Hassan Bevrani

Professor, University of Kurdistan

Fall 2023

H. Bevrani University of Kurdistan 1

Contents

1. Example: Sinning Satellite

2. Multivariable Feedback Control
3. Robust Control of Multivariable Systems
4. frequency Response of MIMO System
5. Singular Value Decomposition

6. Internal Stability
7. Relative Gain Array

H. Bevrani University of Kurdistan 2

 Reference

1. S. Skogestadand I. Postlethwaite, Multivariable Feedback Control; Analysis and

Design, Second Edition, Wiley, 2005.
2. M. Fujita, Lecture Notes on Feedback Control Systems, Tokyo Institute of
Technology, 2019.
3. R. Smith, Lecture Notes on Control Systems, ETH Zurich, 2020.

4. H. S. Tsien, Engineering Cybernetics, McGraw-Hill, 1954.

H. Bevrani University of Kurdistan 3

Robust Control of Multivariable Systems

Example: Spinning Satellite

(Ref 4)
H. Bevrani University of Kurdistan 4
 Spinning Satellite

(Ref 4)
H. Bevrani University of Kurdistan 5

Robust Control of Multivariable Systems

1. Diagonal PID Controller (Decentralized Control)

P(s)

pidtune
(Ref 1, pp. 91-93)

H. Bevrani University of Kurdistan 6

 Continue

P(s)

(Ref 1, pp. 91-93)

H. Bevrani University of Kurdistan 7

Robust Control of Multivariable Systems

2. Dynamic Decoupling (Inverse-based Controller )
Inverse-based

Loop Shaping Design

Target Loop (Desired Loop) √

H. Bevrani University of Kurdistan 8

 Robust Control of Multivariable Systems

Impact of Uncertainty

H. Bevrani University of Kurdistan 9

Robust Control of Multivariable Systems

2. Robust Controller

Robust
Controller

√ X

H. Bevrani University of Kurdistan 10

 Frequency Response for SISO Systems
Example:

H. Bevrani University of Kurdistan 11

Frequency Response for MIMO Systems

Example:

H. Bevrani University of Kurdistan 12

 Singular Value
Definition: The singular values of an m×n matrix A are the square roots of the
non-zero eigenvalues of the symmetric n×n matrix ATA or AAT listed with their
multiplicities in decreasing order σ1 ≥ σ2 ≥ · · · ≥ σn ≥ 0.

Example: Find the singular values for matrix A.

Solution:

l1 = 3
det(A T A - l I 2 ) = A T A - l I 2 = (2 - l )2 - 1 = 0. s1 = 3 s2 = 1
l2 = 1

H. Bevrani University of Kurdistan 13

Singular Value Decomposition (SVD)

Theorem: Let A be an m × n matrix with m ≥ n . Then there exist orthogonal
matrices U (m×m) and V (n×n) and a diagonal matrix Σ = #$%& '! , '" , … , '# (m ×
n) with order '! ≥ '" ≥ … ≥ '# ≥0, such that A holds

! = #Σ% !
The column vectors of U = [u1, . . . ,um] are called the left singular vectors and
similarly V = [v1, . . . , vn] are the right singular vectors of matrix A. The columns of
U and V are orthonormal. The matrix Σ is diagonal with positive real entries of '$ ,
and can be represented as:

H. Bevrani University of Kurdistan 14

 Singular Value Decomposition (SVD)
Example:

(Ref 1, A.3)
H. Bevrani University of Kurdistan 15

SVD

Matlab function: > [U, S, V ] = svd(A)

If !! is the smallest singular value greater than zero then the matrix A has rank r, and s r > 0 . In
this case U and V can be partitioned as U=[U1,U2] and V=[V1,V2], where U 1 = [u1, u2 ,..., ur )
and V 1 = [v1, v2 ,..., vr ) have r columns. Then A can be represented as reduced form of SVD as
follows

> [U, S, V ] = svd(A,0)

Example: Find full and reduced SVD for matrix A.

