Swarm and Evolutionary Computation最新文献_第3页

A hybrid evolutionary algorithm for 2D variable-sized bin packing with guillotine constraint in manufacturing 制造中带有断头台约束的二维变尺寸装箱的混合进化算法

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-02-13 DOI: 10.1016/j.swevo.2026.102317

Qiang Luo , Zuogan Tang , Chunrong Pan , Yunqing Rao

This paper addresses a challenging variant of the two-dimensional variable-sized bin packing problem, characterized by multiple bin sizes, the guillotine-cut constraint, rotatable items, and few item types with high demand quantities. This problem is motivated by industrial application of insulating pressboard cutting in transformers. A hybrid evolutionary algorithm is proposed to solve the problem. The method encodes only sheet types to optimize their usage sequence and generates packing patterns by a heuristic algorithm via processing the encoding. Packing pattern is represented as multitree and constructed using a cutting-based principle. During item placement, a randomization strategy selects from top candidate items to avoid local optima, while compactness is enhanced by replicating individual items. Extensive experiments on three datasets show the proposed approach’s superiority in obtaining high-quality and robust solution. The algorithm substantially outperforms competing metaheuristics, achieving reductions of 4% to 25% across minimum, average, and maximum metrics. Crucially, it exhibits unprecedented solution stability, with standard deviations an order of magnitude lower than competitors. In addition, the randomization strategy yields significant improvements (2.9%-7.2%) over a deterministic variant. Statistical tests confirm all advantages are significant.

本文解决了二维变尺寸箱包装问题的一个具有挑战性的变体，其特征是多个箱尺寸，断头台切割约束，可旋转的物品，以及高需求数量的少数物品类型。这一问题是由变压器绝缘压板切割的工业应用引起的。提出了一种混合进化算法来解决这一问题。该方法对单片类型进行编码以优化其使用顺序，并通过对编码进行处理，采用启发式算法生成包装模式。包装模式被表示为多树，并使用基于切割的原则构造。在物品放置过程中，随机化策略从最优候选物品中进行选择，以避免局部最优，同时通过复制单个物品来增强紧凑性。在三个数据集上的大量实验表明，该方法在获得高质量和鲁棒性解方面具有优势。该算法大大优于与之竞争的元启发式算法，在最小、平均和最大度量上实现了4%到25%的减少。至关重要的是，它表现出前所未有的溶液稳定性，其标准差比竞争对手低一个数量级。此外，随机化策略比确定性变体产生显著的改进（2.9%-7.2%）。统计测试证实所有的优势都是显著的。

{"title":"A hybrid evolutionary algorithm for 2D variable-sized bin packing with guillotine constraint in manufacturing","authors":"Qiang Luo , Zuogan Tang , Chunrong Pan , Yunqing Rao","doi":"10.1016/j.swevo.2026.102317","DOIUrl":"10.1016/j.swevo.2026.102317","url":null,"abstract":"<div><div>This paper addresses a challenging variant of the two-dimensional variable-sized bin packing problem, characterized by multiple bin sizes, the guillotine-cut constraint, rotatable items, and few item types with high demand quantities. This problem is motivated by industrial application of insulating pressboard cutting in transformers. A hybrid evolutionary algorithm is proposed to solve the problem. The method encodes only sheet types to optimize their usage sequence and generates packing patterns by a heuristic algorithm via processing the encoding. Packing pattern is represented as multitree and constructed using a cutting-based principle. During item placement, a randomization strategy selects from top candidate items to avoid local optima, while compactness is enhanced by replicating individual items. Extensive experiments on three datasets show the proposed approach’s superiority in obtaining high-quality and robust solution. The algorithm substantially outperforms competing metaheuristics, achieving reductions of 4% to 25% across minimum, average, and maximum metrics. Crucially, it exhibits unprecedented solution stability, with standard deviations an order of magnitude lower than competitors. In addition, the randomization strategy yields significant improvements (2.9%-7.2%) over a deterministic variant. Statistical tests confirm all advantages are significant.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"102 ","pages":"Article 102317"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146193173","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 hyper-curvature balanced indicator and adaptive phase exploration co-driven evolutionary algorithm for many-objective optimization 一种多目标优化的超曲率平衡指标与自适应相位探测协同驱动进化算法

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-02-17 DOI: 10.1016/j.swevo.2026.102313

Xuezhi Yue , Wenlong Wen , Yanshen Jiang , Ye Tian , Hu Peng

In many-objective optimization (MaOPs), complexity grows with the number of objectives. Most evaluation metrics aim for both convergence and diversity, yet often face high computational costs or reliance on ideal reference points in high-dimensional spaces, hindering diversity preservation. Traditional algorithms use fixed evolutionary phases, limiting temporal adaptability and causing optimization mismatches. For this reason, this study introduces a hyper-curvature balanced indicator and adaptive phase exploration co-driven evolutionary algorithm for many-objective optimization (MaOEA-HAP). Firstly, an innovative hyper-curvature balance indicator is developed, utilizing selected populations characterized by strong convergence and well-distributed diversity, to efficiently direct the evolutionary process. Meanwhile, to eliminate dominant resistance solutions, an adaptive angular dominance pruning strategy is designed to enhance the algorithm’s capability for solution screening. Furthermore, the elite profile mechanism preserves solutions with strong convergence properties, effectively guiding the evolutionary search toward convergence enhancement. Finally, an adaptive stage exploration strategy is formulated to strengthen population exploration ability. To verify its effectiveness, MaOEA-HAP is compared against eight advanced many-objective optimization algorithms using 31 benchmark instances and five real-world optimization problems. Empirical results confirm that the designed algorithm demonstrates strong capability in addressing multiple challenges across many-objective optimization tasks.

