag S()
(0)py +(0)P +(0)Py,
=0 (8)

4pis -4p +(0)P)3

From equation =4 (9)
(7), p,, =
In equation 0.25.
(9). P,,-1.25and let p,,=
0.
-1.25)
Hence P =0.25
0
 Similarity Transformations 157

TRANSIFORMATION TO DIAGONAL
45
DIAGONALIZATION CANONICAL FORM OR

a
system matrix are distinct,
values of
the Eigen
If then the system can be transformed into
diagonal caanonical form. Consider a
state model

X= AX + Br

y=CX.
Caneider that it is transformed into diagonal canonical form with
the transformation
X = PZ.
Then from equations (4.4) and (4.5), the new state model is

Z=[P'AP]Z +[PB] ,
where
[P'AP] is the new system matrix and [P"B]is the new input matrix.

y=[CP]Z
The transformation matrix P, composed of Eigen vectors, transforms
model into diagonal
the given state
canonical form. Consider that for a third-order system, the Eigen values
are distinct
and are , and h. The Eigen vectors associated with each Eigen value can
be obtained
from [2,I-AJ P=[0], where P is an Eigen vector associated with an
.,. Eigen
value

From this equation, .

AP =
Let the Eigen vectors associated with the Eigen values ,
4, and be
|P1
P21 ,P=|P2 and P
=P3 respectively.
LP31
LP32 LP33J
Now AP =hR, AP, =
h, and AR =43.

P2
A
P22
LP32
 Theory
Modern Control
158

Combining these cquations

P21 P22 P23

A|

0 0
|Pu P2 D3 ||
=|P21 P22 l23
LP31 P2 P3 0 0

Pi P2 P3
P22 P23 (4.6)
=|P21
LP31 P32 P33

The matrix A is a diagonal matrix with the diagonal elements being the Eigen values:

A=0 0

From equation (4.6),

PA= AP or A=[P'AP] (4.)

This shows that a system
matrix A can be transformed
formation matrixP into diagonal form using the trau

P1 P12
P3
P=|21 P22 P23
LP31 P32
From equation P3
(4.7) can
of Eigen it
vectors associated be observed
that usingthe transformation
diagonal form. with each composed
Each column of Eigen value, a matrix P,

be transformed Pis an Eigen system matrix can into

into diagonal vector. be transformed
matrix is also canonical form. Thus using Eigen model can
known as a The matrix P which vectors, a
state
modal matrix. system
diagonalizes given a
 Similarity Transformations 159

DETERMINATION
4.6 O
|OFSTATE TRANSITION MATRIX USING
TIIE METTIOD DIAGONALIZATION
a
third-order nystem, let Atate cquntion be

XAX (4.8)
The solution to thin state cquation is

X
=eX(0). (4.9)
Le the system matrix A be transformed into a diagonal
matrix A by the transformation

X= PZ (4.10)

Pisthe transformation matrix composcd of Eigen vectors.

Consider that the transformed state cquation is

Ž= AZ. (4.11)

The solution of this cquation is

Z=eh z(0), (4.12)
where

[eo 0
(4.13)
e

The method of obtaining eM is explained in Section 3.5.l.

From equations (4.9) and (4.10),

PZ=ex(0) = e"PZ0)a sty

Or
Z =(PePZ(0). (4.14)

Comparing cquations (4.12) and (4.14),

 Theory
160 Modern Control

o p-!
eA=|Pe^p-'}= Pl0 (4.19)
Or 0
0

Using equations (4.15),

the state transition
matrix ecan be
determined, For an
n-h
system with distinct Eigen
values
order
0---0
0 0---0
0 es 0- --0

|0 0 0---e
Example 4.9 A state model for a linear system is

y=[2 -1]X.
Transform the system into diagonal canonical
form. Also find the state transition matrix.
Solution:
The Eigen values are obtained
by solving the equation

Ja1-A|=0 or
2+2
or
|-1 a+3=0 2+51+4 = 0 .

The Eigen values are

canonical form.
,
4 =-1 and =-4, and are distinct. Hence
transformed to diagonal the system matrix can u
The Eigen vectors are
determined from

The matrix
of cofactors of
the second row is
2
The Eigen value
corresponding
to 4 =-1is
The Eigen vector
associated with
the Eigen value , =-4 is
 Similarity Transformations 161

For the transformation matrix

Check:

The state model in diagonal canonical form

is1
Z =P'APZ +[P'By
and y = (CPIZ

CP -(2

Hence

and y=(3
3]Z. e o 1
e

The state
transition matrix
is

-2e-e
 162 Modern Control Theory

tnodel for a system is

Hxample 4.10 A state

diagonal canonical forrn. Also find the state tranitmna

Transform the given system into
trix

Solution:
The Eigen values are obtained by solving the equation

or 2 +32+2 =) or (2+12+2) =
0.

The Eigen values are -] and h-2, and are dístinct, Hence, the systen matrix can b
transformed to diagonal canonical form.
The Eigen vectors are determined from

The matrix of cofators of the first row is

The Eigea value coresponding to a,

=-iish-
The Eipen vector associated
-R
with the Eigen value
h=2 is

For the transformation

natrix

Check:
 Sinilarity Transformations 163

The state model in diagonal canonical form is

y=(CP]Z

CP =

Hence the state model in diagonal canonical form is

and y=|1 0Z. bo io

For finding the state transition matrix

2e

-3e21

6e--6e24
-2e- +3e-t
Example 4.11 A linear time-invariant system is represented by a state model

-4 1
0
X= 0 -3 1
|X+|
0
-2 2|

y=[2 -1 1]X.

anstorm the model into diagonal canonical form. Also find the state transition matrix.
Solution:
The Bigen
values are obtained by solving the equation
t64 Moaem Contnol "heory

1() (or (2+4)(1+3)(7+2) =).

|-A-0
) A+2|

Theiigcn values are d, -2,dy

3 and A, 4
and arc distinct. Hence,the system
e
matri\ an transrmed to diagonalcanonical form.
I+{a+3)4 + 2)) -{-A+2)) +{1)

Alàl-A]=-40} +{(a+2)(A+4)) -{Ha+4))

-{0) +{(2+3)(Q+4))|

f(a+3)(2+2) (2+2) 1

0 (a+2)(2+4) (1+4)est
0 0 (a+3)(4+4)

The matrix of cofactors of the third row is P =(4+4)

L(a+3)(4 +4)|

1
Eigen vector corresponding to =-2 is P =| (2+4) =|2
|(2+3)(4+4)Ja=-2

Eigen vector corresponding to , , =-3 is =|(2+4)

(2+3)(2+4)
l-3
Eigen vector corresponding to , =4 is =|(2+4)
(a+3)(2+4)
0
Hence P
=(R B P]=|2 1
11
0
P=2 1
o=-2,
2 0 0
2 0 o

+{0} -{0) +{-1)|

-{0} +(-2}
0 -1] To 0 1

-(-2} 0
ia2 o22
|+{-2} -(-2}
+{-1)-2 2-1 N
2 -2 1
 (0

Check: 0 22
2 -2 || 2 I 0
2 -2
(0 1
|-4 1

() 2
2 -2 (0
-3
2 -2 0 0 -2||2 0 0

1
1| -2 -3 -4
() 2 -2 | -4
- -3
2 -2 1
|-4 0 0

-2 0 0
0 -3 0 A.
0 0 -4
The state model in diagonal canonical form is

and
y=(CP]Zz.
[-2 0
P'AP=0 -3 0
0 0 -4 |2 -2
[1 1

CP=[2 -1 1
2].
2 0 0
Hence, the state model in diagonalcanonical form
isae Hsh l
(0

0 -2
0 -3
y= [2 1
2]Z

e0 0

) 0
e
 166 Modern Control Theory

matrix is
The state transition
0 11
(0
2 -2
1
0
e- |2 -2
2 0 0jo
1 1 0
|1
1
0|0
2e 3-2e-31
|2 0 0 2e- -2e

-e 2

=|0

0 e-2

4.7 TRANSFORMATION TO JORDAN CANONICAL FORM

If the system matrix has repeated Eigen values, then it can be transformed into Jordan canoni
cal form. The system can be represented as

X=JX,
where J is the system matrix
in Jordan canonical form.
The transformation matrix P is composed
These are found out using the methods of Bigen vectors and generalized Eigen vector.
discussed in Sections 4.3 and 4.4 respectively.
transition matrix for the Jordan form is e The stae
in Section 3.5.2. and can be determined using the method explaneu
The state transition
matrix 1s given
byois ifaoa a1
e =
Pe'p-l, sl09
Example 4.12 For a
state model

,
y=[1 2]x
determine
the Eigen
into Jordan canonicalvalues andthe associated
form. Hence, Eigen vectors. model
find the state Transform the given State
transition matrix.
 Similarity Transformations
167

Solution: The Eigen values are obtained by

solvingthe equation
+3
J1-A|= 0 2 or A +
22+1= 0 or (+1)
=0.
values are å
The Eigen
transformed to
=-l
and h=-1, and are
repeated. Hence, the system
can be
Jordan canonical form. matrix
The Eigen vectors are determined from

The matrix of cofactors of the first row is P. =

The Eigen value corresponding to =-1 is P

The generalized Eigen vector corresponding
to =-1 is

2p2 -2p2 =2 and 2p2-2p, =2.

