Digital Control Systems

(EP-5511)

Lecture-11
Pole Placement and Observer Design

Belaynesh Belachew (capt.)

Defence Engineering College

1
 Introduction
Controllability and observability are two important
properties of state models which are to be studied prior
to designing a controller.
Controllability deals with the possibility of forcing the
system to a particular state by application of a control
input. If a state is uncontrollable then no input will be
able to control that state.
On the other hand whether or not the initial states can be
observed from the output is determined using
observability property. Thus if a state is not observable
then the controller will not be able to determine its
behavior from the system output and hence not be able
2 to use that state to stabilize the system.
 Controllability
Before going to any details, we would first formally
define controllability. Consider a dynamical system
x(k  1)  Gx(k )  Hu (k )
y (k )  Cx(k )  Du (k )
where G ϵ Rn×n, H ϵ Rn×m, C ϵ Rp×n, D ϵ Rp×m.
Complete State Controllability: The state equation (or
the pair (G, H) ) is said to be completely state
controllable or simply state controllable if for any
initial state x(0) and any final state x(N), there exists
an input sequence u(k), k = 0, 1, 2, · · · ,N, which
transfers x(0) to x(N) for some finite N. Otherwise
3 the state equation is state uncontrollable.
 Continued
Complete Output Controllability: The system given in
equation is said to be completely output controllable or
simply output controllable if any final output y(N) can
be reached from any initial state x(0) by applying an
unconstrained input sequence u(k), k = 0, 1, 2, · · · ,N,
for some finite N. Otherwise it is not output
controllable.
Theorems on controllability
State Controllability:
1. The state equation or the pair (G,H) is state controllable
if and only if the n × nm state controllability matrix
UC = [H GH G2H ...... Gn−1H] has rank n, i.e., full row
rank.
4
 Continued
2. The state equation is controllable if the n × n
controllability grammian matrix
N 1 N 1
Wc   G i HH T (G i )T   G N 1i HH T (G N 1i )T
i 0 i 0

is non-singular for any nonzero finite N.

3. If the system has a single input and the state model is
in controllable canonical form then the system is
controllable.
4. When G has distinct eigenvalues and in Jordan/
Diagonal canonical form, the state model is
controllable if and only if all the rows of H are
5 nonzero.
 Continued
5. When G has multiple order eigenvalues and in Jordan
canonical form, then the state model is controllable if
and only if
i. each Jordan block corresponds to one distinct
eigenvalue and
ii. the elements of H that correspond to last row of each
Jordan block are not all zero.
Output Controllability: The system in equation is
completely output controllable if and only if the p ×
(n + 1)m output controllability matrix
UOC = [D CH CGH CG2H ...... CGn−1H]
has rank p, i.e., full row rank.
6
 Continued
Controllability to the origin and Reachability
There exist three different definitions of state
controllability:
1. Input transfers any state to any state. This definition
is adopted in this course.
2. Input transfers any state to zero state. This is called
controllability to the origin.
3. Input transfers zero state to any state. This is referred
as controllability from the origin or reachability.
Above three definitions are equivalent for continuous
time system. For discrete time systems definitions
(1) and (3) are equivalent but not the second one.
7
 Continued
Example 1: Consider the system x(k+1) = Gx(k) + Hu(k),
y(k) = Cx(k). where
G   2 1  , H  1 and C   0 1
 1 2 1
Show if the system is controllable. Find the transfer
function Y(z)/U(z). Can you see any connection
between controllability and the transfer function?
Solution: The controllability matrix is given by
UC   H GH   1 1
1 1
Its determinant ||UC|| = 0 => UC has a rank 1 which is less
than the order of the matrix, i.e., 2.
Thus the system is not controllable. 8
 Continued
The transfer function
1
Y ( z)  z  2 1  1 1
G( z )   C ( zI  G ) H  0 1 
1

