Discrete-Time Linear Systems: Automatic Control 1

Lecture: Discrete-time linear systems
Automatic Control 1
Discrete-time linear systems
Prof. Alberto Bemporad
University of Trento
Academic year 2010-2011
Prof. Alberto Bemporad (University of Trento) Automatic Control 1 Academic year 2010-2011 1 / 32
Lecture: Discrete-time linear systems Introduction
Introduction
0 1 2 3 4 5
1.5
2
2.5
3
3.5
y(t), y(kT
s
)
time t
0 1 2 3 4 5
1
1.5
2
2.5
3
3.5
4
u(kT
s
)
time t
Sampling of a continuous signal Discrete-time signal
Discrete-time models describe relationships between sampled variables
x(kT
s
), u(kT
s
), y(kT
s
), k = 0, 1, . . .
The value x(kT
s
) is kept constant during the sampling interval [kT
s
, (k +1)T
s
)
A discrete-time signal can either represent the sampling of a continuous-time
signal, or be an intrinsically discrete signal
Discrete-time signals are at the basis of digital controllers (as well as of digital
lters in signal processing)
Lecture: Discrete-time linear systems Difference equations
Difference equation
Consider the rst order difference equation (autonomous system)
_
x(k +1) = ax(k)
x(0) = x
0
The solution is x(k) = a
k
x
0
0 1 2 3 4 5 6 7 8 9 10
0
0.5
1
1.5
2
2.5
3
x
(
k
)
k
a>1
a=1
0<a<1
Difference equation
First-order difference equation with input (non-autonomous system)
_
x(k +1) = ax(k) +bu(k)
x(0) = x
0
The solution has the form
x(k) = a
k
x
0
._.
natural response
+
k1
i=0
a
i
bu(k 1 i)
. _ .
forced response
The natural response depends on the initial condition x(0), the forced response
on the input signal u(k)
Example - Wealth of a bank account
k: year counter
: interest rate
x(k): wealth at the beginning of year k
u(k): money saved at the end of year k
x
0
: initial wealth in bank account
Discrete-time model:
_
x(k +1) = (1 +)x(k) +u(k)
x(0) = x
0
x
0
10 k
u(k) 5 k
10 %
x(k) = (1.1)
k
10 +
1 (1.1)
k
1 1.1
5 = 60(1.1)
k
50
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
5
10
15
20
25
30
35
40
45
50
Stored amount of money (keur)
x
(
k
)
k (years)
Linear discrete-time system
Consider the set of n rst-order linear difference equations forced by the
input u(k)
_
_
x
1
(k +1) = a
11
x
1
(k) + . . . +a
1n
x
n
(k) +b
1
u(k)
x
2
(k +1) = a
21
x
1
(k) + . . . +a
2n
x
n
(k) +b
2
u(k)
.
.
.
.
.
.
.
.
.
x
n
(k +1) = a
n1
x
1
(k) + . . . +a
nn
x
n
(k) +b
n
u(k)
x
1
(0) = x
10
, . . . x
n
(0) = x
n0
In compact matrix form:
_
x(k +1) = Ax(k) +Bu(k)
x(0) = x
0
where x =
_
_
x
1
.
.
.
x
n
_
_

