Dynamical System Model

CHAPTER
1
Dynamical Systems
and Modeling
To catch a mouse alive, try this trick: When you see the beast
running around a room, open a black umbrella and hold it on
the oor at an angle that allows the mouse to run in (it helps
if someone else can chase the mouse). Your prey, drawn to
dark places, will probably head straight for the umbrella. Then
quickly close it, go outside, and set your prisoner free.
Readers Digest Practical Problem Solver [241]
1.1 What Is a System?
Asystemis a collectionof interactingcomponents. Anelectric motor, anairplane, anda biological
unit such as the human armare examples of systems. Asystemis characterized by two properties,
which are as follows:
1. The interrelations between the components that are contained within the system
2. The system boundaries that separate the components within the system from the compo-
nents outside
The systemboundaries can be real or imagined. They are elastic in the sense that we may choose,
at any stage of the systemanalysis, to consider only a part of the original systemas a systemon its
own. We call it a subsystem of the original system. On the other hand, we may decide to expand
the boundaries of the original system to include new components. In Figure 1.1, we represent
a systems boundaries using a box. The interactions between the system components may be
governed, for example, by physical, biological, or economical laws. In dealing with systems, we
are interested in the effects of external quantities upon the behavior of the system quantities. We
refer to the external quantities acting on the system as the inputs to the system. The condition
or the state of the system is described by the state variables. The state variables provide the
information that, together with the knowledge of the system inputs, enables us to determine the
future state of the system. A dynamical system consists of a set of possible states, together with
a rule that determines the present state in terms of past states [7, p. 1]. An axiomatic description
of a dynamical system is presented in Section 1.3. In practice it is often not possible or too
expensive to measure or determine the values of all of the state variables. Instead, only their
subset or combination can be measured. The system quantities whose behavior can be measured
or observed are referred to as the system outputs.
1
2 CHAPTER 1

DYNAMICAL SYSTEMS AND MODELING
System
Input Output
Figure 1.1 Representation of a system.
In engineering applications, when dealing with dynamical systems, we are interested in
specifying the system inputs that force the system states or outputs to behave with time in some
prespecied manner. That is, we are interested in controlling the systemstates or outputs. This is
accomplished by means of a controller whose task is to produce the required systems inputs that
in turn result in the desired systems outputs. An interconnection of the system and a controller
is called a control system. In Figures 1.2 and 1.3, we show two different types of control systems
that we discuss in the next section. Constructing a controller is a part of the control problem. The
essential elements of the control problem, as described by Owens [218, p. 181] are as follows:
1. A specied objective for the system
2. A model of a dynamical system to be controlled
3. A set of admissible controllers
4. A means of measuring the performance of any given control strategy to evaluate its effec-
tiveness
We now examine these elements one by one. The objective of a control system is to complete
some specied task. This can be expressed as a combination of (a) constraints on the output
or state variables and (b) limits on the time available to complete the control objective. For
example, the objective of a controller might be to force the system output to settle within a
certain percentage of the given value after prespecied time.
We now discuss the modeling part of the control problem. The rst step in the controller
design procedure is the construction of a truth model of the dynamics of the process to be
controlled. The truth model is a simulation model that includes all the relevant characteristics of
the process. The truth model is too complicated for use in the controller design. Thus, we need
to develop a simplied model that can be used to design a controller. Such a simplied model
is labeled by Friedland [91] as the design model. The design model should capture the essential
features of the process. A common model of a dynamical system is the nite set of ordinary
differential equations of the form
x(t ) = f (t, x(t ), u(t )), x(t
0
) = x
0
,
y(t ) = h(t, x(t ), u(t )),
where the state x R
n
, the input u R
m
, the output y R
p
, and f and h are vector-valued
functions with f : RR
n
R
m
R
n
and h : RR
n
R
m
R
p
. Another common model
of a dynamical system is the nite set of difference equations,
x(k +1) = f (k, x(k), u(k)), x(k
0
) = x
0
,
y(k) = h(k, x(k), u(k)),
where x(k) = x(kh), u(k) = u(kh), h is the sampling interval, and k 0 is an integer.
In our subsequent discussion, we regard the dynamical system and its model as equivalent.
During the modeling process, properties of the system, physical constraints, and technical re-
quirements should be taken into account. By itself the pair ( f (t, x, u), h(t, x, u)) has little
1.2 OPEN-LOOP VERSUS CLOSED-LOOP 3
engineering signicance until constraints on the state x, the control u, and the output y are
also given. Control input signals are obtained from physical devices capable of providing only
a limited amount of energy. We refer to the class of controllers that can be considered for the
given control design problem as the set of admissible controllers.
We can evaluate the performance of any given control lawby visual inspection of the transient
characteristics of the controlled system after the design is complete. But this method of assess-
ment of the effectiveness of the given control strategy is inaccurate. We therefore try to quantify
the means of evaluating performance in the form of a performance index or cost functional at
the beginning of the design process. We construct a numerical index of performance whose
value reects the quality of any admissible controller in accomplishing the system objective.
This performance index is assumed to have the property that its numerical value decreases as the
quality of the controller increases. This is the reason why we sometimes refer to the performance
index as the penalty functional. The best controller then is the one that produces the smallest
value of the performance index. We call an admissible controller that simultaneously ensures the
completion of the system objective and the minimization of the performance index an optimal
controller.
1.2 Open-Loop Versus Closed-Loop
We distinguish between two types of control systems. They are:
Open-loop control systems
Closed-loop control systems

An open-loop control system usually contains the following:
1. A process to be controlled, labeled plant
2. The controlling variable of the plant, called the plant input, or just input for short
3. The controlled variable of the plant, called the plant output, or just output for short
4. A reference input, which dictates the desired value of the output
5. A controller that acts upon the reference input in order to form the system input forcing
the output behavior in accordance with the reference signal
Note that the plant and controller themselves can be considered as systems on their own. A
schematic representation of an open-loop system is depicted in Figure 1.2. In an open-loop
control systemthe output has noinuence onthe input or reference signal. The controller operates
without taking into account the output. Thus, the plant input is formed with no inuence of the
output. A household appliance such as an iron is a simple example of an open-loop control
system. In this example, we consider the iron itself as the plant or process to be controlled.
Here we set the temperature control mechanism for the desired temperature. This settingfor
example, very hotis the reference input. We can view the potentiometer, inside the iron, as a
controller, or a control actuator. It regulates the temperature of the iron by allowing the necessary
amount of resistance to produce the desired temperature. The plants output is the temperature
of the heated iron. Not every control system is an open-loop system. Another type of a control
system is a closed-loop system. We can convert an open-loop system into a closed-loop system
by adding the following components:
4 CHAPTER 1

6. The feedback loop where the output signal is measured with a sensor and then the measured
signal is fed back to the summing junction
7. The summing junction, where the measured output signal is subtracted from the reference
(command) input signal in order to generate an error signal, also labeled as an actuating
signal
A schematic of a closed-loop control system is shown in Figure 1.3. In a closed-loop system the
error signal causes an appropriate action of the controller, which in turn instructs the plant to
behave in a certain way in order to approach the desired output, as specied by the reference
input signal. Thus, in the closed-loop system, the plant output information is fed back to the
controller, and the controller then appropriately modies the plant output behavior. Acontroller,
also called a compensator, can be placed either in the forward loop, as in Figure 1.3, or in the
feedback loop.
If the goal of the controller is to maintain the output at a constant value, then we have a
regulator control system. A centrifugal speed regulator, or the yball governor, commonly
known as Watts governor, is an example of a regulator in a closed-loop system. A schematic of
Watts governor is shown in Figure 1.4. The task of the centrifugal speed regulator is to control
the setting of a throttle valve to maintain the desired engine speed. The nominal speed is set by
adjusting the valve in the throttle. As the engine speeds up, the weights are thrown outwards,
the throttle partially closes, and the engine slows down. As the engine slows down below its
nominal speed, the throttle is opened up and the engine gains its speed. The operation of an
engine equipped with a centrifugal speed regulator can be represented using the block diagram
of Figure 1.3. The reference input to the closed-loop system is the desired engine speed. The
desired engine speed is obtained by appropriately setting the valve. The system output is the
actual engine speed. The yball governor is our controller, while the engine represents the plant.
The control signal generated by the controller is the setting of the throttle valve.
We mention at this juncture that Watts application of the centrifugal speed governor in
1788, to regulate the speed of the steam engine, marks the starting point for the development
of automatic controland, in particular, feedback controlas a science [192]. Before 1788,
Controller Plant
Reference
input
Desired
output
Actual
output
Controller
output
Plant
input
Figure 1.2 Open-loop control system.
Controller
Sensor
Plant
Actuating
signal
Desired
output
Reference
input
Actual
output
Controller
output
Measured output
Plant
input
Figure 1.3 Closed-loop control system.
1.3 AXIOMATIC DEFINITION OF A DYNAMICAL SYSTEM 5
Collar
L
m
N
L
gov
Weights
m
Flywheel
To engine
close
open
Valve
Throttle
Figure 1.4 A simplied schematic of the yball governor.
steam engines were controlled manually using the throttle valve. The yball governor spread
widely and quickly. It has been estimated that by 1868 there were about 75,000 yball governors
working in England alone [192]. The yball governor is credited for the hatching of the industrial
revolution. We devise a mathematical model of this revolutionary invention in Subsection 1.7.1,
and in Section 4.7 we perform a quantitative analysis of its dynamical behavior.
If the goal of the controller is to force the output to follow a desired trajectory, then we have
a servomechanism or tracking control system. As we mentioned before, to design a controller,
we need rst to analyze the plant quantitatively. The analysis requires a mathematical, or lin-
guistic, description of the interrelations between the system quantities themselves as well as
the interrelations between system quantities and system inputs. The mathematical description
of these interrelations is called mathematical modeling. We devote Section 1.4 to mathematical
modeling. Constructing models using linguistic description of the process to be controlled will
be discussed in Subsection 8.5.1.
1.3 Axiomatic Denition of a Dynamical System
Following Kalman [147], we can dene a dynamical system formally using the following
axioms:
1. There is given a state space X; there is also T, an interval in the real line representing time.
2. There is given a space U of functions on T that represent the inputs to the system.
3. For any initial time t
0
in T, any initial state x
0
in X, and any input u in U dened for
t t
0
, the future states of the system are determined by the transition mapping
: T X U X
written as
(t
1
; t
0
, x(t
0
), u(t )) = x(t
1
).
6 CHAPTER 1

