Algoritmo de busca por aceleração: um novo método populacional de otimização global aplicado a problemas de otimização restrita e sistemas de equações

Abstract

In this work, a new metaheuristic is proposed to be used in solving restricted optimization problems and systems of equations that can be represented by an objective function of unimodal or multimodal nature. The Accelerated Search Algorithm (ABA) is a stochastic algorithm inspired by the Hooke & Jeeves method, which is a direct search optimization algorithm. ABA, which is a unique method, was implemented in C Programming Language and tested in optimization problems with constraints and systems of equations present in the literature. The results produced indicate that the proposed algorithm can, for example, solve 256-dimensional problems with an average of just 523 iterations, in addition to finding solutions for systems of non-linear equations with an average of just fifteen iterations. Given this, it can be concluded that ABA proved to be efficient, robust, precise and accurate, in addition to being easily implementable.