Modern Control System

ECEg 4510

By: Ethel.C
Chapter 1: Mathematical Modeling of
Dynamic Systems in State Space

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space
A control engineer’s role is to
❑ Modern engineering systems are more complex to handle precise develop a system model, check its
stability, and if the system is unstable,
tasks. Since 1960, modern control theory has used the concept of design a controller(to increase system
state(helps describe a system's condition at any given moment) performance) while also evaluating
performance. For stability analysis,
to analyze and design these systems, leveraging computers for better various techniques are available,
performance. including Routh-Hurwitz criteria,
Root Locus, Nyquist Plot, Bode
❑ Complex systems may have multiple inputs and multiple outputs Plot, and Nichols Chart. These are
classical techniques. In classical
and may be time varying. control design, commonly used
controllers include the PID controller
Modern Control Theory Versus Conventional Control Theory and lag-lead controller, both based
on the transfer function approach,
developed between 1920 and 1950.
❑ The conventional control theory is used for an LTI system and Research in this area is still ongoing.
With the advancement of digital
SISO , where as in modern control theory state space approach computers, the state-space
can be used to represent linear and non-linear system , time technique emerged after 1960, based
on the time-domain approach.
variant or time invariant system and MIMO or SISO However, before 1920, all
methodologies were also time-domain
❑ Classical approach use complex frequency domain approach based, as transfer functions had not
❑ Modern control theory use time domain approach yet been developed. After 1978, both
state-space and classical
❑ State space approach handle non zero- initial condition approaches were combined, leading
to the development of robust control
❑ State equations are first order equations irrespective of system techniques.
order, inputs and outputs.
Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space

Modern Control Theory Versus Conventional Control Theory

❑ A state space model helps us predict how a system will behave in the future based
on its current state and input.
Space station orientation
❑ If we want to predict its future orientation we need to know :

❖ Current State – The current position, velocity, and

rotation of the space station(rotational acceleration).

❖ Control Input – Any external forces or torques applied to

adjust its orientation.

❖ Before we proceed further, we must define state, state

variables, state vector, and state space.

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space

State.
❑ The state of a dynamic system is the smallest set of variables (minimum) (called state
variables) such that the knowledge of these variables at t=t o , together with the
knowledge of the input for t  t o,completely determines the behavior of the system for any
time t  t o .

Note that the concept of state is by no means limited to physical systems. It is applicable to
biological systems, economic systems, social systems, and others.
State Variables.
The state variables of a dynamic system are the variables making up the smallest set of
variables that determine the state of the dynamic system. If at least n variables x,, x2,..
.,X, are needed to completely describe the behavior of a dynamic system (so that
once the input is given for t  t o and the initial state at t = to is specified, the future state of
the systems completely determined), then such n variables are a set of state variables.

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space

❑ Consider a mass-spring-damper system where the

input is an applied force. To predict the output,
which is the position, we need to know the current
state of the system—specifically, the current
position and the current speed (velocity) of the
mass(t=t0).

❑ These two variables, position and speed, are the

state variables of the system.
•• •
f (t) = m x + b x + kx ❑ The input to the system is the applied force, and the
output is the future state of the system, which
we choose the position and velocity as our state variables includes the position and speed at any given
x  time(t>=to).
x = • ❑ We can now define these state variables as a vector
 x 
that represents the state of the system.

The position variable captures the potential energy stored in the spring, while the velocity variable captures the kinetic energy stored by the mass.
The damper only dissipates energy, it doesn't store energy. Often when choosing state variables it is helpful to consider what variables capture the energy
stored in the system.

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space

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 the n components of a vector x. Such a vector is
called a state vector. A state vector is thus a vector that determines uniquely the system state x(t)
for any time t  t o ,once the state at t = to is given and the input u(t) for t >= to is specified.

State Space.
The n-dimensional space whose coordinate axes consist of the xl axis, x2 axis, ...,x, axis,
where xl,x2,.. .,x, are state variables; is called a state space. Any state can be represented
by a point in the state space. • If the mass is at rest at x=2m and v=0, the state is:
 x  2 v
•• •
f (t) = m x + b x + kx x = • =  
 x  0 
we choose the position and x
x-axis → Position
velocity as our state variables • If the mass is moving at x=2m with v=3m/s, its state is:
y-axis → Velocity
x  x   x  2
x = • =   x = • =  
 x   v   x  3 

❑ The system evolves over time as the mass moves, tracing a path in state space.
Ethel.C
 State Space Representation

Input Variables (r) Output Variables (r)

u1 y1

u2 Nth order system y2

Control system state variable (n)
ur ym

Input Vector(rx1) State Vector(nx1) Output Vector(mx1)

 u1   x1   x1 
u  x  x 
 2  2  2
.  .  . 
  X1 X2 X3 Xn    
.  .  . 
u r  xn  xn 
  System Block Diagram     Ethel.C
 State Space Representation

❑ In state-space analysis we are concerned with three types of

variables that are involved in the modeling of dynamic
systems:
1. input variables Input Vector(rx1) State Vector(nx1) Output Vector(mx1)
2. output variables,  u1   x1   y1 
u  x  y 
3. State variables.  2  2  2
u(t)= .  x(t)= .  y(t)= . 
     
.  .  . 
u r  xn 
     ym 
 
The state x, input u and output y interacts with each other with two equations
1. State Equation: In dynamical systems theory, a state equation is a first order differential
equation governing a state variable.
❑ It is a set of n simultaneous first-order differential equation with n variables.
❑ Define how x(t) changes over time based on the current state and the input applied to the
system.
2. Output Equation: Is the linear combination of state variables and the inputs.
dx •
State Equation  = x = f (x, u, t) Output Equation  y = g(x, u, t)
dt Ethel.C
 State Space Representation
State Equation Output Equation

In the form of n number of first order differential The output y1,y2 ….ym(t) of the system
equations as shown below: may be given as

dx • Output Equation  y = g(x, u, t)

In Vector Notation  = x = f (x, u, t)
dt

Ethel.C
 State Space Representation

State Model of Linear System

❑ For LTI systems, the first derivatives of state variables can be expressed as a linear combination of
sate variables and inputs

State Equation A: is state matrix of size (n×n): Defines how

states evolve over time.

