Swarm and Evolutionary Computation最新文献_第3页

River-land multi-modal bulk cargo transportation problem with containerization 河陆多式联运散货运输的集装箱化问题

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-14 DOI: 10.1016/j.swevo.2025.102261

Wei Wu , Lijun Fan , Ruiyou Zhang , Qianli Ma , Peng Jia

This research addresses a river-land multi-modal bulk cargo transportation problem with containerization. It involves three transportation modes: inland waterway, railway, and road transportation. While heterogeneous vessels are commonly employed in inland waterway transportation, few studies have focused on the allocation of these vessels within the context of river-based multimodal transportation. Consequently, introducing decisions on container usage for bulk shipments, identifying containerization locations, and assigning heterogeneous ships to riverine channel in multimodal transportation presents significant challenges. An integer nonlinear programming model based on a directed graph, which incorporates constraints such as water depth, the availability of road and railway vehicles, and the capacity of containerization equipment throughout the planning horizon, is formulated and subsequently linearized. The objective is to minimize the total cost, including transportation, containerization, and cargo damage costs. A multiple ant colony algorithm embedded by a mathematical model is developed to solve the problem. Experiments conducted on numerous near-practical instances demonstrate the effectiveness of the solution methods. The results indicate that for medium- and large-scale instances, the methodology can achieve optimal or high-quality feasible solutions within a reasonable computation time.

本文研究了河陆多式联运散货的集装箱化问题。它涉及三种运输方式：内河运输、铁路运输和公路运输。虽然异质船舶通常用于内河运输，但很少有研究关注这些船舶在河流多式联运背景下的分配。因此，在多式联运中，引入关于散装货物集装箱使用的决定，确定集装箱化地点，并将异质船舶分配到内河航道，都是重大挑战。建立了一个基于有向图的整数非线性规划模型，该模型包含了诸如水深、公路和铁路车辆的可用性以及整个规划范围内集装箱化设备的容量等约束条件，并随后进行了线性化。目标是使总成本最小化，包括运输、集装箱化和货物损坏成本。提出了一种嵌入数学模型的多蚁群算法来解决这一问题。在大量接近实际的实例上进行的实验证明了求解方法的有效性。结果表明，对于大中型实例，该方法能在合理的计算时间内得到最优或高质量的可行解。

{"title":"River-land multi-modal bulk cargo transportation problem with containerization","authors":"Wei Wu , Lijun Fan , Ruiyou Zhang , Qianli Ma , Peng Jia","doi":"10.1016/j.swevo.2025.102261","DOIUrl":"10.1016/j.swevo.2025.102261","url":null,"abstract":"<div><div>This research addresses a river-land multi-modal bulk cargo transportation problem with containerization. It involves three transportation modes: inland waterway, railway, and road transportation. While heterogeneous vessels are commonly employed in inland waterway transportation, few studies have focused on the allocation of these vessels within the context of river-based multimodal transportation. Consequently, introducing decisions on container usage for bulk shipments, identifying containerization locations, and assigning heterogeneous ships to riverine channel in multimodal transportation presents significant challenges. An integer nonlinear programming model based on a directed graph, which incorporates constraints such as water depth, the availability of road and railway vehicles, and the capacity of containerization equipment throughout the planning horizon, is formulated and subsequently linearized. The objective is to minimize the total cost, including transportation, containerization, and cargo damage costs. A multiple ant colony algorithm embedded by a mathematical model is developed to solve the problem. Experiments conducted on numerous near-practical instances demonstrate the effectiveness of the solution methods. The results indicate that for medium- and large-scale instances, the methodology can achieve optimal or high-quality feasible solutions within a reasonable computation time.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102261"},"PeriodicalIF":8.5,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980706","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Adaptive pattern learning particle swarm optimization for large-scale optimization 大规模优化的自适应模式学习粒子群算法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-13 DOI: 10.1016/j.swevo.2025.102268

Zhi-Tao Lai , Zi-Jia Wang , Shuai Liu , Zong-Gan Chen , Zhi-Hui Zhan , Sam Kwong , Jun Zhang

Large scale optimization problems (LSOPs) are an important topic in the field of evolutionary computing (EC), and many researchers have designed various learning strategies to try to solve LSOPs more effectively. However, most of the learning strategies are with the fixed learning pattern during the whole evolution process and lack the adaptive adjustment mechanism according to individual property. In fact, different individuals are with different exploitation or exploration abilities, and are suitable for different learning patterns. Therefore, in this paper, we propose adaptive pattern learning particle swarm optimization (APLPSO) to solve LSOPs. In APLPSO, several learning patterns based on different numbers of learning exemplars are first generated to enrich the learning diversity of population. Then, each individual will evaluate the learning patterns and adaptively select its own appropriate learning pattern for updating. The experimental results on two widely used large-scale optimization test suites, CEC2010 and CEC2013, show that APLPSO significantly outperforms other state-of-the-art large-scale optimization algorithms, including the winners of the CEC2010 and CEC2012 competitions. Moreover, we apply APLPSO to a real-world large-scale portfolio optimization application to show its practical applicability.

