Introduction to State Space Analysis - GATE

Study Material in PDF

In these free GATE 2018 Study Notes, we introduce a new topic in Control Systems –
Introduction to State Space Analysis. In the previous few GATE Preparation Notes,
all the analysis techniques studied had a requirement that the physical system should
be in transfer function form. Although a powerful and simple tool, transfer function
methods have certain draws backs. They are only defined for LTI systems. Also, a
transfer function comes from assuming zero initial conditions. It is also restricted to
only single input output systems and provides no insight about the internal state of
the system.

Thus we require a more general mathematical representation of systems which can

yield information about state of systems variables along with the flow the signals. For
this purpose, we use state space analysis. It eliminates all the disadvantages of
transfer function approach. There are tools like controllability and observability to
check the validity of state variables.

These GATE Notes can be downloaded in PDF for your preparation of GATE EE and
GATE EC. They are also useful for other exams like IES, BSNL, DRDO, BARC. But
before you get started, here are some articles you must read.

Recommended Reading –

Block Diagram Algebra

Signal Flow Graph
Properties of Matrices
Rank of a Matrix
Laplace Transforms

Some basic definitions related to state space analysis are being given as follows:

State of the Systems

1|Page
State of a system describes about the future behaviour of a system by using present
(input variables) and past (State variables).

State Variable
These are the set of variables which along with input define the behaviour of the
system (i.e. future values).

Number of state variable in a system should be minimum equal to the order of the
system. Also, the system output and its derivatives can be considered as state
variables.

State Space
Set of possible values that the state variables can assume is called state space.

Now, suppose we have a multi input multi output system with m number of inputs, n
number of state variables and p number of output variables. The systems will be
represented through state equations as follows

ẋ = Ax + Bu

y = Cx + Du

Hence u is m × 1 input vector, x is n × 1 state variable vector and y is p × 1 output

vector.

u1 x1 y1
u2 x2 y2
u = [ : ],x = [ : ],y = [ : ]
um xn yp

A, B, C, D are matrices with their dimensions as A → n × n, B → n × m, C → p × n,

D→p×m

2|Page
 dx1
a11 a12 … a1n x1 u
dt
a21 a22 … a2n x2 b11 b12 … b1m u1
dx2 2
∴ dt = : .. .. .. : + [b21 .. .. ..
] :
: : .. .. .. : : .. .. .. :
dxn
[ a .. . . ann ] [xn ] bn1 .. .. bnm [u ]
[ ]
dt
n1 m

c11 c12 … c1n x1 u1

y1 d11 d12 … d1m
c21 … … … x2 u2
y2
∴[ :]= : … … . . x 3 + [ d4 .. .. ..
:
: … … .. : .. .. .. ]
: :
yp bp1 .. .. dpm
c
[ P1 … … cpn ] [xn ] [um ]

The state diagram of a standard system can be given as

State models can be derived from the physical system itself or from differential
equation form (phase variable method).

First, we will take a look at physical system itself. Suppose we have a system with
circuit given as follows.

3|Page
In the above circuit E is the input variable. It has three energy storage elements L1, L2
and C. The state variables of the system are x1(t) = v, x2(t) = I1, x2(t) = I2. If we have
knowledge of initial conditions for v(0), I1(0), I2(0) and input signal E for t ≥ 0, then
the behaviour of the network is completely specified for t ≥ 0

Now, applying KVL is the circuit, we get

Cdv
I1 + I2 + =0
dt

dI1
L1 + R1 I1 + E − v = 0
dt

dI2
L2 + R 2 I2 − v = 0
dt

They can also be written as

dv −1 1
= I − I2
dt C 1 C

dI1 1 R1 1
= v− I − E
dt L1 L1 1 L1

dI2 1 R2
= v− I
dt L2 L2 2

In terms of state equation, they can be written as

1 1
0 − −
ẋ 1 C C x1 0
1 R1 1
[ẋ 2 ] = − L1 −
L1
0 [x2 ] + [− L ] u
1
ẋ 3 x3
1
0 −
R2 0
[ L2 L2 ]

4|Page
For y1 = I2 R 2 , y2 = I2

x1
y1 0 0 R2 x
[y ] = [ ] [ 2]
2 0 0 1 x
3

In this way, we derive state equation from the physical systems.

Now, if we look at phase variable method, suppose we have a system represented in

the form of equation as

dn y dn−1 y dy
+ a1 + ⋯ + an−1 + an y = bu
dtn dtn−1 dt

Let x1 = y, ẋ 1 = x2 , ẋ 2 = x3 , … , ẋ n−1 = xn

Now the above equation can be written as

ẋ n + a1 x1 + a2 xn−1 + ⋯ + an−1 x2 + an x1 = bu

ẋ n = −an x1 − an−1 x2 … − a2 xx−1 − a1 xn + bu

Writing is state equation form, we get

ẋ 1 0 1 0 … x1 0
0 0 1 … x2 0
ẋ
[ 2] = [ ][ : ] + 0 u
: 0 … … 1
−an −an−1 … −an xn :
ẋ n [ b]

x1
x2
y = [1 0 0 … 0] [ : ] + [0]u
xn

This form of matrix is also called Bush form or companion form or controllable
canonical form. Now, instead of being a constant; if the input is also varying, then we
will get transfer function in the form of

y(s) b0 sn +b1 sn +⋯+bn

=
u(s) sn +a1 sn−1 +⋯+an−1 s+an

5|Page
(In this example we are considering numerator and denominator having equal
degree).

