Scalable network planning for elastic optical orthogonal frequency division multiplexing (OFDM) networks

We propose to achieve scalable O-OFDM network planning with a genetic algorithm (GA) that operates in an adaptive way for high optimization efficiency. When the network topology and lightpath requests are given, the GA encodes the routing, spectrum and modulation assignments (RSMA) as genes and optimizes them iteratively. Both the crossover and mutation schemes in our proposed GA operate adaptively based on the fitness of individuals. Specifically, when the individuals are not fit yet, their genes can be modified significantly with crossover and mutation. On the other hand, for individuals that are already fit, we limit the crossover and mutation rate to avoid chromosomal disruption. The simulation results with the NSFNET topology and up to 1000 lightpath requests show that the proposed adaptive GA converges faster with better optimization performance, when comparing to a non-adaptive one. For the 1000-request case, the proposed algorithm can converge with a relatively small population size (e.g. 50) within 80 generations.