大规模优化问题（Large scale optimization problems, lsop）是进化计算领域的一个重要课题，许多研究者设计了各种学习策略来尝试更有效地解决lsop问题。然而，大多数学习策略在整个进化过程中都是固定的学习模式，缺乏针对个体属性的适应性调整机制。事实上，不同的个体具有不同的开发或探索能力，适合不同的学习模式。因此，本文提出了自适应模式学习粒子群优化（APLPSO）来解决lsop问题。在APLPSO中，首先基于不同数量的学习范例生成多种学习模式，以丰富群体的学习多样性。然后，每个个体将评估学习模式，并自适应地选择适合自己的学习模式进行更新。在CEC2010和CEC2013两个广泛使用的大规模优化测试套件上的实验结果表明，APLPSO显著优于其他最先进的大规模优化算法，包括CEC2010和CEC2012比赛的获奖者。最后，将该算法应用于一个实际的大规模投资组合优化应用中，以验证其实用性。

{"title":"Adaptive pattern learning particle swarm optimization for large-scale optimization","authors":"Zhi-Tao Lai , Zi-Jia Wang , Shuai Liu , Zong-Gan Chen , Zhi-Hui Zhan , Sam Kwong , Jun Zhang","doi":"10.1016/j.swevo.2025.102268","DOIUrl":"10.1016/j.swevo.2025.102268","url":null,"abstract":"<div><div>Large scale optimization problems (LSOPs) are an important topic in the field of evolutionary computing (EC), and many researchers have designed various learning strategies to try to solve LSOPs more effectively. However, most of the learning strategies are with the fixed learning pattern during the whole evolution process and lack the adaptive adjustment mechanism according to individual property. In fact, different individuals are with different exploitation or exploration abilities, and are suitable for different learning patterns. Therefore, in this paper, we propose adaptive pattern learning particle swarm optimization (APLPSO) to solve LSOPs. In APLPSO, several learning patterns based on different numbers of learning exemplars are first generated to enrich the learning diversity of population. Then, each individual will evaluate the learning patterns and adaptively select its own appropriate learning pattern for updating. The experimental results on two widely used large-scale optimization test suites, CEC2010 and CEC2013, show that APLPSO significantly outperforms other state-of-the-art large-scale optimization algorithms, including the winners of the CEC2010 and CEC2012 competitions. Moreover, we apply APLPSO to a real-world large-scale portfolio optimization application to show its practical applicability.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102268"},"PeriodicalIF":8.5,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980787","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An efficient binary ant colony evolutionary algorithm for feature selection 一种高效的二元蚁群特征选择进化算法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-10 DOI: 10.1016/j.swevo.2025.102258

Peichen Xiong , Zhen Liu , Weiqing Xiong

Feature selection is an optimization problem of finding the optimal feature subset with high computational complexity and belongs to the NP-hard problem. Efficient direct solutions are typically unavailable, necessitating the design of algorithms to tackle such issues. To address this, we propose an efficient binary ant colony evolutionary algorithm to solve the feature selection problem, called EBACEA. Drawing on nucleic acid coding principles and the double helix structure of DNA, a binary network for ant traversal is developed, simplified into a one-dimensional chain by adding complement operators, and a corresponding pheromone-update operator was designed. Compared to the complete graph space, this algorithm requires low intelligence of individual agents, and the set of feasible solution nodes need not be explicitly recorded. Moreover, based on binary coding and inspired by DNA genetic principles, the algorithm incorporates three genetic operators: crossover, mutation and selection, to yield richer sets of solutions, can mitigate the risks of premature convergence and overcome the limitations of evolutionary computation in local search. To validate the performance of the proposed algorithm, we evaluated twelve public datasets and compared the results with those of eight swarm intelligence algorithms. The findings demonstrate that the proposed algorithm exhibits favorable execution time and achieves superior accuracy. Additionally, further comparative analysis confirms the proposed algorithm’s effectiveness.

特征选择是一个计算复杂度很高的寻找最优特征子集的优化问题，属于np困难问题。通常没有有效的直接解决方案，因此需要设计算法来解决此类问题。为了解决这个问题，我们提出了一种有效的二元蚁群进化算法来解决特征选择问题，称为EBACEA。利用核酸编码原理和DNA的双螺旋结构，建立了蚂蚁遍历的二元网络，通过添加补算子将其简化为一维链，并设计了相应的信息素更新算子。与完全图空间相比，该算法对个体智能体的要求较低，且不需要明确记录可行解节点集。该算法以二进制编码为基础，受DNA遗传原理的启发，结合交叉、突变和选择三种遗传算子，产生更丰富的解集，减轻了早熟收敛的风险，克服了进化计算在局部搜索中的局限性。为了验证所提出算法的性能，我们评估了12个公共数据集，并将结果与8种群体智能算法的结果进行了比较。结果表明，该算法具有较好的执行时间和较高的精度。进一步的对比分析验证了算法的有效性。

{"title":"An efficient binary ant colony evolutionary algorithm for feature selection","authors":"Peichen Xiong , Zhen Liu , Weiqing Xiong","doi":"10.1016/j.swevo.2025.102258","DOIUrl":"10.1016/j.swevo.2025.102258","url":null,"abstract":"<div><div>Feature selection is an optimization problem of finding the optimal feature subset with high computational complexity and belongs to the NP-hard problem. Efficient direct solutions are typically unavailable, necessitating the design of algorithms to tackle such issues. To address this, we propose an efficient binary ant colony evolutionary algorithm to solve the feature selection problem, called EBACEA. Drawing on nucleic acid coding principles and the double helix structure of DNA, a binary network for ant traversal is developed, simplified into a one-dimensional chain by adding complement operators, and a corresponding pheromone-update operator was designed. Compared to the complete graph space, this algorithm requires low intelligence of individual agents, and the set of feasible solution nodes need not be explicitly recorded. Moreover, based on binary coding and inspired by DNA genetic principles, the algorithm incorporates three genetic operators: crossover, mutation and selection, to yield richer sets of solutions, can mitigate the risks of premature convergence and overcome the limitations of evolutionary computation in local search. To validate the performance of the proposed algorithm, we evaluated twelve public datasets and compared the results with those of eight swarm intelligence algorithms. The findings demonstrate that the proposed algorithm exhibits favorable execution time and achieves superior accuracy. Additionally, further comparative analysis confirms the proposed algorithm’s effectiveness.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102258"},"PeriodicalIF":8.5,"publicationDate":"2026-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980786","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Evolutionary algorithm with domain-specific operators for UAV path planning 基于特定域算子的无人机路径规划进化算法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-08 DOI: 10.1016/j.swevo.2025.102267

