Professional Documents
Culture Documents
Fr'
by
Wayne F. Bialas
Mark H. Karwan
Wayne F. Bialas t
Mark H. Karwan
may impact the objective functions of several, if not all, levels. The
level-two planner controls the effective resource space for the level-
one planner. This produces a solution structure with the feasible region
•
Introduction
faithful analysis of the tradeoffs among competing goals, and assist a planner
energy policies set forth by the Federal government affect the objectives and
options, and hence the strategies, of state officials. This process continues
economic units such as firms and households. Each unit in the hierarchy wishes
exercised at other levels. In the example, actions at the state level affect
the benefits sought by the Federal government which can, in turn, constitutionally
planning process. A planner at one level of the hierarchy may have his
levels. However, his control instruments may allow him to influence the
policies at other levels and thereby improve his own objective function.
and Norton [2]. Unfortunately, their general problem was imprecisely defined
and they did not realize that the effective feasible regions for higher
1
levels were nonconvex sets. This causes the Candler and Norton algorithm
(i) Control of oil imports: With the desire to limit oil imports, the
Federal government can impose import quotas and duties. This action
will affect the consumption of oil by economic units which are maximizing
profits. The decisions made by these units would then impact on the
objectives of the Federal government including a change in consumption
patterns and levels of oil imports.
game with fixed leaders and a continuous decision space could be defined to
does not consider the activity space of one player to be a function of the
problems [4,5,6]. For example, Haimes, Foley and Yu [5] employ lagrangian
duality to decompose and efficiently solve a large model for the control of
example, in some cases, coalitions of levels can improve the objective functions
level could exercise complete control over all levels even though controlling
only a proper subset of the variables. This methodology can assess the value
The next section will provide the formal, general definition of a multilevel
programming problem. Then the discussion will focus on a special case of this
general problem, the two-level linear resource control problem, and its
set and some general algorithmi c approaches to solve the problem. Finally, a
nk • 4 x k k k k,
R = (x 1 ' 2 "-- ,xn k=1,...,r ,
1
k
r n
where nk = n. Denote the maximization of a function f(x) over R by
k=1 n n
k k+1 k+2 r nk +l k+2 x
varying only x E R k given fixed x, x in R ^ R
by max f(x)
k k+1 k+2
x ix ,x ,...x .
max f (x)
r
r
x
st: max f (x)
r-1
r-1 1 r
x Ix
st:xESCR
^ Sk-1 :
S = {t x
x E - fk-1 (x)= max fk-1 (x) }
k-1 ^k ^r
x ix ,...,x
k-1 r
Note that x is a function of x lc ,...,x . Furthermore, the problem at
max f (x)
k
k k+1
x ix ,...,x
( Pk )
x E Sk
k+1 r r
which is a function of x , and (P ): max f (x) defines the entire
x€S r r
problem.
4
The Two-Level Linear Resource Control Problem
2
max c x
x
1
st: max c x
2
(P ) 4 x
(P l 2
A x + A x <b
1 2 -
x > 0 .
2
Here, level 2 controls x which, in turn, varies the resource space of level
l 2
one by restricting A x < b - A x .
1 2
Ill-Defined Problems
k
Care must be taken when P results in alternative optimal solutions for
k+1r
fixed x . Although not affecting the value of the level-k objective
function, f k (x), these solutions can have drastically varying impact on the
objectives of other levels. Therefore, control over the choice among alternative
ill defined.
problem:
5
max + 1 x2
1 2 x
x
2 max x
2 + x
st: 3
x i , x3 2
1x
2 x + x = 4
2 3
( P )1 1 x > 1
(P ) 2
- x + 2x < 2
1 2
x + x < 4
1 2
which accomplishes this. Given this side payment scheme, x 2 = 1 is the optimal
Nonconvexity
r 2
S 1 = l x > 0 : A x l + A2 x < b ,
1
2 1 l
S = Ix f S : c x = max c 1 xi1
x11i x2
6
2
need not be. Therefore, P , which can be written as
2
max c x
x(S 2
max x + x
2 3
x
2
st: max x
3
xi ,x3 x 2
x + x - x > 2
1 2 3
x - x + x < 2
1 2 3
x + x + x < 6
1 2 3
x + x + x > 2
1 2 3
x < 1 ,
3 -
2
where S is the hatched region shown in Figure 2.
every level, the order and independence with which decisions are made prevent
such cooperation. This rules out any algorithmic approach which seeks only
leads to the best choice of x to be x = 0 with the optimal solution at point (c).
2 2
7
A Sufficient Condition for Complete Control
Consider the two-level linear resource control problem. Given any basis
B C A for the set of constraints Ax = b, one can write the equivalent set of
constraints on x:
Bx + Nx = b
B N
-1 -1
Or xB = B b - B N xN .
2
Further Characterization of S and P2
problem.
Theorem I
2 1
(i) S C S
2 1 2
Proof: (i) s S by the definition of S .
1
(ii) (By contradiction) Let y1 , y ..' y f S with
2" r
r
1 2 2
x = (x ,x ) = X A y , and
t t S
t=1
r
> 0 A l > 0, X A = 1.
at - t
t=1
8
k optimal
solution
Figure 1
2
Example of NOnconvexity of S
Figure 2
10
1 2 -1 2 _2
Suppose y
= (Y1' Y1 ) f S
Then there exists y1
1
such that = Y 1 Y1 ) (
1 I 1 1
with c 17 1 > c y l .