在多目标优化（MaOPs）中，复杂度随着目标数量的增加而增加。大多数评价指标的目标是收敛性和多样性，但往往面临较高的计算成本或依赖于高维空间的理想参考点，阻碍了多样性的保存。传统算法使用固定的进化阶段，限制了时间适应性并导致优化不匹配。为此，本研究引入了一种超曲率平衡指标和自适应相位探测协同驱动的多目标优化进化算法（MaOEA-HAP）。首先，提出了一种创新的超曲率平衡指标，利用具有强收敛性和均匀分布多样性的选择种群，有效地指导进化过程；同时，为了消除优势阻力解，设计了自适应角优势剪枝策略，增强了算法的解筛选能力。此外，精英剖面机制保留了具有强收敛性质的解，有效地指导了进化搜索的收敛性增强。最后，制定了自适应阶段勘探策略，以增强种群勘探能力。为了验证其有效性，使用31个基准实例和5个实际优化问题，将MaOEA-HAP与8种先进的多目标优化算法进行了比较。实证结果表明，所设计的算法在解决多目标优化任务中的多重挑战方面表现出较强的能力。

{"title":"A hyper-curvature balanced indicator and adaptive phase exploration co-driven evolutionary algorithm for many-objective optimization","authors":"Xuezhi Yue , Wenlong Wen , Yanshen Jiang , Ye Tian , Hu Peng","doi":"10.1016/j.swevo.2026.102313","DOIUrl":"10.1016/j.swevo.2026.102313","url":null,"abstract":"<div><div>In many-objective optimization (MaOPs), complexity grows with the number of objectives. Most evaluation metrics aim for both convergence and diversity, yet often face high computational costs or reliance on ideal reference points in high-dimensional spaces, hindering diversity preservation. Traditional algorithms use fixed evolutionary phases, limiting temporal adaptability and causing optimization mismatches. For this reason, this study introduces a hyper-curvature balanced indicator and adaptive phase exploration co-driven evolutionary algorithm for many-objective optimization (MaOEA-HAP). Firstly, an innovative hyper-curvature balance indicator is developed, utilizing selected populations characterized by strong convergence and well-distributed diversity, to efficiently direct the evolutionary process. Meanwhile, to eliminate dominant resistance solutions, an adaptive angular dominance pruning strategy is designed to enhance the algorithm’s capability for solution screening. Furthermore, the elite profile mechanism preserves solutions with strong convergence properties, effectively guiding the evolutionary search toward convergence enhancement. Finally, an adaptive stage exploration strategy is formulated to strengthen population exploration ability. To verify its effectiveness, MaOEA-HAP is compared against eight advanced many-objective optimization algorithms using 31 benchmark instances and five real-world optimization problems. Empirical results confirm that the designed algorithm demonstrates strong capability in addressing multiple challenges across many-objective optimization tasks.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"102 ","pages":"Article 102313"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147423704","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

Dynamic distributed production scheduling problem with simultaneous job arrivals and machine breakdowns: Learning to learn iterated greedy algorithm 同时到达作业和机器故障的动态分布式生产调度问题：学习学习迭代贪婪算法

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-02-14 DOI: 10.1016/j.swevo.2026.102327

Qing Zhou , Weishi Shao , Zhongshi Shao , Dechang Pi , Jiaquan Gao

Dynamic distributed production scheduling poses significant challenges to modern manufacturing systems due to environmental complexity and the need to respond to real-time disruptions. This paper studies the dynamic distributed hybrid flow shop problem (DDHFSP) with simultaneous dynamic job arrivals and machine breakdowns, aiming to minimize makespan. To address this problem, a learnable iterated greedy algorithm (LIG) is proposed within the classical IG framework. LIG is designed to learn to learn how to adaptively configure its operational components by extracting knowledge from search trajectories, thereby effectively handling dynamic scheduling environments. Based on the IG structure, four rescheduling strategies for dynamic events, four destruction-reconstruction strategies, and four local search strategies are designed. The algorithm employs a long short-term memory (LSTM) network to extract temporally dependent production features, and adopts proximal policy optimization (PPO) to build an agent that adaptively selects the most promising combination of strategies according to the current state for execution within the IG framework. Experimental results demonstrate that LIG outperforms several state-of-the-art metaheuristic and reinforcement learning-based methods across various problem scales and dynamic scenarios, exhibiting faster convergence, better solution quality, higher stability, and stronger generalization capability. Ultimately, the proposed learnable mechanism establishes a new paradigm for tackling dynamic disturbances in DDHFSP.