Daison Darlan , Oladayo S. Ajani , Rammohan Mallipeddi

Multi-objective UAV path planning is critical in practical applications such as surveillance, search-and-rescue missions, and environmental monitoring. However, the inherent complexity stemming from conflicting objectives, dynamic operational environments, and stringent mission constraints severely limits the efficacy of conventional evolutionary algorithms. Standard evolutionary operators typically fail to adequately respect domain-specific constraints, leading to infeasible or inefficient flight trajectories. Motivated by these limitations, this paper proposes specialized evolutionary operators tailored explicitly for multi-objective UAV path planning. We introduce a novel crossover operator that strategically employs the A* algorithm to generate feasible offspring paths between selected waypoints from parent solutions. Additionally, we present an adaptive polynomial mutation mechanism that dynamically controls exploration and exploitation by adjusting the mutation factor progressively across generations. Complementing this, we propose a secondary mutation operator utilizing A* to refine path segments effectively. Comprehensive ablation studies demonstrate the synergistic advantage of these innovations. Extensive evaluations on a realistic benchmark environment illustrate that our approach achieves significant enhancements, validated through substantial improvements in the hypervolume metric. Our findings confirm that embedding domain-specific intelligence into evolutionary operators markedly advances the state-of-the-art in multi-objective UAV path planning.

多目标无人机路径规划在监视、搜救任务和环境监测等实际应用中至关重要。然而，冲突的目标、动态的作战环境和严格的任务约束所产生的内在复杂性严重限制了传统进化算法的有效性。标准进化算子通常不能充分尊重特定领域的约束，导致不可行或低效的飞行轨迹。基于这些局限性，本文提出了专门针对多目标无人机路径规划的进化算子。我们引入了一种新的交叉算子，该算子策略性地使用a *算法在父解的选定路径点之间生成可行的子代路径。此外，我们提出了一种自适应多项式突变机制，通过逐步调整突变因子来动态控制探索和开发。作为补充，我们提出了一个利用a *的二次变异算子来有效地细化路径段。综合消融研究证明了这些创新的协同优势。在现实基准测试环境中进行的广泛评估表明，我们的方法实现了显著的增强，并通过对hypervolume度量的实质性改进进行了验证。我们的研究结果证实，在进化算子中嵌入特定领域的智能显著提高了多目标无人机路径规划的水平。

{"title":"Evolutionary algorithm with domain-specific operators for UAV path planning","authors":"Daison Darlan , Oladayo S. Ajani , Rammohan Mallipeddi","doi":"10.1016/j.swevo.2025.102267","DOIUrl":"10.1016/j.swevo.2025.102267","url":null,"abstract":"<div><div>Multi-objective UAV path planning is critical in practical applications such as surveillance, search-and-rescue missions, and environmental monitoring. However, the inherent complexity stemming from conflicting objectives, dynamic operational environments, and stringent mission constraints severely limits the efficacy of conventional evolutionary algorithms. Standard evolutionary operators typically fail to adequately respect domain-specific constraints, leading to infeasible or inefficient flight trajectories. Motivated by these limitations, this paper proposes specialized evolutionary operators tailored explicitly for multi-objective UAV path planning. We introduce a novel crossover operator that strategically employs the A* algorithm to generate feasible offspring paths between selected waypoints from parent solutions. Additionally, we present an adaptive polynomial mutation mechanism that dynamically controls exploration and exploitation by adjusting the mutation factor progressively across generations. Complementing this, we propose a secondary mutation operator utilizing A* to refine path segments effectively. Comprehensive ablation studies demonstrate the synergistic advantage of these innovations. Extensive evaluations on a realistic benchmark environment illustrate that our approach achieves significant enhancements, validated through substantial improvements in the hypervolume metric. Our findings confirm that embedding domain-specific intelligence into evolutionary operators markedly advances the state-of-the-art in multi-objective UAV path planning.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102267"},"PeriodicalIF":8.5,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145915141","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Evaluation-based multi-objective optimization for responsive-resilient supply chain network under hybrid uncertainty 混合不确定性下响应弹性供应链网络基于评价的多目标优化

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-08 DOI: 10.1016/j.swevo.2025.102274

Yurong Guo, Keping Zhou

Designing supply chains that remain responsive and resilient under hybrid uncertainty constitutes a complex multi-objective optimization challenge. This study introduces an evaluation-driven evolutionary framework that integrates metaheuristic search with performance feedback learning. A multi-objective mixed-integer programming model is formulated to simultaneously minimize total cost and maximize responsiveness and resilience. To effectively manage stochastic and fuzzy uncertainty, an improved fuzzy robust stochastic programming (IFRSP) model is developed, incorporating adaptive penalty terms to enhance feasibility robustness. A novel hybrid optimization–evaluation mechanism couples multi-objective optimization framework with a CRITIC–TOPSIS evaluator, where real-time performance feedback dynamically guides the evolutionary process toward high-quality Pareto solutions. Four representative multi-objective meta-heuristic algorithms are embedded and benchmarked. Computational results reveal that the proposed framework exhibits superior convergence diversity, stability, and decision robustness compared to classical approaches. The proposed methodology establishes a self-adaptive learning mechanism between optimization and evaluation, advancing the integration of metaheuristic evolution, performance assessment, and hybrid uncertainty modeling. This work offers a generalizable paradigm for intelligent decision-making in large-scale, uncertainty-driven optimization environments.

