Professional Documents
Culture Documents
TA4 11:15
A ROBUST DISCRETE-TIME ADAPTIVE CONTROLLER+
ABSTRACT
INTRODUCTION
Recently,severalattempts
have been made t o analyzetherobustnessproperties
of adaptivecontrolalgorithmswithrespectto
bounded disturbances and unmodeled dynamics. I n [2]-[9] i t i s shown t h a t u n modeled dynamics o r even small bounded disturbances can
go u n cause mostof theadaptivecontrolalgorithmsto
s t a b l e . The outcome of t h e s ei n s t a b i l i t ys t u d i e si s
t h a t most of the adaptive controllers
need t o be modifiedorredesignedinordertocounteractinstability
and improve robustness.
Severaldifferentapproachesforrobustness
have
been recently proposed i nt h el i t e r a t u r e .
In thecase
of bounded disturbances, the basic idea
of most ofthe
modificationsistopreventinstability
by eliminating
theintegralactionoftheadaptivelaws.This
can be
achieved by theuse of a deadzone [2,4,10-131, a-modif i c a t i o n [6,9] and similarothertechniques[3,10].In
anotherapproach[14]forhandlingdisturbances,the
referenceinputsignal
is requiredto have enough f r e quencies for the measurement vector t o be p e r s i s t e n t l y
exciting and be l a r g e r e l a t i v e t o t h e d i s t u r b a n c e .
Persistent excitation guarantees exponential stabi 1 i t y
and thereforerobustnesswithrespecttodisturbances.
When unmodeled dynamics a r ep r e s e n t ,g l o b a ls t a b i l i t y
cannot be guaranteed by simplyeliminatingtheintegral
action of theadaptivelaws.
The unmodeled dynamics
a c t a s an externaldisturbanceintheadaptiveerror
equation and can no longer be assumed t o be bounded.
Despite this d i f f i c u l t y , however, severallocalresults
have been obtainedintheliterature[19,15,16]
by
using similarmodificationsasinthecase
of bounded
disturbances.Theseresults,
however, a r ea p p l i c a b l e
Globalrobustnessresultsarerecentlyobtained
f o r d i s c r e t e - t i m e [17-191 and continuous-time[20,21]
p l a n t s whose modeled p a r t s may have a r b i t r a r y r e l a t i v e
degree and whose unmodeled p a r t sa r es t a b l e .
A normalizingsignal, which bounds the modeling e r r o r s i g n a l , and a projection which keeps theparameterestimates bounded and within a chosensphereare
used in
[17,18].Ifthedesiredcontrollerparametervector
i s withinthesphere,thenglobalstabilityisguaranteedinthesense
t h a t f o r any bounded i n i t i a lc o n d i tionsallthesignalsintheadaptiveloopare
bounded.
I n anotherapproach,theidea
of normalizationtogether
with a r e l a t i v e dead zone and a projectionare used i n
theadaptive law toachieverobustnessfor
a discretetimealgorithm.
The approachof[19]requires
bounds
on the unknown plant and controllerparametersaswell
A similarapproach as in [19]
as on theplantzeros.
i s used t o proverobustnessinthecase
of an i n d i r e c t
continuous-timeadaptivecontrollerin[20].
It is
shown t h a t i f t h e unknown parameters of t h e p l a n t l i e
in a known convex setthroughout which no unstable
pole-zerocancellationoccurs,thentheuse
of normalization together with
a suitably designed relative
dead zone g u a r a n t e e s s t a b i l i t y i n t h e p r e s e w e o f
small
p l a n tu n c e r t a i n t i e s .