n
.
Linear discrete-time system
The solution is
x(k) = A
k
x
0
._.
natural response
+
k1
i=0
A
i
Bu(k 1 i)
. _ .
forced response
If matrix A is diagonalizable, A = TT
1
=
_
_
_
1
0 ... 0
0
2
... 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 ...
n
_
_
_
A
k
= T
_
_
_
_
k
1
0 ... 0
0
k
2
... 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 ...
k
n
_
_
_
_
T
1
where T = [v
1
. . . v
n
] collects n independent eigenvectors.
Lecture: Discrete-time linear systems Modal response
Modal response
Assume input u(k) = 0, k 0
Assume A is diagonalizable, A = TT
1
The state trajectory (natural response) is
x(k) = A
k
x
0
= T
k
T
1
x
0
=
n
i=1
k
i
v
i
where
i
= eigenvalues of A
v
i
= eigenvectors of A
i
= coefcients that depend on the initial condition x(0)
=
_
_
1
.
.
.
n
_
_
= T
1
x(0), T = [v
1
. . . v
n
]
The system modes depend on the eigenvalues of A, as in continuous-time
Lecture: Discrete-time linear systems Modal response
Example
Consider the linear discrete-time system
_
_
_
x
1
(k +1) =
1
2
x
1
(k) +
1
2
x
2
(k)
x
2
(k +1) = x
2
(k) +u(k)
x
1
(0) = 1
x
2
(0) = 1
_
_
_
x(k +1) =
_
1
2
1
2
0 1
_
x(k) +
_
0
1
_
u(k)
x(0) =
_
1
1
_
Eigenvalues of A:
1
=
1
2
,
2
= 1
Solution:
x(k) =
_
1
2
1
2
0 1
_
k
_
1
1
_
+
k1
i=0
_
1
2
1
2
0 1
_
i _
0
1
_
u(k 1 i)
=
_
1
2
k
1
1
2
k
0 1
_
_
1
1
_
+
k1
i=0
_
1
1
2
i
1
_
u(k 1 i)
=
_
1
_
1
2
_
k1
1
_
. _ .
natural response
+
k1
i=0
_
1
1
2
i
1
_
u(k 1 i)
. _ .
forced response
0 1 2 3 4 5 6 7 8 9 10
1.5
1
0.5
0
0.5
1
1.5
x
1
(k),x
2
(k)
step k
simulation for u(k) 0
Lecture: Discrete-time linear systems Linear difference equations
n
th
-order difference equation
Consider the n
th
-order difference equation forced by u
a
n
y(k n) +a
n1
y(k n +1) + +a
1
y(k 1) +y(k) =
b
n
u(k n) + +b
1
u(k 1) +b
0
u(k)
Equivalent linear discrete-time system in canonical state matrix form
_
_
x(k +1) =
_
_
_
_
_
_
_
_
0 1 0 . . . 0
0 0 1 . . . 0
.
.
.
.
.
.
.
.
.
0 0 0 . . . 1
a
n
a
n1
a
n2
. . . a
1
_
_
_
_
_
_
_
_
x(k) +
_
_
_
_
_
_
_
_
0
0
.
.
.
0
1
_
_
_
_
_
_
_
_
u(k)
y(k) =
_
(b
n
b
0
a
n
) . . . (b
1
b
0
a
1
)
_
x(k) +b
0
u(k)
There are innitely many state-space realizations
MATLAB
tf2ss
n
th
-order difference equations are very useful for digital lters, digital
controllers, and to reconstruct models from data (system identication)
Lecture: Discrete-time linear systems Discrete-time linear systems
Discrete-time linear system
_
_
_
x(k +1) = Ax(k) +Bu(k)
y(k) = Cx(k) +Du(k)
x(0) = x
0
Given the initial condition x(0) and the input sequence u(k), k , it is
possible to predict the entire sequence of states x(k) and outputs y(k), k
The state x(0) summarizes all the past history of the system
The dimension n of the state x(k)
n
is called the order of the system
The system is called proper (or strictly causal) if D = 0
General multivariable case:
x(k)
n
u(k)
m
y(k)
p
A
nn
B
nm
C
pn
D
pm
Example - Student dynamics
Problem Statement:
3-years undergraduate course
percentages of students promoted, repeaters, and dropouts are roughly constant
direct enrollment in 2nd and 3rd academic year is not allowed
students cannot enrol for more than 3 years
Notation:
k Year
x
i
(k) Number of students enrolled in year i at year k, i = 1, 2, 3
u(k) Number of freshmen at year k
y(k) Number of graduates at year k
i
promotion rate during year i, 0
i
1
i
failure rate during year i, 0
i
1
i
dropout rate during year i,
i
= 1
i
i
0
3
rd
-order linear discrete-time system:
_
_
_
x
1
(k +1) =
1
x
1
(k) +u(k)
x
2
(k +1) =
1
x
1
(k) +
2
x
2
(k)
x
3
(k +1) =
2
x
2
(k) +
3
x
3
(k)
y(k) =
3
x
3
(k)
Example - Student dynamics
In matrix form
_
_
x(k +1) =
_
_
_
1
0 0
1