x(t
1
) x(t
2
) x(t
0
)
Figure 1.5 Illustration of the semigroup axiom.
Moreover, the following axioms are satised:
4. The identity property of the transition mapping, that is,
(t
0
; t
0
, x(t
0
), u(t
0
)) = x(t
0
).
5. The semigroup property of the transition mapping, that is,
(t
2
; t
0
, x(t
0
), u(t )) = (t
2
; t
1
, (t
1
; t
0
, x(t
0
), u(t )), u(t )).
The semigroup property axiom states that it is irrelevant whether the system arrives at
the state at time t
2
by a direct transition from the state at time t
0
, or by rst going to an
intermediate state at time t
1
, and then having been restarted from the state at time t
1
and
moving to the state at time t
2
. In either case the system, satisfying the semigroup axiom,
will arrive at the same state at time t
2
. This property is illustrated in Figure 1.5.
6. The causality property, that is,
(t ; t
0
, x(t
0
), u
1
(t )) = (t ; t
0
, x(t
0
), u
2
(t )) for t
0
, t T
if and only if u
1
(t ) = u
2
(t ) for all t T.
7. Every output of the system is a function of the form
h : T X U Y,
where Y is the output space.
8. The transition mapping and the output mapping h are continuous functions.
9. If in addition the system satises the following axiom, then it is said to be time-invariant:
(t
1
; t
0
, x(t
0
), u(t )) = (t
1
+; t
0
+, x(t
0
), u(t )),
where t
0
, t
1
T.
Thus, a dynamical system can be dened formally as a quintuple
{T, X, U, , Y}
satisfying the above axioms.
Our attention will be focused on dynamical systems modeled by a set of ordinary differential
equations
x
i
= f
i
(t, x
1
, x
2
, . . . , x
n
, u
1
, u
2
, . . . , u
m
), x
i
(t
0
) = x
i 0
, i = 1, 2, . . . , n,
together with p functions
y
j
= h
j
(t, x
1
, x
2
, . . . , x
n
, u
1
, u
2
, . . . , u
m
), j = 1, 2, . . . , p.
The systemmodel state is x = [x
1
x
2
x
n
]
T
R
n
. The systeminput is u = [u
1
u
2
u
m
]
T
R
m
, and the system output is y = [y
1
y
2
y
p
]
T
R
p
. In vector notation the above system
1.3 AXIOMATIC DEFINITION OF A DYNAMICAL SYSTEM 7
model has the form
x = f (t, x, u), x(t
0
) = x
0
,
y = h(t, x, u),
where f : R R
n
R
m
R
n
and h : R R
n
R
m
R
p
are vector-valued functions.
In further considerations, we regard a dynamical system under considerations and its model
represented by the above equations as equivalent. We now illustrate dynamical system axioms
on a simple example.
Example 1.1
We consider a dynamical systemwith no external inputs modeled by the rst-order
linear differential equation
x( t ) = ax( t ) , x
0
= x( t
0
) , (1.1)
where a is a constant, x X = R, and T = R. The solution to the above differential
equation has the form
x( t ) = e
a( t t
0
)
x( t
0
) .
The above system model can be characterized by the triple
{T, X, },
where
: T X X
is the transition mapping. In this example,
( t ; t
0
, x( t
0
) ) = e
a( t t
0
)
x( t
0
) .
It is clear that is a continuous function. The transition function = e
a( t t
0
)
x( t
0
)
satises the semigroup property. Indeed,
( t
2
; t
1
, ( t
1
; t
0
, x( t
0
) ) ) = e
a( t
2
t
1
)
_
e
a( t
1
t
0
)
x( t
0
)
_
= e
a( t
2
t
0
)
x( t
0
)
= ( t
2
; t
0
, x( t
0
) ) .
The identity property axiom is also satised:
( t
0
; t
0
, x( t
0
) ) = e
a( t
0
t
0
)
x( t
0
)
= x( t
0
) .
Finally, observe that the system model given by (1.1) is time invariant because
( t
1
; t
0
, x( t
0
) ) = e
a( t
1
t
0
)
x( t
0
)
= e
a( t
1
+( t
0
+) )
x( t
0
)
= ( t
1
+; t
0
+, x( t
0
) ) .
8 CHAPTER 1

1.4 Mathematical Modeling
The goal of mathematical modeling is to provide a mathematical description of the interrelations
between the system quantities as well as the relations between these quantities and external
inputs. Amathematical model can be viewed as a mathematical representation of the signicant,
relevant aspects of an existing or proposed physical system. (Signicance and relevance being
in relation to an application where the model is to be used) [181, p. xi]. Another denition of
mathematical model that we found in reference 76, p. 1, is: A mathematical model is a system
of equations that purports to represent some phenomenon in a way that gives insight into the
origins and consequences of its behavior. (For the controller design purposes, we rst construct
a truth model of the given plant and then its design model.) In general, the more accurate the
mathematical model is, the more complex the modeling equations are. This means that during
the mathematical modeling process we may be forced to nd a middle course between accuracy
of the model and its complexity. Our goal is to generate the most accurate possible description
of the given system while still being able to carry out the model analysis and computations in a
reasonable amount of time. At rst, we tend to make oversimplifying assumptions in order to
obtain a very simple model. However, we should follow the principle of maximum simplicity,
which states that everything should be as simple as possiblebut no simpler. After extracting
some information about the system behavior based on this rst, possibly crude model, it may be
desirable to construct a more accurate model. We try to create a mathematical description of a
given physical process that incorporates realistic assumptions and physical constraints. We then
extract information from this model and compare it with the experimental data. We perform
this comparative study to determine the value of the obtained model. If the model appears to
be inadequate for our needs, then it must be adjusted. The process of mathematical modeling
is depicted in Figure 1.6. In the process of mathematical modeling, one can use different types
of equations and their combinations like algebraic, ordinary and partial differential equations,
Evaluate and interpret
Solve equations
Simplify equations
Construct equations
Formulate problem
Is model adequate?
Design controller
No
Yes
Revise
Figure 1.6 Diagram of mathematical modeling process
and its use in the controller design process.
1.5 REVIEW OF WORK AND ENERGY CONCEPTS 9
difference, integral, or functional equations. Inthis book, onlydifference andordinarydifferential
equations are used to model dynamical systems.
In the following, we discuss the Lagrange equations of motion that are useful in mathematical
modeling of dynamical systems. We rst review relevant concepts that are needed to derive the
Lagrange equations. We then construct mathematical models of a few dynamical systems.
1.5 Review of Work and Energy Concepts
The purpose of this section is to review and reacquaint ourselves with some basic notions
from physics: work and energy. After a brief review of work and energy concepts, we derive
the Lagrange equations of motion and use them to devise mathematical models of dynamical
systems.
Suppose we are given a particle of a constant mass m subjected to a force F. Then, by
Newtons second law we have
F = m
dv
dt
= ma. (1.2)
Here the force F and the velocity v are vectors. Therefore, they can be represented as
F =
_
F
x
1
F
x
2
F
x
3
T
and v =
_
v
x
1
v
x
2
v
x
3
T
= [ x
1
x
2
x
3
]
T
.
Using the above notation, we represent equation (1.2) as
_
_
F
x
1
F
x
2
F
x
3
_
_
= m
_
_
dv
x
1
dt
dv
x
2
dt
dv
x
3
dt
_
_
. (1.3)
Suppose now that a force F is acting on a particle located at a point A and the particle moves
to a point B. The work W done by F along innitesimally small distance s is
W = F
T
s
=
_
F
x
1
F
x
2
F
x
3
_
_
x
1
x
2
x
3
_
_
= F
x
1
x
1
+ F
x
2
x
2
+ F
x
3
x
3
.
where s = [x
1
x
2
x
3
]
T
. The work W
AB
done on the path from A to B is obtained by
integrating the above equation:
W
AB
=
_
B
A
_
F
x
1
dx
1
+ F
x
2
dx
2
+ F
x
3
dx
3
_
.
We now would like to establish a relation between work and kinetic energy. For this observe
that
x dx =
d x
dt
dx = x d x.
10 CHAPTER 1

Using the above relation and Newtons second law, we express W
AB
in a different way as
W
AB
=
_
B
A
(m x
1
dx
1
+m x
2
dx
2
+m x
3
dx
3
)
=
_
B
A
m( x
1
d x
1
+ x
2
d x
2
+ x
3
d x
3
)
= m
_
x
2
1
2
+
x
2
2
2
+
x
2
3
2
_
B
A
= m
_
v
B
2
2

v
A
2
2
_
=
mv
B
2
2

mv
A
2
2
, (1.4)
where v
A
= [ x
1A
x
2A
x
3A
]
T
, v
B
is dened in an analogous manner, and v = v
T
v. Observing
that mv
B
2
/2 is the kinetic energy of the particle at the point B and mv
A
2
/2 is its kinetic
energy at the point A, we obtain
W
AB
= K
B
K
A
.
The above relation can be used to dene the kinetic energy of a particle as the work required to
change its velocity fromsome value v
A
to a nal value v
B
. The relation W
AB
= K
B
K
A
= K
is also known as the workenergy theorem for a particle.
A force is conservative if the work done by the force on a particle that moves through any
round trip is zero. In other words, a force is conservative if the work done by it on a particle that
moves between two points depends only on these points and not on the path followed.
We now review the notion of potential energy, or the energy of conguration. Recall that if
the kinetic energy K of a particle changes by K, then the potential energy U must change by
an equal but opposite amount so that the sum of the two changes is zero; that is,
K +U = 0. (1.5)
This is equivalent to saying that any change in the kinetic energy of the particle is compensated
for by an equal but opposite change in the potential energy U of the particle so that their sum
remains constant; that is,
K +U = constant.
See Exercise 1.1 for a rigorous proof of the above relation. The potential energy of a particle
represents a form of stored energy that can be recovered and converted into the kinetic energy.
If we now use the workenergy theorem and equation (1.5), then we obtain
W = K = U.
The work done by a conservative force depends only on the starting and the end points of motion
and not on the path followed between them. Therefore, for motion in one dimension, we obtain
U = W =
_
x
x
0
F(s) ds. (1.6)
1.5 REVIEW OF WORK AND ENERGY CONCEPTS 11
We thus can write
F(x) =
dU(x)
dx
. (1.7)
A generalization of equation (1.6) to motion in three dimensions yields
U =
_
x
1
x
10
F
x
1
ds
_
x
2
x
20
F
x
2
ds
_
x
3
x
30
F
x
3
ds,
and a generalization of equation (1.7) to motion in three dimensions is
F(x) =
_
U
x
1
U
x
2
U
x
3
_
T
= U(x), (1.8)
where x = [x
1
x
2
x
3
]
T
.
For a formal proof of the property of a conservative vector eld that the work done by it
on a particle that moves between two points depends only on these points and not on the path
followed, see Exercise 1.2.
We now represent Newtons equation in an equivalent format that establishes a connection
with the Lagrange equations of motion, which are discussed in the following section. To proceed,
note that
K =
m x
2
2
=
m x
T
x
2
. (1.9)
Hence,
K
x
i
= m x
i
, i = 1, 2, 3. (1.10)
Combining (1.8) and (1.10), we represent Newtons equations as
d
dt
_
K
x
i
_
+(U)
i
= 0, i = 1, 2, 3. (1.11)
Let
L = K U. (1.12)
The function L dened above is called the Lagrangian function or just the Lagrangian. Note
that
L
x
i
=
K
x
i
, (1.13)
and
L
x
i
=
U
x
i
. (1.14)
Substituting (1.13) and (1.14) into (1.11), we obtain
d
dt
_
L
x
i
_
L
x
i
= 0, i = 1, 2, 3. (1.15)
Equations (1.15) are called the Lagrange equations of motion, in Cartesian coordinates, for a
single particle. They are just an equivalent representation of Newtons equations given by (1.3).
12 CHAPTER 1