B: is the input matrix of size (n×r): Defines

In Vector Matrix Form Input Vector(rx1) State Vector(nx1) how inputs affects the states.
 u1   x1 
• u  x 
x(t) = Ax(t) + Bu(t)  2  2
u(t)= .  x(t)= . 
   
.  . 
u r  xn 
   
Ethel.C
 State Space Representation

State Model of Linear System

State Equation
In Vector Matrix Form
•
x(t) = Ax(t) + Bu(t)

Output Equation:

y(t) = Cx(t) + Du(t)

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space

c11 c12 ........c1n   x1  d11 d12 ........d1r   u1 

c c ........c   x  d d ........d  u 
Output Equation: y(t) = Cx(t) + Du(t) =  21 11 2n   2  +  21 11 2r  2
.  .  .  . 
     
c m1 c m2 .........c mn   x n  d m1 d m2 .........d mr   u r 

C : Output matrix of size (m×n): Defining how the system states contribute to the output
D: Transmission matrix of size (m×r): Describes any direct relationship between inputs and outputs.
whether the input has an instantaneous (direct) effect on the output y.

❑ If D=0, the output depends only on the state x, meaning there is no direct effect of the input on the
output.
❑ If D≠0, the input u directly contributes to the output without delay, meaning the system has a
feedthrough (or direct transmission) component.

❖ X(t): is the state vector of size (n×1)

❖ Y(t): is the output vector of size (m×1)
❖ U(t): is the input vector of size (r×1)

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space

State Model of Linear System

❑ If vector functions f and/or g involve time t explicitly, then the system is called a time varying
system.
•
State Equation: x(t) = A(t)x(t) + B(t)u(t)
Output Equation: y(t) = C(t)x(t) + D(t)u(t)

❑ If vector functions f and g do not involve time t explicitly then the system is called a time-invariant
system.
•
State Equation: x(t) = Ax(t) + Bu(t)
Output Equation: y(t) = Cx(t) + Du(t)

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space

Block diagram representation of state model of a linear multi-input multi-output system

•
x(t) = A(t)x(t) + B(t)u(t)
y(t) = C(t)x(t) + D(t)u(t)

Selection of state variables

1.Number of state variables should be minimum.
2. The number of state variables must be linearly independent.
V=iR
3. The number of state variables is equal to number of independent energy storage element.
4. The number of state variables is equal to the order of differential equation.
5. The number of state variables are equal to number of integrators.

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space
Example

•
x(t) = A(t)x(t) + B(t)u(t)
y(t) = C(t)x(t) + D(t)u(t)
The output as net force

❑ It means that the system has an immediate response to the input

force.
❑ The output depends not just on the internal state (position & velocity)
but also directly on the external forcing function f.
❑ In physical terms, if a force is applied, the system reacts
instantaneously rather than just through its internal state evolution.
 Mathematical Modeling of Dynamic Systems in State Space
Example. RLC

Step.1 Develop relevant differential equations from given systems

di(t)
Vi (t) = i(t)R + L + Vo (t)
dt
Step. 2 Identify or select state variables. (Number of storage element =2(Inductor and
Capacitor)
State Variable 1: x1=VC (capacitor voltage) di (t) 1 dV
VL = L L , Vo =  i c (t)dt,i c (t) = i L (t) = C o
State Variable 2: x2=iL (inductor current) dt C dt

These variables must describe how energy is stored in the system. They must appear in first-order differential equations to track the system’s
evolution over time. In electrical circuits, the best choice for state variables comes from components that store energy.
 Mathematical Modeling of Dynamic Systems in State Space
•
• • • •
dV i dV x(t) = A(t)x(t) + B(t)u(t) y(t) = C(t)x(t) + D(t)u(t)
x1 = Vc find x1 and x1  x1 = Vc = c = L .....where i c = C c
dt C dt
 1 y(t) = Vc (t)
 
• •   0
•

C   Vc  0 
di di(t)
x 2 = iL x 2 = i L = L  Vi (t) = i(t)R + L + Vo (t)  Vc    Vc (t) 
dt dt     y(t) = [1 0]   + 0.u(t)
 =    +   Vi i L (t) 
• • di di (t) −i (t) •  
x 2 = i L = L  L = L R − VC (t) + Vi (t)
i   - 1 R  i L   1   Vc (t) 
dt dt L L  -    y(t) = Vc (t) = [1 0] 
 L L L 
i L (t) 
• iL
x1 =
C
• • di L (t) -i L (t)
x2 = i L = = R-VC (t)+Vi (t)
dt L

Exercise: Given the electrical network of Figure, find a state-space representation if the output is the
current through the resistor.

Norman Nise. Page.126-127

Ethel.C
 Mathematical Modeling of Dynamic Systems in State Space

Exercise: Determine the state-space model for a translational body mass with spring and
damper, upon which an input force F acts.

Exercise: Derive a state space model for the system shown.

