18AN62 – Control Systems – Unit 5 Lecture Notes

Introduction to State Space Analysis

(For private circulation only)

Unit 5 Syllabus
Concepts of state, state variable, state space representation of Linear Time-Invariant (LTI)
Systems, canonical representation of state models, solution of state space equations, state
transition matrix and its properties, concept of controllability & observability and tests to
ascertain the same.

Introduction
In the earlier chapters, we introduced the conventional control theory. Conventional control
theory is in general applicable to linear time invariant systems having a single-input and single-
output. In these approaches the central point is the formulation of transfer function which is
defined for zero initial conditions. The methods explained, such as root locus and frequency
response methods such as Bode, Nyquist, etc., cannot be used for time varying and/or non-
linear systems.
In this section we shall present introductory material on modern state space analysis of control
system.
Modern Control Theory: The modern trend in engineering systems is toward greater
complexity, due mainly to the requirements of complex tasks and good accuracy. Complex
systems may have multiple inputs and multiple outputs and may be time varying. Because of
the necessity of meeting increasingly stringent requirements on the performance of control
systems, the increase in system complexity, and easy access to large scale computers, modern
control theory, which is a new approach to the analysis and design of complex control systems,
has been developed around 1960. This new approach is based on the concept of state. The
concept of state by itself is not new since it has been in existence for a long time in the field of
classical dynamics and other fields.
Modern Control Theory Vs Conventional Control Theory: Modern control theory is
contrasted with conventional control theory in that the former is applicable to multiple-input-
multiple-output systems, which may be linear or nonlinear, time invariant or time varying,
while the latter is applicable only to linear time-invariant single-input-single-output systems.
Also, modern control theory is essentially a time-domain approach, while conventional control
theory is a complex frequency-domain approach.

Advantages and Disadvantages

Conventional Control Theory

Advantages:
(1) Simple techniques.
(2) Requires less computation.
(3) Provides good physical insight into the system.

1
Disadvantages:
(1) Useful only for Single Input Single Output (SISO) systems. For Multiple Input Multiple
Output [MIMO] systems, it is very complex.
(2) They are useful for linear time invariant systems and cannot be used for non-linear systems,
time varying systems etc.
(3) As the transfer function is defined for zero initial conditions, it is not valid for other initial
conditions.
(4) While a system may be overall stable, some parameters of the system may be exceeding
their ranges. We cannot know about such internal states.
(5) Design is basically a trial-and-error approach and designs may not be optimal.

Modern Control Theory

Advantages:
(1) Can analyze time varying or time invariant, linear or non-linear, single or multiple input-
output systems.
(2) It is possible to ascertain the state of the system parameters also and not merely input -
output relations.
(3) The state space methods tend to optimization of systems and optimal design.
(4) Inclusion of initial conditions is possible.
Disadvantages:
(1) Complex techniques.
(2) Computation is more.
Before we proceed further, we must define state, state variables, state vector, and state space.
The state of a system refers to the past, present, and future conditions of the system. The state
variables must satisfy the following conditions:
• At any initial time, t = to, the state variables x1(t0), x2(t0); • • •, xn(t0) define the initial states
of the system.
• Once the inputs of the system for t ≥ to and the initial states as defined above are specified,
the state variables should completely define the future behavior of the system.
It is the smallest set of variables which determines the condition of a dynamic system.

Important Terminology

(1) State: The state of a dynamic system is the smallest set of variables, called state variables,
such that the knowledge of these variables at t = t0 together with the inputs for t ≥ t0, and the
equations describing the dynamics completely determines the behavior of the system at t ≥ t 0.

(2) State Variables: The state variables of a system are defined as a minimal set of variables,
x1(t), x2(t), ... ,xn(t), such that knowledge of these variables at any time t0 and information on
the applied input at time t0 are sufficient to determine the state of the system at any time t ≥ 0.

