Upload
nguyendang
View
215
Download
0
Embed Size (px)
Citation preview
UNCLASSIFIED
AD NUMBER
AD424023
NEW LIMITATION CHANGE
TOApproved for public release, distributionunlimited
FROMDistribution authorized to U.S. Gov't.agencies and their contractors;Administrative/Operational Use; SEP 1963.Other requests shall be referred to ArmyBallistic Research Laboratory, AberdeenProving Ground, MD.
AUTHORITY
USAARADCOM ltr, 24 Feb 1981
THIS PAGE IS UNCLASSIFIED
THIS ' EORT IhA ENDELIMITED
* M#D *CL-EARED .F.OR PUBL3IC 'RELEASE
UNE LD IfET'JVE' 5 00.2 AM DNORE SUTRCtIOwSARE IPEtS): UPON1 . I SE N*.1 VI MC OSJE
D DISTRIB9UT-ION -STATERM T A.:-
* APPROVED FOR PuBLI C REL2ASE;
DISTRIBU10O. VAfLMITE~o..
UNCLASSIFIED
AD 424023
DEFENSE DOCUMENTATION CENTERFOR
SCIENTIFIC AND TECHNICAL INFORMATION
CAMERON SIATION. ALEXANDRIA., VIRGINIA
I
UNCLASSIFIED
1
NOTICE: When government or other drawings, speci-ficationf or other data are used for any purposeother than in connection with a definitely relatedgovernment procuremenL -- ration, the U. S.Government thereby incurs no responsibility, nor anyobligation whatsoever; and the fact that the Govern-ment may have formulated, furnished, or in any waysupplied the said drawings, specifications, or otherdata is not to be regarded by implication or other-wise ae in any manner licensing the holder or anyother person or corporation, or conveying any rightsor permission to manufacture, use or sell anypatented invention that may in any way be relatedthereto.
J i llJ'
MEMORANDUM REPORT NO. 1509SEPTEMBER 1963
A DYNAMIC PROGRAMMING SOLUTION OF A
MISSILE ALLOCATION PROBLEM
William Sacco
RDT & E Project No. I 01O5O1AOO3and 1W232O1A098
BALLISTIC RESEARCH LABORATORIES
ABERDEEN PROVING GROUND, MARYLAND
DDC AVAILABILITY NOTICE
Qualified requesters may obtain copies of this report from DDC.
Foreign announcement and dissemination of this report by DDC isnot authorized.
The findings in this report are not to be construedas an official Department of the Army position.
BALLI ST IC RE SEARC1 LAB ORAT0R IES
MEMORANDUM REPCRT NO. 1509
SEPTEMBER 1965,,
A DYNAMIC PROGRAMMING SOLUTION OF AMISSILE ALLOCATION PROBLEM
William Saf~co
Computing Laboratory
PJ)T & E Project No. 1MO105OIA005 and JMO25201AO98
A B E R D E E N P R O V I N G G :1 O UN D, M A R Y L A N
BALLISTIC RESEARCCH LABORATORIES
1"N0RANDUM REPORT NO. 1509
WSacco/rhgAberdeen Proving Ground, Mi.September 1965
A DYNAMIC PROGRAMMING SOLUTION OF A
MISSILE ALLOATLION PIXOBLEMA-
ABSTRACT
A generalization of a missile allocation problem proposed by
Piccariello is formulated using dynamic programming techniques. The
formulation leads tc an efficient algorithm for computing integer solutions
to the discrete problem.
5
INTRODUCTION
H. PiccarieiLo1 has considered the following interesting missile alloca-
tion problem. Consider a target complex consisting of N missile launch
sites and M control centers. It is assumed that if one (or more) control
center survives an attack, then all surviving missiles can be launched and
if all control centers are destroyed chen none of the surviving missiles can
be launched. Furthermore the assumption is made that each of the N + M missile
sites and control centers presents a separate target to an enemy missilu
attack. hiis is equivalent to the assumption that a:n enemy missile can destroy
no target other than the one agaiast which it is directed. The problem is to
find an allocation of T identical missiles that when attacking the target
complex will minimize tne expected nwrber of' missiles capable of being launched.
Piccariello gives a solution to the problem when it is considered in its
continuous form (allocations are not restricted to integer values). He also
investigates the discrete form of the problem and shows by example, tnat, in
general, the solution in the continuous case is not a solution for the discrete
case. He then derives a necessary condition for the existence of a solution
for the discrete case to differ from the continuous case.
In this paper E generalization of Piccarie.lo's problem is formulated
using dynamic programming techniques., The formulation leads to an efficient
algorithm for computing integer solutions to the discrete problem.