2
0
0
2

3
_
_
_
x(k) +
_
_
_
1
0
0
_
_
_
u(k)
y(k) =
_
0 0
3
_
x(k)
Simulation
1
= .60
1
= .20
2
= .80
2
= .15
3
= .90
3
= .08
u(k) 50, k = 2012, . . .
2012 2014 2016 2018 2020
0
20
40
y(k)
step k
2012 2014 2016 2018 2020
49
50
51
u(k)
step k
Example - Supply chain
y(k)
u(k)
x1(k)
x2(k) x3(k)
S
P R
!1x1(k)
"1x1(k)
!2x2(k)
"2x2(k)
#3x3(k)
$3x3(k)
Problem Statement:
S purchases the quantity u(k) of raw material at each month k
a fraction
1
of raw material is discarded, a fraction
1
is shipped to producer P
a fraction
2
of product is sold by P to retailer R, a fraction
2
is discarded
retailer R returns a fraction
3
of defective products every month, and sells a
fraction
3
to customers
Mathematical model:
_
_
_
x
1
(k +1) = (1
1
1
)x
1
(k) +u(k)
x
2
(k +1) =
1
x
1
(k) + (1
2
2
)x
2
(k) +
3
x
3
(k)
x
3
(k +1) =
2
x
2
(k) + (1
3
3
)x
3
(k)
y(k) =
3
x
3
(k)
k month counter
x
1
(k) raw material in S
x
2
(k) products in P
x
3
(k) products in R
y(k) products sold to customers
Lecture: Discrete-time linear systems Stability of discrete-time linear systems
Equilibrium
Consider the discrete-time nonlinear system
_
x(k +1) = f (x(k), u(k))
y(k) = g(x(k), u(k))
Denition
A state x
r

