Manyobjective Optimization to Design Physical Topology of Optical Networks with Undefined Node Locations

In this paper, we propose a new approach to handle the design of optical networks in scenarios related to the discovery of the most suitable node locations. In general, the node locations of backbone networks are previously defined, and they are posed as input for optimization algorithms. However, different node locations can offer better tradeoffs regarding capital expenditure and network performance. According to our knowledge, this is the first study to propose node locations as decision variables in network optimization together with the physical topology design process. We offer an approach based on clustering to select suitable group of node locations and a many objective memetic evolutionary algorithm that performs a local search to find better node locations during the optimization process. According to our results, our approach presented better than the previous proposals concerning hypervolume and can be pointed as a new way to modeling network planning tools, especially for long-haul optical networks.