 1 z  1
U ( z)   1 z  2  
Although state model is of order 2, the transfer function
has order 1.
The eigenvalues of G are λ1 = −1 and λ2 = −3.
This implies that the transfer function is associated with
pole-zero cancellation for the pole at −3.
Since one of the dynamic modes is cancelled, the system
became uncontrollable.
9
 Observability
The state model (or the pair (G,C) ) is said to be
observable if any initial state x(0) can be uniquely
determined from the knowledge of output y(k) and
input sequence u(k), for k = 0, 1, 2, · · · ,N, where N
is some finite time. Otherwise the state model is
unobservable.
Theorems on observability
1. The state model or the pair (G,C) is observable if the
np × n observability matrix
UO = [C CG CG2 ...... CGn−1]T
has rank n, i.e., full column rank.
2. The state model is observable if the n×n observability
10 grammian matrix
 Continued
N 1 N 1
WO   (G i )T CC T G i   (G N 1i )T C T CG N 1i
i 0 i 0

is non-singular for any nonzero finite N.

3. If the state model is in observable canonical form then
the system is observable.
4. When G has distinct eigenvalues and in Jordan/
Diagonal canonical form, the state model is observable
if and only if none of the columns of C contain zeros.
5. When G has multiple order eigenvalues and in Jordan
canonical form, then the state model is observable if
and only if
i. each Jordan block corresponds to one distinct eigenvalue ;
ii. the elements of C that correspond to first column of each
11 Jordan block are not all zero.
 Continued
Theorem of Duality: The pair (G,H) is controllable if
and only if the pair (GT,HT) is observable.
Loss of controllability or observability due to pole-zero
Cancellation: We have already seen through an
example that a system becomes uncontrollable when
one of the modes is cancelled.
Let us take another example.
Example 2: 0 1 0
x(k  1)  x(k )  u (k );
 1 2 1 
y (k )  1 1 x(k )
The controllability matrix
U C  0 1 
1 2
12
implies that the state model is controllable.
 Continued
On the other hand, the observability matrix
UO   1 1 
 1 1
has a rank 1 which implies that the state model is
unobservable.
Now, if we take a different set of state variables so
that, ̅x1(k) = y(k), then the state variable model will
be: x (k  1)  y(k  1)
1

x1 (k  2)  y (k  2)   y (k )  2 y (k  1)  u (k  1)  u (k )

Lets us take ̅x2(k) = y(k+1)−u(k). The new state variable

model is: x (k  1)  x (k )  u (k )
1 2
13
x2 (k  1)   x1 (k )  2 x2 (k )  u (k )
 Continued
which implies
G   0 1  , H   1  , C  1 0
 1 2  1
The controllability matrix
U C   1 1
 1 1 
implies that the state model is uncontrollable.
The observability matrix 1 0 
UO 
0 1 
implies that the state model is observable.
The system difference equation will result in a transfer
function which would involve pole-zero
cancellation. Whenever there is a pole zero
cancellation, the state space model will be either
14 uncontrollable or unobservable or both.
 Design via Pole Placement
State Feedback Controller: Consider the state-space
model of a SISO system
x(k + 1) = Gx(k) + Hu(k)
y(k) = Cx(k)
In state feedback design, the states are fedback to the
input side to place the closed poles at desired
locations.
Regulation Problem: When we want the states to
approach zero starting from any arbitrary initial state,
the design problem is known as regulation where the
internal stability of the system, with desired
transients, is achieved.
Control input: u(k) = −Kx(k) 15
 Continued
Tracking Problem: When the output has to track a
reference signal, the design problem is known as
tracking problem.
Control input: u(k) = −Kx(k) + Nr(k).
where r(k) is the reference signal.
First designing a state feedback control law using pole
placement technique for regulation problem will
discussed.
By substituting the control law in the system state model,
the closed loop system becomes x(k+1)=
(G−HK)x(k).
If K can be designed such that eigenvalues of G−HK are
16
within the unit circle then the problem of regulation
will be solved.
 Continued
The control problem can thus be defined as: Design a
state feedback gain matrix K such that the control
law given places poles of the closed loop system
x(k+1) = (G−HK)x(k) in desired locations.
• A necessary and sufficient condition for arbitrary pole
placement is that the pair (G,H) must be controllable.
• Since the states are fedback to the input side, we assume
that all the states are measurable.
Designing K by Ackermann’s Formula: Consider the
state-space model of a SISO system.
The control input is u(k) = −Kx(k)
Thus the closed loop system will be
17 x(k + 1) = (G−HK)x(k) = Ĝx(k)
 Continued
Desired characteristic Equation:
|zI − G + HK| = |zI − Ĝ| = 0
or, (z − z1)(z − z2) · · · (z − zn) = 0
or, zn+α1zn-1+...+ αn−1z+ αn = 0
Using Cayley-Hamilton Theorem
Ĝn+α1Ĝn-1+...+αn−1 Ĝ+ αnI = 0
Consider the case when n = 3.
Ĝ = G − HK
Ĝ2=(G−HK)2=G2−GHK−HKG−HKHK=G2−GHK−HKĜ
Ĝ3 = (G − HK)3 = G3 −G2HK − GHKĜ − HKĜ2
Rewritten as α3I+α2Ĝ+α1Ĝ2 +Ĝ3 = 0
Or α3I + α2(G − HK) + α1(G2 − GHK − HKĜ) + G3 − G2HK
18
− GHKĜ − HKĜ2 = 0
 Continued
Thus:
Φ(G)=H(α2K+α1KĜ+KĜ2) +GH(α1K+KĜ)+ G2HK
 2 K  1 KGˆ  KGˆ 2 
  H GH G 2 H    K  KGˆ 
 1
K 
 