=AY + XAyc
1 1 t=2 t t
- ,
2
since S1 is convex. Noting that x = x2 and A. >
1
1 1 1 1 1 1
CX c y < cA Y + cAy
= c 1A 1y1 t t t 1 1 1 ttt
t=2 t=2
1 2 2
Given, x = (x ,x ) S and A l > 0, we have established an x with the
following properties:
sl
(a) x=
- 2 2
(b) x =
1 1 < c1-1
x .
(c) c x
2 2 1 1
This contradicts the definition of S since x t S should have maximized c x
2 2
for the fixed value of x . Therefore Al > 0 implies y S . Since the
1
choice of y l 'among the y's was arbitrary, we have proven A t > 0 implies yt c S
Any point which positively contributes in any convex combination
2
forming a point in S , must also be in S2. Since this is true of any point,
x y
t=1
From Theorem 1, this implies .,Yr t S and hence x cannot be an extreme
2
point of S a contradiction.
11
2
Recalling that P may be formulated as max c x and noting the
x(S 2
control problem (if one exists) occurs at an extreme point of the constraint
Algorithmic Approaches
2
(P2 ) max c x
st: x ( [S 2 ]
is an equivalent formulation for the two-level linear resource control problem.
This suggests a search for cutting plane procedures to approximate the convex
2
hull of S as a direction for future research..
12
Any desirable algorithm for the two level linear resource control problem
^1 ^2
Consider the solution, x = (x ,x ) to the following problem:
2
max c x
6; )
st: Ax=b, x>0
In (P), the level two planner is given full control over all variables. Now
2 ^2
fix x = x and solve the following problem with solution x to determine if
x f S2 : 1
max c x
1 2
st: = (b-A x )
A1 x
1
x > 0
2 is an optimal solution to the overall problem. For
If x = x, then x t S
changes in the second level objective, and, hence quite different choices of c 2
can produce an optimal solution after solving (P). For the example shown in
2
Figure 1, two particular choices of c which lead to such a condition for the
1 2
level-one objective shown are both c2= c and c = -c 1 . Thus both highly
may lead to solutions after solving the two linear programming problems (P) and
(P). Any reasonable algorithm should have the ability to easily solve any
13
2 2 2
xl RHS
.._
2 2
r 22 2
r r r
1 k
x Y1
B Y 11 Y lk
1
x -b2
1-;
B Y21 Y22 — - — '''''' " Y2k
. 2 .. .
_
xB Yml Y m2 b
m Ymk m
2
The variables, x 2 x
k represent the nonbasic level-two variables which are
at nonzero values, and r i2 ,...,rk2 represent the reduced costs of these variables
basisBCA,b=B_ b- y yx 2 where (y ,Y ,t
= Y, = B
-1
(a ,a
j lj 2j mj l j 2j
j= 1
and xB
x denotes the ith basic variable.
2
Step 2. Set x2= x and solve the following problem via bounded simplex
14
Step 3a. If all nonbasic variables are equal to zero, go to step 4 with current
Step 3b. If b. > 0 for all i, go to step 3c. Otherwise, without loss of
Step 3c. Consider any nonbasic variable which is at a strictly positive value,
2 2 2 2
say x.. If r. < 0, increase xj until it enters the basis. If r, %P. 0,
3 -
decrease x 2 until either it reaches zero or it must enter the basis.
Go to step 3a.
Step 4. Beginning with tableau T solve the following problem via a modified
simplex procedure:
2
max c X
st: Ax=b
x>0
1
max c x
2
st: A i < (b-A x )
1x 2
1 2 2
x > 0, x = x
1
via dual simplex on repeated applications of step 4. If x = x
then enter the candidate into the basis. Repeat step 4 until no more
15
Validation and Convergence
The algorithm begins by finding the maximum of the second level objective
over the entire feasible region, S 1 . A check in step 2 is then made to determine
global optimal solution and has solved what was previously termed an easy
problem. If termination does not occur in step 2, the resulting solution from
brought into the basis at its current positive level and x B becomes nonbasic
at its current value of zero. Thus the number of basic variables at level zero
such a pivot is always possible, that is, yk 0 for some j=1,...,k. Suppose
that y 2, = 0 for all j=1,...,k. Then repeated applications of step 3c would
xB will remain zero no matter how x 2l ,...,x are varied. This contradicts the
16
2
If all b.> 0 but there are still nonbasic level-two variables, x2 . .. ,x k'
2
at nonzero values,. then step 3c is entered. Any variable x j , j=1,...,k, is
is assumed when step 4 is entered. Recall that 1;1 0 for all i as a result
of step 3b. Thus there exists two scalars, 0 1 > 0 and 0 2 > 0 such that any
increase or decrease in x, by an amount less than or equal to 01 and 02
By Theorem I, such points must also be in S 2 . Thus each point resulting from
The algorithm terminates with an extreme point solution in S 2 which has the
facts:
17
(i) The feasible region defined by S Ax = b, x > 0t is bounded
(ii) Steps 1, 2 and 4 are finite since the simplex, bounded simplex and
variables.
variables by one.
Conclusions
serve as useful tools in modelling structured economic units. Such models can
This paper has proposed a general mathematical structure for such problems,
For this problem, Theorem I illustrates a key property of the nonconvex feasible
techniques and obviates the need for methods to establish the convex hull of the
level two feasible region. Towards this goal, this paper has offered an adjacent
extreme point method which can find local, and sometimes global, optimal solutions
Acknowledgement
18
Figure Captions
2
FigUre 1. Example of Nonconvexity of S
2
Figure 2. Example of S in Three Dimensions
19
References
20