Scholarlink

Optimization problems are prevalent in real-world contexts and require effective strategies to navigate complex, high-dimensional landscapes characterized by numerous local optima. Existing algorithms frequently face challenges due to their limitations in addressing multiple constraints and complexities. This study introduces the enhanced adaptive crossover-based smell agent optimization (EACB-SAO) algorithm, which draws inspiration from the olfactory senses of living organisms. EACB-SAO offers two significant innovations: firstly, a novel hybrid operator known as Fusion Besiege Attack, which is influenced by the hard-besiege and soft-besiege techniques inherent to Harris Hawks Optimization, thus considerably improving exploitation capabilities. Secondly, EACB-SAO incorporates a probabilistic switching mechanism within its trailing mode, facilitating adaptive transitions between the traditional trailing operator and the newly developed Fusion Besiege Attack operator. This approach aims to improve adaptability across different optimization contexts. To evaluate its effectiveness in complex numerical benchmarks and engineering design problems, EACB-SAO is compared against nine competing algorithms in 23 classical benchmark functions, 29 CEC2017, 12 CEC2022 benchmark functions and 13 real-world applications. The results of a scoring system demonstrate that EACB-SAO achieved the highest score of 100 for the CEC2017, CEC2022, and real-world scenarios, thereby establishing its superiority over other algorithms and advancing the capabilities of previous adaptive SAO approaches. These findings underscore EACB-SAO’s proficiency in addressing practical optimization challenges, highlighting its efficiency and advantages in resolving complex real-world problems. MATLAB code for the algorithm is available in the Code Ocean repository https://doi.org/10.24433/CO.0873412.v1 .

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced Adaptive Smell Agent Optimization: Leveraging Fusion Besiege Attack and Probabilistic Switching Mechanism for Constrained Optimization Tasks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Poomin Duankhan,
  • Khamron Sunat,
  • Chitsutha Soomlek

摘要

Optimization problems are prevalent in real-world contexts and require effective strategies to navigate complex, high-dimensional landscapes characterized by numerous local optima. Existing algorithms frequently face challenges due to their limitations in addressing multiple constraints and complexities. This study introduces the enhanced adaptive crossover-based smell agent optimization (EACB-SAO) algorithm, which draws inspiration from the olfactory senses of living organisms. EACB-SAO offers two significant innovations: firstly, a novel hybrid operator known as Fusion Besiege Attack, which is influenced by the hard-besiege and soft-besiege techniques inherent to Harris Hawks Optimization, thus considerably improving exploitation capabilities. Secondly, EACB-SAO incorporates a probabilistic switching mechanism within its trailing mode, facilitating adaptive transitions between the traditional trailing operator and the newly developed Fusion Besiege Attack operator. This approach aims to improve adaptability across different optimization contexts. To evaluate its effectiveness in complex numerical benchmarks and engineering design problems, EACB-SAO is compared against nine competing algorithms in 23 classical benchmark functions, 29 CEC2017, 12 CEC2022 benchmark functions and 13 real-world applications. The results of a scoring system demonstrate that EACB-SAO achieved the highest score of 100 for the CEC2017, CEC2022, and real-world scenarios, thereby establishing its superiority over other algorithms and advancing the capabilities of previous adaptive SAO approaches. These findings underscore EACB-SAO’s proficiency in addressing practical optimization challenges, highlighting its efficiency and advantages in resolving complex real-world problems. MATLAB code for the algorithm is available in the Code Ocean repository https://doi.org/10.24433/CO.0873412.v1 .