A quasi-reflected and Gaussian mutated arithmetic optimisation algorithm for global optimisation

Abstract

Arithmetic optimisation algorithms (AOAs) rely heavily on a structured framework that is inadequate for all the complexity levels; therefore, they cannot explore the full search space effectively. In this study, these shortcomings of AOAs were addressed by incorporating quasi-reflection-based learning and a Gaussian mutation strategy into the basic AOA. Gaussian mutation enhances the searchability of the basic AOA, whereas quasi-reflection-based learning enables the AOA to jump from the local optima to the global optimum in each iteration, and it is adaptively updated based on the knowledge gained from the offspring. The efficacy of the proposed method was tested using basic benchmark functions, CEC2018 test functions, and real-world applications. The effect of quasi-reflection-based learning on the AOA and the combined effect of quasi-reflection-based learning and Gaussian mutation on the AOA were investigated and compared with other cutting-edge optimisation algorithms. The results of the comparison revealed that the proposed optimisation algorithm outperformed the other algorithms.