由于环境的复杂性和对实时中断的响应需求，动态分布式生产调度对现代制造系统提出了重大挑战。研究了同时存在动态作业到达和机器故障的动态分布式混合流水车间问题（DDHFSP），以最小化完工时间为目标。为了解决这个问题，在经典IG框架内提出了一种可学习的迭代贪婪算法（LIG）。LIG旨在学习如何通过从搜索轨迹中提取知识来自适应地配置其操作组件，从而有效地处理动态调度环境。基于IG结构，设计了4种动态事件重调度策略、4种破坏重建策略和4种局部搜索策略。该算法采用长短期记忆（LSTM）网络提取时间依赖的生产特征，并采用近端策略优化（PPO）构建agent，根据当前状态自适应选择最有希望的策略组合在IG框架内执行。实验结果表明，在各种问题尺度和动态场景下，LIG优于几种最先进的基于元启发式和强化学习的方法，具有更快的收敛速度、更好的解质量、更高的稳定性和更强的泛化能力。最后，提出的可学习机制为处理DDHFSP中的动态干扰建立了一个新的范例。

{"title":"Dynamic distributed production scheduling problem with simultaneous job arrivals and machine breakdowns: Learning to learn iterated greedy algorithm","authors":"Qing Zhou , Weishi Shao , Zhongshi Shao , Dechang Pi , Jiaquan Gao","doi":"10.1016/j.swevo.2026.102327","DOIUrl":"10.1016/j.swevo.2026.102327","url":null,"abstract":"<div><div>Dynamic distributed production scheduling poses significant challenges to modern manufacturing systems due to environmental complexity and the need to respond to real-time disruptions. This paper studies the dynamic distributed hybrid flow shop problem (DDHFSP) with simultaneous dynamic job arrivals and machine breakdowns, aiming to minimize makespan. To address this problem, a learnable iterated greedy algorithm (LIG) is proposed within the classical IG framework. LIG is designed to learn to learn how to adaptively configure its operational components by extracting knowledge from search trajectories, thereby effectively handling dynamic scheduling environments. Based on the IG structure, four rescheduling strategies for dynamic events, four destruction-reconstruction strategies, and four local search strategies are designed. The algorithm employs a long short-term memory (LSTM) network to extract temporally dependent production features, and adopts proximal policy optimization (PPO) to build an agent that adaptively selects the most promising combination of strategies according to the current state for execution within the IG framework. Experimental results demonstrate that LIG outperforms several state-of-the-art metaheuristic and reinforcement learning-based methods across various problem scales and dynamic scenarios, exhibiting faster convergence, better solution quality, higher stability, and stronger generalization capability. Ultimately, the proposed learnable mechanism establishes a new paradigm for tackling dynamic disturbances in DDHFSP.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"102 ","pages":"Article 102327"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147423706","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

Risk-aware hierarchical multi-objective planning of RSU deployment and UAV scheduling for urban V2X networks 城市V2X网络RSU部署和无人机调度的风险感知分层多目标规划

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-03-06 DOI: 10.1016/j.swevo.2026.102348

Weian Guo , Yao Xiao , Shi Cheng , Hui Lu

Urban Vehicle-to-Everything (V2X) services demand infrastructure plans that remain effective under pronounced spatiotemporal demand shifts and operational uncertainty. Existing studies often optimize roadside unit (RSU) deployment or unmanned aerial vehicle (UAV) operations in isolation, rely on scalar objectives, and simplify feasibility constraints, which limits their value for actionable planning. This paper develops a hierarchical multi-objective planning framework that jointly decides (i) where and how many RSUs to deploy and (ii) how to schedule a UAV fleet over time to complement the static backbone. The upper level searches for non-dominated infrastructure configurations that balance total cost, spatiotemporal coverage, and robustness, while the lower level evaluates each configuration through a state-aware greedy dispatch policy that explicitly models UAV operational modes and battery/charging dynamics. Robustness is optimized via a weighted composite of redundancy, coverage variability, and mission balance, and scenario-based Conditional Value-at-Risk (CVaR) is reported as a secondary tail-risk indicator under demand perturbations and component failures. Experiments on realistic urban traffic data, together with policy validation on reduced instances, baseline comparisons, and multi-seed stability analysis, demonstrate that the framework yields diverse and interpretable planning trade-offs.

城市车辆到一切（V2X）服务需要在明显的时空需求变化和运营不确定性下保持有效的基础设施规划。现有的研究通常是孤立地优化路边单元（RSU）部署或无人机（UAV）操作，依赖标量目标，简化可行性约束，这限制了它们对可操作规划的价值。本文开发了一个分层多目标规划框架，该框架共同决定(i)在何处以及部署多少个rsu，以及（ii）如何随着时间的推移调度无人机机队以补充静态主干。上层搜索平衡总成本、时空覆盖和鲁棒性的非主导基础设施配置，而下层通过状态感知贪婪调度策略评估每个配置，该策略明确建模无人机操作模式和电池/充电动态。鲁棒性通过冗余、覆盖可变性和任务平衡的加权组合来优化，基于场景的条件风险值（CVaR）被报告为需求扰动和组件故障下的次要尾部风险指标。对现实城市交通数据的实验，以及对减少实例的策略验证、基线比较和多种子稳定性分析表明，该框架产生了多种可解释的规划权衡。

{"title":"Risk-aware hierarchical multi-objective planning of RSU deployment and UAV scheduling for urban V2X networks","authors":"Weian Guo , Yao Xiao , Shi Cheng , Hui Lu","doi":"10.1016/j.swevo.2026.102348","DOIUrl":"10.1016/j.swevo.2026.102348","url":null,"abstract":"<div><div>Urban Vehicle-to-Everything (V2X) services demand infrastructure plans that remain effective under pronounced spatiotemporal demand shifts and operational uncertainty. Existing studies often optimize roadside unit (RSU) deployment or unmanned aerial vehicle (UAV) operations in isolation, rely on scalar objectives, and simplify feasibility constraints, which limits their value for actionable planning. This paper develops a hierarchical multi-objective planning framework that jointly decides (i) where and how many RSUs to deploy and (ii) how to schedule a UAV fleet over time to complement the static backbone. The upper level searches for non-dominated infrastructure configurations that balance total cost, spatiotemporal coverage, and robustness, while the lower level evaluates each configuration through a state-aware greedy dispatch policy that explicitly models UAV operational modes and battery/charging dynamics. Robustness is optimized via a weighted composite of redundancy, coverage variability, and mission balance, and scenario-based Conditional Value-at-Risk (CVaR) is reported as a secondary tail-risk indicator under demand perturbations and component failures. Experiments on realistic urban traffic data, together with policy validation on reduced instances, baseline comparisons, and multi-seed stability analysis, demonstrate that the framework yields diverse and interpretable planning trade-offs.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"103 ","pages":"Article 102348"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147426628","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