In [21] a directadaptivecontrol
algorithm which isapplicabletocontinuous-timeplants
w i t ha r b i t r a r yr e l a t i v ed e g r e e
i s proposed. The algowhich i s
rithm i s designedforthereducedorderplant
assumed t o beminimum phase and of known r e l a t i v e degree and order, b u t i s shown t o s t a b i l i z e t h e f u l l o r derplantwhich,
due t o t h e unmodeled dynamics, maybe
nonminimum phase and of unknown order. The algorithm
usestheideaofnormalization
and switchinga-modification. The onlyaprioriinformationrequiredforthe
implementation of thealgorithm i s an upper bound for
the norm of the desired controller parameter vector as
in[10,15].
of[21]
areextendedto
In this papertheresults
discrete-time plants with additive
and m u l t i p l i c a t i v e
p l a n tu n c e r t a i n t i e s .
The s t r u c t u r e of t h ec o n t r o l l e r
b u t i s d i f f e r e n t from t h a t
issimilartothatin[l]
of[17-19,221.
The adaptivelawsforadjustingthe
controller parameters employ theidea of normalization
a n d switchingu-modification and a r e new. I t i s shown
t h a t t h e proposedalgorithmguaranteesboundedness
and
smallresidualtrackingerrorsfor
any bounded i n i t i a l
conditions.Intheabsence
of p l a n tu n c e r t a i n t i e st h e
residualtrackingerrorreducestozero.
The onlyaprioriinformationrequiredforimplementation
of the
algorithm i s an upper bound f o r t h e norm of thedesired controller parameter vector.
2 . PLANT AND THE CONTROL OBJECTIVE
tocontinuous-timeplants
whose dominant p a r t s a r e minimum phase and of relative degree (n*) one and whose
unmodeled p a r t s a r e s t a b l e and f a s t .
Consider a single-inputsingle-output
838
(SISO) plant
f o l l o w i n ga u x i l i a r ys i g n a l sa r eg e n e r a t e d :
representedby
(2. 1)
= kp
i s t h e n o m i n a lt r a n s f e rf u n c t i o n
o ft h ep l a n t .G ( z )i ss t r i c t l yp r o p e r ;u A 2 ( z ) ,p A l ( z )
i s an a d d i t i v e and a m u l t i p l i c a t i v e p l a n t p e r t u r b a t i o n
r e s p e c t i v e l y .F o rc l a r i t yo fp r e s e n t a t i o n
and w i t h o u t
l o s so fg e n e r a l i t yb o t hA l ( z ) ,A 2 ( z )a r er a t e db yt h e
Then t h e a d a p t i v e l a w i s g i v e n b y
same p o s i t i v es c a l a rp a r a m e t e r
u . F o rt h en o m i n a lp a r t
o f t h e p l a n t we make t h ef o l l o w i n ga s s u m p t i o n s :
(i)N(z)is
a monic mth o r d e r p o l y n o m i a l i n
z withall
rootsinsidetheunitcircle;(ii)D(z)is
o r d e rp o l y n o m i a li n
z with n>m; ( i i i ) k
a monic n t h
i s anunknown
a s t r i c t l yp r o p e rt r a n s f e rf u n c t i o n ;( v i )
bound 1 > p o > 0 o n t h e s t a b i l i t y m a r g i n p >
an upper
0 for which
(3.5)
where yo, 61, 62, u o , Mo, y a r e p o s i t i v e d e s i g n c o n s t a n t ss u c ht h a t
P
c o n s t a n tw i t h known s i g n ; ( i v ) t h e c o e f f i c i e n t s o f t h e
p o l y n o m i a l sN ( z ) ,D ( z )a r e
unknown.For
t h ep l a n tp e r Al ( z ) i s
t u r b a t i o n sA l ( z ) ,A 2 ( z )
we assumed t h a t :( v )
t h ep o l e so fA 1 ( z / p ) ,A 2 ( z I p )a r es t a b l ei s
The p l a n t o u t p u t
(3.6)
e*
(2.2)
= km
m),1
same
s i g n as k
I nt h e absence o pf l a n ut n c e r t a i n t i e s ,
P'
i . e . , p = 0, t h e a d a p t i v e c o n t r o l l e r i n
[ l ] guarantees
boundedness o f a l l s i g n a l s i n t h e c l o s e d l o o p
and
The i n p u t u k
and t h e
(217-1) d i m e n s i o n a l a u x i l i a r y v e c t o r
a c o n s e r v a t i v e bound f o r
5sBk
term.
