Digital Control

CSE421

Asst. Prof. Dr.Ing.

Mohammed Nour A. Ahmed
[email protected] ω0 = 0.5π/T
@ ζ = 0.2
https://mnourgwad.github.io @
R ✁ ω0 = 0.3π/T
✁ ✙✟
✟
✁☛✁
Lecture 14: Pole placement (Regulator
Problem) ζ = 0.5
!
!
!
✠

Copyright ©2016 Dr.Ing. Mohammed Nour Abdelgwad Ahmed as part of

the course work and learning material. All Rights Reserved.
Where otherwise noted, this work is licensed under a Creative Commons
Attribution-NonCommercial-ShareAlike 4.0 International License.

Zagazig University | Faculty of Engineering | Computer and Systems Engineering Department | Zagazig, Egypt
 Lecture 14
Pole placement
(Regulator Problem)

• Introduce the concept of controllability of dynamic systems.

• Explain pole placement using the state feedback approach.

1
Controllability

The system (Φ, Γ) is controllable if there is a control input sequence

u(0), u(1) . . . u(n-1) that can move the system from an arbitrary
initial state x(0) = xi to an arbitrary final state x(n) = xf.

Controllable system uncontrollable system: there is no input

to move the system from state xi to xf

2
• The system state equation is

x(k + 1) = Fx(k ) + Gu (k )
• Iterating this equation for n steps, we find

x(1) = Fxi + Gu (0)

x(2) = Fx(1) + Gu (1)
= F 2 xi + FGu (0) + Gu (1)
!
éu (n - 1)ù
ê ! ú
x(n) = F n xi + G FG " F n -1G ê
[ ] ú
ê u (1) ú
ê ú
ë u (0) û
3
• To have x(n) = xf, we must be able to solve the following n equations for the control
sequence u(k).
éu (n - 1)ù
ê ! ú
[G FG " F n-1G]êê u(1) úú = x f - F n xi
ê ú
ë u (0) û
• Note that, for single-input systems, the following matrix is of size n x n.

Dc = [G FG ! F G n -1
]
• The system is controllable if the above equations have a solution for the control
sequence. This is achieved if the matrix Δc , called the controllability matrix, is
nonsingular or, more generally, has full rank n.

• Matrix rank = the number of linearly independent columns or rows.

4
Example 1

• Check the controllability of the system

æ1 0ö æ1ö
x(k + 1) = çç ÷÷ x(k ) + çç ÷÷u (k )
è0 2ø è1ø
The controllability matrix :
æ1 1 ö
D c = [G FG ] = çç ÷÷ Þ rank = 2 Þ system is controllable
è1 2 ø

• MATLAB functions to compute the controllability matrix

and check its rank:
CO = ctrb(A,B)
rank(CO)
5
Example 2
• Check the controllability of the system
æ1 0ö æ1ö
x(k + 1) = çç ÷÷ x(k ) + çç ÷÷u (k )
è0 1ø è1ø
æ1 1ö
D c = [G FG ] = çç ÷÷ Þ rank = 1 Þ system is not controllable
è1 1ø
• In order to understand why this system is not controllable, rewrite the
state equation as:
x1 (k + 1) = x1 (k ) + u (k )
x2 (k + 1) = x2 (k ) + u (k )
• If x1(0) = x2(0), then x1(k) = x2(k) for all k. Suppose that we want x1f ≠ x2f,
there is no input sequence to achieve this final state. Hence, this system is
not completely state controllable. 6
State feedback & Pole Placement

• In the state feedback approach, instead of using controllers with fixed

configuration in the forward path,
o control signal u(k) is a calculated as a linear combination of the measured state
variables.

• Using state feedback, the poles or eigenvalues of the closed-loop

system can be placed at specified locations (also called pole
assignment or allocation).

• Poles can arbitrarily be placed if and only if the system is controllable.

7
State feedback block diagram
• A linear system (Φ,Γ,C) with constant state feedback gain matrix K :

• The control signal u(k) is simply a linear combination of all state variables:
u (k ) = - Kx(k ).

• It is assumed, for now, that the reference input r(k) is zero (i.e. a regulator problem)
and that all the states are available for feedback – that is, we have access to the
complete state x(k) for all k.

• The objective of the state feedback design is to determine the matrix K

Ø for single-input systems, K is a row vector of size 1 x n. 8

• The linear system and the feedback control law are given by:
x(k + 1) = Fx(k ) + Gu (k ),
y (k ) = Cx(k )
u (k ) = - Kx(k )

• Substituting by u(k) in the state equation yields the closed-loop state

equation
F cl
$!# !
"
x(k + 1) = (F - GK ) x(k )
y (k ) = Cx(k )

Where Φcl = Φ-ΓK, is the state matrix of the closed-loop system.

9
Theorem: State feedback

If the pair (Φ,Γ) is controllable, then

there exists a feedback gain matrix K that arbitrarily assigns the closed-loop
system poles to any set [λ1, λ2, …, λn].

