Multiobjective programming in optimization of the interval objective function (Q2639774)

This paper considers the mathematical programming problem whose constraint set \(S\subset R^ n\) is known with certainty, but whose linear objective function Z has interval coefficients. Clearly, for any \(x\in S\), Z(x) is also a nonempty interval. Two natural notions of domination of intervals are proposed; \((\geq_{LR})\) if both endpoints of interval A are at least as large as those of B, then A is (weakly) preferred, and \((\geq_{CW})\) if the center of A is at least as great as that of B while the radius of A is no larger, A is weakly preferred. The disjunction of these orderings is a natural candidate for the ordering of intervals in the absence of further structure. The paper shows that (1) the two partial orderings are different; (2) the two orderings do not conflict (in the sense that there is no pair of intervals A, B such that \(A\neq B\), \(A\geq_{LR}B\), and \(B\geq_{CW}A)\); (3) the disjunction of the orderings (A is weakly preferred to B iff \(A\geq_{LR}B\) or \(A\geq_{CW}B)\) is equivalent to \((\geq_{LC})\) if the lower endpoint of A is no less than that of B, and the center of A is no less than B's, A is weakly preferred; (4) since Z is linear, we may take the solution set for \(\max (\geq_{LC})Z(x)\) s.t. \(x\in S\), as the Pareto optimal set of the multicriteria program \(\max \{z_ L(x),z_ C(x)\}\) s.t. \(x\in S\), where \(z_ L\) is the objective formed by taking the lower limit coefficient-by-coefficient, and \(z_ C\) is that formed by taking the center of each interval coefficient; and (5) interval objective programs satisfy the natural equivalence that the solution to \(\min (\geq_{LC})Z(x)\) s.t. \(x\in S\) is identical to that for \(\max (\geq_{LC})-Z(x)\) s.t. \(x\in S\).