Particle swarm optimization with discrete crossover

Many adaptations to the original particle swarm optimization algorithms have been developed to improve performance with respect to the quality of the solutions found, convergence speed, and robustness. One class of such adaptations incorporates evolutionary operators within the particle swarm optimization algorithm cycle. To date, selection, mutation, and crossover operators have been incorporated within particle swarm optimizers with varying degrees of success. This article focuses on particle swarm optimizers that utilize discrete crossover operators, with the main objective to show if any performance gains can be achieved by incorporating discrete crossover. Six discrete crossover operators are proposed for incorporation into a global best particle swarm optimizer. The performance of these discrete crossover operators are compared with that of the global best particle swarm optimizer and amongst one another to identify the best performing discrete crossover operators. The best operators are then compared with particle swarm optimizers that make use of blending crossover operators. Empirical evidence obtained from an extensive benchmark suite shows that two of the proposed discrete crossover operators perform significantly better than the global best particle swarm optimizer and all of the other crossover operators.