Let
=0 so that p,,=1. Hence P,
p,
=

The transformation
matrix is
P -2
11
JP|=2

Check:

1]:

CP =1
 168 Modern Control Theory

The state model in diagonal canonical form

is

y=[CP]Z=[-6 1]2

2te-e-2te]
-2e-!
e-2te 2te-!
-2te-! e+2re
Example 4.13 For a state model

y=[2 -1]X,
determine the Eigen
values and the associated
into Jordan canonical Eigen vectors. Transform
form. Also find the given state mode
the state transition matrix.
Solution:
The Eigen values are
obtained by solving
the equation
a1-A|=0
or
2+42+4=0 or (2+2) =0.
The Eigen values are
be transformed h
=-2 and , =-2
to Jordan canonical and are repeated.
The Eigen vectors form. Hence matrix can
are determined thesystem
from

The cofactors
of the first row
is P=
 Similarity Transformations 169

The Eigen
vector corresponding to h
=-2 is R

gencralized Eigen vector corresponding

The to =-2 is

The transtormation matrix is

|P|=1

Check:

The state model in diagonal canonical form is

Z=[P'APZ +[PBr
y= [CP]Zz

CP =(2

Hence,

le21
|0
The state
transition matrix for the given state model is

e=PeP -e--te21|

te-21
 170 Modern Control Theory

Example4.14 A state model for a lincar system is

-6
X-12
-8

2 -1]X.
Transform the given state model
into Jordan canonical form. Also find the statetransition
matrix.

Solution:
The Eigen values are obtained
by solving the equation

J+6 -1
al-A|=0 or 12 or
-1=0 2+62 +122+8=0 or (A+2 =0.
8 0

The Eigen values are

matrix can be transformed h=-2,h=-2 and y =-2 and are repeated.
to Jordan canonical Hence the system
The Eigen vectors are form.
determined from

[a+6 -1
(1-A]=| 12

8
The matrix
of cofactors +0)
of thethirdrow is 1

P =-{-(+6)}
=+6
+62+12)
+62+12
The Eigen vector
corresponding ,
to =-2 is
P =+6
Generalized +62+12
Eigen vector
corresponding
to =-2 is

P,=2+6
da
2+62+12
24+6|
 Similarity Transformations 171

generalized Eigen vector

Another corresponding to h
=-2 is R=
24+6Ji
0
transformation matrix is
P=4 1
The t
4 2

0
+() -(0) +(0)| 1
01
1
=-(4) +(1) -(0)=-4 1 o
4 2
|+(4) -(2) +(1)) 4 -2 1

1 0 0|1 0 011
Check:
P'P=-4 1
0 4 1

|4 -2 1||
L
4
2
lo0 1|
1
0 0|-6 0|1 0 0

1
0| -12 0 4 1
0
4 -2 -8 0||4 2

|1 0 0|-2 1
0 -2 1
01
1
|-8 2

-8 Lo0 -2
Ihe state model in
diagonal canonical form is

Ž =(P-'APJZ +[P'B,
y=[CP]Z.

-2 1
0 1
0
0 -2 pB=4 I

0 0 -2| 4 -2

07
CP =[1
2 |4
1

4 2 1
 172 Modern Control Theory

-2
Hence Z=| 0 -2
0 0

-i]2

=0 le-2
e
0

The state transition matrix is e4 = Pe'p-l

-21
2! 1
0 0
4 1
olo -21 te2 -4 1
0
4 2
4 -2 1

07 e-44e +2e te--24

0|4e 44te
1
e-221e-2 le
4ea
-2es
e-241e-2 +21-21
-12te- 4 8
te-te
e2 e
+2te24
-81e- +8¢²e-21
-41?e-2
-4e2 le2 +21e-2
+
2te-+ 21e-21
4.8 TRANSFORMATION
FORM TO DIAGONAL FROM CONTROLLABLE
A
state model FORM CANONICAL
in controllable
If the Eigen canonical
values are form can
sociated distinct, be directly transformed
with each Eigen the transformation forn.
betransformed value. In matrix linto diagonal as
this case is
into diagonal of distinct Eigen composed of Eigen Vectorscan
form. If some values, the
Eigen values given state model
are repeated, transformation
the
 Simílarity Transformations 173

composed of Eigen vectors and generalized Eigen vectors,

matrix,is in whích case the given
can be transformed into Jordan canonical form. Consider a state model a system
state
model of
controllable canonical form
in 1
0
() 1

-1
[2l-A]=|0 -1

The matrix of cofactors of the first row of [aI -A] =|-{4,}

|4-2+4,)| (2+a,)
The matrix of cofactors of the second row of [I-A]=|+{l+a,)} =| 2+az)
-tq2+a}
+1) 1

The matrix of cofactors of the third row of [A/-A]=|-(-)|=| 2|

+(a)|
row
It can be observed that it is convenient to select the matrix of cofactors of the third
of
[/ -A] as the Bigen vectors.

Case 1: Distinct Eigen values

When
the Eigen values are distinct, the Eigen vectors are given by

P=4 ;i=1,2,3,.,n.
 Modern Control
Theory
174

Eigen values 4, and a, the transformation

system with distinct
For a thid-order
matrix is 1 1

P=| d·

Vander Monde matrix. For an n-th order system

transformation matrix
This
P
is called or
the transformation matrix, the Vander Monde
with distinct Eigen values ,2,
matrix, is
1
... 1
1

Example 4.15 Show that a state

model in controllable canonical
form with istinct Eigen values can form or in phase variable
Vander Monde matrix.
be transformed into diagonal canonical form using the
Solution: Consider a
third-order system
sented either in the
controllable
with distinct Eigen values ,,
and a. If it is repre
will be of the form canonical form or in
phase variable form.,
the system matrx

4 1

The characteristic
equation is

-A|=0
-1
Or A(2 +a,2) + a}+ao
Or =0.
The Vander
Monde matrix
is ()
 Similarity Transformations 175

0 1 1 1 1

AP=| 0

=|4

Using equation (1),

|4 1 1
|| 0 0
4
AP=| 0=PA.

AP = PA or A= P'AP
Hence, a Vander Monde matrix transforms a state model in controllable canonical formn
or in phase variable form to the diagonal canonical form.

Case 2: Repeated Eigen values

A System with repeated Eigen values can be transformed into Jordan canonical form. Con
Sider, for example, a fifth-order systemwith four repeted Eigen values. Let the Eigen values
be A4s A4,
4 and As the Vander Monde matrix is given by P=[R B P B].

Let P, =|1
 170

a system in controllable
canonical form is
Example 4.16 A state nmodel of

y'=(3 1]X.

Transform the state model into diagonalcanonical

form.

Solution: The Eigen values are obtaincd by solving the equation

-1 or +31+2 =0 or (2+1)(2+2) = 0.
|al-A=0
The Eigen values are , y
=-l and =-2 and are distinct. Hence, the system matriv
be transformed to diagonal canonical form. Since the given state model is in controlilt
canonical form, the transformation matrix is the Vander Monde matrix, It is given hy

[-2

Check:

CP =[3
The state model in diagonal
canonical form is

Example 4.17
A state model
of a system in phase
variable form
is

y=[3 -1]X.
Transform
the state model
Solution: into diagonal
The Eigen values canonical form.
are obtained
by solving
the equation
|a1-A|=0 or -1
2+6 2 = 0.
+62+5=0 or (2+1)(2+5)
 IOns 177

are
The Eigen values =-l and ,=-5 and are distinct. Hence,
transformed Ito diagonal canonical form.
Since the given state the system matrix can
be
transformation matrix is the Vander Monde model is in phase variable
forn, the matrix. It is given by

p-l

Check:

r CP =[3
1

The state model in diagonal canonical form is

2]
y=[4 81Z.