2
For example, if n variables xl (t), x2 (t), x3 (t), …, xn (t) are needed to completely describe the
behavior of a dynamic system, such that once the input is given for t ≥ t0 and the initial state at
t = t0 is specified, the future state of the system is completely determined, then such n variables
xl (t), x2 ( t), x3 ( t), …..., xn (t) are called set of state variables. These state variables need not
be physically measurable or observable quantities.

(3) State Space: The n dimensional space whose co-ordinate axes consist of the xl axis, x2
axis, …..., xn axis is called a state space. Any state can be represented by a point in the state
space. Typically, for a two-dimensional state space can be represented as shown below.

(4) State Vector: If 'n' state variables are needed to completely describe the behavior of a given
system, then these n state variables can be considered to be the n components of a vector x(t).
Such a vector is called a state vector. A state vector is thus a vector which determines uniquely
the system state x (t) for any time t ≥ t0, once the input u(t) for t ≥ t0 is specified.
The two-dimensional state vectors corresponding to time instant t1 and t2 are represented below:

(5) State trajectory: It is the locus of the tip of the state vectors, with time as the independent
variable.

3
State Equations of continuous time system: In state-space analysis we are concerned with
three types of variables that are involved in the modeling of dynamic systems: Input variables,
Output variables, and State variables.
It may have to be remembered that, the state-space representation for a given system is not
unique, except that the number of state variables the same for any of the different state-space
representations of the same system.
The dynamic system must involve elements that memorize the values of the input for t ≥ tl.
Since integrators in a continuous-time control system serve as memory devices, the outputs of
such integrators can be considered as the variables that define the internal state of the dynamic
system. Thus, the outputs of integrators serve as state variables.
The number of state variables to completely define the dynamics of the system is equal to the
number of integrators involved in the system. .
Assume that a multiple-input-multiple-output system involves n integrators. Assume also that
there are r inputs u1(t), u2(t)..., ur(t) and m outputs y1(t), y2(t)..., ym(t).
Define ‘n’ outputs of the integrators as state variables: x1(t), x2(t), ..., xn(t) Then the system
may be described by

x 1 (t )  f 1 ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t )

x 2 (t )  f 2 ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t )
 (Eq 1)



x n (t )  f n ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t )
The m outputs y1(t), y2(t), ... ,ym,(t) of the system may be given by
y1 (t )  g1 ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t )
y 2 (t )  g 2 ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t )

(Eq 2)


y m (t )  g m ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t )

4
Now, we can define:
 x1 (t )   u 1 (t )   y1 (t ) 
 x (t )   u (t )   y (t ) 
 2   2   2 
.  .  . 
x (t )    (n x 1); u (t )    (r x 1) ; y (t )    (m x 1)-
 .   .   . 
.  .  . 
     
 x n (t )  u r (t )   y m (t ) 
which are state, input and output vectors respectively
Now, define:
 f 1 ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t ) 
 f ( x , x ,......, x ; u , u ,...., u ; t ) 
 2 1 2 n 1 2 r 
 
f ( x, u , t )   
 
 
 
 f n ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t ) 

 g1 ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t ) 
 g ( x , x ,......, x ; u , u ,...., u ; t ) 
 2 1 2 n 1 2 r 
 
g ( x.u.t )   
 
 
 
 g m ( x1 , x 2 ,......, x n ; u1 , u 2 ,...., u r ; t )
Then Equations (1) and (2) become

x (t )  f ( x , u , t ) (3)
y  g ( x, u , t ) (4)
where Equation (3) is the state equation and Equation (4) is the output equation.
If vector functions f and / or g involve time t explicitly, then the system is called a time- varying
system.
If Equations (3) and (4) are linearized about the operating state, then we have the following
linearized state equation and output equation:

X ( t )  A ( t ) X ( t )  B ( t )U ( t ) (5)
Y ( t )  C ( t ) X ( t )  D ( t )U ( t ) (6)
where A(t) is called the system matrix or evolution matrix, B(t) the input matrix, C(t) the output
matrix, and D(t) the direct transmission matrix.
A block diagram representation of Equations (5) and (6) is shown in Figure 4.