Ethel.C
 State Space from Transfer Function

❑ Let us represent a general, nth-order, linear differential equation with constant coefficients in
state space in the phase-variable form. We will then show how to apply this representation to
transfer functions. Consider the differential equation

❑ A convenient way to choose state variables is to choose the output, y(t), and its (n - 1 )
derivatives as the state variables. This choice is called the phase-variable choice

Ethel.C
 State Space from Transfer Function

In vector-matrix form

Since the solution to the differential equation is y(t), or X1, the output equation is

Ethel.C
 State Space from Transfer Function

Example: Find the state-space representation in phase-variable form for the transfer function shown

Cross-multiplying yields

❑ Since the output is c = x1, the combined state and output equations are

Ethel.C
 State Space from Transfer Function

Example: Find the state-space representation in phase-variable form for the transfer function shown

Ethel.C
 State-Space Representation in Canonical Forms.

1. Controllable Canonical Form

Consider a system defined by:

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

State Equation Output Equation

•
x(t) = A(t)x(t) + B(t)u(t) y(t) = C(t)x(t) + D(t)u(t)

Ethel.C
 State-Space Representation in Canonical Forms.

Consider a system defined by:

Example: Controllable Canonical Form

Controllable Canonical Form

Ethel.C
 State-Space Representation in Canonical Forms.

2. Observable Canonical Form

Consider a system defined by:

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

❑ Used for designing observers

A T , BT , CT of controllable canonical form
Output Equation
State Equation
•
x(t) = A(t)x(t) + B(t)u(t) y(t) = C(t)x(t) + D(t)u(t)

Ethel.C
 State-Space Representation in Canonical Forms.

3. Diagonal Canonical Form

❑ Here we consider the case where the denominator polynomial involves only distinct root
Y(s) a 2s 2 + a1s + a o c1 c2 c3
= = + +  Distinict roots
U(s) b3s + b 2s + b1s + b 0 s + 1 s +  2 s +  3
3 2 1

Y(s) c1 (s +  2 )(s +  3 ) + c 2 (s + 1 )(s +  3 ) + c3 (s + 1 )(s +  2 )

=
U(s) (s + 1 )(s +  2 )(s +  3 )
Lets compare equations
a 2s 2 + a1s + a o = c1 (s +  2 )(s +  3 ) + c 2 (s + 1 )(s +  3 ) + c3 (s + 1 )(s +  2 )
Now,lets find c1 , c 2 and c3  lets s = −1 , s = − 2 , and s = − 3
for s = − 2
a 2  2 2 − a1 2 + a o = c 2 (− 2 + 1 )(− 2 +  3 )
a 2  2 2 − a1 2 + a o a 2 12 − a11 + a o a 2  32 − a1 3 + a o
c2 = , c1 = , c3 =
(− 2 + 1 )(− 2 +  3 ) (−1 +  3 )(−1 +  3 ) (− 3 + 1 )(− 3 +  2 )
Ethel.C
 State-Space Representation in Canonical Forms.

3. Diagonal Canonical Form

Y(s) c c c c c c
= 1 + 2 + 3  Y(s) = 1 U(s) + 2 U(s) + 3 U(s)
U(s) s + 1 s +  2 s +  3 s + 1 s + 2 s + 3
u(s) u(s) u(s)
Now let  x1 = , x2 = and x 3 =  y(s) = c1x1 + c 2 x 2 + c3 x 3
s + 1 s + 2 s + 1
x1s + x11 = u(s)
• •   
x1s = − x11 + u(s) → Inverse laplace → x1 = − x11 + u x
   −1 0
1 x
0   1  1
•    
x =Ax+Bu   x 2  = 0 -1 0   x 2  + 1 u
•
•
x1 = − x11 + u  •  0 0 -    1
x3   2 
x3   
•    
x 2 = − x 2 2 + u  
•  x1 
x 3 = − x 3 3 + u y=Cx+Du  y = [c1 c 2
 
c3 ]  x 2 
y(s) = c1x1 + c 2 x 2 + c3 x 3  
x3 
 
Ethel.C
 State-Space Representation in Canonical Forms.

3. Diagonal Canonical Form

❑ Consider a system defined by:

State Equation Output Equation

•
x(t) = A(t)x(t) + B(t)u(t) y(t) = C(t)x(t) + D(t)u(t)

Ethel.C
 State-Space Representation in Canonical Forms.

4. Jordan Canonical Form

❑ The state model having repeated roots in the denominator is called Jordan canonical form
❑ This method is based on partial fraction method

Y(s) b3s3 + b 2s 2 + b1s1 + b 0

= 3  Assume that the roots of → s 3
+ a s 2
+ a 1 + a 0 are repeated
s1

s + a 2s + a1s + a 0
2 1 2
U(s)
b3s3 + b 2s 2 + b1s1 + b 0
Y(s) = U(s)
(s + 1 ) (s +  3 )
2

c1 c2 c3
Y(s) = b3 U(s) + U(s) + U(s) + U(s)
(s + 1 ) 2
(s + 1 ) s + 3
Y(s) = b3 U(s) + c1x1 (s) + c 2 x 2 (s) + c3 x 3 (s)
u(s) u(s) u(s)
x1 (s) = , x 2 (s) = and x 3 (s) =
(s + 1 ) 2
(s + 1 ) (s +  3 )

Ethel.C
 State-Space Representation in Canonical Forms.

4. Jordan Canonical Form

u(s)
x1 (s) (s + 1 ) 2 1 x (s)
= = → x1 (s) = 2
x 2 (s) u(s) s + 1 s + 1 •   
x
 1   −1 1  x1 
(s + 1 ) 0 0 
•   
•
    
x1 (s)s + x1 (s)1 = x 2 (s) → x1 (s)s = − x1 (s)1 + x 2 (s) x =Ax+Bu   x 2  = 0 -1 0   x 2  + 1  u
•  •  0 0 -    1 
x1 = − x11 + x 2 x3   3 
x3   
   
•
x 2 = − x 2 2 + u  
•  x1 
x 3 = − x 3 3 + u  
y=Cx+Du  y = [c1 c 2 c3 ]  x 2  + b 3 u
y = c1x1 + c2 x 2 + c3 x 3 + b3u  
x3 
 

Ethel.C
 State-Space Representation in Canonical Forms.

