Journal of Frontiers of Computer Science and Technology ›› 2024, Vol. 18 ›› Issue (2): 403-424.DOI: 10.3778/j.issn.1673-9418.2211001

• Theory·Algorithm • Previous Articles     Next Articles

Multi-objective Dwarf Mongoose Optimization Algorithm with Leader Guidance and Dominated Solution Evolution Mechanism

ZHAO Shijie, ZHANG Hongyi, MA Shilin   

  1. 1. Institute of Intelligence Science and Optimization, Liaoning Technical University, Fuxin, Liaoning 123000, China
    2. Institute for Optimization and Decision Analytics, Liaoning Technical University, Fuxin, Liaoning 123000, China
  • Online:2024-02-01 Published:2024-02-01

领导者引导与支配解进化的多目标矮猫鼬算法

赵世杰,张红易,马世林   

  1. 1. 辽宁工程技术大学 智能科学与优化研究所,辽宁 阜新 123000
    2. 辽宁工程技术大学 运筹与优化研究院,辽宁 阜新 123000

Abstract: In the face of the increasingly complex multi-objective optimization problems, it is necessary to develop novel multi-objective optimization algorithms to meet the challenges. This paper proposes a multi-objective dwarf mongoose optimization algorithm (MODMO) with leader guidance and dominated solution dynamic reduction evolution mechanism. In the leader guidance mechanism, a dynamic trade-off factor is introduced to regulate the search radius of the scout mongoose exploring the mound. At the same time, an external archive is constructed with a non-inferior solution set and the leader is determined according to the non-dominated ranking level, and then the scout mongoose is guided to advance to the multi-objective frontier to improve the convergence of the algorithm. The dominant solution dynamic reduction evolution strategy is constructed to overcome the redundancy problem in the process of maintaining the external archive of non-inferior solutions. It dynamically selects the dominant solutions based on the dominance relationship and crowding distance and stores them in the external archive. The dominant solution information is integrated into the population evolution to realize the mining of multi-objective potential frontier and enhance the diversity of the algorithm. Compared with five representative algorithms on ZDT, DTLZ and WFG benchmark functions, experimental results show that MODMO algorithm has significant advantages in convergence and diversity.

Key words: multi-objective optimization, dwarf mongoose optimization algorithm, leader guidance mechanism, external archive, dominated solution dynamic reduction evolution strategy

摘要: 面对现实中日益复杂的多目标优化问题,需要发展新型多目标优化算法应对挑战。提出一种基于领导者引导与支配解动态缩减进化的多目标矮猫鼬优化算法(MODMO)。领导者引导机制通过引入动态权衡因子以调控侦察猫鼬探寻土丘的搜索半径,同时以非劣解集构建外部存档并根据非支配排序层级确定出领导者,进而引导侦察猫鼬向多目标前沿面推进以改善算法的收敛性;支配解动态缩减进化策略是为克服非劣解外部存档维护过程中的解冗余问题而构建,其以支配关系和拥挤距离动态筛选支配解并存入外部存档,以支配解信息融入种群进化实现多目标潜在前沿的挖掘并增强算法的多样性。在ZDT、DTLZ与WFG基准函数上,与5种代表性比较算法的实验结果表明MODMO算法在收敛性与多样性上均具有显著优势。

关键词: 多目标优化, 矮猫鼬优化算法, 领导者引导机制, 外部存档, 支配解动态缩减进化策略