IC302: Modern Control Theory

Unit 1: Models in State-space

Associate Professor Dept. of Instr. & Control Engg., NIT Tiruchirappalli

[email protected];gmail.com

Dr. R. Ayyagari

Recap:

`State' Variables

sf we know how iL nd vC re hngingD we know everything out the iruitF he two oFdFes reX d R I vC C I u iL@tA a iL dt L L L d I iL vC @tA a dt C orD equivlentlyD x@tA a "
R L

a A x@tA C b u@tA "

I C

I L

x t

"@ A C

I L

u t

@A

he solution to this mtrixEvetor dierenE til eqution @wht is the ordercA is given in terms of the eigenvlues of the mtrix AF
Q

A Wonderful Insight

romogeneityX vR @tA a edditivityX vL@tA a a @A

" H x@tA

vC t R i t R

@ A C u@tA " I x@tA C I u@tA

endD linerityX vC @tA H I H " R H x C H u@tA vR @tA a vL@tA R I I

R

Let's have some fun making faces

he pir of equtions my e lued s I I  C R v C LC v a LC u v @IA L nd I I  C R i C LC i a L u i @PA L hese two equtions look dierentD ut they re similar yserve { iqnF @PA hs derivtive of input
S

A simpler choice - intuitive???

Eqn (1)...

hene the stte vetor xX " xI a v nd xP a xI so tht H I H x@tA a " " x@tA C I
LC R L

u t

@A

a A x@tA C B u@tA stte eqution4 "

v

LC

a I H x@tA " a C x@tA " output eqution4

T

Controllable Canonical Form (CCF)

Eqn (1)... not so intuitive!

por revityD write  C aIv C aHv a bHu v hene the stte vetor xX " xI a v nd xP a xI C aIxI so tht xP a aHxI C bHu nd H aI I x@tA a " " u@tA x@tA C bH aH H
v

a I H x@tA "

Observable Canonical Form (OCF) U

Eqn (1)... Where is the intuition?

qet two numers pI, pP @ cA suh tht aI a pI C pP nd aH a pI pP hene the stte vetor x suh thtX " v a x I C xP v a pIxI pPxP equivlentlyD pP I v xI a v C pP pI pP pI pI I v xP a v pP pI pP pI
V

Eqn (1)...

prom the preeding defFD  a pIxI pPxP v aI v C aH v a @pI C pPA@pIxI pPxPA C pIpP @xI C xPA a
p x p x

PI I

PP P

ewriting eqn @IAD we now hve pI @xI C pIxIA pP @xP C pPxPA a bHu
W

Eqn (1)...

tuggling little moreD pI @xI C pIxIA pP pI pP @xP C pPxPA a pP pI

at

b p p

yrD @I tA @xI C pIxIA t @xP C pPxPA a ku errngingD @xI C pIxI kuA t @xI C pIxI C xP C pPxPA a H

P I ak

pinllyD dd nd sutrt ku to the seond term on vr

IH

Eqn (1)...

sntuitive @cA hoieX xI a pIxI C ku xP a pPxP ku so tht k pI H " x@tA a " u@tA x@tA C k H pP
v

a I I x@tA "

Diagonal Canonical Form (DCF) Simplest of all!

ss pI a pP lwysc
II

Ater all, It's the same system

e hve the following stte vetorsX i x a " v , xC a v , xO a aIvvC v " " v nd v C v xD a " @I Av v ih of these vetors my e linearly transformed into nother using n pproprite nonEsingulr mtrixD for eFgF I H xC xO a " " aI I
aT
O C

ehnillyD we change the basis of the stteEspe4 to suit to our onveniene

IP

Assignment

ytin the nonil forms using iqn @PA row do you hndle the derivtive on r inumerte ll the stte vetors ytin ll possile mtries Tij elte the stte vetors rising out of iqn @PA to those rising out of iqn @IA row do these T mtries look likec
Ater all It's the same system!

