An Improved Differential Evolution Algorithm Based on the Biological Characteristics of the Squirting Cucumber

Abstract

Differential evolution (DE) is widely used in numerical optimization, but fixed parameters and a single mutation strategy make it difficult to coordinate exploration and exploitation across different search stages. To address this issue, this paper proposes Bio-EbDE, a DE algorithm with stage-aware mutation strategy switching, subpopulation-specific parameter adaptation, and a local optimum escape mechanism. Its main distinction lies in three coordinated designs: mutation strategies are selected according to both evolutionary stage and subpopulation role; the scaling factor F is generated from different distributions for different subpopulations; and a spray-melon-inspired perturbation generates offspring with adaptive dispersal when premature convergence is detected. Experiments on CEC2013, CEC2022, and a constrained engineering problem show that Bio-EbDE achieves competitive optimization accuracy and convergence behavior.