1.6 The Lagrange Equations of Motion
Asingle particle with no constraints is free to take up any position in the three-dimensional space.
This is the reason we need three coordinates to describe its position. In the case of a system of
p particles, there are 3p degrees of freedom. In general, if there are constraints imposed on the
particles, then the number of the degrees of freedom is
n = 3p number of constraints.
For example, if the systemconsists of only one particle in space and if the particle is restricted to
move along a line, then n = 32 = 1. We can use many different coordinate systems to describe
a location of a particle in space. We will use the letter q as a symbol for a coordinate regardless of
its nature. Ageneralized coordinate may represent, for example, angular displacement or electric
charge. This indicates that the results we are going to obtain will be of general nature and will
not be restricted to mechanical systems. We refer to q as a generalized coordinate. Thus, if
x = x(q
1
, q
2
, q
3
),
y = y(q
1
, q
2
, q
3
),
z = z(q
1
, q
2
, q
3
),
then
x =
x
q
1
q
1
+
x
q
2
q
2
+
x
q
3
q
3
,
and so on. We assume that the functions x, y, z C
2
; that is, they are twice continuously differen-
tiable, which implies that their second partial derivatives are equal. Suppose now that a particle
is conned to move on a surface. Therefore, to describe the location of the particle, we need only
two generalized coordinates. The position of the particle on a surface in a three-dimensional
space is then described as
x = x(q
1
, q
2
),
y = y(q
1
, q
2
),
z = z(q
1
, q
2
),
and hence
x =
x
q
1
q
1
+
x
q
2
q
2
,
y =
y
q
1
q
1
+
y
q
2
q
2
, (1.16)
z =
z
q
1
q
1
+
z
q
2
q
2
.
We know that the work W done by the force F is W = F
x
x + F
y
y + F
z
z. On the other
hand, this work is equal to the change in the kinetic energy. Therefore,
m( x x + y y + z z) = F
x
x + F
y
y + F
z
z (1.17)
The above equation is known as DAlemberts equation.
1.6 THE LAGRANGE EQUATIONS OF MOTION 13
If we have a system of particles acted upon by forces F
1
, . . . , F
p
and if each of the particles
undergoes displacements s
1
, . . . , s
p
, then the total work done by the forces is
W =
p
i =1
_
F
x
i
x
i
+ F
y
i
y
i
+ F
z
i
z
i
_
. (1.18)
Substituting (1.16) into DAlemberts equation and rearranging, we obtain
W = m
_
x
x
q
1
+ y
y
q
1
+ z
z
q
1
_
q
1
+m
_
x
x
q
2
+ y
y
q
2
+ z
z
q
2
_
q
2
=
_
F
x
x
q
1
+ F
y
y
q
1
+ F
z
z
q
1
_
q
1
+
_
F
x
x
q
2
+ F
y
y
q
2
+ F
z
z
q
2
_
q
2
. (1.19)
Let
F
q
1
= F
x
x
q
1
+ F
y
y
q
1
+ F
z
z
q
1
and F
q
2
= F
x
x
q
2
+ F
y
y
q
2
+ F
z
z
q
2
.
Then, (1.19) can be represented as
W = F
q
1
q
1
+ F
q
2
q
2
.
To proceed further, note that
d
dt
_
x
x
q
1
_
= x
x
q
1
+ x
d
dt
_
x
q
1
_
.
Hence
x
x
q
1
=
d
dt
_
x
x
q
1
_
x
d
dt
_
x
q
1
_
(1.20)
The time derivative of x = x(q
1
, q
2
) is
x =
x
q
1
q
1
+
x
q
2
q
2
. (1.21)
Differentiating the above partially with respect to q
1
yields
x
q
1
=
x
q
1
(1.22)
Finally, we will show that
d
dt
_
x
q
1
_
=
x
q
1
(1.23)
14 CHAPTER 1

Indeed, because x = x(q
1
, q
2
) and the partial derivative x/q
1
is in general a function of q
1
and q
2
, we have
d
dt
_
x
q
1
_
=

q
1
_
x
q
1
_
q
1
+

q
2
_
x
q
1
_
q
2
. (1.24)
Taking the partial derivative of (1.21) with respect to q
1
yields
x
q
1
=

q
1
_
x
q
1
_
q
1
+

q
1
_
x
q
2
_
q
2
=

q
1
_
x
q
1
_
q
1
+

q
2
_
x
q
1
_
q
2
, (1.25)
because by assumption x, y, z C
2
and hence
q
1
_
x
q
2
_
=

q
2
_
x
q
1
_
.
With the above in mind, comparing the right-hand sides of (1.24) and (1.25), we obtain (1.23).
We now use (1.20), (1.22), and (1.23) to arrive at the Lagrange equation of motion for a single
particle. We rst substitute (1.22) and (1.23) into (1.20) to get
x
x
q
1
=
d
dt
_
x
x
q
1
_
x
x
q
1
. (1.26)
Note that
q
1
_
x
2
2
_
= x
x
q
1
(1.27)
and
q
1
_
x
2
2
_
= x
x
q
1
. (1.28)
Substituting (1.27) and (1.28) into (1.26) gives
x
x
q
1
=
d
dt
_
( x
2
/2)
q
1
_
( x
2
/2)
q
1
.
We can obtain similar expressions for y and z. Taking this into account, when q
2
= 0, equa-
tion (1.19) becomes
W =
_
d
dt
_

q
1
m( x
2
+ y
2
+ z
2
)
2
_

q
1
m( x
2
+ y
2
+ z
2
)
2
_
q
1
=
_
F
x
x
q
1
+ F
y
y
q
1
+ F
z
z
q
1
_
q
1
= F
q
1
q
1
. (1.29)
Let K =
1
2
m( x
2
+ y
2
+ z
2
) denote the kinetic energy of the particle. Then, we can repre-
sent (1.29) as
d
dt
_
K
q
1
_
K
q
1
= F
q
1
(1.30)
The equation above is called the Lagrange equation of motion for the q
1
coordinate. Using the
same arguments as above, we can derive the Lagrange equation of motion for the q
2
coordinate.
In general there are as many the Lagrange equations of motion as there are degrees of freedom
of the particle.
Recall that the expression W = F
x
x + F
y
y + F
z
z is a work done by a force F =
[F
x
F
y
F
z
]
T
for a general displacement s =[x y z]
T
. In terms of generalized coordinates
and when, say, q
2
= 0, we have
W =
_
F
x
x
q
1
+ F
y
y
q
1
+ F
z
z
q
1
_
q
1
= F
q
1
q
1
= W
q
1
.
A generalized force F
q
r
is of such nature that the product F
q
r
q
r
is the work done by driving
forces when q
r
alone is changed by q
r
. We mention that a generalized force does not have to
be a force in the usual sense. For example, if q
r
is an angle, then F
q
r
must be a torque in order
that F
q
r
q
r
be work.
We now derive a more general version of the Lagrange equations of motion for a single
particle. In our derivation we will need the notion of nonconservative forces, which we discuss
next. First recall that we interpret the kinetic energy of a particle as its ability to do work by
virtue of its motion. This motion is the result of conservative and nonconservative forces acting
upon the particle. A force is nonconservative if the work done by the force on a particle that
moves through any round trip is not zero. Thus a force is nonconservative if the work done by
that force on a particle that moves between two points depends on the path taken between these
points.
Let us now suppose that in addition to the conservative forces, a single nonconservative force
due to friction acts on the particle. Denote by
W
c
the sumof the work done by the conservative
forces, and denote by W
f
the work done by friction. We then have
W
f
+
W
c
= K. (1.31)
Thus, if a nonconservative force, like friction, acts on a particle, then the total mechanical energy
is not constant, but changes by the amount of work done by a nonconservative force according
to the relation (1.31). Let us now denote by E the nal mechanical energy of the particle and
denote by E
0
its initial mechanical energy. Then, we can write
E E
0
= E = W
f
.
The work done by friction, the nonconservative force, is always negative. Hence, the nal
mechanical energy E is less than the initial mechanical energy. The lost mechanical energy is
transformed into internal energy U
int
. In other words, the loss of mechanical energy equals the
16 CHAPTER 1

gain in internal energy and we write
E +U
int
= 0.
In general, we have
K +
U +U
int
+(change in other forms of energy) = 0.
This means that the total energy, kinetic plus potential plus internal plus all other forms of
energy, does not change. This is the principle of the conservation of energy. To proceed further,
recall that any generalized force F
q
i
acting on our particle can be expressed as
F
q
i
= F
x
x
q
i
+ F
y
y
q
i
+ F
z
z
q
i
.
Assuming that the forces are conservative, we get
F
q
i
=
_
U
x
x
q
i
+
U
y
y
q
i
+
U
z
z
q
i
_
.
We thus have
F
q
i
=
U
q
i
.
Therefore, the Lagrange equations of motion for a single particle for conservative forces can be
represented as
d
dt
_
K
q
i
_
K
q
i
=
U
q
i
.
We can rearrange the above equations as
d
dt
_
K
q
i
_

q
i
(K U) = 0.
We now introduce the Lagrangian function L, dened as
L = K U.
Note that because U is not a function of q
i
, we have
K
q
i
=
L
q
i
.
Taking the above into account, we can represent the Lagrange equations of motion for conser-
vative forces as
d
dt
_
L
q
i
_
L
q
i
= 0.
If some of the forces acting on the particle are nonconservative, the Lagrange equations of
motion for our particle will take the form
d
dt
_
L
q
i
_
L
q
i
=