THE FORMUlATION OF THE PROBLEM
We shal) use the following notations and definitions:
P i probability of survival of missile site i; i = 1,2,..when attacked by a single incoming missile,
P. probability of survlval of control center j; j = 1,2,...MJ when attacked by a single incoming missile,
a.: number of missiles at missile site i,1
T: total number of attacking missiles;
E(X,Y): expected number of surviving missiles after an attack byT missiles, where Y = T - X missiles are allocated to thecontrol centers and X missiles are allocuted to +he
missile sites.
The following expression for E(X,Y) ca1 be given:-
N [ M _Y j)
E(XY) a i 7T (1- ) ()1i L j=l
where x is the number of missiles allocated to the ith launcher site, yj is
the number uf missiles allocated to the jth control center and
N
Sxi =X (a)
\~y..Y=T - X. (0)
j=1
The objectiv" is to find the minimum values of E(X,Y) subject to the
constraints a) and b) where each xi and each yj is an integer.
The above problem reduces to the Piccariello problem when al=a2 ... =N=3,PI=P2.... =P,, and P =P =....P
TYNAMIC PROGRAMMUNG FORMUIATION
An examination of equation (f.), subject to the constraints (a) and (b)
indicates that we are faced with an N t M dimensional problem. In order to
circumvent thin dimensional difficulty., we will refconulate the problem as a
multi-stage problem and apply the techniques of dynamic programming to obtain
a feasible computational schemu. lt us defin-,
9k(X) = the minimum expected number of surviving missiles
for an allocation of X attacking missiles to k missile
sites,
and f (Y) = the minimnun value of the probability of the survival of at
least one control center for an allocation of Y attacking missiiefs
to I control centers,
Wi,
Then we have
mn E(XY) = Min m (x)fM(y) (2)
subject to the constraint X + Y = T, where
N x.
M, M i P i X1 + x 2 (3)(x 1=1
and
M v.fM(Y) = Min 1 - TT ( i -') bubject to the condition
(yj j - -l
Yl + Y 2 +'".+YM = Y"
Using the Principle of Optimality 2 we can express the g1 recursively as
F= M a P xk +g, k_ = 2 ,3,...,N, where Xk is permittedgk() M"'k tkk +gl k (5)to vary over the set 0,I,2,...XI. For k - 1 we have
gl(x) = a1 1 (6)
To obtain fM(Y) we first observe that fm(Y) = 1 - hM(Y), where (Y)
Max M (1 - PJ).jT=l - j)
P.ploying the Optimality Principle once again, we obtain
h,(y) = M.• 1 - Pj) h•_ (Y-Y;) , £ 2 ,5,...,M where yE
O,1, 2 ,...,Y) and (7)
h1 (Y) (1 - IY) (8)
The value Y is also permitted to range over the set of integers[O.I,2,...,T].
COMPUTATIONAL PROCEDURE
The dynamic programming formulation imbeds the original problem within
a family of analogous problems in whLch the basic parameters N, N, and T
assume sets of values which permit u2 to obuain, in the course of the compu-
tation, the solution to a variety of sub-problems. Because of the structure
of the process we are able to use the information obtained from sub-problems
of the originial problem to obtain th! solution of the original problem. The
input information that is required is the knowledge of N, M, and T and the
values of the PiIs, P.'s, and a 's.
We begin the computatiun by ubtaining the sequences (gk(X') and [Xk(X)]
from f-qustions (5) and (6), and the sequences [h,(Y)j and (y,(Y)) from
equations (7) and (8). Given this information, we are then prepared to use
equati.ou (2) to compute Min E(X,!').
NUMERICAL EXAMPLE
Lot N=, M= 2, T 5, P 1 ý .50, P2 = .30, P5 = .60, a1 =10, a2 = 8 ,
a. = 12, P1 = .60, P2 = .50. From equations 5, 6, 7, and 8, we obtain the
relations:
gý(X) a •,11
g2(X) M M P + gl(X-x{) +
g:(X) Min {aPx5 + g,(X-x,)}
hl(Y) = (1 - Y),
and Y y2h2 (Y) = Ma (1 - •2) hl(yy 2 )J.
Y"8
Using the previous relations we obtain the values,
gl(O) = 10
g1(l) = 5
91 (2) 2.5
g1(3) = 1.25
91(4)= .625
g1 (5) = .3125
g2(0) = a2 + g1(0) =8 + lo 18
8'.3010 + gl U
g9 i) = Min - 12.4
( + gl(O) 2.4 + 10 = 12.4
8 + g1 (2) = 1.0.5
g2 (2) = Min 2 = g1(1) = 7.4 Y 7.4
L.72 + g1(0) = 10.721
(8 + g1 (5) = 9.25
9 )= M 2.4 + gl(2) = 4.9 L72 + gl(1) = .72
,216 + gl(O) = 10.216
"8 + gl(4) = 8.625
2.4 + g1 (3) = 5.65
g2 (4) : Min .72 + gl( 2 ) = 5.52 ) ).52
.216 + gl(1) 5.216
.o648 + (o) n 10.0618
9
91(g5)
2.4 + g1(4) = 5.025
.72 + g1 (3) = 1.97 .97
.216 + 4.(2) = 2.716
,.064,8 + g1.(1) = 5.064-8
S.01944 + g 1 (o) = 10.01944.