设计在混合不确定性下保持响应和弹性的供应链是一个复杂的多目标优化挑战。本研究引入了一个评估驱动的进化框架，将元启发式搜索与绩效反馈学习相结合。建立了一个多目标混合整数规划模型，以使总成本最小化，同时使响应能力和弹性最大化。为了有效地管理随机和模糊不确定性，提出了一种改进的模糊鲁棒随机规划（IFRSP）模型，并引入自适应惩罚项来增强可行性鲁棒性。一种新的混合优化-评估机制将多目标优化框架与critical - topsis评估器相结合，其中实时性能反馈动态引导进化过程走向高质量的Pareto解。嵌入了四种具有代表性的多目标元启发式算法并对其进行了基准测试。计算结果表明，与经典方法相比，该框架具有更好的收敛多样性、稳定性和决策鲁棒性。该方法在优化与评价之间建立了一种自适应学习机制，促进了元启发式进化、绩效评估和混合不确定性建模的融合。这项工作为大规模、不确定性驱动的优化环境中的智能决策提供了一个可推广的范例。

{"title":"Evaluation-based multi-objective optimization for responsive-resilient supply chain network under hybrid uncertainty","authors":"Yurong Guo, Keping Zhou","doi":"10.1016/j.swevo.2025.102274","DOIUrl":"10.1016/j.swevo.2025.102274","url":null,"abstract":"<div><div>Designing supply chains that remain responsive and resilient under hybrid uncertainty constitutes a complex multi-objective optimization challenge. This study introduces an evaluation-driven evolutionary framework that integrates metaheuristic search with performance feedback learning. A multi-objective mixed-integer programming model is formulated to simultaneously minimize total cost and maximize responsiveness and resilience. To effectively manage stochastic and fuzzy uncertainty, an improved fuzzy robust stochastic programming (IFRSP) model is developed, incorporating adaptive penalty terms to enhance feasibility robustness. A novel hybrid optimization–evaluation mechanism couples multi-objective optimization framework with a CRITIC–TOPSIS evaluator, where real-time performance feedback dynamically guides the evolutionary process toward high-quality Pareto solutions. Four representative multi-objective meta-heuristic algorithms are embedded and benchmarked. Computational results reveal that the proposed framework exhibits superior convergence diversity, stability, and decision robustness compared to classical approaches. The proposed methodology establishes a self-adaptive learning mechanism between optimization and evaluation, advancing the integration of metaheuristic evolution, performance assessment, and hybrid uncertainty modeling. This work offers a generalizable paradigm for intelligent decision-making in large-scale, uncertainty-driven optimization environments.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102274"},"PeriodicalIF":8.5,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145915142","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A multi-objective evolutionary algorithm with clustering-based archiving and adaptive search mechanism 基于聚类归档和自适应搜索机制的多目标进化算法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-08 DOI: 10.1016/j.swevo.2025.102277

Yiting Zeng, Peng Shao, Shaoping Zhang

To address the challenges of unstable archiving mechanisms and the difficulty of balancing diversity and convergence in multi-objective evolutionary algorithms, this paper proposes a multi-objective evolutionary algorithm with clustering-based archiving and an adaptive search mechanism based on Harris Hawks optimization (MOCAS/HHO). Building on the framework of the Harris Hawks Optimization (HHO), MOCAS/HHO employs k-medoids clustering to update the archive, where representative solutions at the cluster centers are preserved to improve solution diversity. Subsequently, MOCAS/HHO identifies ‘valuable solutions’ from the archive to guide the population toward the correct search direction. Based on the proportion and saturation of the ‘valuable solutions’, a regulatory factor is introduced to perturb the escape energy E, enabling the algorithm to adaptively adjust its search direction. Moreover, leaders are randomly selected from the valuable solutions to enhance stability and the global search capability of MOCAS/HHO. For the performance evaluation, MOCAS/HHO is compared with 9 algorithms on 25 benchmark functions, using IGD and HV metrics and statistical analysis. MOCAS/HHO outperforms MOHHO on approximately 88 % of the selected 2–3 objective functions, while achieving superior performance on all chosen 4-objective high-dimensional functions. For the Car side impact design problem, MOCAS/HHO improves IGD by 24.3 % over MOEDO; for the Liquid-rocket single element injector design, it improves IGD by 65.95 % over MOGWO; and for Conceptual marine design, it ranks second in IGD to MOEA/D. Overall, these results indicate that MOCAS/HHO achieves a good balance between convergence and diversity across both benchmark test functions and practical engineering applications.

