CHAPTER ONE

STATE SPACE ANALYSIS

Learning Outcomes after completing this chapter will be to enable the reader:
• Appreciate the benefits of state-space analysis.
• Understand the concepts of state, state variables, and state model.
• Model the state-space equation of a control systems.
• Obtain the Transfer Function of a dynamic system from its State space model.

1.1 Introduction
Basically, the main objective of feedback control is to ensure that variables of interest in a
process or a system, thought of as the output signal is kept under control. The variables of
interest can the tracking reference trajectories (known as tracking or servo) which are
maintained close to their set-points (known as regulation). The fundamental reasons for
feedback are (a) to counteract disturbance signal affecting the output, (b) to improve system
performance in the presence of model uncertainty, and (c) To stabilise an unstable plant.
State-space methods are simple tools applied in the analysis and designing of feedback
control systems. The state-space technique provides a convenient formulation procedure for
the modelling and analysis of systems with non-linearities like the Multiple-Input Multiple-
Output (MIMO) systems such that the mathematical equation representing the system can
be reduced. The conventional method of system modelling represents the Single-Input
Single-Output (SISO) system with the algebraic input-output relationship or transfer
function of the system’s variables, while the modern (state-state) approach uses the
differential equation to represent linear, non-linear, time-invariant and time-varying systems.
State-space technique enables the faster observation of the internal behaviour of system as
well as the determination of the outputs of the system in future when some inputs are given
to the system.
The conventional method of system modelling like the transfer function method suffers some
drawbacks such as:
• The computation of output for some system input and no information about the
systems internal state.
• The transfer function of a system is defined at zero initial condition and not at any
other time after the zero-initial condition.
• It is only applied to SISO Linear Time-invariant Systems.
In contrast, the benefits of state space analysis method are;

1
 • The responses of non-linear and time-varying systems like the real-life systems can
be determined using state-space analysis.
• A unified representation of digital systems with different sampling scheme can be
achieved with state-variable approach.
• The effect of all the initial conditions present is a system are represented in state-
space analysis.
• The state space approach can be used to obtain a unified representation and analysis
of SISO and MIMO systems.
• The simulation block representation for an equation of a system can be obtained
directly using state space mode, hence the mathematical representation of a system
is simplified.
• The computer solutions of both continuous as well as discrete-data system are
obtained easily since the state-space formulation of these system is natural and
convenient for computer simulation.
• The state equations of a system involve matrix algebra, therefore the state variable
equations are highly compatible for computer simulation analysis using
programming languages.

1.2 Concepts of State, State Variables and State Space Model

State: The state of a continuous-time dynamic system at time 𝑡 = 𝑡0 (or 𝑛 = 𝑛0 for a
discrete-time system) is the smallest set of information (called the State Variables) required
such that the additional knowledge of the inputs to the system for time 𝑡 = 𝑡0 are sufficient
to predict the future behaviour (system state) and the outputs of the system for all time 𝑡 ≥
𝑡0 (or 𝑛 ≥ 𝑛0 ). An ideal (linear) representation of a system using a state-determined system
model is shown in figure 1.1.

Input Vector 𝑢 Output Vector 𝑦

𝑢1 ሺ𝑡ሻ 𝑦1 ሺ𝑡ሻ
. System represented by .
. the State Variable .
. .
𝑢2 ሺ𝑡ሻ ሼ𝑥1 ሺ𝑡ሻ, 𝑥2 ሺ𝑡ሻ, … , 𝑥𝑛 ሺ𝑡ሻሽ 𝑦2 ሺ𝑡ሻ
𝑢𝑚 ሺ𝑡ሻ 𝑦𝑘 ሺ𝑡ሻ

Figure 1.1: State variable representation of a linear time-invariant systems

From figure 1.1, the state variables 𝑥1 ሺ𝑡ሻ, 𝑥2 ሺ𝑡ሻ, … , 𝑥𝑛 ሺ𝑡ሻ are the information-bearing
variables (energy storage elements like capacitance and inductance) of a system. The
2
sufficient information contained by the state variables with the past history, input/output
relationship and future inputs to a system can be used to predict all future states of the system.
State Variables: This is a set of variables that explains the present configuration of a system
and determines the future behaviour of the system when the present state of the system, the
excitation signals and the equation describing the dynamics (input/output relationships and
future inputs) are given. The state variables are the information-bearing variables of a system
chosen in a way that correspond to the number of the physically measurable quantities such
as the capacitance inductance, current through a resistance amongst others in an electrical
system. In state space method, a system is represented in an 𝑁-dimensional space in which
every coordinate axes is defined by a corresponding state variable 𝑥1 ሺ𝑡ሻ, 𝑥2 ሺ𝑡ሻ, … , 𝑥𝑛 ሺ𝑡ሻ,
where 𝑛 is the number of the state variable of the system.
State Space: The 𝑁-dimensional space that represents a system is called the state space. Its
𝑛 number of state variables are selected such that the behaviour of the system can be
completely be determined.
State Vector: The state vector contains an 𝑛 state variables ሼ𝑥1 , 𝑥2 , … , 𝑥𝑛 ሽ which
completely determines the behaviour of the dynamic system. The state vector can be written
as an 𝑛 𝑥1 matrix as:
𝑥1
𝑥2
𝑋=[⋮ ] ሺ1.1ሻ
𝑥𝑛
where 𝑥1 , 𝑥2 , … , 𝑥𝑛 are functions of time 𝑡.
State Space Model: A linear time-invariant system’s state space model can be referred to
as state space equation. The state space equation is made up of a set of 𝑛 state equations
expressed in terms of matrices 𝐴 and 𝐵, and a set of output equations which relate output
variable of interest to the state variables and inputs variables. The output equations are
expressed in terms of matrices 𝐶 and 𝐷. The state space model conical form can be written
as:
𝑥̇ ሺ𝑡ሻ = 𝐴𝑥ሺ𝑡ሻ + 𝐵𝑢ሺ𝑡ሻ
ሺ1.2ሻ
𝑦ሺ𝑡ሻ = 𝐶𝑥ሺ𝑡ሻ + 𝐷𝑢ሺ𝑡ሻ

For a linear time-invariant SISO system which can be represented by state space model given
in equation (1.2), the vector block diagram can be drawn as shown in figure 1.2. The vectors
and matrices multiplication are represented with the blocks while the double lines imply a
multiple variable signal path.

𝐷
𝑥̇ ሺ𝑡ሻ 𝑥ሺ𝑡ሻ
𝑢ሺ𝑡ሻ 𝐵 ++ න 𝑑𝑡 𝐶 ++ 𝑦ሺ𝑡ሻ
3

𝐴
 Figure 1.2: Vector block diagram for a LTI continuous time system in state-space system
dynamics.
The Signal Flow Graph (SFG) for the LTI continuous system’s state space model block
diagram in figure 1.2 is shown in figure 1.3.
𝐷

න 𝑑𝑡
𝐵 𝑥̇ ሺ𝑡ሻ 𝑥 𝐶
𝑢ሺ𝑡ሻ 𝑦ሺ𝑡ሻ

Figure 1.3: Signal flow graph representation of the state space model of a LIT
continuous time system.

1.3 LTI Continuous System Representation in State Space Form

Apart from the use of conventional approaches like Laplace transfer and 𝒵-tranforms
methods to analyse system, the modern approach which utilises the state-variable
formulation can be used for system analysis because of the advantages of the state-variables
method in the analysis and design of both continuous and discrete control systems over the
conventional approaches.
A LTI continuous system is represented in state space form by a set of first-order differential
equations known as the state equations. But for a discrete-data system containing both
continuous-data and digital-data components defined in discrete time 𝑘𝑇, a set of first-order
difference equations and differential equations are used to expressed its state equations. A
LTI continuous time system with single nonlinearity and 𝑛 state variables, 𝑚 inputs can be
represented by an 𝑛𝑡ℎ -order linear differential equation with constant coefficients. The
general form of an 𝑛𝑡ℎ -order linear differential equation relating the input 𝑢ሺ𝑡ሻ and output
𝑦ሺ𝑡ሻ of an LTI continuous time system with initial conditions 𝑦ሺ0ሻ, 𝑦̇ ሺ0ሻ, 𝑦̈ ሺ0ሻ, … , 𝑦 ሺ𝑛−1ሻ ሺ0ሻ
is given by:

𝑦 ሺ𝑛ሻ + 𝑎1 𝑦 ሺ𝑛−1ሻ + ⋯ + 𝑎𝑛−1 𝑦̇ + 𝑎𝑛 𝑦 = 𝑏0 𝑢ሺ𝑚ሻ + ⋯ + 𝑏𝑚−1 𝑢̇ + 𝑏𝑚 𝑢 ሺ1.3ሻ

4
where 𝑦 ሺ𝑛ሻ and 𝑢ሺ𝑚ሻ denote the 𝑛𝑡ℎ and 𝑚𝑡ℎ order derivatives of the outputs and inputs
respectively with respect to time, 𝑎𝑖 (for 𝑖 = 1, 2, . . . , 𝑛) and 𝑏𝑗 (for 𝑗 = 0, 1, 2, . . . , 𝑚) are
constants of the output and inputs respectively, 𝑚 and 𝑛 are integers such that 𝑚 ≤ 𝑛.
Equation (1.3) can be easily expended to a set of 𝑛 coupled first-order linear differential
equations with constants for a system with 𝑛 state-variables, 𝑚 inputs and 𝑘 outputs. The
system also has a 𝑘 output equations with coefficients. The process is achieved by choosing
the system output and its 𝑛 − 1 derivatives as the state-variables. This can be written as:
𝑑𝑥1 ሺ𝑡ሻ
= 𝑎11 𝑥1 ሺ𝑡ሻ + 𝑎12 𝑥2 ሺ𝑡ሻ + ⋯ + 𝑎1𝑛 𝑥𝑛 ሺ𝑡ሻ + 𝑏11 𝑢1 ሺ𝑡ሻ + 𝑏12 𝑢2 ሺ𝑡ሻ + 𝑏1𝑚 𝑢𝑚 ሺ𝑡ሻ
𝑑𝑡
𝑑𝑥2 ሺ𝑡ሻ
= 𝑎21 𝑥1 ሺ𝑡ሻ + 𝑎22 𝑥2 ሺ𝑡ሻ + ⋯ + 𝑎2𝑛 𝑥𝑛 ሺ𝑡ሻ + 𝑏21 𝑢1 ሺ𝑡ሻ + 𝑏22 𝑢2 ሺ𝑡ሻ + 𝑏2𝑚 𝑢𝑚 ሺ𝑡ሻ ሺ1.4ሻ
𝑑𝑡
⋮
𝑑𝑥𝑛 ሺ𝑡ሻ
= 𝑎𝑛1 𝑥1 ሺ𝑡ሻ + 𝑎𝑛2 𝑥2 ሺ𝑡ሻ + ⋯ + 𝑎𝑛𝑛 𝑥𝑛 ሺ𝑡ሻ + 𝑏𝑛1 𝑢1 ሺ𝑡ሻ + 𝑏𝑛2 𝑢2 ሺ𝑡ሻ + 𝑏𝑛𝑚 𝑢𝑚 ሺ𝑡ሻ
𝑑𝑡
and
𝑦1 ሺ𝑡ሻ = 𝑐11 𝑥1 ሺ𝑡ሻ + 𝑐12 𝑥2 ሺ𝑡ሻ + ⋯ + 𝑐1𝑛 𝑥𝑛 ሺ𝑡ሻ + 𝑑11 𝑢1 ሺ𝑡ሻ + 𝑑12 𝑢2 ሺ𝑡ሻ + 𝑑1𝑚 𝑢𝑚 ሺ𝑡ሻ
𝑦2 ሺ𝑡ሻ = 𝑐21 𝑥1 ሺ𝑡ሻ + 𝑐22 𝑥2 ሺ𝑡ሻ + ⋯ + 𝑐2𝑛 𝑥𝑛 ሺ𝑡ሻ + 𝑑21 𝑢1 ሺ𝑡ሻ + 𝑑22 𝑢2 ሺ𝑡ሻ + 𝑑2𝑚 𝑢𝑚 ሺ𝑡ሻ
ሺ1.5ሻ
⋮
𝑦𝑘 ሺ𝑡ሻ = 𝑐𝑘1 𝑥1 ሺ𝑡ሻ + 𝑐𝑘2 𝑥2 ሺ𝑡ሻ + ⋯ + 𝑐𝑘𝑛 𝑥𝑛 ሺ𝑡ሻ + 𝑑𝑘1 𝑢1 ሺ𝑡ሻ + 𝑑𝑘2 𝑢2 ሺ𝑡ሻ + 𝑑𝑘𝑚 𝑢𝑚 ሺ𝑡ሻ
where 𝑥𝑖 ሺ𝑡ሻ(for 𝑖 = 1, 2, . . . , 𝑛) are the state variables, 𝑢𝑖 ሺ𝑡ሻ (for 𝑖 = 1, 2, . . . , 𝑚) are the
system inputs, 𝑦𝑖 ሺ𝑡ሻ (for 𝑖 = 1, 2, . . . , 𝑘) are the system outputs and 𝑎𝑛𝑛 , 𝑏𝑛𝑚 , 𝑐𝑘𝑛 , and 𝑑𝑘𝑚
are the system coefficients.
Equation (1.4) (called the state equations) and equation (1.5) (called the output equations)
constituted the state-space model of a system. The compact matrix notation of the state-space
model which can be easily manipulated can be written using the laws of linear algebra as
follows:
𝑥1 𝑎11 𝑎12 𝑎1𝑛 𝑥1 𝑏11 𝑏12 𝑏1𝑚 𝑢1
𝑑 𝑥2 𝑎21 𝑎22 ⋯ 𝑎21 𝑥2 ⋯
𝑏21 𝑏22 𝑏2𝑚 𝑢2
=[ ⋮ ]=[ ⋮ ⋱ ⋮ ][ ⋮ ] + [ ][ ⋮ ] ሺ1.6ሻ
𝑑𝑡 ⋮ ⋱ ⋮
𝑥𝑛 𝑎𝑛1 𝑎𝑛2 ⋯ 𝑎𝑛𝑛 𝑥𝑛 𝑏𝑛1 𝑏𝑛2 ⋯ 𝑏𝑛𝑚 𝑢𝑟
and
𝑦1 𝑐11 𝑐12 𝑐1𝑛 𝑥1 𝑑11 𝑑12 𝑑1𝑚 𝑢1
𝑦2 𝑐21 𝑐22 ⋯ 𝑐21 𝑥2 ⋯
𝑑21 𝑑22 𝑑2𝑚 𝑢2
[⋮]=[ ⋮ ⋱ ⋮ ][ ⋮ ] + [ ][ ⋮ ] ሺ1.7ሻ
⋮ ⋱ ⋮
𝑦𝑘 𝑐𝑘1 𝑐𝑘2 ⋯ 𝑐𝑘𝑛 𝑥𝑛 𝑑𝑘1 𝑑𝑘2 ⋯ 𝑑𝑘𝑚 𝑢𝑟

5
Comparing the compacted form of the compacted matrix notation form of the state space
model given in equation (1.2) and its expanded form given in equation (1.6) and equation
(1.7), 𝑥 is an 𝑛-dimensional vector called the state vector, 𝐴 is a matrix of order 𝑛 × 𝑛 known
as the system matrix, 𝑢 is an 𝑟-dimensional vector called the control vector, 𝐵 is a 𝑛 × 𝑚
matrix known as the control matrix, 𝐶 is a 𝑘 × 𝑚 matrix known as the output matrix, 𝑦 is a
𝑘-dimensional vector known as the output vector and 𝐷 is a 𝑘 × 𝑚 null matrix of constant
coefficients that weight the system inputs. The 𝐷𝑢ሺ𝑡ሻ implies a direct coupling of the input
to output, however, the output power level is normally higher that the input power level in
control system, with exception in communication system where the reverse is the case and
the presence of the 𝐷𝑢ሺ𝑡ሻ is often common. For many physical LTI continuous system, state-
space model equations reduce to:
𝑥̇ ሺ𝑡ሻ = 𝐴𝑥ሺ𝑡ሻ + 𝐵𝑢ሺ𝑡ሻ
ሺ1.8ሻ
𝑦ሺ𝑡ሻ = 𝐶𝑥ሺ𝑡ሻ
The different domains for representing the state-space models of both time-variant and time-
invariant continuous and discrete time systems are given in table 1.1.

