GAIA - Uma proposta de agente inteligente acoplado a algoritmos evolucionários para ajuste dinâmico de parâmetros através de conceitos de design inteligente e lógica fuzzy

Abstract

For many years Evolutionary Techniques have been successfully applied in several computational optimization problems. For the matter of obtaining “best results” and a wide exploration of the search surface, the choices for tuning those methods can be exponentially complex and require a large human intervention. Those traditional Darwinian models rely only on randomness and heuristic choices, which are kept fix during the entire algorithm execution without following the individuals’ variability and any eventual changes needed. For that matter, the present work introduces the combination of Intelligent Design Theory aspects with a hybrid approach of an evolutionary algorithm, through the implementation of a Fuzzy Intelligent Designer agent, which dynamically monitors and regulates six chosen algorithms’ parameters, adjusting their values for any given situation. In order to testify these proposals, a Brazilian’s capitals symmetric TSP problem genetic algorithm prototype was also built, allowing the execution of 580 tests, plus several simulations and comparisons between many proposed set of configurations, regular techniques and other recent works. The Intelligent Design intervention deliveries over-expectation results, opening a wide new space for research of the: “Darwin-less Evolutionary Algorithms” or “The Hybrid Self-Adaptable Evolutionary Algorithms”.