Improving exploration and exploitation via a Hyperbolic Gravitational Search Algorithm

Abstract

Finding the best compromise between exploration and exploitation phases of a search algorithm is a hard task. The Gravitational Search Algorithm (GSA) is an evolutionary algorithm based on the Newton’s universal law of gravitation. Many studies show that GSA suffers from slow exploitation which generates premature convergence. This paper proposes a Hyperbolic Gravitational Search Algorithm (HGSA) able to find an optimal balance between exploration and exploitation. The main contributions of this work are: the definition of suitable hyperbolic acceleration coefficients, the dynamic regulation of the gravitational constant coefficient through hyperbolic function and the definition of a decreasing hyperbolic function for the best agents which attract the other agents. The proposed algorithm is compared with well-known search algorithms on classical benchmark functions, CEC-06 2019 benchmark functions and three-bar truss design problem. The results show that HGSA is better than other algorithms in terms of optimization and convergence performances.