0% found this document useful (0 votes)
70 views70 pagesELEC4632 Notes Part 1
notes
Uploaded by
Johnny JobbyCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
70 views70 pagesELEC4632 Notes Part 1
notes
Uploaded by
Johnny JobbyCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 70
Computer Control Systems: List of References
Lecture Notes, Tutorials and Project Manual
‘The following sources have been used in preparation BLEC4632 lecture notes, tutorials and
project manual:
KJ. Astrem and B. Wittenmark, Computer-Controlled Systems. ‘Third Edition, Prentice
Hall, 1997.
GF. Franklin, J.D. Powell and M.L, Workman. Digital Control of Dynamic Systems. Second
Edition. Addison-Wesley, 1990.
iC.Ogata, Modern Control Engineering, Second Edition, Prentice Hall Inc., 1990.
K.Ogata, Modern Control Engineering, ‘Third Edition, Prentice Hall Inc., 1997
B.C. Kuo, Digital Control Systems, Saunders College Publishing, 1992.
M.E. ElHawary, Control System Engineering, Prentice Hall Inc, 1984. 7
R.C. Dorf. Modern Control Systems, Fifth Edition, Addison-Wesley, 1989.
hth Edition, Addison-Wesley, 1998.
R.C. Dorf and R.H. Bishop. Modern Control Systems,
LR. Petersen and A.V. Savkin, Robust Kalman Filtering for Signals and Systems with Large
Uncertainties. Birkhauser, Boston, 1999.
LRPetersen, V.A. Ugrinovskii, and A.V. Savkin. Robust Control Design via H° Methods.
Springer-Verlag, London, 2000.
IEEE Control Systems Magazine.
IEEE Transactions on Control Systems Technology.
In particular, the material of Sections 1.3-1.4, 2.5 (second part), 3.3, Chapter 4 and most
of tutorial questions were taken from the book of Astrem and Wittenmark. Many examples in
Chapters 1-4 are from the book of Dorf and Bishop. 1 INTRODUCTION TO DIGITAL CONTROL SYSTEMS
1A Introduction
‘The term automatic digital control system is intended to be somewhat self-explanatory. ‘Lhe
‘word system implies not just one component but a number of components that work together in
a prescribed fashion to achieve a particular goal. ‘This goal is the control of some physical
quantity, and the control is to be achieved in an automatic fashion, often without the
aid of human supervision. The word digital means that we use a digital computer for real-time
control of a dynamic process (chemical process, vehicle, electromechanical servomechanism, ete.)
‘The purpose of the introductory material in this first chapter is to give an intuitive feeling
for what is meant by an automatic digital control system.
Practically all-control’ systems that: are implemented today is based on computer control
‘oclectronics is that practically all control
‘A consequence of the revolutionary advances in
systems constructed today are based ou microprocessors and sophisticated microcontrollers. It is
therefore important to understand digital control systems well. Such systems may be viewed as
approximations of analog-control systems, but this is a poor approach because the full potential
of computer control is not used. ‘There are also phenomena that occur in digital control systems
le to
that have no correspondence in analog systems. By using digital control systems it is pos
obtain higher performance than with analog systems. It is important for a control engineer to
understand this.
‘The main goal of this course is to provide a solid background for understanding,
analyzing, and designing digital control systems.
~1- A typical digital control system can be described schematically as in Figure 1.1
The output of the plant(process) y(t) is a continuous-time signal. ‘The output is converted to
into digital form by the analog-to-digital (A-D) converter. ‘The conversion is done at the sampling
times ¢,. ‘The computer interprets digital signal {y_} as @ sequence of numbers, processes the
measurements using an algorithm, and gives a new sequence of numbers {uj}. This sequence is
converted to an analog signal by a digital-to-analog (D-A) converter. ‘The D-A converter produces
@ continuous-time signal. This is normally done by keeping the control signal constant between
the conversions, The events are synchronized by the real-time clock in the computer.
A typical digital control system contained both continuous-time signals and sampled or discrete-
time signals, The mixture of different types of signals sometimes causes difficulties. In many cases,
however, it is sufficient to describe the behaviour of the system at the sampling instants. ‘The
signals are then of interest only at discrete times. Such systems will be called diserete-
systems. They can be described by difference equations.
We will consider two types of problems:
1, Analysis problems - to analyze the digital control system shown in Pig. 1.1.
