Combining genetic local search into multi-population evolutionary algorithms for the capacitated vehicle routing problem

• Autor(a) principal: Babak Rezaei
• Data de Publicação: 2023
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: http://hdl.handle.net/1843/60098
• topic: Vehicle routing problem; Evolutionary computation; Imperialist competitive algorithm; Hybrid genetic search; Island genetic algorithm; Multi-population genetic algorithm; Engenharia elétrica; Veículos; Algoritmos genéticos; Genética; Algoritmos; Inteligência artificial; Ciência da computação; Cálculos numéricos

The Vehicle Routing Problem (VRP) is one of the most significant problems in operational research today. VRP has a vast range of application fields such as transportation, logistics, manufacturing, relief systems and communication. To suit the needs of different real-world VRP scenarios, many models of VRP have been developed - CVRP (Capacitated VRP) being the classical form. In this study, at first a hybrid algorithm (ICAHGS) for solving CVRP is proposed, combining a refined ICA (Imperialist Competitive Algorithm) as the primary evolutionary and multi-population method, and a Hybrid Genetic Search (HGSCVRP) algorithm as an enhanced local search and population management strategy within the ICA framework. ICAHGS has been compared to several state-of-the-art algorithms from literature. The results of this comparison, which include both classical benchmark instances and real-world applications, demonstrate the competitive performance of the proposed algorithm. Afterwards, Dynamic Population Island GA and HGS (DPIGA-HGS) is introduced, which is a novel hybrid metaheuristic model. DPIGA-HGS integrates a specialized island model (DPIGA) and a refined HGS as its local search engine within each island. The primary objective of DPIGA-HGS is to contribute to the advancement of the field by proposing a new variant of Island GA and simultaneously achieving improved optimization results in comparison to ICAHGS. The results of the comparative analyses revealed the superior performance of DPIGA-HGS when pitted against other state-of-the art algorithms, including ICAHGS. Across multiple benchmark datasets, DPIGA-HGS showcased its prowess by achieving a significant number of BKS (Best Known Solution), outperforming its competitors in various instances.