Example 4.18 A linear system has a state model in controllable canonical form given by
o
01 0]
1|+oy=1 2
x.
|-15 -23 -9
Transform the state model into diagonal canonical form.

Solution: The Eigen values are obtained by solving the equation

-1 0

|a-4=l0 -1 =0
|1s 23 2+9

or 1
+92 +232+15 =0.
a
+9+ 23) +15 =0
an values are 4 ==-1,4 =-3 and =-5 and are distinct, Hence, the state model
The Eigen
be transformed
Jable into diagonal canonical form. Since the given state model is in control-
canonical
form, the transformation matrix is the Vander Monde matrix,

1 1

P=|4 -3 -5
1
 1

-s|=(-75+45)-(–25 +5) +
(-9+3) = -16
9 25|

[+(-30) -(16) +(-2)1] 30 16 21

p=-(-20)
16
+(24) -(-4) -20 -24
l+(-6) - (8) +(-2)|
16
6 8

30 16 2 || 1 1
|1 0 01
Check: PP= 16 -20 -24 -4| -1 -3 -5=0 1 0
6 8 1
9 1

p-'AP=
30 16 21 0 1
0 1 1
1]
-20 -24 -4 0 0
16 -1 -3 -5
6 8 2||-1S -23 -9 1 9

30 16 21-1 -3
25t
-5 -1 0 0
-20
16 -24 -4|| 1
9 25
6 -1
0-3
2|| -27 -125|
-5
30 16 2|01 1/8
-20 -24
16|
6 8
1/8

CP = [1 2 1] -1 -3 -5 =[0 4 16].
L1 9 25
The state model
in diagonal canonical
form is
0 07
Z= 0
1/8 1

-3 0
0 -s
|Z+|-1/4|r y=[0 4 16]Z.
Example 4.19 1/8
A state model
of a system in controllable
canonical form
is trgt
y=[2
Transform
Solution:
the state model
into Jordan
-1]X.12
The Eigen values canonical form.
are obtained
by solving
the equation
 -1
2i-A=0 a+ 4| =0 or
1+ 41+ 4 = (0 or (a+ 2)
=0.
The Eigen
values are =
-2 and hy =-2 and are repeated.
transformed to Jordan canonical form. Hence, the system matrix
can be Sincethe given state model
is in controllable
canonical|form, the transformation matrix is the Vander
Monde matri.

Check:

CP =[2

The state model in diagonal canoniçal form is

[-2 1]
y=[4 -1]2.

Lxample 4.20 A state model of a system in phase variable form is

-
y=[1 2]X.
Transform
the state model into Jordan canonical form.
Solution:
The Eigen values are obtained by solving the equation

a1-A|=0 or = 0 or (4+3) =0.

=0 1+61+9
|9 +6
The , Hence, the system matrix
can
Eigen values are
bettransformed h=-3 and =-3 and are repeated.
form, to Jordan canonical form. Since the given state model is in phase variable
the
transformation matrix is the Vander Monde matrix.
 180 Modern Control Theory

Check:

CP=[1

The state model in diagonal canonical form is

y=[-5 2]Z.
Example 4.21 A state model a
of linear system is
01
X- 0
|+o y=4 -2 )x.
-18 -21 -8
Transform the state model
into Jordan canonical form.
Solution: The Eigen
values are obtained by solving
the equation

Ja1-A|=0
=0
18 21 2+8
2 +82 +212+18 =0 or (2+2)(2+3
The Eigen values are =0.
matrix can be =-2,h=-3
transformed to and
a=-3 and are
lable canonical Jordan canonical repeated. Hence, the system
form, the transformation form. Since the given
matrix is state model is in controk
the Vander Mondee matrix.
1

9)
P=-2 -3
4 9 -6| -3 l|=18–9-(12-4)=1
9 -6
 +9
-(-6) +1 9 66 1]
p-!.
=-(8) +(-6) -1 |=-8
+(-6) -(5)
-6 -1|
+(-1)|-6 -5 -1|
9 6 1 1
0] 0
Check: P'P=-8 -6 -1 -2 -3 1
0
-6 -5 -1|| 4 9 -6 0 1

9 6 1
0 0|1 1
0
p'AP =-8 -6 -1 0 1|| -2 -3 1

-6 -5 -1|| -18 -21 -8 4 9 -6|

9
-8 -6-1
-2 4
6 -3 -2 01
9 -6 0 -3 1

-6 -5 -1|| -8 -27 27 0
0-3
6 1

p'B=-8 -6 CP = [4 -2 -1]|-2 -3 1=[41 4].

-6 -5 -1 4 9-6|it
The state model in diagonal canonical form is

-2 0 0 1|7
Z=| 0
-3 y=[4 1
4]Z. st
0 0 -3|

4.9 TRANSFORMATION FROM OBSERVABLE CANONICAL FORM

T0 DIAGONAL FORM
A State model in observable canonical form can be transformed into diagonal formn, if the
Eigen values are distinct. If some Bigen values are repeated, the given state
model can be
transformed into Jordan canonical form.
Case 1: Distinct
Eigen values
Considera
fifth-order system, the state model of which in observable canonical form can be
Tepresented
as
1
0 0
1 (0

X =\-dh 1 0
X+b, r y=[| 0 0 0 0Jx.
0 1
by

L-4 0 0 0 0|
182 Modern Control Theory

is
The characteristic cquation
= 0.
2+ a,a' + a,l' +a,a' al+ay
+

are distinct. The transformation matrix that transiorms

P

Consider that the Eigen valucs

model in observable canonical form to diagonal canonical form can be determined
the state
as explained below:

P,=+al+a,
a+a, +ayi' +aydta
P, =(Pla=4 P; =[Plies

Case 2: Repeated Eigen values

For a fifth-order system, let the characteristic equation be

+a42 + a,
+a,1 +ad+aj =0:b ai isbom stsz T

Consider that the Eigen values are A,,4,4, l, and

ds. Four Eigen values are repeatel.
The transformation matrix P that transforms the state model
can be determined as explained
in observable canonical form
below:

P,=|+ayl+ a,

|A+a,' +a,l' +ay+a

da

P=(R P
Example 4.22 The state P P].
model of a system
in observable canonical
form is

y=[| 0]x.
Convert this model
into diagonal form.
 The given
state model is in observable canonical form.
Sohution:

[-3
[a/ - A]= 2+3
2

The characteristic
equation is
1? +31+2 = 0= 1' + a,+ uy

TheEigen values are A,

=-l and , =-2.
The Eigen values are distinct
sothat the given state model can
be transformed into diago-
form.
Rlanonial

The transformation matrix is

P=|

P'B= CP =[1 1].

The state model in diagonal canonical form is

y=[1 1]X.

Eumple
4.23 The state model of a system in observable canonical form is
[-6 1
X=|-11 0 +|2 r y=[1 0 0]x.
Convert
|-6 0 0 3
this model
Salution:
into diagonal | form.

|-6 11 07 [a+6 -1
A=|-11 0 1
[a1-A]=|11 -1
-6 0 0 6 0
 Theory
184 Modern Control

cquation is
The characteristic 1' +ay1
+
a,l + a
A'+61'+||A+6=)=

= -1,
,
-2 and h =-3.
Eigen values are A, state model can be transformed into diago-
The given
values are distinct sothat the
The Eigen
nal canonical form.

1+6
|A+64+11) |+62+11J

B=+6
+64+11-2
The transformation matrix is

4 |P|=-2.
3

H-1} -(-1) +{-1} 1

-1 1

4
+{-9} -(-3} +{-1} -3
1
-1 -6 |1 1
1|
1
4 5 4 3
JLO
1
11|-1 -2 -3 0
4 -2 2| -5 -8 -2 0
-3 1J-6 -6 0
-3|
1
-1
4
9 -3
 Sirrilarity TrarsioTratiors
15

0] 5
CP=1 0 4 3=( 1)
6 3

-1 2]
=0 -2 0Z+-6|r
0 0 -3
Eunple 4.24 The state model
of a system in observable canonical fotm ís

y=[|

Convert this model into diagonal form.