n
and an input u
r

m
are an equilibrium pair if for
initial condition x(0) = x
r
and constant input u(k) u
r
, k ,
the state remains constant: x(k) x
r
, k
Equivalent denition: (x
r
, u
r
) is an equilibrium pair if f (x
r
, u
r
) = x
r
x
r
is called equilibrium state, u
r
equilibrium input
The denition generalizes to time-varying discrete-time nonlinear systems
Stability
Consider the nonlinear system
_
x(k +1) = f (x(k), u
r
)
y(k) = g(x(k), u
r
)
and let x
r
an equilibrium state, f (x
r
, u
r
) = x
r
Denition
The equilibrium state x
r
is stable if for each initial conditions x(0) close
enough to x
r
, the corresponding trajectory x(k) remains near x
r
for all k
a
a
Analytic denition: > 0 > 0 : x(0) x
r
< x(k) x
r
< , k
The equilibrium point x
r
is called asymptotically stable if it is stable and
x(k) x
r
for k
Otherwise, the equilibrium point x
r
is called unstable
Stability of rst-order linear systems
Consider the rst-order linear system
x(k +1) = ax(k) +bu(k)
x
r
= 0, u
r
= 0 is an equilibrium pair
For u(k) 0, k = 0, 1, . . ., the solution is
x(k) = a
k
x
0
The origin x
r
= 0 is
unstable if |a| > 1
stable if |a| 1
asymptotically stable if |a| < 1
0 2 4 6 8 10
0
0.5
1
1.5
2
2.5
3
x
(
k
)
k
x
r
x
0
a>1
a=1
0<a<1
a=0
Stability of discrete-time linear systems
Since the natural response of x(k +1) = Ax(k) +Bu(k) is x(k) = A
k
x
0
, the stability
properties depend only on A. We can therefore talk about system stability of a
discrete-time linear system (A, B, C, D)
Theorem:
Let
1
, . . .,
m
, m n be the eigenvalues of A
nn
. The system
x(k +1) = Ax(k) +Bu(k) is
asymptotically stable iff |
i
| < 1, i = 1, . . . , m
(marginally) stable if |
i
| 1, i = 1, . . . , m, and the eigenvalues with unit
modulus have equal algebraic and geometric multiplicity
a
unstable if i such that |
i
| > 1
a
Algebraic multiplicity of
i
= number of coincident roots
i
of det(I A). Geometric
multiplicity of
i
= number of linearly independent eigenvectors v
i
, Av
i
=
i
v
i
The stability properties of a discrete-time linear system only de-
pend on the modulus of the eigenvalues of matrix A
Stability of discrete-time linear systems
Proof:
The natural response is x(k) = A
k
x
0
If matrix A is diagonalizable
1
, A = TT
1
,
=
_
_
_
_
1
0 ... 0
0
2
... 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 ...
n
_
_
_
_
A
k
= T
_
_
_
_
_
_
k
1
0 ... 0
0
k
2
... 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 ...
k
n
_
_
_
_
_
_
T
1
Take any eigenvalue = e
j
:
|
k
| =
k
|e
jk
| =
k
A is always diagonalizable if algebraic multiplicity - geometric multiplicity
1
If A is not diagonalizable, it can be transformed to Jordan form. In this case the natural
response x(t) contains modes k
j
k
, j = 0, 1, . . . , alg. multiplicity = geom. multiplicity
Example 1
_
_
_
x(k +1) =
_
0 0
1
1
2
_
x(k)
x(0) =
_
x
10
x
20
_
eigenvalues of A:
_
0,
1
2
_
solution:
_
x
1
(k) = 0, k = 1, 2, . . .
x
2
(k) =
_
1
2
_
k1
x
10
+
_
1
2
_
k
x
20
, k = 1, 2, . . .
0 2 4 6 8 10
0.5
0
0.5
1
x
1
(
k
)
k
0 2 4 6 8 10
1
0.5
0
0.5
1
1.5
x
2
(
k
)
k
asymptotically stable
Example 2
_
_
_
x(k +1) =
_
0 1
1 0
_
x(k)
x(0) =
_
x
10
x
20
_
eigenvalues of A: {+j, j}
solution:
_
x
1
(k) = x
10
cos
k
2
+x
20
sin
k
2
, k = 0, 1, . . .
x
2
(k) = x
10
sin
k
2
+x
20
cos
k
2
, k = 0, 1, . . . ,
0 2 4 6
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
x
1
(
k
)
k
0 2 4 6
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
x
2
(
k
)
k
marginally stable
Example 3
_
_
_
x(k +1) =
_
1 1
0 1
_
x(k)
x(0) =
_
x
10
x
20
_
eigenvalues of A: {1, 1}
_
x
1
(k) = x
10
+x
20
k, k = 0, 1, . . .
x
2
(k) = x
20
, k = 0, 1, . . .
Note: A is not diagonalizable !
0 2 4 6
3
2
1
0
1
2
3
4
5
6
7
x
1
(
k
)
k
0 1 2 3 4 5
2
1.5
1
0.5
0
0.5
1
1.5
2
x
2
(
k
)
k
unstable
Example 4
_
_
_
x(k +1) =
_
2 0
1 0
_
x(k)
x(0) =
_
x
10
x
20
_
eigenvalues of A: {0, 2}
solution:
_
x
1
(k) = 2
k
x
10
, k = 0, 1, . . .
x
2
(k) = 2
k1
x
10
, k = 1, 2, . . .
0 2 4 6
40
20
0
20
40
60
80
x
1
(
k
)
k
0 1 2 3 4 5
40
20
0
20
40
60
80
x
2
(
k
)
k
unstable
Zero eigenvalues
Modes
i
=0 determine nite-time convergence to zero.
This has no continuous-time counterpart, where instead all converging
modes tend to zero in innite time (e
i
t
)
Example: dynamics of a buffer
y(k)
u(k)
x
2
(k) x
1
(k) x
3
(k)
_
_
_
x
1
(k +1) = x
2
(k)
x
2
(k +1) = x
3
(k)
x
3
(k +1) = u(k)
y(k) = x
1
(k)
_
_
x(k +1) =
_
_
_
0 1 0
0 0 1
0 0 0
_
_
_
x(k) +
_
_
_
0
0
1
_
_
_
u(k)
y(k) =
_
1 0 0
_
x(k)
Natural response: A
3
x(0) = 0 for all x(0)
3
For u(k) 0, the buffer deploys after at most 3 steps !
Summary of stability conditions for linear systems
system continuous-time discrete-time
asympt. stable i = 1, . . . , n (
i
) < 0 |
i
| < 1
unstable i such that (
i
) > 0 |
i
| > 1
stable i, . . . , n (
i
) 0 |
i
| 1
and
i
such that (
i
) = 0 |
i
| = 1
algebraic = geometric mult.
Lecture: Discrete-time linear systems Sampling continuous-time systems
Exact sampling
Consider the continuos-time system
_
_
_
x(t) = Ax(t) +Bu(t)
y(t) = Cx(t) +Du(t)
x(0) = x
0
We want to characterize the value of x(t), y(t) at the time instants
t = 0, T
s
, 2T
s
, . . . , kT
s
, . . ., under the assumption that the input u(t) is
constant during each sampling interval (zero-order hold, ZOH)
u(t) = u(k), kT
s
t < (k +1)T
s
x(k) x(kT
s
) and y(k) y(kT
s
) are the state
and the output samples at the k
th
sampling
instant, respectively
0 2 4 6 8 10
1
0.5
0
0.5
1
1.5
y(t), y(kT
s
)
time t
0 2 4 6 8 10
2
1.5
1
0.5
0
0.5
1
1.5
2
u(t), u(kT
s
)
time t
Exact sampling
Let us evaluate the response of the continuous-time system between time
t
0
= kT
s
and t = (k +1)T
s
from the initial condition x(t
0
) = x(kT
s
) using
Lagrange formula:
x(t) = e
A(tt
0
)
x(t
0
)+
_
t
t
0
e
A(t)
Bu()d = e
A((k+1)T
s
kT
s
)
x(kT
s
)+
_
(k+1)T
s
kT
s
e
A((k+1)T
s
)
Bu()d
Since the input u(t) is piecewise constant, u() u(k), kT
s
< (k +1)T
s
.
By setting = kT
s
we get
x((k +1)T
s
) = e
AT
s
x(kT
s
) +
_
_
T
s
0
e
A(T
s
)
d
_
Bu(kT
s
)
and hence
x(k +1) = e
AT
s
x(k) +
_
_
T
s
0
e
A(T
s
)
d
_
B u(k)
which is a linear difference relation between x(k) and u(k) !
Exact sampling
The discrete-time system
_
x(k +1) =