Table 1.1: State-space model Representations in Different Domain

System Type State-space Model
Continuous Time-invariant 𝑥̇ ሺ𝑡ሻ = 𝐴𝑥ሺ𝑡ሻ + 𝐵𝑢ሺ𝑡ሻ
𝑦ሺ𝑡ሻ = 𝐶𝑥ሺ𝑡ሻ + 𝐷𝑢ሺ𝑡ሻ

Continuous Time-variant 𝑥̇ ሺ𝑡ሻ = 𝐴ሺ𝑡ሻ𝑥ሺ𝑡ሻ + 𝐵ሺ𝑡ሻ𝑢ሺ𝑡ሻ

𝑦ሺ𝑡ሻ = 𝐶ሺ𝑡ሻ𝑥ሺ𝑡ሻ + 𝐷ሺ𝑡ሻ𝑢ሺ𝑡ሻ

Discrete Time-invariant 𝑥ሺ𝑘 + 1ሻ = 𝐴𝑥ሺ𝑘ሻ + 𝐵𝑢ሺ𝑘ሻ

𝑦ሺ𝑘ሻ = 𝐶𝑥ሺ𝑘ሻ + 𝐷𝑢ሺ𝑘ሻ
Discrete Time-variant 𝑥ሺ𝑘 + 1ሻ = 𝐴ሺ𝑘ሻ𝑥ሺ𝑘ሻ + 𝐵ሺ𝑘ሻ𝑢ሺ𝑘ሻ
𝑦ሺ𝑘ሻ = 𝐶ሺ𝑘ሻ𝑥ሺ𝑘ሻ + 𝐷ሺ𝑘ሻ𝑢ሺ𝑘ሻ

Laplace Domain of Continuous 𝑠𝑋ሺ𝑠ሻ = 𝐴𝑋ሺ𝑠ሻ + 𝐵𝑈ሺ𝑠ሻ

Time-Invariant 𝑌ሺ𝑠ሻ = 𝐶𝑋ሺ𝑠ሻ + 𝐷𝑈ሺ𝑠ሻ

𝒵-domain of Discrete Time-Invariant 𝑧𝑋ሺ𝑧ሻ = 𝐴𝑋ሺ𝑧ሻ + 𝐵𝑈ሺ𝑧ሻ

𝑌ሺ𝑧ሻ = 𝐶𝑋ሺ𝑧ሻ + 𝐷𝑈ሺ𝑧ሻ

1.4 State-Space Model Transfer Function

Given the LIT continuous time system state-space model and its Laplace Transform
domain representation in table 1.1, the input and output relationship of the Laplace
transform domain representation can be obtained as follows;

6
 𝑠𝑋ሺ𝑠ሻ − 𝑥ሺ0ሻ = 𝐴𝑋ሺ𝑠ሻ + 𝐵𝑈ሺ𝑠ሻ ሺ1.9ሻ
𝑌ሺ𝑠ሻ = 𝐶𝑋ሺ𝑠ሻ + 𝐷𝑈ሺ𝑠ሻ ሺ1.10ሻ
Where 𝑥ሺ0ሻ = 0 in equation (1.9) because the transfer function is defined with zero initial
condition. Rewriting equation (1.9) with the introduction of an identity matrix 𝐼, we have
[𝑠𝐼 − 𝐴]𝑋ሺ𝑠ሻ = 𝐵𝑈ሺ𝑠ሻ
Therefore,
𝑋ሺ𝑠ሻ = [𝑠𝐼 − 𝐴]−1 𝐵𝑈ሺ𝑠ሻ (1.11)
The 𝑠𝐼 term creates a square or an 𝑛 × 𝑛 matrix with 𝑠 on the leading diagonal and zeros on
other positions. The 𝑠𝐼 term gives matrix of the same dimension which easily permits
addition and subtraction.
Substituting equation (1.11) in equation (1.10), we have
𝑌ሺ𝑠ሻ = 𝐶[𝑠𝐼 − 𝐴]−1 𝐵𝑈ሺ𝑠ሻ + 𝐷𝑈ሺ𝑠ሻ
𝐶[𝑎𝑑𝑗ሺ𝑠𝐼−𝐴ሻ]
Recall that, 𝐶[𝑠𝐼 − 𝐴]−1 = |ሺ𝑠𝐼−𝐴ሻ|
It is evident that |ሺ𝑠𝐼 − 𝐴ሻ| is the characteristics polynomial of the transfer function 𝐺ሺ𝑠ሻ,
thus the poles of transfer function 𝐺ሺ𝑠ሻ are the eigen-values of matrix 𝐴.
𝑌ሺ𝑠ሻ
Therefore, the transfer function 𝐺ሺ𝑠ሻ = 𝑈ሺ𝑠ሻ is given as

𝑌ሺ𝑠ሻ
= 𝐶[𝑠𝐼 − 𝐴]−1 𝐵 + 𝐷 ሺ1.12ሻ
𝑈ሺ𝑠ሻ
Since the 𝐷 is a null matrix as explained in section 1.3, the transfer function becomes
𝑌ሺ𝑠ሻ
= 𝐶[𝑠𝐼 − 𝐴]−1 𝐵 ሺ1.13ሻ
𝑈ሺ𝑠ሻ
When the differential equations of transfer function of a control system are given, the state-
space model of the system can be obtained through three different ways namely;
(i) The Physical Variable Form.
(ii) The Phase Variable Form.
(iii) The Canonical Form.

1.5 State-Space Modeling using Physical Variable Form

In the method, the physical variables of the system are used and considered as state
variables when developing the state-space equations. The state variables differ from system
to system as described in table 1.2.
Table 1.2: State Variables for Electrical and Mechanical Systems
System State Variables

7
 Electrical System Voltage across the Resistance
Voltage across the Capacitor
Inductance Current
Capacitance Current
Current flowing through the Resistance
Current flowing through the Inductance
Current flowing through the Capacitance
Mechanical Translational System Linear Displacement
Liner Velocity
Linear Acceleration
Mechanical Rotational System Angular Displacement
Angular Velocity
Angular Acceleration

1.5.1 State-Space Modeling of Electrical System using Physical Variables Method

Let us look at the procedure for modeling the state space equation of electrical systems by
working on a simple RLC electrical networks shown in the following examples.
Example 1.1: Develop the state-space model for the RLC network shown in figure E1.1

Figure E1.1: A RLC Electrical System

Solution: Since the state variables are not given in the question, lets choose the current
𝑖ሺ𝑡ሻ through the inductor and the voltage 𝑣𝑐 ሺ𝑡ሻ across the capacitor as the state variables
𝑥1 ሺ𝑡ሻ and 𝑥2 ሺ𝑡ሻ respectively. The input to the system 𝑒ሺ𝑡ሻ is taken as 𝑢ሺ𝑡ሻ.

Applying the Kirchhoff’s Voltage Law (KVL) to the RLC network gives;
𝑑𝑖ሺ𝑡ሻ
𝑒ሺ𝑡ሻ = 𝐿 + 𝑅𝑖ሺ𝑡ሻ + 𝑣𝑐 ሺ𝑡ሻ ሺ1ሻ
𝑑𝑡
Making the derivative the subject of the equation and divide through with the inductance 𝐿,
we have;
𝑑𝑖ሺ𝑡ሻ 1 𝑅 1
= 𝑒ሺ𝑡ሻ − 𝑖ሺ𝑡ሻ − 𝑣𝑐 ሺ𝑡ሻ ሺ2ሻ
𝑑𝑡 𝐿 𝐿 𝐿

8
Also, we write the equation that relates the voltage 𝑣𝑐 ሺ𝑡ሻ across the capacitance and the
current 𝑖ሺ𝑡ሻ through the inductance. This is chosen because the same current 𝑖ሺ𝑡ሻ flow
through the capacitance.
1
𝑣𝑐 ሺ𝑡ሻ = න 𝑖ሺ𝑡ሻ 𝑑𝑡 ሺ3ሻ
𝐶
Differentiating equation (3) gives,
𝑑𝑣𝑐 ሺ𝑡ሻ 1
= 𝑖ሺ𝑡ሻ ሺ4ሻ
𝑑𝑡 𝐶
Substituting the chosen state variables 𝑖ሺ𝑡ሻ = 𝑥1 ሺ𝑡ሻ and 𝑣𝑐 ሺ𝑡ሻ = 𝑥2 ሺ𝑡ሻ together with the
input to the system 𝑒ሺ𝑡ሻ = 𝑢ሺ𝑡ሻ in equation (2) and equation (4) gives;
𝑑𝑥1 ሺ𝑡ሻ 𝑅 1 1
= − 𝑥1 ሺ𝑡ሻ − 𝑥2 ሺ𝑡ሻ + 𝑢ሺ𝑡ሻ ሺ5ሻ
𝑑𝑡 𝐿 𝐿 𝐿
𝑑𝑥2 ሺ𝑡ሻ 1
= 𝑥1 ሺ𝑡ሻ ሺ6ሻ
𝑑𝑡 𝐶
The output 𝑦ሺ𝑡ሻ of the RLC network is the voltage 𝑣𝑐 ሺ𝑡ሻ across the capacitance and that
capacitance voltage is one of the state variables, i.e. 𝑣𝑐 ሺ𝑡ሻ = 𝑥2 ሺ𝑡ሻ, thus;

𝑦ሺ𝑡ሻ = 𝑥2 ሺ𝑡ሻ ሺ7ሻ

Therefore, equation (5) and equation (6) form the compact state equation matrix, while
equation (7) forms the output equation matrix. Hence, the state-space model of the RLC
network is written as;
𝑅 1 1
𝑥̇ 1 ሺ𝑡ሻ − − 𝑥1 ሺ𝑡ሻ
[ ]=[ 𝐿 𝐿 ][ ] + [𝐿] 𝑢ሺ𝑡ሻ ሺ8ሻ
1
𝑥̇ 2 ሺ𝑡ሻ 0 𝑥2 ሺ𝑡ሻ 0
𝐶
𝑥1 ሺ𝑡ሻ
𝑦ሺ𝑡ሻ = [0 1] [ ] ሺ9ሻ
𝑥2 ሺ𝑡ሻ
Example 1.2: Develop the state-space model for the electrical circuit shown in figure E1.2
given that the state variables are voltage 𝑣1 ሺ𝑡ሻ across capacitance 𝐶1 , current 𝑖ሺ𝑡ሻ through
the inductance 𝐿1 and voltage 𝑣2 ሺ𝑡ሻ across capacitance 𝐶2 .

9
 Figure E1.2(a): RLC Electrical Network
Solution: The branch currents and voltages are shown in the redrawn circuit of figure
E1.2(b).

Loop 1 Loop 2 Loop 3

Figure E1.2(b)
The electrical input 𝑢ሺ𝑡ሻ to the circuit is 𝑣𝑖 ሺ𝑡ሻ and the node voltages are 𝑣1 ሺ𝑡ሻ and 𝑣2 ሺ𝑡ሻ.
Applying KVL to loops 1, 2 and 3 in circuit to get equations (1), (2) and (3).

𝑣𝑖 ሺ𝑡ሻ = 𝑖1 ሺ𝑡ሻ𝑅1 + 𝑣1 ሺ𝑡ሻ ሺ𝑖ሻ

𝑑𝑖ሺ𝑡ሻ
𝑣1 ሺ𝑡ሻ = 𝐿1 + 𝑣2 ሺ𝑡ሻ ሺ2ሻ
𝑑𝑡
𝑣2 ሺ𝑡ሻ = [𝑖ሺ𝑡ሻ − 𝑖2 ሺ𝑡ሻ]𝑅2 ሺ3ሻ
𝑑𝑥1 ሺ𝑡ሻ
To solve for the first-order derivative of the first state variable ( ), we applying
𝑑𝑡
Kirchhoff’s Current Law (KCL) at voltage 𝑣1 ሺ𝑡ሻ node gives,
1
𝑣1 ሺ𝑡ሻ = න[𝑖1 ሺ𝑡ሻ − 𝑖ሺ𝑡ሻ]𝑑𝑡 ሺ4ሻ
𝐶1
Differentiating equation (4) with respect to time 𝑡 gives;
𝑑𝑣1 ሺ𝑡ሻ 1
= [𝑖1 ሺ𝑡ሻ − 𝑖ሺ𝑡ሻ] ሺ5ሻ
𝑑𝑡 𝐶1
Eliminating non-state variable 𝑖1 ሺ𝑡ሻ term, by making 𝑖1 ሺ𝑡ሻ the subject in equation (1) and
substitute in equation (5) to get
𝑑𝑣1 ሺ𝑡ሻ 1 𝑣𝑖 ሺ𝑡ሻ − 𝑣1 ሺ𝑡ሻ 1
= [ ] − 𝑖ሺ𝑡ሻ ሺ6ሻ
𝑑𝑡 𝐶1 𝑅1 𝐶1
Simplifying and rearranging equation (6) gives
𝑑𝑣1 ሺ𝑡ሻ 1 𝑣1 ሺ𝑡ሻ 1 1 𝑣𝑖 ሺ𝑡ሻ
=− − 𝑖ሺ𝑡ሻ + ሺ7ሻ
𝑑𝑡 𝐶1 𝑅1 𝐶1 𝐶1 𝑅1
10
Substituting the state variable terms as 𝑥1 = 𝑣1 , 𝑥2 = 𝑖ሺ𝑡ሻ, 𝑥3 = 𝑣2 , and 𝑢ሺ𝑡ሻ = 𝑣𝑖 ሺ𝑡ሻ gives
𝑑𝑥1 ሺ𝑡ሻ 1 1 1
=− 𝑥1 ሺ𝑡ሻ − 𝑥2 ሺ𝑡ሻ + 𝑢ሺ𝑡ሻ ሺ8ሻ
𝑑𝑡 𝑅1 𝐶1 𝐶1 𝑅1 𝐶1
𝑑𝑥2 ሺ𝑡ሻ
To solve for the first-order derivative of the second state variable ( ), we rearrange
𝑑𝑡
equation (2) to get
𝑑𝑖ሺ𝑡ሻ 1 1
= 𝑣1 ሺ𝑡ሻ − 𝑣2 ሺ𝑡ሻ
𝑑𝑡 𝐿1 𝐿1
Substituting the state variable terms as 𝑥1 = 𝑣1 , 𝑥2 = 𝑖ሺ𝑡ሻ, 𝑥3 = 𝑣2 , and 𝑢ሺ𝑡ሻ = 𝑣𝑖 ሺ𝑡ሻ gives
𝑑𝑥2 ሺ𝑡ሻ 1 1
= 𝑥1 ሺ𝑡ሻ − 𝑥3 ሺ𝑡ሻ ሺ9ሻ
𝑑𝑡 𝐿1 𝐿1
𝑑𝑥3 ሺ𝑡ሻ
To solve for the first-order derivative of the third state variable ( ), we write the
𝑑𝑡
expression for the current 𝑖2 ሺ𝑡ሻ through the capacitance 𝐶2 and the voltage 𝑣2 across the
capacitance 𝐶2 as
𝑑𝑣2 ሺ𝑡ሻ
𝑖2 ሺ𝑡ሻ = 𝐶2
𝑑𝑡
𝑑𝑣2 ሺ𝑡ሻ 1
= 𝑖2 ሺ𝑡ሻ ሺ10ሻ
𝑑𝑡 𝐶2
Eliminate non-state variable term 𝑖2 ሺ𝑡ሻ by making 𝑖2 ሺ𝑡ሻ the subject in equation (3)
𝑣2 ሺ𝑡ሻ
𝑖2 ሺ𝑡ሻ = 𝑖ሺ𝑡ሻ −
𝑅2
Substituting the above equation (3) in equation (10) gives
𝑑𝑣2 ሺ𝑡ሻ 1 𝑣2 ሺ𝑡ሻ
= [𝑖ሺ𝑡ሻ − ] ሺ11ሻ
𝑑𝑡 𝐶2 𝑅2
Substituting the state variable terms as 𝑥1 = 𝑣1 , 𝑥2 = 𝑖ሺ𝑡ሻ, 𝑥3 = 𝑣2 , and 𝑢ሺ𝑡ሻ = 𝑣𝑖 ሺ𝑡ሻ in
equation (10) gives
𝑑𝑥3 ሺ𝑡ሻ 1 1
= 𝑥2 ሺ𝑡ሻ − 𝑥 ሺ𝑡ሻ ሺ12ሻ
𝑑𝑡 𝐶2 𝑅2 𝐶2 3
Expressing equations (8), (9), and (12) in matrix form give the state space model equation
of the electrical network, thus
1 1
𝑥̇ 1 ሺ𝑡ሻ − − 0 𝑥1 ሺ𝑡ሻ 1
𝑅1 𝐶1 𝐶1
1 1 𝑅1 𝐶1
𝑥̇ 2 ሺ𝑡ሻ = 0 − 𝑥2 ሺ𝑡ሻ + 0 𝑢ሺ𝑡ሻ
𝐿1 𝐿1
[𝑥̇ 3 ሺ𝑡ሻ] 1 1 [𝑥 ሺ𝑡ሻ]
0 − 3 [ 0 ]
[ 𝐶2 𝑅2 𝐶2 ]
11
1.5.2 State-Space Modeling of Mechanical System using Physical Variables Method
Let us look at the procedure for modeling the state space equation of mechanical systems by
working on some translational and rotational mechanical systems shown in this section.
The following steps should be followed to develop the state space model equation.