Soutioa:

The characteristic equation is

2+42+4=0= 2 +42+a, t 4 =4 =4.

Eigen values are
The
h=2 andso,=-2.
The Eigen values are repeated that the given state model can be transformed into
Jordan canonical form.

The transformation matrix is

IPl=1
P=(R

CP =[1

Ine state model in Jordan canonical form is

y=[1 0]x.
186 Modern Control Theory

a system in observable canonical form is

Example 4.25 The state model of
1
-7
0 y=[1 0 0Jx.
-15
|-9 0

Convert this model into diagonal form.

Solution:

-7 [2+7 -1 0
4=-15 [J-A]=|15
|-9 0 0 9
The characteristic equation is

a'+72'+152 +9 =0= +a,)'+al+ao a, =7,4 =15 and a, =9

,
The Eigen values are =-1,, =-3 and =3.
Two Eigen values are repeated so that the given state
model can be transformed into
Jordan canonical form.

P, =|+a, 2+7
2+a,d+a 2+72+1s

R=|1+7
P=2+7
|2+72+15
2+72+15

d |0
B=
d |22 +7
al'+12+15)|
Jiz-3
The transformation matrix
is

0
P=(R P ]= 69 4
1|Pl=(4-3)-(6-9)=4
3
 Tansrormations 187

+{1}) -{1} +{1}

p' -{-3) + {1} -1 11]
4 -{1)
+{-18} -{-6}
1 -1
+(-2} |
-18 6
1
-1 1||-7 0 |1 1 0
=
p'AP 3 1
-1 -15 1
4 4
-18 6 -2||-9 0||9 3
1
-1 1
|-1 -3 |-1 0 0
3 1
-1|-6
4
|-18 6 -2|| -9
0 -3 1=J
-9 0 -3
1
-1
p-'B= - 3

-18 6

1 1

CP =[1 0 01| 6 4 0].

9 3

The state model in Jordan canonical form is

0
|-1
0LX. i
X= 0-3 1
|X+| y=[1 1

0 0 -3|
Example 4.26 The state model of a system in observable canonical form is
[-6 0
X=-12 1
|X+|2|r y=1 0 0]x.
-8 0
Convert this model into diagonal form.
Solution:

-6 1 07 [a+6 -1 0

A=-12 0 1 [2/-A]=|12 -1
-8 0 0| 8
The characteristic
equation is
+62? an
+122 +8 =0=2 +a, +
aj2+do ay
=6,, =12 and =8.
188 Modern Control 1nory

are å; -2,, =-2 and =-2

(2+2)' =0. The Eigen values model can
are repeated so that the given state be transtormed
three Eigen values
The
Jordan canonical form. ina
1
P, =|+
az =|+6
[4+a,2 +4 +62+12|

P=+6
a+61+12Jj--2

d
+6
da
a+64+12-2 |22+6J;
1
1
d
2+6
2! da? 2! da
2+62+12Ja-3 22+6J

1 0
P=[R R]==4 1
0
4 2

+{1} -0} +{0}

p=4} +{1} -{0} - -4
L+{4} -{2} +{1} L 4 -2
0
p'AP=-4
1
0 1
1
0| -12 0
4 -2 1| -8 2
1
0 0|-2 1
0
-4 1
-8 2
4 -2
-2 1=J.
-8 0
o-2
0 )
p'B=4 CP =
0 0
=[! 0
0).

4 -2 ([10 1 0

2
 Similarity Transformations 189

staate model in Jordan canonical form is

The

-2
-2
y=[l 0 0]x.
0

PROPERTIES OF SIMILARITY TRANSFORMATION

40
Eigen
values: The Eigen values of a system are not
L lwarisne of afected by similarity
rmation. For a system matrix A, the Eigen values are obtained by solving the equation

-4=0
The system matrix fromthe transformed state model is [P'APland the Eigen values are
akulated by solving the equation

Ja-pr4P|=0.

-pP-PAP|
p=p'IP-P'AP|

-l4/-a||

invariance of Eigen values in similarity transformation.

Ihis shows the
samne even after simi
hvariance of transfer function: The transfer function remains the
laritytransforimation.
Consider a state model
 190 Modern Control Theory

The transfer function is given by

Y() =
C[sl- 4]B.
R(s)
The state nnodel for a transformed system can be represented as

Z
=[P-'APZ+P-Br and y =(CPJZ.

The transfer function is given by

[CPIsl-p-AP)PB]2o .rotaa
=
[CP|sP'P-p'AP]-'PB]
=[CPI[P-'sP- P'AP]-p-'Bzhn msie sdt

=[CP[P"'sIP-P"AP]P-'B]
=[CP[P(s-A)PT'[P'B]
=[CP
P(s1-'PP-'B]
=
C[sl-A'B.
This shows the invariance
of transfer function in
similarity transformation.

4. The trace
of [PAP]is equal to the trace
of A.
SUMMARY
State space models
are generally
venience
of design and analysis transformed into canonical
The method of control systems. forms for simplicity
as similarity of transforming and cuu
non-canonical
transformation. state models
The matrix intò canonical
that transforms a forms 15 ku
transformation given state
matrix. model
into a canonical
The same
system can state model is called the
Eigen values have different
and the same state space
For a square transfer function. models,
matrixA but these models have the same
AP = with an
P is known as Eigen Eigen value , a
For every
Eigen value, vector. column vector relation
P
If an there that satisfies the
n
xn matrix has is an associated
pendent
Eigenvectors. distinct Eigen alues,Eigen vector.
it
is said to have a inde-
full set of n linearly
 matrix has distinct Eigen values, it can be
transformed into diagonal
form by
similarity transfornation,
a matrix has a set of m repcated Eigen values, m<n,
n Nn
Ifanlinearly it does not possess a full set
independent Eigen vectors. Therc are (n
otn
+
-m
1) Eigen vectors associated
distinet Eigen values. The remaining (m- 1) Eigcn vectors are called
withthe the general-
vectors,
izd Eigen
transformation matrix that is composcd of Eigen vectors
The transforms state model
into diagonal canonicalform.
A
square matrix with repated Eigen values can be transformed into Jordan canonical
of state model.
fornn

he transformation matrix that transtorms a state model in controllable canonical form

diagonal formis the Vander Monde matrix,
to

PRACTICE PROBLEMS
pP41 Determine the Eigen values and the associated Eigen vectorsfor the square matrix

0
PP 4.2 Find the Eigen values and the associated Eigen vectors for

PP 4.3 Determine the Eigen values and the associated Eigen vectors forgnii Si,11
1 0
3

PP 4.4 For a system matrix find the state transition matrix.

-1 -1
PP 4.5
For a square matrix determine the Eigen values and the associated
Eigen
vectors. Diagonalize given matrix, and hence find the
state transition matrix.
the matrix
PP 4.6
Find the Eigen values and theassociated Eigen vectors for the system

PP 4.7 matrix
Find Eigen vectors for the system
the Eigen values and the associated

[-7 1
0
1

A=-16 0

-12 0
 Solution of State
3 Equations
31 SOLUTION OF STATE EQUATION IN S-DOMAIN
Consider that a system is represented by a state equation

X() = A
X()+BR(t).
Taking Laplace transform
of the state equation and assuming X(0) as initial state vector,
sX(s)- X(0) = A X(s)+ B R(s)

[sI-AJX(s) = X(0)+BR(s)
X(s) =[sI-AJ X(0)
+[sI-AT'B R(9). (3.1)
Taking inverse Laplace transforms, the state vector
X(t) can be obtained.
The output that can be obtained fromn the
equation is
Y() = CX(t)+ D
R().
Taking Laplace transforms,

Y($) =CX(9) +D
R(s).eel sat (3.2)
D)
Substituting equation (3.1) in equation (3.2),
esxiu so 0 h
HS
Y(s) =C[s I-
A' x(0)+{C[sI-A' B+D} R(S).l 45 (3.3)
Taking inverse Laplace transforms, the output
y(t) can be obtained. The first term on
gnt hand side of equation (3.3) is generated by the initialstate X(0)only and is independent
e
inputr(t). Hence, it is called the Zero Input Response (ZIR).
right hand The second term on the
side of equation (3.3) is due to
state
X(0). Hence,
the input r(t) only and is independent of the initial
it is called the Zero State Response (ZSR). If D=0, then
Y(s) = C[sl-AJ X(0) +
C[sl-AJBR($).
 =
Zero Input Response: YzR ($) C[sl-A]' x(0)
=
Zero State Response: YzsR ($) C[sI-AJB R(s).