Ile*ll
can be
mk
The usek t e n r e t a r d sa d a p t a t i o n
whenever B k grows t o l a r g e v a l u e s
and, t h e r e f o r e ,
ADAPTIVE CONTROLLER ( k p = 1)
and o u t p u t yk a r eu s e dt og e n e r a t e
[l].
The twoterms i nt h ea d a p t i v el a w( 3 . 5 )w h i c ha r e
c r u c i a lf o rr o b u s t n e s sa r et h en o r m a l i z i n gs i g n a l
t oz e r o .
When p # 0, however, t h e s t a b i l i t y p r o p e r t i e s
ofthealgorithmin
[l]cannolonger
be guaranteed no
w a t t e r how small IJ i s . The c o n t r o lo b j e c t i v ei st o
dev i s e an a d a p t i v e c o n t r o l l e r f o r t h e p l a n t ( 2 . 1 ) s o t h a t
t h ec l o s e dl o o ps y s t e mi sg l o b a l l ys t a b l e
and t h e
t r a c k i n g e r r o r i s as s m a l l as p o s s i b l e f o r a l l p o s s i b l e
p l a n t u n c e r t a i n t i e s Al ( z ) , A 2 ( z ) s a t i s f y i n g a s s u m p t i o n s
( V I ,( v i ) .
3. THE DESIGN OF
THE
a r e s t a b l e and
l a r g e ,t h e r e f o r e ,
s a f e l yu s e d .
-yk
m
[e~T,e;T,ej]
- Fl-'
and km i s a c o n s t a n t o f t h e
[(z/qo)I
v e c t o rd e f i n e di n
i s a m o n i c( n - m ) t ho r d e rp o l y n o m i a lw i t hr o o t si n -
sidetheunitcircle
aresatisfied
Wm(z/qo)and
model
y k = Wm(z)rk
D,(z)
0<62, d0<1, m a x [ ~ ~ , q ~ l ~ 6 ~ 6
known.
yk i s r e q u i r e d t o t r a c k t h e o u t -
ofthereference
p u t y:
y + 2 u 0 < 1 , Mo)2/18*11,
u,, as
L e m a3 . 1 .L e tx k R n
be t h es t a t eo ft h ef o l l o w i n g
system
F E R (n-l)x(n-l)
are strictly
wheretheeigenvaluesof
w i t h i nt h eu n i tc i r c l e ,( F , g )i s
a c o n t r o l l a b l ep a i r ,
T
and uk = [wk( ' I T 'Ok(2)T,yk]
For ease of presentation
xktl
where B k =
i n p u tt ot h e
(3.2)
wheregk i s a g e o m e t r i c a l l y decaying t e n whichdepends on t h e i n i t i a l cond i t i o n s .
is a (2n-1)
dimensional
ad-
[l]and i s g i v e n below.
13,7)
that
j u s t a b l ec o n t r o pl a r a m e t e rv e c t o r .
The c o n t r o l l e r
s t r u c t u r e( 3 . 1 ) - ( 3 . 2 )i st h e
same as t h e one t h a tw o u l d
known. The
be used i f t h ec o e f f i c i e n t s o f Go(z)were
a d a p t i v el a wf a ra d j u s t i n gt h ec o n t r o l l e rp a r a m e t e r
Bk
i sd i f f e r e n tf r o mt h a to f
BGk
= Axk +
Proof.Since
O < 60,62< 1 , t h e n t h e r e e x i s t s
s t a n t c o > 0 s u c ht h a t
( 3 . 7 ) we have
The
a39
k
(\A \( 2
a con-
~ ~ T6 h e~r e f .o r e f, r o m
k-1
i =O
From ( 3 . 4 ) we have
m I,r,
1 <k 1
(3.10)
< 1
-
osek@k
T
mk
- 0 forall
1
- i s a wellmk
collBll
k 5 0 . Taking N 1 -2- 6,
definedpositivequantityfor
(3.8)followsdirectly
'
Vk,Vo
where Vo i s a f i n i t ec o n s t a n t .T h e r e f o r e ,
a uniformly bounded sequence.Furthermore,there
existpositiveconstants
g o , g1 such t h a t
@k i s
from ( 3 . 9 ) ,( 3 . 1 0 ) .