Heterogeneous evolutionary reinforcement learning with mixed attention and diffusion model for dynamic seru formation 基于混合注意和扩散模型的异质进化强化学习动态血清形成

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-03-11 DOI: 10.1016/j.swevo.2026.102353

Hongbo Jin, Dongni Li, Yuqing Jin, Yao Lyu, Yaoxin Zhang, Zihao Gao

The increasing uncertainty in manufacturing presents significant challenges to conventional production systems. Seru production systems (SPSs) are emerged as an effective approach for handling fluctuating product varieties and volumes. In dynamic scenarios, orders with uncertain demands arrive sequentially, requiring real-time decisions for seru formation. This study addresses the dynamic seru formation problem with the objective of maximizing revenue. To tackle this challenge, a hybrid algorithm named Hetero-ERL is proposed, integrating evolutionary exploration, reinforcement learning (RL) refinement, diffusion modeling, and a worker–order mixed attention mechanism. Evolutionary exploration promotes solution diversity, diffusion modeling captures multimodal formation policies, and RL refinement fine-tunes policies. The mixed attention mechanism enhances the algorithm’s ability to learn both intra-type patterns and cross-type worker–order relationships. The algorithm is evaluated under several dynamic uncertainty settings and compared with three categories of baselines: evolutionary reinforcement learning (ERL) algorithms, deep reinforcement learning (DRL) algorithms, and an approximation method. Results show that Hetero-ERL achieves performance gains ranging from 16.18% to 56.83% over ERL algorithms, 6.84% to 29.91% over DRL algorithms, and 69.47% to 78.45% over the approximation method. These findings confirm that Hetero-ERL provides strong generalization in dynamic seru formation and offers practical potential for uncertain manufacturing scenarios.

制造业日益增加的不确定性对传统生产系统提出了重大挑战。Seru生产系统（SPSs）作为处理波动的产品品种和数量的有效方法而出现。在动态场景中，需求不确定的订单顺序到达，需要实时决策来形成服务。本研究以收益最大化为目标，研究动态的血清形成问题。为了应对这一挑战，提出了一种名为Hetero-ERL的混合算法，该算法集成了进化探索、强化学习（RL）改进、扩散建模和工作顺序混合注意机制。进化探索促进了解决方案的多样性，扩散建模捕获了多模态形成策略，RL细化微调了策略。混合注意机制增强了算法学习类型内模式和跨类型工序关系的能力。该算法在几种动态不确定性设置下进行了评估，并与三类基线进行了比较：进化强化学习（ERL）算法、深度强化学习（DRL）算法和近似方法。结果表明，与ERL算法相比，Hetero-ERL算法的性能提高幅度为16.18% ~ 56.83%，与DRL算法相比，性能提高幅度为6.84% ~ 29.91%，与近似方法相比，性能提高幅度为69.47% ~ 78.45%。这些发现证实了Hetero-ERL在动态血清形成中具有很强的通用性，并为不确定的制造场景提供了实用潜力。

{"title":"Heterogeneous evolutionary reinforcement learning with mixed attention and diffusion model for dynamic seru formation","authors":"Hongbo Jin, Dongni Li, Yuqing Jin, Yao Lyu, Yaoxin Zhang, Zihao Gao","doi":"10.1016/j.swevo.2026.102353","DOIUrl":"10.1016/j.swevo.2026.102353","url":null,"abstract":"<div><div>The increasing uncertainty in manufacturing presents significant challenges to conventional production systems. <em>Seru</em> production systems (SPSs) are emerged as an effective approach for handling fluctuating product varieties and volumes. In dynamic scenarios, orders with uncertain demands arrive sequentially, requiring real-time decisions for <em>seru</em> formation. This study addresses the dynamic <em>seru</em> formation problem with the objective of maximizing revenue. To tackle this challenge, a hybrid algorithm named Hetero-ERL is proposed, integrating evolutionary exploration, reinforcement learning (RL) refinement, diffusion modeling, and a worker–order mixed attention mechanism. Evolutionary exploration promotes solution diversity, diffusion modeling captures multimodal formation policies, and RL refinement fine-tunes policies. The mixed attention mechanism enhances the algorithm’s ability to learn both intra-type patterns and cross-type worker–order relationships. The algorithm is evaluated under several dynamic uncertainty settings and compared with three categories of baselines: evolutionary reinforcement learning (ERL) algorithms, deep reinforcement learning (DRL) algorithms, and an approximation method. Results show that Hetero-ERL achieves performance gains ranging from 16.18% to 56.83% over ERL algorithms, 6.84% to 29.91% over DRL algorithms, and 69.47% to 78.45% over the approximation method. These findings confirm that Hetero-ERL provides strong generalization in dynamic <em>seru</em> formation and offers practical potential for uncertain manufacturing scenarios.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"103 ","pages":"Article 102353"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147426630","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