IQ

Example
y

fy inspetionX nd

@QA C Wy@PA C PUy@IA C PRy a u

H I H AC a H H I PR PU W

W I H AO a PU H I PR H H

roweverD it is not so esy to get P H H AD a H Q H H H R

IR

Assignment...

ytin the nonil forms for the followE ing iruitsX

IS

Technicalities

vet9s look t the system mtrixX H aI I AO a , aH H

A

R L

I C

I L

AC

a I LC a

H I
R L

AD

yserve tht tre a aID nd a P C aI C aH for ll the four mtries

IT

Technicalities...

gonsider ggp nd ygpD where xO a T x C x C a T I x O st follows tht xO a T xC a T @AC xC C BC uA a T AC T I xO C T BC u

AO BO

eny trnsformtion of the form Aj T a T Ai using nonEsingulr @idellyD orthonormlA mtrix T is dened s similarity transformationF st tkes lots of ingenuity to onstrut T for given Ai suh tht it is trivil to hndle Aj Y fter llD it9s the sme system3
IU

Technicalities...

gonsider ggp nd ygpD where I H O TC a aI I st is esy to verify tht

AO T

H I a aH H a T AC

purtherD the hrteristi polynomil of the mtrix is preserved upon trnsformE tionF c ht re the eigenvlues
IV

Technicalities...

gonsider n ritrry A nn sf there exists T a " " " tI tP tn suh tht AT a T ADD then AT a A" A" A" tI tP tn nd T AD a pI" pP" pn" tI tP tn iquting the mtries olumnwiseD we my redily identify tht ti is the eigenvector orresponding to the eigenvalue pi of A sn other wordsD the mtrix T is the ompoE sition of ll the eigenvetors of A
IW

Example

a H I P Q fy inspetionD a P C Q C P nd hene pI a I nd pP a P @rel nd distintA st is esy to otin T s I I T a I P nd esier to verify tht I H T IAT a H P
A

gonsider

PH

Example

a H I I P rereD a P C P C I nd pI a I a pP algebraic multiplicity of pI is P e my @howcA otin I H T a I I ndD quite interestinglyD I I T IAT a H I Jordan Form orD lokEdigonl4
A

gonsider

PI

Example

gonsider

H I H A a H H I P S R fy inspetionD a Q C RP C S C P nd hene pI a I a pP, nd pQ a P algebraic multiplicity of pI is P

ytin T nd ompute T IAT Y onvine yourself why it is lled lokE digonl4

Exercise:

PP

Example

gonsider

I H H Q H Q Q H A a H. S Q Q H. S Q H H Q ith some diulty we my otin pI a P a pP nd pQ a R a pR

Exercise:

ytin T nd ompute T IAT

PQ

Diagonalizability

ivery squre mtrix hs set of eigenvlE ues CY some of these might e repeting his repitition is dened s the algebraic multiplicity of tht eigenvlue he numer of independent4 eigenveE tors for given eigenvlue is dened s the geometric multiplicity of tht eigenvlue por n eigenvlue whih hs am > I not ll eigenvetors re independent st my e formlly shown tht
gm am

PR

Diagonalizability...

Lemma:

e squre mtrix A is similr to stritly digonl mtrix if nd only if gm a am for every eigenvlue of A sf one existsD n eigenEsis is the most desirle tordn form is lokEdigonl sn generl seD the similrity trnsformE tion results in @upperGlowerA tringulr mtrix
PS

Eigenvectors

en eigenvector " for squre mtrix A is e nonzero vetor suh tht A" a " e e for some slr F he slr is lled the eigenvalue of A ssoited with the eigenvetor "F e e sy tht n eigenvetor elongs to the null-space of I AF purtherD if the rnk of iI A a n r then there would e r independent eigenE vetors ssoited with iF he integer r is lled the degeneracy of AF
PT

Eigenvectors...

st is interesting tht the rottion opertor in the iulidenEP spe os sin A a sin os does not rry ny vetor to prllel oneD iFeFD A hs no eigenvetor unless a H or F yn the other hndD every liner opertor on omplex vetor spe hs t lest one eigenvetorD nd in most ses the eigenE vetors form sisF
PU

Eigenvectors...