g3(0) = [12 + 20-] 30
9 () =Min 12 + g2 (1) --24.4• 24.
g3(l 7.2 + g2(0) =25.21I 2).= m } i
1A-7.= 19.4+
g3(2) = Min 7.2 + 12.4 9.6 19.4i
4.32 + 18 =22.32
12 + 4.9 - 16.9
7.2 + 7.h = 14+.6
(5) Min 7 14.634.32 + 12.4 = 16.72
2.592 + 18 20.592
12 + 5.52 15.52
g93(4) t$Min 4.E2 + 7.4 1172 = 11.72
2.592 1 12.4 14.992
1.5552 + 18 19.5552
10
"12 + 1.97 15.97
7.2 + 5.52 = 10.72
4.32 + 4.9 = 9.222.992 + 7.4 = 9.992
1.5552 ý 12.4 = 13.9552
-.93r5±1 + 18 = 18.9'j12
h1 (O) = 0
h1 (1) = .40
h1 (2) = .64
h1 (3) = .784
h1 (4) = .8804
1h(5) = .92224
h 2 (0) 0 =
h,(1) = Max 4 0 =
h2 (2) = Ma{(.5o)(.4o) = .20j = .200
(.5o)(.64) = .32
h2(3) = Max 3 .32
(.75)(.40) .30
0
0
(.50)(.784) = .392h 2(I) = Max (.75)(.64) = .48 = .48
1 (.875)(.4o) = .35
1,o
'-0(-50)(,8804) = .440o2
(15' (.75)(.7810 = .588h2 (87564 = .56 .588
(.9s5)(.14o) =.575
We are finally ready to make use of~ eq. (2)
Min E(Y,X) =Min g,,(X) f (ý
where
=1-hM(Y),
and
X + Y =T.
Fo-r tuI example N 5, Nv = 2 so that
imi E(Y/X) Min g3i(X) mm(Y) i.e.,tx,Y)
g3 (0/) f2,(5) =(30)(.412) 12.;56
9.(1) f2(4) = (24.4))(.42) = 12.688
g.(2) f(3) = (19.4)(.68) = 13.192
Mi n .(Y,X) =Mi n -,9x).22
li E(X Y ,X Mii 8(5) f2(2) = (14.6)(.830) = 11.68 ~ 2txY 11y72
g i f,)(O) (-.22)(l) = 9.22
The minimum vrt1iip~ is 9.22. The allocation of missiles which yieldsthe minimum value is = Y, Y = 0 What remains is the determination of the
optimal values of the x;1 I = 1,2,5:
S= M x(O) 2,
= x2 (X-x.) = ',C(3) 1
xI =xl(X-~xj -2','• - I(2) : 2
Thtref ort thu optLmJil allocatioui po] jy Is giwv n by
(x 1! x. yI' y•) = (2,1,2,0,0).
DISCUSSION
The author is grateful to Mr. fi. L. Morritt for many helpful discussions
during the formulation and solution of this problem. Mr. Me-rrtt hbs pointed
out to the author •evei-al problem areas of interest in which this procedure
would have direct application. For instance it would be possible by utilizing
these methods to decide upon an optimum ratio of vuutrol ctn(:!r L, missile
sites. It would Aso be possible to use this methodology to ,ietermine whether
it is desirablL to harden ICBM siteS, o- to utilize the additional money
planned to be spent on active or passive defense by building more imdefended
launching situs.
WILLIAM SACC"
p.,
BIBLIOGIjAPHY
1. Piccariello, H. A Missile Allocation Problem, Operations Research,
Vol. 10, No. 6. November - December 1962, p. 795.