Ax(k) +

B u(k)
y(k) =

Cx(k) +

D u(k)
depends on the original continuous-time system through the relations
A e
AT
s
,

B
_
_
T
s
0
e
A(T
s
)
d
_
B,

C C,

D D
If u(t) is piecewise constant, (
A,
B,
C,

D) provides the exact evolution of state
and output samples at discrete times kT
s
MATLAB
sys=ss(A,B,C,D);
sysd=c2d(sys,Ts);
[Ab,Bb,Cb,Db]=ssdata(sysd);
0 1 2 3 4 5 6 7 8 9 10
0.4
0.2
0
0.2
0.4
y(t), y(kT
s
)
time t
0 1 2 3 4 5 6 7 8 9 10
1
0.5
0
0.5
1
u(t), u(kT
s
)
time t
Exact sampling
Rule of thumb: T
s
1
10
of the rise time = time to move from 10%
to 90% of the steady-state value, for input u(t) 1, x(0) = 0
0 1 2 3 4 5 6 7 8 9 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
rise time
time t
o
u
t
p
u
t
rise time
10 %
90 %
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
0.2
0.4
0.6
0.8
1
y(t), y(kT
s
)
time t
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
0.5
1
1.5
2
u(t), u(kT
s
)
time t
More on the choice of sampling time in the second part of the course ...
Approximate sampling - Eulers method
x(kT
s
)
x((k +1)T
s
) x(kT
s
)
T
s
_ x(t)
x((k+1)T)x(kT)
T
x((k+1)T)
x(kT)
t
kT (k+1)T
x(t)
T
x((k + 1)T
s
)
x(kT
s
)
(k + 1)T
s kT
s
x(kT
s
)
x((k + 1)T
s
) x(kT
s
)
T
s
T
s
Leonhard Paul Euler
(1707-1783)
For nonlinear systems x(t) = f (x(t), u(t)):
x(k +1) = x(k) +T
s
f (x(k), u(k))
For linear systems x(t) = Ax(t) +Bu(t):
x((k +1)T
s
) = (I +T
s
A)x(kT
s
) +T
s
Bu(kT
s
)
A I +AT
s
,

B T
s
B,

C C,

D D
Note that e
T
s
A
= I +T
s
A +. . . +
T
n
s
A
n
n!
+. . .
Therefore when T
s
is small Eulers method and exact sampling are similar
Example - Hydraulic system
Continuous time:
_
_
_
d
dt
h(t) =
a
2g
A
_
h(t) +
1
A
u(t)
q(t) = a
_
2g
_
h(t)
Discrete time:
_
_
_
h(k +1) =

h(k)
T
s
a
2g
A
_
h(k) +
T
s
A
u(k)
q(k) = a
_
2g
_
h(k)
h
u
q
0 5 10 15 20 25 30
0
1
2
3
4
5
6
7
level h(t) (m)
time (s)
continuous time
Euler approximation
continuous time
Euler approximation
continuous time
Euler approximation
English-Italian Vocabulary
discrete-time linear systems sistemi lineari a tempo discreto
sampling interval tempo (o intervallo) di campionamento
difference equation equazione alle differenze
zero-order hold mantenitore di ordine zero
piecewise constant costante a tratti
rise time tempo di salita
Translation is obvious otherwise.

Discrete-Time Linear Systems: Automatic Control 1

Uploaded by

Document Information

Original Title

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

Discrete-Time Linear Systems: Automatic Control 1

Uploaded by

Copyright:

Available Formats

Lecture: Discrete-time linear systems

You might also like