Pheromone particle swarm optimization of stochastic systems
Pheromone particle swarm optimization (PSO) of stochastic systems tests the impact of adjustments to algorithm parameters on algorithm performance when searching for optimal solutions to stochastic simulations. To test the benefit of adjusting PSO, the tuned algorithm is compared to the results from the commercial optimization software, OptQuest. In addition, two modifications to pheromone PSO are proposed. These include utilizing orthogonal arrays as an initial position for the algorithm and biasing the release of pheromones in the first iteration based on the relative strength of the objective function. These modifications are shown to improve the average objective functions found as well as the time to convergence in the optimization of some problem types. This paper also highlights the applicability of using pheromone PSO to optimize stochastic simulations compared to commercial optimization software.