A metaheuristic approach for optical network optimization problems

Two of the most complex optimization problems encountered in the design of third generation optical networks are the dynamic routing and wavelength assignment (DRWA) problem under the assumptions of ideal and non-ideal physical layers. Both these problems are NP-complete in nature. These are challenging due to the presence of multiple local optima in the search space. Even heuristics-based algorithms fail to solve these problems efficiently as the search space is non-convex. This paper reports the performance of a metaheuristic, that is, an evolutionary programming algorithm in solving different optical network optimization problems. The primary motivation behind adopting this approach is to reduce the algorithm execution time. It is demonstrated that the same basic approach can be used to solve different optimization problems by designing problem-specific fitness functions. Also, it is shown how the algorithm performance can be improved by integrating suitable soft constraints with the original constraints. Exhaustive simulation studies are carried out assuming the presence of different levels of linear impairments such as switch and demultiplexer crosstalk and non-linear impairments like four wave mixing to illustrate the superiority of the proposed algorithms.

. Comparison of fitness deviation of evolutionary programming algorithm, genetic algorithm, and simulated annealing for a 14 node network.Figure optionsDownload as PowerPoint slideHighlights
► A set of algorithms for solving difficult optimization problems is proposed.
► Only mutation is sufficient for generating good solutions.
► Algorithms shown to exhibit lower computational complexity for the DRWA problem.
► The algorithm involves only online computation.
► Novel wavelength assignment techniques are utilised to improve network performance.