5
 Figure 4 Block diagram of the linear, continuous time control system
represented in state space.
A time-varying control system is a system in which one or more of the parameters of the system
may vary as a function of time.
If vector functions f and g do not involve time t explicitly then the system is called time-
invariant system. In this case, Equations (5) and (6) can be simplified to

x (t ) = A x(t) + B u(t) (7)

y(t) = C x(t) + D u(t) (8)

Equation (7) is the state equation of the linear, time-invariant system. Equation (8) is the output
equation for the same system.
In this course we shall be concerned mostly with systems described by Equations (7) and (8).

Note on output variables:

One should not confuse the state variables with the outputs of a system. An output of a system
is a variable that can be measured, but a state variable does not always satisfy this requirement,
as stated earlier. For instance, in an electric motor, such state variables as the winding current,
rotor velocity, and displacement can be measured physically, and these variables all qualify as
output variables. On the other hand, magnetic flux can also be regarded as a state variable in
an electric motor, because it represents the past, present, and future states of the motor, but it
cannot be measured directly during operation and therefore does not ordinarily qualify as an
output variable.
In general, an output variable can be expressed as an algebraic combination of the state
variables as explained earlier.
Representation of systems by state space equations:
Some example cases for deriving a state equation and output equation will be presented below.
1. For the circuit system shown in Fig. 1 find state model using the physical variable for state
variables.

6
 Fig 1
The dynamic system must involve energy storing elements. There are three energy storing
elements, two inductors and one capacitor. Therefore three state variables are needed.
Let us choose i1, i2 the two currents through the inductors and the voltage across capacitor as
the three state variables.
dv
Writing the KCL at node (1) i1  i 2  C (1)
dt
di 1
Applying KVL to the two loops e  R 1 i1  v  L1 0 (2)
dt

di 2
L2  R 2 i2  v  0 (3)
dt

di1 R 1 1
From Eq (2)   1 i1  v  e (4)
dt L1 L1 L1

di2 R 1
From Eq (3)   2 i1  v (5)
dt L2 L2

dv 1 1
From Eq (1)  i1  i 2 (6)
dt C C
As already indicated, define the state variables i1 = x1; i2 = x2 and v = x3; and input variable u =
e and the output variables as y1 = voltage drop across R1; y2 = current i2. Then the state equation
(7) and output equation (8) representation of the physical system are shown below:

 R1 1
    L 0
L1   x   
 1
 x1   1  1 L  x1 
 x  =   R2 1    1   y1  R1 0 0  
 2   L2
0 x
L2    
2  0 e (7)  y    0 1 0  x2  (8)
 2   
 x3   1 1   x3   0   x3 
   0  
 
 C C 

7
2) Obtain the state variable representation of an armature-controlled DC servo motor.

The above figure 11.3 shows the armature controlled DC servo motor.
The governing equations, with usual notations, are

𝑣 = 𝑅 𝑖 + 𝐿 + 𝑒 Eq (1)

𝑒 = 𝐾 = 𝐾 𝜔 Eq (2)

𝐽 +𝐵 = 𝑇 = 𝐾 𝑖 Eq (3)

Eq (3) can be written as 𝐽 +𝐵𝜔 = 𝐾 𝑖 Eq (4)

= 𝜔 Eq (5)

Now, we define the state variables as listed below:

x1(t) = ia
x2(t) = ω
x3(t) = θ
Input u(t) = vi
Output y = θ = x3(t)
Substituting for eb(t) from Eq (4) in Eq (2) and simplifying, we get
𝑥̇ = − 𝑥 − 𝜔+ (6)

Rearranging eq (4) 𝑥̇ = − 𝑥 − 𝑥

From eq (5) 𝑥̇ = 𝑥

Expressing in vector matrix form, the state equations are:

𝑅 𝐾
⎡− − 0⎤ 𝑥 1
𝑥̇ 𝐿 𝐿
⎢ ⎥
𝑥̇ = ⎢ 𝐾 𝐵 𝑥 + 𝐿 𝑢
− 0 ⎥ 0
𝑥̇ ⎢ 𝐽 𝐽 ⎥ 𝑥
0
⎣ 0 1 0⎦

8
 𝑥
The output equation y = [0 0 1] 𝑥
𝑥

3) Consider the mechanical system shown in Figure 2. We assume that the system is linear.
The external force u( t) is the input to the system, and the displacement y( t) of the mass is the
output. The displacement y(t) is measured from the equilibrium position in the absence of the
external force. This system is a single-input-single-output system.
 