4. Jordan Canonical Form

State Equation Output Equation

•
x(t) = A(t)x(t) + B(t)u(t) y(t) = C(t)x(t) + D(t)u(t)

Ethel.C
 State-Space Representation in Canonical Forms.

Example

Ethel.C
 Converting from State Space to Transfer Function

❑ In the last section, we have seen how to convert transfer functions into state-space
representations. Now we move in the opposite direction and convert the state-space
representation into a transfer function.

•
x(t) = Ax(t) + Bu(t)
−1 adj ( sI − A )
y(t) = Cx(t) + Du(t) (sI − A) =
det ( sl − A )
Take the Laplace transform assuming zero initial conditions
•
x(t) = Ax(t) + Bu(t)  sX(s) = AX(s) + BU(s) Y(s) = C(sI − A) −1 BU(s) + DU(s)
y(t) = Cx(t) + Du(t)  Y(s) = CX(s) + DU(s)
Y(s) = [C(sI − A) −1 B]U(s)
Solving for X(s)
X(s)[sI − A] = BU(s) → I identity matrix Y(s)
= C(sI − A) −1 B + D
X(s) = [sI − A]−1 BU(s) U(s)

Ethel.C
 Converting from State Space to Transfer Function

Example: Determine transfer function Y(s)

Step 3: G(s) = = C(sI-A)−1B + D =
0 1 0 U(s)
x=  x +   u, y= 1 0 x
-1 -2  1  s + 2 1  0  1
[1 0]     [1 0]  
s -1  =  -1 s  1 
+0= 2  s
Step 1 (sI-A)=   s + 2s + 1
2
s + 2s + 1
1 s+2 
Y(s) −1 1
s + 2 1 G(s) = = C(sI-A) B + D = 2
-1  U(s) s + 2s + 1
adj ( sI − A )
= 2 
−1 s
Step 2 (sI-A) =
det ( sl − A ) s + 2s + 1 Exercise: Convert the state and output transfer function

s + 2 -1
cofactor ( sI − A ) = 
1 s 
s + 2 1
adj ( sI − A ) = (cofactor ( sI − A )) = 
T

-1 s 
det ( sl − A ) = s(s + 2) − (−1) = s 2 + 2s + 1
Ethel.C
 Eigenvalues and Eigenvectors in State-Space Stability Analysis

Eigenvalues of an n x n Matrix A. The eigenvalues of an n X n matrix A are the

roots of the characteristic equation and satisfies the equation Ax i = x i

The eigenvalues are also called the characteristic roots. Consider, for example, the
following matrix A

The characteristic equation is

Ethel.C
 Eigenvalues and Eigenvectors in State-Space Stability Analysis

Properties of Eigenvalues

❑ If the coefficient of the Matrix [A] is real, then its eigen values are always
real or in complex conjugate pair
3 1 2
x = 1 2 3  x
1 2 5 

❑ If the Matrix A is real symmetric(row and columns elements are same,

the eigenvalues are always real (no complex conjugate pair)

1 2 4
x =  2 5 8  x
4 8 7 

Ethel.C
 Eigenvalues and Eigenvectors in State-Space Stability Analysis

❑ The eigenvalues of [A] and its transpose matrix [𝑨𝑻 ] are always the same.

[A] → 1 ,  2 ,  3
[A T ] → 1 ,  2 ,  3

❑ Eigenvalues of [A] and it inverse [𝑨−𝟏 ]: [A] having eigenvalues

 i , i=1,2,3,4....n
then 1 are eigenvalues of [𝑨−𝟏 ]
, i=1,2,3,4....n
i
Note:  A  must be singular matrix (detA  0)

❑ Matrix is singular if and only if it has an eigenvalues of zero.

❑ Eigenvalues can be zero but not eigenvectors
Ax i = x i
1 2   −2   0 
 2 4  1  =  0 
    
 −2 
=   =0
1 
Ethel.C
 Cayley-Hamilton Theorem

❑ Every square matrix(A) satisfies its own characteristic equation.

 n + a n −1 n −1 + .... + a 2  2 + a1 + a 0 = 0  Characterstice equation

❑ According to the Cayley-Hamilton theorem, if we replace λ in the characteristic
equation with the matrix A, the result will be the zero matrix:
P(A)=0

❑ Where P(A) is the characteristic polynomial evaluated at A.

A n + a n −1A n −1 + .... + a 2 A 2 + a1A + a 0 I = 0

3 2 
A=  then det(A-I) = 0
1 4 
det(A-I) =  2 − 7 + 10 = 0
Then substitute A into the charactristic polynomial
3 2  3 2  3 2  1 0  0 0 
=A 2 − 7A + 10 = 0 =  •
   − 7 1 4  +10 0 1 = 0 0 
1 4  1 4       
Ethel.C
 Eigenvalues and Eigenvectors in State-Space Stability Analysis

❑ Same eigenvector can not be associated with different eigenvalues

Ax = 1x, Ax =  2 x
❑ These eigenvalues determine how the system behaves
1 −  2 = 0 dynamically
Stability
❑ Eigenvalues play a key role in stability analysis because they reveal how different modes
of the system evolve over time. When analyzing the stability of a linear system, the
eigenvalues of its characteristic matrix determine the behavior of the system.
Y(s) = C(sI − A) −1 BU(s) + DU(s) Example:
Y(s) = [C(sI − A) −1 B]U(s)
 −2 0  0 1  2 1
Y(s)
= C(sI − A) −1 B + D A=  A=  A=
U(s)
0 -3 -1 0  0 1
Y(s) adj ( sI − A )
=C B+D
U(s) det ( sl − A )
❑ Stable (all real parts are negative).
Y(s) Cadj ( sI − A ) B + det ( sI − A ) D ❑ Unstable System(One or more eigenvalue is positive)
=
U(s) det ( sI − A ) ❑ Marginally stable (Imaginary)

Ethel.C
 The Concept of Diagonalization

Introduction to Diagonalization
❖ In system design, analyzing the original model can be complex.
❖ Diagonalization simplifies analysis by transforming a system into a more manageable form while
preserving its properties.
❖ This transformation is particularly useful in control system applications.
❖ The method involves converting the state-space model into a canonical (diagonal) form, where
the system matrix A has only diagonal elements with zero elsewhere.