2. Synthesis problems - to design a suitable “Algorithm”,
1.2 Examples of digital control systems
Example 1.1
Consider the satellite attitude control system shown in Fig. 1.2. The diagram shows the
control of only the angle @ (In the actual system there are controls about three axes). Smiall
Jets apply reaction forces to rotate the satellite body into the desired attitude. ‘The two skew
symmetrically placed jets denoted by A or B operate in pairs. Assume that each jet thrust is F/2
and a torque T = FI is applied to the system. The jets are applied for a certain time duration
-- Computer
1
i ee —
AD ftsoritna [a
we
Process Communications satellite. (Couresy Ford Aerospace and Communi
cations Corporation.)
iter of mass
Reference
Fig. 4.2, and thus the torque can be rewritten as T(t). ‘The moment of inertia about the axis of rotation
at the center of mass is J.
We assume that torque 7(¢) is the input, and the angular displacement 0(t) of the satellite is
the output. We consider the motion only in the plane of the page.
Applying Newton's second Jaw to the system under assumption that there is no friction in
the environment of the satellite, we have
(.)
Now we obtain the transfer function for the satellite attitude control system.
We have established that this system can be described by the polynomial equation (1.1). We
now take Laplace transforms to calculate the transfer function from the input u(t) = T(t) to the
(i) (we assume zero initial conditions):
output y(t)
Js*¥(s) = U(s).
From this equation, we-obtain that the transfer function of the system is given by
202
OO) = Fe) = Fs
‘This control system will serve asa sequential design example for our course. We will design
various digital controllers for this system.
Example 1.2
‘The space shuttle, with its robot arm, is shown in Fig. 1.3. An astronaut controls the robot
arm and gripper by using a window and the TV cameras. He/she uses information from the
computer generated data display and visual sensory data from the TV monitor and by looking out the window. He/she commands the robot, arm via joystick command to the computer. The
block diagram of this system is shown in Fig.14.
Example 1.3
Computer control of a robot to spray-paint an automobile is shown in Fig. 1.5.
1.3 A naive approach to digital control
We may expect that a computer-controlled system behaves as a continuous-time system if the
sampling period is sufficiently small. ‘This is true under very reasonable assumptions. We will
illustrate this with an example.
We will design a computer system to control the satellite attitude (see Example 1.1).
First step: we design a continuous-time controller to control this system.
‘The block diagram of this continuous-time control system is shown in Fig. 1.6. Here u. is the
command signal (reference input). A simple controller can be described by
OK sth
= Uels) — KV (6) (1.2)
Us)
where a,b, and K are some positive parameters. The system with this continuous-time controller
has a reasonable behaviour (see Fig. 1.7).
Second step: we re-write the continuous-time control law as a system of differential equa-
tions.
Indeed, the control Jaw given by (1.2) may be re-written as
U(s) ox, fe(s) = KY) += ¥ (0) (1.3) Fig. 1-3.
(2) Space shutte
nd robot arm and
@) astronaut
conttotof the arn,
[zm }-—-
sever | Revot arm
‘& motorsis}
TV monitor sae
avwindow view Fig. 4.8:
Automobile spray-
paint system, Controller 4] Anptier- || Phat [2
Fig 46.
Output ‘This control law can be re-written in a state-space form as
u(t) = Keel) — v(0) +210)
(0) = a2 (0) + (a - Bult). (14)
‘Third step: we design a digital controller to approximate the continuous-time controller.
‘To obtain an algorithm for digital control, we approximate the derivative #(2) by a difference
as follows:
£(t) = —ax(t) + (a - b)y(t) =
where h is the period of sampling. ‘The following approximation of the continuous-time control
law (1.4) is then obtained:
u,) = K Qua) ~ alta) +206)
@(te41) = (ah + 1x(te) + h(a ~ Bute) (1.5)
where te = kh, teer = (k + 1h.
‘The system with this digital controller has a reasonable behaviour when the period of sampling
is very short (see Fig. 1.8). Notice that the control signal for the computer-controlled system
is constant between the sampling instants. When the sampling period increases the computer-
controlled system will, however, deteriorate (see Fig. 1.9). ‘The response is quite reasonable for
short sampling periods, but the system becomes unstable for longer sampling periods.
We have shown that it is straightforward to obtain an algorithm for digital control simply
by writing the continuous-time control law as as a differential equation and approximating the
derivatives by differences. ‘This “Naive Approach” Procedure seems to work well if the sampling
period is sufficiently small
- 40- Output
3 10
o 5 10 15 0 3 10 15
‘Time (ot) Time (ot)
Fig. bt Fig. o>
Output
Output
0 10
Time Time
Beste Fig. 4.£0. step responses (dots!
