Documente Academic
Documente Profesional
Documente Cultură
OPTIMIZATION
L. C e s a r i D e p a r t m e n t of M a t h e m a t i c s of M i c h i g a n - A n n A r b o r (Michigan)
University
We p r e s e n t timal
here
existence
theorems
for m u l t i d i m e n s i o n a l calculus
problems
of op-
control
and
for p r o b l e m s
of the
of v a r i a t i o n s
concerning We theore-
integrals
of an e x t e n d e d
Lagrangian in the
domain.
interested
led to t h e s e
problems
of o p t i m a l
control
and
to e q u i v a l e n t
problems
of the
calculus
extendfor theo-
ed L a g r a g i a n s , the existence
and u n d e r theorems.
essentially Second,
equivalent
assumptions
the u n d e r l y i n g theorems,
semicontinuity
or e q u i v a l e n t
lower
closure
by an i n t e r p l a y Kuratowski's
of u p p e r property as sole
semicontinuity
sets:
(K) , or e q u i v a l e n t beside
requirement
convexity, sets.
property
(Q) of s u i t a b l e
auxiliary
I. A n e x t e n d e d We m a y consider
problem
of
the c a l c u l u s of m i n i m u m
of v a r i a t i o n s of the functional
the p r o b l e m
I[x~
f G
S,
LI (G),
(1)
A, domain
t6G
b > I, t =
and w h e r e
variable space
x is t h o u g h t possibly S (L P linear.
of as an e l e m e n t a Sobolev space
(X,~), of M:
X on G w i t h
~ . We not
think
(G)) s, L:
necessarily
84
: In
(yl,...,yS),
z(t)
(Lx) (t) =
(I) A is a g i v e n is G, a n d Thus, in
set of the
whose Q(t,y)
t-space
for a n y
(t,y) ~ A
a subset
z-space taken
is g i v e n . functions If w e
constraints
on the v a l u e s
b y the
(Fix) (t) a n d by M
the s e t of all (t,y,z) ~ I R u + s + r w i t h t ~ G , (t,y)6A, o z6Q(t,y), it is c o n v e n i e n t to take F = ~ ~ f o r all (t,y,z) 6 ~ u + s + r - M , o o U+S thus Fo(t,y,z ) = +~whenever t ~ u _ G, w h e n e v e r t ~ G , (t,y)6]R -A, and whenever constraint Indeed, t ~ G, (t,y) 6 A , z ~IRr-Q(t,y). W i t h this convention the
denote
the r e m a i n i n g t ~G, F
constraints. ( t , y ( t ) ) ~ A, o
for a.a.
hence
z(t) ~ Q ( t , y ( t ) ) will be u s e f u l
in G. T h e in the
convention
formulation
o of c o n d i t i o n s that
= + ~ in I R U + s + r - M on F o A. .
b y A(t)
the
(t,y) C
think
of S as a s u b s e t
space N,
X on G i n v o l v i n g 0 ! l~I ~ N, below
deriva-
D ~ X of o r d e r s be simply the
u p to a m a x i m a l
say,
D e x of o r d e r s the
order,
or 0 ~
lel
X whose given,
boundary or s a t i s f y
values given
~G,
or t r a c e s
y D ~ x, 0 ~
conditions,
An existence we state p.