Applications

✓ Investigation of system properties.

✓ Time response evaluation.
✓ Checking controllability and observability.
✓ Control design.
✓ Model order reduction techniques.

Ethel.C
 The Concept of Diagonalization

Motivations for diagonalization

Lets take a homogeneous state equation

•
x(t) = Ax(t)
❖ x is a state vector with multiple components (X₁, X₂, …, Xₙ).
❖ A is an n × n matrix that defines how the system evolves.

Decoupled (Diagonal) Systems

Consider the simplest case where A is diagonal:

•
x(t) = Dx(t)
where D is a diagonal matrix:

Ethel.C
 The Concept of Diagonalization

Decoupled System
•
x(t) = Dx(t)

➢ Each equation takes the form

➢ The solution is:

➢ In matrix form, this can be written as: The solution to the state equation becomes simpler when the
state matrix A is diagonalized with D. For decoupled
systems, this process is particularly straightforward
because they are already diagonalized. However, in real-
world systems, the matrix A might not be diagonal,
indicating that the system is coupled, which adds
Where: complexity to the solution.

This is the matrix exponential of D.

Ethel.C
 The Concept of Diagonalization

Coupled system

❖ Lets take coupled system represented in the state-space form as

❑ The system matrix A is not diagonal, and it is beneficial to transform it into a diagonal structure.
To achieve this, lets define a new state vector z by using a transformation matrix P

x = Pz where P is nxn matrix •

−1 −1
•
z = P APz + P Bu  z = Az + Bu
−1
z=P x
P −1AP = A : a diagonal matrix of eigenvalues
• • •
z = Az + Bu
−1 −1
z = P x = P (Ax + Bu) −1
P B = B : is the transformed input matrix.
• •
For the output equation, we substitute x = Pz
z = P x = P −1 (APz + Bu)
−1
y = Cz + Du
• • y = (CP ) z + Du  y = Cz + Du
z = P x = P −1APz + P −1Bu
−1
C = CP:is the transformed output matrix
Ethel.C
 The Concept of Diagonalization

Modal Matrix
The transformation matrix P is chosen such that P −1AP is diagonal. This is accomplished by selecting P
as the Modal Matrix, which consists of the columns of eigenvectors of A.
•
z = Az + Bu

P=[v1 v 2 v3 ....v n ] y = Cz + Du

Vande Monde Matrix

❑ If A is nxn matrix with distinct eigenvalues and it is in canonical form

Ethel.C
 The Concept of Diagonalization

Example Consider the following state-space representation of a system where A is in canonical form

Ethel.C
 The Concept of Diagonalization

Simplifying gives

The output equation, is modified to

Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION

1. Solution of Homogeneous State Equations and State Transition Matrix :

1.1 Homogeneous state equations: refer to a system of state-space equations in which the
external input u(t) is zero. These equations describe the behavior of the system based solely on
its initial state and the internal dynamics, without any external forcing or input.
•
The solution of → x(t) = Ax(t) is x(t)=e At x(0)
Homogeneous State Equation We can write the solution of the homogeneous state quation as:
• x(t)=e At x(0) → x(t)=(t)x(0)
x(t) = Ax(t)
where :
y(t) = Cx(t)
(t) = e At :
x(0) : Intial states of state variables as a state vector
(t) : State Transtion Matrix
e At → Matrix Exponential

Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION

1. Solution of Homogeneous State Equations and State Transition Matrix :

1.2 State Transition Matrix (STM)

❖ Describes how the state of a system evolves over time when the input is zero, i.e., it captures the
evolution of the system’s internal state without external influence. It is derived from the homogeneous
state equations and typically takes the form
(t) = e At where A is the system matrix and t is time.
❖ Zero-Input Response(Natural response): The STM is used to model the system’s behavior in
response to initial conditions (ignoring the effect of external inputs). It is referred to as the free
response because it is solely driven by the initial state vector x(0), without considering any forcing
inputs.

In particular, if the matrix A is diagonal, then:

Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION

Properties of State-Transition Matrices

We shall now summarize the important properties of the state-transition matrix Φ(t).For the time-
invariant system

Solution of Homogeneous State Equations

❑ If the initial states of state variables is given as initial state •

x(t) = Ax(t)
vectors x(0) the solution of homogeneous state equation can be
x(t)=e At x(0)
obtained by solving the state transition matrix Φ(t) or Matrix
x(t)=(t)x(0)
exponential ( (t) = e At )

Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION

Methods
3. Diagonal Method
1. Power Series Method 4. Caley Hamilton method
• x(t)=(t)x(0) = e At x(0)
x(t) = Ax(t) → x(t)=e At x(0)
1 2 2 1 k k
x(t)=(t)x(0) (t) = e = I + At + A t + .... + A t
At

2! 2!