‘of a digital computer implementation cf a first-order lag for different dela;
jin the input step (dashed) compared with the first sampling instant. For
comparison the response of the corresponding continuous-time system (solid)
is also shown,
- Na 1.4 Analog and digital systems: the differences and the similar
In this section we will give examples that illustrate the differences and the similarities of
analog and digital control systems. It will be shown that essential new phenomena that require
theoretical attention do indeed occur.
‘Time dependence
‘The presence of sampling makes digital control systems time-varying. The step response of a
continuous-time system and the corresponding digital system is shown in Fig. 1.10. ‘The figure
clearly shows that the digital system is not time-invariant because the response depends on the
time when the step occurs. Ifthe input is delayed, then the output is delayed by the same amount,
only if the delay is multiple of the sampling period.
Because sampling is usually periodic, digital control systems will often result in closed-loop
systems that are linear periodic systems.
Deadbeat control
‘The period nature of the control action can be actually used to obtain control strategies with
superior performance.
Consider the digital control system shown in Fig. 1.11.
‘This system can be represented in a state-space form as,
a(t) = u(t) (1.6)
y(t) = 2(0). (L727)
Suppose that h > 0 is the sampling period of the system. Consider the following digital
control law
uf) = =) (19)
-4A2- Ye Digskal PL ust
| =
Gotreler Goye &
e
+
Fig. & Aa
xe]
ate —
0 n ,
4 who AL th
Gig. AAD, where ti = kh. Let us study step response of the system (1.6), (1.8). In other words, the
solution to equations (1.6), (1.8) with ue(ty,) = 1 for all k = 0,1,2,.... It can be easily seen, that
ystom is shown in Fig, 1.12. ‘The
(te) = u(t,) = 1 for all k =1,2,.... The step response of the s;
control strategy (1.8) is called deadbeat control because the system is at rest when the desired
position is reached. Such a control scheme cannot be obtained with a continuous-time controller
because the solutions to any continuous-time system are sums of functions that are products of
polynomials and exponential functions
A general procedure to design a deadbeat control law will be given in this course.
Sampling creates new frequencies
Stable linear time-invariant continuous-time systems have the property that the steady-state
response to sinusoidal excitations is sinusoidal with the frequency of the excitation signal. Digital
control systems behave in a much more complicated way because sampling will create signals
with new frequencies. This can drastically deteriorate performance if proper precautions are not
taken,
Consider the satellite attitude control system from Exdmple 1.1. Assume that there is a
sinusoidal measurement noise with a high frequency. Fig. 1.13 shows interesting variables for
the continuous-time system and the corresponding computer control system. ‘There is a clearly
drastic difference botween the systems. For the continuous-time system, the measurement noise
has very little influence on the attitude. It does, however, create substantial control action with
the frequency of the measurement noise. ‘The high-frequency measurement noise is not noticeable
in the control signal for the digital control system, but there is also a substantial low-frequency
‘component.
To understand what happens, we can consider Pig. 1.14, which shows the control signal and
the measured signal on an expanded scale. ‘The figure shows that there is a considerable variation
of the measured signal over the sampling period and the low-frequency variation is obtained by
sampling the high-frequency signal at a slow rate.
- Lu We have thus made the observation Uat sampling czcates signals with new frequencies. This
is clearly a phenomena that we must uriderstand in order to deal with digital control systems
-—is- a )
o 10 20
Time
SPP frequency is ist
is a measurement noise 1
sampled-data system.
ine sampling period is & = 0.5. and shere
O.1sin 12, (a1 Continuous-time system: 1b:
° 7 i
Time
an o Rig Gay.
Frequency ay is 1, the sampling peried ts
noise n = O.sin 12
0.5, and there-is a me
= iG 2 DISCRETE-TIME SYSTEMS
2.1 Introduction
Discrete-time systems are those in which one or more variables can change only at discrete
instants of time. These instants may specify the time at which some physical measurement is
performed or the time at which the memory of digital computer is read.
‘Many modern industrial control systems are discrete-time systems since they include some
rete in time. Sometimes, however, the sampling
elements whose inputs and outputs are
operation, or discretezation, may be entirely fictitious and introduced only to simplify the analysis
of a control system which actually contains only continuous elements.