(i) When a translational/rotational mechanical system is given, there is always a 𝑖

number of state variables and 𝑘 number of output variables present in the system.

(ii) If an 𝑛 number of state variables and 𝑚 number of output variable are given, then, 𝑛
state variables and 𝑚 output variables are assumed, otherwise the total state variables
𝑖 and output variables 𝑘 present in the system are assumed.
(iii) Obtain the differential equation of the system using D’Alembert’s principle.
(iv) Develop the first-order derivatives of the assumed state variables from the system’s
differential equation obtained in the previous step such that the number of the first-
order differential equations equals the order of the differential equation of the system.
(v) Write the state space model matrices of the mechanical system form the arrangement
of the first-order differential equations obtained in the previous step.
Example 1.3: Develop the state space model of the translational mechanical system shown
in figure E1.3.

Figure E1.3: A Spring-Mass-Damper Mechanical System

Solution: The output of the translational mechanical system shown in figure E1.3 is
displacements 𝑦1 ሺ𝑡ሻ and 𝑦2 ሺ𝑡ሻ, the input 𝑓ሺ𝑡ሻ is 𝑢ሺ𝑡ሻ and the velocities are 𝑣1 ሺ𝑡ሻ, 𝑣2 ሺ𝑡ሻ.

In the absence of the state variables in the question, we selected the state variables for the
system as

𝑥1 ሺ𝑡ሻ = 𝑦1 ሺ𝑡ሻ
12
𝑥2 ሺ𝑡ሻ = 𝑦2 ሺ𝑡ሻ
𝑑𝑦1 ሺ𝑡ሻ
𝑥3 ሺ𝑡ሻ = 𝑣1 ሺ𝑡ሻ =
𝑑𝑡
𝑑𝑦2 ሺ𝑡ሻ
𝑥4 ሺ𝑡ሻ = 𝑣2 ሺ𝑡ሻ =
𝑑𝑡
Using D’Alembert’s principle to masses 𝑀1 and 𝑀2 yield equation (1) and (2) respectively

𝑑 2 𝑦1 ሺ𝑡ሻ 𝑑(𝑦1 ሺ𝑡ሻ − 𝑦2 ሺ𝑡ሻ)

𝑓ሺ𝑡ሻ = 𝑀1 2
+ 𝐵2 + 𝐾1 𝑦1 ሺ𝑡ሻ ሺ1ሻ
𝑑𝑡 𝑑𝑡
𝑑 2 𝑦2 ሺ𝑡ሻ 𝑑(𝑦2 ሺ𝑡ሻ − 𝑦1 ሺ𝑡ሻ)
0 = 𝑀2 2
+ 𝐵2 + 𝐾2 𝑦2 ሺ𝑡ሻ ሺ2ሻ
𝑑𝑡 𝑑𝑡
Rearrange equations (1) and (2) by making the second derivatives the subject

𝑑 2 𝑦1 ሺ𝑡ሻ 𝐾1 𝐵2 𝑑𝑦1 ሺ𝑡ሻ 𝐵2 𝑑𝑦2 ሺ𝑡ሻ 1

2
= − 𝑦1 ሺ𝑡ሻ − + + 𝑓ሺ𝑡ሻ ሺ3ሻ
𝑑𝑡 𝑀1 𝑀1 𝑑𝑡 𝑀1 𝑑𝑡 𝑀1

𝑑2 𝑦2 ሺ𝑡ሻ 𝐾2 𝐵2 𝑑𝑦1 ሺ𝑡ሻ 𝐵2 𝑑𝑦2 ሺ𝑡ሻ

2
=− 𝑦2 ሺ𝑡ሻ + − ሺ4ሻ
𝑑𝑡 𝑀2 𝑀2 𝑑𝑡 𝑀2 𝑑𝑡
Substituting the selected state variables in equations (3) and (4)
𝐾1 𝐵2 𝐵2 1
𝑥̇ 3 ሺ𝑡ሻ = − 𝑥1 ሺ𝑡ሻ − 𝑥3 ሺ𝑡ሻ + 𝑥4 ሺ𝑡ሻ + 𝑢ሺ𝑡ሻ ሺ5ሻ
𝑀1 𝑀1 𝑀1 𝑀1
𝐾2 𝐵2 𝐵2
𝑥̇ 4 ሺ𝑡ሻ = − 𝑥2 ሺ𝑡ሻ + 𝑥3 ሺ𝑡ሻ − 𝑥 ሺ𝑡ሻ ሺ6ሻ
𝑀2 𝑀2 𝑀2 4
Representing equations (5) and (6) in matrix form give the state equation, while using the
displacements of the masses that were chosen as one of the state variables 𝑥1 ሺ𝑡ሻ = 𝑦1 ሺ𝑡ሻ and
𝑥2 ሺ𝑡ሻ = 𝑦2 ሺ𝑡ሻ give the output equation of the system model.

0 0 1 0
𝑥̇ 2 ሺ𝑡ሻ 0 0 0 1 𝑥2 ሺ𝑡ሻ 0
𝑥̇ 2 ሺ𝑡ሻ 𝐾1 𝐵2 𝐵2 𝑥 ሺ𝑡ሻ 0
= −𝑀 0
𝑀1
− 2
𝑀1 𝑥3 ሺ𝑡ሻ + 1 𝑢ሺ𝑡ሻ
𝑥̇ 3 ሺ𝑡ሻ 1 ⁄𝑀
1
[𝑥̇ 4 ሺ𝑡ሻ] 𝐾2 𝐵2 𝐵2 [𝑥4 ሺ𝑡ሻ] [ 0 ]
0 − −
[ 𝑀2 𝑀2 𝑀2 ]

13
 𝑥2 ሺ𝑡ሻ
𝑦1 ሺ𝑡ሻ 1 0 0 0 𝑥2 ሺ𝑡ሻ
[ ]=[ ] +0
𝑦2 ሺ𝑡ሻ 0 1 0 0 𝑥3 ሺ𝑡ሻ
[𝑥4 ሺ𝑡ሻ]
Example 1.4: Develop the state equation and output equation of the rotational mechanical
system shown in figure E1.4.

Figure E1.4: Rotational Mechanical System.

Solution: Let’s choose the angular displacements 𝜃1 ሺ𝑡ሻ, 𝜃2 ሺ𝑡ሻ, 𝜃3 ሺ𝑡ሻ and angular
velocities 𝜔1 ሺ𝑡ሻ, 𝜔2 ሺ𝑡ሻ, 𝜔3 ሺ𝑡ሻ as the state variables
𝑑𝜃1 ሺ𝑡ሻ 𝑑 2 𝜃1 ሺ𝑡ሻ
𝑥1 ሺ𝑡ሻ = 𝜃1 ሺ𝑡ሻ 𝑥4 ሺ𝑡ሻ = 𝜔1 ሺ𝑡ሻ = 𝑥̇ 4 ሺ𝑡ሻ =
𝑑𝑡 𝑑𝑡 2
𝑑𝜃2 ሺ𝑡ሻ 𝑑 2 𝜃2 ሺ𝑡ሻ
𝑥2 ሺ𝑡ሻ = 𝜃2 ሺ𝑡ሻ 𝑥5 ሺ𝑡ሻ = 𝜔2 ሺ𝑡ሻ = 𝑥̇ 5 ሺ𝑡ሻ =
𝑑𝑡 𝑑𝑡 2
𝑑𝜃3 ሺ𝑡ሻ 𝑑 2 𝜃3 ሺ𝑡ሻ
𝑥3 ሺ𝑡ሻ = 𝜃3 ሺ𝑡ሻ 𝑥6 ሺ𝑡ሻ = 𝜔3 ሺ𝑡ሻ = 𝑥̇ 6 ሺ𝑡ሻ =
𝑑𝑡 𝑑𝑡 2
The input variable is 𝑇ሺ𝑡ሻ = 𝑢ሺ𝑡ሻ
Applying D’Alembert’s principle to the system gives

𝑑 2 𝜃1 ሺ𝑡ሻ 𝑑𝜃1 ሺ𝑡ሻ

𝑇ሺ𝑡ሻ = 𝐽1 + 𝐵1 + 𝐾1 [𝜃1 ሺ𝑡ሻ − 𝜃2 ሺ𝑡ሻ] ሺ1ሻ
𝑑𝑡 2 𝑑𝑡
𝑑 2 𝜃2 ሺ𝑡ሻ 𝑑𝜃2 ሺ𝑡ሻ
0 = 𝐽2 + 𝐵3 + 𝐾2 [𝜃2 ሺ𝑡ሻ − 𝜃3 ሺ𝑡ሻ] + 𝐾1 [𝜃2 ሺ𝑡ሻ − 𝜃1 ሺ𝑡ሻ] ሺ2ሻ
𝑑𝑡 2 𝑑𝑡
𝑑 2 𝜃3 ሺ𝑡ሻ 𝑑𝜃3 ሺ𝑡ሻ
0 = 𝐽3 2
+ 𝐵3 + 𝐾2 [𝜃3 ሺ𝑡ሻ − 𝜃2 ሺ𝑡ሻ] + 𝐾3 𝜃3 ሺ𝑡ሻ ሺ3ሻ
𝑑𝑡 𝑑𝑡
Substituting the selected state variables in equations (1), (2), (3) and rearranging the resultant
equation yield
𝐾1 𝐾1 𝐵1 1
𝑥̇ 4 ሺ𝑡ሻ = − 𝑥1 ሺ𝑡ሻ + 𝑥2 ሺ𝑡ሻ − 𝑥4 ሺ𝑡ሻ + 𝑢ሺ𝑡ሻ ሺ4ሻ
𝐽1 𝐽1 𝐽1 𝐽1

14
 𝐾1 𝐾1 + 𝐾2 𝐾2 𝐵2
𝑥̇ 5 ሺ𝑡ሻ = 𝑥1 ሺ𝑡ሻ − ( ) 𝑥2 ሺ𝑡ሻ + 𝑥3 ሺ𝑡ሻ − 𝑥 ሺ𝑡ሻ ሺ5ሻ
𝐽2 𝐽2 𝐽2 𝐽2 5
𝐾2 𝐾2 + 𝐾3 𝐵3
𝑥̇ 6 ሺ𝑡ሻ = 𝑥2 ሺ𝑡ሻ − ( ) 𝑥3 ሺ𝑡ሻ − 𝑥 ሺ𝑡ሻ ሺ6ሻ
𝐽2 𝐽3 𝐽3 6
From equations (4), (5), and (6), the state equation of the system is obtained, while the output
equation is obtained from the angular displacements 𝜃1 ሺ𝑡ሻ, 𝜃2 ሺ𝑡ሻ, 𝜃3 ሺ𝑡ሻ which are the
outputs of the system. Thus

0 0 0 1 0 0
0
𝑥̇ 1 ሺ𝑡ሻ 0 0 0 0 1 0 𝑥1 ሺ𝑡ሻ
0
𝑥̇ 2 ሺ𝑡ሻ 0 0 0 0 0 1 𝑥2 ሺ𝑡ሻ
−𝐾1 𝐾1 0
𝑥̇ 3 ሺ𝑡ሻ 0 −𝐵1 𝑥3 ሺ𝑡ሻ
= 𝐽1 𝐽1 0 0 + 1 𝑢ሺ𝑡ሻ
𝑥̇ 4 ሺ𝑡ሻ 𝐽1 𝑥4 ሺ𝑡ሻ 𝐽1
𝐾1 𝐾1 + 𝐾2 𝐾2 𝐵2
𝑥̇ 5 ሺ𝑡ሻ −( ) 0 0 − 𝑥5 ሺ𝑡ሻ 0
𝐽2 𝐽2 𝐽2 𝐽2
[𝑥̇ 6 ሺ𝑡ሻ] 𝐾2 𝐾2 + 𝐾3 0 𝐵3 [𝑥6 ሺ𝑡ሻ]
0 −( ) 0 − [0]
[ 𝐽2 𝐽3 𝐽3 ]

𝑥1 ሺ𝑡ሻ
𝑥2 ሺ𝑡ሻ
𝜃1 ሺ𝑡ሻ 1 0 0 0 0 0 𝑥 ሺ𝑡ሻ
3
[𝜃2 ሺ𝑡ሻ] = [0 1 0 0 0 0]
𝑥4 ሺ𝑡ሻ
𝜃3 ሺ𝑡ሻ 0 0 1 0 0 0
𝑥5 ሺ𝑡ሻ
[𝑥6 ሺ𝑡ሻ]
1.5.3 State-Space Modeling of Electromechanical System using Physical Variables
Method
In this section, we will model the state-space model of a Direct Current (DC) machine given
in example E1.5. The DC motor is an electromechanical system that combines both electrical
and mechanical units.
Example E1.5: Develop the state equation and output equation of the Armature-controlled
DC motor shown in figure E1.5(a).

15
 Figure E1.5(a): Armature-controlled DC motor
Solution: The armature-controlled DC motor shown in figure E1.5(a) has the armature
conductor winding and field conduct winding that control the motor. The speed of the DC
motor is directly proportional to the armature voltage 𝑉𝑎 ሺ𝑡ሻ and inversely proportional to the
flux.
The mechanical part of the system has the rotor and the road connected to the shaft while the
armature winding and the field winding forms the electrical system. In this system, he
armature voltage is varied to reach the anticipated speed while the field voltage is kept
constant. The DC motor has the following parameters;
𝑅𝑎 is the armature resistance in ohms Ω
𝐿𝑎 is the armature inductance in henry ሺ𝐻ሻ
𝑖𝑎 is the armature current in ampere ሺ𝐴ሻ
𝑣𝑎 is the armature voltage in volts ሺ𝑉ሻ
𝑒𝑏 is the back emf in volts ሺ𝑉ሻ
𝑇 is the torque developed by motor, in newton-metre in 𝑁-𝑚
𝜃 is the angular displacement of shaft in radian ሺ𝑟𝑎𝑑ሻ
𝜔 is the angular velocity of the shaft in 𝑟𝑎𝑑/𝑠𝑒𝑐
𝐽 is the moment of initial of motor and load in 𝑘𝑚-𝑚2 /𝑟𝑎𝑑
𝐵 is the frictional coefficient of motor and load in 𝑁-𝑚/ሺ𝑟𝑎𝑑/𝑠𝑒𝑐ሻ
The equivalent circuit of the electrical section (armature) and the mechanical sections of the
DC motor are given in figure E1.5 (b) and (c) respectively.