F
q
i
(1.32)
where

F
q
i
are nonconservative forces acting on the particle.
Example 1.2
Consider the simple pendulum shown in Figure 1.7. The simple pendulum is an
idealized body consisting of a point mass M, suspended by a weightless inexten-
sible cord of length l. The simple pendulum is an example of a one-degree-of-
freedom system with a generalized coordinate being the angular displacement .
The pendulum kinetic energy is
K =
1
2
Ml
2
2
.
We assume that the potential energy of the pendulum is zero when the pendulum
is at rest. Hence, its potential energy is
U = Mgl( 1 cos ) ,
where g = 10 m/sec
2
is the acceleration due to gravity. The Lagrangian function
is
L = K U
=
1
2
Ml
2
2
Mgl( 1 cos )
= 6
2
60( 1 cos ) .
The Lagrange equation describing the pendulum motion is
d
dt
_
L
= 0.
l 2 m
M 3 kg
Figure 1.7 A simple pendulum.
18 CHAPTER 1

In our example the above equation evaluates to
12
+60sin = 0,
or, equivalently,
= 5sin.
Example 1.3
In Figure 1.8, the r robot manipulator is shown. This example was adapted
from Snyder [266, Chapter 10]. A lumped mass representation of the r robot
manipulator is shown in Figure 1.9. The mass m
1
=10 kg represents the mass of
the outer cylinder positioned at its center of mass. The constant distance r
1
=1 m
designates the xed distance between the center of the mass of the outer cylinder
and the center of rotation. The mass of the load is represented by m
2
=3 kg and is
assumed to be located at the end of a piston of a telescoping arm that is a variable
radial distance r from the hub or center of rotation. The angle of rotation of the
manipulator arm is . The inputs to the system are assumed to be (a) a torque T
Figure 1.8 The r robot manipulator.

y
r
x
g
m
2
m
1
r
1
Figure 1.9 A lumped mass

representation of the r robot
manipulator.
applied at the hub in the direction of and (b) a translational force F
r
applied in the
direction r . Finally, g = 10 m/sec
2
is the gravitational constant. We construct the
Lagrange equations of motion for the system, and then we represent the obtained
model in a state-space format. We rst nd the total kinetic energy of the system.
The position of the mass m
1
is
x
1
= r
1
cos ,
y
1
= r
1
sin.
Note that r
1
is a constant. Differentiating x
1
and y
1
with respect to time yields
x
1
= r
1
sin,
y
1
= r
1
cos .
The magnitude squared of the velocity vector of the mass m
1
is
v
2
1
= x
2
1
+ y
2
1
= r
2
1

2
sin
2
+r
2
1

2
cos
2
= r
2
1

2
.
Therefore, the kinetic energy of the mass m
1
is
K
1
=
1
2
m
1
v
2
1
=
1
2
m
1
r
2
1

2
.
We now derive an expression for the kinetic energy of the second mass. The
position of the mass m
2
is
x
2
= r cos ,
y
2
= r sin.
Note that the distance r is not a constant. Differentiating x
2
and y
2
with respect
to time yields
x
2
= r cos r

sin,
y
2
= r sin +r

cos .
The magnitude squared of the velocity vector of the mass m
2
is
v
2
2
= ( x
2
)
2
+( y
2
)
2
= ( r cos r

sin)
2
+( r sin +r

cos )
2
= r
2
+r
2
2
.
Hence, the kinetic energy of the mass m
2
is
K
2
=
1
2
m
2
v
2
2
=
1
2
m
2
( r
2
+r
2
2
) .
The total kinetic energy of the system is
K = K
1
+ K
2
=
1
2
m
1
r
2
1

2
+
1
2
m
2
( r
2
+r
2
2
) .
20 CHAPTER 1

We now nd the potential energy of the system. The potential energy of the mass
m
1
is
U
1
= m
1
gr
1
sin.
The potential energy of the mass m
2
is
U
2
= m
2
gr sin.
The total potential energy of the system is
U = U
1
+U
2
= m
1
gr
1
sin +m
2
gr sin.
The Lagrangian function for the system is
L = K U =
1
2
m
1
r
2
1

2
+
1
2
m
2
r
2
+
1
2
m
2
r
2
2
m
1
gr
1
sin m
2
gr sin.
The manipulator has two degrees of freedom. Hence, we have two Lagrange equa-
tions of motion for this system:
d
dt
_
L
= T
,
d
dt
_
L
r
_
L
r
= F
r
.
The rst Lagrange equation yields
m
1
r
2
1

+m
2
r
2
+2m
2
r r

+ g cos ( m
1
r
1
+m
2
r ) = T
.
The second Lagrange equation gives
m
2
r m
2
r

2
+m
2
g sin = F
r
.
We next represent the above modeling equations in state-space format. (One
can use, for example, MATLABs Symbolic Math Toolbox to perform needed
calculations.) Let the state variables be dened as
z
1
= , z
2
=

, z
3
= r , z
4
= r .
Then, we can write
_
_
z
1
z
2
z
3
z
4
_
_
=
_
_
z
2
2m
2
z
2
z
3
z
4
g cos z
1
( m
1
r
1
+m
2
z
3
) +u
1
m
1
r
2
1
+m
2
z
2
3
z
4
z
2
2
z
3
g sin z
1
+u
2
/m
2
_
_
, (1.33)
where u
1
= T
and u
2
= F
r
.
In the following section, we develop models of a few more dynamical systems.
1.7 MODELING EXAMPLES 21
1.7 Modeling Examples
1.7.1 Centrifugal Governor
In this subsection, we model the centrifugal governor. In our modeling, we follow Pontryagin
[235, pp. 213220]. For an alternative approach to modeling the governor see Maxwell [197] and
Fuller [92]. Somefor example, MacFarlane [192, p. 250]claim that Maxwell founded the
theory of automatic control systems in his 1868 paper On governors. However, this landmark
paper is written in a terse and cryptic style. Fuller [92] does a wonderful job elucidating,
correcting, and expanding Maxwells paper. Mayr [199, Chapter X] gives a fascinating historical
account of the application of the centrifugal governor to the speed regulation of the steamengine
and tracks the origins of the yball governor.
A schematic of the governor is depicted in Figure 1.4, while forces acting on the governor
are shown in Figure 1.10. For simplicity of notation, we assume that the length of the arms, on
which the weights are fastened, are of unity length. If the weights rotate with angular velocity
gov
, then the arms form an angle with the vertical position. Let v
weight
denote the linear
velocity of the weights, and let r be the radius of the circle along which the weights rotate. Then,
v
weight
=
gov
r. Note that r = sin . The resulting centrifugal force acting on each weight is
m
v
2
weight
r
= m
2
gov
r
= m
2
gov
sin .
Simultaneously, a gravitational force mg acts on each weight. In addition, a frictional force in
the hinge joints acts on each weight. We assume that the frictional force is proportional to

L 1 L 1
sin
mg
m
m
g
s
i
n
2
g
o
v
s
i
n

c
o
s
m
2
gov
sin
gov
Figure 1.10 Forces acting on
the weights in the governor.
22 CHAPTER 1

with the coefcient of friction b. Summing forces along the axis perpendicular to an arm yields
an equation of motion of the weight:
m L = m
2
gov
sin cos mg sin b
=
1
2
m
2
gov
sin 2 mg sin b . (1.34)
Note that m L = m because by assumption, L = 1. The governor is connected to the engine
via a gear with the gear ratio N such that
gov
= N, (1.35)
where is the angular velocity of the ywheel. Substituting (1.35) into (1.34) gives
m =
1
2
mN
2
2
sin 2 mg sin b .
Let I be the moment of inertia of the ywheel,
1
the torque delivered by the engine, and
l
the
load torque. Then, the equation modeling the motion of the ywheel is
I =
1

l
. (1.36)
The collar of the governor is connected with the valve so that at the nominal value of =
0
the engine delivers the nominal torque
d
that results in the desired angular velocity =
d
of the ywheel. We model the interaction of the collar and the valve using the equation
1
=
d
+(cos cos
0
), (1.37)
where >0 is a proportionality constant that depends on the type of the governor. Sub-
stituting (1.37) into (1.36) yields
I =
1

l
=
d
+(cos cos
0
)
l
= cos ,
where =
l

d
+ cos
0
. We thus have two equations that model the governor:
m =
1
2
mN
2
2
sin 2 mg sin b ,
I = cos .
We use the state variables
x
1
= , x
2
= , and x
3
=
to represent the above model in state-space format, i.e., as a system of rst-order differential
equations:
x
1
= x
2
,
x
2
=
1
2
N
2
x
2
3
sin 2x
1
g sin x
1

b
m
x
2
, (1.38)
x
3
=

I
cos x
1

I
.
1.7.2 Ground Vehicle
In this subsection, we model a ground vehicle in a turning maneuver, such as a step lane-change
maneuver. Our modeling process follows that of Wong [303] and of Will and

Zak [299]. During
such a step lane-change maneuver, the vehicle undergoes translational as well as rotational mo-
tion. When modeling a moving vehicle, it is convenient to use a reference coordinate frame
attached to and moving with the vehicle. This is because with respect to the vehicle coordinate
frame, moments of inertia of the vehicle are constant. We use the Society of Automotive En-
gineers (SAE) standard coordinate system as shown in Figure 1.11. We use lowercase letters
to label axes of the vehicle coordinate system. Following the SAE convention, we dene the
coordinates as follows:
The x coordinate is directed forward and on the longitudinal plane of symmetry of the
vehicle.
The y coordinate has a lateral direction out the right-hand side of the vehicle.
The z coordinate is directed downward with respect to the vehicle.
The yaw velocity about the z axis is denoted

.
In developing our simple model, we neglect the effects of roll and pitch. The origin of the vehicle
coordinate system is located at the vehicle center of gravity (CG). We also use a global, or xed,
coordinate system. Its axes are labeled using uppercase letters as shown in Figure 1.12. To form
y
z
x
Roll
Yaw
Pitch
Figure 1.11 The Society of
Automotive Engineers (SAE)
coordinate system attached to the
vehicle.
0
x
x
v
x
v
y
y
v
y
v
x
v
x
v v
v
y
v
y
X
Y
at time t
at time t t
Path
0
0
Figure 1.12 Analyzing vehicle
motion using coordinate frame
attached to the vehicle; snapshots of
the vehicle velocity at times t and
t +t .
24 CHAPTER 1

x
v
x
v
x
v
y
v
y
y
0
Figure 1.13 Computing the change in the longitudinal and
lateral components of the vehicle velocity.
the equations of motion of the vehicle, we need an expression for the vehicle CG acceleration.
We rst compute the change in the longitudinal component of the velocity. Using Figures 1.12
and 1.13 and assuming small angles, we obtain
(v
x
+v
x
) cos (v
y
+v
y
) sin v
x
v
x
v
y