Unit Impulse Response

Consider a SISO system. For unit impulse input, r() = 6() so that R(s) =l.
Hence, the zero state response of the system for unit impulse input is
Yzsa(s) = C[sI-Af' B.
Response of the system is
Y() = Yyus (s) + Yrse (s) = C[sI-A' x(0)+C[sI-AJ'B= C(sl-Afl [X(0)+A1
Taking inverse Laplace transforms, the response y(t) can be obtained.

Unit Step Response

Consider aSISO system. For unit step input, r() =1 so that R()=O OTD0E
S
Hence, the zero state response of the system for unit step input is s
et b

Ysa($) =
C[s-Af'B.
Response of the system is

Y($) = C[sl-Af' X(0)+C[sI-A' B

=C[sl-AJ" *x09-cu-ara (3.4

YzIR (S) = C[s[-Ax(0)A 8 o

ie B5

36
Taking inverse Laplace
transform of equation (3.5) and equation (3.6), the ZIR and
ZSR in time domain can be a
obtained for unit step input
and hence the total respons
Example 3.1 The system
matrix of a linear time invariant system is given as
and the initial state is
A=-3
X(0) =|
2
. Determine
the state respo
response of the system for zero input
Solution: The resolvent
matrix is

-1 1
[s+4 1

s+4| s +4s +3
 Solution of State Equations 105

is
response for zero input
The state
XzIR() =|+l - AJ xo)
2s +7
(s+1)(s+3)
(s+6)
(s+1)(s+3)
time domain solution can be obtained by taking inverse Laplace transform of this, using
partial fraction
expansion.
(5/2) (1/2)
XzIR(s) = S+1 S+3
-512) (3/2)
S+1 s+3
Taking inverse Laplace transform of cach element.

Example 3.2 A linear time invariant system is described by the state equation

y)=[1

If the initial state vector is X(0) = find the Zero Input Response and the Zero State

Response due to a unit impulse input. Also find the total response.
Solution:

The resolvent matríx

is

[s+3 2

+4s+5-1
1
2

Cp-AT.l!
 1

s+3]
(s+2)2 j+2
The zero input response is

Yzn (s) = C[sl-Af (0)

2.s +4 2(s+2)
(s+2)++2 (s+2) +1 (s+ 2)+1I
Taking inverse Laplace transforms, the time domain solution is

zir() = 2e-21 COst.

For a unit impulse input r() = 8(t), R(s) =1. The zero state response is
YzsR(s) = C[sI-AJ" B.
1
2s +5
(s+2)? (s+2) +1

2(5+2) +1_2(s+2)
2(s+2) 1

(s+2) +1 (s+2) +1 (s+2) +1

Taking inverse Laplace transforms,

'zsp() = 2e-2 cost+e-21 i

sint.
Total response of the system is

y0) = 2e-21 cos! 42:-2t Cost +e~21 sint = 4e-21

sint.
Example 3.3 For the system
of Example 3.2, determine the Zero Input
State Response due to a Response and the Zer
unit step input. Also find response.
the total
Solution:

i]x X(0) =
The resolvent matrix
is

)-1-4' -" s+4s+S-1

[s+3 2

s+1|
 Solution of State Equations

2
s+3]

The zero input responsc is

/Yzn (8) = C|sl- Af'x(o)
1 2] 2s +4 2(5+2)
(s+2)' +1 (s+2 +1
is
Taking inverse Laplacc transforms, thc time domain solution
cost.
Vznl) = 2e

The input is r() = 1, a unit step input so that R(G) = 1/s. The zero state response is

YzsR(s) = C[sI-Af' B R(9)

2s +5 1 s+2
s{(s+2) +1} S (s+2)' +1 gt. irs
Taking inverse Laplace transforms, the zero state response is

YSR() =1-ecost
The total response is
= cost+1-e21 cOst
= YaR
() yzsr() 2e
+
v)
=1+e-2 cost

Example 3.4 A linear time-invariant system is described by the state equation

If the initial state vector is x0 find the Zero Input Response and the Zero State
Kesponse due to a unit impulse input. Also find the total response.
Solution:
The resolvent matrix is
1 1

[s+3
s+3s+2|-2 s+3
1 1

c[sl -A]=R4s+2 S+3 (s+1)s+2)

The ZIR is
=
C(sl- A x(0)
Yr
1 S-1
(s+)(s+2)ls (s+1)(s +t 2) 4

2+
-2 3
s
S+1 +2
Taking inverse Laplace transforms, the ZIR is

Va() =-2e-+3e-.
For a unit impulse input r(t) = ö), so that R(9) =1. The zero state response is
=
YysR(s) C[sI-ATB n
zest alqsa
1
1

(s+i)(s+2)
(s+1)(s +2)
1 1

(s+1) (s+2)
Taking inverse Laplace transforms,

=
Vzsn(1)
e-e2
Total response of the system is

yl) =(-2e+3e)+(e-e-2)=-e+2e-2
Example 3.5 For the system
of Example Zer
State Response due to a 3.4, determine the Zero the
unit step input. Also Input Response,
find the total response.
Solution:
 matrix is
The resolvent
|S+3 1 1

oG)=(s1-AJ-2 2+3s +2-2

1 1

s+3s+2! s+3 (s+1(s+2)

The ZIR is
=
Yn C(sl- Aj x(0)
1
s-1
(s+I)(6 +2) (s+1)(s+2)
S-1 -2 + 33
+3s +2 s+1 s+2
Taking inverse Laplace transforms, the ZIR is

VR() =-2e+3e-2
The input isr() = 1, a unit step input so that R(s) = 1/s. The zero state response is
YzsR =C[sl-Af' BR(S).

s(s+1)(s +2)

1 11 1;
11
s(s+1)(s +2) 2s (s+1) 2 s+2
Taking inverse Laplace transforms, the ZSR is

sal) =-ette 2 2
Total response is
1
=
y() Yn (9)+yasn() (-2e"
=
+3e")+(;-+e)
1

Example 3.6 A state model for a linear system is given by

1 01

X= yt) =[4 1
ox.
-18 -21 -8|
 110 Modern Control Theory

a unit step input,given that X(0) =[| 0 0r

Determine the response due to

Solution:

-1S -21 -8

)=[4 ox X(0) =[ o
o.
The resohent matrix is

-1
(sl-A' =o S

1s 21 s+8

1
+(s +&s + 21) -{-(s+8)) +1
+ -18 +s(s+8) -{-s}
+&s 2ls +18
-18s}-{21s+18} +s²
(s+&s+21) (s+8) 1

-18 s(s+8)
(s+2\(s +3)
-18s -(21s+18) s

(+8s+21) (s+8)
c[J-af' = (s+2)(6+3j?l 1
0|-18 s(s+8)
-18s -(21s+18)
(s+2)(s+ 3)2|(4s +32s+ 66) (+12s+32) (s+4)|·
The ZIRis

YR = C{sl- Af X(0)

sts+2)(6+3)4s +32s+ 66) (s+12s +32)

4(s +8s+16.5)
(s+ 2)(s+3)
18 14
6
S+2 S+3 (s+3)?
 Solution of State Equations 111

is
Taking inverse Laplace transforms, the ZIR
zr() = 18e-2-14e-316e-31
The input is r() = l, a unit step input so that R(s) = 1/s. The zero state response is
YzSR = C[sl- Aj' BR(s)

321(4s+32s+ 66) (s+12s+ 32)

s(s +2)(s+

S+4
s(s+2)\(s +3)

21
9 s (s+2)*9(6+3)"3 (s+3
Taking inverse Laplace transforms, the ZSR ise )Y

Vsal)=-ee.
The complete response is

= + Pzsa() =(18e--14¢-3,
J) Vn ()
G+17e-2-19,-3 17,

3.2 SOLUTION OF STATE EQUATION IN TIME-DOMAIN

The state equation is

X()= AX()+B R()

or

X()-A X() = B R().

s3 bitsi yds lhdt erve st

Multiplying es
both sides by e,etth (hi
e-Al X()-eAAXx() = eAB R().iotesl ssic ot tron
This can be written as
112 Modern Control Theory

t, within limits 0 to t,
Integrating both sides with respect to

-Al xo -f-aop R(O) d

e-^x()-X(0) = je^B R(Ø)d0

0

e-A X)= x(0)+je-A B R(O)

d0.
0

From this equation, the solution for the state vector is

X(0)
=e^X(0)+jA(-9)B R(O)d0.
0

The first term on the right hand side is the ZIR and the second term is the zSR.
The output is
Y(t) =CX(t) +D
R().Si glas23a
With D =0, the output equation is

Y() =CX() = Ce^ x(0)+CjeAl-0) B R(0) d0 .