Moment adaptation evolution strategy based on an aging leader and learned challenger mechanism 基于老化领导者和习得挑战者机制的时刻适应进化策略

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-03-12 DOI: 10.1016/j.swevo.2026.102351

Qiming Liu , Chenxu Yang , Tao Li

In this study, a moment adaptation evolution strategy based on an aging leader and learned challenger mechanism (MoA-ES-ALLC) is proposed to enhance the optimization performance of the covariance matrix adaptation evolution strategy (CMA-ES). In contrast to the CMA-ES, the proposed algorithm defines a well-performing individual as the leader to retain advantageous information. It achieves the moment adaptation by guiding the mean using an influence coefficient and modifying the covariance matrix with a new meaningful evolution path, thereby accelerating convergence. Learned challengers, generated by a dual-center learning strategy, increase population diversity. A restart strategy, coupled with learned challengers, helps avoid local optima. To further demonstrate the generality of these strategies, they are also integrated into the matrix adaptation evolution strategy (MA-ES), resulting in a second algorithm termed moment MA-ES based on an aging leader and learned challenger mechanism (MoMA-ES-ALLC). Moreover, numerical examples on various landscapes demonstrate that both proposed algorithms perform similarly and outperform the variants of CMA-ES in terms of convergence speed and search accuracy. Subsequently, a comprehensive evaluation is conducted focusing on the MoA-ES-ALLC. The comparison results on the CEC-2022 benchmarks show that it achieves better results than several state-of-the-art algorithms. Finally, the practicability of the proposed MoA-ES-ALLC is verified through applications to eight engineering problems and trajectory planning for a 6-DOF robot manipulator. The proposed MoA-ES-ALLC presents an efficient solution to the critical challenges of convergence speed and local optima in evolutionary computation, making it particularly effective for complex optimization problems.

为了提高协方差矩阵适应进化策略（CMA-ES）的优化性能，提出了一种基于老化领导者和学习挑战者机制的时刻适应进化策略（MoA-ES-ALLC）。与CMA-ES相比，该算法将表现良好的个体定义为领导者，以保留优势信息。利用影响系数引导均值，用新的有意义的进化路径修正协方差矩阵，实现矩适应，加快收敛速度。双中心学习策略产生的学习型挑战者增加了种群多样性。一个重新开始的策略，加上有经验的挑战者，有助于避免局部最优。为了进一步证明这些策略的通用性，它们还被集成到矩阵适应进化策略（MA-ES）中，从而产生了基于老化领导者和学习挑战者机制（MoMA-ES-ALLC）的第二种算法，称为矩MA-ES。此外，在不同场景下的数值算例表明，两种算法在收敛速度和搜索精度方面都优于CMA-ES的变体。随后，对MoA-ES-ALLC进行了综合评价。在CEC-2022基准测试上的对比结果表明，该算法比几种最先进的算法取得了更好的效果。最后，通过对6自由度机器人机械臂的8个工程问题和轨迹规划的应用，验证了所提出的MoA-ES-ALLC的实用性。提出的MoA-ES-ALLC算法有效地解决了进化计算中收敛速度和局部最优的关键问题，对复杂的优化问题尤其有效。

{"title":"Moment adaptation evolution strategy based on an aging leader and learned challenger mechanism","authors":"Qiming Liu , Chenxu Yang , Tao Li","doi":"10.1016/j.swevo.2026.102351","DOIUrl":"10.1016/j.swevo.2026.102351","url":null,"abstract":"<div><div>In this study, a moment adaptation evolution strategy based on an aging leader and learned challenger mechanism (MoA-ES-ALLC) is proposed to enhance the optimization performance of the covariance matrix adaptation evolution strategy (CMA-ES). In contrast to the CMA-ES, the proposed algorithm defines a well-performing individual as the leader to retain advantageous information. It achieves the moment adaptation by guiding the mean using an influence coefficient and modifying the covariance matrix with a new meaningful evolution path, thereby accelerating convergence. Learned challengers, generated by a dual-center learning strategy, increase population diversity. A restart strategy, coupled with learned challengers, helps avoid local optima. To further demonstrate the generality of these strategies, they are also integrated into the matrix adaptation evolution strategy (MA-ES), resulting in a second algorithm termed moment MA-ES based on an aging leader and learned challenger mechanism (MoMA-ES-ALLC). Moreover, numerical examples on various landscapes demonstrate that both proposed algorithms perform similarly and outperform the variants of CMA-ES in terms of convergence speed and search accuracy. Subsequently, a comprehensive evaluation is conducted focusing on the MoA-ES-ALLC. The comparison results on the CEC-2022 benchmarks show that it achieves better results than several state-of-the-art algorithms. Finally, the practicability of the proposed MoA-ES-ALLC is verified through applications to eight engineering problems and trajectory planning for a 6-DOF robot manipulator. The proposed MoA-ES-ALLC presents an efficient solution to the critical challenges of convergence speed and local optima in evolutionary computation, making it particularly effective for complex optimization problems.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"103 ","pages":"Article 102351"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147451526","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

Adversarial autoencoder enhanced evolutionary guidance for multi-objective multi-task optimization 对抗性自编码器增强的多目标多任务优化进化制导

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-02-19 DOI: 10.1016/j.swevo.2026.102335

