Completion of the squares in the finite horizon $H^\infty$
control problem by measurement feedback
Citation for published version (APA):
Trentelman, H. L., & Stoorvogel, A. A. (1989). Completion of the squares in the finite horizon $H^\infty$ control problem by measurement feedback. (Memorandum COSOR; Vol. 8927). Technische Universiteit Eindhoven.
Document status and date: Published: 01/01/1989
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne Take down policy
If you believe that this document breaches copyright please contact us at: openaccess@tue.nl
providing details and we will investigate your claim.
EINDHOVEN UNIVERSITY OF TECHNOLOGY
Department of Mathematics and Computing Science
Memorandum COSOR 89-27
Completion of the squares in the finite horizon Hoo control problem by measurement feedback
H.L. Trentelman & A.A. Stoorvogel
Eindhoven, October 1989
COMPLETION OF THE SQUARES IN THE
FINITE HORIZON
HOOCONTROL
PROBLEM BY MEASUREMENT FEEDBACK
by
H.L. Trentelman& A.A. Stoorvogel
Department of Mathematics and Computing Science P.O. Box 513
5600 MB Eindhoven, The Netherlands
ABSTRACT
In this paper we study the finite horizon version of the standardHoo control problem by measure-ment feedback. Given a finite-dimensiona1linear, time-invariant system, together with a positive real numbery, we obtain necessary and sufficient conditions for the existence of a possibly time-varying dynamic compensator such that theL2[O,T]-induced norm of the closed loop operator is smaller thany.These conditions are expressed in terms of a pair of quadratic differential inequali-ties. generalizing the well-known Riccati differential equations that were introduced recently in the context of finite horizonHoocontrol.
-
2-1. THE FINITE HORIZON HooCONTROL PROBLEM
Consider the linear time-invariant system x(t)
=
Ax (t)+
Bu (t)+
Ew (t) ,y(t)=Ctx(t)+ D\w(t) ,
z(t)=C2x(t)+D2u(t) ,
(1.1)
where x E /Rn is the state, U E /Rm the control input, WE /Rl an unknown disturbance input,
y E JRP the measured output and zE /Rq the output to be controlled. A, B, E, Ct ,D \ ,C2 and
D2 are constant real matrices of appropriate dimensions. In addition, we assume that some fixed
time interval [0,T] is given. We shall be concerned with the existence and construction of dynamic compensators of the form
p(t)=K(t)p(t)+L(t)y(t) ,
u(t) =M(t)p(t) +N(t)y(t) , (1.2)
whereK, L, M andNare real, matrix valued, continuous functions on [0,T].The feedback inter-connection of LandLF is the linear time-varying system Lcl described by
[
X(t)] _ [A +BN(t)C\ BM(t)] [X(t)] [E+BN(t)Dt]
pet) - L(t)C t K(t) pet)
+
L(t)D t wet) ,[ X
(t)]
z(t)=(C2+D2N(t)C t D2M(t) pet) +D2N(t)D t wet) .
Let us denote the matrices appearing in these equations byAe(t), Ee(t), Ce(t) andDeCt), respec-tively. Obviously, if we put
x
(0)=
0,p(0)=
0, then the closed loop systemLel defines a convo-lution operatorGel:L~[0,T] ~ L~[0,T] given byI
(Gclw)(t)=z(t)=
J
Ce(t) <I>eCt, 't)Ee('t) w('t)d't+DeCt) w(t) ,o
where<De(t,t) is the transition matrix ofAe(t).Thus, the influence of disturbancesw E L~[0,T]
on the output z can be measured by the operator norm ofGel,giveninthe usual way by IIGFwll2
IIGelll :=sup { IIwll2 10*-W E L~[0,T]} .
Here,IIxll2 denotes theL2[O, Tjnorm of the functionx. The problem that we shall discuss in this
paper is the following:
giveny> 0, find necessary and sufficient conditions for the existence of a dynamic compensator LF such that IIGelli
<
y.(2.1)
(2.2) -
3-The problem as posed here will be referred to as thefinite horizonHoo control problem by meas-urement feedback. This problem was studied before in [61 and [21. In the latter rererencesit is however assumed that the rollowing conditions hold: DI is surjective, D2 is injective. In the
present paper we shall extend the results obtained in [6J and [2] to the case that01 andD2 are arbitrary.