For a SISO system, the output
equation is
y0) = Ce^l x(0)+CjeAl(-) B r(0)de

y()= zn()+ Vzsa()

Ynl) = Ce^ x(0)
and

zsal) = CieAt-0B r(O)do= jceh-0B

r0)de.
Consider equation
(3.7) with zero input. Then this equation can
be written >
X(0)=e
This shows
x(0).
that the initial state X(0)
sition in state is driven thistr
known as state
is carried out by to a state X(0) at
the matrix exponential any time t and
transition matrix. e. Because of
this
property

3.2.1 Unit
Impulse Response
For unit impulse
input, r() = St).Hence, input
the response of impulse
unit
the system for
 J0) -Ce^ x(O)+jceN-90 B Ø)40.

Cex(0)+C eB=CeX(0) + B]
(3.11)

3.2.2 Unit Step Response

=1
For unit step input, r(4) :1;0.Hencc, thc response of the system for unit step input is

D) =Ce^x(0)+{CeN-9)B do

=CeAi x(0)+C[-A-jeAleA-jB
ten
=Ce X(0)+CA-!eAl-Alp
-CeAl x(o) +CA-'e -JB (3.12)

-YzIR() +YzSR(9)
Example 3.7 For the system of Example 3.4, find the state transition matrix and hence the
unit impulse response.

Solution:

y() =[1 o]x

The resolvent matrix is

1]
s+3s+2-2 s+3|
1
1 2 1
+
(s+1)(s+2) (s+1)(s+2) (s+1)' (s+2) (s+1) (s+2)
= . 2
-2 s+3 -2 2
L(s+1)(s+2) (s+1)(s +2) (s+1) (s +2) (s+1) (s+2)
The state transition matrix is

+2e-24 2e-e-2
Kesponse due to unit impulse
input is
y() =Ce4 x(0)+ce4p
 N)=CeN(0)+ B]
+2e-2r
-2t

+2e-21
-2e +2-21

=+2e-2
Example 3.S A linear time-invariant system is represented by a state
diagram shown in Figure
E3.S. Determine the state transition matrix, the
ZIR, the ZSR due to a unit step input and
the complete response. Assume that X(0) 1].Obtain the solution in the
time domain

2
+

Figure E3.8 State diagram for

Example 3:8
Solution: Select the output
of each integrator as a state variable as shown
From this diagram, the state in Figure E3.8(a)
equation and the output equation can
be written as

) =[2

X(0)
=[ '.
The system matrix A is
diagonal and the Eigen
values are a,
=-2and
h=-.
S+2 1
[s+3 S+2
(s+2)(s +3)| 0
s+2
 3

Figure E3.8(a) State diagram for Example 3.8

Tiking invere Laplace transforms, the state transition matrix is

=
L'{sI-AT}=| e0
-3
0

The ZIR is

FR)=Ce'x(0)
-(2

Ihe input is r() = 1, unit step. Hence the ZSR is

'zsa (t)=CjeAt-aB u(e) de

0

e-2(1-0) 0

-i2-4-9,-9)do

=e-)+;e-) 3
1

---ez0.
 116 Modern Control Theory

Alternate method:
Zero state response can also be obtaincd as y =
() CA-e-IB

e-3t

--646-2e2

The result is the same as

that obtained earlier.
Example 3.9 A linear system
is represented by the state model

xo9-[
Find the ZIR and
the complete solution
for the state equation
Solution: The resolvent for a unit step input.
mnatrix is

s+3 1

s+4s + 4-1 s+1

s+3
(s+2)? (s+2) (s+2) (s+2)²
-1 S+1
(s+2)?
L(s+2)2 -1
(s+2)2
(s+2) (s + 2)
Taking inverse Laplace transforms.
le-21

= Ce x(0)=(2
J'a)
e-2-e-21
-217
-[2

rsa)=C4 [e-i]e

+te-1 le2i

+3te-1 11

2e-2 -3te- -2|

Complete response is

11 11
)= Yzn()+ PzsR)=e +-
4 4 2 4 2

Example 3.10 A state model of a linear system is given as

Obtain, by the time domain method, the ZIR and the complete response for a unit impulse
input.
Solution:
 Modern Cori
118
2
[s+3
resolvent matrix is -27
The s+1 S+1
+4s +5|-1
s+3
2
S+3
+1 (s+ 2) +1
2] (s+2)
1 fs+3 S+1
-1
(s+2) +1-1 s+l G+2)°
(s+2)?
+1 (s+2) +1|

1 2
[s+2-+
+1 (s+2) +1
(s+ 2) +1
(s+2) 1
S+2
-1
+1 (s+2) +i (s+2)+1|
(s+2)
Taking inverse Laplace transforms,
,-21
e-2 cos t+e- sint 2e sin t
e- cos
-e sin t-esint
The zero input response is

V) =CeA X(0)

e2 cost+e sin/
2esint
ecost-esint
e2 cost +esint =e cost.
-l sint
The zero state response is

sal) =Ce Je-20 cost+e2 sint 2e-2/ sint

B-!
sint e cos-e-20 sint
e
cost+e sint
=l
1

sin(+2e
+4esint i ictO: R1
t+ cost-2e sin
=3e- cost+2e-2 sint
Complete response
is

y) =
zn() +yzse() =e cost+ (3e-21 sint
cost +2e-
sin) =
4e- cost+2e-2
Evsmple
equation,for a unit
step input.

Solution:

xo

zero input response was obtained in Example 3.10 as

The

PzR() =ecost.
The zero state response is

Jaa)-Ca"|e-]8
-3 -21| e- cost+e sint-12e-2 sint aups re
sint e-2 cost-e-21gsint-1

e-2 cost +e sint

sint-1+4e
2 -3]
-sint+2e- cost-2e sint -2)

|-1+e2, cost +5e-2 sintt

-2+2e cost-3e-2 sint

-2e cos! -2e sint +2-8e-20 sint+3e-24 sint -6e cost+6esint +6

--cost -e
8 -8
sint.
5

33 METHODS OF COMPUTING STATE TRANSITION MATRIX

3
For
the solution a of state transition matrix eA? is neces
sary,especialy of state equation, the computation
in the time domain solution. Many methods are available for determining the
Sate transition
matrix:
I. Laplace
transform method
2. Power
series method
3. Using
Sylvester's
Interpolation Formula
*Using Cayley-Hamilton
D.Diagonalization Theorem
These
method
methods
are discussed
here one by one.
120 Modern Control Theory

Method
3.3.1 Laplace Transform
From equation (3.5),
YzIR (s) = C[sI- A' x(0).

Taking inverse Laplace transforms,

Vaa (9 = Icisl-Af'x(0))

-c{u'ul-ar}x0).
From equation (3.9),

Vzn() =CeAx(o). (B.l4

Comparing equations (3.13) and (3.14), the state transition matrix is

eAl
=L'sI-Af'. (3.5)

d At
Example 3.12 From the result e =AeA , devise a method of finding the state trans
tion matrix. dt

Solution:

Let F() = eAl F(0) =

e^0 =1.
It is known that

d FO) = A
F().
Taking Laplace transforms,

sF(S)- F(0) = AF(s).

Or sF(s)-I = AF($) or =
[sl- AJF(S)
I.
From this, F($) =
(sl- AJ.
Taking inverse Laplace
transforms,
 --
Solutionof Stato
iquallone 121

Hence, eA

4.3.2 Power Series Method

The state transition matrix ecan be
written in a powet
sories as

2! 3! 4 k
=I+At+
2! 3! 41
k!
This method is
practically very time consuming
and is scldom uscd,
Example 3.13 Find the state transition matrix for
Solution: The state transition matrix is given by

-l+At+ArAr,A.AC+..
2! 3! 4! k!

011 01-1

Hence, e

Example
3.14 Find
the state transition matrix for
Solution:
The state
transition matrix is given by

eAl
I+Al++A' 2!
41 t.+
 122 Modern Control Theory

-8 12|
0
-8
e4 3! 0 +..
-8|
+...