 2 K  1 KGˆ  KGˆ 2 
 UC   K  KG ˆ 
 1
K 
 
where ϕ(.) is the closed loop characteristic polynomial
and UC is the controllability matrix.
Since UC is nonsingular.
19
 Continued
 2 K  1 KGˆ  KGˆ 2 
U C 1 (G )    K  KG ˆ 
 1
K 
 
 2 K  1KGˆ  KGˆ 2 
or  0 0 1U C 1 (G)  0 0 1   K  ˆ
KG 
 1
K 
 
or K   0 0 1U C 1 (G)
Extending the above for any n,
or K   0 0 ... 1U C 1 (G );
20 where U C   H GH G 2 H ... G n 1H 
 Continued
Example 3: Find out the state feedback gain matrix K
for the following system using two different methods
such that the closed loop poles are located at 0.5, 0.6
and 0.7.
0 1 0 0 
x(k  1)   0 0 1  x(k )  0  u (k )
 1 2 3 1 
Solution 0 0 1 
U C  0 1 3
1 3 7 
The above matrix has rank 3, so the system is
controllable.
Open loop characteristic equation:
21 z3 + 3z2 + 2z + 1 = 0
 Continued
Desired characteristic equation:
(z − 0.5)(z − 0.6)(z − 0.7) = 0
or, z3 − 1.8z2 + 1.07z − 0.21 = 0
Since the open loop system is already in controllable
canonical form.
K=[α3−a3 α2−a2 α1 − a1]
where, α3=−0.21, α2=1.07, α1=−1.8 and a3=1, a2=2, a1= 3.
Thus
K = [−1.21 −0.93 −4.8]
Using Ackermann’s formula:
1  2 3 1  2 3 1 
U C 1   3 1 0    3 1 0
1  1 0 0  1 0 0
22
 Continued
Φ(G)=G3 − 1.8G2+1.07G − 0.21I
 1 2 3   0 0 1.8 
 3 5 7    1.8 3.6 5.4 
 7 11 10  5.4 9 10.6
 0 1.07 0   0.21 0 0 
 0 0 1.07    0 0.21 0 
 1.07 2.14 3.21  0 0 0.21
 1.21 0.93 4.8 
  4.8 8.39 13.47 
 13.47 22.14 30.02 
Thus  2 3 1
K   0 0 1U C 1 (G)  0 0 1 3 1 0   (G)
1 0 0
23
 1 0 0 (G)   1.21 0.93 4.8
24