= x y .
We form the Newton quotient using the above. Then, taking the limit, we obtain the total
longitudinal acceleration component a
x
:
a
x
= lim
t 0
v
x
v
y

t
= v
x
v
y
= x y
z
, (1.39)
where
z
=

. Using an argument similar to the one above, we obtain the lateral acceleration
component a
y
to be
a
y
= v
y
+v
x

= y + x
z
. (1.40)
We assume that the vehicle is symmetric with respect to the xz plane. We use the so-called
bicycle model that is shown in Figure 1.14, where
f
denotes the front wheels steering angle
and
r
is the rear wheels steering angle. We denote by 2F
x f
lumped longitudinal force at the
front axle and denote by 2F
xr
lumped longitudinal force at the rear axle. The lumped lateral
force at the front axle is denoted 2F
y f
, while lumped lateral force at the rear axle is denoted
2F
yr
. Summing the forces along the x axis and assuming small angles, we obtain
m( x y
z
) = 2F
x f
cos
f
2F
y f
sin
f
+2F
xr
cos
r
+2F
yr
sin
r
2F
x f
+2F
xr
. (1.41)
y
2F
yr
l
2
l
1
f
2F
xr
2F
xf
2F
yf
0
x
X
Y
0
Figure 1.14 Bicycle model of a
vehicle.
Next, summing the forces along the y axis and assuming small angles yields
m( y + x
z
) = 2F
x f
sin
f
+2F
y f
cos
f
+2F
yr
cos
r
2F
xr
sin
r
2F
y f
+2F
yr
. (1.42)
The third equation of motion is obtained by summing the torques with respect to the z axis.
Let I
z
denote the moment of inertia of the vehicle about the z axis. Then, referring again to
Figure 1.14 and assuming small angles, we obtain
I
z

z
= 2l
1
F
y f
cos
f
+2l
1
F
x f
sin
f
2l
2
F
yr
cos
r
+2l
2
F
xr
sin
r
2l
1
F
y f
2l
2
F
yr
. (1.43)
In our further analysis, we assume v
x
= x = constant; that is, the driver is neither accelerating
nor braking the vehicle. This means that we need to use only the last two equations of motion
in our model. To proceed, we need to dene tire slip angles and cornering stiffness coefcients.
The tire slip angle is the difference between the tires heading direction, which is the steer
angle, and the tires actual travel direction. In other words, the tire slip angle is the angle
between the desired direction of motion of the tire specied by the steer angle and the actual
direction of travel of the center of the tire contact. We denote by C
f
the cornering stiffness
coefcient of the front tires and denote by C
r
the cornering stiffness coefcient of the rear tires.
The coefcients C
f
and C
r
are measured experimentally. A typical relation between a tire slip
angle and a cornering force is shown in Figure 1.15. It is a nonlinear relation, where the cornering
coefcient is
C
w
=
dF
yw
d
w
,
where w = f or w = r. The lateral forces F
y f
and F
yr
that appear in (1.41)(1.43) are functions
of the corresponding tire slip angles
f
and
r
and cornering stiffness coefcients C
f
and C
r
.
In our simple linearized model, we assume that C
f
= C
r
= constant, and model dependencies
between the lateral forces and tire slip angles are expressed as
F
y f
= C
f
f
(1.44)
26 CHAPTER 1

0
0
500
1000
1500
2000
2500
3000
0.05 0.1 0.15 0.2 0.25 0.3
Slip (rad)
L
a
t
e
r
a
l

f
o
r
c
e

(
N
)
Figure 1.15 Typical cornering
characteristic.
0
X
Y
r
v
x
r
l
1
z
v
y
l
2
z
v
y
v
y
v
x
Figure 1.16 Computing tire slip angles.
and
F
yr
= C
r
r
. (1.45)
We formulate expressions for the tire slip angles using Figure 1.16. Assuming small angles, we
form an expression for the front tire slip angle:
f
=
f
tan
1
_
l
1
z
+v
y
v
x
_

f

l
1
z
+v
y
v
x
=
f

l
1
z
+ y
x
. (1.46)
The rear tire slip angle is
r
=
r
tan
1
_
l
2
z
v
y
v
x
_
=
r
+tan
1
_
l
2
z
v
y
v
x
_

r
+
l
2
z
v
y
v
x
=
r
+
l
2
z
y
x
. (1.47)
Hence
F
y f
= C
f
_
f

l
1
z
+ y
x
_
, (1.48)
and
F
yr
= C
r
_
r
+
l
2
z
y
x
_
. (1.49)
Substituting the above two equations into (1.42) and assuming small angles, we obtain
m y +m x
z
= 2C
f
f
2C
f
l
1
z
+ y
v
x
+2C
r
r
+2C
r
l
2
z
y
v
x
.
Rearranging terms and performing some manipulations, we get
y =
_
2C
f
+2C
r
mv
x
_
y +
_
v
x

2C
f
l
1
2C
r
l
2
mv
x
_
z
+
2C
f
m

f
+
2C
r
m

r
. (1.50)
Substituting the expressions for F
y f
and F
yr
into (1.43) and assuming small angles, we obtain
I
z

z
= 2l
1
C
f
_
f

l
1
z
+ y
v
x
_
f
2l
2
C
r
_
r
+
l
2
z
y
v
x
_
.
Rearranging and performing manipulations yields

z
=
_
2l
1
C
f
2l
2
C
r
I
z
v
x
_
y
_
2l
2
1
C
f
+2l
2
2
C
r
I
z
v
x
_
z
+
2l
1
C
f
I
z
f

2l
2
C
r
I
z
r
. (1.51)
Later, we will analyze a vehicle performance in the xed coordinates. In particular, we will be
interested in an emergency lane-change maneuver. This requires that we transform the vehicle
lateral variable into the xed coordinates. Assuming small angles and referring to Figure 1.17,
we obtain
Y = x sin y cos
v
x
y. (1.52)
28 CHAPTER 1

0
y
x
Y
X
Heading angle
Figure 1.17 Computing the velocity of a
vehicle with respect to the xed coordinates.
Note that
X = x cos y sin .
For the step lane change, the yaw rate of the vehicle is equal to the yaw rate error; that is,
z
=

.
We further assume that
= . (1.53)
We can now represent the vehicle modeling equations in a state-space format. Let
x
1
= y, x
2
= , x
3
=
z
, x
4
= Y.
Then, using (1.50), (1.51), (1.52), and (1.53), we obtain
_
_
x
1
x
2
x
3
x
4
_
_
=
_
2C
f
+2C
r
mv
x
0 v
x

2C
f
l
1
2C
r
l
2
mv
x
0
0 0 1 0
2l
1
C
f
2l
2
C
r
I
z
v
x
0
2l
2
1
C
f
+2l
2
2
C
r
I
z
v
x
0
1 v
x
0 0
_
_
_
_
x
1
x
2
x
3
x
4
_
_
+
_
_
2C
f
m
2C
r
m
0 0
2l
1
C
f
I
z
2l
2
C
r
I
z
0 0
_
_
_
r
_
. (1.54)
A possible vehicle parameters are given in Table 1.1.
1.7.3 Permanent Magnet Stepper Motor
In this subsection, we develop a model of a permanent magnet (PM) stepper motor. Stepper
motors can be viewed as digital positioning devices. Specically, stepper motors are electrome-
chanical motion devices that convert input information in digital form into output that has
Table 1.1 Typical Vehicle Data
C
f
= C
r
30,000.00 N/rad
l
1
1.20 m
l
2
1.22 m
m 1280 kg
I
z
2500 kgm
2
v
x
18.3 m/sec
Rotor
N
S
N
S
N
S
Coil C
Coil B
Coil A
Coil C
Coil B
Coil A
Stator windings
Figure 1.18 A simplied schematic of an operation of a
stepper motor.
mechanical form. The stepper motor can be driven by digital pulses of the following form: move
clockwise or move counterclockwise. We now describe an idea of an operation of a PM stepper
motor and then derive its modeling equations.
Asimplied organization of a stepper motor is shown in Figure 1.18. The armature, or rotor, is
a magnet. At any instant of the operation, exactly one pair of windings of the stator is energized.
A pair of windings represents a phase. Each pair of the stator windings is wound in such a
way that their magnetic elds are collinearfor example, windings A and A
. The rotor will

then rotate until its magnetic eld is aligned with the elds of the energized pair of the stator
windings. Figure 1.18 shows the situation where the pair of windings A and A
is energized. If
the rotor is at rest, as shown in Figure 1.18, and the pair A, A
is de-energized while the pair B,

B
is energized, the rotor will rotate to the new position so that its magnetic eld is aligned with
the magnetic eld of the pair B, B
. In this example of the three-phase motor, a single step results

in the rotor motion of 360
/6 = 60
. We next give a more detailed description of the electrical

organization of the stepper motor and construct its model.
The basic PM stepper motor consists of a slotted stator and a rotor. There can be two sets of
teeth on the rotor. These two sets of teeth are out of alignment with each other by a tooth width,
as shown in Figure 1.19. Specically, the rotor teeth of the left end-cap are displaced one-half
pitch from the teeth on the right end-cap. A magnet, or an electromagnet, that is mounted on
30 CHAPTER 1

N
N
N
S
S
S
N
Figure 1.19 A possible rotor
structure of a PM stepper motor.
Stator
Stator
N
S Magnet
a b
Iron
end-caps
Stator
windings
Air gap
(c) Side, cross-sectional view
(a) Axial view at a (b) Axial view at b
Stator windings
N1
N5
N2
N3 N4
S1
S2
S3 S4
S5
Figure 1.20 Construction features of a four-pole stator and a ve-pole rotor PM stepper motor.
the rotor magnetizes the iron end-caps. The left end-cap is magnetized as a north pole, while
the right end-cap is magnetized as a south pole. The slotted stator is equipped with two or more
individual coils. Figure 1.20 shows a four-pole stator and a ve-pole rotor structure.
We now discuss the way in which the PM stepper motor can be used to perform precise
positioning. Assume that the starting point is as represented in Figure 1.21. Here, coil AA
is energized so that the upper stator pole is given an S-pole polarity and the corresponding
S
N
S
N
A
B
A
A
B B B
A
Stator windings
N1
N5
N2
N3 N4
S1
S2
S3 S4
S5
Figure 1.21 Coil AA
energized for an SN orientation, while coil BB
de-energized.
S N S N
S
N
B B
B B
A
N5
N2
N3
N4
S1
S2
S3
S4
S5
A
A
A
Figure 1.22 Coil AA
de-energized, while coil BB
energized for an SN orientation.