Mathematical models for discrete-time systems are introduced and studied in this chapter.
A key idea of this chapter is to show how a continuous-time system can be transformed into
a discrete-time system by considering the behavior of the signal at the sampling instants.
Roughly speaking, in this chapter the control system is studied as seen from computer. The
computer receives measurements from the process at discrete times and transmits new control
signals at discrote times. The goal is to describe the change in the signals from sample to sample
and disregard the behavior between the samples. The use of difference equations then becomes
a natural tool. it should be pointed out, that discrete-time mathematical models only give the
inuous-time system.
behavior at the sampling points whereas the physical process is still a com 2.2 Mathematical models of discrete-time systems
We will study the following to mathematical models in terms of difference equations for
discrete-time control systems:
1. Polynomial difference equations.
2. State-space difference equations.
Polynomial difference equations
Consider the following difference equation:
coy (+n) bay (k-tn—1) ++ tan1y(K-+ 1) Fany(k) = bou(k-+m)+byte(k--m—1) +--+ a(t] +b a te(h)
where w is the input, y is the output, ¢= kh, = 0,1,2,..., and h is the sampling period.
‘This equation can be re-written as
a(g)y(-) = 6(g)ue(-)
where q is the shift operator, « and 6 are the corresponding polynomials.
State-space models
‘The most general state-space represesitation of a linear discrete-time systems is given by
a(k +1)
Gk)x(k) + H(B)u()
u(h) = C(k)2(k) + D(Au(k) where u(k) is the input vector y(f) is the output vector 2(K) is the state vector = kh, k=
0,1,2,-.. fe is the sampling period.
time-invariant, then it can by represented by the following
If the linear discrete-time system
state-space equations:
a(k-+1) = Ga(k) +Hu(k)
wk) = Cx(k) + Duk)
Consider the following polynomial difference equation:
y(k-+n) + ayy(k-+n 1) + aay(k +m 2) +. ban aye +1) + any(k) = bulk)
‘We would like to re-write this equation in state-space form. Let us define:
ai(k) = u(k)
a(k+1) = 22(k)
tax(k-+1) = x3(k)
trailk+1) = an(h)
an(k- 1) = ay 2_(K) ~ aty—4(h) —-.. — ayer(h) + bu(h) 0 1 0 0
i(k +1) alk) 0
0 0 1
a(k +1) x(t) | | o
ta(k-+ 1) an(k)
Oy Ont yn on
ai(k)
w(K)
ui) = [1ro0. ae
a(t)
cr
a(k+1) = Ga(k) + Hu(k)
uk) = Gxt)
where
alk)
alk)
a(t) =
n(k) aya
Ou.
tna 2.3 Transfer function of a discrete-time system
See your old lecture notes for Z-transform!
‘The z-Transformation
‘The z-transform approach has the same relationship to linear time-invariant discrete-time sys-
tems as the Laplace transform approach bears to linear time:
wwariant continuous
ne systems.
Let {s"(kT)} where
0,1,2,... be a given sequence. ‘Then, the z-transform of 2*(+) is
defined as
&lz"()| =X)
Sours m
io
Solving difference equations by the z-transform method
It is obvious, that
Blot + 1)] = 2X(z) — 22(0).
Consider the following example
a(k+2) +42(k +1) +32(4) =0, 2(0) =0, 2(1) =~2
Solve this equation by using the z-transform method. We have
a(k +2) = 2?X(z) — 20(1) — 270(0).
From this and the above equations, it follows that
(2? +42 +3)X(z) — 22(0) — 4z2(0) - zx(1) Hence,
22
4z+3
Furthermore,
2z
X(2)=-— 2
)=-ayars
Since
we have that
afk) = —(-1)' + (-3)
‘The inverse z-transform
Let the z-transform of a sequence x(k) be given as
bz
*O= Gaara
Determine x(k) = Z-'[X(z)}
Since
X@)=
G-HEFa ~
we have
5
a(k) = ge = (-4)4.
Solving the Diserete-time State Equations
Consider the following state-space and output equations:
(k+l) = Ga(k) + Hulk) (2a)
u(k)
Gx(k) + Du(k) (2.2) ‘The solution of eqn. (1) for any k > 0 may be obtained directly by recursion as follows:
(1) = Ge(0) + Hu(0)
(2) = Ge(1) + Hu(1)
= G*x(0) + GHu(0) + Hu(1)
2@) = Ge(3)+Hu(3)
= Gx(0) + G?Hu(0) + GHu(l) + Hu(2)
By repeating this procedure we obtain
a(k) = Gkz(
Clearly x(k) consists of two parts one representing the contribution of the initial state «(0) and
the other the contribution of the input u(y), 7 =0,1,2,...,k—1.