N-1
I C
sup
N+m
A V k < 0 whenever
ROBUSTNESS ANALYSIS
4.
D e = {ek:lim
k 1 0 , (4.5)impliesthat
'
"
N> 0
[ e k [(q6$+q7ul
(4.1)
k=O
and
Hence
($L<k)'
-2
mk
Step 2:
1 @kWk1
.
-
The "Smallness" of
I n orderto
mk
I
I @kwkI i s smallinthe
7
prove that(4.6)impliesthat
mk
0> i s an
mk
Proof of Lema4.1.Since
have
uk
[u:'
T
) T , ~ : 2 ) T , y k ] we
Corollary4.1.
In theabsence of p l a n tu n c e r t a i n t i e s ,
i . e . , u=O, thecontroller(3.1)-(3.9)guaranteeszero
r e s i d u a lt r a c k i n ge r r o r ,i na d l i t i o nt os t a b i l i t y .
Proofof
Theorem 4.1. The proof of Theorem 4.1 i s
completed i n thefollowingthreesteps:
(e)'
Step 1 : Boundednessof
i n[ l ] ,t h et r a c k i n ge r r o re k
Wm(z)@lukt
T
Noting t h a t u k = @kuk+
r k +e t T u k , i t can be shown t h a t
Using t h e d e f i n i t i o n of e* given
t osek@k.
T
ek
ek and "Smallness"of
pnk
nk = A ( Z ) ' J k
- 6*'(zI - F)-'g]uk-
Wm(z)[B; +
6 * T ( ~ -I F)-lg]yk + uA(z)uk
(4.8)
can be expressed[24]as
yk = W,(z)[l
(4.2)
Lemma 4.2.
- F)-'~II.
n * = n-m i s t h e r e l a t i v e d e g r e e
T
Then from (3.6) and (4.2) we have ~ ~ = @ ~ and,
< ~ + m ~
t h e r e f o r e ,( 3 . 5 ) may
be
writtenas
from
mk
(4.3)
We consider the following positive definite function
7
Then A V k = Vktl
- Vk
7
maybe
Proofof
Lemma 4 . 2 . Writing ckti = z ' W , ( z ) ~ ; ~ , we note
t h a t z'W,(z) i s a proper transfer function for
i = 0,
1 ,...,n* and uk i s bounded by mk. The proof of (4.19)
thenfollowsas
an immediate application of Lemma 3.1.
Lemma 4.3.
written as
If
"
(4.5)
840
(4.10)
is satisfied for
some E ~ [0,1]
t h e r ee x i s tp o s i t i v ec o n s t a n t sc 3 ,c 4s u c ht h a t
Il@k+l- Okl!
'
f '3%
P r o o fo fL e m a4 . 3 .
(4.11)
c4u
From (4.3) we o b t a i n
R4 = (N}\R3
(4.21)
we have
Then t h e f o l -
l o w i n gr e l a t i o n s h i p sh o l d :
Q3cnl and
M
3
M
iR2
N
i
c
l
R4
,
and M 2 ( M 4 5 N
(4.22)
(4.13)
f o r some k > 0.
