JOURNAL OF MATHEMATICAL ANALYSIS AND APPLICATIONS 114, 176-177 (1986)
Solution of Lanchester Equation Models
for Combat
G. ADOMIAN
Center for Applied Mathematics, University of Georgia,
Athen.7, Georgia 30602
Submitted by E. Stanley Lee
This paper provides a methodology for the study of quantitative behavior of
solutions to the extended system of Lanchester-type differential equations for the
prediction of the outcome of battles under different condition. The method is suf-
liciently general for use with nonlinear and/or stochastic adaptations as well.
‘$3 1986 Academic Press. Inc.
In military operations research activities, Lanchester-type combat
models are widely used (discussed most recently in [l] by Taylor who
gives a number of referencesto the literature).
These equations are given as
dx/dt = -a(t) y - fl( t) x,
(1)
dy/dt = -b(t) x-~(t) y,
with x(0) and y(O) as given constants, t = 0 denoting the time at which the
battle begins, and fFthe time at which it ends, x(t) and v(t) representing the
force levels at time t, and the coefficients a(t), h(t), P(t), x(t) denoting
attrition rates which indicate effectiveness of enemy fire. Clearly x(t) > 0,
y(t) > 0 for 0 < t < t,. The coefficients a(t), h(t) represent aimed fire between
the x and y infantry units or forces while cc(t), b(t) represent supporting
artillery tire not subject to attrition.
Let us write L = d/dt and write (1) in the form
Lx = -a(t) y - B(t) x,
(2)
Ly= -b(t)x-cc(t) y.
Operating from the left by Lp ’ = j:, [.I dt, we have
x=x(O)-Lp’a(t)y-L-‘p(t)x,
(3)
y=y(0)-Lm’b(t)x-L-‘cr(t)y.
176
0022-247X/86 $3.00
Copyright c 1986 by Academic Press, 1~
All rights of reproduction in any form reserved
LANCHESTER EQUATION MODELS FOR COMBAT 177
Since these equations are linear, there is no need for the (Adornian) A,
polynomials; any nonlinear adaptation of the model could proceed as we
do here by substitution of these polynomials for the particular nonlinear
terms. Now write x = C,“=Ox,, and y = C,“=O y, with x,, = x(0) and
y, = y(O) and using the decomposition method of Adomian [2,3], write
x, = -L-b(t) y,- L-‘/?(t) x0,
y, = -L-lb(t) x0-L-‘a(t)y,,
x2= -L-b(t) y, -L-‘/?(t)x,,
y,= --L-‘b(t)x,-L~‘a(t)y,, (4)
x,= -L-‘a(t)y,_,-L-lP(t)x,_,,
y,= -L-‘b(t)x,~,-L-‘cr(t)y,_,,
which determines components of x and y in successive pairs in terms of
preceding components. The n-term approximation for x is now given as
n- 1
XII= c xn
i=O
and for y by
n-l
yn= c Yn. (5)
i=o
It is observed that not only can nonlinear terms be considered if we wish to
modify the model but one can consider stochastic attrition specified by giv-
ing appropriate expectations, covariances, etc., for stochastic processes,in
which case, X,, Y, are random series whose ensemble averages,etc., can be
taken to provide statistics of the solutions. The processesshould of course
not allow physically unrealistic assumptions such as delta-function
correlations, Wiener processes, martingales, etc., but use measurable
statistics. Now specification of the conditions for battle termination such as
fixed time, complete annihilation of one force, attainment of a maximum
allowable level of causalities on one side, etc., allow complete calculation.
REFERENCES
1. J. G. TAYLOR, Battle outcome prediction for an extended system of Lanchester-type dif-
ferential equations, J. Math. Anal. Appl. 103 (1984), 371-379.
2. G. ADOMIAN, “Stochastic Systems,” Academic Press, New York, 1983.
3. G. ALXJMIAN, “Nonlinear Stochastic Operator Equations,” Academic Press, New York,
1986.