A Robust Optimization approach to DVB-T Network Design

We address the question of defining a robust optimization approach to model and solve a DVB-T network design problem, while taking into account the uncertainty that naturally affects the propagation of wireless signals. The robust counterpart of the Mixed Integer Linear Programming model that represents the design problem exploits multiband uncertainty, a cardinality-constrained uncertainty model that employs multiple deviation bands. Since the robust counterpart may prove challenging to solve also for state-of-the-art optimization solvers, we propose a matheuristic for its solution. The matheuristic combines a variable fixing procedure exploiting suitable (tight) linear relaxations of the model with exact large neighborhood search. Results of computational tests considering realistic instances are reported to assess the performance of the approach, showing that the matheuristic can generate solutions of higher value than a commercial optimization solver within the available time budget.