Example:

a H I H H he eigenvlues re I a H nd P a HF roweverD A is not similr to H H AD a H H sine "I a I H T e is the only independent eigenvetor
A

vet

e my lso verify y ontrditionX por ny nonEsingulr T H H aA T AD T I a H H

PV

Generalized Eigenvectors...

ywing to the m E gm inequlityD let us sk c "i a " U X A"i a j "i e H e e nd A"iCI a j "iCI C "i, i a I, P, , e e e for given eigenvlue j whose lgeri multipliity is greter thn IF sf the nswer is yes9D it mens tht "i is n e eigenvetor nd tht the mtrix A would simply sle the vetor "iCI in its own diE e retion nd dds it to "i @fF prllelogrm e lwA
PW

Generalized Eigenvectors...

uppose tht j hs n lgeri multipliE ity m nd geometri multipliity qF e rst nd ll the q independent solutions to @j I AA"i a " e H por eh "i thus determinedD we onstrut e generlized eigenvetor9 using @j I AA"iCI a "i e e sf the result "iCI is linerly independent of e ll the vetors lredy foundD then it is vlid generlized eigenvetorF sf still more generlized eigenvetors re neededD then we solve @j I AA"iCP a "iCI e e nd so on
QH

Generalized Eigenvectors...

e Jordan Block Jk@A is k k upper tringulr mtrix of the form I I Jk @A a FFF I here re k I terms CI4 in the superE digonlD nd the slr ppers k times on the min digonlF ell other entries re zeroD nd JI@A a F

QI

Generalized Eigenvectors...

e Jordan matrix J is diret sum of torE dn loks JnI @IA JnP @PA J a FFF Jn @k A in whih the orders ni my not e distint nd the vlues i need not e distintF
k

xotie tht if eh tordn flok Jn @iA is oneEdimensionlD iFeFD ll ni a I nd k a nD then the tordn mtrix J is stritly digonlF
i

QP

Generalized Eigenvectors...

vet A e given omplex mE trixF here is nonsingulr mtrix T suh tht JnI @IA JnP @PA I A a T T FFF Jn @k A
Theorem:
k

QQ

Generalized Eigenvectors...

I P Q A a H I R H H I his mtrix hs a I with am a Q nd gm a IF reneD there is only one indeE pendent eigenvetor nd two generlized eigenvetors re neededF he indF eigenE vetor is I "I a H e H he genF eigenvetors re I I S "P a I nd "Q a IT e e P I H V

Example:

vet

QR

Generalized Eigenvectors...

Example...

gonsequentlyD we get
T AT

I I H a H I I H H I

QS

Generalized Eigenvectors...

I H H Q H I Q H A a H. S Q I H.S Q H H I he eigenvlues re P, P, R, ndRF ith a PD we my oserve tht the mtrix Q H H Q Q I A a H.H Q Q H.H S Q S Q H H Q hs rnk P nd hene there would e P independent eigenvetorsF hey re H I I H "I a I nd "P a H e e H I FFF

Example:

vet

QT

Generalized Eigenvectors...

roweverD with a RD Q H H Q H Q Q H I A a H.S Q Q H.S Q H H Q hs rnk Q nd hene there would e only I independent eigenvetorF st is H I "Q a I e H e need one generlized eigenvetorX P I Q "R a H e P
QU

Generalized Eigenvectors...

gonsequentlyD we get
T AT

P H H H

H P H H

H H R H

H H I R

QV

Assignment...
A

is S S mtrix for whih the following informtion is villeX por pI a pP a P, rnk PI A a R por pQ a pR a pS a P rnk PI A a Q ytin the eigenvetors nd determine the tordn form of e ytin the tordn form of R P I A a H T I H R P
QW

Assignment...

rove the following sttementF sf " is n eigenvetor with eigenvlue D e it follows tht " is n eigenvetor with e eigenvlue F sn ftD the nonempty set Vq a {" V X A" a q"} e e e is suspe of V for ny q

RH

Generalized Eigenvectors...

ytining similr digonl mtrix @or torE dn mtrixA for given liner opertor A is not strightforwrd gomputing eigenvlues oils down to solvE ing polynomil equtionD whih ould e formidle tskF nlike ny liner sysE tem of equtionsD solving for eigenvlues requires n innite lgorithmD nd hene ll prtil eigenvlue omputtions re E omplished y itertive methods
RI