2. QUADRATIC DIFFERENTIAL INEQUALITIES
A central role in our study of the problem posed is played by what we shall call the qua-dratic differential inequality. let 'Y>
°
be given. For any differentiable matrix function P:[0,T] ~ /Rnxn,defineFy(P):[0,T] ~ JR(n+m)x(n+m) by[
P+A'P+PA+C2'C2+
~PEETP
PB+C2'D2]Fy(P):= B'P+D 2'C 2 D 2 ' D 2 ·
IfF iP)(t)~
°
for all tE [0,T], then we shall say thatP satisfies the quadratic differential ine-quality (at y). Also a dual version of (2.1) will be important to us: for any differentiableQ:[0,T] ~ /Rnxndefine Gy(Q): [0,T] ~ JR(n+p)x(n+p) by:
[
-Q
+AQ +QA' +EE'+
~
QC 2'C 2Q QC 1' +ED1']Gy(Q):= C 1Q +D1E' D1 D1"
IfGy(Q)(t)~
°
forall t E [0,T] then we shall say thatQsatisfies the dual quadratic differentialinequality (aty). In the sequel let
G(s):=C 2(ls-Ar1B+D2 ,
H(s):=C1(ls -A)-IE +D1
denotc the open loop transfer matrices from uto z andw to y,respectively. Furthermore, denote by normrank(G) and normrank(H) the ranks of these transfer matriccs considered as matrices with entries in the field of real rational functions. We are now ready to state our main result:
THEOREM 2.1. Let'Y
>
0. The following two statements are equivalent: (i) There exists a time-varying dynamic compensatorLFsuch that IIGcllI<
y,(ii) There exist differentiable matrix functionsP andQ: [0,T] ~ JRnxn such that
-4-(b) rankF y(P)(t)=nonnrank (G) \;ItE [0,T] , (2.4)
(c) Gy(Q)(t)~O\;ItE [O,T]andQ(O)=O, (2.5)
(d) rankG..,(Q)(t)=nonnrank(H) \;ItE [0,T] , (2.6)
(e)
Yl-
Q(t)P(t)is invertible \;ItE [0,T]. (2.7)The aim of this paper is to outline the main steps and ideas involved in a proof of the latter theorem. For a more detailed discussion we would like to refer to[5].
It can be shown that, in general, ifF ..,(P)~
°
on[0,T] thenrankF..,cP)(t)~nonnrank(G) on [0,T]
and, likewise,ifG..,(Q)~ Oon[0,T] then
rankGy(Q)(t)~nonnrank(H)on lO,TJ
This means that the pair of conditions (2.3), (2.4) can be refonnulated as:Pis a rank-minimizing solution of the quadratic differential inequality at y, satisfying the end-condition P (T)
=
0. A similar restatement is valid for the conditions (2.5), (2.6).Itcan also beshown that ifP satisfies (2.3) and (2.4) then it is unique. Also, this unique solution turns out to be symmetric for all t E lO,T].The same holds forQsatisfying (2.5) and (2.6).We will now show that for the special case thatD1andDz are assumed to be surjective and injective, respectively, our Theorem 2.1 specializes to the results obtained before in [6] and [2]. Indeed, ifD
z
is injective then of coursenonnrank(G)
=
rankDz=
mDenote
Ry(P)
:=P
+A'P+
PA+
C z' C z+
~
PEE'P - (PB+
C z' Dz)(D z' D zr
1(B'p +Dz'C z)Clearly,R ..,(P) is the Schur-complement of D
z'
Dz
inF ..,(P). Therefore we haverankF ..,(P)(t)
=
m+
rankR..,(P)(t)for all t E lO,T]. This implies that the pair of conditions (2.3), (2.4) is equivalent to the condi-tion: P is the solution of the Riccati differential equation Ry(P)
=
°
with terminal condition P (T)=
0. A similar statement holds forQsatisfying (2.5) and (2.6). Thus we obtainCOROLLARY 2.2. Let y> 0. Assume D1is surjective and D
z
is injective. Then the following statements are equivalent:5
-(i) There exists a time-varying dynamic compensator}:.F such that IIGe/1i
<
y.(ii) There exist differentiable matrix-functions P and Q: 10,T] ~ /Rnxn such that for all
t E [0,T]
-p
=A'P+
PA+
C z' C z+
~
PEE'P - (PB+
C z' Dz)(D z' DZ)-l(B'P+
D z' Cz), P(T)=
0 andwith, in addition,
y.l -
Q(t)P (t) invertible for allt E [0,T]o
Itcan be shown that if the conditionsinthe statement of Theorem 2.1 (ii) indeed hold, then it is always possible to find a suitable compensator with dynamic order equal to
n,
the dynamic order of the system to be controlled.3. COMPLETION OF THE SQUARES
In this section we shall outline the proof of the implication (i) =:> (ii) of Theorem 2.1. Con-sider the system }:.. For given
u
andw, letzu,w denote the output to be controlled, withx
(0)= O. Our starting point is the following lemma:LEMMA 3.1. Lety
>
O. Assume that for all 0:t:-wE L~ [0,T] we haveinf{lIzu,wllz-yllwllz I
u
E L~[O,T]}<
0 . (3.1)Then there exist a differentiable matrix function P: [0,T] ~ /Rnxn such that Fr<P)(t)~0 'titE [0,T], P(T) =0 andrankF y(P)(t) =normrank(G)'tit E [0,T].
