A Novel Genetic Algorithm-Based Methodology for Large-Scale Fixed Charge Plus Routing Network Design Problem With Efficient Operators

Abstract

In this paper, we present a novel approach that addresses the problem of large-scale network topology design and routing. There are research works that used exact methodologies based on Integer Linear Programming (ILP) models to develop potential solutions for this problem. However, this problem is computationally NP-hard, thus solving it is hugely demanding on computational power for large-scale networks, and in many cases, it is not even possible to generate a solution with a reasonable optimality gap. This paper presents a hybrid algorithm based on the Genetic Algorithm with efficiently designed genetic operators. This algorithm aims to design the topology of large-scale networks and generate a routing configuration for a set of predefined traffic demands on the networks while keeping the total cost of design and routing at a minimum. The results have been compared to an exact ILP model, a relaxed ILP model, and a customized GA as benchmarks for validation purposes. These comparisons showed that the proposed algorithm significantly outperforms the ILP solutions in all of the large-scale network configurations that were used as case studies.INDEX TERMS Genetic algorithm, large-scale network, optimal topology, routing.

Algoritmo evolutivo generador de rutas eficientes para el transporte público

“…Información Tecnológica -Vol. 34 Nº 1 -2023 utilizando datos precisos de la red de carreteras locales y datos representativos de la demanda de transporte público pueden generar escenarios plausibles que son entre un 10% y un 20% más eficientes que las redes existentes, utilizando los mismos parámetros; así mismo Doostie et al (2021) presentan un algoritmo híbrido basado en el Algoritmo Genético con operadores genéticos diseñados eficientemente. Este algoritmo tiene como objetivo diseñar la topología de redes a gran escala y generar una configuración de enrutamiento para un conjunto de demandas de tráfico predefinidas en las redes, manteniendo al mínimo el costo total de diseño y enrutamiento; sin embargo, estos autores no presentan resultados en la red Swiss Road de Mandl.…”