theorem
can n o w be
stated
assumption
function
[I~,
Fo(t,y,zj that,
is a g i v e n
function
above,
for e v e r y
e > O,
is a c o m p a c t
(a) m e a s (G-K) < e; (b) the e x t e n d e d f u n c t i o n F res+r o to K x IR is B - m e a s u r a b l e ; and (c) for a l m o s t all t ~ G the function F o (y,z) has v a l u e s finite or + ~ and is
is o f t e n o j u s t as g e n e r a l as t h o s e of the
function
and
(C~)
is
recently stated
by Ioffe, condition
conventions
is a l s o
on A and t h e s e t s Q ( t , x ] . this o assumption, the N e m i t s k i i maps operator appearing in (I), n a m e l y into measur-
(y,z) + E
(t,y(t),z(t))
measurable
functions
y(t],z(t)
65
able We
functions
say t h a t
an o p e r a t o r xk x graph
if X k ~ has the
S, x ~ S , closed x ~S,
Px = y. We
say t h a t P
property say
if x k ~ that
Px = y. We x kx
P has
x k~S, that Px k
x ~X,
in X,
implies
that
there
is a s u b s e q u e n c e
y. If S = X, c l o s u r e and c l o s e d g r a p h p r o p e r t i e s coins The m o s t u s u a l p r o p e r t y that Xk~S , x ~ S , x k x in X, i m p l i e s in Y is r e f e r r e d (A l o w e r to as the continuity Under of P on S. condition function (C~), F
O
Px k P x (I.i) that
theorem). , the
assume
for a.a.
all y ~ A ( t )
extended
(t,y,z),
z~-IRris k=
convex
in z. Let bk t,,/,~_j, ~k(t) , ~(t) , Ik(t) , l(t) , t 6 G , functions on G, w i t h ~k~ ~, ~k 6 ( L p ( G ) ) r ' l ' l k ~ weakly in (L I (G))r,I~>A assume that
1,2,..., such
be m e a s u r a b l e that
L I (G), weakly
yk y in m e a s u r e Let
in G,
in LI (G).
~k (t)=Fo(t'yk(t) and
and
that
t ~ G,
~ ~ LI(G) , such
/G~(t)dt
for the e x t e n d e d t ~G
problem
(I)).
Under
condi-
that
for a.a. in
and y e A(t), us a s s u m e
the e x t e n d e d that F
O
Fo(t,y,z), O, t ~G,
z ~ I R r, is c o n v e x ~ 6 LI(G), and
z. Let
--
(I) for
--
c > O we h a v e
(t,y,z)
xEX
property,
the c o n v e r g e n c e T h e n the
in m e a s u r e functional
property, ICx ] in
property.
an a b s o l u t e shall sketch
(A T o n e l l i - t y p e where
as b e f o r e , 0 ~
is a s c a l a r ~ +
function
~ < + ~, b o u n d e d
%(6)/{
+ + ~ as
66
such S~ = also
that
F
0
(t,y,z)
and
(c) is r e p l a c e d relatively
by
(ce) in
any
subclass
{x} ~ of S w i t h
(weakly)
compact
(L (G)) s is
(weakly)
relatively
c o m p a c t i n X.
2. A p r o b l e m We m a y
of o p t i m a l
control
and a s s o c i a t e d with
La~ran~ian
equation and
consider
the p r o b l e m
of m i n i m u m
differential
constraints I[x,u~ = /G fo (t' (Mx) (t),u(t))dt, (Mx) (t), u(t)) , x6S, (2) t~G,
t 61 G,
(Mx) (t)),
(., (Mx) (.) ,u(')) ~- L I (G) domain variable space in the RU space,u
m
G is a b o u n d e d the s t a t e
> I,
t =
(t1 , . . . , t U ) 6 ' G, of a s u b s e t
and w h e r e
x is t h o u g h t possibly we think
of as an e l e m e n t a Sobolev space
S of a t o p o l o g i c a l its w e a k p = topology
(X,~),
X on G w i t h , L: S (Lp(G))r, y(t) A =
. As b e f o r e not
of M;
S (Lp(G))s
operators,
linear,
and we w r i t e
is a s u b s e t For every u =
II~~+s w h o s e U(t,y)
projection
is G.
(t,y) 6
of the u - s p a c e denotes
IR m ~s a s s i g n e d , function If M descalar
O
above
on U ( t , y ( t ) ) , (t,y,u) with
or u ( t ) E (t,y)~A,
U(t,y(t)), u~iU(t,y),
set of all
(fl,...,fr) consider = =
are the
functions
define,~ on M.