From the diagram, the system equation is m y  b y  ky  u (2.1)
This system is of second order. This means that the system involves two integrators. Let us

define state variables x1(t) and x2(t) as x1(t) = y(t) and x2(t) = y (t )
(Energy elements: Spring – potential energy; Mass – Kinetic energy)

Fig 2 Mechanical System

Writing the equations in a vector-matrix form,

   0 1  x   0 
 x 1    k b  1   1 u
 x 2   m    x2   
m m

The output equation can be written as

 x1 
y  1 0  
 x2 

Expressing the state equation and the output equation for the system, in the standard form:


x  Ax  Bu
y  Cx  Du
where

9
Figure 3 is the block diagram for the system. Notice that the outputs of the integrators are state
variable

3) Obtain a state-space representation of the system shown in Figure 3–22.

Figure 3–22 Mechanical system.

Solution.

The system equations are

The output variables for this system are y1 and y2 .

Define state variables as

Then we obtain the following equations:

10
Hence, the state equation is

and the output equation is

11
Problem 5

Correlation between Transfer Functions and State-Space Equations: In what follows we

shall show how to derive the transfer function of a single-input-single-output system from the
state-space equations. Let us consider the system whose transfer function is given by
( )
= 𝐺(𝑠) (2-22)
( )
This system may be represented in state space by the following equations:

where x is the state vector, u is the input, and y is the output. The Laplace transforms of
Equations (2-23) and (2-24) are given by

Since the transfer function was previously defined as the ratio of the Laplace transform of the
output to the Laplace transform of the input when the initial conditions were zero, we set x(0)
in Equation (2-25) to be zero. Then we have
sX(s) – A X(s) = B U(s)
or (sI – A) X(s) = B U(s)
By pre-multiplying with (sI - A)-1 to both sides of this last equation, we obtain
X(s) = (sI – A)-1 B U(s) (2-27)
12
By substituting Equation (2-27) into Equation (2-26),
we get Y(s) = [C(sI - A)-1 B + D]U(S)
Upon comparing Equation (2-28) with Equation (2-22), we see that
G(s) = C (sI – A)-1B + D
This is the transfer-function expression of the system in terms of A, B, C, and D.
Note that the right-hand side of Equation (3-29) involves (s1 - A)-1. Hence G(s) can be written
as
( )
𝐺(𝑠) = | |
where Q(s) is a polynomial in s. Therefore, │ sI - A │ is equal to the characteristic polynomial
of G(s).
In other words, the eigenvalues of A are identical to the poles of G(s).

State space representations of transfer-function systems

Different Representations of a State Model:

State space representation using phase variables in Controllable Canonical Form (CCF):

13
Phase Variable CCF form for numerator terms having derivatives of forcing function:

14
Equations (12.13) and (12.14) are state model in phase variable form (CCF).

(Normally degree of numerator is less than denominator, so that xn term is absent giving easy
equation).

State space representation using phase variable in Observable Canonical Form:

Consider the transfer function

Cross multiplying

Define the state variables as shown below:

The state and output equations are given by

15
Many techniques are available for obtaining state-space representations of transfer-function
systems such as representations in the controllable, observable, diagonal, or Jordan canonical
forms.
Consider a system defined by (system involving derivatives of input)

where u is the input and y is the output. This equation can also be written as

(9 – 2)
Controllable Canonical Form: The following state-space representation is called a
controllable canonical form:

16
The controllable canonical form is important in discussing the pole-placement approach to
control systems design.
Observable Canonical Form: The following state-space representation is called an
observable canonical form:

