FB-GSA: A fuzzy bi-level programming based gravitational search algorithm for unconstrained optimization

Abstract

The Gravitational Search Algorithm (GSA) which is a prominent nature-inspired computing technique outperforms in the exploration stage, but its performance degrades in the exploitation stage. A fuzzy bi-level programming based gravitational search algorithm (FB-GSA) is proposed in this study. The basic concept to create FB-GSA is the iterative fuzzy decision-making operation. FB-GSA accompanies the algorithms such as Chaotic Gravitational Search Algorithm (CGSA), and the proposed local search using spectral Polak-Ribire-Polyak-3 (spectral PRP-3) method. Initially, the adaptive parameters, for the fuzzy decision-making process, are determined. Then, the controlled operation of constituent algorithms is executed using fuzzy Bi-level logic, which leads to an optimal solution. Experimental evaluation of FB-GSA is performed using several unimodal and multi-modal benchmark functions. Experimental results illustrate that FB-GSA outperforms other state-of-art works for most benchmarks. The simulation results for FB-GSA also presents a significant improvement in the convergence speed. The fuzzy-based adaptive control employed in FB-GSA makes it devoid of premature convergence.