© All Rights Reserved

4 views

© All Rights Reserved

You are on page 1of 4

6.245: MULTIVARIABLE CONTROL SYSTEMS

by A. Megretski

Problem Set 5 Solutions

1

Problem 5.1

Use KYP Lemma to find (analytically) the set of all a R such that the

Riccati equation

P A + A

P = (C

PB)(C B

P ),

where (A, B) is controllable, (C, A) is observable, and

B = (s + a)

1000

C(sI A)

1

,

has a stabilizing solution P = P

.

This is a Riccati equation of the form

+ P +

P = P P,

where

= C

C, = A + BC, = BB

.

Since the pair (A, B) is controllable, so is the pair (A + BC, B). According to the KYP

Lemma, a stabilizing solution of the Riccati equation exists if and only if

|w|

2

|Cx|

2

0 for jx = (A + BC)x + Bw, R.

Substitution v = w + Cx yields an equivalent condition

|v|

2

2Re(v Cx) 0 for jx = Ax + Bv, R.

1

Version of April 26, 2004

2

Again, according to the KYP lemma,this is equivalent to

1 > Re G(j) R, G(s) = C(sI A)

1

B.

Since the maximal real part of G(j) is achieved at = 0, a stabilizing solution of the

> 2

0.001

Riccati equation exists if and only if |a| .

Problem 5.2

Using the generalized Parrots theorem, write down an algorithm for

finding matrix L which minimizes the largest eigenvalue of

+ 2L

M = M(L) =

2L

+

+ L

L

,

where =

, , and =

are given matrices.

First, let us nd the lower bound

for the functional to be minimized. Note that

max

(M(L)) < r if nd only if the quadratic form

2

r

(w, u, y) = w w + 2Re w (y + 2u) + y y + |u|

2

r(|w| + |y|

2

)

is negative denite for u = Ly. Conditions for existence of such L are given by the

generalized Parrots theorem (which can be applied because

r

is convex with respect to

u):

(a) < rI (i.e.

r

(w, 0, 0) 0);

4I b

(b) < rI (i.e. the minimum of

r

(w, u, y) with respect to u is negative

denite).

Hence

= max

max

(),

max

4I

b

.

Now, for r =

, let u

= c

1

y + c

2

w be the argument of minimum of

r

(w, u, y) with

respect to u (it is easy to see that, in our case, c

1

= 0 and c

2

= 2). Let

(w, y) =

r

(w, c

1

y + c

2

w, y) = w w + 2Re w y + y y 4|w|

2

(|w|

2

+ |y|

2

)

be the minimum itself. Let w

= c

3

y be the argument of maximum of

r

(w, y) with

respect to w (since <

I,

r

maximum is well dened). It is easy to see that, in our case,

c

3

= (4I +

I )

1

.

3

To complete a solution, let us prove that

L = L

= c

1

+ c

2

c

3

= 2(4I +

I )

1

is an optimal value of L. Indeed, according to the way c

1

, c

2

, c

3

are dened,

r

(w, u, y) = |u c

1

y c

2

w|

2

(w c

3

y)

(4I +

I )(w c

3

y) +

(y),

r

where

(y) = max

r

r

(w, u, y) = y (

I +

(4I +

I )

1

)y 0.

r

w w u

When u = (c

1

+ c

2

c

3

)y, we have

I)w+

r

(w, u, y) = |c

2

(wc

3

y)|

2

(wc

3

y)

(4I+

I)(wc

3

y)+

(y) = w (

(y) 0.

r r

Problem 5.3

Use the KYP Lemma to write a MATLAB algorithm for checking that a

given stable transfer function G = G(s), available in a state space form,

satisfies the condition

|G(j)| > 1 R {}.

The algorithm should be exact, provided that the linear algebra op-

erations involved (matrix multiplications, eigenvalue calculations, com-

parison of real numbers) are performed without numerical errors. In

particular, checking that |G(j

k

)| > 1 at a finite set of frequencies

k

is

not acceptable in this problem

2

.

Assume that a minimal state space model of G is given by

A B

G := .

C D

Note that condition |G(j)|

2

> 1 is equivalent to

|Cx + Dw|

2

|w|

2

being positive denite subject to jx = Ax + Bw for all real , including = , in

which case the linear constraint takes the form x = 0. According to the KYP Lemma,

2

Of course, frequency sampling may be acceptable in many practical applications

4

this is equivalent to the inequality D

P = P

of the Riccati equation

+ P +

P = PP,

where

= C

(I D(D

D I)

1

D

)C, = A B(D

D I)

1

D

C, = B(D

D I)

1

B

.

The second condition is equivalent to the absence of purely imaginary eigenvalues of the

associated Hamiltonian matrix

H = .

The M-function ps5 3.m implements the algorithm. When its argument d is less than

one, it either reports the D condition D

small (numerically indistinguishable from zero) minimal absolute value of the real part of

eigenvalues of the associated Hamiltonian matrix.