It may be noted that the n x n state matrix of the state equation given by Equation (9–5) is the
transpose of that of the state equation defined by Equation (9–3).
Diagonal Canonical Form: Consider the transfer-function system defined by Equation (9–2).
Here we consider the case where the denominator polynomial involves only distinct roots. For
the distinct-roots case, Equation (9–2) can be written as

The diagonal canonical form of the state-space representation of this system is given by

17
Jordan Canonical Form: Next we shall consider the case where the denominator polynomial
of Equation (9–2) involves multiple roots. For this case, the preceding diagonal canonical form
must be modified into the Jordan canonical form. Suppose, for example, that the pi’s are
different from one another, except that the first three pi’s are equal, or p1=p2=p3. Then the
factored form of Y(s)/U(s) becomes

The partial-fraction expansion of this last equation becomes

A state-space representation of this system in the Jordan canonical form is given by

18
Obtaining state model from differential equation representation or transfer function
representation or block diagram representation;

Typical example problems:

A–2–6. Show that for the differential equation system:

19
20
21
This is one of the state models.

A–2–8. Obtain a state-space model of the system shown in Figure 2–26.

(Fig 2.26)

Solution. Let us define the output of the plant as x1, the output of the controller as x2, and the
output of the sensor as x3. Then we obtain, from block diagram algebra,

22
Note: It is important to note that this is not the only state-space representation of the system.
Infinitely many other state-space representations are possible. However, the number of state
variables is the same in any state-space representation of the same system. In the present
system, the number of state variables is three, regardless of what variables are chosen as state
variables.

A–2–9. Obtain a state-space model for the system shown in Figure 2–27(a).

Fig 2 - 27
Noticing that (as + b)/ s2 involves a derivative term, and such a derivative term may be avoided
if we modify (as + b) /s2 as

Using this modification, the block diagram of Figure 2–27(a) can be modified to that shown in
Figure 2–27(b). Define the outputs of the integrators as state variables, as shown in Figure 2–
27(b).Then from Figure 2–27(b) we obtain

Fig 2 – 27

23
which can be rewritten as

Taking the inverse Laplace transforms of the preceding three equations, we obtain

Rewriting the state and output equations in the standard vector-matrix form, we obtain

Solutions to state equations: Direct solutions (any type of system) or through Laplace
transform method for LTI
Direct solutions: Numerical integration approach – Euler integration; Runge – Kutta second
order and RK fourth order methods; etc.
Transform methods: by obtaining transfer function and then solving.
A–2–11. Obtain the transfer function of the system defined by

Solution. The transfer function G(s), for a state model representation of the system, with usual
notation, is given by,

G(s) = C (SI - A)-1 B + D

In this problem, matrices A, B, C, and D are

Hence

24
 ( )
(sI - A)-1 = ( )

Det (sI - A) = (s + 1)2 (s + 2)

(𝑠 + 1)(𝑠 + 2) (𝑠 + 2) 1
Adj (sI - A) = 0 (𝑠 + 1)(𝑠 + 2) (𝑠 + 1)
0 0 (𝑠 + 1)

Therefore G(s)

Simplifying,

From the above equation, time domain response can be obtained for different types of input.

B–2–10. Consider the system described by

Obtain the transfer function of the system.

Solution:

25
26
27
State Transition Matrix (STM)
Preliminaries:
Laplace Transform Approach to the Solution of Homogeneous State Equations (Zero Input
solution).
Before we solve vector-matrix differential equations, let us review the solution of the scalar
differential equation

In solving this equation, we may assume a solution x(t) of the form

We shall now solve the vector-matrix differential equation

where x = n state vector; A = n x n constant matrix

By analogy with the scalar case, we assume that the solution is in the form of a vector power
series in t, or

By substituting this assumed solution into Equation (9–28), we obtain

28
If the assumed solution is to be the true solution, Equation (9–30) must hold for all t. Thus, by
equating the coefficients of like powers of t on both sides of Equation (9–30), we obtain

By substituting t = 0 into Equation (9–29), we obtain x(0) = b0

Thus, the solution x(t) can be written as

The expression in the parentheses on the right-hand side of this last equation is an n x n matrix.
Because of its similarity to the infinite power series for a scalar exponential, we call it the
matrix exponential and write

In terms of the matrix exponential, the solution of Equation (9–28) can be written as

Laplace Transform Approach to the Solution of Homogeneous State Equations.

