Genetic Algorithm, Extremal Optimization, and Particle Swarm Optimization Applied to the Discrete Network Configuration Problem

Genetic Algorithm, Extremal Optimization, and Particle Swarm Optimization are applied to the discrete (integer-based) communications network configuration problem. In this preliminary study, each heuristic search scheme is used to determine the optimal (or near optimal) set of communications equipment components needed to satisfy user-specified radio subscriber and wire subscriber requirements. The network configuration problem is based on the U.S. Army’s Mobile Subscriber Equipment communications networking system. Although MSE is no longer a major asset in the U.S. Army’s communications inventory, it continues to provide an excellent platform for configuration optimization due to its non-permutation yet discrete nature. The comparison of the heuristics includes their overall reliability as well as the number of fitness evaluations needed to converge on the optimal solution. General results indicate that all approaches achieve relatively high reliability using some novel operators and adaptations, but the GA does so using considerably fewer fitness evaluations.