2. Bellman, R. Dynamic Programming. PrintuLon University Press: 1957.
14
DISTR.?ZFLTION LIST
No. of No. of
copies Organization Copies Organization
20 Commander I Commanding OfficerDefense Documentation Ccnter Harry Diamond LaboratoriesAT-T: TIPCR ATTN: Technical Informati6fiCameron Station Office, Branch 012Alexandria, Virginia 22514 Washington, D. C. 20438
Director 1 PresidentAdvanced Research Projects U. S. Army Air Defense Board
Agency Fort Bliss, Texas
Department of DefenseWashington, D. C. 20301 1 Chief of Research and Deve16pmdnt
Army Research Office
1 Director Department of the ArmyTDA /Wea•on S•ystems Evaluation W.ashin.gto-+n, D. C. 20X1o
GroupRoom IE875, The Pentagcn "3 Chief, Bureau of Naval Weap6h§Washington, D. C. 20301. ATTN!. DIS-33
Department of the Navy
1 Commanding General Washington, D. C. 20360U. S. Army Materiel CommandATTN. AMCRD-RP-P 1 CommanderWashington, D. C. 20315 Naval Ordnance Laboratory
White Oak1 Commandi-g General Silver Spring 19, Maryland
U. S. Army Materiel CommandATTN: AMCflD-DS 1 CommanderWashington, D. C. 20515 U. S. Naval Ordnance Test Station
AT91: Technical Library
1 Redstone Scientific Information China lake, California 93557Center
ATTN: Chief, Documents Section 1 Chief of Naval OperationsU. S. Army Missile Command ATIT: Op03EGRedstone Arsenal, Alabama 35809 Department of the Navy
Washington, D. C. 20360Commana i ng mGe- ]
U. S. Army Missile Command 1 DirectorRedstone Arsenal, Alabama 35809 U. S. Naval Research Laboratory
ATTN: Technical Information DivisionCommanding Officer Washington, D. C. 20390Picatinny ArsenalATIN: Feltman Reseaiý,h and 1 APGC (PGAPI)
Engineering Laboratories Eglin Air Force BaseDover, New Jersey Florida 32511P
15
DISTRIBUTION LIST
zU. uf No. ofCopi~s Organization Copies Organization
Director, Project RAND 1 General Electric CompanyATTN: Librarian Missile & Space Vehicles Departm~fitDepartment of the Air Force 3198 Chestn'.t Street1700 Main Street Philadelphia, PennsylvaniaSanta Monica, California
1 Institute for Defense AnalysesSSD (SSRri1) Research & Fngineering SupportAF Unit Post Office DivisionLos Angeles 45, California ATN: Technical Information
OfficeBSD (BSR) 1825 Connecticut Avenue, N.W.Norton Air Force Base Washington, D. C.California
I The Martin CompanyDirector Orlando, FloridaNational Aeronautics and
Space Administration 1 United Aircraft CorporationAmes Research Center Missiles and Space DivisionMoffett Field, California 400 Main Street
East Hartford 8, ConnecticutCentral Intelligence Agency 1243 E sree, ~1 Applied Physics LaboratoryWashington 25, D.C. The Johns Hopkins University
86k1 Georgia Avenue
Aerojet General Corporation Siler Spring, Maryland11711 South Woodruff Avenue I Stanford Research InbtituteDowney, California ATIs: Dr. Irving Yabroff
AVCO Manufacturing Corporation Menlo Park, California
Research and Development Division 4 Australian Group201 Lowell Street c/o Military AttacheWilmington, Massachusetts Australian Embassy
2001 Connecticut Avenue, N. W.Battelle Memorial Institute Washington, D. C.ATTN: Battelle - Defender505 King Street 10 The Scientifiu information OfficerColumbus 1, Ohio Defence Research Staff
British EmbassyBoeing Airplane Company 5100 Massachusetts Avenue, N.W.Aerospace Division Washington 8' 'D ".Seattle 214, Washington 4 Defence Research Member
Canadian Joint Staff21450 MasacihusLL:s Avenue, N.W.
WashLngto. 8, D. C.
16
4ý4)
I') p*0 V, 0
() 4. A
0 .1% U) /200/a
/14 b . 04.ýC) 00.04) 4) ~44
a4) /0 00
0 0., a
4)
CO 0 0- .0
"440 *fo )
0 43.C I 0344)C)j ) 0 0l U) 4)
O44C 0 fC) W4)1:10 0 WW t
12 C) 0~4 1 40;
to 0)10) H 0ý It040
005
C)j 01 44 0) /1440 40 4.1,4
< 0 'A CA A )) 0 p Hx 0 )m ' 0) ) 05,24 00
0.30 4)34 0 4> 4
~ 4 92 - 4 04> )0 4,I 1 0 'a .bod" m to '
0 00 0 004) 01 c
Is C) 0
MH) w4 E H)u
M 0 ) 0- C) 0 .
'4H4 4)
0 0 u1
H 04) 4 4
04)410 4 ) 10 04f A ý. . .. 4)
It 0 0 00
P4 0. o)
H/4)
H,. .54." I
0 0.0.4/ 0 8044i (44 0 It~ 6l 0
0 Q,'404 .40 . 00 t.- 0 4 d) to
1'4)4 0 4)4
.40(4It H 100 H
0 0l 0 I 8It 41' P4 43 20 .
0400
0 o,
S*-t 0 . A
.08) 01 4). 0 -4
0 r.W~/ )0w~
"'' o .4 04
01.: It___ _ _ __ _ _U, r,'tX