Application of differential evolution algorithm to optimization problems in optical networks

Abstract

It is well-known that telecommunications are developing almost exponentially worldwide in response to the ever-increasing bandwidth demand and transmission distances required in communication networks. Wavelength division multiplexing (WDM) optical networks have led to substantial research, which has eventually emphasized the modifications required in the optical network architectures to achieve their full potential. Optical networks are a sheld quite rich of optimization problems ranging from simple to multiobjective combinatorial ones. In WDM networks, the routing and wavelength assignment (RWA) and the survivable virtual topology mapping (SVTM) issues are of paramount importance in network optimization. With the evolution of optical WDM networks to a more exible architecture such as OFDM optical networks, new problems such as routing and spectrum allocation (RSA) arises. RWA, SVTM and RSA problems in an arbitrary mesh network are known to be NP-complete. Computational intelligence emerges as a crucial tool to deal with those complex optimization problems. In computational intelligence, nature-inspired algorithms encompass a set of heuristics that base their operation on the imitation of nature's behavior. It has been proved that those algorithms can be applied to a wide range of optimization problems in diverse areas of the engineering field obtaining near-optimal solutions in an acceptable amount of time. In this doctoral dissertation we present the application of di erential evolution (DE) algorithm to the RWA, SVTM and RSA problems in optical networks. We also propose the analysis of the control parameters of the DE algorithm on the system performance's improvement. Additionally, we propose strategies to improve the e ciency of the algorithm. We present experiments that demonstrate the e ectiveness and e ciency of the algorithm.