PROOF. A proof of this can be given by combining the result of [2, Theorem 2.3] with ideas
used in the proof of [4, Theorem 5.4].
0
Now, assume that the condition (i) in the statement of Theorem 2.1 holds, Le. assume there exists a dynamic compensator }:.,.. such that liCe/II <y. Then condition (3.1) holds: let
w E L~[0,T] and w :t:-0 and let zbe the closed loop output with x(0)
=
0 andp(0)=
O. Then-
6-and hence IIzu.wll - yIIwllz <O. Then also the infimum in (3.1) is less than O. We may then con-clude that, indeed, a differentiable matrix functionPexists that satisfies(2.3)and(2.4).
The fact that also (2.5)and(2.6)hold can be proven by the following dualization argument. Consider the dual system
~=A'l;+CI'V+ Cz'd ,
r:
T\=B'l;+
Dz'd ,C=E'l;+D1'v ,
and apply to
r:
the time-varying compensatorq
=K'(T - t)q +M'(T - t)T\ , v=L'(T-t)q+N'(T-t)T\ .It can be shown that if we denote by
Gel
the closed loop operator ofr:
xLF' (withC(O)
=
0,q(0)=
0), and it G~l denotes the adjoint operator ofGel then the following equality holds:(3.2) where R denotes the time-reversal operator (Rx)(t):= x(T - t). Now, if IIGelll <y then also IIG~lll
<
Yand therefore, by 3.2,IIGelII<
y. We can therefore conclude that the quadratic differen-tial inequality associated withr:
has an appropriate solution, say P(t), on [0,T]. by definingQ(t):=P(T - t)we obtain a functionQthat satisfies(2.5)and(2.6).
Finally, we have to show that condition(2.7)holds. We shall need the following lemma: LEMMA 3.2. Assume that there exists P:[0,T] ~ JRnxn such that F y(P)(t)'2:.0, "iItE [0,T], and rankFy(P)(t)
=
normrank(G),"iIt E [0,T].Then there exist continuous matrix functions Cz.p andDp such that for allt[
c
ZP'(t)]
FiP)(t)=
Dp~(t)
(Cz.p(t) Dp(t)) .AssumeFiP)is factorized as in(3.3). Introduce a new system, sayLp,by
(3.3)
-
7-Xp
=
(A+
~
EE'P)xp+
Bup+
Ewp ,yp =(C}
+
~D}E'P)XP
+
D}wp, Zp= C2,pXp+
Dpup .(3.4)
We stress that ~p is a time-varying system with continuous coefficient matrices. If ~F is a dynamic compensator of the form (1.2).letGp,cI denote the operator from wp tozp obtained by interconnecting~pand~F'
The crucial observation now is thatIIGclll
<
Yif and only ifIIGp,clll<
Y.that is.a compensator~F"works" for ~
if
and onlyif
it "works" for ~p! A proof of this can be based on the following "completion of the squares" argumentLEMMA 3.3. Assume thatP satisfies (2.3) and (2.4). Assumexp(O)=X(0)=0, up(t)=u(t) for all t E [0,T] and suppose that Wp and ware related by wp(t)
=
wet) - y-2E'P (t)x (t) for all t E [0,T].Then for alltE lO,T] we haveliz(t)1I2- yllw (t)1I2
=
.!L.(x'(t)P (t)x (t))+
IIzp (t)1I2 - yllwp(t)1I 2dt Consequently:
IIzll~ - yllwll~
=
IIzpll~- yllwpll~ (3.5)PROOF: This canbeproven by straightforward calculation, using the factorization (3.3).