1+(-2)+-21
+- +.

+...

0
1+(-2)+2n'2+ 1

Hence, e^=
Example 3.15 Find the state transition
matrix for
Solution: The state transition
matrix is given by

eAl=I+At+A,A'
2! 3!
A#
+
4!
-+...+ +
k!

1-34++...
2

It can be observed
-21+3f
-t...
that it is very difficult
1-2t-++..
in this matrix,
and hence this method is to obtain closed form
a series
not in practice.
of expression for each
 Solution of State Equations 123

es Using Syvester's Interpolation Formula

Csse 1: System Matrix with Distinct Eigen Values
or sn mth the n Eigen values be
onder system, let ,.,.a.. Then Sylvester's interpola
omputing the state transition matrix
thn ormula for
is given by
1

=0 (3.16)

A!

Sotbing this equation, the state transition matrix e can be determined.

Case 2: With Repeated Eigen values

For an r-th order system, assume that there are three repeated Eigen values and let these be
i=,.For such a system matrix, Sylvester's interpolation formula can be written as

0 0 1 I
1
A
2! d2
d 1
A²

d A

=0 (3.17)
 Theory
Modern Control
124

This reduces
to
1
-
0 <A
A2
1

1 A3
24
3A, (3.10)
34 =0

(n-)(n-2)1-3
|24
(n-I)4-2

2
matrix for
using Sylvester's interpolation formula, the state transition
Example 3.16 Find,

Solution: The Eigen values are obtained by solving the characteristic equation
|2+2 -2
or
|21-A|=0 240.
or =-1 and =-4.
2+52+ 4=0
The Eigen values are distinct. Hence, the Sylvester's interpolation formula is

1 1
I
-4 A =0.
e e e

Expanding this determinant,

-4e^l -et A+(eAl-e I)+e(A+41) = 0

3ehl =e (A+)+e(A+41)
 Solution of State Equations 125

+2e -2e-41 +2e

te! 2e41 +e!

The state transition matrix is

Example 3.17 Find, using Sylvester's interpolation formula, the state transition matrix for

Solution: The Eigen values are obtained by solving the characteristic equation

-2
|al-A|=0 A+4|
=0.

2+42+ 4 =0 or
l=-2 and A
=-2;repeated roots.
The Eigen valus are repeated. Using the Sylvester's interpolation formula,

0 1

A =-2 A =0.
le2
d

Expanding this
determinant,
A+(-21e -e)l= 0.
eA-te2
eA
=te(A+21)+eI
 126 Modern Control Theory

2te-21
e+2t e-21
--2e-21

3.3.4 Cayley-Hamilton Theorem

cvery square matrix satisfies its own characteristig
Cayley-Hamilton theorem states that
equation.
ca
Consider a third order system with system matrix A. Its characteristic equation
written as
P(a) = 4' +a,a +
ajd+a =0.

The Cayley-Hamilton theorem states that

F(A) = A'+ ay A+a, A+ ao I=0.
To illustrate this, consider a third-order system matrix

0 1
0
A=0 1

-6 -11 -6|
Its characteristic equation is

-1
|-A|=0 lo
Or -1]=0.
11 +6|
Or P(2) =1'+62 +112+6=0

1
0 0 0| 0
A= 0 0 1
1|=|-6 -11 -6
-6 -11 -6|-6 -11 -6 36 60 25

0| 0 0 1

=0 -6 -11
-6 -11 -6
-6=| 36 60 25
-6 -11 -6|| 36 60 25 -150 -239 -90
P(A) = A'+ 6A'+11A+61 = -636 -11 -6 0 0 6
60
25+--36 -66 -36
-150 -239 -90 360 150|
|216
 Solution of State Equations 127

0 6 (0
0
11|+0 6
-66 -121 -66| 0 6|0

strate the Cayley-Hamilton theorem,

$4
Applications of Cayley-Hamilton Theorem

sA1 Natrìx Inversion

mN-Hamilton theorem can be used to find the inverse of a square matrix.
vaystem matrix of size n X n, let the characteristic equation be
+..+adt a (3.19)

the system matrixA satisfies its own characteristic equation,

A"+a,-jA+a- A. +a A+ ag I =0. (3.20)

Asme that Aexists Pre-multiplying equation (3.20) with A,

A=+a- A-+a2 AR-t...+a I+ 4g A-'=0.

(3.21)
0
3
Etample 3.18 Find the inverse of using Cayley-Hamilton theorem.
Soletion: The Eigen values are obtained by solving the cháracteristic equation

|21-A=0 or3
or = 0 or =-2 and
a
=-3.
2+52+6
The matrix
A
satisfies the characteristic equation and hence,
aso
A-+5A+6I=0 or A+5I+6A-!0.B6
 120 Modern Control T'heory

-2
using Cayley-fHamilton theorem
ExAple J,19 Find the inverNe of -2
0 -!
Solutlon: The Bigen values are obtained by solving the
characteristic equation

|a+2
|aI-A-0 +2
)

or a' +51'
(a+ 2)(4 +34 +
2)+(2+1) = ) +92+5 =0.i0
The matrix equation is A'+5A'+9A+SI = 0 or A'+5A+91+5A- =0.

[-2 01-2 01 3 -4
A'=- -2 1|-1-2 3-3|
4

0 -1| 0 1

|3 -4 1| -2 01 1 0 0|
-3+s|-1 -2 1|+9|0
lo 0 0
-1|
2 1]
Hence, 2

34,2 Reduction of the Degree of a Matrix Polvnemial oukoy. 5y51 t

Using Cayley-Hamilton
theorem, a matrix polynomial
degree (n-1) or less.
Consider a system matrix of degree n can be reduced to one
then the characteristic of size n X n, valus
polynomial If represents any Eigen
associatcd with the
matrix A is
P(A) = 1"

Consider another
+n-jt Cn-22t.+4+e0:Ate
polynomial
F(2) of degree m; m>n
F(2) = +
ay-j" +
am-22"+..+a
+ 40
 Solutlon of Blale Bquations 129

F(a) a)
A) Q) P(a)

where QA) is a polynomiallof dogreo (m-n) and R(2) intho romainder polynomlal of degree

(n-).
() (2)Q(0)+ R(a).
-
Now (3.22)

)consider
Also
a matrix polynonial in tornmN of A
of dogreo . Thls can be written as

F(A) -
A"+a- A+-2 Att A+ yl.
laa sinilar way to cquation (3,22), this matrix polynomial can be written as

F(A) P(A)QA)+ R(A),et (3.23)

Atthe Eigen values, P(a) 0, which is tho charnctoristlo cquation.

By Cayley-Hamilton theorem, every square matrix satisfics its own charncteristic equa
tion. Hence, P(A) = 0
From equation (3.23),

F(^) m R(A)
The polynomial R(A) is, at most, of degrco (m- 1) and henco F(A) can be written as

F(4) = R(A) = a l+ a A+ ag A+.+ ah- A, (3.24)

1 using Cayley
Example 3.20 Ir A
: compute F(A) = A'+4^'+3A'+2 A+ I,
Hamilton theorem.

Solution: The characteristic polynominl is

J21-A|=0
on0 5or -34+2 =0. ()

The polynomial F(A) = A44^'+34'+2 A+I is of degreo 4.

From the Cayley-Hamilton
theorem,
A'-3A+2l =0 or A 3A-21 nai
3A'-2A 3(3A-21)- 2A =7A-61
= m

A'= AA' A(3A-21)

A' A?A?
=
(3A-21)(3A-21) =9A'-12A+41
 130 Modern Control Theory

Hencc, F(A) = (15A-14)+4(7A-6I)+3(3 A-21)+2A+1 = 54 A-431

1.
Now F(A) is of' degrec
01 -43 54]
119|

3.4.3 Evaluation of the Analytic Functions of a Square Matrix

Ananalytic function of a square matrix is defined by a
similar series as its scalar c0ounterpan.
For cxample,

sin.x =- ".**
.

If A is a square matrix, then an analytic function similar to this series is

A
sin A = A'
A-a 7!
Similarly cos.x =
1+.2! 4! 6!