Let us first consider the scalar case:
𝑥̇ = 𝑎𝑥 Eq (1)
Taking the Laplace transform of the above equation (1) we obtain
sX(s) - x(0) = aX(s) Eq (2)
where

Solving Equation (2) for X(s) gives

The inverse Laplace transform of this last equation gives the solution
x(t) = eat x(0)
The foregoing approach to the solution of the homogeneous scalar differential equation can be
extended to the homogeneous state equation: (bold face indicates vector and matrix)
𝑿̇(t) = A X(t) Eq (3)
Taking the Laplace transform of both sides of Equation (3), we obtain
sX(s) - x(0) = AX(s)
29
where

Hence, (sI-A) X(s) = x(0) Eq (4)

Pre-multiplying both sides of this last equation by (sI-A)–1,

we obtain X(s) = (sI-A)–1 x(0)
The inverse Laplace transform of gives the solution.
Thus

Eq (5)

Eq (6)
(Note: The inverse Laplace transform of a matrix is the matrix consisting of the inverse Laplace
transforms of all elements.)
From Eq (5) and Eq (6), solution to Eq (1) is obtained as
x(t) = eAt x(0) Eq (7)
State-Transition Matrix.
We can write the solution of the homogeneous state equation 𝑿̇ = 𝑨 𝑿 as
X(t) = eAt x(0) = Φ(t) x(0)
This solution of homogeneous state equation shows that the initial state at x0 at t = 0 is driven
to a state x(t) at time t.
Since this transition in state is carried out by the matrix exponential eAt, it is known as the state
transition matrix and is denoted by Φ(t) i.e. Φ(t) = eAT.
It can be noted that the STM depends only on A.
If the initial time is t0, the STM becomes Φ(t – t0) = eA(t – t0)
The STM depends only on the length of time (t – t0) and not on the initial time t0, so the initial
time can be conveniently regarded as zero.
Properties of State-Transition Matrices.
We shall now summarize the important properties of the state-transition matrix For the time-
invariant system
𝑿̇ = 𝑨𝑿 for which Φ(t) = eAT
we have the following:

30
Problems on STM:

31
Prob 2) Find the STM of the state equation given below:

Using partial fraction expansion for all elements, the required STM is

Prob 3) For a system when

32
 Eq (1,2)

(Eq 3,4)

33
Computation of STM:

Prob 4

34
 It is known from one of the properties of STM, i.e. , we obtain the
inverse of the state-transition matrix as follows:

Solution of Nonhomogeneous State Equations.

We shall begin by considering the scalar case, as before,

Let us rewrite Equation (9–39) as

Multiplying both sides of this equation by e–at, we obtain

Integrating this equation between 0 and t gives

Or

The first term on the right-hand side is the response to the initial condition and the second term
is the response to the input u(t).
Let us now consider the nonhomogeneous state equation described by

35
By writing Equation (9–40) as
𝒙̇ − 𝑨 𝒙(𝒕) = 𝑩 𝒖(𝒕)
and pre-multiplying both sides of this equation by e–At , we obtain

Integrating the above equation between 0 and t gives

Or

Equation (9–41) can also be written as

where Equation (9–41) or (9–42) is the solution of Equation (9–40). The solution x(t) is clearly
the sum of a term consisting of the transition of the initial state and a term arising from the
input vector.