0
THEOREM 3.4. Let P satisfy (2.3) and (2.4). Let ~F be a dynamic compensator of the form (1.2).Then
PROOF. AssumeIIGp •c/1i
<
yand consider the interconnection of~and~F' p Up .::-'.:- w~p
c..--p~F
y..::-
'1/1~
-y 1~F
Let
°
:tow E L~ [0,T], let x be the corresponding staLe trajectory of ~ and defineWp :=w -
y2
E'px. Then clearlyyp=y, xp=x and thereforeUp=u.This implies that the equal-ity (3.4) holds. Also, we clearly have8
-IlzpJI~- .yzllwpll~S (IIGp.c/112 - .yz~lwpll~ (3.6) Next, note that the mapping
lIJ>
H wp+y-2 E'Pxpdcfines a bounded operator fromLi
[0.T) toLi[o.
T). Hence there exists a constant ~>
0 such that IIwpll~ >~lwll~. DefineS>
0 by 52:=r
-IIGp,clI12. Combining (3.4) and (3.5) then yieldsIIzll~- .yznwll~S-52~lwll~ .
Obviously. this implies thatllGc/1I2
sf -
()2~<
f.
IJ
We will now prove that(2.7)holds. Again assume that:Ep yields 1IGclll
<
y. By applying a versionof Lemma
3.1
for time-varying systemsitcan then be proven that thedualquadratic differential inequality associated with:Ep :_
[-1'
+ApY+
YAp' +EE'+
_~
YC2,P'C2,pYG (Y)'- T
. , . - CI,PY +D E'I
has a solutionyet)on [0.T].satisfyingYeO)
=
0 and_
.
[IS -
Ap(t) -E]rankG.,(Y)(t)
=
nonnrank CI.P(t) DI
-n
(3.7)for allt e [0,T). Here, we have denotedA p =A +y-2EE'Pand CI,P
=
CI +y-2DIE'P. Further-more. it can be shown that Yisunique
on each interval [0,tI](tI Sn.
Ontheother hand. it canbe proven that on each interval [0,
ttl
on which 1-QP is invertible, the functionY:=
(1-QPrlQ satisfiesGy(Y)(t)~
0,YeO)
=0 and the rank condition (3.7). Thus on any such interval [0, td we must haveY(t)=yet). aeady, since Q(O)=O, theree.xistsO<
tlS Tsuch thatI - QP is invertible on [0,tI). assume now that tI
>
0 is the smallest real number such that1- Q(tl)P(tl)is not invertible. Then on [0,tl)we have
Q(t)=(/- Q(t)P(t)) yet)
and hence, by continuity
-Q(tI)=(/-Q(tl)P(td)Y(tI) .
(3.8)
There exists x
*
0 such that x'(I - Q(tI)P(tI»=
O. By (3.8) this yields x'Q(tt>
=
0 whencex'
=
0, which is a contradiction. We must conclude that I - Q (t)P (t) is invertible for all t e [0,T].- 9-4. EXISTENCE OF COMPENSATORS
In the present section we will sketch the main ideas of our proof of the implication (ii) ~
(i) of Theorem 2.1. The main idea is as follows: starting from the original system L we shall define a ncw system,Lp,Q'which has the following important properties:
(1) Let Lp be any compensator. The closed loop operator Gel of the interconnection LX Lp
satisfies IIGclll
<
Yif
and onlyif
the closed loop operator of Lp,Qx
Lp, say Gp,Q,ch satisfiesIIGp,Q,clll
<
y.(2) The system LP,Q is almost disturbance decoupable by dynamic measurement feedback, i.e. forallE
>
0there existsLpsuch thatIIGp,Q,clll<
E.Property(1)states that a compensatorLF"works" forL.if and only if it "works" forLp,Q' On the other hand, property(2)states that,indeed, there exists a compensatorLF that "works" forL.P,Q:
take any£$ yand take a compensatorLFsuch thatIIGp,Q,clll
<
E. Then by, property(1),IIGclll<
ysoL.F works forL..This would clearly establish a proof of the implication(ii) ~ (i) in Theorem 2.1.
