Estudio e implementación de optimización gravitatoria y desarrollo de distintas metaheurísticas generadas a partir de él

Abstract

In this thesis we propose the comprehensive study of the heuristic: "Space Gravitational Optimization", S.G.O., designed for global optimization of continuous functions. We study its foundations and parameters to determine their values universally. We subsequently fulfill our goal of achieving the optimum in 40 benchmark functions, common tests of effectiveness and efficiency in global optimization, with different topologies and sizes between two and thirty, several multimodal. For achieving, we join S.G.O. with algorithms of different nature: local search (Nelder-Mead and Gradient), concentration (Segmentation) and intensification (Worm Hole and Very Simple Optimization). With them metaheuristic of high effectiveness and efficiency are generated. Two of them, Segmentation and Worm Hole, are unpublished, designed and developed by us. Last but not least our work involves several tests of the different metaheuristic generated with a real instance: Cassini 2, representing the actual track made by the eponymous unmanned spacecraft on its journey to Saturn.