针对多目标进化算法中归档机制不稳定以及难以平衡多样性和收敛性的问题，提出了一种基于聚类归档和基于Harris Hawks优化的自适应搜索机制的多目标进化算法（MOCAS/HHO）。在Harris Hawks Optimization （HHO）框架的基础上，MOCAS/HHO采用k- medioids聚类更新存档，其中保留了集群中心的代表性解决方案，以提高解决方案的多样性。随后，MOCAS/HHO从档案中识别出“有价值的解决方案”，以引导人们朝着正确的搜索方向前进。根据“有价解”的比例和饱和度，引入调节因子扰动逃逸能E，使算法能够自适应调整搜索方向。此外，从有价值的解决方案中随机选择领导者，以增强MOCAS/HHO的稳定性和全局搜索能力。为了进行性能评价，利用IGD和HV指标和统计分析，在25个基准函数上比较了MOCAS/HHO与9种算法的性能。在所选的2-3个目标函数中，MOCAS/HHO在约88%的目标函数上优于MOHHO，而在所有所选的4-目标高维函数上均取得了优异的性能。对于汽车侧面碰撞设计问题，moas /HHO比MOEDO提高了24.3%的IGD；对于液体火箭单元件喷射器设计，IGD比MOGWO提高65.95%；船舶概念设计IGD排名第二，仅次于MOEA/D。总体而言，这些结果表明，在基准测试功能和实际工程应用中，MOCAS/HHO在收敛性和多样性之间取得了良好的平衡。

{"title":"A multi-objective evolutionary algorithm with clustering-based archiving and adaptive search mechanism","authors":"Yiting Zeng, Peng Shao, Shaoping Zhang","doi":"10.1016/j.swevo.2025.102277","DOIUrl":"10.1016/j.swevo.2025.102277","url":null,"abstract":"<div><div>To address the challenges of unstable archiving mechanisms and the difficulty of balancing diversity and convergence in multi-objective evolutionary algorithms, this paper proposes a multi-objective evolutionary algorithm with clustering-based archiving and an adaptive search mechanism based on Harris Hawks optimization (MOCAS/HHO). Building on the framework of the Harris Hawks Optimization (HHO), MOCAS/HHO employs <em>k</em>-medoids clustering to update the archive, where representative solutions at the cluster centers are preserved to improve solution diversity. Subsequently, MOCAS/HHO identifies ‘valuable solutions’ from the archive to guide the population toward the correct search direction. Based on the proportion and saturation of the ‘valuable solutions’, a regulatory factor is introduced to perturb the escape energy <em>E</em>, enabling the algorithm to adaptively adjust its search direction. Moreover, leaders are randomly selected from the valuable solutions to enhance stability and the global search capability of MOCAS/HHO. For the performance evaluation, MOCAS/HHO is compared with 9 algorithms on 25 benchmark functions, using IGD and HV metrics and statistical analysis. MOCAS/HHO outperforms MOHHO on approximately 88 % of the selected 2–3 objective functions, while achieving superior performance on all chosen 4-objective high-dimensional functions. For the Car side impact design problem, MOCAS/HHO improves IGD by 24.3 % over MOEDO; for the Liquid-rocket single element injector design, it improves IGD by 65.95 % over MOGWO; and for Conceptual marine design, it ranks second in IGD to MOEA/D. Overall, these results indicate that MOCAS/HHO achieves a good balance between convergence and diversity across both benchmark test functions and practical engineering applications.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102277"},"PeriodicalIF":8.5,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145915143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A hybrid surrogate-assisted evolutionary algorithm for technician routing and scheduling problems 技术人员路由调度问题的混合代理辅助进化算法

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-01 DOI: 10.1016/j.swevo.2025.102280

Engin PEKEL

Technician routing and scheduling problems (TRSP) that involve time windows and skill-based requirements represent complex optimization tasks frequently encountered in areas such as maintenance and field service operations. Because these problems belong to the class of NP-hard combinatorial challenges, obtaining exact solutions within realistic scales is computationally prohibitive. To address this challenge, the present study proposes a hybrid surrogate-assisted evolutionary algorithm (HSAEA) that combines several artificial intelligence (AI) strategies to obtain high-quality solutions more efficiently. At its core, the framework employs a Genetic Algorithm (GA) that is specifically designed to allocate and schedule service jobs among available technicians. To enhance this GA, the paper incorporates: a surrogate model – a Multi-Layer Perceptron (MLP) regressor trained on domain-specific features – to approximate the expensive fitness evaluation of candidate schedules; a reinforcement learning (RL) based adaptive mutation operator selection mechanism that dynamically chooses mutation operators based on their past performance, aiming to balance exploration and exploitation; and a simulated annealing (SA) procedure applied as a local search for intensification, refining individual solutions by exploiting neighborhood improvements. The proposed method is formalized and each component’s role is detailed. To sum up, the proposed hybrid framework combines several complementary ideas: surrogate models are used to cut down computational effort, reinforcement learning helps direct the search process more intelligently, and simulated annealing refines the solutions at a local level. When tested on standard benchmark datasets, the method consistently produced better-quality solutions than the baseline algorithms, and it did so within practical computation times.

