Plataforma Web Hybroo: ambiente experimental voltado à hibridização de algoritmos de otimização

• Ano de defesa: 2019
• Autor(a) principal: Ancioto Junior, Edgar Marcos
• Orientador(a): Dantas, Maria José Pereira
• Banca de defesa: Vieira, Flávio Henrique Teles , Souza, Marcos Antônio
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica de Goiás
• Programa de Pós-Graduação: Programa de Pós-Graduação STRICTO SENSU em Engenharia de Produção e Sistemas
• Departamento: Escola de Engenharia::Curso de Engenharia de Produção
• País: Brasil
• Palavras-chave em Português: Ambiente Web experimental; CVRP; TSP; Hibridização de metaheurísticas; Otimização não linear
• Palavras-chave em Inglês: Experimental Web environment; CVRP; TSP; Hybridization of metaheuristics; Nonlinear optimization
• Área do conhecimento CNPq: ENGENHARIA DE PRODUCAO::GERENCIA DE PRODUCAO
• Link de acesso: https://tede2.pucgoias.edu.br/handle/tede/4361

In optimization there are methods applicable to a variety of minimization problems. When defining the method to be applied to a real problem, it is based on the scientific production seeking similarity with the problem and the efficiency of its results. There is a gap in the challenge of extracting information from experiments in scientific works that make possible the reproduction of the study. This work aims to develop an online experimental environment focused on optimization, with the functionality of performing experimental tests with metaheuristic algorithms, with the possibility of hybridization, validating the concepts developed through the solution of benchmark problems and allowing the comparison with other methods developed in the literature. All codes are developed in the Python programming language and are open source. The environment will allow the configuration of the algorithm and will demonstrate the resolution of the problem, allowing the understanding of the application of methods, the evaluation of hybridizations and the comparison between all the methods. For metaheuristic methods, Genetic Algorithm (GA), Simulated Annealing (SA) and Ant Colony (ACO) were developed for routing problems. For non-linear optimization GA and SA methods were applied to the functions and benchmark instances to evaluate the efficiency of the methods. Hybridizations of the methods were evaluated using retransmitted collaborative work logic. The results demonstrated that the hybrid techniques have a superior capability to the conventional methods to solve the problems of instances for the Traveling Salesman Problem (TSP) and Vehicle Routing Problem (CVRP). The ACO + GA hybridization was the combination which reached the best average efficiency values for the problems with values of 97.5% for the TSP and 96.66% for the CVRP, highlighting the executions involving the biggest problems among the selected ones. An environment was made available on the web, http://hybroo.hopto.org:5000/ to access the methods and the execution of the tests, allowing the data conference of this research