A modified Gravitational Search Algorithm and its application

Gravitational Search Algorithm (GSA) is a population-based optimization algorithm based on Newton's law of gravitation and the rules of mass interactions. Despite good exploration, GSA suffers from improper exploitation ability. This is mainly due to relatively big movements of agents, even the qualified ones, in the entire search process. In this paper, in order to improve the balance between exploration and exploitation of GSA, the quality of a current solution is considered in the processes of computing its velocity and acceleration. The experiments are conducted on standard unimodal and multimodal benchmark functions and their shifted and rotated versions. The obtained results are compared with those of five well-known algorithms in this field. In addition, the proposed algorithm is applied to clustering of wireless sensor network to find near-optimum cluster heads in a way that energy consumption would be the minimum. The obtained results show the high performance of the proposed algorithm in both fields.