lower stator is given an N-pole polarity. The rotor remains in this position as long as stator coil
voltageand hence stator polaritiesremains unchanged. This is because the net torque acting
on the rotor is zero. If we now de-energize coil AA
and energize coil BB
so that the left

pole becomes an S-pole and the right pole assumes an N-polarity, then this results in the new
steady-state position shown in Figure 1.22. We can proceed by alternatively energizing the two
stator coils. The PM stepper motor can thus be controlled by digital pulses applied to its two
inputs: move clockwise or move counterclockwise.
Our next objective is to devise a mathematical model of the PM stepper motor. In our dis-
cussion we use the following notation. We denote by R
t
the number of rotor teeth. The angular
displacement between rotor teeth is called the tooth pitch and denoted T
p
. The angular displace-
ment of the rotor as we change the input voltage from one pair of windings to the other is called
the step length and denoted S
l
. The number of phasesthat is, the number of pairs of stator
windingsis denoted N
p
. In our example, we have N
p
= 2, R
t
= 5, and T
p
= 2/R
t
= 72
.
In this example, it takes four input voltage switchings, and hence steps, to advance the rotor one
tooth pitch. Indeed, because
T
p
= 2N
p
S
l
=
2
R
t
, (1.55)
32 CHAPTER 1

we obtain
S
l
=

R
t
N
p
= 18
. (1.56)
We denote the rotor angular displacement by .
Following Krause and Wasynczuk [169, pp. 350351], we nowtrace the main path of the ux
linking the AA
windings for the rotor position shown in Figure 1.23. The ux leaves the left
end-cap through the rotor tooth at the top that is aligned with the stator tooth having the A part
of the AA
winding. The ux then travels up through the stator tooth in the stator iron. Next,
the ux splits and travels around the circumference of the stator, as shown in Figure 1.23(a).
The ux returns to the south pole of the rotor through the stator tooth positioned at the bottom
of the A
part of the AA
winding, as shown in Figure 1.23(b). We denote the peak ux linkage

produced by the rotor permanent magnet in the AA
stator winding by n
m
, where n is the
number of turns of the winding. For = 0
, corresponding to the rotor position in Figure 1.23,

the ux linkage in the AA
winding of the stator is maximal and is approximately equal to n

m
.
S3
S4
S5
S2
(b) Axial view at b (a) Axial view at a
A
B
A
B
N
S
(c) Side, cross-sectional view
a b
N5
N2
N3 N4
S2
S3 S4
S5
Figure 1.23 The main path of the ux linking the AA
winding.
Table 1.2 Symbols Used in Modeling Equations of the PM Stepper Motor
T
p
Tooth pitch
S
l
Step length
R
t
Number of teeth on rotor
N
p
Number of phases, that is, stator winding pairs
Angular displacement of the rotor
Angular velocity of the rotor
v
a
(t ) Voltage applied to phase a, that is, AA
stator winding
v
b
(t ) Voltage applied to phase b, that is, BB
stator winding
i
a
(t ) Current in phase a
i
b
(t ) Current in phase b
R Stator winding resistance per phase
L Stator winding inductance per phase
B Viscous friction coefcient
I Rotor inertia
K
b
Back emf constant
For = 90
/R
t
, corresponding to the rotor position in Figure 1.22, the ux linkage in the AA
winding of the stator is minimal, approximately equal to zero. We model the ux linkage,
a
,
produced by the permanent magnet in the AA
stator winding as
a
= n
m
cos(R
t
). (1.57)
The (main) ux, linking the BB
winding for the rotor position shown in Figure 1.22, would

enter the stator tooth on which the B part of the BB
winding is located from the rotor tooth on

the left. The ux then would travel around the circumference of the stator and return to the rotor
through the stator pole on which the B
part of the BB
winding is wound. For = 90
/R
t
,
corresponding to the rotor position in Figure 1.22, the ux linkage in the BB
winding of the
stator is maximal and is approximately equal to n
m
. For = 0
, corresponding to the rotor

position in Figure 1.23, the ux linkage in the BB
winding is minimal, approximately equal

to zero. We model the ux linkage in the BB
phase of the stator by
b
= n
m
sin(R
t
). (1.58)
In the following development, we use symbols that are described in Table 1.2. Applying
Kirchhoffs voltage law to the circuit of phase a yields
v
a
(t ) = Ri
a
(t ) + L
di
a
(t )
dt
+e
a
(t ),
where e
a
is the back emf induced in phase a; that is,
e
a
=
d
a
dt
=
d
a
d
d
dt
= nR
t
m
sin(R
t
)
d
dt
= K
b
sin(R
t
),
34 CHAPTER 1

where K
b
= nR
t
m
. Hence, the voltage equation for phase a can be represented as
v
a
= Ri
a
+ L
di
a
dt
K
b
sin(R
t
).
Applying now Kirchhoffs voltage law to the circuit of phase b yields
v
b
(t ) = Ri
b
(t ) + L
di
b
(t )
dt
+e
b
(t ),
where e
b
is the back emf induced in phase b; that is,
e
b
=
d
b
dt
=
d
b
d
d
dt
= nR
t
m
cos(R
t
)
d
dt
= K
b
cos(R
t
),
where, as before, K
b
= nR
t
m
. Hence, the voltage equation for phase b can be represented as
v
b
= Ri
b
+ L
di
b
dt
+ K
b
cos(R
t
).
Let T
a
and T
b
denote torques produced by the currents i
a
and i
b
, respectively. Then, applying
Newtons second law to the rotor yields
I
d
2
dt
2
+ B
d
dt
= T
a
+ T
b
.
The PM stepper motor is an electromechanical device in which electric power is transformed
into mechanical power. The mechanical power, P
a
, due to the torque T
a
is
P
a
= T
a
d
dt
. (1.59)
Assuming that there are no losses, this mechanical power is equal to the power developed in
phase a:
P
a
= i
a
d
a
dt
= i
a
d
a
d
d
dt
. (1.60)
Equating the right-hand sides of (1.59) and (1.60) and performing simple manipulations, we
obtain
T
a
= i
a
d
a
d
= i
a
K
b
sin(R
t
).
Applying similar arguments to phase b; we get
T
b
= i
b
d
b
d
= i
b
K
b
cos(R
t
).
Taking the above equations into account, we represent the equation of motion of the rotor as
I
d
2
dt
2
+ B
d
dt
= T
a
+ T
b
= i
a
K
b
sin(R
t
) +i
b
K
b
cos(R
t
).
Combining the above equations, we obtain a model of a two-phase (i.e., a two-input) PMstepper
motor. We have
d
dt
= ,
d
dt
=
K
b
I
i
a
sin(R
t
) +
K
b
I
i
b
cos(R
t
)
B
I
,
di
a
dt
=
R
L
i
a
+
K
b
L
sin(R
t
) +
1
L
v
a
,
di
b
dt
=
R
L
i
b

K
b
L
cos(R
t
) +
1
L
v
b
.
We now dene the state variables: x
1
=, x
2
=, x
3
=i
a
, and x
4
=i
b
. We let K
1
= B/I , K
2
=
K
b
/I , K
3
= R
t
, K
4
= K
b
/L, K
5
= R/L, u
1
=
1
L
v
a
, and u
2
=
1
L
v
b
. Using the above notation,
we represent the PM stepper motor model in state space format:
_
_
x
1
x
2
x
3
x
4
_
_
=
_
_
x
2
K
1
x
2
K
2
x
3
sin(K
3
x
1
) + K
2
x
4
cos(K
3
x
1
)
K
4
x
2
sin(K
3
x
1
) K
5
x
3
K
4
x
2
cos(K
3
x
1
) K
5
x
4
_
_
+
_
_
0 0
0 0
1 0
0 1
_
_
_
u
1
u
2
_
. (1.61)
1.7.4 Stick Balancer
In this subsection we develop a nonlinear model of the stick balancer, also known as the cart
with an inverted pendulum. In our derivation of the model, we use Newtons laws, in a fashion
similar to that of Kwakernaak and Sivan [173] and Ogata [216], where linearized models of the
stick balancer were developed. An alternative derivation, using the method of DAlembert, can
be found in Cannon [41, Section 22.4].
A free-body diagram of a cart on which a stick (i.e., an inverted pendulum) is mounted is
shown in Figure 1.24. Let H = H(t ) and V = V(t ) be horizontal and vertical reaction forces,
respectively, in the pivot. That is, the force H is the horizontal reaction force that the cart
exerts upon the pendulum, whereas H is the force exerted by the pendulum on the cart.
Similar convention applies to the forces V and V. The x and y are the coordinates of the
xed, nonrotating coordinate frame xy. The angular displacement of the stick from the vertical
position is = (t ). The mass of the cart is M, while the mass of the stick is m. The length of
the stick is 2l, and its center of gravity is at its geometric center. The control force applied to
the cart is labeled u. We assume that the wheels of the cart do not slip. We model the frictional
force of the cart wheels on the track by
f
c
=
c
sign( x), (1.62)
where
c
is the cart friction coefcient. Friedland [91, p. 201] refers to this model of friction as
the classical Coulomb friction model. Let (x
G
, y
G
) be the coordinates of the center of gravity
of the stick. Then,
x
G
= x +l sin(),
y
G
= l cos().
(1.63)
36 CHAPTER 1

V
H H
V
x
0
mg
l
l
y
u
Figure 1.24 Free-body diagram of a
cart with an inverted pendulum.
We are now ready to write down equations modeling the system. The equation that describes the
rotational motion of the stick about its center of gravity is obtained by applying the rotational
version of Newtons second law. Summing the moments about the center of gravity of the stick,
we obtain
I
d
2
dt
2
= Vl sin() Hl cos(), (1.64)
where
I =
_
l
l
r
2
dm =
ml
2
3
(1.65)
is the moment of inertia of the stick with respect to its center of gravity.
We next write the equation that describes the horizontal motion of the center of gravity of
the stick. Applying Newtons second law along the x axis yields
m
d
2
dt
2
(x +l sin()) = H. (1.66)
Performing the required differentiation, we get
m( x +l[
2
sin() +

cos()]) = H. (1.67)
The equation that describes the vertical motion of the center of gravity of the stick is obtained
by applying Newtons second law along the y axis. We have
m
d
2
dt
2
(l cos()) = V mg. (1.68)
Performing the needed differentiation yields
ml(
2
cos()

sin()) = V mg. (1.69)
Finally, we apply Newtons second law to the cart to get
M
d
2
x
dt
2
= u H f
c
. (1.70)
Substituting (1.70) into (1.67) gives
m x +ml

cos() ml
2
sin() + f
c
= u M x. (1.71)
Now substituting equations (1.69) and (1.70) into (1.64) yields
I

