We present mixed integer linear programming models that are robust in the face of uncertain traffic demands known to lie in a certain polyhedron for the problem of dense wavelength division multiplexing network routing and provisioning at minimal cost. We investigate the solution of the problem in a set of numerical experiments for two models of polyhedral uncertainty: (a) the hose model, (b) a restricted interval uncertainty model. We report the results of these numerical experiments in comparison with an alternative model of robustness due to Kennington et al.
Report no. 2005-05. Bilkent University, Department of Industrial Engineering, Bilkent 06800 Ankara, Turkey. April 2005.