‘The state-trar
oth) = Gt
ka
afk) = Ofk)0(0) + 0 dk ~ 7 1I)He(s)
jm
‘The output equation can be written as
j — 1) Hu(;) + Du(k)
ket
ak) = Cd(h)z(0) +O > oe
r=
ka
= Co(k)x(0) +6 > o(7)Hu(k — 5 — 1) + Duk)
jo
Fi Computation of state-transition matrix
‘The state-transition matrix can be computed using the following formula:
= ot) =27 [ef -G)72]
‘The-transfer function of a discrete-time system
‘The transfer function of a discrete-time system is defined as the following ratio:
4 Zloutput]
‘Transfer Function = =~
Zlinpul}
with zero initial condition.
For the polynomial system
o(a)u() = H@)u(),
Ge)= 3.
Consider the state-space equations given by
a(e+1) = Gali) + Hu(k)
yk) = Cx(k) + Du(k)
‘Taking Z-transforms
2X(2) = GX(z)+HU(2)
Y(z) = CX(z)+DU(2) Simplifying we get
¥(2) = [eter -oy Hn +0] U(z)
‘Therefore the transfer function is given by
YO _ g—@-ned
68) = FG
= ho- 2.4 Sampling continuous-time systems
Sampling in control engineering means that a continuous-time signal is replaced by a sequence
of numbers, which represents the values of the signal at certain times.
‘Types of sampling operations:
L Periodic sampling: the sampling instants are equally spaced, or ty = kh(t=0,1,2,-.-)
2.Multiple-order sampling: the pattern of t is repeated periodically, ot thyr—tk = constant Vk.
3.Multiple-rate sampling: t = phy and gh where p,q are integers.
4.Random sampling.
‘We will study only the case where the sampling is periodic.
Samplers or holding devices
‘The essential element of a discrete-time system is a sampler.
A periodic sampler converts a continuous signal into a train of pulses occurring at the sampling
instants 0, h,2h,....
A holding device converts the sampled signal into one which is constant between two consec-
utive sampling intervals
ay(kh+0)=2(bh) for OSt 0 using the Laplace transform
(t).
method. We use the following formula
(st - Ay = ce").
Here £.is the Laplace transform, and J is the unity matrix. The Laplace transform method gives
1
(ray = s+1 0 1 s 0 _{#r 0
rou
ahs es
Hence
0
l-e* 1
and
eh
d=
h-1+e*
~At- Fig. %n%.
~1g— Sampling a system with time delay
‘The theory of
‘Time delays are common in mathematical models of industrial proces:
continuous-time systems with time delays is complicated because the systems are infinite-dimensional.
It is, however, easy to sample systems with time delays because the control signal is constant
between sampling instants, which makes the sampled-data system finite-dimensional.
‘The following differential equation is a common mathematical model for control systems with
a time delay
‘#(t) = Ax(t) + Bu(t — 7). (2.4)
Here T’is the time delay. It is assumed initially that 0 <7" 0. ‘The transfer function of this system is defined by
G,(2) = C(I — G(h))“"7() + D
(sce Section 2.3). From this formula, we obtain that if = 1 then
“1
, z-1 05 (2-41)
Gl=(1 0 =—opF
° ( ) x4 1 @-1P
On the other hand, this system ean be represented in input-output form as it is shown in Fig.
. We have that
2.7. Obtain the pulse transfer function of this system for h =
Hence we can use two methods to obtain transfer functions of discrete-time syst
(i) From sampled state-space models
(ii) From input-output representations
Pulse transfer functions of cascaded elements
Consider the systems shown in Fig, 2.8 and determine their pulse transfer functions.
For the system (a)
2(e-2h — ethy
AG) = Cee Ne
Thy
For the system (b)
a
GAG = age
It is important to note that G,G2(z) # Gi(z)G2(z). Hence, we must be careful and observe
~ Ut- Ah zero 4 i" >
. ore
heel S(s+a)
Fy. 2.6
ue) OK [eee 4 ch 30
held |. _ st
Fig. 2.4
— 279- a i
UE ae
i
Poa
@ -
a
2 er SS
wo Ee
Fig. 2.8
= 302 whether or not there is a sampler between cascaded elements.
