Uma metaheurística baseada em interação social para otimização não-linear de domínios contínuos

Abstract

This work presents a metaheuristic model based on intelligent agents social interaction, using the social learning of a population of agents to achieve problems optimization. The method described here is based on humans social interaction, such as: comparison of social status, trend of society based on the most influential people, knowledge exchange, less influential people following the ones with most influential in their group and the search of a person where he achieves its best performance. An agent’s influence is measured through its social status, therefore, each agent have an influence radius proportional to its status. These concepts were modeled to create the technique, the learning among agents occurs when a less influential agent is located inside the influence region of an agent with more social status than him. Tests results, benchmark and real life problems, are presented and discussed. Those tests indicate that the model is a promising population metaheuristic.