An optimal policy based on the genetic algorithm for the dynamic threshold of the optical network

The complete partitioning policy (CP) for the wavelength resource in optical networks is now widely focused on. The dynamic threshold is one of the ways to make CP policy more efficient. Furthermore, an optimized threshold will be better for reducing the blocking probability and improving the utilization of the wavelength resource. Hence, the genetic algorithm is selected as the optimal policy on virtue of its excellent global search performance for getting optimized value of the dynamic threshold. Moreover, a maximal threshold as the high limit for the dynamic threshold is needed to be decided for making wavelengths shared between different wavelength classes, because the class with higher priority can share its wavelengths with the lower one after its own call setups are satisfied. Therefore, a neural network predictor that can predict the number of the next call setup is designed on the basis of the genetic algorithm to solve this problem. The values of the dynamic threshold and the maximal threshold are calculated, and the simulation results show that they take good effect in reducing the blocking probability and improving the utilization of the wavelength resource.