(b) (c)
Figure E1.5: Armature DC motor. (b) equivalent electrical circuit, (c) mechanical
circuit.
Let us choose the state variables to be
16
𝑥1 ሺ𝑡ሻ = 𝑖𝑎 ሺ𝑡ሻ
𝑑𝜃ሺ𝑡ሻ
𝑥2 ሺ𝑡ሻ = 𝜔ሺ𝑡ሻ =
𝑑𝑡
𝑥3 ሺ𝑡ሻ = 𝜃ሺ𝑡ሻ
The input variable is the armature voltage 𝑣𝑎 ሺ𝑡ሻ = 𝑢ሺ𝑡ሻ
Applying KVL to the equivalent circuit in figure 1.5 (b) gives
𝑑𝑖𝑎 ሺ𝑡ሻ
𝑣𝑎 ሺ𝑡ሻ = 𝑅𝑎 𝑖𝑎 ሺ𝑡ሻ + 𝐿𝑎 + 𝑒𝑏 ሺ𝑡ሻ ሺ1ሻ
𝑑𝑡
In the mechanical section of the DC motor shown in figure E1.5(c), the differential
equation representing the rotational action is given by
𝑑 2 𝜃ሺ𝑡ሻ 𝑑𝜃ሺ𝑡ሻ
𝐽 + 𝐵 = 𝑇ሺ𝑡ሻ ሺ2ሻ
𝑑𝑡 2 𝑑𝑡
Since the torque output of the DC motor is proportional to armature current as the flux
produced by the field voltage is constant, we can write

𝑇ሺ𝑡ሻ ∝ 𝑖𝑎 ሺ𝑡ሻ
Therefore,
𝑇ሺ𝑡ሻ = 𝐾𝑡 𝑖𝑎 ሺ𝑡ሻ ሺ3ሻ
where 𝐾𝑡 is the torque constant in 𝑁-𝑚/𝐴

Substituting equation (3) in equation (2) to get

𝑑2 𝜃ሺ𝑡ሻ 𝑑𝜃ሺ𝑡ሻ
𝐽 2
+𝐵 = 𝐾𝑡 𝑖𝑎 ሺ𝑡ሻ ሺ4ሻ
𝑑𝑡 𝑑𝑡
The back emf developed in the DC motor is proportional to the speed or the angular
velocity of the shaft, i.e.
𝑑𝜃ሺ𝑡ሻ
𝑒𝑏 ሺ𝑡ሻ ∝
𝑑𝑡
hence, the back emf,
𝑑𝜃ሺ𝑡ሻ
𝑒𝑏 ሺ𝑡ሻ = 𝐾𝑏 ሺ5ሻ
𝑑𝑡
where 𝐾𝑏 is the back emf constant in 𝑉/ሺ𝑟𝑎𝑑/𝑠𝑒𝑐ሻ

substituting equation (5) in equation (1) gives

𝑑𝑖𝑎 ሺ𝑡ሻ 𝑑𝜃ሺ𝑡ሻ
𝑣𝑎 ሺ𝑡ሻ = 𝑅𝑎 𝑖𝑎 ሺ𝑡ሻ + 𝐿𝑎 + 𝐾𝑏 ሺ6ሻ
𝑑𝑡 𝑑𝑡
Equations (4) and (6) are the differential equations governing the armature control DC motor.
Therefore, substituting the chosen state variables in equations (4) and simplifying gives
𝑑𝑥1 ሺ𝑡ሻ
𝑢ሺ𝑡ሻ = 𝑅𝑎 𝑥1 ሺ𝑡ሻ + 𝐿𝑎 + 𝐾𝑏 𝑥2 ሺ𝑡ሻ
𝑑𝑡
17
 𝑅𝑎 𝐾𝑏 1
𝑥̇ 1 ሺ𝑡ሻ = − 𝑥1 ሺ𝑡ሻ − 𝑥2 ሺ𝑡ሻ + 𝑢ሺ𝑡ሻ ሺ7ሻ
𝐿𝑎 𝐿𝑎 𝐿𝑎
and substituting the state variables in equation (6) and simplifying gives
𝑑2 𝑥2 ሺ𝑡ሻ 𝑑𝑥2 ሺ𝑡ሻ
𝐽 2
+𝐵 = 𝐾𝑡 𝑥1 ሺ𝑡ሻ
𝑑𝑡 𝑑𝑡
𝐾𝑡 𝐵
𝑥̇ 2 ሺ𝑡ሻ = 𝑥1 ሺ𝑡ሻ − 𝑥2 ሺ𝑡ሻ ሺ8ሻ
𝐽 𝐽
and 𝑥̇ 3 ሺ𝑡ሻ = 𝑥2 ሺ𝑡ሻ ሺ9ሻ
The output variables are chosen as follow
𝑦1 ሺ𝑡ሻ = 𝑖𝑎 ሺ𝑡ሻ
𝑑𝜃ሺ𝑡ሻ
𝑦2 ሺ𝑡ሻ = 𝜔ሺ𝑡ሻ =
𝑑𝑡
𝑦3 ሺ𝑡ሻ = 𝜃ሺ𝑡ሻ

Mapping the output variables to the state variables give

𝑦1 ሺ𝑡ሻ = 𝑥1 ሺ𝑡ሻ ሺ10ሻ

𝑦2 ሺ𝑡ሻ = 𝑥2 ሺ𝑡ሻ ሺ11ሻ
𝑦3 ሺ𝑡ሻ = 𝑥3 ሺ𝑡ሻ ሺ12ሻ

Finally, the state space model of the armature-controlled DC motor is written by expressing
equations (7), (8) and (9) in matrix form yield the state equation and equations (10), (11) and
(12) yield the output equation as follow
𝑥̇ 1 ሺ𝑡ሻ 𝑅𝑎 𝐾𝑏 𝑥 ሺ𝑡ሻ 1
− − 0 1
𝐿𝑎 𝐿𝑎 𝐿𝑎
𝑥̇ 2 ሺ𝑡ሻ = 𝐾𝑡 𝐵 𝑥2 ሺ𝑡ሻ + 0 𝑢ሺ𝑡ሻ
− 0
𝐽 𝐽
[𝑥̇ 3 ሺ𝑡ሻ] [ 0 1 0] [𝑥3 ሺ𝑡ሻ] [ 0 ]
𝑦1 ሺ𝑡ሻ 1 0 0 𝑥1 ሺ𝑡ሻ

𝑦2 ሺ𝑡ሻ = 0 1 0 𝑥2 ሺ𝑡ሻ

[𝑦3 ሺ𝑡ሻ] [0 0 1] [𝑥3 ሺ𝑡ሻ]

The block diagram representation of the armature-controlled DC motor from the
relationships of the state variables in equations (7), (8), (9), (10), (11) and (12) is shown in
figure E1.5d.

18
Figure E1.5d: Block diagram representation of the state-space model of an armature-
controlled DC motor

1.6 1.7 Transfer Function from State Space Model Representation

The transfer function and the input-output differential equation of a system can be found
from the state space representation of that system is given. This operation can be done by
obtaining the Laplace transform and matrix operation as presented in section 1.4. The
following examples illustrates the general method for obtaining transfer function of a
dynamic LTI continuous time system from the state space model.
Example E1.7: Obtain the transfer function of the system defined by the following
state-space equations:
𝑥̇ 1 ሺ𝑡ሻ 𝑥 ሺ𝑡ሻ 0
0 1 1
[ ]=[ ][ ] + [ ] 𝑢ሺ𝑡ሻ
−10 −7
𝑥̇ 2 ሺ𝑡ሻ 𝑥2 ሺ𝑡ሻ 1
𝑥1 ሺ𝑡ሻ
𝑦ሺ𝑡ሻ = [1 0] [ ]
𝑥2 ሺ𝑡ሻ
Solution: The system has only one input 𝑢ሺ𝑡ሻ and one output 𝑦ሺ𝑡ሻ because matrix 𝐵
has just one column and matrix 𝐶 has one row. From the given state-space model, we obtain
0
0 1
𝐴=[ ], 𝐵 = [ ], 𝐶 = [1 0] and 𝐷 = [0]
−10 −7
1
The Transfer function is given by,
𝑌ሺ𝑠ሻ
𝐺ሺ𝑠ሻ = = 𝐶[𝑠𝐼 − 𝐴]−1 𝐵 + 𝐷 ሺ1ሻ
𝑈ሺ𝑠ሻ

1 0 𝑠 0
where 𝑠𝐼 = 𝑠 [ ]=[ ]
0 1 0 𝑠

19
 𝑠 −1
therefore ሺ𝑠𝐼 − 𝐴ሻ = [ ]
10 𝑠 + 7
the inverse of matrix ሺ𝑠𝐼 − 𝐴ሻ is found as
𝑠+7 1
adjሺ𝑠𝐼 − 𝐴ሻ [ ]
[𝑠𝐼 − 𝐴]−1 = = −10 𝑠
detሺ𝑠𝐼 − 𝐴ሻ 𝑠 2 + 7𝑠 − 10
From equation (1), The transfer function 𝐺 is obtained as
0
𝑠+7 1
[1 0] [ ][ ]
−10 𝑠
𝐶[𝑠𝐼 − 𝐴]−1 𝐵 = 1
2
𝑠 + 7𝑠 − 10
1
𝐺=
𝑠2 + 7𝑠 − 10

Example E1.8: Obtain the transfer function for the state-space representation of a
system given by
𝑥̇ 1 ሺ𝑡ሻ −2 1 0 𝑥1 ሺ𝑡ሻ 0
ሺ𝑡ሻ ሺ𝑡ሻ
[𝑥̇ 2 ] = [ 0 −3 1 ] [𝑥2 ] + [0] 𝑢ሺ𝑡ሻ
𝑥̇ 3 ሺ𝑡ሻ −3 −4 −5 𝑥3 ሺ𝑡ሻ 1

𝑥1 ሺ𝑡ሻ
𝑦ሺ𝑡ሻ = [0 1 0] [𝑥2 ሺ𝑡ሻ]
𝑥3 ሺ𝑡ሻ
Solution: From the given state-space model, we have that

−2 1 0 0
𝐴=[ 0 −3 1 ], 𝐵 = [0], 𝐶 = [0 1 0] and 𝐷 = [0]
−3 −4 −5 1
The Transfer function is given by,
𝑌ሺ𝑠ሻ
𝐺ሺ𝑠ሻ = = 𝐶[𝑠𝐼 − 𝐴]−1 𝐵 + 𝐷 ሺ1ሻ
𝑈ሺ𝑠ሻ
𝑠+2 −1 0
[𝑠𝐼 − 𝐴] = [ 0 𝑠+3 −1 ]
3 4 𝑠+5
1
[𝑠𝐼 − 𝐴]−1 = × adjሺ𝑠𝐼 − 𝐴ሻ
detሺ𝑠𝐼 − 𝐴ሻ

20
 𝑠 2 + 8𝑠 + 19 𝑠+5 1
1
= 3 [ −3 𝑠 2 + 7𝑠 + 10 𝑠+2 ]
𝑠 + 10𝑠 2 + 35𝑠 + 41
−3𝑠 − 9 −4𝑠 − 11 𝑠 2 + 5𝑠 + 6
From equation (1), the transfer function is
𝑠 2 + 8𝑠 + 19 𝑠+5 1 0
1
𝐺ሺ𝑠ሻ = [0 1 0] 3 2 [ −3 𝑠 2 + 7𝑠 + 10 𝑠+2 ] [ 0]
𝑠 + 10𝑠 + 35𝑠 + 41
−3𝑠 − 9 −4𝑠 − 11 𝑠 2 + 5𝑠 + 6 1

1
𝐺ሺ𝑠ሻ =
𝑠 3 + 10𝑠 2 + 35𝑠 + 41

CHAPTER TWO

SOLUTION OF STATE EQUATION

Learning Outcomes after completing this chapter will be to enable the reader:
• solve for eigenvalues and eigenvectors.
• obtain the solution of state equation of a continuous time system.
• compute the state transition matrix form state equations.

2.1 Eigenvalues and Eigenvectors

For a system of equations given by equation (2.1), where 𝐀 is a square matrix of the order 𝑛
with 𝐱̇ and 𝐱 been column vectors. The solution of the equation exits if 𝐱̇ and 𝐱 have the
same direction in space but differ only in magnitude by a scaler factor 𝜆.
21
 𝐱̇ = 𝐀𝐱 ሺ2.1ሻ
Thus, the solution must have the form
𝐱̇ = 𝜆𝐱 ሺ2.2ሻ
The diagonal matrix 𝜆 is a scaler factor or quantity (i.e. a number),
Substituting equation (2.2) in equation (2.1) and rearranging terms gives
[𝜆𝐈 − 𝑨]𝐱 = 0 ሺ2.3ሻ
Equation (2.3) has a nontrivial solution ሺ𝐱 ≠ 0ሻ i.e. only if 𝐱 is not zero and as such matrix
[𝜆𝐈 − 𝑨] do not have full rank and the determinant of the coefficients of 𝐱 must be zero, i.e.
𝑑𝑒𝑡ሺ𝜆𝐈 − 𝑨ሻ = |𝜆𝐈 − 𝑨| = 0 ሺ2.4ሻ
The unit matrix 𝐈 was introduced since we can subtract only a matrix from another matrix.
When the matrix 𝐀 is of the order 𝑛, the resulting homogeneous equation is written as
ሺ𝑎11 − 𝜆ሻ𝑥1 + 𝑎12 𝑥2 + 𝑎12 𝑥3 + ⋯ + 𝑎1𝑛 𝑥𝑛 = 0
𝑎21 𝑥1 + ሺ𝑎22 − 𝜆ሻ𝑥2 + 𝑎23 𝑥3 + ⋯ + 𝑎2𝑛 𝑥𝑛 = 0
⋮ ሺ2.5ሻ
ሺ𝑎
𝑎𝑛1 𝑥1 + 𝑎𝑛2 𝑥2 + 𝑎𝑛3 𝑥3 + ⋯ + 𝑛𝑛 − 𝜆ሻ𝑥𝑛 = 0
and the resulting set of equations are simplified in equation (2.6)
ሺ𝑎11 − 𝜆ሻ 𝑎12 ⋯ 𝑎1𝑛 𝑥1 0
𝑎21 ሺ𝑎22 − 𝜆ሻ ⋯ 𝑎 𝑥
) ( ⋮ ) = (0)
2𝑛 2
( ሺ2.6ሻ
⋮ ⋮ ⋯ ⋮ ⋮
𝑎𝑛1 𝑎𝑛2 ⋯ ሺ𝑎𝑛𝑛 − 𝜆ሻ 𝑥 𝑛 0

The roots 𝜆𝑖 of the characteristic equation which may be distinct (simple), repeated with a
multiplicity 𝑝, real or complex are known as the eigenvalues, characteristic values or
latent roots of matrix 𝑨. The corresponding solutions of the given equations 𝐀𝐱 = 𝜆𝐱 are
called the eigenvectors, or the characteristics vectors of matrix 𝐀. The polynomial 𝑄ሺ𝜆ሻ
may be written in factored form as
𝑄ሺ𝜆ሻ = ሺ𝜆 − 𝜆1 ሻሺ𝜆 − 𝜆2 ሻ … ሺ𝜆 − 𝜆𝑛 ሻ ሺ2.7ሻ
It is important to note that the product of eigenvalues of a matrix 𝑨 is equal to its determinant,
that is
𝜆1 𝜆2 … 𝜆𝑛 = |𝑨| ሺ2.8ሻ
and the sum of the eigenvalues is equal to the sum of the elements on the main diagonal (the
trace) of 𝑨, that is
𝑛 𝑛