Denition of State

he state x X nI of system t time to is the minimum set of internl vriles tht is suient to uniquely speify the system outputs given the input signl over to, AF ypillyD is the underlying eld for the lger yften the hoie of sttes is nturl ndGor oviousY howeverD trnsformtions eome mndtoryD nd the theory hs propensity for mthemtil rigor rther thn physil intuition
RP

Building Blocks

here re severl grphil symols tht help usX

RQ

Building Blocks...

vet us rst onsider simple rst order homogeneous system z C ao z a H or z a aoz

Schematically

yserve how we used the initil ondition e shll now justify this shemti
RR

Building Blocks...

snitillyD nextD
z z

zo

a aozo@I aotAdt C zo aPtP a zo@I aot C oP A e my ontinue this to otin aPtP aQtQ o z @tA a zo I aot C o P3 q Q3 q q ao t C C @IA q3 a zo ea t
z
o

nextD

a aozodt C zo a zo@I aotA

RS

Building Blocks...

sing the vrileEseprle methodD d z @tA a aodt z @tA or ln z@tA a aot C onstnt or z@tA a zo ea t whih mthes with our shemti his is how we solve dierentil equtions y employing analog simulationD typilly irE uit with severl opEempsF Exercise: olve z C aHz a bHv using n pE proprite shemtiD nd formlly justify tht it works
o

RT

Building Blocks...

vet us generlize some moreF gonsider z C aoz a bIv C bov e shemti demnds some ingenuityF vet v zI so tht zI C aozI a v vinerity implies z a bIzI C bozI

he output eqF is z a @bo bIaoAzI C bIv

RU

Building Blocks...

st is esy to get n equivlent shemtiX z a bIv C bov aoz z a bIv dt C @bov aoz Adt a bIv C @bov aozAdt he orresponding shemti FFF

RV

Building Blocks...

he stteEspe modelX zI a aozI C @bo bIaoAv z a zI C bIv gompring the twoD the B nd C mtries re interhnged he D mtrix is lso present in oth { in generlD D ppers whenever the oFdFeF ontins nth order derivtive of the input
RW

Building Blocks...

vet9s next tke up y @nA C anIy @nIA C C aoy a por revity nd lrityD let9s rst onsider y @QA C aPy @PA C aIy @IA C aoy a H nd dene xI a y xP a xI xQ a xP
SH
bnu n

@ A C bnIu@nIA C C bou

Building Blocks...

FFFresulting in xI H I H H xP a xQ ao aI y a I H H x sf we dd n input bouD xI H I H H xP a xQ ao aI

y

H xI I xP

H xI I xP

H C H u

bo

a I H H

C H u
SI

Building Blocks...

pinllyD let9s tke up y @QA C aPy @PA C aIy @IA C aoy a bQu@QA C bPu@PA C bIu@IA C bou essuming tht u xI

resulting in FFF
SP

Building Blocks...

the stteEspe modelX xI H I H xI H H H I xP C H u xP a I xQ ao aI aP xQ nd y a @bo bQaoA @bI bQaIA @bP bQaPA x C bQ u

Controllable Canonical Form

SQ

Building Blocks...

he ygpX

resulting in

SR

Building Blocks...

the stteEspe modelX bP bQaP aP I H A a aI H I B a bI bQaI bo bQao ao H H nd C a I H H , D a bQ

SS

Building Blocks...

xextD let9s ssume tht the prtil frtion expnsion of the trnsfer funtion is vilE le so tht the shemti looks like

resulting in
ST

Building Blocks...

the stteEspe modelX pI H H kI A a H pP H , B a kP H H pQ kQ nd C a I I I , D a k o orD T B a I I I with C a kI kP kQ , D a ko

Diagonal Canonical Form

SU

Example

vet
y

@QA C T y@PA C IIy@IA C Ty a Su@IA C Uu

CCF:

H I H H A a H H I BaH T II T I C a U S H D a H

OCF:

T I H H A a II H I B a S T H H U C a I H H D a H

SV

Example...

DCF: he trnsfer funtion is

y s u s

@A a Ss C U @ A @s C IA@s C PA@s C QA I Q a s C I C s C P C s RQ C reneD I I H H A a H P H Ba Q R H H Q C a I I I D a H orD FFF

SW

Example...