y(s) b0 +b1 s−1 +b2 s−2 +⋯+bn s−n

=
u(s) 1+a1 s−1 +⋯+an−1 s−(n−1) +an s−n

Multiplying both numerator and denominator by variable X(s), we get

y(s) (b0 +b1 s−1 +b2 s−2 +⋯+bn s−n )X(s)

=
u(s) (1+a1 s−1 +⋯an−1 s−(n−1) +an s−n )X(s)

y(s) = (b0 + b1 s−1 + ⋯ bn s−n )X(s)

u(s) = (1 + a1 s −1 +. . . an−1 s−(n−1) + an s −n )X(s)

X(s) = u(s) − [a1 s−1 + a2 s −1 … an−1 s −(n−1) + an s−n ]X(s)

⇒ ẋ n = u − [a1 s −1 + a2 s−2 +. . an−1 s −(n−1) + an s −n ]ẋ n

ẋ n = u − a1 xn − a2 xn−1 − ⋯ − an−1 x2 − an x1

Now, y = (b0 + b1 s −1 + ⋯ bn s −n )ẋ n

= b0 ẋ n + b1 xn + b2 xn−1 + ⋯ bn x1

y = bn x1 + bn−1 x2 + ⋯ + b1 xn + b0 [u − a1 xn − a2 xn−1 … − an−1 x2 − ax x1 ]

y = (bn − b0 an )x1 + (bn−1 − b0 an−1 )x2 + ⋯ (b1 − b0 a1 )xn + b0 u

ẋ 1 0 1 0 … 0 x1 0
x2̇ 0 0 1 … 0 x2 0
[ ]=[ ][ : ] + [ ]u
: : :
ẋ n −an −an−1 .. .. −a1 xn 1

x1
x2
y = [bn − b0 an bn−1 − b0 an−1 … b1 − b0 a1 ] [ : ] + [b0 ]u
xn

6|Page
Another form of representation is observable canonical form. Consider a system as
below

y(s) b0 s3 +b1 s2 +b2 s+b3

T(s) = =
u(s) s3 +a1 s2 +a2 s+a3

b b b
b0 + 1 + 22 + 33
T( s ) = s s s
a1 a2 a3
1−(− − 2 − 3 )
s s s

Above transfer function can lead to a signal flow graph with four forward paths and
three loops all touching each other

From above diagram, we get

y = x1 + bo u

ẋ 1 = −a1 (x1 + b0 u) + x2 + b1 u

= −a1 x1 + x2 + (b1 − a1 b0 )u

ẋ 2 = −a2 x1 + x3 + (b2 − a2 b0 )u

ẋ 3 = −a3 x1 + (b3 − a3 b0 )u

ẋ 1 −a1 1 0 x1 b1 − a1 b0
[ẋ 2 ] = [−a2 0 1] [x2 ] + [b2 − a2 b0 ] u
ẋ 3 −a3 0 0 x3 b3 − a3 b0

x1
y = [1 0 x
0] [ 2 ] + b0 u
x3
7|Page
Apart from these, we can also represent system in diagonal canonical form

If the system is

y(s) b0 sn +b1 sn−1 +⋯+bn−1 s+bn

=
u(s) sn +a1 sn−1 +⋯+an−1 s+an

It can also be written as

y(s) ci
= b0 + ∑ni=1
u(s) s−λi

y = c1 x2 + c2 x2 + ⋯ + cn xn + b0 u

ẋ 1 = λ1 x1 + u

ẋ 2 = λ2 x2 + u

Here λi are the poles of transfer function and ci are the respective residue of the
poles.

The state equation can be written as

ẋ 1 λ1 0 0 … 0 x1 1
ẋ 0 λ2 0 … 0 x2 1
[ 2] = [ ][ : ] + [ ]u
: : :
ẋ n 0 0 … … λn x n 1

x1
x2
y = [c1 c2 … cn ] [ : ] + bo u
xn

Diagonalization of State Equation

We have a state equation

ẋ = Ax + Bu

y = cx + Du

And its other form is

̃u
v̇ = Λv + B
8|Page
y = C̃v + D
̃u

Conversion from first to second form is called diagonalization of system.

𝜆1 0 … 0
0 𝜆2 … 0
Here Λ = [ ]
0 0 𝜆3 …
0 … … 𝜆𝑛

Where λ1, λ2, … λn are Eigen values of the matrix A

x = Mv

Here M is n × n non-singular constant matrix also known as modular matrix

Mv̇ = AMv + Bu

v̇ = M −1 AMv + M−1 Bu

̃ = M −1 B
Λ = M−1 AM, B

y = Cx + Du

= CMv + Du

C̃ = CM, D
̃=D

Until now, we discussed the definition and forms of state variable system. Further,
we will discuss system analysis using state space model.

Liked this article on Introduction to State Space Analysis? Let us know in the
comments. Also read –

System Analysis through State Space Model

Shortcut to find Controllability & Observability

9|Page