That is the eigenvalues of the closed-loop state matrix

F cl = F - GK
can be arbitrarily assigned.

10
pole placement by equating coefficients
1. Evaluate the desired characteristic polynomial from the specified
eigenvalues using the expression
n

Õ (l - l ) = 0
i =1
i

2. Evaluate the closed-loop characteristic polynomial using the expression

det[lI - (F - GK )] = 0.

3. Compare the two polynomials in 1 & 2 to get the entries of the gain matrix
K.

11
Example 3
Assign the eigenvalues {0.3± j0.2} to the pair
æ0 1ö æ0ö
F = çç ÷÷, G = çç ÷÷
Solution: è 3 4ø è1ø
• For the given eigenvalues, the desired characteristic polynomial is

D = ( l - 0.3 + j 0.2)(l - 0.3 - j 0.2) = l2 - 0.6l + 0.13

• The closed-loop state matrix is

æ0 1ö æ0ö æ 0 1 ö
F cl = F - GK = çç ÷÷ - çç ÷÷(k1 k 2 ) = çç ÷÷
è 3 4ø è1ø è 3 - k1 4 - k2 ø

12
• The closed-loop characteristic polynomial is
l -1
| lI - Acl |= =0
- (3 - k1 ) l - (4 - k 2 )
Þ l2 - (4 - k 2 )l - (3 - k1 ) = 0
• Comparing with the desired characteristic polynomial,
D = l2 - 0.6l + 0.13
• Gives
k1 = 3.13, k 2 = 3.4

• To check, use MATLAB commands:

>> A = [0, 1; 3, 4];
>> B = [0; 1];
>> poles = [0.3+j*0.2, 0.3–j*0.2];
>> K = place(A, B, poles)
K = 3.13 3.40 13
Controllable canonical form

• The algebra for finding the state feedback gain matrix K for systems
with n > 2 becomes quite tedious.
• However, it is specially simple if the system matrices happen to be in
the controllable canonical form.

• For a third order system, this form is:

éa1 a2 a3 ù é1ù
F = êê 1 0 0 úú, G = êê0úú, C = [b1 b2 b3 ]
êë 0 1 0 úû êë0úû

14
Example 4

Design a feedback controller for the pair

æ 1 2 3ö æ1ö
ç ÷ ç ÷
F = ç 1 0 0 ÷, G = ç 0 ÷
ç0 1 0÷ ç0÷
è ø è ø

to obtain the eigenvalues {0.1, 0.4±j0.4}.

15
Answer
• For the given eigenvalues, the desired characteristic polynomial is
D = (l - 0.1)(l - 0.4 + j 0.4)(l - 0.4 - j 0.4)
= l3 - 0.9l2 + 0.4l - 0.032.

• The closed-loop state matrix is

æ 1 2 3ö æ 1ö æ1 - k1 2 - k 2 3 - k3 ö
ç ÷ ç ÷ ç ÷
F cl = F - GK = ç 1 0 0 ÷ - ç 0 ÷(k1 k 2 k3 ) = ç 1 0 0 ÷
ç0 1 0÷ ç0÷ ç 0 1 0 ÷ø
è ø è ø è
l - 1 + k1 k 2 - 2 k3 - 3
lI - F cl = -1 l 0 = l2 (l - 1 + k1 ) + l (k 2 - 2) + (k3 - 3)
0 -1 l
= l3 + (k1 - 1)l2 + (k 2 - 2)l + (k3 - 3)
Þ k1 = 0.1, k 2 = 2.4, k3 = 2.968
16
Selecting desired pole locations

• The locations of the poles (eigenvalues) in the z-plane are

directly related to the transient response of the system.

• For example, the smaller is the distance of the pole from the
origin, the faster is the response associated with it.

• Also, the contours of constant damping ratio (which

determines percent overshoot) are spirals as shown.

17
Note on selecting desired pole locations!

• Seeking a closed-loop response that is much faster than the slowest

component response (often the plant) will lead to high gains for the state
feedback matrix and consequently to a high control effort which may
saturate one or more components.

• For example, if all closed-loop eigenvalues are placed at the origin of the
complex plane (i.e. deadbeat control), the resulting control law can
assume unacceptably high values.

18
Thanks for your attention.
Questions?

Asst. Prof. Dr.Ing.

Mohammed Nour A. Ahmed
[email protected]
Robotics Research Interest Group (zuR2 IG)
https://mnourgwad.github.io Zagazig University | Faculty of Engineering |
Computer and Systems Engineering Department
| Zagazig, Egypt

Copyright ©2016 Dr.Ing. Mohammed Nour Abdelgwad Ahmed as part of the course work and learning material. All Rights Reserved.
Where otherwise noted, this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Mohammed Ahmed (Asst. Prof. Dr.Ing.) Digital Control 2/1