I n v i e wo f( 4 . 1 3 )
The i n e q u a l i t y ( 4 . 6 ) t o q e t h e r w i t h t h e f a c t t h a t
kd+N-1 - T 2
($kck)
T
M
2
C
~
o
ti
usek@kf o r any koO
,
-2
k= ko
mk
we c a nw r i t e
2 Al
A0
M < + 1-1 - N.
a t m o s tn * + lp o i n t so f
i n (4.12)andusingthe
/ m k < 1, we o b t a i n
R1 f r o mb e l o n g i n gt oR 3 ,
l o w st h a tM 4 i M 2 ( n * * + 1 )a n dt h e r e f o r e
(4.15)
N - - (n*tl)
~ T ~ 5 e , k~ = n,n+l
@ ,. .~.n+n*
it fol-
M3,N-M2(n*t1)
we o b t a i n
i n D2 canprevent
E0
2-Eo
4,13),(4.14)
Sinceeachpoint
i m p l yt h a t
Hence
(4.16)
mk
issatisfiedfor
some
Then
[0,1].
E
T h e r e f o r e ,t h e r ee x i s t sp o s i t i v ec o n s t a n t s
3,4 s u c ht h a t
ko+N-1
(4.17)
'
'
holdsfork=n,n+l,..
c59 c 6 '
P r o o fo f
Lemma 4.4.
.n+n*and
some p o s i t i v e c o n s t a n t s
k=ko
T
1 @kwk
I
c fl$t
- -
pf2+u2
mk
fi,
2
0
fq
i = 1,2,
(4.24)
NEO
The sequence uk c a nb ew r i t t e n
as
I @kok1
mean f o r
mk
s m a l l p and c0.
(4.18)
f o r some s c a l a r s di,
i = 0,1,.
..n*.
Step3:
Boundedness o fS i g n a l sa n d" S m a l l n e s s "
I t can be shown t h a t ek may be expressedasthe
k'
o u t p u to ft h ef o l l o w i n gn o n - m i n i m a ls t a t er e p r e s e n t a ti on
T
e k + l = Acekt bc@k@k pbcl?lkt
pbc2n2k 9
of
-
Hence
'
- T-
(4.25)
ek - hcek
pn1k
where ~ I ~ ~ = A ~ ( Z ) U ~ , n 2 k = A 2 ( ~ ) ~ k , A ~ ( Z ) i sand
proper
t h ep o l e so fT 2 ( z / p o )a r es t a b l e ,
Ac i s s t a b l e and
where dm = max di
~
I.
Using ( 4 . 2 6 )a n dt h er e s u l t so f
Lemmas 4.2,4.3
i n ( 4 . 1 9 ) , ( 4 . 1 7 )f o l l o w sd i r e c t l y .
L e t u s now c o n s i d e r a t i m e i n t e r v a l c o n s i s t i n g o f
samplesand d e f i n e t h e f o l l o w i n gs e t s
hTc ( z I - A c ) -1 b = W,(z).
C
Similarly,
ck
can be represen-
t e da st h es t a t eo ft h es y s t e m
N
(4.26)
841
k tN-1
us now choose t h e
where A i s a s t a b l em a t r i x .L e t
positivedefinitefunction
f o r some p o s i t i v ec o n s t a n t sq 5 ,q 6 ,q 7 .L e t t i n g
P > 0,
where a , B a r e p o s i t i v e c o n s t a n t s a n d P =
T
T
P1 = P1 > 0 s a t i s f y t h e Lyapunov equations AcPAc P = -I,
A TPIAT
IlZ I] tu-r In
1 t T2
o1 kl k
UsingLema3.1
f o r some c o n s t a n t s
P r o o f o f C o r o l l a r y 4.1. The p r o o fo fC o r o l l a r y
P o l l o w s d i r e c t l y from t h e p r o o f o f Theorem4.1by
+ lykl 5
P1 = -I r e s p e c t i v e l y . N o t i n g t h a t l u k l
T ~ T, ~ ,T ~ .