We shall now describe how the new system L.P,Q is defined. Assume that there existPand
Qsatisfying(2.3)to(2.7). Apply Lemma3.2 to obtain a continuous factorization(3.3)ofFyep) and let the system L.pbe defined by(3.4).Next, consider the dual quadratic differential inequality GlY)2 0 associated with the systemLp,together with the conditions YeO)
=
0 and the rank con-dition (3.7). As was already noted in the previous section, the conditions (2.5), (2.6) and (2.7) assure that there exists a unique solution Yon [O,T]. (In fact, yet)=(1'
I-Q(t)p(or1Q(t).) Now, it can be shown that there exists a factorizationwithEp,Q andDp,Q continuous on[O,T].Denote
Ap,Q(t):=Ap(t)+yet) C2,P'(t)C2,P(t) , Bp,Q(t) :=B
+
Y(t) C 2,P'(t) Dp(t).Then, introduce the new systemLp,Q by:
Xp,Q
=
Ap,Q xp,Q+
Bp,Q up,Q+
Ep,Q wp,Q 'YP,Q
=
Cl,pXp,Q+
Dp,Q wp,Q 'zp,Q
=
C 2,pxp,Q +Dp up,Q.Again,L.P,Q is a time-varying system with continuous coefficient matrices. We note that L.P,Q is in fact obtained by first transforming L into Lp and by subsequently applying the dual of this transformation toL.p. We shall now first show that property (1)above holds. IfL p is a dynamic compensator, then letGp,Q,cIbe the closed loop operator fromwp,Q tozp,Q in the interconnection
-10-P,Q Up,Q 1.:.£ f -1/1
"
~1',Q Q'B
F z YP,-:::-
1/1 f-"
~ y U"
~FRecall thatGel denotes the closed loop operator from wto z in the interconnection of LandLF'
We have the following: THEOREM4,1.
PROOF. Assume LF yields IIGelll
<
y. By Theorem 3.4 then alsoIlGp,elll<
y, Le.• LF intercon-nected withLp (given by 3.4) also yields a closed loop operator with norm less than y.Itis easily seen that the dual compensator"£/
(see section 3), interconnected with the dual ofLp:~=Ap'(T -t)~
+
Cl,p'(T -t)v+
C2,p'(T -t)dLp' T]=B'~+ DP'(T-t)d
~ =E'~+OI'V
yields a closed loop operator
6
P,el(fromdto~) with116
P,elll<
y. Now, the quadratic differential inequality associated withLp' is the transposed. time-reversed version of the inequality G-yCY);:: 0 and therefore has a unique solutionyet)
=yeT -
t)such thatyeT)
=0 and the corresponding rank condition (3.7) holds. By applying Theorem 3.4 to the system LP' we may then conclude that the interconnection of L/ with the dual LP,Q' of LP,Q yields a closed loop operator with norm less thany. Again by dualization we then conclude that IIGp,Q,clll <y. The converse implication ispro-ven analogously.
0
Property (2) is stated formally in the following theorem:
THEOREM 4.2. For all e>0 there exists a time-varying dynamic compensator LF such that
IIGp ,Q,clll
<
y.0
Due to space limitations, For a proof of the latter theorem we refer to [5]. By combining theorems 4.1 and 4.2 we immediately obtain a proof of the implication (ii) ~ (i) in Theorem 4.2.
11 -5. CONCLUDING REMARKS
In this paper we have studied the finite horizonHoo control problem by dynamic measure-ment feedback. We have noted that the results obtained can be specialized to re-obtain results that were obtained before [6] and [2]. The development of our theory runs analogously to the theory developed in [4] and [3] around the standardHoo control problem (theinfinite horizon version of the problem studied in the present paper). In the latter references the main tools are the so-called quadratic matrix inequalities, the algebraic versions of the differential inequalities used inthe present paper. For the special case that D1 is surjective and D2 is injective these quadratic matrix inequalities reduce to the algebraic Riccati equations that were also obtained in [6] and [ll.