Ezlz=f(t,y,u),
z = f(t,y,u),
where ]R
r
(t,y) ~ A g a n d
Q(t,y)
is the p r o j e c t i o n to d e f i n e = 0,
of Q(t,y) sets
. Actually,
it is c o n v e n i e n t Q(t,y)
these
= 0, ~(t,y)
the e m p t y
87
We
define
T(t,y,z),
-~ < T(t,y,z)
< +~
, associated
to p r o b l e m T(t,y,z)
[zI ( z , z ) 6 Q(t,y)]
(3)
[zIz>f
- -
(t,y,u)
0 '
z = f(t,y,u)
'
u~U(t,y)].
where, A,
for
(t,y) A and
z~Q(t,y)
we h a v e
- ~ < T < + ~,
and
for
z ~
] R r - Q ( t , y ) , we have
T = + ~,
and c e r t a i n l y
we h a v e
T = + ~
(t,y) -C ]lgu+S-A. In o t h e r w o r d s , T is an e x t e n d e d f u n c t i o n d e f i n e d U+s+r ]R .We shall now a s s o c i a t e to p r o b l e m (2) of o p t i m a l c o n t r o l , new problem Six] = of ~ i n i m u m I s T(t,(Mx) ( t ) , ( L x ) ( t ) ) d t , ~ L I (G) , x ~ S,
(4) T(-, (Mx) (.) , (Lx) (-)) with the i m p l i e d constraints ( L x ) ( t ) Q(t,(Mx) (t)), calculus t6~G (a.e.). of the type (I). We say
of the
of v a r i a t i o n s (2).
deparametrized
problem
assumptions, in m o r e
problems
(2) and
as we on the
details. general
First,
some b a s i c on f ,f o
assumption
p.
385)
is a c o m p a c t sets A K = are
subset
K of G such ] , ~
that
(a)
(G-K)
< e;
[(t,y)~- A I t ~ K on
= ~(t,y,u) 6 M
are closed;
(c) fo,f
continuous
Under
this
asstunption
the N e m i t s k i i
operators
appearing
in
functions. we h a v e some
for a l m o s t
T(t,y,z)
also $tate
general
properties
the s e t s ~ or p r o p e r t y
~, y) h a v e (K) w i t h
the K u r a t o w s k i respect
property
of u p p e r
semicontinui(to,Yo) we
to y, p r o v i d e d
for e v e r y
68
'(~>0 cl
~ y~N6(Yo)Q(to'Y)
(5)
(t,y) Q(t,y) closed. have property (graph) has this
are necessarily
it is well known
(K) with respect to y, if and only if the sets of points [(y,z,z) I (z,z)~Q(t,y), y~A(t)~ are closed. Also,
(K) with respect to y if and only if the extended (t,y,z) ~ I R u+s+r, is lower semicontinuous are closed, T is finite, in
(y,z) . If
this is the case, then all sets Q(t,y) by Min in the definition T(t,y,z), that is, Q(t,y) of T whenever
or Q(t,y)
= then,
=[(z,z) Iz O ~ T(t,y,z)3.
(C '~) holds,
and of McShane's
and Weinberg's of m i n i m u m
problems
the assumptions
that we
is an extended
O
function
satisfying
in Section
the calculus
of variations
(2.i). (A lower closure theorem) Let G be a bounded region in the t-space u I~ ,u > I, let A be a subset of the ty-space ]R u+s whose projection
on
the t-space is G, and for any (t,y) ~ A let Q(t,y) be a subset of the z o z-space IR r+1 with the property that (zO,z ) ~ ( t , y ) , z o' > z o im(z',z)~Q(t,y). Take Q(t,y)=~ 9,or (t,y)~R~+s-K,and convex and have p r o p e r t y ( K ) ~ i t h lk(t), 6 l(t), qk(t), (Lp(G))r, t~G, assume that or respect to y.
plies
~(t),
k=1,2 ..... be
~ L I (G) , such that Yk(t) 6 A ( t ) , k=1,2,..., (~k(t), ~k(t))e Q(t,Yk(t)), ~k(t)>Ik(t), t6G,
< + ~.
