open access

https://hdl.handle.net/2115/97571

Competitive differential evolution with knowledge inheritance for single-objective human-powered aircraft design

英語

Human-powered aircraft (HPA) design

2025-04-10

Springer

Journal of Supercomputing

81

721

This paper introduces a novel and efficient optimizer derived from differential evolution (DE): Competitive differential evolution with knowledge inheritance (CDEKI). CDEKI is developed by introducing the competitive mechanism and incorporates a novel DE/winner-to-best/1 mutation strategy and a hybrid local search operation. Moreover, we emphasize the significance of a commonly neglected component in DE: the repair operation. We propose a novel repair operation with knowledge inheritance to accelerate optimization convergence, particularly when the generated offspring exceeds the search domain. Through comprehensive numerical experiments conducted on the CEC2020 benchmark functions, competing against sixteen state-of-the-art optimizers and advanced DE variants, our proposed CDEKI demonstrates significant superiority. Additionally, the ablation experiments are conducted to independently investigate the performance of the proposed strategies. Furthermore, we extend the application of CDEKI to real-world human-powered aircraft design tasks, showcasing its extraordinary performance in practical scenarios. As an effective and efficient optimizer, CDEKI presents a compelling alternative evolutionary approach for addressing real-world applications across diverse domains. The source code of CDEKI is available in https://github.com/RuiZhong961230/CDEKI.

学術雑誌論文

Accepted Manuscript

This version of the article has been accepted for publication, after peer review (when applicable) and is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: http://dx.doi.org/10.1007/s11227-024-06859-3

0920-8542