技术人员路线和调度问题（TRSP）涉及时间窗口和基于技能的要求，是维护和现场服务操作等领域经常遇到的复杂优化任务。由于这些问题属于NP-hard组合挑战的范畴，在实际尺度内获得精确解在计算上是禁止的。为了应对这一挑战，本研究提出了一种混合代理辅助进化算法（HSAEA），该算法结合了几种人工智能（AI）策略，以更有效地获得高质量的解决方案。该框架的核心是采用遗传算法（GA），该算法专门用于在可用的技术人员之间分配和调度服务工作。为了增强这种遗传算法，本文采用了：一个代理模型-一个基于特定领域特征训练的多层感知器（MLP）回归器-来近似候选时间表的昂贵适应度评估；基于强化学习（RL）的自适应突变算子选择机制，该机制根据突变算子过去的表现动态选择突变算子，旨在平衡探索和利用；模拟退火（SA）程序应用于局部搜索强化，通过利用邻域改进来改进单个解决方案。对所提出的方法进行了形式化描述，并详细说明了每个组件的作用。综上所述，所提出的混合框架结合了几个互补的思想：使用代理模型来减少计算工作量，强化学习有助于更智能地指导搜索过程，模拟退火在局部级别改进解决方案。当在标准基准数据集上进行测试时，该方法始终产生比基线算法质量更好的解决方案，并且在实际计算时间内做到了这一点。

{"title":"A hybrid surrogate-assisted evolutionary algorithm for technician routing and scheduling problems","authors":"Engin PEKEL","doi":"10.1016/j.swevo.2025.102280","DOIUrl":"10.1016/j.swevo.2025.102280","url":null,"abstract":"<div><div>Technician routing and scheduling problems (TRSP) that involve time windows and skill-based requirements represent complex optimization tasks frequently encountered in areas such as maintenance and field service operations. Because these problems belong to the class of NP-hard combinatorial challenges, obtaining exact solutions within realistic scales is computationally prohibitive. To address this challenge, the present study proposes a hybrid surrogate-assisted evolutionary algorithm (HSAEA) that combines several artificial intelligence (AI) strategies to obtain high-quality solutions more efficiently. At its core, the framework employs a Genetic Algorithm (GA) that is specifically designed to allocate and schedule service jobs among available technicians. To enhance this GA, the paper incorporates: a surrogate model – a Multi-Layer Perceptron (MLP) regressor trained on domain-specific features – to approximate the expensive fitness evaluation of candidate schedules; a reinforcement learning (RL) based adaptive mutation operator selection mechanism that dynamically chooses mutation operators based on their past performance, aiming to balance exploration and exploitation; and a simulated annealing (SA) procedure applied as a local search for intensification, refining individual solutions by exploiting neighborhood improvements. The proposed method is formalized and each component’s role is detailed. To sum up, the proposed hybrid framework combines several complementary ideas: surrogate models are used to cut down computational effort, reinforcement learning helps direct the search process more intelligently, and simulated annealing refines the solutions at a local level. When tested on standard benchmark datasets, the method consistently produced better-quality solutions than the baseline algorithms, and it did so within practical computation times.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"100 ","pages":"Article 102280"},"PeriodicalIF":8.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145883719","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Opposition based learning for metaheuristic algorithms: Theory, variants, applications, and performance evaluation 基于对立学习的元启发式算法：理论、变体、应用和性能评估

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-01 DOI: 10.1016/j.swevo.2025.102271

Rebika Rai , Buddhadev Sasmal , Arunita Das , Totan Bharasa , Krishna Gopal Dhal , Prabir Kumar Naskar

Despite the widespread popularity of Metaheuristic Algorithms (MAs) in solving complex optimization problems, several challenges persist. These challenges include slow convergence rates, limited exploration capabilities, and premature convergence. Opposition-Based Learning (OBL) is a moderately novel and promising method that has the potential to boost the effectiveness of different soft computing algorithms, including Fuzzy Logic, Artificial Neural Networks, and Evolutionary Algorithms. OBL has demonstrated its ability to enhance the performance of these algorithms effectively. OBL has shown that it can effectively improve the performance of these algorithms. The process of OBL entails producing a group of solutions that are opposite in nature to each proposed solution, which helps to broadens the range of population and makes it easier to explore the search area. In this article, we give a thorough analysis of the OBL technique and how it is used in MAs. We talk about the theoretical underpinnings of OBL and the various variations that have been put forth in the literature. Additionally, we provide an overview of recent research studies that have applied OBL to a broad spectrum of optimization problems. We conclude by exploring into the possible future research paths for OBL and its potential applications in diverse fields. To accomplish this, a total of 168 research papers on MAs with OBL, published in various journals and conferences between 2021 and 2025, were reviewed. From the study, it is clear that OBL is an effective and efficient strategy for enhancing the performance of MAs and has the potential to advance the field of optimization further.

尽管元启发式算法（MAs）在解决复杂优化问题方面广泛流行，但仍存在一些挑战。这些挑战包括收敛速度慢、勘探能力有限和过早收敛。基于对立的学习（OBL）是一种比较新颖和有前途的方法，有可能提高不同软计算算法的有效性，包括模糊逻辑、人工神经网络和进化算法。OBL已经证明了它能够有效地提高这些算法的性能。实验表明，OBL可以有效地提高这些算法的性能。OBL的过程需要产生一组解决方案，这些解决方案在性质上与每个提出的解决方案相反，这有助于扩大人口范围，并使探索搜索区域变得更容易。在本文中，我们对OBL技术及其在MAs中的应用进行了全面的分析。我们讨论了OBL的理论基础以及文献中提出的各种变体。此外，我们还概述了最近将OBL应用于广泛的优化问题的研究。最后，我们探讨了OBL未来可能的研究路径及其在不同领域的潜在应用。为了实现这一目标，共审查了2021年至2025年期间在各种期刊和会议上发表的168篇关于具有OBL的MAs的研究论文。从研究中可以清楚地看出，OBL是一种有效和高效的策略，可以提高MAs的性能，并有可能进一步推动优化领域的发展。