PCF:

@A a S sCU I I S @A @s C IA @s C PA @s C QA he shemti for this is

y s u s

from whih we my uild the mtries FFF

TH

Example...

Q I A a H P H H C a I H H

H H P BaI S I I D a H

TI

Example

vet

@Aa I I @ A sP C Ps C I a @s C IAP ply lever trik nd rewrite s

y s u s

so tht

y u

I bo CI a ssC I a s C I

hgp hs ollpsed to gp c itD there is moreFFF

TP

Example

vet

y s u s

@ A a @s C IA@s C QA @A @s C PAP kP kI C s CP a ko C s C P

TQ

Example...

prom the shemtiD we otin the followE ing equtionsX x I a Px I C x P C k I u x P a Px P C k P u y a xI C k o u so tht P I , B a kI A a kP H P

C

a I H

a k o

TR

Example

vet

T s

@ A a @s C IAI@s C PA P kP kI C s CI kQ a sCI CsCP

TS

Example

he stteEspe model is I I H kI A a H I H , B a kP kQ H H P C a I H I riting the mtries BD C D nd D is trivE ilF sf the residues re ko knD then the B @or C A mtrix onsists of kI kn in the olumn @rowAD nd ko ppers s the D mE trixF sf we write C mtrix extly elow the A mtrixD the olumns of C ontin Is orE responding to the distint poles Hs orreE sponding to the repeted polesF
TT

Example...

e stritly dhere to IF the shemti of eh of the prtil frE tionsD nd PF enumerte the stte vriles right to leftF
xi

from fr

TU

Example

vet

I T @sA a @s C IA@s C PAP@s C QAQ

TV

Example...

prom the shemtiD xI a xI C kIu x P a Px P C x Q C k P u x Q a Px Q C k Q u x R a Qx R C x S C k R u x S a Qx S C x T C k S u x T a Qx T C k T u

y

a x I C x P C xR

TW

Example...

nd heneD kI I H H H H H kP H P I H H H kQ H H P H H H , B a A a k H H H Q I H R H H H Q I kS H kT H H H H H Q C a I I H I H H xotie the mtries B nd C F

UH

Assignment ...

qiven verify tht

T s

@ A a sP@s IC PA

H I H A a H H H H H P

ytin the tgp if I T @sA a P s @s C QAQ@s C IA

UI

Example

vet

s @ A a sP C Cs IC T U eny reful oservtionc T s

uikly @rtherD mehnillyA H C I T @sA a s C I s C T so tht I H ,Ba H , Ca I I A a H T I

UP

Example...

es fr s modelling is onernedD this zeroE pole nelltion n e ignored roweverD suh system lks n importnt property lled controllability { you n onvine youself tht you do not hve ny control over xID ut it dds to xPD nd hene inuenes the outputF

UQ

Some Interesting Connections

vet us resort to vple trnsformtionX L{x a Ax C Bu} s x@sA x@HA a Ax@sA C Bu@sA orD @s I AAx@sA a x@HA C Bu@sA nd x@sA a @sI AAI{x@HA C Bu@sA} vterD we hve ertin importnt questions regrding this new mtrix sI A nd its inverseF
UR

Connections...

sf it wereI slr rst order systemD then @sI AA would look like I sa whose inverse is eatF pollowing the sme onvention LI{@sI AAI} a eAt a H I P Q edily we hve I s I @sI AAI a P s C Q nd with little eort Pet ePtPt et ePtPt At a e Pet C Pe et C Pe
A

vet

US

Connections...

yserve tht A is given in the ggp { y @PA C Qy @IA C Py a u@tA essuming zero initil onditions sP C Qs C P y @sA a u@sA orD I u@sA y @sA a P s C Qs C P trF fnF sn ftD we n generlize these resultsD nd sy { the eigenvlues of the system mtrix A re identil to the poles of the system
UT

Connections...

st is now esy to see tht y@sA a C dj @sI AA B C D H @sA a u@sA det@sI AA in terms of the mtries A, B, C D nd D sf A were digonl mtrixD eAt ws simple nd strightforwrd sf A is in the ggpD then omputing eAt is it involved in terms of nding the eigenE vluesD nd otining prtil frtions nd the result looks muh messier sn ftD the eigenvlues re smeY while At we do not other out the nl shpe of e D we re indeed othered out the ompuE ttion involved
UU

MIMO Systems

vook t this @PA @IA @IA y I C QyI C PyP a u I C PuP C PuP @PA @IA @PA @IA yP C Ry I C Qy P a uP C QuP C uI e my go out using uilding loks nd get nonil modelsF fut FFF

UV

MIMO Systems...

