Otimização por nuvem de partículas aplicada ao problema de atribuição de tarefas dinâmico

• Ano de defesa: 2012
• Autor(a) principal: Pierobom, Jean Lima
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Programa de Pós-Graduação em Computação Aplicada
• Palavras-chave em Português: Inteligência coletiva; Otimização combinatória; Otimização matemática; Inteligência artificial; Algorítmos computacionais; Simulação (Computadores); Swarm intelligence; Combinatorial optimization; Mathematical optimization; Automatic theorem proving; Computer algorithms; Computer simulation; Artificial intelligence; Demonstração automática de teoremas
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/205

Swarm Intelligence searches for solutions to optimization problems using computational techniques inspired in the emerging social behavior found in biology. The metaheuristic Particle Swarm Optimization (PSO) is relatively new and can be considered a metaphor of bird flocks. PSO has shown good results in some recent works of discrete optimization, despite it has been originally designed for continuous optimization problems. This paper deals with the Task Assignment Problem (TAP), and presents an application: the optimization problem of allocation of taxis and customers, whose goal is to minimize the distance traveled by the fleet. The problem is solved in a static scenario with two versions of the discrete PSO: the first approach that is based on a binary codification and the second one which uses permutations to encode the solution. The obtained results show that the second approach is superior than the first one in terms of quality of the solutions and computational time, and it is capable of achieving the known optimal values in the tested instances of the problem. From this, the algorithm is adapted for the optimization of the problem in a dynamic environment, with the application of different strategies to respond to changes. The new results show that some combination of approaches enables the PSO algorithm to achieve good solutions along the occurrence of changes in decision variables problem, in all instances tested, with different sizes and scales of change.