A one-phase algorithm for semi-infinite linear programming (Q911454)

The primal problem of semi-infinite linear programming is defined as ``minimize \(c^ Tx\), subject to \(a(u)^ Tx-b(u)\geq 0\) for all \(u\in U''\), where c, \(x\in R^ n\), \(U\subseteq R^ m\) is an infinite parameter set, a(u): \(U\to R^ n\), and b(u): \(U\to R^ 1.\) An algorithm for solving this problem is given by using a succession of cuts which force the solution towards feasible x and u, by solving intermediate problems which find approximations by alternately solving a linear programming problem in x for fixed u and a nonlinear problem in u for fixed x. A duality theorem is given, and if a solution exists for the primal then the algorithm also solves the dual problem. Convergence is proved and bounds are given for an \(\epsilon\)-optimal solution. The algorithm is generalized to nonlinear semi-infinite programming, and is also applied to ordinary convex programming.