A novel integer linear programming model for routing and spectrum assignment in optical networks

The routing and spectrum assignment problem is an NP-hard problem that receives increasing attention during the last years. Existing integer linear programming models for the problem are either very complex and suffer from tractability issues or are simplified and incomplete so that they can optimize only some objective functions. The majority of models uses edge-path formulations where variables are associated with all possible routing paths so that the number of variables grows exponentially with the size of the instance. An alternative is to use edge-node formulations that allow to devise compact models where the number of variables grows only polynomially with the size of the instance. However, all known edge-node formulations are incomplete as their feasible region is a superset of all feasible solutions of the problem and can, thus, handle only some objective functions.Our contribution is to provide the first complete edge-node formulation for the routing and spectrum assignment problem which leads to a tractable integer linear programming model. Indeed, computational results show that our complete model is competitive with incomplete models as we can solve instances of the RSA problem larger than instances known in the literature to optimality within reasonable time and w.r.t. several objective functions. We further devise some directions of future research.