Regenerative population strategy-I: A dynamic methodology to mitigate structural bias in metaheuristic algorithms

Structural bias in metaheuristic algorithms is a critical issue, characterized by an inherent tendency to excessively exploit certain regions of the search space, even when unsupported by the objective function. This bias can distort the exploration process, negatively impacting the efficiency and effectiveness of the algorithms. Although many studies focus on understanding and identifying structural bias, effective mitigation strategies remain scarce. To address this gap, this study introduces the Regenerative Population Strategy-I (RPS-I), a methodology designed to counteract structural bias by dynamically redistributing the population. RPS-I integrates seamlessly into existing metaheuristic frameworks without altering their core mechanisms, providing a practical solution to reduce structural bias. The effectiveness of RPS-I is demonstrated by applying it to six metaheuristic algorithms: Genetic Algorithm, Differential Evolution, Particle Swarm Optimization, Grey Wolf Optimizer, Whale Optimization Algorithm, and Harris Hawks Optimization. The Generalized Signature Test is used to quantify the structural bias of these algorithms after incorporating RPS-I. The results indicate that the RPS-I-enhanced versions exhibit a significant reduction in bias compared to their original counterparts. Furthermore, these enhanced versions are validated using the IEEE CEC 2022 benchmark functions and two classical engineering problems, where they demonstrate improved search capabilities. This study highlights the importance of mitigating structural bias in metaheuristic algorithms, preserving the strengths of existing methods while extending their applicability and robustness. The source code for RPS-I is publicly available at https://github.com/kanchan999/RPS-I_Code.git.