{"title":"Opposition based learning for metaheuristic algorithms: Theory, variants, applications, and performance evaluation","authors":"Rebika Rai , Buddhadev Sasmal , Arunita Das , Totan Bharasa , Krishna Gopal Dhal , Prabir Kumar Naskar","doi":"10.1016/j.swevo.2025.102271","DOIUrl":"10.1016/j.swevo.2025.102271","url":null,"abstract":"<div><div>Despite the widespread popularity of Metaheuristic Algorithms (MAs) in solving complex optimization problems, several challenges persist. These challenges include slow convergence rates, limited exploration capabilities, and premature convergence. Opposition-Based Learning (OBL) is a moderately novel and promising method that has the potential to boost the effectiveness of different soft computing algorithms, including Fuzzy Logic, Artificial Neural Networks, and Evolutionary Algorithms. OBL has demonstrated its ability to enhance the performance of these algorithms effectively. OBL has shown that it can effectively improve the performance of these algorithms. The process of OBL entails producing a group of solutions that are opposite in nature to each proposed solution, which helps to broadens the range of population and makes it easier to explore the search area. In this article, we give a thorough analysis of the OBL technique and how it is used in MAs. We talk about the theoretical underpinnings of OBL and the various variations that have been put forth in the literature. Additionally, we provide an overview of recent research studies that have applied OBL to a broad spectrum of optimization problems. We conclude by exploring into the possible future research paths for OBL and its potential applications in diverse fields. To accomplish this, a total of 168 research papers on MAs with OBL, published in various journals and conferences between 2021 and 2025, were reviewed. From the study, it is clear that OBL is an effective and efficient strategy for enhancing the performance of MAs and has the potential to advance the field of optimization further.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"100 ","pages":"Article 102271"},"PeriodicalIF":8.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145883645","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Economic optimization scheduling of combined cooling, heat, and power systems utilizing a multi-strategy enhanced PSO algorithm 利用多策略增强型粒子群算法的冷、热、电联合系统经济优化调度

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-01 DOI: 10.1016/j.swevo.2025.102278

Beichen Chen , Haoyuan Lv , Wenpeng Hong , Qingyu Su , Jian Li

Under the dual carbon goals, enhancing the energy efficiency of combined cooling, heating, and power (CCHP) systems while reducing carbon emissions is crucial. As the physical carrier of the energy internet, CCHP systems deliver comprehensive benefits, including energy savings, environmental improvement, and enhanced power supply reliability. This paper establishes a CCHP system model based on the principle of energy gradient utilization. Under time-of-use electricity pricing, the model comprehensively considers electricity consumption, heating, and cooling demands. Subsequently, a hybrid algorithm BC-GWOPSO is proposed, combining an improved Gray Wolf Optimization (GWO) algorithm with Particle Swarm Optimization (PSO). The specific improvement strategy involves using the

β

- distribution strategy to adjust the inertia weight in PSO and applying the cosine law to modify the convergence factor in GWO. For typical summer and winter days, total cost is adopted as the objective function for optimization scheduling, enabling more rational power allocation among units within the CCHP system and minimizing system costs. Finally, the BC-GWOPSO algorithm was experimentally compared with four other optimization algorithms. Friedman and Wilcoxon test results show that BC-GWOPSO algorithm is superior to the other four algorithms. The CCHP system optimization operation results demonstrate that the proposed method effectively reduces total operating costs, environmental costs, and load loss costs. Compared to other algorithms, it exhibits faster convergence speed and better stability, providing an effective scheduling solution for CCHP systems.

在双碳目标下，提高冷热电联产系统的能源效率，同时减少碳排放至关重要。作为能源互联网的物理载体，CCHP系统具有节约能源、改善环境和提高供电可靠性等综合效益。本文建立了基于能量梯度利用原理的热电联产系统模型。在分时电价下，该模型综合考虑了用电量、供热需求和制冷需求。随后，将改进的灰狼优化（GWO）算法与粒子群优化（PSO）算法相结合，提出了BC-GWOPSO混合算法。具体改进策略包括利用β-分布策略调整粒子群中的惯性权值，利用余弦定律修正粒子群中的收敛因子。对于典型的夏天和冬季，采用总成本作为优化调度的目标函数，使热电联产系统内各机组之间的功率分配更加合理，使系统成本最小化。最后，将BC-GWOPSO算法与其他四种优化算法进行了实验比较。Friedman和Wilcoxon测试结果表明，BC-GWOPSO算法优于其他四种算法。热电联产系统优化运行结果表明，该方法有效降低了总运行成本、环境成本和负荷损失成本。与其他算法相比，该算法具有更快的收敛速度和更好的稳定性，为热电联产系统提供了有效的调度解决方案。

