Physical Topology Design of Optical Networks Aided by Many-Objective Optimization Algorithms

In this paper, we investigate the performance of two many-objective evolutionary algorithms to design optical networks. Many-objective algorithms are a particular class of multi-objective algorithms whose goal is to solve problems with four or more conflicting objectives. We compared the state of the art algorithm, called NSGA-III, with a recently proposed swarmbased approach, named MaOPSO. We consider four important objectives to design optical networks: network blocking probability, capital expenditures, energy consumption and robustness. According to our results, the new many-objective based on the particle swarm optimisation algorithm outperformed the NSGAIII for this challenging problem and this study suggests that MaOPSO can be advantageous to tackle real world problems.