Matrix Exponential:

2. Laplace Transform
•
Steps
x(t) = Ax(t) → sX(s) − X(0) = AX(s) → sX(s) − AX(s) = X(0)
X(s)(sI − A) = X(0) adj(sI-A)
1.Find (s)=(sI-A) −1 =
X(s) = (sI − A) −1 X(0) det (sI-A)
X(s) = (s)X(0) → (s) = (sI − A) −1 2.X(s) = (s)X(0)
 L−1 3.x(t) = L−1 (X(s)) = (t)x(0)
x(t) = (t)x(0) and y(t)=Cx(t) 4.y(t0 = Cx(t)

Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION

Example: Obtain the state transition matrix , solution of the state equations and output
equation(time response of the system for initial state variables without input)
•   
• x1 0 1  x1  1 
x(t) = Ax(t) =  •  =  =
-3    0 
, x(0)
   −2  
x2  x2 
 
x1
y(t) = Cx(t) = [1 0]  
 
x2 
Solution
1. STM

Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION

2. State equations

x(t)=(t)x(0) = e At x(0)
 2e − t − e −2t e − t − e 2t  1   2e − t − e −2t 
x(t) =  −t 2t   
=  −t 2t 
 −2e + 2e 2t
-e + 2e     −2e + 2e 
t
0

3. Response
 2e − t − e −2t  −t −2t
y(t) = Cx(t) = [1 0]   =2e − e ❖ System time response because of
−t
 −2e + 2e  initial conditions x(0) or zero input
2t

Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION

3. Diagonal Method

In particular, if the matrix A is diagonal(Decoupled), then: For coupled

•
x = Ax(t)
x = Pz where P is nxn matrix
z = P −1x
• •
−1
z = P x = P −1 (Ax)
• •
−1
z = P x = P −1 (APz)
• •
−1
z = P x = P −1APz
4. Caley Hamilton method

Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION

2. Solution of Non-Homogeneous State Equations

•
x(t) = Ax(t) + Bu(t)
y(t) = Cx(t) + Du(t)

Laplace Transform Method

•
x(t) = Ax(t) + Bu(t) L → sX(s) − X(0) = AX(s) + BU(s)
sX(s) − AX(s) = X(0) + BU(s) X(s)(sI − A) = X(0) + BU(s)
X(s) = (sI − A) −1 X(0) + (sI − A) −1 BU(s)
X(s) = (s)X(0) + (s)BU(s) X(s) = (s)X(0) + (s)BU(s)
Multiplication in s domain is
Y(s) = CX(s) + DU(s) convolution integral in time Y(s) = CX(s) + DU(s)
domain
 L−1 t

t t x(t) = e At x(0) +  e A(t − ) Bu( )d

x(t) = (t)x(0) +  (t − )Bu( )d x(t) = e At x (0) +  e A(t − ) Bu( )d o

o o
Ethel.C
 SOLVING THE TIME-INVARIANT STATE EQUATION
Steps
1.Find (s)=(sI-A) −1 : State Transition Matrix
2.Compute (s)x(0) if x(0)=0,(s)x(0) = 0
3.Compute (s)BU(s)
Steps
4.X(s)=(s)x(0) + (s)BU(s)
1.Find (s)=(sI-A) −1 : State Transition Matrix
5.Y(s) = CX(s) + DU(s)
−1  s+3 1 
6.X(s) → L X(s) = x(t)  (s + 1)(s + 2) (s + 1)(s + 2) 
7.y(t) = Cx(t) + Du(t) or Y(s) → L−1Y(s) → x(t) (s) =  
 −2 s 
 (s + 1)(s + 2) (s + 1)(s + 2) 

Example Obtain the time response of the following
system: where u ( t )is the unit-step function occurring
at t = 0 2.Compute (s)x(0) if x(0)=0,
•
• 
x1 0
 
1  x1  0  0   s+3 1 
x(t) = Ax(t) =  •  =   (s + 1)(s + 2) (s + 1)(s + 2) 
+ =
-3    1` 0 
u(t), x(0)
   −2  
x2  x2    x(0) = 0
   −2 s 
x1
y(t) = Cx(t) = [1 0]    (s + 1)(s + 2) (s + 1)(s + 2) 
  
x2  Ethel.C
SOLVING THE TIME-INARIANT
SOLVING THE TIME-INVARIANT
STATE EQUATIONSTATE EQUATION
1
3.Compute (s)BU(s):u(s) → Lu(t)=
s 5.Y(s) = CX(s) + DU(s)
 s+3 1   1 
 (s + 1)(s + 2) 
(s + 1)(s + 2)  0  1  s(s + 1)(s + 2) 
   Y(s) = [1 0]  = 1
 −2 s  1  s  1  s(s + 1)(s + 2)
 (s + 1)(s + 2) 
(s + 1)(s + 2)   (s + 1)(s + 2) 
  
4.X(s)=(s)x(0) + (s)BU(s) → X(s)=0 + (s)BU(s) 1 − t 1 −2t
y(t) = −e + e
2 2
 s+3 1 
 (s + 1)(s + 2) 0 
(s + 1)(s + 2)   
X(s) =   1
 −2 s  
 (s + 1)(s + 2) s 
 (s + 1)(s + 2) 
 1 
 s(s + 1)(s + 2) 
X(s) =   Find inverse laplace transform x(t)
 1 
 (s + 1)(s + 2) 
 
x(t)=_________
Ethel.C
 Controllability and Observability

Concept

❖ In control systems engineering, two fundamental questions must be addressed:

1. Is the system controllable?
2. Is the system observable?

At a high level, control system design follows a structured process. It begins with identifying the physical system to
be controlled, known as the plant. Actuators, such as DC motors, are then used to manipulate the system by
applying control forces or energy. Sensors measure key system states, such as voltage, position, or temperature,
providing essential outputs. The controller processes this sensor data along with a reference command to generate
appropriate control signals for the actuators, ensuring the system operates as intended.
 Controllability and Observability

Concept

However, a controller cannot work properly if:

❖ The actuators cannot influence the necessary parts of the system → the system is not
controllable.