∑ 𝜆𝑖 = ∑ 𝑎𝑖𝑖 ≡ 𝑡𝑟𝑎𝑐𝑒 𝑨 ሺ2.9ሻ

𝑙 𝑖=1

Example E2.1: Find the eigenvalue and the eigenvector for the matrix A given by
1 0 4
𝐴 = [0 2 0 ]
3 1 −3

22
Solution: From the determinant of the matrix, |𝑨 − 𝜆𝐈| = 0, we have the characteristics
equation to be
𝜆 0 0 1 0 4 1−𝜆 0 4
[0 𝜆 0 ] − [0 2 0 ] = [ 0 2−𝜆 0 ]
0 0 𝜆 3 1 −3 3 1 −3 − 𝜆
Expanding this gives
ሺ1 − 𝜆ሻ[ሺ2 − 𝜆ሻሺ−3 − 𝜆ሻ − 0] + 4[0 − 3ሺ2 − 𝜆ሻ] = 0
ሺ2 − 𝜆ሻ[ሺ1 − 𝜆ሻሺ−3 − 𝜆ሻ − 12] = 0
therefore 𝜆=2 or 𝜆2 + 2𝜆 − 15 = 0
hence the Eigenvalues are 𝜆 = 2, 3 or − 5
The Eigenvectors are obtained as follow
The first eigenvector 𝐱1 is found with 𝜆1 = 2, where [𝑨 − 𝜆𝐈]𝐱 = 0 becomes
1 0 4 1 0 0 𝑥1 0
{[0 2 0 ] − 2 [ 0 1 0 ]} [ 𝑥2 ] = [ 0]
3 1 −3 0 0 1 𝑥3 0
−1 0 4 𝑥1 0
[ 0 0 0 ] [𝑥2 ] = [0]
3 1 −5 𝑥3 0
1
−𝑥1 + 4𝑥3 = 0 then 𝑥3 = 4 𝑥1
5 7
3𝑥1 + 𝑥2 − 5𝑥3 = 0; 3𝑥1 + 𝑥2 − 𝑥1 = 0; then 𝑥2 = − 𝑥1
4 4
7 1
Therefore, 𝑥1 , 𝑥2 , 𝑥3 are the ratio 1: − 4 : 4, i.e. 4: -7:1 when the largest component of the
eigenvector is of unity magnitude.
Thus, the eigenvector corresponding to the first eigenvalue 𝜆1 = 2 is given by
𝑥1
𝑥1 7 4
− 𝑥
𝐱1 = [𝑥2 ] = 4 1 = [−7]
𝑥3 1 1
[ 4 𝑥1 ]
Similarly, the second eigenvector 𝐱 2 is found with 𝜆2 = 3, where [𝑨 − 𝜆𝐈]𝐱 = 0 becomes
1 0 4 1 0 0 𝑥1 0
{[0 2 𝑥
0 ] − 3 [0 1 0 ]} [ 2 ] = [0]
3 1 −3 0 0 1 𝑥3 0
−2 0 4 𝑥1 0
[ 0 −1 0 ] [𝑥2 ] = [0]
3 1 −6 𝑥3 0
1
−2𝑥1 + 4𝑥3 = 0 then 𝑥3 = 2 𝑥1
23
−𝑥2 = 0; 𝑥2 = 0; then 𝑥2 = 0
Thus, the eigenvector corresponding to the second eigenvalue 𝜆2 = 3 is given by
2
𝐱 2 = [0 ]
1
Finally, the third eigenvector 𝐱 3 is found with 𝜆3 = −5, where [𝑨 − 𝜆𝐈]𝐱 = 0 becomes
1 0 4 1 0 0 𝑥1 0
{[0 2 𝑥
0 ] + 5 [0 1 0 ]} [ 2 ] = [0]
3 1 −3 0 0 1 𝑥3 0
6 0 4 𝑥1 0
𝑥
[0 7 0 ] [ 2 ] = [0]
3 1 2 𝑥3 0
3
6𝑥1 + 4𝑥3 = 0 then 𝑥3 = − 2 𝑥1
7𝑥2 = 0; 𝑥2 = 0; then 𝑥2 = 0
Thus, the eigenvector corresponding to the second eigenvalue 𝜆2 = 3 is given by
2
𝐱3 = [ 0 ]
−1
In summary, we have
4 2 2
𝜆1 = 2, 𝐱1 = [−7]; 𝜆2 = 3, 𝐱 2 = [0]; 𝜆2 = 3, 𝐱 3 = [ 0 ]
1 1 −1

2.3 State Transition Matrix

The performance of a system can be examined by evaluating the time response of the state
variables of the control system. The time response of a control system comprises of the
transient response and steady state response. The transient response can be obtained by
solving for the solution of the state vector differential equation. The solution of the
homogenous or non-homogenous state vector differential equation requires that the State
Transition Matrix (STM) denoted with 𝜙ሺ𝑡ሻ be determined.
For a homogenous state equation,
𝑥̇ ሺ𝑡ሻ = 𝐴𝑥ሺ𝑡ሻ ሺ2.23ሻ
Taking Laplace transform of equation (2.23) gives
𝑠𝑋ሺ𝑠ሻ − 𝑥ሺ0ሻ = 𝐴𝑋ሺ𝑠ሻ ሺ2.24ሻ
Therefore, 𝑋ሺ𝑠ሻ = [𝑠𝐼 − 𝐴]−1 𝑥ሺ0ሻ ሺ2.25ሻ
Taking Inverse Laplace transform gives

24
 𝑥ሺ𝑡ሻ = ℒ −1 ሼ[𝑠𝐼 − 𝐴]−1 ሽ𝑥ሺ0ሻ ሺ2.26ሻ
Compering equation (2.26) with the solution of homogenous state equation of equation
(2.18) gives the state transition matrix, and the Laplace transform gives the resolvent matrix
denoted with 𝛷ሺ𝑠ሻ. Hence
𝛷ሺ𝑡ሻ = 𝑒 𝐴𝑡 = ℒ −1 ሼ[𝑠𝐼 − 𝐴]−1 ሽ
and in the s-domain, we have
𝛷ሺ𝑠ሻ = [𝑠𝐼 − 𝐴]−1
For a non-homogenous state equation,
𝑥̇ ሺ𝑡ሻ = 𝐴𝑥ሺ𝑡ሻ + 𝐵𝑢ሺ𝑡ሻ ሺ2.27ሻ
Taking Laplace transform, we get
𝑠𝑋ሺ𝑠ሻ − 𝑥ሺ0ሻ = 𝐴𝑋ሺ𝑠ሻ + 𝐵𝑈ሺ𝑠ሻ
or
[𝑠𝐼 − 𝐴]𝑋ሺ𝑠ሻ = 𝑥ሺ0ሻ + 𝐵𝑈ሺ𝑠ሻ ሺ2.28ሻ
Multiply both sides with [𝑠𝐼 − 𝐴]−1 gives
𝑋ሺ𝑠ሻ = [𝑠𝐼 − 𝐴]−1 𝑥ሺ0ሻ + [𝑠𝐼 − 𝐴]−1 𝐵𝑈ሺ𝑠ሻ
Taking inverse Laplace transform, we yield
𝑋ሺ𝑡ሻ = ℒ −1 ሼ[𝑠𝐼 − 𝐴]−1 ሽ𝑥ሺ0ሻ + ℒ −1 ሼ[𝑠𝐼 − 𝐴]−1 𝐵𝑈ሺ𝑠ሻሽ ሺ2.29ሻ
Comparing equation (2.29) with the solution of non-homogenous state equation of
equation (2.22), we have that
ℒሼ Φሺ𝑡ሻሽ = ℒሼ𝑒 𝐴𝑡 ሽ = Φሺ𝑡ሻ = [𝑠𝐼 − 𝐴]−1
Φሺ𝑡ሻ is called the state transition matrix because it relates the system state at time 𝑡 to
another time zero.

2.3.1 Properties of State Transition Matrix

Six useful properties and proofs of the STM 𝛷ሺ𝑡ሻ for the time-invariant system are stated
below.
(i) 𝛷ሺ0ሻ = 𝑒 𝐴ሺ0ሻ = 𝐼
−1
(ii) 𝛷ሺ𝑡ሻ = 𝑒 𝐴𝑡 = 𝑒 −ሺ−𝐴𝑡ሻ = [ 𝛷ሺ−𝑡ሻ]−1
(iii) 𝛷−1 ሺ𝑡ሻ = ሺ𝑒 𝐴𝑡 ሻ−1 = 𝑒 −𝐴𝑡 = 𝛷ሺ−𝑡ሻ
(iv) 𝛷ሺ𝑡1 + 𝑡2 ሻ = 𝑒 𝐴𝑡1 ∙ 𝑒 𝐴𝑡1 = 𝛷ሺ𝑡1 ሻ 𝛷ሺ𝑡2 ሻ = 𝛷ሺ𝑡2 ሻ 𝛷ሺ𝑡1 ሻ
(v) [ 𝛷ሺ𝑡ሻ]𝑘 = ሺ𝑒 𝐴𝑡 ሻ𝑘 = 𝑒 𝐴ሺ𝑘𝑡ሻ = 𝛷ሺ𝑘𝑡ሻ
(vi) 𝛷ሺ𝑡2 − 𝑡1 ሻ 𝛷ሺ𝑡1 − 𝑡0 ሻ = 𝛷ሺ𝑡2 − 𝑡0 ሻ = 𝛷ሺ𝑡1 − 𝑡0 ሻ 𝛷ሺ𝑡2 − 𝑡1 ሻ

Prove
25
 𝛷ሺ0ሻ = 𝐼 and 𝛷 −1 ሺ𝑡ሻ = 𝛷ሺ−𝑡ሻ
Solution:
Given the STM equation
𝛷ሺ𝑡ሻ = 𝑒 𝐴𝑡 ሺ2.30ሻ
−𝐴𝑡
Multiplying both sides by 𝑒 , we get
𝛷ሺ𝑡ሻ𝑒 −𝐴𝑡 = 𝑒 𝐴𝑡 𝑒 −𝐴𝑡 = 1
Multiplying both sides of equation (2.30) by 𝛷−1 ሺ𝑡ሻ, we get
𝛷 −1 ሺ𝑡ሻ 𝛷ሺ𝑡ሻ = 𝛷 −1 ሺ𝑡ሻ𝑒 𝐴𝑡 ሺ2.31ሻ
−1 ሺ𝑡ሻ𝑒 𝐴𝑡
𝐼= 𝛷
−𝐴𝑡
Therefore, 𝑒 = Φ−1 ሺ𝑡ሻ ሺ2.32ሻ
−1 ሺ𝑡ሻ −𝐴𝑡
Thus, Φሺ−𝑡ሻ = Φ =𝑒
The result from this proof shows that
𝑥ሺ0ሻ = Φሺ−𝑡ሻ𝑥ሺ𝑡ሻ ሺ2.33ሻ
which means that the STM process can be bilateral in time or the transition in time can take
place in either direction.
Prove:
Φሺ𝑡2 − 𝑡1 ሻ Φሺ𝑡1 − 𝑡0 ሻ = Φሺ𝑡2 − 𝑡0 ሻ 𝑓𝑜𝑟 𝑎𝑛𝑑 𝑡0 , 𝑡1 , 𝑡2
Solution:
Φሺ𝑡2 − 𝑡1 ሻ Φሺ𝑡1 − 𝑡0 ሻ = 𝑒 𝐴ሺ𝑡2−𝑡1 ሻ 𝑒 𝐴ሺ𝑡1−𝑡0ሻ
= 𝑒 𝐴ሺ𝑡2−𝑡1 ሻ = Φሺ𝑡1 − 𝑡0 ሻ ሺ2.34ሻ
This property of STM implies that a STM process can be divided into a number of sequences.
Prove:
[ Φሺ𝑡ሻ]𝑘 = Φሺ𝑘𝑡ሻ for 𝑘-positive integer
Solution:
[ Φሺ𝑡ሻ]𝑘 = 𝑒 𝐴𝑡 𝑒 𝐴𝑡 … 𝑒 𝐴𝑡 ሺ𝑘 𝑡𝑒𝑟𝑚𝑠ሻ ሺ2.35ሻ
= 𝑒 𝑘𝐴𝑡 = Φ𝑘ሺ𝑡ሻ

2.3.2 Computation of State Transition Matrix

Several methods of determining STM are discussed on this section.
(i) The Inverse Laplace Transform Approach
Through the method, we solve for the resolvent matrix Φሺ𝑠ሻ before obtaining it’s inverse
Laplace transform.
𝐴𝑑𝑗[𝑠𝐼 − 𝐴]
Φሺ𝑠ሻ = [𝑠𝐼 − 𝐴]−1 =
|𝑠𝐼 − 𝐴|
26
Taking the inverse Laplace transform yields,
Φሺ𝑡ሻ = ℒ −1 ሼ Φሺ𝑠ሻሽ

Example E2.1: Determine (a) the State Transition Matrix (b) the Zero-input response, (c)
the Zero-state response for 𝑢 = 1, (d) the total response, and (e) the inverse of state transition
matrix for a system with the vector-matrix equations given below and initial condition
𝑥ሺ0ሻ = [1 1]𝑇 .
𝑥̇ 1 ሺ𝑡ሻ 𝑥 ሺ𝑡ሻ 0
0 1 1
[ ]=[ ][ ] + [ ] 𝑢ሺ𝑡ሻ
−4 −5
𝑥̇ 2 ሺ𝑡ሻ 𝑥2 ሺ𝑡ሻ 1
Solution: The state equation is expressed as
𝑥̇ = 𝐴𝑥 + 𝐵𝑢
0
0 1
Thus, 𝐴=[ ] and 𝐵=[ ]
−4 −5
1
(a) the State Transition Matrix 𝜱ሺ𝑡ሻ.
1 0 0 1 𝑠 −1
𝑠𝐼 − 𝐴 = 𝑠 [ ]−[ ]=[ ]
0 1𝑠 −4 −5 4 𝑠+5
𝑠 −1
Therefore, |𝑠𝐼 − 𝐴| = | | = 𝑠 2 + 5𝑠 + 4
4 𝑠+5
Then,
𝑎𝑑𝑗ሺ𝑠𝐼 − 𝐴ሻ
ሺ𝑠𝐼 − 𝐴ሻ−1 =
|𝑠𝐼 − 𝐴|
1 𝑠+5 1
= [ ]
𝑠2 + 5𝑠 + 4 −4 𝑠

𝑠+5 1
𝑠2 2
+ 5𝑠 + 4 𝑠 + 5𝑠 + 4
=
−4 𝑠
[𝑠 2 + 5𝑠 + 4 𝑠 2 + 5𝑠 + 4]

4 1 1 1
− −
3ሺ𝑠 + 1ሻ 3ሺ𝑠 + 4ሻ 3ሺ𝑠 + 1ሻ 3ሺ𝑠 + 4ሻ
=
−4 4 −1 4
+ +
[3ሺ𝑠 + 1ሻ 3ሺ𝑠 + 4ሻ 3ሺ𝑠 + 1ሻ 3ሺ𝑠 + 4ሻ]
27
The STM is given by 𝜱ሺ𝑡ሻ = 𝑒 𝐴𝑡 = ℒ −1 ሼ[𝒔𝑰 − 𝑨]−1 ሽ
Therefore,
4 −𝑡 1 −4𝑡 1 −𝑡 1 −4𝑡
𝑒 − 𝑒 𝑒 − 𝑒
3 3 3 3
𝜱ሺ𝑡ሻ =
4 −𝑡 4 −4𝑡 1 4
[ 3𝑒 − 3𝑒
− − 𝑒 −𝑡 + 𝑒 −4𝑡 ]
3 3

(b) the Zero-input response.