FFFlook t the simultion digrm of the rst equtionD rewriting it s @IA yI a @Pu@IA QyI AC P @uI C PuP PyPAdt dt

UW

MIMO Systems...

imilrlyD

he two digrms n e joined to mke igger oneD ut s long s we re ler out the vrilesD we shll void mking it further omplited
VH

MIMO Systems...

gomining the twoD

y

I a xI y P a xQ C u P

I a Qx I C x P C P u P x P a Py P C u I C P u P a Px Q C u I x Q a Ry I C x R C Q u P a Rx I C x R C Q u P x R a Qy P C u I a Qx Q C u I Qu P
x

VI

MIMO Systems...

husD

Q I H H H P A a R H H H H Q I H H H C a H H I H

H H P H , B a I H H I Q H I Q nd D a H H H I

his model does not onform to ny of the erlier nonil forms st does not mtter how A looks like { we n lwys4 mke similarity trnsformE tions
VP

MIMO Systems...

vooking t the system from lssil onE trol point of viewFFF yI@sA yI@sA yP@sA y @sA , , , nd P uI@sA uP@sA uI@sA uP@sA yrD equivlently yI@sA uI@sA a HII@sA HIP@sA uP@sA yP@sA HPI@sA HPP@sA
H s

@A

VQ

MIMO Systems...

vet us now otin the trnsfer mtrixF I s C Q I H H I H H H H s P H H @sA a H H I H R H s I H H Q s H P I H C H H H Q H I I Q sP C I sQ C T s @s IA s @sQ C TsP C s VA a s @s C IAQ
VR

MIMO Systems...

sn generlD if there re r inputs nd m outE putsD the trnsfer mtrix is of size m r sing superposition it is esy to interpret wht goes on sn other wordsD if ll the r re ting on the systemD then eh of the outputs re r yi a Hij @sAuj @sA
j

aI

wost of the lssil ides of sy sysE tems my e extended to wswy systemsY n exeption to this is the ide of phse mrgin
VS

MIMO Systems...

hile pir of sy trnsfer funtions HI@sA nd HP@sA my e multiplied in ny wyD this is not the se with pir of wswy trnsfer funtions { the order of the trnsfer mtries is the reverse of the order in whih they re sdedF he losedEloop trnsfer mtrixD with p refE erene inputsD r ontrol inputs nd m outE putsD looks like y @sA a P C @I C F P C AI r@sA xotie tht we need to ompute the inverse of mtrixF he denomintor of ll the mr trnsfer funtions is the smeD nd it turns out to e the hrF polyF of AF
VT

MIMO Systems...

vet us look t

he trnsfer mtrix from r to y my e written y inspetion s followsX y @sA a PQPP@I C CPPAIPI r@sA

herive the trnsfer funtion sE suming tht this is sy systemD nd ompre the resultsF
Exercise:

VU

MIMO Systems...

here re severl resons for using trnsfer funtions nd the ssoited frequeny doE minF irtullyD ll prtil feedk onE trol systems re nlyzed nd designed this wy roweverD the min dierene is the presE ene of directions in the wswy systemsF en lgeri tool lled h provides useful wy of looking t the diretionlity of wswy systemsF
VV

MIMO Systems... Assignment ...

ytin stteEspe modelD nd the trnsE fer mtrix for the wswy system given y @QA @PA @IA @IA y I C Q y I C S y I C S y P C U y I Uy Q a uI
y

@PA C yP yI C PyQ C QyP RyI @IA @IA @IA P a uP @IA y Q C Py Q y I a uQ

VW

MIMO Systems...