❖ The sensors cannot provide enough information about the system’s state → the system is
not observable.
Ethel.C
 Controllability and Observability

Concept
Example: Consider a car
Assume that the states of the system include:

• X and Y position
• X and Y velocity
• Yaw angle (orientation)
• Yaw rate (rate of change of orientation)
The system has:

• Sensors: A speedometer (measuring velocity) and human

vision (providing additional information).

• Actuators: A steering wheel (controlling yaw) and gas and

brake pedals (controlling velocity).

Ethel.C
 Controllability and Observability

Concept
Concept  •  xx 
x  
Example: Consider a car •

  xy 
u w 
• v   vx   
x(t) = Ax(t) + Bu(t) →  • = A  + B u g 
   vy  u 
•     b 
  
 
  
  
 xx 
 
 xy  u w 
 vx   
y(t) = Cx(t) + Du(t) → y = C   + D u g 
 vy  u 
   b 
  

  


❖ With this setup, the driver can control the car’s position, velocity, yaw angle, and yaw rate
making it controllable. It means that if the system is controllable by using the steering wheel, gas
pedal, and brakes appropriately(actuators), you can move the vehicle from its initial state at
t=to to its final state at t=tf or to any desired position, at any speed.

Ethel.C
 Controllability and Observability

Concept
Example: Consider a car

❖ let’s assume we remove all sensor feedback. This is equivalent to zeroing out the C matrix in the
state-space model, meaning we have no sensor data.
❑ We do not know our position, velocity, or orientation. However, we can still press the pedals
and turn the steering wheel.
Although the car remains mathematically controllable, it is effectively uncontrollable in
practice, since the driver cannot perceive its motion. This example shows that a lack of
observability can make a system practically unusable, even if it is technically controllable.
❖ Now, let’s assume we remove the car’s steering wheel and pedals—essentially zeroing out the B
matrix in the state-space model.
❑ The car continues to move but the driver cannot change its direction or speed.
❑ The system remains observable (the driver can still see the speedometer and surroundings).

Ethel.C
 Controllability and Observability

Concept
Example: Consider a car

1. A system must be both controllable and observable to function properly.

2. Even partial uncontrollability or unobservability can lead to failure.

For example, if a car’s steering wheel is detached but the pedals still work, the car is technically
controllable in one direction, but it is still functionally uncontrollable, because the yaw state
cannot be influenced.

Estimating States Instead of Measuring Them

❑ Observability does not require direct measurement of every state. Instead, states can often be
estimated from available data.

Ethel.C
 Controllability and Observability

Concept
Estimating States Instead of Measuring Them

For instance, in the car example:

❑ If a position sensor is available, velocity can be estimated by differentiating position.
❑ If a speedometer is available, position can be estimated by integrating velocity.

✓ Thus, a system can still be observable even if we do not measure every state directly. This is why
observers (like the Kalman filter) are used in control systems they allow us to estimate states
while reducing the number of required sensors.

Example: radar tracking of an airplane.

What is the state of an airplane given noisy
radar measurements of the airplane’s
position?
This can be addressed using Kalman filter.

Ethel.C
 Controllability and Observability

Concept
A well-designed control system must be both controllable and observable.
❖ Controllability ensures the system can reach any state.
❖ Observability ensures we can infer the system’s state from outputs

❑ Consider the following dynamics for the car example, where the steering is not part of the system.
This represents a car moving straight without control over the steering:
•
 
•
  x = −x
•
 x   −1 0  x  0 
x(t) = Ax(t) + Bu(t) →  •  =   + 1  u P
-2   v
•
v  
0
    v = −2v + u p
   
❑ This system describes a car where the position x is completely independent of the control input, and
only the velocity v is affected by the throttle (uncontrollable).
Now, lets modify the system to: •
•   x = −x + u w
•
 x   −1 0  x  1 0  u W 
x(t) = Ax(t) + Bu(t) →  •  =   + 0
-2   v 1  u P 
•
v    
0
 

 
v = −2v + u p

The car can both move and steer (Controllable).

Ethel.C
 Controllability and Observability

Concept

Lets take another scenario

•  
•
 x   −1 1  x  0 
x(t) = Ax(t) + Bu(t) →  •  =   + 1  u
-2   v
v     
0

   

❑ Now the x equation depends on x and v. So, we can simultaneously stabilize v, and by the
schedule of how we manipulate v, we can also stabilize x.
❑ If these variables are coupled, sometimes it becomes controllable, which is really cool, and
it's useful for control, because it means with one actuator.

Ethel.C
 Controllability and Observability

Definition: Controllability
State Controllability: A system is said to be controllable at time t0, if it is possible by means
of an unconstrained control vector to transfer the system from any initial state x(t0) to any
other state in a finite interval of time x(tf).

Methods for Testing Controllability

In literature there are two tests, one is called Kalman test and the other is called Gilbert test,
both have been used to check the controllability of the system.

Kalman Test for Controllability

The Kalman test assesses controllability using the controllability matrix:
A nxn
= [B AB A 2 B.......A n −1B]
And check the rank or singular (det (0) or non-singular(det(not-0)

Ethel.C
 Controllability and Observability

Example

Gilbert Test for Controllability

1. System distinct eigen values
2. Diagonal •  
3. If we miss B matric value
•
 x   −1 0  x  0 
x(t) = Ax(t) + Bu(t) →  •  =   + 1  u P
-2   v
v     
0

   

Check for those and if all the three conditions are meat system is Uncontrollable

Ethel.C
 Controllability and Observability

Output Controllability
A system is output controllable if you can drive its output (what you measure or care about) from
any initial value to any desired value in finite time, regardless of the states.