The solution of state equation is given by
𝑡

𝑥ሺ𝑡ሻ = 𝑒 𝐴𝑡 𝑥ሺ0ሻ + න 𝑒 𝐴ሺ𝑡−𝜏ሻ 𝐵𝑢ሺ𝜏ሻ 𝑑𝜏

0

For a zero-input response denoted here by 𝑥1 ሺ𝑡ሻ, we make 𝑢ሺ𝜏ሻ = 0 and use the first part of
the solution of the state equation. Therefore,
𝑥1 ሺ𝑡ሻ = 𝑒 𝐴𝑡 𝑥ሺ0ሻ
Considering that the STM 𝜱ሺ𝑡ሻ = 𝑒 𝐴𝑡 and the initial condition 𝑥ሺ0ሻ = [1 1]𝑇 , then
𝑥1 ሺ𝑡ሻ = 𝜱ሺ𝑡ሻ𝑥ሺ0ሻ
4 −𝑡 1 −4𝑡 1 −𝑡 1 −4𝑡
𝑒 − 𝑒 𝑒 − 𝑒 1
3 3 3 3
𝑥1 ሺ𝑡ሻ = [ ]
4 −𝑡 4 −4𝑡 1 −𝑡 4 −4𝑡 1
−
[ 3 𝑒 + 𝑒 − 𝑒 + 𝑒 ]
3 3 3
1 1 −𝑡
ሺ4𝑒 −𝑡 − 𝑒 4𝑡 ሻ ሺ𝑒 − 𝑒 4𝑡 ሻ 1
3 3
𝑥1 ሺ𝑡ሻ = [ ]
4 −𝑡 −4𝑡
1 −𝑡 −4𝑡 1
[3 ሺ−𝑒 + 𝑒 ሻ 3 ሺ−𝑒 + 4𝑒 ሻ]
1
ሺ5𝑒 −𝑡 − 2𝑒 −4𝑡 ሻ
3
𝑥1 ሺ𝑡ሻ =
4
ሺ−5𝑒 −𝑡 + 8𝑒 −4𝑡 ሻ
[3 ]
(c) the Zero-state response for 𝑢 = 1.
The zero-state response denoted here with 𝑥2 ሺ𝑡ሻ is obtained by considering only the second
part of the solution of the state equation. That is
𝑡

𝑥2 ሺ𝑡ሻ = න 𝑒 𝐴ሺ𝑡−𝝉ሻ 𝑩𝑢ሺ𝝉ሻ 𝑑𝝉

0

28
 1 1 −ሺ𝑡−𝝉ሻ
(4𝑒 −ሺ𝑡−𝝉ሻ − 𝑒 4𝑡 )
𝑡 (𝑒 − 𝑒 4ሺ𝑡−𝝉ሻ ) 0
3 3
=න [ ] 𝑑𝝉
4 1
0
[3 (−𝑒
−ሺ𝑡−𝝉ሻ
+𝑒 −4ሺ𝑡−𝝉ሻ
) −𝑡
(−𝑒 + 4𝑒 −4ሺ𝑡−𝝉ሻ
)] 1
3
1 −ሺ𝑡−𝝉ሻ
𝑡 (𝑒 − 𝑒 −4ሺ𝑡−𝝉ሻ )
3
=න 𝑑𝝉
0 1 −ሺ𝑡−𝝉ሻ
[3 (−𝑒 + 4𝑒 −4ሺ𝑡−𝝉ሻ )]
𝑡
1 1 1
න (𝑒 −ሺ𝑡−𝝉ሻ − 𝑒 −4ሺ𝑡−𝝉ሻ ) 𝑑𝝉 (1 − ) − (𝑒 −𝑡 − 𝑒 −4𝑡 )
3 4 4
0
= 𝑡 =
1
න (−𝑒 −ሺ𝑡−𝝉ሻ + 4𝑒 −4ሺ𝑡−𝝉ሻ )𝑑𝝉 [ሺ−1 + 1ሻ − ሺ−𝑒 −𝑡 + 𝑒 −4𝑡 ሻ]
3
[0 ]
3 1
− 𝑒 −𝑡 + 𝑒 −4𝑡
4 4
=
[ 𝑒 −𝑡 − 𝑒 −4𝑡 ]
(d) the total response.
The total response is given by
𝑡

𝑥ሺ𝑡ሻ = 𝑥1 ሺ𝑡ሻ + 𝑥2 ሺ𝑡ሻ = 𝑒 𝐴𝑡 𝑥ሺ0ሻ + න 𝑒 𝐴ሺ𝑡−𝜏ሻ 𝐵𝑢ሺ𝜏ሻ 𝑑𝜏

0
1 3 1
ሺ5𝑒 −𝑡 − 2𝑒 −4𝑡 ሻ − 𝑒 −𝑡 + 𝑒 −4𝑡
3 4 4
= +
4
ሺ−5𝑒 −𝑡 + 8𝑒 −4𝑡 ሻ [ 𝑒 −𝑡 − 𝑒 −4𝑡 ]
[3 ]
3 2 −𝑡 5
+ 𝑒 − 𝑒 −4𝑡
4 3 12
=
2 −𝑡 5 −4𝑡
[ − 𝑒 + 𝑒 ]
3 3
Therefore,

29
 3 2 −𝑡 5 −4𝑡
𝑥1 ሺ𝑡ሻ + 𝑒 − 𝑒
4 3 12
[ ]=
𝑥2 ሺ𝑡ሻ 2 −𝑡 5 −4𝑡
[ −3𝑒 + 3𝑒 ]
(e) the inverse of state transition matrix.
This is obtained as 𝜱−1 ሺ𝑡ሻ = 𝑒 −𝐴𝑡 = 𝜱ሺ−𝑡ሻ
1 1 𝑡
ሺ4𝑒 𝑡 − 𝑒 4𝑡 ሻ ሺ𝑒 − 𝑒 4𝑡 ሻ
3 3
=
4 𝑡 1
ሺ𝑒 + 𝑒 4𝑡 ሻ ሺ−𝑒 𝑡 + 4𝑒 4𝑡 ሻ
[− 3 3 ]

Example E2.2: A system is given by the following state equation

𝑥̇ ሺ𝑡ሻ = 𝐴𝑥ሺ𝑡ሻ + 𝐵𝑢ሺ𝑡ሻ
1
2 0
where 𝐴=[ ] and 𝐵=[ ]
−1 −1
1
(a) Determine the STM of the system by Inverse Laplace and by series summation
methods.
(b) Determine the solution of the system (State Vector 𝑥ሺ𝑡ሻ) with a unit step function
1
input for 𝑡 ≥ 0 and initial vector 𝑥ሺ0ሻ = [ ].
0
Solution:
(a) To solve for the STM of the system
(i) Using the Inverse Laplace Transform Method
The STM is given by 𝜱ሺ𝑡ሻ = 𝑒 𝐴𝑡 = ℒ −1 ሼ[𝒔𝑰 − 𝑨]−1 ሽ
Therefore,
𝑠−2 0
[𝒔𝑰 − 𝑨] = [ ]
1 𝑠+1
And the resolvent matrix 𝜱ሺ𝑠ሻ = [𝒔𝑰 − 𝑨]−1 is
𝑎𝑑𝑗ሺ𝑠𝐼 − 𝐴ሻ
[𝒔𝑰 − 𝑨]−1 =
|𝑠𝐼 − 𝐴|
1 𝑠+1 0
= [ ]
ሺ𝑠 − 2ሻሺ𝑠 + 1ሻ −1 𝑠−2

30
 1
0
𝑠−2
=
−1 1
[ሺ𝑠 − 2ሻሺ𝑠 + 1ሻ 𝑠 + 1]

Taking the Inverse Laplace Transform of the resolvent matrix 𝜱ሺ𝑠ሻ gives the STM
𝜱ሺ𝑡ሻ = ℒ −1 ሼ𝜱ሺ𝑠ሻሽ
1
0
𝑠−2
ℒ −1
−1 1
{[ሺ𝑠 − 2ሻሺ𝑠 + 1ሻ 𝑠 + 1]}
Individually, the inverse transform is
1 −1 1 1 1
ℒ −1 {𝑠−2} = 𝑒 2𝑡 , ℒ −1 ሼ0ሽ = 0, ℒ −1 {ሺ𝑠−2ሻሺ𝑠+1ሻ} = − 3 𝑒 2𝑡 + 3 𝑒 −𝑡 and ℒ −1 {𝑠+1} = 𝑒 −𝑡

Hence,
𝑒 2𝑡 0
𝜱ሺ𝑡ሻ = [ 1 ]
− ሺ𝑒 2𝑡 + 𝑒 −𝑡 ሻ 𝑒 −𝑡
3

(ii) Using the Series Summation Method

By series summation method, we have that
𝐴𝑡 𝐴2 𝑡 2 𝐴3 𝑡 3 𝐴𝑛 𝑡 𝑛
Φሺ𝑡ሻ = 𝐼 + + + + ⋯+
1! 2! 3! 𝑛!
2 0
Given that 𝐴=[ ]
−1 −1
2 0 2 0 4 0
Then 𝐴2 = [ ][ ]=[ ]
−1 −1 −1 −1 −1 1
2 0 4 0 8 0
Also 𝐴3 = [ ][ ]=[ ]
−1 −1 −1 1 −3 −1
Considering the series summation up till 𝑛 = 3 while ignoring the higher order terms, we
have
𝐴𝑡 𝐴2 𝑡 2 𝐴3 𝑡 3
Φሺ𝑡ሻ = 𝐼 + + + +⋯
1! 2! 3!
Substituting 𝐴, 𝐴2 , 𝐴3 in the above equation yield

31
 2 3
1 0 2 0 4 0 𝑡 8 0 𝑡
𝛷ሺ𝑡ሻ = [ ]+[ ]𝑡 + [ ] +[ ] +⋯
0 1 −1 −1 −1 1 2 −3 −1 6
2 4𝑡 3
2𝑡 0 0
1 0 2𝑡 0 3
=[ ]+[ ] + [ 𝑡2 𝑡2] + +⋯
0 1 −𝑡 −𝑡 − 𝑡3 𝑡3
2 2 [− 2 − ]
6
4𝑡 3
2
1 + 2𝑡 + 2𝑡 + +⋯ 0
3
=
𝑡2 𝑡3 𝑡2 𝑡3
[ −𝑡 − 2 − 2 + ⋯ 1 − 𝑡 + − + ⋯]
2 6
𝑒 2𝑡 0
=[ 1 1 ]
− 𝑒 2𝑡 + 𝑒 −𝑡 𝑒 −𝑡
3 3

(b) The solution of the system

The solution of the state equation of the system has two parts as given below
𝑡

𝑥ሺ𝑡ሻ = 𝑒 𝐴𝑡 𝑥ሺ0ሻ + න 𝑒 𝐴ሺ𝑡−𝜏ሻ 𝐵𝑢ሺ𝜏ሻ 𝑑𝜏

0

Therefore, by substituting values for 𝐵, initial vector 𝑥ሺ0ሻ, unit step function input 𝑢ሺ𝑡ሻ =
1, and STM 𝛷ሺ𝑡ሻ in the total solution of the system, we obtain
𝑒 2𝑡 0 1 𝑡 𝑒 2ሺ𝑡−𝜏ሻ 0 1
𝑥ሺ𝑡ሻ = [ 1 1 ][ ] + න[ 1 1 ] [ ] [1] 𝑑𝜏
− 𝑒 2𝑡 + 𝑒 −𝑡 𝑒 −𝑡 − 𝑒 2ሺ𝑡−𝜏ሻ + 𝑒 −ሺ𝑡−𝜏ሻ 𝑒 −ሺ𝑡−𝜏ሻ
3 3 0 0 3 3 0
𝑡
𝑒 2𝑡 𝑒 2ሺ𝑡−𝜏ሻ
−
2
=[ 1 1 ]+
− 𝑒 2𝑡 + 𝑒 −𝑡 1 2ሺ𝑡−𝜏ሻ 2 −ሺ𝑡−𝜏ሻ
3 3 [6 𝑒 + 𝑒
3 ]0

𝑒 2𝑡 1 𝑒 2𝑡
− +
2 2
=[ 1 1 −𝑡 ] + 1 1
2𝑡
− 𝑒 + 𝑒 2𝑡
2 2 −𝑡
3 3 −
[6 6 𝑒 + − 𝑒 ]
3 3

32
 1 3𝑒 2𝑡
− +
2 2
=
5 1 2𝑡 1 −𝑡
[6 − 6 𝑒 − 3 𝑒 ]
Therefore, the solution with State Vectors 𝑥1 ሺ𝑡ሻ and 𝑥2 ሺ𝑡ሻ are
1 3𝑒 2𝑡
𝑥1 ሺ𝑡ሻ = − +
2 2
and
5 1 2𝑡 1 −𝑡
𝑥2 ሺ𝑡ሻ = − 𝑒 − 𝑒
6 6 3
Example E2.4: Use 𝜆 matrix method to solve for the eigenvalues, eigenvectors, STM and
the solution of the system, that is, state vector 𝑥ሺ𝑡ሻ for the system described by 𝑥̇ ሺ𝑡ሻ =
0 1
[ ] 𝑥ሺ𝑡ሻ. The system has a unit step function input for 𝑡 ≥ 0 and the initial vector
−2 −3
1
𝑥ሺ0ሻ = [ ].
0
Solution:
0 1
Given that 𝐴 = [ ], then
−2 −3
For Eigenvalues:
|𝜆𝐼 − 𝐴| = |𝜆 −1
|
2 𝜆+3
= 𝜆2 + 3𝜆 + 2
Solving the characteristics 𝜆2 + 3𝜆 + 2 = 0 give the eigenvalues of the matrix 𝐴 as 𝜆1 =
−1 and 𝜆2 = −2.
For Eigenvectors:
Let 𝑅1 be the eigenvector corresponding to the eigenvalue 𝜆1 = −1, then [𝜆1 𝐼 − 𝐴]𝑅1 = 0
will be satisfied. Hence,
1 0 0 1 𝑝11
[−1 [ ]−[ ]] [ ] = 0
0 1 −2 −3 𝑝21
The above matrix gives simultaneous equations as
𝑝11 + 𝑝12 = 0 ሺ1ሻ
−2𝑝11 − 2𝑝12 = 0 ሺ2ሻ
Solving equations (1) and (2) simultaneously yield 𝑝11 = −𝑝21
If 𝑝11 = 𝑝21 = 1, then, the eigenvector for the eigenvalue 𝜆1 = 1 is
1
𝑃1 = [ ]
−1
33
Let 𝑅2 be the eigenvector corresponding to the eigenvalue 𝜆2 = −2, then [𝜆2 𝐼 − 𝐴]𝑅2 = 0
will be satisfied. Hence
1 0 0 1 𝑝12
[−2 [ ]−[ ]] [ ] = 0
0 1 −2 −3 𝑝22
The above matrix gives simultaneous equations as
2𝑝12 + 𝑝22 = 0 ሺ3ሻ
−2𝑝11 − 𝑝22 = 0 ሺ4ሻ
Solving equations (3) and (4) simultaneously yield 𝑝22 = −2𝑝12
If 𝑝12 = 1 and 𝑝22 = −2 , then, the eigenvector for the eigenvalue 𝜆2 = −2 is
1
𝑃2 = [ ]
−2

For the State Transition Matrix:

The model matrix is
𝑀 = [𝑃1 𝑃2 ] = [ 1 1
]
−1 −2
Taking the inverse of the modal matrix, we get
2 1
𝑀−1 = [ ]
−1 −1
For a well-formed model matrix 𝑀, it should diagonalise matrix 𝐴 using transformation such
that the diagonal elements are similar to the obtained eigenvalues values. Therefore, It is
seen that
2 1 0 1 1 1
𝑀−1 𝐴𝑀 = [ ][ ][ ]
−1 −1 −2 −3 −1 −2
2 1 −1 −2
=[ ][ ]
−1 −1 1 4
−1 0
=[ ]=𝜆 ሺ5ሻ
0 −2
Equation (5) shows that the determined model matrix is correct because the diagonal
elements are same as the obtained eigen values.
−1 0 −𝑡
Since 𝜆 = [ ], then 𝑒 𝜆𝑡 = [ 𝑒 0 ]
−2𝑡
0 −2 0 𝑒
Therefore, STM, 𝛷ሺ𝑡ሻ = 𝑀𝑒 𝜆𝑡 𝑀−1
1 1 𝑒 −𝑡 0 2 1
𝛷ሺ𝑡ሻ = [ ][ −2𝑡 ] [−1 ]
−1 −2 0 𝑒 −1
1 1 2𝑒 −𝑡 −𝑡
=[ ][ −𝑒 ]
−1 −2 −𝑒 −2𝑡 𝑒 −2𝑡
−𝑡 −2𝑡
= [ 2𝑒−𝑡 − 𝑒 −2𝑡 −𝑒 −𝑡 + 𝑒 −2𝑡 ]
2𝑒 − 2𝑒 −𝑒 −𝑡 + 2𝑒 −2𝑡
34
For the Solution of the System:
The system is a homogenous type, i.e. 𝑥̇ ሺ𝑡ሻ = 𝐴𝑥ሺ𝑡ሻ, hence the solution of the system is
given by
𝑥ሺ𝑡ሻ = 𝑒 𝐴𝑡 𝑥ሺ0ሻ ሺ5ሻ
Substituting the given parameters in equation (5), we get
−𝑡 −2𝑡
𝑥ሺ𝑡ሻ = [ 2𝑒−𝑡 − 𝑒 −2𝑡 −𝑒 −𝑡 + 𝑒 −2𝑡 ] [1]
2𝑒 − 2𝑒 −𝑒 −𝑡 + 2𝑒 −2𝑡 0
2𝑒 −𝑡 − 𝑒 −2𝑡
=[ ]
2𝑒 −𝑡 − 2𝑒 −2𝑡
Therefore, the solution with state vectors are
𝑥1 ሺ𝑡ሻ = 2𝑒 −𝑡 − 𝑒 −2𝑡
and 𝑥1 ሺ𝑡ሻ = 2𝑒 −𝑡 − 2𝑒 −2𝑡

35
 CHAPTER THREE

STATE VARIABLE FEEDBACK SYSTEM DESIGN

Learning Outcomes after completing this chapter will be to enable the reader:
• Learn the basics of controller design and procedure for pole placement
• Appreciate the concept of Controllability and Observability.
• Design of State Feedback Design Controllers and Observers.
• Be familiar with the Model Control Observer Design.
• Understand the concept of Adaptive Control Design.
• Analyse Multivariable Control System.

3.1 Introduction
This chapter presents the design of controllers using state feedback. Design may be seen as
the act of modifying the control system’s parameters to meet the specified stability,
performance, and robustness objectives of a plant. A feedback control system is needed when
we do not desire the operation sequence of a plant and wishes modify the behaviour of that
plant to meet our requirements. We have two based design method; the Classical and the
Modern design. In Classical design, we vary the controller transfer function until a desired
closed-loop performance is achieved. Chapter 1 presented the state-space analysis and
system modeling where it was shown that state-space methods are basic tools for analysing
and designing feedback control systems. But for systems with nonlinearities, multiple-input
and multiple-output, a compensator in cascade with the plant or in the feedback path, having
correct placement of poles and zeros is designed to give a desired transient response and
steady-state error. A control system compensation scheme can be designed using the time-
domain system expressed in terms of state variables. This is particularly seen in a control
system whose control signal 𝑢ሺ𝑡ሻ is a function of some measurable state variables. Then, a
state variable controller whose operation depends on the measured information from the
input can be developed. This state variable controller is a kind of system compensation (a
compensator) that is useful for system optimisation.
The design of the state variable feedback control takes three major design steps. The first
step uses all the measurable state variable in a full-state feedback control law. However, full-
state feedback is not feasible practically since all the states are not measurable. In the second
step, the states that are not available at the output are estimated using an observer that is
constructed for the purpose of state estimation. This is because, not all the state may be
directly sensed and seen at the output. Some states need not be estimated owing to the fact
that the states are already available as the system output. When the states are not available,
the observer is a Full-state observer, while, when the states are already available, the
observer is a Reduced-order observer. The third design step is to interface the observer to
the full-state feedback control law and (optionally) consider the reference inputs to form a
Compensator.

36
3.2 Controller Design
The purpose of controller design is to use feedback to steer the state/output of the
system/plant towards some intended steady-state value or some reference/set point. The
assignment of poles of the closed-loop transfer function to any desired location along the
roots of the plant equation can be controlled through the introduction of additional
parameters. This is a design procedure for generating control laws, hence known as Pole
Assignment or Pole Placement.
In order to introduce the pole assignment technique in a controller design, we consider the
𝑛th-order feedback control system have an 𝑛th-order closed-loop characteristic equation of
the form
𝑠 𝑛 + 𝑎𝑛−1 𝑠 𝑛−1 + ⋯ + 𝑎1 𝑠 + 𝑎0 = 0 ሺ3.1ሻ
We determine the closed-loop pole locations of the system shown in equation (3.1) by
identifying the highest power of 𝑠 (which is unity). If we introduce 𝑛 adjustable parameters
that are related to the coefficients in equation (3.1), then, we can set the poles of the system’s
closed-loop to any location of choice.
Let’s consider a state-space representation of a plant given in equation (3.2a) and (3.2b) as
𝐱̇ ሺ𝑡ሻ = 𝐀𝐱ሺ𝐭ሻ + 𝐁𝐮ሺ𝐭ሻ ሺ3.2ሻ
𝑦ሺ𝑡ሻ = 𝐂𝐱ሺ𝐭ሻ + 𝐃𝐮ሺ𝐭ሻ ሺ3.3ሻ
Usually in a typical feedback control system shown in figure 3.1(a), there is a feedback of
the output 𝑦 to the summing junction. However, in order to have the required adjustable
closed-loop poles values, all the state variables need to be fed back to the control, 𝑢, through
a gain, 𝑘𝑖 . Therefore, there would be an 𝑛 gains, 𝑘𝑖 , that correspond to the order of the
characteristics equation of the system. The feedback gain matrix is indicated in figure 3.1(b)
by the feedback vector −𝐊. The thin lines in figure 3.1 are scalers and the thick lines are
vectors.

(a)

37
 (b)
Figure 3.1: (a) State-space representation of a feedback system; (b) State-space
representation of a system with state-variable feedback.
From figure 3.1(b), we select a control law of the form
𝐮 = ሺ𝒓 − 𝑲𝒙ሻ ሺ3.4ሻ
𝒓 is a vector of desired state variables.
we write the state equations for the closed-loop system as
𝐱̇ = 𝐀𝐱 + 𝐁𝑢 = 𝐀𝐱 + 𝐁ሺ𝐫 − 𝐤𝐱ሻ
therefore,
𝐱̇ = ሺ𝐀 − 𝐁𝐤ሻ𝐱 + 𝐁𝐫 ሺ3.5ሻ
𝐲 = 𝐂𝐱 ሺ3.6ሻ
For the system described by equation (3.5) and (3.6), the characteristics equation is given by
|ሺ𝑠𝐈 − 𝑨ሻ| = 0 ሺ3.7ሻ
Then the roots of the characteristics equation (3.7) are the open-loop poles or eigenvalues of
the system. While the closed-loop system described by equation (3.5) has the characteristic
equation as equation (3.8) and the roots of the equation are the closed-loop poles or the
eigenvalues.
|ሺ𝑠𝐈 − 𝐀 + 𝐁𝐊ሻ| = 0 ሺ3.8ሻ
The SFG implementation of the feedback system shown in figure 3.1(b) is drawn by making
an assumption of a plant SFG in phase-variable form as shown in figure 3.2(a). Each state
variable is then fed back to the plant’s input, 𝑢, through a gain, 𝑘𝑖 , as depicted in figure
3.2(b).

(a)

38
 (b)
Figure 3.2: (a) Phase-variable representation of the SFG of the feedback system; (b) SFG
representation of a system with state-variable feedback.
From figure 3.2(a), we can design the state-variable feedback for close-loop pole assignment
by equating the characteristic equation and solving for the values of the feedback gains, 𝑘𝑖
.
The phase-variable representation of the state-space expression of the system given in
equation 3.2 can be written as follows. Let the components of the system be written as
0
0 1 0 ⋯ 0
0 0 1 ⋯ 0 0
𝐀= ⋮ ⋮ ⋮ ⋮ ⋮ ; 𝐁= ; 𝐂 = [𝑐1 𝑐2 ⋯ 𝑐𝑛 ]
∙ ∙ ∙ ∙ ∙ ⋮
∙
[ −𝑎0 −𝑎1 −𝑎2 ⋯ −𝑎𝑛−1 ]
[1]
The characteristic equation of the system is 𝑠 𝑛 + 𝑎𝑛−1 𝑠 𝑛−1 + ⋯ + 𝑎1 𝑠 + 𝑎0 = 0
When each state-variable of the closed-loop system if fed back to the summing point 𝑢, we
have that
𝑢 = −𝐊𝐱 ሺ3.9ሻ
where
𝐊 = [𝑘1 𝑘2 ⋯ 𝑘𝑛 ] ሺ3.10ሻ
The feedback gains for the phase variables are the 𝑘𝑖 ’s, therefore, the system matrix for the
closed-loop system can be written in the phase-variable form as

39
 0 1 0 ⋯ 0
0 0 1 ⋯ 0
𝐀 − 𝐁𝐊 = ⋮ ⋮ ⋮ ⋮ ⋮ ሺ3.11ሻ
∙ ∙ ∙ ∙ ∙
[ −ሺ𝑎0 + 𝑘1 ሻ −ሺ𝑎1 + 𝑘2 ሻ −ሺ𝑎2 + 𝑘3 ሻ ⋯ −ሺ𝑎𝑛−1 + 𝑘𝑛 ሻ]
By inspection, we can write the characteristic equation of equations 3.5 and 3.6 as
𝑑𝑒𝑡(𝑠𝐈 − ሺ𝐀 − 𝐁𝐊ሻ) = 0. Hence, we can write,
𝑠 𝑛 + ሺ𝑎𝑛−1 + 𝑘𝑛 ሻ𝑠 𝑛−1 + ሺ𝑎𝑛−2 + 𝑘𝑛−1 ሻ𝑠 𝑛−2 + ⋯ + ሺ𝑎1 + 𝑘2 ሻ𝑠 + ሺ𝑎0 + 𝑘1 ሻ = 0 ሺ3.12ሻ
Comparing equations (3.1) and (3.12), we see that the closed-loop characteristics of a system
represented in the phase-variable form can be obtained from the open-loop characteristics
equation by inserting the supposed 𝑘𝑖 to each coefficient.
An example for a controller design for in the phase-variable form is shown in example 3.1.
Example 3.1: Design the phase-variable feedback gains of the given plant to obtain 9.5%
overshoot and a settling time of 0.74 seconds.
20ሺ𝑠 + 5ሻ
𝐺ሺ𝑠ሻ = ሺ3.13ሻ
𝑠ሺ𝑠 + 1ሻሺ𝑠 + 4ሻ
The needed closed-loop poles characteristic equation is −5.4 ± 𝑗7.2
We draw the signal-flow diagram for the system to be

Figure 3.3: Phase-variable representation for the plant of example 3.1

Feeding back all the state variables to the control, 𝑢, through gains 𝑘𝑖 , gives signal-flow
graph in figure 3.4.

40
Figure 3.4: Signal-flow graph representation of the plant with state-variable feedback of
example 3.1
We write the closed-loop system’s equations from figure 3.4 as
0 1 0 0
𝐱̇ = [ 0 0 1 ] 𝐱 + [0] 𝑟 ሺ3.14ሻ
−𝑘1 −ሺ 4 + 𝑘 2 ሻ −ሺ 5 + 𝑘3 ሻ 1
𝑦 = [100 20 0]𝐱 ሺ3.15ሻ
The closed-loop system’s characteristics equation is obtained using equation (3.8) as
𝑑𝑒𝑡ሺ𝑠𝐈 − 𝐀 + 𝐁𝐊ሻ = 0 ሺ3.16ሻ
𝑠 3 + ሺ5 + 𝑘3 ሻ𝑠 2 + ሺ4 + 𝑘2 ሻ𝑠 + 𝑘1 = 0 ሺ3.15ሻ
Comparing equation (3.15) and the desired characteristic equation (3.15) formed from poles
−5.4 ± 𝑗7.2 and −5.1.
𝑠 3 + 15.9𝑠 2 + 136.08𝑠 + 413.1 = 0 ሺ3.16ሻ
Equating the coefficients of equations (3.15) and (3.16), we have
𝑘1 = 413.1; 𝑘2 = 132.08, 𝑘3 = 10.9
The zero term of the closed-loop transfer function is the same as the zero term of the open-
loop system ሺ𝑠 + 5ሻ.
With equations (3.4) and (3.5), we have that the state-space representation of the closed-loop
system is:
0 1 0 0
𝐱̇ = [ 0 0 1 ] 𝐱 + [0] 𝑟 ሺ3.17ሻ
−413.1 −136.08 −15.9 1
41
 𝑦 = [100 20 0]𝐱 ሺ3.18ሻ
The transfer function is
20ሺ𝑠 + 5ሻ
𝑇ሺ𝑠ሻ = ሺ3.19ሻ
𝑠3 + 15.9𝑠 2 + 136.08𝑠 + 413.1

3.3 Controllability and Observability

The concept of controllability and observability was introduced in the design of modern
control system in state space because of the important role they play.

3.3.1 Controllability
The control signal 𝑢 can control the behaviour of each state variables of a system by
adjusting the pole location of that closed-loop system to any desired position. However, in
systems, some state variables are not controllable or may exhibit an unstable response
because of a nonzero initial condition. This means that a state-feedback design in not feasible
in some systems. Based on these facts, a system is controllable if an input to that system can
be seen to move every state variable from a desired initial state to a desired final state, while
a system is uncontrollable if the system’s state variables cannot be taken from an initial state
to a desired final state.
The controllability of a system whose state equation is given can be check by inspection.
Thus, a system having a distinct eigenvalue and a diagonal system matrix is controllable if
the input coupling matrix 𝐁 does not have any rows that are zero. However, when a system
does not have a diagonal or parallel form with distinct eigenvalues or has multiples poles
even when it is represented in parallel form, its controllability cannot be determined by
inspection.
The controllability of a plant in any representation or choice of state variables can be
determined by setting up the Controllability Matrix. For and 𝑛𝑡ℎ -order plant whose state
equation is given by
𝐱̇ ሺ𝑡ሻ = 𝐀𝐱ሺ𝐭ሻ + 𝐁𝐮ሺ𝐭ሻ ሺ3.20ሻ
𝒚ሺ𝑡ሻ = 𝐂𝐱ሺ𝐭ሻ + 𝐃𝐮ሺ𝐭ሻ

is completely controllable if the controllability matrix 𝐐𝐌 of the order 𝑛 × 𝑛 given in

equation (3.21) has the rank 𝑛.
𝐐𝐌 = [𝐁: 𝐀𝐁: 𝐀𝟐 𝐁 ∶ ∙∙∙ ∶ 𝐀𝒏−𝟏 𝐁] ሺ3.21ሻ

42
3.3.2 Observability
Observation is sometimes called Estimation, and it is used to calculate the state variables
when some or all the states are not available or not reachable from the system. The ability
to reach the state variables of a system depend on the system hardware. But most times it is
impractical to us the hardware to access the states variables because of cost, accuracy, or
availability. This can be seen in Inertial Measurement Unit (IMU) that can measure and
report body’s specific force, angular rate, and sometimes the orientation of the body using a
combination of accelerometers, gyroscope, and sometimes magnetometers. After some
period of usage without recalibration, the alignment of these measurement devices
deteriorates; thus, some alternative means of measuring the parameters may be desirable. In
some other applications, some of the state variables may not be available at all because of
the system configuration, while in some other application, the cost of measuring the state
variables and communicating them to the controller are costly. For control systems whose
state variables are not accessible from the plant, it is possible to estimate the state, and the
estimated state are then fed to the controller for the generation of the controlled signal for
the plant. The estimated state variables are calculated by an Estimated or an Observer which
is a model of the plant.
The state-space model of the system is used to check whether the system is observable or
not by determining the Observation Matrix. If a system is represented by matrix 𝐴, then, the
observability of the system is determined by solving the 𝑛 × 𝑛𝑚 composite matrix 𝐐𝑶 given
by equation (3.22) whose rank is 𝑛;
𝟐 ሺ𝒏−𝟏ሻ 𝐓
𝐐𝑶 = [𝐂 𝐓 : 𝐀𝐓 𝐂 𝐓 : (𝐀𝐓 ) 𝐂𝐓 ∶ ∙∙∙ ∶ (𝐀𝑻 ) 𝐂 ] ሺ3.22ሻ

Example E3.1: Write the state equation for the block diagram of the system shown below
in Fig(a), in which 𝑥1 ሺ𝑡ሻ, 𝑥2 ሺ𝑡ሻ, and 𝑥3 constitute the state vector. Also, determine whether
the system is completely controllable and observable.