69
theorem t ~
for p r o b l e m
(2)). the
Under
condition are
sets
Q(t,y) that
convex for
property _> O,
(K) w i t h
to y. L e t u s constant
assume
(l~)
fo(t,y,u)
- c[f(t,y,u)[. x ~ X with
~ = {(x,u)} (weakly)
elements
I[x,u~
relative
in X. A s s u m e one has
that
closure
at least
the
closed
t h a t M has
in m e a s u r e
property, in
the w e a k
convergence minimum in ~.
property.
functional general
I[x,u~
an a b s o l u t e
A very as
Tonelli-type
theorem
for p r o b l e m s
(2) can
now be
stated
follows: theorem existence by (~) theorem there ++~ for p r o b l e m is a s c a l a r as ~ ~+ by ~, (c~) (2)). Same
is r e p l a c e d below,
with (c)
(~)/~
fo(t,y,u) class S =
is r~eplaced relatively
(weakly) in X.
compact
is also
(weakly)
compact
As m e n t i o n e d , p r o b l e m s
For
(2) and
and
the
above o~ proof
problems
optimal
control
to the
Section
I for e x t e n d e d (I.i) we
problems sketch
of the
of v a r i a t i o n s . a proof sets
of t h e o r e m
rem
shall
in S e c t i o n in
7 is a l s o for the of
of theo
(2.i).
The p r o p e r t y to y)
(K) we
(2.i)
~(t,y) semicontinui-
(with r e s p e c t ty in (y,z)
corresponds function F
assumption
lower
for the
(or T ( t , y , z ) ) .
3. Some We m a y
more
general
problems
of o p t i m a l
consider
the p r o b l e m
of m i n i m u m
with
70
(Jx) (T) = g(T, (Kx) (T),V(T)), (t, (Mx) (t)) & A, (T, (Kx) (T)) ~ B, u(t) v(T) 6
T ~
G is a d o m a i n the state
(tl,...,t o) ~
variable
x is t h o u g h t possibly
as an e l e m e n t space
of a s u b s e t its
space
(X,~),
a Sobolev
X on G w i t h
~. H e r e M: S ~ (L (G)) s, L: S + (Lp(G)) r, K: S + (Lp(F)) r' P S +(Lp(~)) , p ~ I, are g i v e n o p e r a t o r s , not n e c e s s a r i l y linear, y(t) = (Mx) (t) ,z(t) r = . Here u(t) = = (Lx) (t) , t ~ G, ~(T) (u I ..... um), control = (Kx) (T) , control ~ is
and we w r i t e ~(T) =
(Jx) (T),
T ~
t ~ G, is the
function
on G, v(T) on
function of the
on F, and
A is a s u b s e t
t-space T-space
and B is a s u b s e t think
r. if we derivatives
of S as a s u b s e t up O ~
X on G, i n v o l v i n g be
of o r d e r Dax,
to a m a x i m a l lel ~ N-I,
then M may
traces
derivatives derivatives
on F. T h e n
L and J m a y be traces on
operators
on G and t h e i r
, of the o r d e r s in S e c t i o n
N and N-I
2 we m a y (T,~)
introduce ~ B, and
R(T,~),
(6) r e d u c e s extended
to a p r o b l e m
on G and theo-
integrands.
We o m i t
the e x i s t e n c e
rems
similar
to t h o s e there
of S e c t i o n s
I and
be t h a t
is no d i f f e r e n t i a l
on G and
that Or
considered boundary
in a p r e v i o u s values
by F i c h e r a .
= O, o it m a y
no d i f f e r e n in S e c t i o n
equation
F, and go = O, w h i c h
considered
71
4. The This
equivalence
theorem
of a n a l y s i s
vergence
in L I. T h e
theorem
for L e b e s g u e with
measures
in IR
or for a b s t r a c t We state it h e r e
measure in its
spaces, simplest
finite,
o-finite, form,
or w i t h o u t
atoms.
in a b o u n d e d
region
namely
for L e b e s g u e
theorem). functions
t ~ G} be region
a family G in ~ u .
on the b o u n d e d
are e q u i v a l e n t : is s e q u e n t i a l l y is e q u i a b s o l u t e l y M and a real with weakly relatively in G. ~ = ~(~), ~ + ~, such that compact in LI (G)-
integrable valued
(c) T h e r e
is a c o n s t a n t
function
0 < ~ < + ~, b o u n d e d
below,
~(~)/~
+ ~ as
~(If(t)
l)dt ~ M
for
all f ~ {f}.
