Enhanced Dung Beetle Optimization Algorithm with Chaotic Initialization and Adaptive Perturbation Strategies

Jinkai Mao, Mingyang Yu, Jing Xu

Keywords: Metaheuristic algorithm, Chaotic Initialization, Adaptive t-Distribution, Enhanced Dung Beetle Optimizer

Abstract: This paper proposes an Enhanced Dung Beetle Optimization (EDBO) algorithm to address the limitations of the original DBO, such as premature convergence, weak global exploration, and low convergence precision. The EDBO integrates three improvement mechanisms: a Cubic Chaotic Map for population initialization, an improved global search strategy, and an adaptive t-distribution perturbation operator. These mechanisms jointly enhance population diversity, strengthen global search capability, and accelerate convergence in later iterations. Extensive experiments conducted on the CEC2017 benchmark suite demonstrate that EDBO achieves superior optimization performance compared to several well-known algorithms, including GWO, WOA, DMO, SO, and the original DBO. Statistical analyses confirm its robustness, stability, and adaptability across diverse optimization scenarios. Furthermore, ablation experiments reveal that each enhancement contributes significantly to performance improvement, while their synergistic integration yields the best overall results. The proposed EDBO algorithm thus provides a reliable and efficient optimization framework, offering promising potential for real-world applications such as control parameter tuning, intelligent robotics, and industrial process optimization.

Submission Number: 36