REFERENCES
[1] 1. Doyle,K. Glover, P.P. Khargonekar, B.A. Francis, "State space solutions to standard H2
andH00 control problems", IEEE Trans. Aut. Contr., Vol. 34, No.8, 1989, pp. 831-847.
[2] D.J.N. Limebeer, B.D.O. Anderson, P.P. Khargonekar, M. Green, "A game theoretic approach toHoo control for time varying systems", preprint, 1989.
[3J A.A. Stoorvogel, "The singularH00control problem with dynamic measurement feedback",
preprint, 1989, Submitted to SIAM J. Contr.& Opt.
[41 A.A. Stoorvogel& H.L. Trentelman, "The quadratic matrix inequality in singularH00
con-trol with state feedback", preprint 1989, To appear in SIAMJ. Contr.& Opt.
r
51
A.A. Stoorvogcl & H.L. Trentelman, "The finite horizonHoo control problem with dynamic measurement feedback", in preparation.l6] G. Tadmor,"H00 in the time domain: the standard four blocks problem", 1988, To appear in
EINDHOVEN UNIVERSITY Of TECHNOLOGY
Depanrnent of Mathematics and Computing Science
PROBABIUTY
THEORY, STATISTICS.
OPERATIONS
RESEARCH
AND SYSTEMS
THEORY
P.O. Box 513
5600 MB Eindhoven - The Netherlands
Secretariate:
Dommelbuilding 0.03
Telephone:
040 - 47 3130
List of COSOR-memoranda - 1989
Number
Month
Author
Title
M 89-01
January
D.A. Overdljk
Conjugate profiles on mating gear teeth
M 89-02
January
A.H.W. Geerts
A priori results in linear quadratic optimal control theory
M 89-03
February
A.A.Stoorvogel,
The quadratic matrix inequality in singular
H _control with stall
H.L. Trentelman
feedback
M 89-04
february
E. WilJekens
Estimation of convolution tail behaviour
N. Veraverbeke
M 89-05
March
H.L. Trentelman
The
totalJy singular linear quadratic problem with indefinite cost
M
89-D6
April
B.G. Hansen
Self-decomposable distributions
andbranching processes
M 89-07
April
B.G. Hansen
Narc on Urbanik's class
L.
M 89-08
April
B.G. Hansen
Reversed self-decomposability
M
89-09April
A.A.Stoorvogel
Thesingular zero-sum differential game with stabiJity using
H _con·
troltheory
M 89-10
April
L.J.G. Langenhoff
Ananalytical theory of multi-echelon production/distribution systems
W.H.M.Zijm
-
2-Number
Momh
Author
Title
M 89-12
May
D.A. Overdijk
De geometric van de kroonwieloverbrenging
M 89-13
May
U.B.F.Adan
Analysis of the shonest queue problem
J. Wessels
W.H.M.Zijm
M 89-14
June
A.A. Stoorvogel
TIlesingular
H _control problem with dynamic measurement
feed-back
M 89-15
June
A.H.W. Geerts
Theoutput-stabilizable subspace and linear optimal control
M.L.J. Hautus
M 89-16
June
p.e.
Schuur
On the asymptotic convergence of the simulated annealing algorithm
in the presence of a parameter dependent penalization
M 89-17
July
A.H.W. Geerts
A priori results in linear-quadratic optimal control theory (extended
version)
M 89-]8
July
D.A. Overdijk
The curvature of conjugate profiles in points of contact
M 89-19
August
A. Dekkers
Anapproximation for the response time of an open CP-disk system
J. van der Wal
M 89-20
August
W.FJ. Verhaegh
On randomness of random number generators
M 89-21
August
P. Zwietering
Synchronously Parallel: Boltzmann Machines: a Mathematical Model
E. Aarts
M 89-22
August
U.B.F. Adan
Anasymmetric shonest queue problem
J. Wessels
W.H.M.Zijm
~
89-23
August
D.A. Overdijk
Skew-symmetric matrices in classical mechanics
~
89-24
September
F.W. Steutel
Thegamma process and
thePoisson distribution
J.G.F. Thiemann
.t
89-25
September
A.A. Stoorvogel
Thediscrete time
H _control problem: the full-information case
I 89-26
OCtober
A.H.W. Geerts
Linear-quadratic problems and the Riccati equation
M.L.J. Hautus
I 89-27