A
and cosA
=
1-AA_
2! 4! 6! *
2A Isccovios s

e'=ltx*
3! 4! +..

and eAl = [+At+ At A

A4 +
+
2! 3! 4!
Computation of an
analytic function of a square
(3.24) and its scalar
counterpart. The method matrix can be carried out
using equau
matrix and the distinct is discussed here.
Eigen values are Consider that A is an
From equation (3.24), 4, ,,,: (A

F(A) = R(A)= do
I+ a A+ a,
The corresponding
scalar equation is
A+...+-A

(329
Equation (3.25)
represents n
values,
and solving these, allsimultancous cquations corresponding n difTerent
Eiget
a the coefficients to determined
Cy,,---,n-1 can be
 Substituting these values in equation (3.24), the analytical function of the matrix A can be
determined. Selecting F(A)= e state transition matrix can be
the evaluated using this
method.

Examplk 321 Evaluate the analytical function cos A

of the matrix

Solution: For the square matrix A,

the Eigen values are obtained by solving

|àl-A=0 or 2' +31+2 =0.

The Eigen values are 4 =-land , =-2.

Le cos A = , I+ a A. The corresponding scalar equations are cos, = a + ah
and cos A =
+a.
Sotving these equations
, cos(2,)-, cos(4,)-cos(-2)+2 cos(-1)
=1.4967
-1+2

=
cos(4)-cos(4)cos(-)-cos(-2).-=0.9565.
-1+2

Hence, cos A
=1.4967 I+0.9565A

1.4967.3 0.95657
1913
-1.3728

Example 3.22 IfA is a (2 x 2) matrix with distinct Eigen values, determine the state transtion
matrix using the Cayley-Hamilton

Solution: Let
theorem.
the Eigen values be 4, and , and the scalar polynomial be f(2) =e.

ra)=e!
Solving these
equations,

and a=
 Theory
132 Modern Control
is
the matrix to be evaluated
Now the analytical function of
S(A)=eAr = a, l+ a A

e -A.

for using Cayley-Hamilto

Exsmple 3.23 Evaluate the state
transition matrix
theorem.
Solution: The Eigen values
are obtained by solving
-2 or
|2I-A|=0 2 +52+4= 0.

The Eigen values are

, =-1l and , 4.
f4)=e1 =a, +a4
e
sa4)=e=a,+ag
Solving these equations,

and q-e-).
Hence, the state transition matrix,

f(A) =e^I = , I+ aA

ao-2a 2a]
ao-3a

Example 3.24 If A
is a (2 x
2) matrix
tion matrix using the Cayley-Hamiltonwith repeated Eigen trans
values, determine the state
theoren.
Solution: Let the Eigen
values be and , and the scalar polynomial
be
 Solution of State Equations 133

(2)
Diflerentiating cquation (l) with respcct to A,

d
da al+aa)
te or a = te4 (3)

Substituting cquation (3) in cquation (2),

(4)

e= a I+ a,A

Evaluate the state transition matrix for A =

2 using Cayley-Hamilton
e3.25
Example.

theorem.

Solution: The Eigen values are obtained by solving

Ja1-A|=0 o2 or (2+2)(A+2) = 0.

The Eigen values are A, =-2and y =

-; repeated Eigen values.
(1)

(2)

10 get another equation, differentiate equation (1) with respect to and substitute 2 =-2.

Ti-2 dA l-2 l--2

(3)
From equations
(2) and (3),

+2a =e+21e-21 (4)

 134

matrix is given by
The state transition
fA)=I = , l+ Aa
a
-2a

le-2/
e-2
p-21
|0
[-2 1
0

transition matrix for A= 0 -2 1by Cayley-Hamiltn:

Example 3.26 Evaluate the state 0 -2
theorem.
values are A =-2,-2,-2 repeated
th
Eigen
Solution: The matrix A is in Jordan form. The

Sa)
=e = y + adt az2'
= +
-2a 4a,.
Differentiating equation(1) with respect to 2,
=
(a + 2a,2)

=(a +2a,2)

te-21 =
a-40ay.
Differentiating equation (1)with respect to A twice,

d
d2?
l=-2 la-2
Pe=2a,
-2 0-2 1
0 1

A'=0 -2 4-4
0-2
1
|0 4 -4
0 -2|| 0 0
-2| 0 0 4
F(A) = e4I
 1
[1 0 0 [-2 1 |4 -4
= n| 0
o 1 o+a 00 -2 1+0 4 -4
0 0 -2
-2 0 4

ap -2a, + 4ca,
o ay -2a, + 4a, -4a,
0 o -2 +4a,
1

=l0 te2

Method of Diagonalization
state transition matrix is discussed in
The method of diagonalization for finding the
Chapter 4.

STATE TRANSITION3
3.5 SPECIAL CASESOF DETERMININGTHE
MATRIX
canonical form, then the
If the system matrix is either in the diagonal formor in the Jordan
state transition matrix can be written down by inspection.

3.5.1 System matrix in Diagonal formsto xisiaq seeSe

Consider that for a third-order system, the state model is

|-4.0
0
0

where
-: -4 and - are the distinct Eigen values.
Ihe corresponding differential equations are

(3.26)
 Theory
Modern Control
136

as
can be obtained
Solutions to these I'x (0)

xy =ex, (0) (33

x (0).
Keeping in matrix form,
, (0) 7
|0 (0) (3
0 0
olssilertogcld to bety
This can be expressed as
X)= ex(0)
Comparing equations(3.32) and (3.33),STAG 0
232A3 J41332E
e 0

o e
0
0

3.5.2 System matrix In Jordan form innopsid ni xisisTm aleve i

a o

Considera state equation in Jordan canonical form l 12b-Laist

=
X JX.
For a third-order system, the state model in Jordan
form can be written as

-
=0 I
0 - 1JL*3

with all the Eigen values at A =

h.
The corresponding differential
cquations are
 Solutiorn of State Equations 137

(3.35)

(3.36)

(3.37)

The solution to cquation (3.37) is

x =e4x,(0). (3.38)

Substituting equation (3.38) in equation (3.36),

X)
=-4x, +ex(0). (3.39)

Solving this first-order linear differential equation,

=eg (0)+te(0). (3.40)

Substituting equation (3.40) in equation (3.35),

(0)+te-Alx (0).
x=-h+e (3.41)

Solving this first-order linear differential equation,

(3.42)
2

Keeping equations (3.38), (3.40) and (3.42) in matrix form,

tet (0)7
(0) (3.43)
x=0

This can be written as

=
x() eX(0). (3.44)

COmparing equations (3.43) and (3.44), the state transition matrix is given by

|0
 138 Modern ContrUI

as
This can be written

e 2! da?
da
d
=0 da

system, with n
repeated Eigen values,
In general for an n-th order

2! 3! n!
1

0 2!
utsns ti (7-1)!
Ar =l0 (n-2)ei

0 0

Example 3.27 Repeat Example 3.25, considering that A is in Jordan canonical form.

Solution: A

The Eigen values are A

=-2, repeated twice.
The state transition matrix is given
by

, le21
l=-2

Example 3.28 Determine

the state transition
matrix for the system matrx
-2 0
A =| 0
-2 1

0
-2|
 Solution of State Equations 139

Solution: The given matrix is Jordan canonical

form withrepeated Eigen values d = dy
repeated thrice. In this case, the state transition
matrix is given by -2.
4

2!da?

Al=|0 d
da
21 le2
0

SUMMARY
A state equation can be solved using Laplace transform
method.
A state equation can be solved by
actual integration.
For a SISO system, zero
the input response is VzR !) =CeX(0)
For a SISO system, the zero state response
due to a unit impulse input is
'zse(1) = Ce
B.
For a SISO system, the zero state response due to a unit step input
Jzsa = is
() CA'e-1B.
The resolvent matrix
of a square matrix A is
The state transition sI-A.
matrix for any square matrix can be determined
transform method, power series by using Laplace
method, Sylvester's interpolation formula,
Hamilton theorem or by diagonalization Cayley
method.
Power series method
of finding the state transition matrix of a square matrix is very
consuming and quite often may not yield a time
closed form of solution.
Cayley-Hamilton theorem states that every square
matrix satisfies its own characteristic
equation.
Cayley-Hamilton theorem can be used to
find the inverse of a square matrix,
USing Cayley-Hamilton if it exists.
a
theorem, matrix polynomial of degree n can
of degree be reduced to one
(n-1) or less.
An analytic function
of a square matrix is defined by a similar transcendental
scalar counterpart. series as its
Cayley-Hamilton theorem can
matrix. be used to evaluate an analytical
function of a square
Cayley-Hamilton theorem can
matrix. be used to evaluate the state
transition matrix of asquare