(d) T h e r e
is a real ~(~/
valued
function
~ =~(~), such
below with
~ + ~ as f
~ + + ~,
that
~ {f},
is e q u i a b s o l u t e l y to a s s u m e in ~ ,~
integrable nonnegative
restrictive and c o n v e x
continuous, called of
EO,+~).
Functions
or ~ as above
equivalence
(b)
by De La V a l l ~ e by C a n d e l o r o
Poussin.
implicaimplithen An
also
proved
The and
(c) +
proved
by T o n e l l i The
in p a r t i c u l a r (d) +
in the g e n e r a l and d i r e c t
case.
implication
proof
of the w h o l e ~3~.
for u =
recently I. We
by Cesari refer
and P u c c i
can be e x t e n d e d and
u ~
to the
last m e n t i o n e d
for the p r o o f
a number
of r e f e r e n c e s .
72
Cesari's
property
(Q)
as before
(t,y) ~
R u+s
Q ( t , y ) C R r+1, we say that these sets have the property micontinuity with respect to y, provided for every
(to,Y o) we have
Q(t o ,yo ) = ~ 6 > o el co O y ~ N&(y o) Q(to'Y) Equivalently, property. we may say that the set valued map (t,y) Q(t,y)
are necessarily
than property
(Q) in proving
(low-
thus in existence
theorems~
problems
of optimi-
zation
[2]) will be illustrated below. for instance by Castaing theorems, by Schuur in in Banach and
(Q) has been used by many authors, in theoretical the existence questions
in selection
of solutions
of differential by Olech,
equations Lasota,
and Baum in to
by Angell
the existence
of solutions functional
nonlinear Volterra equations with delay and in nonlinear equations; and by Cesari and Hou in proving the existence
of solution
Recently,
property
Suryanarayana
(Q) is a g e n e r a l i z ~ Indeed Z,
tion of the concept of maximal m o n o t o n i c i t y any maxinml monotone map z Q(z) Q(x) C Z) has necessarily property
in a real Hilbert
has proved hhat maximality with respect to any analytical a large class necessarily implies property (Q).
73
6. G r o w t h p r o p e r t y (6.i) T h e o r e m
(~)implies [I], p.
property 333).
(Q).
([2] and
and for e v e r y x ~ A let Q(x) 6 ~ 1+r such ~(~), }(~)/~ X in A, that (a) (y,z) ~
denote Q(x),
L e t A be any set of p o i n t s x ~ h , I a set of p o i n t s (y,z) = (y,z ..... z r) y ~ y', implies (y',z) 6 below, Q(x). such Let that of
function,
bounded
(x) be a n e i g h b o r h o o d implies
and a s s u m e
x e N6(x)
y ~ #(Izl). is convex,
has a n u m b e r
of v a r i a n t s one: x ~ IR
of w h i c h we
mention (6.ii)
the f o l l o w i n g
let M d e n o t e
let T
O
(x,y,z)
on M, and
denote
T O (x,y,z) N
(y,z) ~ Q(x)] C
(x) of x in A a s s u m e ~ (Izl), y ~ L, w h e r e
then sec-
p = 1).