he signine of stteEspe model for multivrile systems is evident nowF yne the mtries re knownD the solution to the stte vetor n e diretly omputedD nd hene the solution to the output vetor hether the output vriles yi re ouE pled does not other us ny longer sn ftD this stteEspe representtionD reE grdless of order nd numer of vrilesD mkes possile uniform set of solution tehniques nd omputer lgorithmsF
WH

MIMO Systems...

vet

I H P P H B a H I A a H H I , I I H Q R IH t x@HA a S nd u@tA a I P

WI

MIMO Systems...

e my proeed routinely to otin x@sA a @sI AAI x@HA C B u@sA nd x@tA a LI{x@sA}
RU IPU ePt SV eQt C I t QT IR C R W T t PW eQt C I t C II a Ue W Q W Uet C PW eQt P Q Q

et

WP

MIMO Systems Assignment...

ytin stteEspe model for the followE ing iruit

erify the results of the ove exmpleF

WQ

Discrete-Time Systems

ht we hve here is m y @k C nA C C aoy @kA a biu@k C iA iaH with m n sf we dene stte vriles y @kA a xI@kA xI@k C IA a xP@kA FF
xn

@k C IA a I

xn k

@A
WR

Discrete-Time Systems...

we my otin the stteEspe representE tion s x@k C IA a A x@kA C B u@kA @ A a C x@kA C D u@kA muh the sme wy we did erlier
y k

he hoie of symols { eDfDgDhDxDnd y { might use some onfusionY the dE vntge is tht mny onepts @ontinuous nd disreteA re reveled s eing essenE tilly identil
WS

Discrete-Time Systems...

yserve tht x in ontinuousEtime systems is repled y x@k C IA sf we integrted dierentil eqution to otin its solutionD we now hve to delay dierene eqution eordinglyD we use unit time delay lok
Z

est of the story is routine E we n hve ll the nonil forms

WT

Discrete-Time Systems...

vet
y

@QA I.UHVSy@PA CH.WRPSy@IA H.ITSQy a H.HIQTTUu@PA C H.HHITUu@IA H.HHSHu

he ggp nd the ygp re stright forE wrdF H I H H Ac a H H I , Bc a H H.ITSQ H.WRPS I.UHVS I Cc a H.HHSH H.HHITU H.HIQTTU I.UHVS I H H.HIQTTU Ao a H.WRPS H I , Bo a H.HHITU H.ITSQ H H H.HHSH Co a I H H
WU

Discrete-Time Systems...

he trnsfer funtion of the system is H.HIQTTUzP C H.HHITUz H.HHSH T @z A a Q z I.UHVSz P C H.WRPSz H.ITSQ H TUIR a z HIQTUW z H..SRVV z C H..TUHQ H.HIQTTU . z H TUHQ z so tht FFF H HRIV H H HPVI a z .H.TUHQ C @z .HHHUR P z .H.QTUW .TUHQA

WV

Discrete-Time Systems...

H.TUHQ H.IPIS I H Ap a H H.TUHQ I , Bp a H H H H.QTUW I Cp a H.HIVQR H.HHITU H.HIQTTU

H.HRIV H.TUHQ I H Ad a H H.TUHQ H , Bd a H.HHUR H H H.QTUW H.HPVI Cd a I H I

WW

Discrete-Time Systems...

es in the ontinuousEtime seD the hgp is tully tgp owing to repeted poles e hve simply expnded T @zA into prtil frtionsD nd not T @zA/zY neverthelessD the result is sme

IHH

Discrete-Time Systems... Assignment...

ytin the shemti digrms nd verify the results of previous exmple por the tgp verify the result using the other prtil frtion expnsion ytin the nonil forms of the disreteE time system given y T @zA a H.HHTURS@z C H.HTUPA@z C I.PRITA @z H.HRWUWA@z H.PPQIQA@z H.THTSQA
IHI

Discrete-Time Systems...

gonsider the QsQy systemX @QA @PA @PA yI C T y I T y Q @IA @IA Py I C y P C y I Py Q a uI C u@IA P
y

@PA C QyP yI @IA @IA P C S y P C yQ a uI C uP C uQ @IA C PyQ yP@kA Q a uQ uP C Uu@IA Q

IHP

he mtries re FFF

Discrete-Time Systems...

