Reformulation‐Linearization Technique for MIPs

Abstract

The reformulation‐linearization technique (RLT) for mixed‐integer programs is an automatic model enhancement approach that generates a hierarchy of relaxations spanning the spectrum from the continuous linear programming (LP) relaxation to the convex hull of feasible solutions. This process is applicable to both 0‐1 as well as general discrete programs, and can be further augmented through the use of classes of semidefinite cuts. Often, the first‐level relaxation itself provides a sufficiently tight model reformulation that significantly improves problem solvability. The RLT procedure also offers a unifying framework for solving continuous nonconvex factorable optimization problems to global optimality.