36
Solution in Terms of x(to): Thus far we have assumed the initial time to be zero. If, however,
the initial time is given by t0 instead of 0, then the solution to Equation (9–40) must be modified
to

Problem:

37
Concepts of Controllability and Observability
In control problems two basic questions need to be answered in deciding whether or not a control
solution exists. These questions may be posed as:
(i) Can we transfer the system from any one state to any other desired state in finite time by
application of a suitable control force?
(ii) Knowing the output vector for a finite length of time, can we determine the initial state of
the system?
The answers to these basic questions were conceptualized by Kalman into what is known as
controllability and observability and are below:
A system is said to be completely state controllable if it is possible to transfer the system
state from any initial state x(to) to any desired state x(t) in specified finite time by a control
vector u(t).
Sometimes it is desired to transfer the system output from an initial value to any other desired
value. For a system whose output does not depend upon control vector but depends upon state
vector only, the dimensionality of output vector is usually less than that of state vector.
Therefore the output control is comparatively easier. In our study, we shall discuss state
controllability only. The conditions for output controllability, if desired, may be derived on
the similar lines.
A system is said to be completely observable, if every state x(t0) can be completely
identified by measurements of the outputs y(t) over a finite time interval.
A system which is not completely observable implies that some of its state variables are
shielded from observation.
The concepts of controllability and observability play an important role in control engineering.
These concepts were originally introduced by Kalman. The mathematical tests of
controllability and observability developed by Kalman, though elegant do not give a physical
feel of the problems involved. The alternative test introduced by Gilbert uses canonical state
model and provides better physical insight into the problem. In this section, we shall first
discuss the Gilbert's method and then Kalman's test without proof.
Test Methods for Controllability
As mentioned earlier, the concept of controllability involves the dependence of state variable
of the system on the inputs. Consider a single input linear time invariant system

(9)
where {x} = n-dimensional state vector; {u} = control signal (control force);
A =n x n matrix; and B = n x 1 matrix.
Let the initial system state be x(0) and the final desired state be x(t). The system described by
eqn. (9) is controllable if it is possible to construct a control signal, which in finite time interval
0 < t ≤ tf will transfer the system state from x(0) to x(tf)

38
Kalman Method for testing for controllability:
The test of controllability due to Kalman which can be applied to any state model (canonical
or otherwise) is stated below.
A general nth order multi-input linear time-invariant system (with an m-dimensional control
vector),

is completely controllable if and only if the rank of the composite (controllability) matrix
Qc = [B : AB : A2B : ….: An-1B] is n. (10)
Since only matrices A and B are involved in (10), we may say that the pair (A, B) is
controlled if rank of Qc = n.
Gilbert's method of testing for controllability.
Let us first assume that the Eigen-values of the matrix A are all distinct to that it can be
transformed into the canonical state variable form

Or (11)

This equation can be written in the component form as

which has the solution

The system described by eqn. (10) is completely controllable if state variable vi can be
transferred from initial state vi(0) to a final state vi(tf) in a finite time tf. In other words, the
system is controllable if it is possible to construct a control signal u(t) such that the following
equation is satisfied :

There are actually, numerous values of u(τ) which satisfy this equation provided
because otherwise the link between input and the corresponding state variable gets broken and
hence it is no longer possible to control that particular state variable.
It therefore follows that the necessary condition of complete controllability is that the
vector B should not have any zero elements. If any element of this vector is zero, then the
corresponding state variable is not controllable. It can be further shown that the condition stated
here is in fact both necessary and sufficient.
The result just obtained, can be extended to the case where the control force u is an m-
dimensional vector. For the system described by

39
 where

The necessary and sufficient condition for controllability is that the matrix must
have no row with all zeros. It is observed from the above equation that if any row of the matrix

is zero, it is not possible to influence the corresponding state variable by the control forces
and hence the particular state variable is uncontrollable. In case A has a Jordan block the

elements of any row of that correspond to the last row of the Jordan block are not all zero.

Example: Consider the system with state equation

which gives the Eigen-values λl = - 1, λ2 = - 2, λ3 = - 3 (Distinct roots)

Choosing Vander Monde matrix as modal matrix, we have

Therefore, the state equation in canonical form is given by

Since no element of 𝐵 is zero, the system is completely controllable.

Let us now test controllability of this system by the Kalman's test.