= (mg ml
2
cos() ml

sin())l sin() + f
c
l cos() (u M x)l cos(). (1.72)
We next substitute u M x from (1.71) into (1.72) and perform manipulations to get
I

= mgl sin() ml
2

m xl cos(). (1.73)
Let
a =
1
m + M
.
Then, we can represent (1.71) as
x = mal

cos() +mal
2
sin() af
c
+au. (1.74)
We substitute (1.74) into (1.73) to obtain
=
mgl sin() m
2
l
2
a
2
sin(2)/2 +mal cos() f
c
mal cos()u
I m
2
l
2
a cos
2
() +ml
2
. (1.75)
Let x
1
= and x
2
=

. Using the expression for I , given by (1.65), we represent (1.75) in
state-space format as follows:
x
1
= x
2
,
x
2
=
g sin(x
1
) mlax
2
2
sin(2x
1
)/2 +a cos(x
1
) f
c
4l/3 mla cos
2
(x
1
)
mal cos(x
1
)u
4ml
2
/3 m
2
l
2
a cos
2
(x
1
)
.
(1.76)
We next substitute

, obtained from (1.73), into (1.74) to get
x =
mag sin(2x
1
)/2 +ax
2
2
sin(x
1
)4ml/3 +(u f
c
)4a/3
4/3 ma cos
2
(x
1
)
. (1.77)
Let x
3
= x and x
4
= x. Then, by combining (1.76) and (1.77) we obtain a state-space model of
the inverted pendulum on a cart of the form
_
_
x
1
x
2
x
3
x
4
_
_
=
_
_
x
2
g sin(x
1
) mlax
2
2
sin(2x
1
)/2
4l/3 mla cos
2
(x
1
)
x
4
mag sin(2x
1
)/2 +al x
2
2
sin(x
1
)4m/3
4/3 ma cos
2
(x
1
)
_
_
+
_
_
0
a cos(x
1
)
4l/3 mla cos
2
(x
1
)
0
4a/3
4/3 ma cos
2
(x
1
)
_
_
(u f
c
).
(1.78)
38 CHAPTER 1

While designing a controller for the stick balancer system, we will use an equivalent represen-
tation of (1.78):
_
_
x
1
x
2
x
3
x
4
_
_
=
_
_
x
2
g sin(x
1
)
4l/3 mla cos
2
(x
1
)
x
4
mag sin(2x
1
)/2
4/3 ma cos
2
(x
1
)
_
_
+
_
_
0
a cos(x
1
)
4l/3 mla cos
2
(x
1
)
0
4a/3
4/3 ma cos
2
(x
1
)
_
_
_
u f
c
+ml x
2
2
sin(x
1
)
_
. (1.79)
1.7.5 Population Dynamics
We rst model dynamics of a single population. Then, we use the obtained model to investigate
two interacting populations.
Let x = (t ) be the population of the given species at a time t . The simplest model of the
population dynamics of the given species is the one resulting from the hypothesis that the net
rate of change of x is proportional to the current value of x. Let r be the rate of growth of the
population. Then, the simple model of the population dynamics is
x = r x. (1.80)
Solving (1.80) subject to the initial condition, x(0) = x
0
, yields
x = e
rt
x
0
. (1.81)
In the above linear model, the growth rate, r, is constant. It does not account for the limitation
of space, food supply, and so on. Therefore, we modify the unrestricted growth rate r to take
into account the fact that the environment of the population can only support a certain number,
say K, of the species. The constant K is called the carrying capacity of the environment or
the saturation level. Specically, if x > K, then because there is not enough food or space to
support x, more species will die than are born, which results in a negative growth rate. On the
other hand, if x < K, then there is an extra food and space and the population growth is positive.
One function that satises the above requirements has the form
r
_
1
x
K
_
.
Using the above model of the growth rate, we obtain the following model of the population
growth:
x = r
_
1
x
K
_
x. (1.82)
Let c = r/K. Then, we represent (1.82) as
x = (r cx)x (1.83)
Equation (1.83) is known as the Verhulst equation or the logistic equation. According to Boyce
and DiPrima [32, p. 60], Verhulst, a Belgian mathematician, introduced equation (1.83) in 1838
to model human population growth. For reasons somewhat unclear, Verhulst referred to it as the
logistic growth; hence (1.83) is often called the logistic equation.
We will use the logistic equation to analyze interacting species. We rst model two interacting
species that do not prey upon each other, but compete for the same resourcesfor example, for
a common food supply. Let x
1
and x
2
be the populations, at time t , of the two competing species.
We assume that the population of each of the species in the absence of the other is governed by
a logistic equation. Thus, a model of two noninteracting species has the form
x
1
= (r
1
c
1
x
1
)x
1
,
x
2
= (r
2
c
2
x
2
)x
2
,
(1.84)
where r
1
and r
2
are the unrestricted growth rates of the two populations, while r
1
/c
1
and r
2
/c
2
are their saturation levels. When both species are present, each will affect the growth rate of the
other. Because x
1
and x
2
use the same resources, the growth rate of x
1
will be lowered by an
amount proportional to the size of x
2
. We modify the growth equation for species x
1
as
x
1
= (r
1
c
1
x
1
a
1
x
2
)x
1
, (1.85)
where a
1
is a measure by which species x
2
lowers the growth rate of x
1
. Similarly, the growth
rate of x
2
is reduced by a
2
x
1
, where a
2
is a measure of the degree by which species x
1
lowers
the growth rate of x
2
. Therefore, the growth equation for species x
2
becomes
x
2
= (r
2
c
2
x
2
a
2
x
1
)x
2
. (1.86)
In summary, equations (1.85) and (1.86) model two species that interact by competing for a
common resource like food or other supply.
We will now study the situation in which one species, the predator, preys upon the other
species, the prey, and there is no competition between the predator and prey for a common
resource. Let x
1
and x
2
denote the populations of the prey and predator, respectively. We will
make the following modeling assumptions:
1. When there is no predatorthat is, x
2
= 0the prey enjoys an exponential growth gov-
erned by
x
1
= r
1
x
1
, r
1
> 0. (1.87)
2. When there is no preythat is, x
1
= 0the predator perishes. We model this scenario
using the equation
x
2
= r
2
x
2
, r
2
> 0. (1.88)
3. The larger the population of the predator, the more prey is eaten. The larger the population
of the prey, the easier hunting is for the predator. We model the growth rate decrease of
the prey by including in (1.87) the term dx
1
x
2
, d > 0. The growth rate increase of the
predator is modeled by adding the term ex
1
x
2
in (1.88), where e > 0.
The resulting model of the predatorprey relationship is
x
1
= (r
1
dx
2
)x
1
x
2
= (r
2
+ex
1
)x
2
(1.89)
Equations (1.89) are known as the LotkaVolterra equations. They were published by Lotka in
1925 and by Volterra in 1926.
40 CHAPTER 1

Notes
A landmark paper on mathematical description of linear dynamical systems from a controls
point of viewis by Kalman [147]. The rst chapter of Sontags book [269] is a nice, easy to read,
and comprehensive introduction to the subject of mathematical control theory. Sontags chapter
is easily available on the web. For more examples of dynamical system models from mechani-
cal and electrical engineering, the reader is referred to Truxal [283], Banks [19], Raven [240],
Mohler [207], and DAzzo and Houpis [57]. Wertz [298] has many excellent real-life models
from space science. A well-written paper on vehicle modeling and control for broad audience
is by Ackermann [2]. In addition to references 303 and 299, vehicle and tire dynamics mod-
eling are investigated in detail by Nalecz [211], Nalecz and Bindemann [212, 213], and Smith
and Starkey [263, 264]. For more on stepper motors, the reader is referred to Acarnley [1],
Kenjo [153], Kenjo and Nagamori [154], Del Toro [61, Chapter 9], Krause and Wasynczuk [169,
Chapter 8], and Zribi and Chiasson [319]. For further analysis of the equations modeling
competing species and the predatorprey equations, we recommend Boyce and DiPrima [32,
Sections 9.4 and 9.5]. Our derivation of the logistic equation follows that of Sandefur [253]. The
logistic equation as well as equations modeling interacting species can be used for mathematical
modeling of tumor dynamics and interactions between tumor and immune system. A survey of
models for tumorimmune system dynamics can be found in the well-edited book by Adam and
Bellomo [4].
Mayr writes on page 109 in reference 199, It is still widely believed that the steam-engine
governor is the oldest feedback device, and that James Watt had not only invented but also
patented it. While both errors are easily refuted, we are still not able to reconstruct the history of
this invention in all desired completeness. Watt did not patent the governor. He did not invent
it either. On page 112 of his book [199], Mayr adds the following: But the application of the
centrifugal pendulum in a system of speed regulation of steam engines was a new breakthrough
for which the rm of Boulton & Watt, if not James Watt himself, clearly deserves the credit.
EXERCISES
1.1 Prove the law of conservation of energy as stated in the following theorem:
Theorem 1.1 Given a conservative vector eld Fthat is, F =U, where U is the
potential energyassume that a particle of mass m, satisfying Newtons law, moves
on a differentiable curve C. Then, the sum of the potential energy and kinetic energy
is constant.
Hint Denote by C(t ) a differentiable curve along which a particle of mass m moves.
Hence,
F(C(t )) = m

C(t ).
Prove that
U(C(t )) +
1
2
m

C(t )
2
= constant.
EXERCISES 41
To showthat the above relation holds, differentiate U(C(t )) +
1
2
m

C(t )
2
with respect
to time and show that the time derivative is zero; hence the sum of the energies is
constant.
1.2 Show that in a conservative vector eld the work done by the eld on a particle that
moves between two points depends only on these points and not on the path followed.
Hint Note that the work done against the conservative force eld F along a curve C
between points C(t
1
) and C(t
2
) is
_
t
2
t
1
F(C(t ))
T

C(t ) dt.
By assumption, F is a conservative vector eld. Hence,
F = U. (1.90)
Combining the above yields
_
t
2
t
1
F(C(t ))
T

C(t ) dt =
_
t
2
t
1
U(C(t ))
T

C(t ) dt.
Use the chain rule to represent the expression inside the integral on the right-hand
side as
U(C(t ))
T