T P I I H H

I H H H H H

H I H H H H

H I H Q S I

H H H I H H

H T H H P , B a I H H I I H P

H RP I H H IR H H I T I IQ

I H H H H H C a H H H I H H , H H H H H I

H H H D a H H H H H U

erify the results of this exmple

IHQ

Discrete-Time Systems... Assignment...

ytin the stte model for the following system yI@k C PA C IHyI@k C IA yP@k C IA CQyI@kA C PyP@kA a uI@kA C PuI@k C IA
y

P@k C IA C RyP@kA RyI@kA a PuP@kA uI@kA

IHR

Discrete-Time Systems... What is eAt?

qiven sequence of input vetors u@HAD u@IAD F F F long with x@HAD we my ompute x@iA t ny generl time iX x@IA a Ax@HA C Bu@HA x@PA a Ax@IA C Bu@IA a APx@HA C Bu@IA C ABu@HA FF x@iA a Aix@HA C B AB APB AiIB u@i IA u@i PA u@i QA FF u@HA vterD we see tht H i n
IHS

Discrete-Time Systems... What is eAt?

epprentlyD the stte trnsition mtrix4 hs totlly dierent shpe { An yserve tht the onvolution integrtion is repled y disrete summtion elsoD oserve the mtrixEvetor produt for the disrete summtion

IHT

Discrete-Time Systems...

elterntivelyD pply EtrnsformtionX Z{x@k C IA a Ax@kA C Bu@kA} x@z A a @zI AAI @z x@HA C Bu@z AA nd x@kA a Z I{x@z A} xotie tht
Z{Ak }

a @zI AAIz
IHU

Discrete-Time Systems...

vet

por ny generl time kD Ak I k k @ I Ak k @P k A@ I AkI @PA P P I Ak I AkP H a @P k@ P H H @ I Ak P

I I P I P A a H P I H H P

I P

Exercise:

erify for ny ritrry k

IHV

Discrete-Time Systems...

vet

I P A a I V

@HA a

I Q

I V , I P

a I H H I
k ek

, u k

@Aa

por k a P x@kA a APx@HA C Bu@IA C ABu@HA I I IU I I H I V eI P TR V I a I IU Q C H I H I I P I V TR V a US I TR C e

UW TR

IHW

Discrete-Time Systems...

e ould not ut help oserve tht for lrge kD omputing Ak would e muh simE pler provided A is in digonl form erify the results of previous exmple y solving the diF eqnF using EtrF sn the previous exmpleD otin x@kA for k a Q, R, S nd verify tht they n e oE tined s liner omintions of x@HAD x@IAD nd x@PAF purtherD generlize this oservE tion for ny k > P
IIH

Discrete-Time Systems... Assignment...

ytin the solutions

S IP A a I IP I P

I IP , x@HA a P I S IP

a H
S T
C

IQ T

I H

I H P I Q

, B

Q I a P H I I

I Q I H I I

I x@HA a P I I
III

u k

@Aa

k k

Summary of Unit 1

sn this unit we delt with severl vriles { inputD outputD nd the stte vriles e strive to hieve uniform model lled the stteEspe model { set of four mE tries yne my sk why to omplite the issue { for etter expression hen we del with more generl systemsD the moment we introdue n dditionl set of stte vriles @just suient in numE erAD we will e le to express the system netlyD preiselyD nd unmiguouslyF
IIP

Summary of Unit 1...

wodern ontrol theory dtes from round IWTHD when frequeny response funtions nd vple trnsforms were elipsed y stteEspe models tteEspe models per se were not relly newY dierentil equtions hve een studE ied sine the dvent of the lulus ht perhps ws new ws the mting of mtrix theory with dierentil equtions to provide n ll enompssing frmework for ontrolD espeilly extensions to nonliner systems FFF
IIQ

Summary of Unit 1...

FFFnd the ility to exploit timeEdomin stility methodsD in prtiulrD vypunov9s methodD whih ly dormnt for hlf enE tury ht the stteEspe provided ws wy to pture the internl workings of sysE temX n inputEoutput model in terms of the vple vrile does not provide muh insight into wht is going on inside sn ontrstD stteEspe model llows us to look t eh stte nd deide whether the ehvior of the stte is desirleF
IIR