7. S k e t c h
(a) C o n c e r n i n g pointed y in
respect
(1.ii)
to be lower
semicontinuous
(y,z) . Thus,
it is e n o u g h
In the n o t a t i o n s [ ( z ) I ~ > z0 x E I xk
of S e c t i o n
I we d e n o t e by Q(t,y)
the sets
Let i = I n f ( I [ x ] , kJ s u c h that
i as
, x k ~ S,and t h e n
74
~ Q(t,Yk(t)) ,
= Fo(t,Yk(t) ,~k(t)) ,
(8)
sequence, weakly
such that Xk x ~ X
property,
there in
LXk = ~k +
~ weakly
is a s u b s e q u e n c e , say still [k], such that r (LI (G)) and Mx k = Yk y in m e a s u r e in G. At L and M has the closed property, g r a p h property, hence
hence
in LI,
-- ~(t) >
= ~ G q k d t ~-II~II1
is b o u n d e d
below,
the e x i s t e n c e (I.i)
proof we need,
semicontheo-
equivalent
lower
(2.i). We s k e t c h
(implications
contains
a subsequence,
Ik is also w e a k l y
vergent These
toward
a scalar lk' ~
of theorems
If Jk = I[Xk] then
Jk i as k + ~ ~k + ~
@s e O as s ~. Here theorem,
implication
(a)
(d) , there
increasing,
+ + ~ as ~
the s e q u e n c e
is e q u i a b s o l u t e l y such
integrable,
h e n c e there
is an e l e m e n t Now,
75 r+2
converges rem
weakly
to p,l,~
in
(LI(G))
. By the B a n a c h - S a k s - M a z u r
theo-
(cf., e.g.
k = I,...,N, ...,N,
N = 1,2,... , w i t h if
and s u c h that,
(S) ~ ) PN
tEG,
N = 1,2 .....
(p,l,~) Thus,
there
is a s e q u e n c e
(S) (s) (s) (PN (t) 'IN~ (t) '~N~ (t)) Let
(p(t),l(t),~(t))
i ~, for t G (a.e.).
qN (t) = ~k c(s)Nk qs+k(t) , t 6 G, and note Pk(t) that ~k(t)~_ Ak(t), t~G, fG ~ k ( t ) d t = J k ' k = 1 , 2 .....
= ~ ( l ~ k ( t ) I) ,
and h e n c e
t & G,
i - 6
with (cf.
6s 0 as s + ~ [I], p. 301),
, and by F a t o u ' s
lemma
relations
(s)
(s)
(t) ~ l(t),
t ~
G (a.e.) t
(s)
dt ~ lim i n f + ~ q
S+~
' t & G,
= lim inf
~ (t) >l(t),
t
O
6 G, and e > O, t h e r e
is n o w an s
O
such that,
for s > s
-O
lYs(to)-YoI<
e where
Yo = Y(to)"
For
k = I ,2 . . . . . (10)
A,
76
Q'~(t
O
(~,z)6 Q ( t o , Y ) i C ~Rr+2
'
(t ,y) ~ A.
O
,y)
have
property
(K) w i t h
respect
to y since
they
are
the
of the sets Q(to,Y ) , which have this property, with the )-I, z ~ IRr~. The sets Q'~(t
O O
fixed set[(n,z)I~>l(t
--
the
sets
t o fixed,
with the sets Q(x) in the yz-space replaced by the sets Q' (t ,y) in o the ~z-space, with the sets Q(x) in the vyz-space replaced by the sets Q'm(to 'y) in the vhz-space, with To(X,y,z) replaced by To(Y,n,z) =
(Iz[) , continuous/ and since the sets Q' (to,Y) have property respect to y, and the sets Q'~(t
O
,y)
have
property
y, and this
G.
( P s + k ( to , n s + k ( t O) '~s+k ( to ) ~ _ Q ' ~ ( t o'Ys+k (to)), ) ) and hence (~kCNk Ps+k(to )' ~kCNk ~s+k
(s) (s) (to) ,z c(S)
S >
--
k Nk ~s+k(to )C co Q
S
O
,~
(to,Yo,~),
~ + ~
,y) for ly-yol < e. o the points in the first member of this relation (p (to) ,n (t o ) ,~ (t)) as an element of accu-
Since ~ (t) o
= lim inf n (s)(t ) as s ~, we derive that o (P(to)' ~(to )' ~(to )) ~ el co Q'~(to,Yo,e).
