Robust Models for Linear Programming with Uncertain Right Hand Side

We propose new robust models for handling right hand side uncertainty in linear problems. A cutting plane-like method could be devised to solve the resulting robust problems, but the subproblem to solve at each step involves a bilinear objective. Upper approximations are thus constructed based on linear decision and zero-order rules on the adjustable variables. Tractable reformulations are given on some uncertainty sets arising in practice. Heuristics are also proposed to compute lower bounds. To assess the methodology we consider its application to the capacity assignment problem. (C) 2016 Wiley Periodicals, Inc.