Pulse transfer functions of closed-loop systems
Consider the closed-loop sampled-data system shown in Fig. 2.9. Then,
= R(s) — H(s)C(s), C(s) = G(s)B"(s)-
Hence,
E(s) = R(s) — G(s) H(s)B"(s).
Furthermore,
BM(s) = RY(s) — GH"(s)B"(s).
It implies that
Rs)
+GIG)
Since
Cs) = G(s) E"(s),
we have
ig) =< LRG)
CO) = TF GIFG)
In terms of z-transform,
_ GR)
CO) = Ty GHey
Hence, the pulse transfer function of the closed-loop system is
oe) ot
Riz) 1+GH(z)"
‘Table 2.1 shows the pulse transfer functions for five typical configurations of closed-loop
—~ %4- z
ao f ao foc oo cw RQ) cay
RG) RG (5) *( RG) iz)
65) i Le al eats of 2 e) Gacy) [2
vm
Roy, RU) cs) C7(s) RG) cur
OG ofa So amas
te
Re RtiGuer GaN co ty RGite) ca
cu AO Leto] LBS of Guay tS
i
Table
~ 32-
2.4. sampled-data systems.
Example 2.9
‘The
Computer control of a robot to spray-paint an automobile is shown in Fig, 2.10(a
block diagram of this system is shown in-Fig. 2.10(b). Obtain the pulse transfer function of the
closed-loop system,
For this case,
GQ) _ _ Cole)
RE) 1+6,@)
and-we have proved above, that the open-loop transfer function is
Kh — K ,_K(z-1)
az—1) @ ety
Gyl2)
Example 2.10
A spacecraft with a camera is shown in Fig. 2.11(a). The closed-loop system for the rotational
speed control for the camera slewing is shown in Fig, 2.11(b). Obtain the pulse transfer function.
Example 2.11
‘There have been significant developments in the application of robotics technology to nuclear
ion of remotely operated devices can signifi-
power plant maintenance problems. The appli
cantly reduce radiation exposure to personnel and improve maintenance-prograin performance.
A remotely controlled robot for nuclear plants is shown in Fig. 2.12 (a) and (b). Obtain the
pulse transfer functi
~ 337 Fig. 2. 10(a)
Automobile spray-
aint system.
= BAS Ry LN Cea)
camera, ema
Wen tere al
Kg der _—_+
arb h old} | Gantsedised)
- Fay. 2. 1 0b)
Cis) Cig. 2.02¢0) Manipestetnm
Remotely /
centotedrobotior
uctear plants,
-36- 3 ANALYSIS OF DISCRETE-TIME SYSTEMS
3.1 Introduction
me systems have been introduced in Chapter 2. Tt has
Mathematical models for discrete-
deen shown how continuous-time control systems are transformed when sampled. In this chapter
‘we will develop the key tools for analyzing discrete-time systems.
3.2 Stability of linear discrete-time contrel systems
he issue of ensuring stability of a closed-loop feedback system is central to control system
design. Knowing that an unstable closed-loop system is generally of no practical value, we seek
methods to help us to analyze and design stable systems.
A stable system is a dynamic system with a bounded response to a bounded
input.
See any undergraduate control engineering textbook,
‘The concept of stability can be illustrated by considering a right circular cone placed on a
plane horizontal surface. If the cone is resting on its base and is tipped slightly, it returns to its
original equilibrium position. ‘This position and response are said to be stable, If the cone rests
on its side and is displaced slightly, it rolls with no tendency to leave the position on its side. ‘This position is designated as the neutral stability. On the other hand, if the cone is placed on its
tip and released, it falls onto its side. This position is said to be unstable. ‘These three position
are illustrated in Fig. 3.1.
‘The stability of a dynamic system is defined in a similar manner. ‘The response to a dis-
placement, or initial condition, will result in either a decreasing, neutral, ae hnexeosing response.
Consider an extremely simple discrete-time control system which is described by the following
first-order equation
a(kh +h) = ax(kh) +u(kh)
|. Then the solution of this system
where @ is @ parameter. Consider the control input u =
(kh)
stable if fal <1
is shown in Fig. 3.2. Hence this system is
neutral if fa
unstable if Ja] > 1.
All our control systems designs must result in a closed-loop stable‘system: Many physical
systems are open-loop unstable, and some systems are even designed to be open-loop unstable.