boundedand
t h ef a c tt h a tI $ i w k / / m ki s
$ 11gk112-r3mi-$
a1 Ba31
a2mk[73
t o a3,
T~
stants.Choosing
uo = min [
1
0
1l<k! 2
- l'kII
ck
71
,13
f - UT5
761
and
one
(4.28)
t o 'c6, y2, s1 , s 2 a r e p o s i t i v e c o n a
2
B = 8'r6, a = 4(al t B + a3and
73
When k i s unknown,two
a d d i t i o n a le s t i m a t e dp a P
rametersneed t o be i n t r o d u c e d : one f o re s t i m a t i n g
[ F -aa~u
f o re s t i m a t i n gl / k
The c o n t r o l a wi sg i v e n
P'
bY
' Y2taslIlekllI$~WkI
tas21@kwkI
where a.
ask+mandthenuse
bounded, i t can
5.
p = 0, + k w k + O
4.1
first
(4.31) t o show t h a t e k + O a s k + - .
be shown t h a t
v k + l -'k
N+m,
( 4 . 1 )f o l l o w s .
u k = ekwk + c
(5.1)
Ok
where
ek,
wk a r ea sd e f i n e d
e s t i m a t eo f
k,/k,
i n s e c t i o n 3 and c
isthe
Ok
The f o l l o w i n g a u x i l i -
a ti n s t a n tk .
arysignalsaregenerated
u E [O,u,I
T
c k = 8k<k
we have
- wm(z)ukt: :oc
; E k = y k -Y;'
$kck
(5.2)
(4.29)
f o r some c o n s t a n t s 0 < y o <1 and y1 > 0.
same stepsas
shown t h a t
where qk i s t h e e s t i m a t e o f
k / ka ti n s t a n t
P m
i sg i v e nb y( 3 . 4 ) .O e f i n i n gt h ee s t i m a t e dp a r a m e t e r
T
T
theupdatelaw
v e c t o r Bk as Fk = [8kYco ,$kl
k
chosenas
F o l l o w i n gt h e
i n t h ep r o o fo fL e m a4 . 4 ,
i t can be
k and ck
is
(4.30)
(5.3)
for each
u E [O,ul)
2
by choosing O < ~~5( 1 - y o ) /16y4
( 1t y o ) q 3 2
= ( 3 t y o ) / 2 and
li*=min[pl
,p2] i t f o l -
therefore
i n t h ec l o s e dl o o pa r e
cky
mk, ekand
allthesignals
bounded.
T m
i k ' m k t2< k <-T-k Y Mo,2(18*I(
where r k = [ckYykY<klTY
max(E;,G*),
and mk, i s as d e f i n e d i n s e c t i o n 3 .
Remark 5.1.
I nt h eo - m o d i f i c a t i o nu s e d
i n (3.5),(5.3)
u s w i t c h e sf r o m 0 t o uo whenever llekl\, llekll exceed
M , No r e s p e c t i v e l y . An a l t e r n a t i v ea p p r o a c hi st o
0
use a d i f f e r e n t u f o r eachelement o f B k o r Bk. Such
anapproach w
i
l n o t change t h e s t a b i l i t y r e s u l t s a n d
i t may have a b e n e f i c i a l
proofsofthealgorithm,but
effectonthetransientofthesignalsintheclosed
1oop.
(4.31)
f o r some p o s i t i v e c o n s t a n t s
3.1and
From (4.2) i t f o l l o w s t h a t
Using(4.24),Lema
e * = [ ~ * T , c ~ , ~ , ~ * ~ T , y m a x ( 1 , P ) + 2 u o2,<
k > 0.
q3, q4andany
o--_
t h e boundednessof mk , i t
Remark 5.2.
I n t h i s paper i t i s assumed t h a tn > m ,
i.e. , therelativedegreeofthenominaltransfer
f u n c t i o no ft h ep l a n ti s
n*,l.