Xia Wang , Hongwei Ge , Yaqing Hou , Liang Sun , Bin Li

Multi-objective multitasking optimization (MOMT) problems are of significant importance in practical applications, particularly in complex engineering fields. These problems require simultaneously optimizing multiple objectives while effectively managing the interactions and information sharing between tasks. However, existing MOMT evolutionary algorithms often face challenges such as insufficient knowledge transfer, negative transfer issues, and difficulties in maintaining the quality and diversity of solutions. To address these challenges, this paper proposes an adversarial autoencoder (AAE) enhanced evolutionary guidance algorithm for multi-objective multi-task optimization (EMT-AAE). First, a new adversarial autoencoder model is proposed to learn the promising evolutionary information and achieve positive transfer. The evolutionary dynamics from source to target task are dynamically implemented by periodically training the model structure. Next, a new AAE-guided offspring reproduction strategy is proposed to improve the quality of generated solutions. Finally, an adaptive environmental selection strategy is proposed to balance solution diversity and convergence. To validate the effectiveness of the proposed algorithm, experiments are conducted on three multi-objective multi-task benchmark test suites. Compared with other state-of-the-art algorithms, EMT-AAE performs excellently in addressing both classic and complex MOMT optimization problems. Additionally, it shows scalability and practicality in three real-world problems.

多目标多任务优化（MOMT）问题在实际应用中具有重要意义，特别是在复杂的工程领域。这些问题需要同时优化多个目标，同时有效地管理任务之间的交互和信息共享。然而，现有的MOMT进化算法经常面临知识转移不足、负迁移问题以及难以保持解的质量和多样性等挑战。为了解决这些问题，本文提出了一种针对多目标多任务优化（EMT-AAE）的对抗性自编码器（AAE）增强进化制导算法。首先，提出了一种新的对抗性自编码器模型来学习有希望的进化信息并实现正迁移。通过对模型结构的周期性训练，动态实现了从源任务到目标任务的演化动态。其次，提出了一种新的aae引导子代繁殖策略，以提高生成解的质量。最后，提出了一种平衡解的多样性和收敛性的自适应环境选择策略。为了验证该算法的有效性，在三个多目标多任务基准测试套件上进行了实验。与其他最先进的算法相比，EMT-AAE在解决经典和复杂的MOMT优化问题方面都表现出色。此外，它在三个实际问题中显示了可伸缩性和实用性。

{"title":"Adversarial autoencoder enhanced evolutionary guidance for multi-objective multi-task optimization","authors":"Xia Wang , Hongwei Ge , Yaqing Hou , Liang Sun , Bin Li","doi":"10.1016/j.swevo.2026.102335","DOIUrl":"10.1016/j.swevo.2026.102335","url":null,"abstract":"<div><div>Multi-objective multitasking optimization (MOMT) problems are of significant importance in practical applications, particularly in complex engineering fields. These problems require simultaneously optimizing multiple objectives while effectively managing the interactions and information sharing between tasks. However, existing MOMT evolutionary algorithms often face challenges such as insufficient knowledge transfer, negative transfer issues, and difficulties in maintaining the quality and diversity of solutions. To address these challenges, this paper proposes an adversarial autoencoder (AAE) enhanced evolutionary guidance algorithm for multi-objective multi-task optimization (EMT-AAE). First, a new adversarial autoencoder model is proposed to learn the promising evolutionary information and achieve positive transfer. The evolutionary dynamics from source to target task are dynamically implemented by periodically training the model structure. Next, a new AAE-guided offspring reproduction strategy is proposed to improve the quality of generated solutions. Finally, an adaptive environmental selection strategy is proposed to balance solution diversity and convergence. To validate the effectiveness of the proposed algorithm, experiments are conducted on three multi-objective multi-task benchmark test suites. Compared with other state-of-the-art algorithms, EMT-AAE performs excellently in addressing both classic and complex MOMT optimization problems. Additionally, it shows scalability and practicality in three real-world problems.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"102 ","pages":"Article 102335"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147423703","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

Integrated scheduling of cargo vessels, research vessels, and marine experiments in multifunctional ports using Q-learning enhanced PSO 基于q -学习增强粒子群算法的多功能港口货船、科考船和海洋实验综合调度

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-02-12 DOI: 10.1016/j.swevo.2026.102315

Xiang-Yang Li , Zhong-Yi Yang , Ming-Wei Li , Wei-Chiang Hong

Multifunctional ports integrating cargo and research operations (CRPs) face unprecedented scheduling complexities due to spatiotemporal conflicts among cargo vessels, research vessels, and marine experiments. To resolve the aforementioned resource conflicts, this study proposes a hierarchical spatiotemporal coordination framework that establishes differentiated operational zones and experiment time windows. Then, a multi-objective joint scheduling model (BCAEA) is formulated to integrate berth allocation, quay crane assignment, and experiment arrangement, simultaneously minimizing shipowners' and operational costs while maximizing experimental efficiency. To solve this large-scale optimization problem, an enhanced particle swarm optimization algorithm (QLEPSO) is developed, incorporating a position update strategy pool, Q-learning-based strategy selection, and adaptive parameter control. Numerical experiments using real operational data from Chinese CRPs demonstrate that QLEPSO outperforms standard PSO by 47.17% in solution quality for large-scale problems. Moreover, the proposed BCAEA_QLEPSO method generates high-quality allocation schemes for instances involving 90 vessels and 18 experiments within 1 minute, validating the effectiveness of integrating reinforcement learning with swarm intelligence for complex port scheduling.

