Otimização multiobjetivo de redes de filas finitas de servidor único por meio de um algoritmo de enxame de partículas

Abstract

Long and time-consuming queues are a common experience and occur in various everyday situations. Mathematical queue models are often applied to understand and to improve these systems. This study focuses on optimizing finite queueing network systems, aiming to maximize throughput, minimize the overall capacity, and to address the trade-off between capacity allocation and throughput. A heuristic method based on the particle swarm algorithm is presented to identify a Pareto set of suboptimal solutions. The approach is illustrated with networks of M/G/1/k queues, in Kendall notation, utilizing a classical performance analysis tool, the generalized expansion method (GEM). Experiments demonstrate efficacy and computational efficiency in terms of high quality solutions and low CPU times. Future investigations could explore diverse network configurations and real-world applications. The study contributes to the optimization of queueing systems and their practical implementations.