H. Bevrani University of Kurdistan 16

 SVD
Solution:
>> [U,S,V] = svd(A,0)
>> A=[0 1;1 1;1 0]
A=
U=
0 1
-0.4082 0.7071
1 1
-0.8165 0.0000
1 0
-0.4082 -0.7071
S=
>> [U,S,V] = svd(A)
1.7321 0
U=
0 1.0000
-0.4082 0.7071 0.5774
V=
-0.8165 0.0000 -0.5774
-0.7071 -0.7071
-0.4082 -0.7071 0.5774
-0.7071 0.7071
S=
1.7321 0
0 1.0000
0 0
V=
-0.7071 -0.7071
-0.7071 0.7071
H. Bevrani University of Kurdistan 17

Continue
Example:

σ-plot is the extension of Bode gain plot to MIMO Systems

SISO: Absolute value

MIMO: Singular value plot

(Ref 1, p. 79)
H. Bevrani University of Kurdistan 18
 Internal Stability

Example (SISO):
Unstable
Stable

Unstable pole/zero cancellation

In order to avoid pole/zero cancellation, consider input

injection and output measurement for each dynamic block
(Gang of Four).
Sensitivity: Complementary Sensitivity:

Load Sensitivity: Noise Sensitivity:

(Ref 1, p. 144)
H. Bevrani University of Kurdistan 19

Internal Stability
Nominal Stability:
: Vectors
: Transfer function matrices
Well-posedness:

(Gang of Four: well-defined and proper)

Theorem: Assume P, K contain no unstable hidden modes. Then, the feedback system in the
figure is internally stable if and only if all four closed-loop transfer matrices are stable.

Theorem: Assume shows the state space representation of above system. The system
is internally stable if and only if is stable.
(Ref 1, p. 145)
H. Bevrani University of Kurdistan 20
 Relative Gain Array (RGA)

Example:

element wise multiplication

Rule 1 √
Rule 2 √
(Ref 1, Sec. 3.4, p. 85)
H. Bevrani University of Kurdistan 21

Control of Multivariable Systems

Steady-State Decoupling

(Ref 1, pp. 91-93)

H. Bevrani University of Kurdistan 22
 Poles
Theorem: The pole polynomial corresponding to a minimal realization of a
system with transfer function G(s) is the least common denominator of all non-
identically zero minors of all orders of G(s).

Example:

The minors of order 1:

The minors of order 2:

The least common denominator of all the minors:

(Ref 1, p. 135)
H. Bevrani University of Kurdistan 23

Zeros
Theorem: The zero polynomial z(s), corresponding to a minimal realization of the
system, is the greatest common divisor of all the numerators of all order-r minors
of G(s), where is the normal rank of G(s), provided that these minors have been
adjusted in such a way as to have the pole polynomial as their denominator.

Example: Normal rank: 2

The minors of order 2:

The greatest common divisor of numerator:

(Ref 1, p. 139)
H. Bevrani University of Kurdistan 24
 Pole/Zero Cancellation

(Ref 1, Sec. 4.5)

H. Bevrani University of Kurdistan 25

Sensitivity as a Feedback Performance Index

Disturbance Attenuation in SISO Systems
Open-loop Closed-loop

Small is a good feedback performance

H. Bevrani University of Kurdistan 26

 Sensitivity as a Feedback Performance Index
Insensitivity to Plant Variations in SISO Systems

Small absolute value of S ( ) is a good feedback performance

(Ref 1, p. 23)
H. Bevrani University of Kurdistan 27

H∞ Norm as a System Gain

In MIMO Systems, System Gain:

: Proper stable system

Example:
σ-plot

dB

(Ref 1, p. 158)
H. Bevrani University of Kurdistan 28
 Difference Between the H∞ and H2 Norms

Minimizing H∞ norm: Push down Minimizing H2 norm: Push down

“peak of maximum singular value” “whole thing (all singular values over
all frequencies)”

Multiplicative property: Multiplicative property:

2 ? 2 2
(Ref 1, pp. 75, 159)
H. Bevrani University of Kurdistan 29

Thank You!

H. Bevrani University of Kurdistan 30