We have

40
Then

The controllability matrix defined in eqn. (11) is given by

Qc = [B: AB: A2B]

It is easily seen that det Qc ≠ 0 i.e., its rank is r = n = 3: The system is therefore completely
controllable.
The state controllability depends on how the state variables are defined for a given system.
Test Methods for Observability
The Kalman's test of observability is as follows:
A general nth order multi-input multi-output linear time-invariant system:
x= Ax + Bu
y=Cx
is completely observable if and only if the rank of composite matrix
Qo = [CT : ATCT : (AT )2 CT :...(AT)n-1CT] is n. (13)
This condition is also referred as the pair (A, C) being observable.
Gilbert’s Method:
Consider the state model of an nth order single-output linear time-invariant system,

The state equation may be transformed to the canonical form by the linear transformation
x = Mv.
The resulting state and output equations are

(12 a)

(12 b)

Since diagonalization decouples the states, no state now contains any information regarding
any other state, i.e., each state must be independently observable. It therefore follows that for
a state to be observed through the output y, its corresponding coefficient in eqn. (12 b)
should be nonzero.
If any particular 𝐶 is zero, the corresponding vi can have any value without its effect showing
up in the output y. Thus the necessary (it is also sufficient) condition for complete state
observability is that none of the 𝐶 ’s (i.e., none of the elements of 𝐶 = CM should be zero).
41
The result may be extended to the case of multi-input-multi-output systems where the output
vector, after canonical transformation is given by
The result may be extended to the case of multi-input-multi-output systems where the output
vector, after canonical transformation is given by

i.e.
The necessary condition for complete observability is that none of the columns of the matrix
𝐶 be zero. This is the Gilbert's test.

Example: Let us examine the observability of the system given below.

(i)

(ii)

The characteristic equation is

i.e. λ (λ-+1 ) (λ+2)= 0

Therefore the Eigen-values of matrix A are

λ1= 0; λ2 = -1; λ3 = -2
The Vander Monde matrix is then
1 1 1
M  0  1  2

0 1 4 

Under the linear transformation x = Mv, the output is given by

It is found that the system is not completely observable, since the state variable v2 is hidden
from observation.
Now applying the Kalman's test to the same system we get from eqns. (i) and (ii)

42
Therefore the composite matrix defined in eqn. (13) is given by

the rank of the matrix Qo is = 2, while n = 3. Hence one of the state variables is unobservable.
Problem:

43
44
Example problems:

45
46
47
(Note: * indicates Transpose)

Problems:
1. Given the simplified pitch response of an aircraft for elevator input which is a second order
differential equation 𝜃̈+ 2 𝜃̇ + 5 θ = – δ
Rewrite the equation in state space form. Also,
(i) Determine the characteristic equation, undamped natural frequency, damping ratio,
time to half / double amplitude and cycles to half / double amplitude
(ii) Find the Eigen values of system
(iii) Plot the Eigen values on the S – plane and describe the motion one might expect
from these Eigen values.
2. The Dutch roll motion of an airplane is approximated using the following equations:

Assuming the flight velocity and corresponding dimensional derivatives of the aircraft to
be:
Yβ = – 2.38 m/s2; Yr = 0.753 m/s; 𝑌 = 1.596 m/s2; u0 = 47 m/s;
Nβ = 0.64 /s2. Nr = – 0.34 /sec; 𝑁 = – 0.616 /s2.
(a) Determine the Dutch roll eigenvalues. (b) What is the damping ratio and undamped
natural frequency? (c) What is the period and time to half amplitude of the motion?
3. The soft landing of a lunar module descending on the moon can be modeled as shown in
Figure P3.8. Define the state variables as X1 = y, X2 = dy/dt, X3 = m and the control as u =
dm/dt. Assume that g is the gravity constant on the moon. Find a state variable model for this
system. Is this a linear model?
48
State space equations are:

𝑋̇ = 𝑋 (vertical velocity of descent)

𝑋̇ = −𝑔 (acceleration = retarding thrust – moon’s gravitational pull)

𝑋̇ = 𝑢 (rate of mass depletion)

This is a set of nonlinear equations

49