Figure E3.5(a)
Solution: Consider the first block of figure E3.5(a) as show in figure E3.5(b).

Figure E3.5(b)
43
From figure E3.5(b), we have that
2
𝑋1 ሺ𝑠ሻ = [𝑋 ሺ𝑠ሻ − 𝑋3 ሺ𝑠ሻ]
𝑠ሺ𝑠 + 1ሻ 2
Then [𝑠 2 + 𝑠]𝑋1 ሺ𝑠ሻ = 2𝑋2 ሺ𝑠ሻ − 2𝑋3 ሺ𝑠ሻ
Taking the inverse Laplace transform, we get,
𝑥̈ 1 ሺ𝑡ሻ + 𝑥̇ 1 ሺ𝑡ሻ = 2𝑥2 ሺ𝑡ሻ − 2𝑥3 ሺ𝑡ሻ ሺ1ሻ
Consider the second block as given in figure E3.5(c)

Figure E3.5(c)
It is seen from figure E3.5(c) that
𝑋3 ሺ𝑠ሻ = 𝑠𝑋1 ሺ𝑠ሻ
Taking the inverse Laplace transform, we get,
𝑥̇ 1 ሺ𝑡ሻ = 3𝑥3 ሺ𝑡ሻ ሺ2ሻ
Consider the third block as given in figure E3.5(d)

Figure E3.5(d)
It is seen from figure E3.5(d) that
2
𝑋2 ሺ𝑠ሻ = ( ) [𝑈ሺ𝑠ሻ − 𝑋1 ሺ𝑠ሻ]
𝑠+3
Rearranging the equation gives
ሺ𝑠 + 3ሻ𝑋2 ሺ𝑠ሻ = 2𝑈ሺ𝑠ሻ − 2𝑋1 ሺ𝑠ሻ
𝑠𝑋2 ሺ𝑠ሻ = −2𝑋1 ሺ𝑠ሻ − 3𝑋2 ሺ𝑠ሻ + 2𝑈ሺ𝑠ሻ
Taking the inverse Laplace transform, we get,
𝑥̇ 2 ሺ𝑡ሻ = −2𝑥1 ሺ𝑡ሻ − 3𝑥2 ሺ𝑡ሻ + 2𝑢ሺ𝑡ሻ ሺ3ሻ
44
Differentiating equation (2) w.r.t 𝑡 , we obtain,
𝑥̈ 1 ሺ𝑡ሻ = 𝑥̇ 3 ሺ𝑡ሻ ሺ4ሻ
Substituting equation (2) and equation (4) into equation (1) and simplifying, we obtain,
𝑥̇ 3 ሺ𝑡ሻ = 2𝑥2 ሺ𝑡ሻ − 3𝑥3 ሺ𝑡ሻ ሺ5ሻ
Let the output be,
𝑦ሺ𝑡ሻ = 𝑥1 ሺ𝑡ሻ ሺ6ሻ
Arranging equations (2), (3), (5), and equation (6) to obtain the required state-space model
of the given system as;
𝑥̇ 1 ሺ𝑡ሻ 0 0 1 𝑥1 ሺ𝑡ሻ 0
[𝑥̇ 2 ሺ𝑡ሻ] = [−2 −3 0 ] [𝑥2 ሺ𝑡ሻ] + [2] 𝑢ሺ𝑡ሻ ሺ7ሻ
𝑥̇ 3 ሺ𝑡ሻ 0 2 −3 𝑥3 ሺ𝑡ሻ 0

𝑥1 ሺ𝑡ሻ
𝑦ሺ𝑡ሻ = [1 0 0] [𝑥2 ሺ𝑡ሻ] ሺ8ሻ
𝑥3 ሺ𝑡ሻ
From the state-space model given in equations (7) and (8), we have that;
0 0 1 0
𝐴 = [−2 −3 0 ], 𝐵 = [2], and 𝐶 = [1 0 0]
0 2 −3 0

Check for Controllability

The order of matrix 𝐴 is 3, therefore, the controllability matrix 𝐐𝐌 is given by
𝐐𝐌 = [𝐁: 𝐀𝐁: 𝐀𝟐 𝐁 ]
Hence,
0
𝐵 = [2]
0
0 0 1 0 0
𝐴𝐵 = [−2 −3 0 ] × [2] = [−6]
0 2 −3 0 4
0 0 1 0 4
𝐴2 𝐵 = 𝐴ሺ𝐴𝐵ሻ = [−2 −3 0 ] × [ −6] = [ 18 ]
0 2 −3 4 −24
Therefore, the controllability matrix 𝐐𝐌 is
45
 0 0 4
𝐐𝐌 = [2 −6 18 ]
0 4 −24
The rank of the controllability matrix 𝐐𝐌 is 3, because;
0 0 4
|𝐐𝐌 | = |2 −6 18 | = 32 ≠ 0
0 4 −24
Since the rank of the controllability matrix and the order of matrix 𝐴 are same, the given
system is completely controllable.

Check for Observability

The order of matrix 𝐴 is 3, therefore, the observability matrix 𝐐𝑶 is given by
𝟐
𝐐𝑶 = [𝐂 𝐓 : 𝐀𝐓 𝐂 𝐓 : (𝐀𝐓 ) 𝐂𝐓 ]

Hence,
1
𝐶 𝑇 = [0]
0
0 −2 0 1 0
𝐴𝑇 𝐶 𝑇 = [0 −3 2 ] [0] = [0]
1 0 −3 0 1
0 −2 0 0 0
ሺ𝐴𝑇 ሻ2 𝐶 𝑇 = 𝐴𝑇 [𝐴𝑇 𝐶 𝑇 ] = [0 −3 2 ] [0] = [ 2 ]
1 0 −3 1 −3
Therefore, the observability matrix 𝐐𝑶 is
1 0 0
𝐐𝑶 = [0 0 2 ]
0 1 −3
The rank of the observability matrix 𝐐𝑶 is 3, because;
1 0 0
|𝐐𝑶 | = |0 0 2 | = −2 ≠ 0
0 1 −3
Since the rank of the observability matrix and the order of matrix 𝐴 are same, the given
system is completely observable.
Example E3.2: Write the state and output equation in matrix form and also test the
controllability and observability of the system characterised by the transfer function

46
 𝑌ሺ𝑠ሻ 2
= 3
𝑈ሺ𝑠ሻ 𝑠 + 6𝑠 2 + 11𝑠 + 6
Solution
Rearranging the transfer function gives
ሺ𝑠 3 + 6𝑠 2 + 11𝑠 + 6ሻ𝑌ሺ𝑠ሻ = 2𝑈ሺ𝑠ሻ
Taking the inverse Laplace transform
𝑦⃛ሺ𝑡ሻ + 6𝑦̈ ሺ𝑡ሻ + 11𝑦̇ ሺ𝑡ሻ = 2𝑢ሺ𝑡ሻ
Let 𝑦ሺ𝑡ሻ = 𝑥1
𝑦̇ ሺ𝑡ሻ = 𝑥̇ 1 = 𝑥2
𝑦̈ ሺ𝑡ሻ = 𝑥̇ 2 = 𝑥3
𝑦⃛ሺ𝑡ሻ = 𝑥̇ 3 = 𝑥4

Therefore 𝑥̇ 1 = 𝑥2
𝑥̇ 2 = 𝑥3
𝑥̇ 2 = 2𝑢ሺ𝑡ሻ − 6𝑥3 − 11𝑥2 − 6𝑥1
𝑥̇ 1 ሺ𝑡ሻ 0 1 0 𝑥1 ሺ𝑡ሻ 0
𝑥̇ ሺ𝑡ሻ
[ 2 ]=[ 0 0 𝑥 ሺ𝑡ሻ
1 ] [ 2 ] + [0] 𝑢ሺ𝑡ሻ ሺ1ሻ
𝑥̇ 3 ሺ𝑡ሻ −6 −11 −6 𝑥3 ሺ𝑡ሻ 2

𝑥1 ሺ𝑡ሻ
𝑦ሺ𝑡ሻ = [1 0 0] [𝑥2 ሺ𝑡ሻ] ሺ2ሻ
𝑥3 ሺ𝑡ሻ
0 1 0 0
𝐴=[ 0 0 1 ], 𝐵 = [0], and 𝐶 = [1 0 0]
−6 −11 −6 2

Check for Controllability

The order of matrix 𝐴 is 3, therefore, the controllability matrix 𝐐𝐌 is given by
𝐐𝐌 = [𝐁: 𝐀𝐁: 𝐀𝟐 𝐁 ]
Hence,
0
𝐵 = [0]
2
47
 0 1 0 0 0
𝐴𝐵 = [ 0 0 1 ] × [0] = [ 2 ]
−6 −11 −6 2 −12
0 1 0 0 2
2
𝐴 𝐵 = 𝐴ሺ𝐴𝐵ሻ = [ 0 0 1 ] × [ 2 ] = [−12 ]
−6 −11 −6 −12 50
Therefore, the controllability matrix 𝐐𝐌 is
0 0 2
𝐐𝐌 = [0 2 −12 ]
2 −12 50
The rank of the controllability matrix 𝐐𝐌 is 3, because;
0 0 2
|𝐐𝐌 | = |0 2 −12 | ≠ 0
2 −12 50
Since the rank of the controllability matrix and the order of matrix 𝐴 are same, the given
system is completely controllable.

Check for Observability

The order of matrix 𝐴 is 3, therefore, the observability matrix 𝐐𝑶 is given by
𝟐
𝐐𝑶 = [𝐂 𝐓 : 𝐀𝐓 𝐂 𝐓 : (𝐀𝐓 ) 𝐂𝐓 ]

Hence,
1
𝐶 𝑇 = [0]
0
0 0 −6 1 0
𝐴𝑇 𝐶 𝑇 = [1 0 −11 ] [0] = [1]
0 1 −6 0 0
0 0 −6 0 0
ሺ𝐴𝑇 ሻ2 𝐶 𝑇 = 𝐴𝑇 [𝐴𝑇 𝐶 𝑇 ] = [1 0 −11 ] [1] = [0]
0 1 −6 0 1
Therefore, the observability matrix 𝐐𝑶 is
1 0 0
𝐐𝑶 = [0 1 0]
0 0 1
The rank of the observability matrix 𝐐𝑶 is 3, because;

48
 1 0 0
|𝐐𝑶 | = |0 1 0| ≠ 0
0 0 1
Since the rank of the observability matrix and the order of matrix 𝐴 are same, the given
system is completely observable.

EXERCISES

E3.1 Consider
−3 1 1
𝐱̇ = [ ]𝐱 + [ ]𝒖
−2 0 −2
𝑦 = [1 0]𝑥

Find the feedback gain 𝐤 in 𝑢 = 𝑟 − 𝐤𝐱 such that the resulting system has eigenvalues at
−2 ± 2𝑗.
E3.2 Design the phase-variable feedback gains for the given open-loop plants (a), (b), and (c) to yield
a closed-loop response of 15%, 10%, 15% overshoot, and settling time of 0.75 second, 2 seconds, and
0.2 second respectively. Place the third pole 10 times, 15 times, and 20 times respectively as far from
the imaginary axis as the dominant pole pair. Use the phase variables for state-variable feedback.
20
(a) 𝐺ሺ𝑠ሻ = ሺ𝑠+2ሻሺ𝑠+4ሻሺ𝑠+8ሻ
20ሺ𝑠+2ሻ
(b) 𝐺ሺ𝑠ሻ = 𝑠ሺ𝑠+5ሻሺ𝑠+7ሻ
ሺ𝑠+2ሻ
(c) 𝐺ሺ𝑠ሻ = ሺ𝑠+5ሻሺ𝑠+9ሻ

E3.3 Given the plant below, design an integral controller to yield a 10% over-shoot, 0.5
second settling time, and zero steady-state error for a step input.
−1 1 0
𝐱̇ = [ ]𝑥 + [ ]𝑢
0 2 1
𝑦 = [1 1]𝑥

E3.4 Write the state and output equation in matrix form and also test the controllability
and observability of the system characterised by the transfer function
𝑌ሺ𝑠ሻ 𝑠+6
= 2
𝑈ሺ𝑠ሻ 𝑠 + 5𝑠 + 6

49
E3.5 For the state-space equation of a system containing 𝐴, 𝐵, and 𝐶 given below, check
if the system is completely state controllable, completely observable and completely output
controllable.
2 0 0 0 1
1 0 0
𝐴 = [0 2 0], 𝐵 = [1 0], and 𝐶 = [ ]
0 1 0
0 3 1 0 1
E3.6 Check the controllability and observability of the following state-variable equation;
−3 1 0 2
𝐱̇ = [−2 0 0] 𝑥 + [4] 𝑢
1 0 0 1
𝑦 = [1 0 0]𝑥

50
 CHAPTER FOUR

PID CONTROL

4.1 Introduction
This chapter presents an introduction to the 3-term controller known as PID Controller. The
PID is an acronym for Proportional-Integral-Derivative that make up the standard three-term
controller. PID is properly the most common and widely used control algorithm by most
industrial feedback system loops. PID and its variations has had a long history of
implemented in many different forms, such as a stand-alone controller or as a part of a Direct
Digital Control (DDC) or a hierarchical distributed process control system. The three-term
PID controller has survived the advancement of technology from analogue era of pneumatics
to electronic tubes, transistors, and integrated circuits, then to the digital computer
(microprocessor) control system era. The older PID controller needs to be tuned for optimum
performance, while the newer PID controllers implemented on microprocessor have been
given opportunities to achieve more features like automatic tunning, gain scheduling, and
continuous adaptation. Automatic tunning of a PID controller is the ability of that controller
to be tunned automatically on demand from an operator or an external signal. Adaptation is
when the parameters of a controller are continuously updated. In spite the well-established
practice of installing, tunning, and using the controllers, there are more potential for
improving PID control. It is found in practice that majority of the PID controllers are
operated in manual mode. But for some PID controllers that are operated in automatic mode,
the derivative action is often shut down because the derivative action is problematic when
trying to achieve proper tunning. However, the assurance for improving the performance of
the control loop lies on the indebt knowledge and understanding the specific process and that
of the controller by the control engineer.
This chapter presents an introduction to PID control, the basic algorithm and various
representation of the controller. The chapter concluded with the use and misuse of PID
control with illustrations of systems where PID control excels and where it does not excel.

51