For example, most modern fighter aircraft are open-loop unstable by design, and without active
feedback control assisting the pilot they cannot fly. Roughly speaking, the more open-loop stable
an aircraft is, the more difficult it is to maneuver. For example, a fighter aircraft ig always less
open-loop stable than a commercial aircraft. Active control is introduced by control engineers to
stabilize the-unstable plant.
_ Let G(2) be the transfer function of a discrete-time control system and
= be be bbz thm _ B(z)
OO) == Co tat. anizre = AG) (2) Sable ( Neutral
Rg. B.U The stability ofa cone. where ai’s and &;’s are constants and m 0, then the polynomial (3.1) has all its roots inside the unit dise if and only if all
a§, k=0,1,...,n—1 are positive. If no af is zero, then the number of negative af is equal to
the number of roots outside the unit disc.
Remark If all af are positive for k = 1,2,...,n—1, then the condition a8 > 0 can be shown
to be equivalent to the conditions
AQ) >0, (-1)"A(-1) > 0.
‘These conditions constitute necessary conditions for stability arid hence can be used before form-
ing the table.
Note The equation (3.1) is called the characteristic equation of the control system-and
the polynomial (3.1) is called the characteristic polynomial of the control system.
Stability of a second-order discrete-time system
Let the characteristic equation be :
A(z) = 2? tayz +092 =0.
Jury’s array for this equation is shown in Fig. 3.6. It can be easily seen from the Jury’s criterion
that all the roots of the equation are inside the unit disc if and only if the following. conditions
hold:
a3 <1, a2>—L+a, a> ‘This stability area is shown in Fig. 3.7.
Example 3.3
Designers até at work to develop small, fast, vertical-takeoff fighter aircraft that are invisible
to radar (stealth aircrat) (see Fig, 3.8). ‘The control system for the heading or direction control
jg a discrete-time negative unity feedback system with the open-loop transfer function
K(—22? +32 —0.5)
zeit)”
a= where
aft = af - yet;
a = ab /oh
‘The first and second rows are the coefficients in (3.4) in forward and re-
verse order, respectively. The third row is obtained by multiplying the second.
row by aq = a,/a9 and subtracting this from the first row. The last element in
the third row is thus zero. The fourth row is the third row in reverse order. The
scheme is then repeated until there are 2n + 1 rows. The last row consists of
only one element.
Ry 3.5, 3.3 Controllability, Reachability, Observability, and Detectability
In this section, two fundamental questions 'for discrete-time control systems are discussed.
‘The first is whether it is possible to steer a system from a given initial state to any other state.
‘The second is how to determine the state of a dynamic system from observations of inputs and
outputs. These questions were posed and answered by Kalman, who also introduced the concepts
of controllability and observability. ‘The systems are allowed to be multiple-input-multiple-output
systems.
Controllability and Reachability
Consider the system
x(k +1) = Ga(k) + Hulk),
x(k) = Cx(h). (3.2)
‘Assume that the initial state z(0) is given. ‘The state at time n, where n is the order of the
system is given by
a(n) = G"2(0) + G*-1Hie(0) +--+ + Hu(n — 1)
=6%2(0) + Wo G3) Here T denotes the vector or matrix transposition.
Tf We has full rank n, then it is possible to find n linear algebraic equations from which the
control signal can be found such that the initial state is tranisferred to the desired final state 2(n).
Notice that the solution is not unique if there is more than one input signal.
Controllability The system (3.2) is said to be controllable if for any initial condition 2(0)
there exists control sequence (0), u(1),...,u(n —1) such that 2(n) = 0. In other words, the
origin can be reached from any initial state.
Reachability The system (3.2) is said to be reachable if for any initial condition 2(0) and
any vector a € R* there exists a control sequence u(0),u(1),..-,u(n —1) such that 2(n) =a. In
other words, any final state can be reached from any initial state.
‘The following statement follows immediately from the definitions.
‘Theorem The system (3.2) is reachable if and only if the matrix W has rank n.
It should be pointed out, that in the ase of continuous-time systems, reachability and con-
trollability are equivalent. In the case of diserete-time systems, controllability obviously follows
from reachability. However, the following example shows that reachability does not follow from
controllability.
Example 3.4 A controllable system which is not reachable _
‘The system
00 0
ak+D= a(h)+ u(t)
is not reachable because the matrix has rank 1