{"title":"Economic optimization scheduling of combined cooling, heat, and power systems utilizing a multi-strategy enhanced PSO algorithm","authors":"Beichen Chen , Haoyuan Lv , Wenpeng Hong , Qingyu Su , Jian Li","doi":"10.1016/j.swevo.2025.102278","DOIUrl":"10.1016/j.swevo.2025.102278","url":null,"abstract":"<div><div>Under the dual carbon goals, enhancing the energy efficiency of combined cooling, heating, and power (CCHP) systems while reducing carbon emissions is crucial. As the physical carrier of the energy internet, CCHP systems deliver comprehensive benefits, including energy savings, environmental improvement, and enhanced power supply reliability. This paper establishes a CCHP system model based on the principle of energy gradient utilization. Under time-of-use electricity pricing, the model comprehensively considers electricity consumption, heating, and cooling demands. Subsequently, a hybrid algorithm BC-GWOPSO is proposed, combining an improved Gray Wolf Optimization (GWO) algorithm with Particle Swarm Optimization (PSO). The specific improvement strategy involves using the <span><math><mi>β</mi></math></span>- distribution strategy to adjust the inertia weight in PSO and applying the cosine law to modify the convergence factor in GWO. For typical summer and winter days, total cost is adopted as the objective function for optimization scheduling, enabling more rational power allocation among units within the CCHP system and minimizing system costs. Finally, the BC-GWOPSO algorithm was experimentally compared with four other optimization algorithms. Friedman and Wilcoxon test results show that BC-GWOPSO algorithm is superior to the other four algorithms. The CCHP system optimization operation results demonstrate that the proposed method effectively reduces total operating costs, environmental costs, and load loss costs. Compared to other algorithms, it exhibits faster convergence speed and better stability, providing an effective scheduling solution for CCHP systems.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"100 ","pages":"Article 102278"},"PeriodicalIF":8.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145883722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Surrogate-assisted evolutionary optimization using on-demand helper task for high-dimensional expensive multi-objective problems 基于按需助手任务的高维昂贵多目标问题代理辅助进化优化

IF 8.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Swarm and Evolutionary Computation

Pub Date : 2026-01-01 DOI: 10.1016/j.swevo.2025.102273

Qian Li, Yuanchao Liu, Jianchang Liu

Surrogate-assisted evolutionary algorithms (SAEAs) have been recognized as well-suited for solving expensive multi-objective optimization problems (EMOPs). However, most existing SAEAs show promising performance on low-dimensional expensive multi-objective optimization, and rarely pay attention to solving high-dimensional EMOPs. Thus, this work proposes an SAEA using on-demand helper task, termed SAEA-DHT, to efficiently address high-dimensional EMOPs. In SAEA-DHT, an on-demand helper task construction mechanism is proposed to create a low-dimensional helper task tailored to the optimization stage. During the convergence urgent stage, the helper task targets convergence-critical decision variables, whereas in the diversity-demand stage, it shifts the focus to the remaining decision variables. Furthermore, a cross-dimensional knowledge transfer approach is incorporated to efficiently transfer high-quality solutions discovered by the helper task to the original task, thereby accelerating convergence towards the true Pareto front. Finally, an adaptive infill selection scheme is proposed. This scheme dynamically selects infill solutions for exact fitness evaluations based on either convergence-driven or diversity-driven criteria. Extensive experiments are conducted on three well-known benchmarks and time-varying ratio error estimation problems containing up to 200 decision variables. The experimental results demonstrate the advantages of SAEA-DHT over six state-of-the-art SAEAs in solving high-dimensional EMOPs.

代理辅助进化算法（saea）被认为非常适合解决昂贵的多目标优化问题（EMOPs）。然而，现有的多目标优化算法大多在低维昂贵的多目标优化问题上表现良好，很少关注高维emop问题的求解。因此，这项工作提出了一个SAEA使用按需助手任务，称为SAEA- dht，以有效地解决高维emop。在SAEA-DHT中，提出了一种按需辅助任务构建机制，以创建适合优化阶段的低维辅助任务。在收敛紧急阶段，辅助任务针对收敛关键决策变量，而在多样性需求阶段，它将重点转移到剩余决策变量。此外，采用跨维度的知识转移方法，将辅助任务发现的高质量解有效地转移到原始任务中，从而加速向真正的帕累托前沿的收敛。最后，提出了一种自适应填充选择方案。该方案基于收敛驱动或多样性驱动准则动态选择精确适应度评估的填充解。在包含多达200个决策变量的三个著名基准和时变比误差估计问题上进行了广泛的实验。实验结果表明，在求解高维emop时，SAEA-DHT算法优于6种最先进的saea算法。

{"title":"Surrogate-assisted evolutionary optimization using on-demand helper task for high-dimensional expensive multi-objective problems","authors":"Qian Li, Yuanchao Liu, Jianchang Liu","doi":"10.1016/j.swevo.2025.102273","DOIUrl":"10.1016/j.swevo.2025.102273","url":null,"abstract":"<div><div>Surrogate-assisted evolutionary algorithms (SAEAs) have been recognized as well-suited for solving expensive multi-objective optimization problems (EMOPs). However, most existing SAEAs show promising performance on low-dimensional expensive multi-objective optimization, and rarely pay attention to solving high-dimensional EMOPs. Thus, this work proposes an SAEA using on-demand helper task, termed SAEA-DHT, to efficiently address high-dimensional EMOPs. In SAEA-DHT, an on-demand helper task construction mechanism is proposed to create a low-dimensional helper task tailored to the optimization stage. During the convergence urgent stage, the helper task targets convergence-critical decision variables, whereas in the diversity-demand stage, it shifts the focus to the remaining decision variables. Furthermore, a cross-dimensional knowledge transfer approach is incorporated to efficiently transfer high-quality solutions discovered by the helper task to the original task, thereby accelerating convergence towards the true Pareto front. Finally, an adaptive infill selection scheme is proposed. This scheme dynamically selects infill solutions for exact fitness evaluations based on either convergence-driven or diversity-driven criteria. Extensive experiments are conducted on three well-known benchmarks and time-varying ratio error estimation problems containing up to 200 decision variables. The experimental results demonstrate the advantages of SAEA-DHT over six state-of-the-art SAEAs in solving high-dimensional EMOPs.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"100 ","pages":"Article 102273"},"PeriodicalIF":8.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145883646","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0