由于货船、科考船和海洋实验船之间的时空冲突，综合货运和科研作业（CRPs）的多功能港口面临着前所未有的调度复杂性。为了解决上述资源冲突，本研究提出了一个分层的时空协调框架，该框架建立了差异化的操作区域和实验时间窗口。然后，将泊位分配、岸机分配和实验安排整合在一起，建立多目标联合调度模型（BCAEA），实现船东成本和运营成本最小化，实验效率最大化。为了解决这一大规模优化问题，提出了一种基于位置更新策略池、基于q学习的策略选择和自适应参数控制的增强粒子群优化算法（QLEPSO）。利用中国CRPs的实际运行数据进行的数值实验表明，QLEPSO在解决大规模问题方面的质量比标准PSO高47.17%。此外，提出的BCAEA_QLEPSO方法在1分钟内生成了涉及90艘船舶和18个实验的高质量分配方案，验证了将强化学习与群体智能相结合用于复杂港口调度的有效性。

{"title":"Integrated scheduling of cargo vessels, research vessels, and marine experiments in multifunctional ports using Q-learning enhanced PSO","authors":"Xiang-Yang Li , Zhong-Yi Yang , Ming-Wei Li , Wei-Chiang Hong","doi":"10.1016/j.swevo.2026.102315","DOIUrl":"10.1016/j.swevo.2026.102315","url":null,"abstract":"<div><div>Multifunctional ports integrating cargo and research operations (CRPs) face unprecedented scheduling complexities due to spatiotemporal conflicts among cargo vessels, research vessels, and marine experiments. To resolve the aforementioned resource conflicts, this study proposes a hierarchical spatiotemporal coordination framework that establishes differentiated operational zones and experiment time windows. Then, a multi-objective joint scheduling model (BCAEA) is formulated to integrate berth allocation, quay crane assignment, and experiment arrangement, simultaneously minimizing shipowners' and operational costs while maximizing experimental efficiency. To solve this large-scale optimization problem, an enhanced particle swarm optimization algorithm (QLEPSO) is developed, incorporating a position update strategy pool, Q-learning-based strategy selection, and adaptive parameter control. Numerical experiments using real operational data from Chinese CRPs demonstrate that QLEPSO outperforms standard PSO by 47.17% in solution quality for large-scale problems. Moreover, the proposed BCAEA_QLEPSO method generates high-quality allocation schemes for instances involving 90 vessels and 18 experiments within 1 minute, validating the effectiveness of integrating reinforcement learning with swarm intelligence for complex port scheduling.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"102 ","pages":"Article 102315"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154423","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 competition-driven two-phase evolutionary algorithm for constrained multi-objective optimization 约束多目标优化的竞争驱动两阶段进化算法

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-02-12 DOI: 10.1016/j.swevo.2026.102322

Shengwei Wang, Mengduo Yu, Chenhao Yuan, Keda Chen, Aobo Guo, Hui Duan

The core challenge of constrained multi-objective optimization problems (CMOPs) lies in achieving an effective balance among feasibility, convergence, and diversity. The introduction of constraints not only restricts the feasible solution space but also increases the difficulty of identifying the Pareto-optimal solution set. To address these challenges of balancing feasibility, convergence, and diversity under complex constraints, this paper proposes a competition-driven two-stage evolutionary algorithm, termed AAPEA. In AAPEA, the evolutionary process is divided into two stages. In the first stage, the main population is selected according to the constraint dominance criterion, guiding solutions to gradually converge toward the feasible solution space. The auxiliary population evolves independently through non-dominated sorting to enhance solution diversity, expand the feasible region, and provide effective search guidance for the main population. In order to reduce computational cost, an adaptive size adjustment mechanism is introduced in the auxiliary population, which dynamically adjusts the population size based on the evolutionary state. In addition, a CSO based competition-driven search operator is incorporated to promote convergence toward high-quality solution regions through competitive learning and updating mechanisms. In the second stage, the auxiliary population is removed, thus only the main population is retained. A feasibility-driven environmental selection strategy is then employed to further enhance convergence within the feasible solution space. Experimental results on three benchmark test suites and three real-world engineering problems show that AAPEA demonstrates superior performance compared with eight state-of-the-art CMOEAs.

约束多目标优化问题的核心挑战在于如何在可行性、收敛性和多样性之间取得有效的平衡。约束条件的引入不仅限制了可行解空间，而且增加了识别pareto最优解集的难度。为了解决在复杂约束条件下平衡可行性、收敛性和多样性的挑战，本文提出了一种竞争驱动的两阶段进化算法，称为AAPEA。在AAPEA中，进化过程分为两个阶段。第一阶段，根据约束优势准则选择主要种群，引导解逐步向可行解空间收敛。辅助种群通过非优势排序独立进化，增强了解的多样性，扩大了可行区域，为主种群提供了有效的搜索指导。为了降低计算成本，在辅助种群中引入自适应大小调整机制，根据进化状态动态调整种群大小。此外，还引入了基于竞争驱动的CSO搜索算子，通过竞争学习和更新机制促进算法向高质量解域收敛。在第二阶段，去除辅助种群，只保留主要种群。然后采用可行性驱动的环境选择策略，进一步增强可行性解空间内的收敛性。在3个基准测试套件和3个实际工程问题上的实验结果表明，与8个最先进的cmoea相比，AAPEA具有优越的性能。