C(t ) =
dU(C(t ))
dt
.
1.3 An example of a closed-loop feedback system is a toilet ushing device connected to a
water tank. Atoilet systemis shown in Figure 1.25. The control objective is to maintain
water in the tank at a constant level. Draw a block diagram of the system.
Float rod Float ball
Trip lever
Lift wire
Tank ball
Valve seat
Outlet valve
Overflow
tube
Handle
Ballcock
Figure 1.25 Toilet-ushing system of Exercise 1.3. (Adapted from reference 241, p. 211.)
42 CHAPTER 1

Table 1.3 Numerical Data for the Two-Tank Pumped
Storage System Model for Exercise 1.4
Tank 1 diameter 2 m
Tank 2 diameter 3 m
R
1
10 sec/m
2
R
2
20 sec/m
2
R
1
A
2
A
1
h
1
q q
o
q
i
R
2
h
2
Pump
Figure 1.26 A two-tank pumped storage system for Exercise 1.4.
1.4 The goal of this exercise is to develop a model of a two-tank pumped storage system.
Figure 1.26 illustrates such a system. A pumped storage system can be used to supply
water to a turbine producing electric power. The two storage tanks are connected
together through a valve that may be modeled as a linear resistance R
1
. The resistance
of the turbine supply pump is modeled by a linear resistance R
2
. Suppose that we wish
to design a controller that would control the inow rate q
i
in order to maintain the
output ow rate q
o
at some desired value q
d
. We assume that the linearized ow rates
q(t ) and q
o
(t ) are proportional to the corresponding water heights; that is,
q(t ) =
h
1
(t ) h
2
(t )
R
1
,
q
o
(t ) =
h
2
(t )
R
2
.
Let A
1
and A
2
represent the surface areas in the two tanks. Then, the principle of
conservation of mass for an incompressible uid yields
q
i
(t ) q(t ) = A
1
h
1
(t ),
q(t ) q
o
(t ) = A
2
h
2
(t ).
Numerical data of a two-tank pumped storage system is given in Table 1.3. Find the
transfer function L(q
o
(t ))/L(q
i
(t )).
1.5 The objective of this exercise is to derive a model of an automotive suspension. The role
of an automotive suspension is to support the vehicle body on the axles. Asimple linear
model of the suspension assembly associated with a single wheel, called the quarter-car
model, is shown in Figure 1.27. The vehicle body is represented by the sprung mass,
m
s
, while the tire and axle are represented by the unsprung mass, m
u
. Model the tire
EXERCISES 43
Table 1.4 Parameters of the Vehicle Suspension of
Exercise 1.5
m
s
355 kg
m
u
46 kg
k
s
14384 N/m
b
s
1860 N/m/sec
k
t
300,000 N/m
m
s
k
s b
s
F
a
z
s
m
u
z
u
z
r
k
t
Figure 1.27 Linear quarter-car model for
Exercise 1.5.
as a spring with spring coefcient k
t
, and assume its damping to be negligible. The
suspension consists of the spring, the shock absorber, and the variable force element.
The role of the variable force element is to generate a force that compensates for the
uneven road, which is the source of the system disturbances. Assume that the variable
force element F
a
can instantaneously provide any desired force. The above described
model of an automotive suspension is referred to as a linear active suspension. Write
down equations modeling the dynamics of the two-degree-of-freedom system depicted
in Figure 1.27 by applying Newtons law to each mass of the system and taking into
account the positive direction of the z coordinates. Dene the state variables to be
x
1
= z
s
z
u
= suspension stroke,
x
2
= z
s
= sprung mass velocity,
x
3
= z
u
z
r
= tire deection,
x
4
= z
u
= unsprung mass velocity.
Let u = F
a
be the system input, and let d(t ) = z
r
(t ) be the disturbance input due
to the road roughness. Represent the obtained linear model of active suspension in
state-space format. Use data of a commercial vehicle given in Table 1.4. The data come
fromSunwoo and Cheok [272]. Substitute the vehicle data into the modeling equations.
A design goal of an automotive engineer may be to design a control law to isolate the
sprung mass from road disturbances. In other words, he or she may wish to design
a control law such that the vertical velocity of the sprung mass x
2
is as close to 0 as
44 CHAPTER 1

possible in spite of road disturbances. Assume that we can measure the vertical velocity
of the sprung mass. Write the output equation.
To investigate the effect of road disturbances on the vertical velocity of the sprung
mass, set u = 0 and (t ) = d sin(6t ). Generate a plot of y versus t for t in the range
[0 sec, 5 sec] and d = 0.5.
1.6 The objective of this exercise is to develop a model of a single-axis machine tool drive in
which a motor drives a lead screw through a gear. Figure 1.28 illustrates such a system.
This model comes from Driels [68, pp. 613619]. The system operates as follows.
The lead screw moves the table with a workpiece on it under the cutter to perform
the desired operation. The table is positioned by a control system. We wish to place the
table in a desirable position as rapidly and accurately as possible. This means that while
modeling the machine tool drive system we should, in particular, take into account the
exibility of the lead screw. The length of the lead screw implies that its rotation at
the gear box will not be the same as its rotation at the table. Thus, the angular twist
along the length of the lead screw cannot be neglected. Model the lead screw exibility
using a torsional spring as illustrated in Figure 1.29. The system input is the motor
torque
m
, and the system output is the rotation of the end of the lead screw labeled
l
. Assume that the damping torque, due to viscous friction, acting on the motor is
proportional to its angular velocity

m
=
m
with the coefcient of viscous friction
denoted c
m
. Denote the motor moment of inertia by I
m
, and denote by
1
the load torque
on the motor gear due to the rest of the gear train. Write the equation modeling the
dynamics of the motor shaft.
Let
2
be the torque transmitted to the load shaft, and let N = N
1
/N
2
be the gear
ratio, where N
1
denotes the number of teeth of the rst gear, and N
2
is the number of
teeth of the second gear. The work done by the rst gear is equal to that of the second.
Motor
Gear
Cutter
Guides
Workpiece
Potentiometer
Lead screw
Table
m
Figure 1.28 A schematic of a
machine tool drive system of
Exercise 1.6. (Adapted from
Driels [68]. )
EXERCISES 45
Table 1.5 Data for the Machine Tool Drive System
Model for Exercise 1.6
I
m
0.0001 Nmsec
2
/rad
I
l
0.001 Nmsec
2
/rad
c
m
= c
l
0.01 Nmsec/rad
k 1 Nm/rad
N 1/5
l
c
m
m
k(
m
N
l
)
m
I
l
I
m
c
l
l
Torsional spring
k
1
N
Figure 1.29 A mechanical model of
the machine tool drive system for
Exercise 1.6.
Model the load shaft exibility with a torsional spring. The torsional spring coefcient
is k. Assume that the moment of inertia of the lead screw is negligible compared to that
of the load. Denote the moment of inertia of the load by I
l
. The torque due to viscous
friction of the load is c
l
l
, where c
l
is the coefcient of the viscous friction of the
load. The angular displacement of the end of the lead screw is
l
. Write the equation
modeling the dynamics of the load. Dene the state variables to be
x
1
=
m
, x
2
=

, x
3
=
l
, x
4
=

l
.
Represent the modeling equations in state-space format. Typical numerical data for a
machine tool drive system are shown in Table 1.5.
1.7 Consider a bicycle model of the front-wheel-steering vehicle shown in Figure 1.30.
Construct a two-dimensional state-space model of the vehicle, where x
1
= y and
x
2
=

. The input u =
f
and the output y = x
2
. In your derivation assume small
angles. Vehicle data are given in Table 1.6.
Table 1.6 Vehicle Data for Exercise 1.7
l
1
= l
2
1 m
v
x
10 m/sec
m 1000 kg
I
z
1000 kgm
2
C
f
= C
r
10,000 N/rad
46 CHAPTER 1

f
F
yr
Tire velocity
Tire velocity
F
yf
r
F
yr
l
2
l
1
f
F
yf
Figure 1.30 Bicycle model of a vehicle for Exercise 1.7.
1.8 Consider the two-massspring system depicted in Figure 1.31.
(a) Form the Lagrangian of the system.
(b) Write the Lagrange equations of motion for the system.
(c) Represent the obtained Lagrange equations of motion in state-space format.
m
1
1 kg
u
x
1
m
2
1 kg
x
2
y
Frictionless
surface
k
Figure 1.31 The two-massspring system for Exercise 1.8.
1.9 Consider the mechanical system shown in Figure 1.32.
(a) Form the Lagrangian for the system.
(b) Write the Lagrange equation of motion for the system.
(c) Represent the system model in state space format.
M
B
F
K
x
Figure 1.32 Mechanical system for
Exercise 1.9.
EXERCISES 47
1.10 Consider a dynamical system that consists of a cart with an inverted pendulum (point
mass on a massless shaft) attached to it as depicted in Figure 1.33.
(a) Write the Lagrange equations of motion for the system.
(b) Represent the obtained model in state-space format using the state variables
x
1
= x, x
2
= x, x
3
= , x
4
=

.
Force u
Frictionless
surface
Distance x
M 10 kg
m 1 kg
g 10 m/sec
2
l 1 m
Figure 1.33 An inverted pendulum moving on a cart for Exercise 1.10.
1.11 Consider a schematic of a eld-controlled DCmotor shown in Figure 1.34. Let x
1
= ,
x
2
=

, x
3
= i
f
, and y = x
1
. Treat the eld inductance L
f
as a parasitic element and
denote it as .
(a) Form the actual plant model P
in state-space format.
(b) Find the reduced-order design model P
0
.
R
f
6
K
m
6 Nm/A
Load inertia
I 1 kgm
2
Damping constant
B 1 kg m
2
/sec
i
f
L
f
u
Figure 1.34 A diagram of a eld-controlled DC motor for Exercise 1.11.

Dynamical System Model

Uploaded by

Copyright:

Available Formats

Dynamical System Model

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dynamical System Model

Uploaded by

Copyright:

Available Formats

CHAPTER

Open-loop control systems

Closed-loop control systems

Figure 1.8 The r robot manipulator.

Figure 1.9 A lumped mass

The z coordinate is directed downward with respect to the vehicle.

The yaw velocity about the z axis is denoted

. The rotor will

is de-energized while the pair B,

. In this example of the three-phase motor, a single step results

. We next give a more detailed description of the electrical

energized for an SN orientation, while coil BB

de-energized, while coil BB

energized for an SN orientation.

and energize coil BB

so that the left

winding, as shown in Figure 1.23(b). We denote the peak ux linkage

, corresponding to the rotor position in Figure 1.23,

winding of the stator is maximal and is approximately equal to n

winding for the rotor position shown in Figure 1.22, would

winding is located from the rotor tooth on

winding is wound. For = 90

, corresponding to the rotor

winding is minimal, approximately equal

phase of the stator by

You might also like