Since e > 0 is a r b i t r a r y ,
by property
(Q) we d e r i v e
that
77
By the d e f i n i t i o n of ~ we have then (q(to), ~(t o ) ) 6 and this holds for a.a. t (a.e.) , and i > /Gq(t)dt > IG F o ( t , y ( t ) , ~ ( t ) ) d t , y(t) = where, or - ~. (Mr) (t) , is m e a s u r a b l e w i t h Lebesgue integral finite
O
Q(to,Y(to))
(11)
t ~ G
~(t)
(Lx)
(t),
(12)
at this point, F
s e m i c o n t i n u i t y argument.
(c) The argument for the existence theorem continues now with the usual remark that q(t) L-integrable > F
-o
(t,y(t)
~
~(t))
> -~(t)
--
- cl~(t) I
'
hence F
o
is
S is closed, x ~ S,
hence I[xq Z i, and then I[x] = i. This completes the proof of the existence theorems (1.ii) and (2.ii). if it is k n o w n that the
there is no need to construct the a u x i l i a r y sets Q'4~(t,y). We refer to ([I], p. 345, Second proof) for the details. This is w h a t happens (1.iii) and in
(2.iii).
8. C r i t e r i a of the F-, G-, H - t y p e s . In a p p l y i n g e x i s t e n c e theorems of Section 2 to given problems the functions f o and f are given, the p r o p e r t y for which
are d i s c u s s e d in detail in
[I] (Chapter 13)for the one d i m e n s i o n a l case, but hold for m u l t i d i m e n sional problems as well. In each of the conditions, we deal w i t h a sec o r r e s p o n d i n g sequence Yk(t)= as k + ~, and the d i f f e r e n c e s
78
6k(t) 6ok(t)
requirements
D, a n a l y t i c a l
in c h a r a c t e r ,
This
suffices
(D) m e n t i o n e d consideration.
For t h e s e
p o s e d by R o t h e
[9], we r e f e r
to C e s a r i
and S u r y a n a r a y a n a in q u e s t i o n s
[4]. C o n d i t i o n of a p p r o x i m a -
(D) w a s p r o v e d by A n g e l l t i o n in t h e c a l c u l u s
[7] to be r e l e v a n t
of v a r i a t i o n s .
Lipschitz-type Y'YK ~
condition F
For
I < p < + ~,
(Lp(G)) s,
IIyk-Yllp +
I~k(t) I , 16ok(t) l<Fk(t)h(lYk(t]_ > O, 0 < ~ < + ~, is a g i v e n = O, and h(~) ~ cl~l Y for constants), and Fk(t) (p' = ~ if
(c,y,~ O g i v e n
functions
w i t h p' = p / ( p - y ) ,
_ < M, a g i v e n
constant
Lipschitz-type = c
condition
Fp : functional
viewpoint.
For
',h as in ( F ) above, and there is a f u n c t i o n F(t,u) > O such '~'~O "p p. P that / G I F ( t , u ( t ) ) dt ~ M for all a d m i s s i b l e c o n t r o l f u n c t i o n s u(t),
t 6 G, and [ f ( t , y l , u ( t ) - f ( t , Y 2 , u ( t ) I < F(t,u(t))h(lyl-Y21) ~ and the same h o l d s For v a r i a n t s 13). oI- ~o" conditions, conditions and condition F here we m e n t i o n ,we refer to []] (Chapter only the g e o m e t r i c
of t h e s e
For the r e m a i n i n g
viewpoint.