{"title":"A competition-driven two-phase evolutionary algorithm for constrained multi-objective optimization","authors":"Shengwei Wang, Mengduo Yu, Chenhao Yuan, Keda Chen, Aobo Guo, Hui Duan","doi":"10.1016/j.swevo.2026.102322","DOIUrl":"10.1016/j.swevo.2026.102322","url":null,"abstract":"<div><div>The core challenge of constrained multi-objective optimization problems (CMOPs) lies in achieving an effective balance among feasibility, convergence, and diversity. The introduction of constraints not only restricts the feasible solution space but also increases the difficulty of identifying the Pareto-optimal solution set. To address these challenges of balancing feasibility, convergence, and diversity under complex constraints, this paper proposes a competition-driven two-stage evolutionary algorithm, termed AAPEA. In AAPEA, the evolutionary process is divided into two stages. In the first stage, the main population is selected according to the constraint dominance criterion, guiding solutions to gradually converge toward the feasible solution space. The auxiliary population evolves independently through non-dominated sorting to enhance solution diversity, expand the feasible region, and provide effective search guidance for the main population. In order to reduce computational cost, an adaptive size adjustment mechanism is introduced in the auxiliary population, which dynamically adjusts the population size based on the evolutionary state. In addition, a CSO based competition-driven search operator is incorporated to promote convergence toward high-quality solution regions through competitive learning and updating mechanisms. In the second stage, the auxiliary population is removed, thus only the main population is retained. A feasibility-driven environmental selection strategy is then employed to further enhance convergence within the feasible solution space. Experimental results on three benchmark test suites and three real-world engineering problems show that AAPEA demonstrates superior performance compared with eight state-of-the-art CMOEAs.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"102 ","pages":"Article 102322"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154424","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 surrogate-based strategy for accelerating convergence speed when solving expensive unconstrained Multi-Objective Optimisation Problems 基于自适应代理的高代价无约束多目标优化算法

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

Swarm and Evolutionary Computation

Pub Date : 2026-03-01 Epub Date: 2026-02-24 DOI: 10.1016/j.swevo.2026.102311

Tiwonge Msulira Banda, Alexandru-Ciprian Zăvoianu

Multi-Objective Evolutionary Algorithms (MOEAs) have proven effective at solving Multi-Objective Optimisation Problems (MOOPs). However, their performance can be significantly hindered when applied to computationally intensive industrial problems. To address this limitation, we propose an adaptive surrogate modelling approach designed to accelerate the early-stage convergence speed of state-of-the-art MOEAs. This is important because it ensures that a solver can identify optimal or near-optimal solutions with relatively few fitness function evaluations, thereby saving both time and computational resources. Our method employs a two-loop architecture. The outer loop runs a (baseline) host MOEA which carries out true fitness evaluations. The inner loop contains an Adaptive Accelerator that leverages data-driven machine learning (ML) surrogate models to approximate fitness functions. Integrated with NSGA-II and MOEA/D, our approach was tested on 31 widely known benchmark problems and a real-world North Sea fish abundance modelling case study. The results demonstrate that by incorporating Gaussian Process Regression, one-dimensional Convolutional Neural Networks, and Random Forest Regression, our proposed approach significantly accelerates the convergence speed of MOEAs in the early phases of optimisation.

多目标进化算法（moea）在解决多目标优化问题（MOOPs）方面已被证明是有效的。然而，当应用于计算密集型工业问题时，它们的性能会受到很大的阻碍。为了解决这一限制，我们提出了一种自适应代理建模方法，旨在加快最先进的moea的早期收敛速度。这很重要，因为它确保求解器可以用相对较少的适应度函数评估来识别最优或接近最优的解决方案，从而节省时间和计算资源。我们的方法采用了双循环架构。外部循环运行一个（基线）主机MOEA，它执行真正的适应度评估。内环包含一个自适应加速器，它利用数据驱动的机器学习（ML）代理模型来近似适应度函数。结合NSGA-II和MOEA/D，我们的方法在31个广为人知的基准问题和一个真实的北海鱼类丰度建模案例研究中进行了测试。结果表明，通过结合高斯过程回归、一维卷积神经网络和随机森林回归，我们提出的方法在优化的早期阶段显著加快了moea的收敛速度。

{"title":"Adaptive surrogate-based strategy for accelerating convergence speed when solving expensive unconstrained Multi-Objective Optimisation Problems","authors":"Tiwonge Msulira Banda, Alexandru-Ciprian Zăvoianu","doi":"10.1016/j.swevo.2026.102311","DOIUrl":"10.1016/j.swevo.2026.102311","url":null,"abstract":"<div><div>Multi-Objective Evolutionary Algorithms (MOEAs) have proven effective at solving Multi-Objective Optimisation Problems (MOOPs). However, their performance can be significantly hindered when applied to computationally intensive industrial problems. To address this limitation, we propose an adaptive surrogate modelling approach designed to accelerate the early-stage convergence speed of state-of-the-art MOEAs. This is important because it ensures that a solver can identify optimal or near-optimal solutions with relatively few fitness function evaluations, thereby saving both time and computational resources. Our method employs a two-loop architecture. The outer loop runs a (baseline) host MOEA which carries out true fitness evaluations. The inner loop contains an Adaptive Accelerator that leverages data-driven machine learning (ML) surrogate models to approximate fitness functions. Integrated with NSGA-II and MOEA/D, our approach was tested on 31 widely known benchmark problems and a real-world North Sea fish abundance modelling case study. The results demonstrate that by incorporating Gaussian Process Regression, one-dimensional Convolutional Neural Networks, and Random Forest Regression, our proposed approach significantly accelerates the convergence speed of MOEAs in the early phases of optimisation.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"102 ","pages":"Article 102311"},"PeriodicalIF":8.5,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147423688","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