- 54 WodzickiUploaded bygotunicorns
- ADAMS Vibration TheoryUploaded bydrkwing
- MTH210F09 ReviewUploaded byschreckk118
- Analytic Network ProcessUploaded byshivanshu_parihar
- EigenvaluesUploaded byrhinemine
- Applied math lecture notesUploaded byJeff Farmer
- LovaszUploaded byShahar Papini
- Linear AlgebraUploaded byAL
- m225Uploaded byAakash Verma
- ACT.pdfUploaded byrs0004
- p504Uploaded byklomps_jr
- 01450747Uploaded byMayssa Rjaibia
- 04654615Uploaded byUtpal Das
- c11-4Uploaded byVinay Gupta
- 13 Methods for the solution of AXD − BXC = EUploaded byWiliams Cernades Gomez
- cwp-593Uploaded byDan Melancon
- 12.pdfUploaded byTalin Machin
- Two Dimensional Systems REPORTUploaded byDavidSanRománTorrubia
- l_11Uploaded bySanjay Dubey
- 1.Mathematical Physics GATEUploaded bynrdnandan
- MIT6_262S11_lec08_Pagina12Uploaded byGago Romero
- LADE14 EigenUploaded byRoumen Guha
- Stability Analysis for VAR systems.docUploaded bycosmin
- Open channel hydra new.pdfUploaded byprakasharmy
- 763313A_solution_1Uploaded byguillherme
- IJDKPUploaded byLewis Torres
- L2_MA_4301Uploaded bySamantha Brown
- Homework #10.pdfUploaded byAnonymous tzYVAD
- 2007_fm_paper1Uploaded byB
- Lecture 2Uploaded by_MerKeZ_

- COMP-I-09(1)Uploaded bysergiovelasquezg
- [PID] PID Control - Good Tuning - A Pocket GuideUploaded byAdil Ahsan
- ps3sol_6245_2004.pdfUploaded bysergiovelasquezg
- ps5_6245_2004.pdfUploaded bysergiovelasquezg
- ps4_6245_2004.pdfUploaded bysergiovelasquezg
- ps6sol_6245_2004.pdfUploaded bysergiovelasquezg
- 325-NCS_Handbook05Uploaded bysergiovelasquezg
- 11_Simulation of the Effect of Data ExchangeUploaded bysergiovelasquezg
- 451-2003-1-PBUploaded bysergiovelasquezg
- 395-1979-1-PBUploaded bysergiovelasquezg
- 60481_91781v00_PID_ControllerUploaded bysergiovelasquezg
- Physical ScienceUploaded bysergiovelasquezg
- Introductory ACCircuit TheoryUploaded bysergiovelasquezg
- 0321294319 Root KitsUploaded bysergiovelasquezg

- Image TransformsUploaded byScindia Scarletsmiles
- Core 3 Notes Maths AQAUploaded bysophie
- MATH1081 T1 AnswersUploaded byJames PPD
- Ordinary and Partial Differential EquationsUploaded byAmit Yadav
- matheng-skript-1213Uploaded bymcrajpura
- MMAN3200 State SpacesUploaded byAl Chemist Shahed
- Hw 3 SolutionsUploaded byCody Sage
- iiser8Uploaded byapi-3695829
- Lecture NotesUploaded bySaid Halilovic
- documents.mx_qt-study-pack-2006-wip (1).pdfUploaded byKafonyi John
- Lew 17139Uploaded byBernardo Cascão
- Chapter 4 Version 1 of Vector Analysis Written by Hameed UllahUploaded bySikandar Khan
- MATLAB Central - Shortcut for Symmetric MatricesUploaded bySepliong
- MATH21 CoursewaresUploaded byPaolo Gochingco
- Maria Farge - Multiscale Representations, Fractals, Self-similar Random Processes and WaveletsUploaded byZachary Lee
- Ch 8 SolutionsUploaded byyaya
- 6024Methods Units 1 and 2 Exam NotesUploaded bykoko
- 131903-2130002-Maths-IIIUploaded byAkash Balani
- Homework151aMain.pdfUploaded byAhmad Imran Rafique
- Capter 2Uploaded byFahmy Alhafidz
- 216_vgGray R. - Probability, Random Processes, And Ergodic PropertiesUploaded bydavid
- A simple proof of the Perron-Frobenius theorem for positive symmetric matrices.Uploaded bybidsitlov
- A Lyapunov Analysis of Stability Robustness for Discrete Linear Descriptor SystemsUploaded bymamos99blue
- Right Circulant Matrices with Fibonacci SequenceUploaded bySundarVaidya
- Appendix A- Gamma and Bessel Functions.pdfUploaded byCedric Stokes
- ConvexityUploaded bylionshare148
- Linear Algebra With ApplicationsUploaded byAzima Tarannum
- Unit1.Inner Product Spaces.Uploaded byPrafulla Patil
- MIT6_241JS11_chap01Uploaded bySAN JUAN BAUTISTA
- Greatest and Least Integer FunctionsUploaded byAbhishek Singh

## Much more than documents.

Discover everything Scribd has to offer, including books and audiobooks from major publishers.

Cancel anytime.