79
Growth-type *:~ m
condition
ll~.m, w i t h
. (i) T h e r e e x i s t s a c o n t i n u o u s f u n c t i o n ~(u) , Pq . I~(u) I ~, such that for I < p , q < + ~, y , y ~. (Lp ( G s)) (G))TM IlYll llYkH < L , ll~(u)ll < L (L , L q ' p' p -- o q -o ~ y(t) in m e a s u r e in G as k + ~, and (ii) there with c,c' 2 O, O that, < ~ ~ p, O fo~ all < ~ ~ q, (t,y,u), and a func(t,z,u) ~ M
~(Uk) e ( L Yk(t)
constants),
constants ~(t) 2 O,
c,c',a, B t 6G,
~ 6 LI (G),
such
we h a v e
+ c(lYk(t) IP-~
We r e f e r
to
[1~for
conditions
~q
, G , G p~ ~,~
conditions
in m e a s u r e there
(ii) ~ O,
H . (i) For I < q < ~' Y' Yk m e a s u r a b l e , Yk(t) q in G as k + ~ , u k ~ (L (G)) m, .IIUkllq < L, a c o n s t a n t , q are o t h e r c o n s t a n t s c', B , 0 < B < q, and a f u n c t i o n $ 6 LI(G), such that for all (t,y,u), (t,z,u) ~ M,
t ~ G,
we have
If(t,y,u)
- f(t,z,u) I
<_
~(t)
+ c'
lul q-8
and We
the refer
same to
holds [I]
for f
O
. H .
for c o n d i t i o n
(Oct.
lecture,
associates P. B r a n d i ,
at the U n i v e r s i t y C. Gori, P. P u c c i ,
in the A. Sal-
(M. Boni,
on the e x i s t e n c e
and p r o p e r t i e s parametric
inteover
calculus
of v a r i a t i o n s , Two detailed
surfaces. in p r o c e s s The
expositions
of this m a t e r i a l
by C. V i n -
ti are n o w
C. V I N T I ,
integrals
of W e i e r s t r a s s - B u r k i l l - C e s a r i .
"Parametric
and Related
80
Topics"
(L. Cesari,
R. Goor, E. Silverman,
C. Vinti).
Springer-
V e r l a g Lecture Notes in Math. (to appear). C. VINTI, N o n l i n e a r i n t e g r a t i o n and W e i e r s t r a s s manifold: integral over a Journal I, S e p t e m b e r
W i t h the agreement of the O r g a n i z i n g Committee, we refer to these publications for the m a t e r i a l above. Also, we refer to L. CESARI, E x i s t e n c e of solutions and existence of optimal soluC o n f e r e n c e in S. M a r g h e r i t a Ligu-
Ital., Section A;
for m a t e r i a l c o n c e r n i n g existence of solutions to b o u n d a r y value problems for n o n l i n e a r ordinary and partial d i f f e r e n t i a l equations.
References [I] L. CESARI, O p t i m i z a t i o n - Theory and Applications. Problems with O r d i n a r y D i f f e r e n t i a l Equations. S p r i n g e r Verlag 1983, xiv+542. [2] L. CESARI, E x i s t e n c e theorems for weak and usual optimal solutions in L a g r a n g e p r o b l e m s w i t h u n i l a t e r a l constraints. I and II. Trans. Amer. Math. Soc. 124, 1966, 369-412, 413-429. [3~ L. CESARI and P. PUCCI, An e l e m e n t a r y proof of an e q u i v a l e n c e t h e o r e m relevant in the theory of optimization. Journ. O p t i m i z a t i o n Theory Appl., to appear. [4~ L. CESARI and M.B. SURYANARAYANA, Closure theorems w i t h o u t semin o r m a l i t y conditions. Journ. O p t i m i z a t i o n Theory Appl. 15, 1975, 441-465.
81
L. CESARI and M.B. SURYANARAYANA, On recent e x i s t e n c e theorems in the theory of optimization. Journ. O p t i m i z a t i o n Theory Appl. 31, 1980, 197-415. L. CESARI and M.B. SURYANARAYANA, Upper s e m i c o n t i n u i t y p r o p e r t i e s of set valued functions. N o n l i n e a r A n a l y s i s 4, 1980, 639-656.
[Tj
T.S. ANGELL, A note on a p p r o x i m a t i o n of optimal solutions of free problems of the calculus of variations. Rend. Circ. Mat. Palermo (2) 28, 1979, 258-272. G.S. GOODMAN, The duality of convex functions and Cesari's property (Q). Journ. O p t i m i z a t i o n Theory Appl. 19, 1976, 17-23. E.H. ROTHE, An e x i s t e n c e t h e o r e m in the calculus of variations, Arch. Rat. Mech. Anal. 21, 1966, 151-162.