Thèse de doctorat en Informatique
Sous la direction de Alain Billionnet.
Soutenue en 2008
à CNAM .
Pas de résumé disponible.
Equipments replacement robust optimization in a decreaseing network
France Telecom is planning a change of facilities in its fixed phone network in a five years period. Because of competition and new offers, clients linked to the network sites are decreasing during the period. This enables to use again an equipment if it had been set up for a client who left afterwards. So the optimization problem consists in deciding for each year, which sites are migrated, how many equipments are bought or reused, in order to minimize the total cost. We present first a formulation by integer linear programming and we build an algorithm to solve optimally a polynomial subproblem. The exact resolution with a solver is efficient even with big instances and we propose various heuristics of relative good quality who provide a quicker answer. We then focused on a robust approach, in order to protect against uncertainty in datas coming from forecast studies. We studied a first method based on a representation by a finite number of scenarios and we obtained good results up to 5 scenarios. We also tried another method who relays on the definition of an uncertainly set around every uncertain data, but we had difficulties in appliying it to our specific problem