A Heuristic to Find a P%-robust Network Design

Given a network with a set of nodes and a set of potential edges, we must decide what edges will be included in the design to satisfy demands between pairs of nodes. Each edge has associated a finite capacity, a fixed cost and routing costs. Uncertainty will be considered in important input parameters such as demands and routing costs and will be modeled through scenarios. The objective is to find a network design that is "good" across all scenarios potentially realizable. A heuristic method based on Benders Decomposition and Tabu Search is proposed that finds an initial solution which will be improved in an iterative fashion until lies within p% of the optimal or near-optimal solution for any realizable input data scenario. Key-Words: Robust optimization, Multicommodity network, Capacitated network design, Uncertain data, Heuristics.