Technical Note—Complementary Programming

Abstract

A mathematical programming problem P : minimize z = d T x + e T u + f T v subject to Ax + Bu + Cv ≧ g , x , u , v ≧ 0, u T v = 0, is called a complementary programming ( CP ) problem. The condition u T v = 0 distinguishes CP problems from ordinary LP problems. A variety of nonlinear programming problems can be formulated in this form, including absolute-value programming problems, 0-1 mixed-integer-programming problems, quadratic-programming problems, and so forth. This paper proposes a branch-and-bound algorithm for CP problems and reports some computational results.