A Novel Steady-state Genetic Algorithm Approach To The Reliability Optimization Design Problem Of Computer Networks

摘要

This paper introduces the development and implementation of a new methodology for optimizing reliability measures of a computer communication network within specified constraints. A genetic algorithm approach with specialized encoding, crossover, and mutation operators to design a layout topology optimizing source-terminal computer communication network reliability is presented. In this work, we apply crossover at the gene level in conjunction with the regular chromosome-level crossover operators that are usually applied on chromosomes or at boundaries of nodes. This approach provides us with a much better population mixture, and hence faster convergence and better reliability. Applying regular crossover and mutation operators on the population may generate infeasible chromosomes representing a network connection. This complicates fitness and cost calculations, since reliability and cost can only be calculated on links that actually exist. In this paper, a special crossover and mutation operator is applied in a way that will always ensure production of a feasible connected network topology. This results in a simplification of fitness calculations and produces a better population mixture that gives higher reliability rates at shorter convergence times.