Enhancing the gravitational search algorithm through adaptive parameter control using linear mapping of fitness:

A Preliminary Investigation

Abstract

This paper presents a novel multimodal optimization algorithm by modifying our previously proposed gravitational particle swarm algorithm (GPSA). The conventional GPSA has demonstrated its ability to generate multiple design patterns for mechanical engineering problems, surpassing traditional methods. However, an important challenge in the conventional GPSA pertains to determining its control parameter. To address this challenge, the study proposes a modified version of GPSA called αGPSA, where the value of control parameter is adjusted based on a linear mapping of particle fitness values. Experimental evaluations were conducted using a simple multimodal optimization problem. The experimental results demonstrate that the proposed αGPSA achieves better performance in terms of peak ratio and success rate. Furthermore, the proposed αGPSA reduced sensitivity to the control parameter, streamlining the process of parameter selection for users compared to the conventional GPSA.