When
n=m,
t w op o s s i b l e
f o l l o w s from ( 4 . 3 1 )t h a t
842
r o b u s ta d a p t i v ec o n t r o l l e r sc a n
be d e s i g n e d w i t h
s i m i l a r p r o p e r t i e s as t h o s eg i v e nb y
Theorem4.1and
Corollary4.1.
S a s t r y , S. S., "ModelReferenceAdaptiveControl:
Stability, Parameter Convergence and Robustness
I.M.A. J o u r n a lo fC o n t r o l
and Information,Vol.1,
PP. 27-66,1984.
The secondone
= 1 andthe
same
f o r m o f f i l t e r s as i n ( 3 . 1 ) , ( 3 . 2 ) b u t w i t h
F b e i n g an
n x nm a t r i xi n s t e a d
o f ( n - 1 ) x ( n - 1 ) .I nt h i sc a s e
c k = u k and t h e p r o o f o f s t a b i l i t y androbustness i s
much s i m p l e r t h a n t h a t
o f Theorem4.1
oftheproofis
no l o n g e r needed.
6.
sinceStep
CONCLUSION
,I'
Ioannou. P. A. " R o b u s tA d a o t i v eC o n t r o l l e rw i t h
ZeroResidualTrackingErrors,"
I E E E Trans.on
AutomaticControl,Vol.
AC-31, No. 8, Aug. 1986.
Kosut, R. L.and
C. R. Johnson, Jr., "An I n p u t OutputView o f Robustness i n A d a p t i v e C o n t r o l ,"
Automatica,Vol.
20, No. 5, Sept.1984.
P r a l y , L. "Robustness o f ModelReferenceAdaptive
C o n t r o l ," 3 r d Y a l e Workshopon A p p l i c a t i o n s o f
A d a p t i v e SystemsTheory,June1983.
REFERENCES
Pralv,L."Robust
ModelReferenceAdaDtive
t r o l j e r s ,P a r t1 :S t a b i l i t yA n a l y s i s ' , "
23rd I E E E CDC, 1984.
Con-
&
K r e i s s e l m e i e r . G. and B . D. 0. Anderson."Robust
," IEEE-Trans.
ModelReferenceAdaptiveControl
on AutomaticControl,Vol.
AC-31, No. 2,
Feb.1986.
E g a r d t , B., " S t a b i l i t y A n a l y s i s o f A d a p t i v e
ConJACC,
t r o l Systems withDisturbances,"Proc.
San F r a n c i s c o , CA, 1980.
Egardt, B., S t a b i l i t y o f A d a p t i v eC o n t r o l l e r s ,
S p r i n g e r - V e r l a g ,B e r l i n ,1 9 7 9 .
K r e i s s e l m e i e r , G., "A R o b u s tI n d i r e c tA d a p t i v e
ControlApproach,"4thYale
WorkshoponTheory
and A p p l i c a t i o n s o f A d a p t i v e C o n t r o l ,
May 1985.
Goodwin, G. and K . S. S i n ,A d a p t i v eF i l t e r i n g
P r e d i c t i o n a n dC o n t r o l ,P r e n t i c e - H a l l ,
1984.
Narendra, K. S., Y. H. L i n andL.
S. Valavani,
" S t a b l eA d a p t i v eC o n t r o l l e rD e s i g n ,P a r t1 1 :
ProofofStability,"
I E E E Trans.onAutomatic
Control,Vol
AC-25, June1980.
yer
R i e d l e , B. D. and P. V . K o k o t o v i c ," D i s t u r b a n c e
IEEE
I n s t a b i l i t i e s i n anAdaptiveSystem,"
Trans. on AutomaticControl,Vol
AC-29,
pp.822-824.
K r e i s s e l n e i e r , G. and K. S. Narendra,"Stable
ModelReferenceAdaptiveControl
i n t h e Presence
o f Bounded Disturbances," I E E E Trans.onAutom a t i cC o n t r o l ,V o l .
AC-27, No. 6, pp.1169-1176,
1982.
843