Example:

Q out = [CB CAB CA 2 B.......CA n −1B] → D=0

Q out = [CB CAB CA 2 B.......CA n −1B:D] → D  0

Ethel.C
 Controllability and Observability

Stabilizability:
❖ It is defined as the uncontrollable modes are stable and unstable modes are controllable.

Definition: Observability

❖ A system is said to be observable at time to if, with the system in state x(to),it is possible to
determine this state from the observation of the output(measure) over a finite time interval.
❖ Observability means you can reconstruct x(t) from y(t).

Kalman Test: Check the observability matrix: y(t)=Cx(t)

O = [CT (CA) T .......(CA n −1 ) T ]
rank(m)

Ethel.C
 Controllability and Observability

Example: Check for output controllability and observability

Ethel.C
 Controllability and Observability

Detectability: Partially observable system

If the unobservable modes are stable and the observable modes are unstable, the system is said
to be detectable.
•
x(t) = Ax(t) + Bu(t)
• -1 0  1
x(t) =   x +   u
0 2  1
y = [0 1]x
0 in C is unobservable mode but it is stable where as 1 is obesrvable but it is unstable

Transfer Function

Pole Cancellation: In pole zero cancellation the system is unobservable and uncontrollable

Ethel.C
 Review Questions on State-Space Representation

1. Why is state-space modeling used in system analysis? Provide two reasons and explain
an advantage of the state space approach over the Transfer function approach.
2. Define and explain the following terms related to state-space representation:
A. State variables
B. State
C. State vector
D. State space

3. What is required to represent a system in state space, and how many state equations are
needed for an eighth-order system?
4. How do the state equations and output equations function in a system? Explain the role of the
output equation in relation to the state variables.

✓ Nise, N. S. (2020). Control systems engineering (8th ed.). Wiley.

Ethel.C
 Review Questions on State-Space Representation

5. Discuss the selection of state variables in system modeling:

A. What factors influence their choice?
B. What is a convenient choice for electrical networks?
C. If an electrical network has three energy-storage elements, can it have more than three
state variables? Explain.
D. What is meant by the phase-variable form of the state equation?

✓ Nise, N. S. (2020). Control systems engineering (8th ed.). Wiley.

Ethel.C
 Exercise

Practical Application:
In practice, the parameters Ka,Kb,K3,K2,K1, and K0 are determined
through a combination of theoretical modeling and empirical testing.
Factors influencing these parameters include the missile's mass
distribution, aerodynamic surface configurations, propulsion system
characteristics, and control surface effectiveness.

https://www.academia.edu/88005368/Modeling_of_Ballistic_Missile_Dynamics_in_Pitch_Plan
e_And_It_s_Stability

✓ Nise, N. S. (2020). Control systems engineering (8th ed.). Wiley.

Ethel.C
 Exercise

✓ Ogata, K. (2010). Modern control engineering (5th ed.). Prentice Hall.

Ethel.C
 Exercise

✓ Ogata, K. (2010). Modern control engineering (5th ed.). Prentice Hall.

Ethel.C
 Exercise

1. A single-axis solar tracker adjusts the angular position of a solar panel to optimize its Given Parameters:
➢ Jm: Motor inertia - [kg·m²]
exposure to solar radiation. The error signal, which represents the difference between the ➢ Dm: Motor damping - [N·m·s]
➢ Jp: Panel inertia - [kg·m²]
current panel position and the optimal position relative to the sun, is first amplified by a ➢ Dp: Panel damping - [N·m·s]
➢ Kt: Motor torque constant - [N·m/A]
preamplifier with a constant gain Ka. The amplified signal is then passed through a power
➢ Kb: Back electromotive force constant -
amplifier, which is modeled as a first-order system with a time constant Ta. This system [V·s/rad]
➢ Ra: Armature resistance - [Ω] (Ohms)
controls an armature DC motor that is rigidly connected to the panel to achieve precise ➢ La: Armature inductance - [H] (Henries)
➢ Ta: Time constant (Amplifier) - [s]
positioning. The system operates with unity feedback to ensure that the panel remains (seconds)
➢ Ka: Preamplifier gain - [V/V] (Volt per
aligned with the sun for maximum energy capture. Volt)

A. Draw the functional block diagram of the solar tracker system(classical approach).

B. Find the state-space representation (state and output equations) in phase-variable form for
each dynamic subsystem

✓ Nise, N. S. (2020). Control systems engineering (8th ed.).

Wiley.
Case study: Antenna Control

Ethel.C
 Reference

✓ Ogata, K. (2010). Modern control engineering (5th ed.). Prentice Hall.

✓ Nise, N. S. (2020). Control systems engineering (8th ed.). Wiley.

✓ https://www.youtube.com/watch?v=BYvTEfNAi38&t=520s
✓ https://www.youtube.com/results?search_query=steve+brunton
✓ https://ocw.mit.edu/courses/16-30-feedback-control-systems-fall-
2010/1bfc976fcead1982d90c5057511e5ef7_MIT16_30F10_lec05.pdf
✓ http://web.mit.edu/2.14/www/Handouts/StateSpace.pdf

✓ https://www.researchgate.net/publication/353845729_An_Improved_Extended_Kalman_Filter_for_R
adar_Tracking_of_Satellite_Trajectories
✓ https://www.researchgate.net/publication/247370753_Reducing_Structural_Coupling_of_a_Flexible_
Beam_Using_Kalman_Filters#pf2
✓ https://slideplayer.com/slide/5184558/
✓ https://www.slideshare.net/slideshow/kalman-filter-statistics/71856684#2
✓ https://www.chegg.com/homework-help/questions-and-answers/slip-angle-vehicle-yaw-rate-model-
captures-vehicle-dynamics-x-y-y-directions-vehicle-param-q75745379