Swarm and Evolutionary Computation最新文献_第2页

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-21 DOI: 10.1016/j.swevo.2026.102297

Yanlin Wu , Xinyu Zhou , Hui Wang , Jia Zhao

This study addresses the computational inefficiency of traditional evolutionary multitasking algorithms in solving many-task capacitated vehicle routing problems (CVRPs) by proposing a knowledge transfer optimization framework driven by dynamic similarity evaluation. Existing approaches predominantly rely on explicit knowledge transfer mechanisms based on transfer matrices, whose computational complexity escalates exponentially with increasing task quantities. To overcome this limitation, a three-phase optimization framework is developed: (1) Common features across multiple tasks are extracted through feature space mapping techniques, establishing a quantifiable similarity evaluation model; (2) An adaptive knowledge transfer feedback system is implemented, integrating a transfer-effect monitoring mechanism and dynamic weight adjustment strategy to ensure real-time optimization of knowledge source quality; (3) A hybrid crossover operation architecture is designed, combining elite solution transfer with local route optimization to reduce computational overhead. Comparative experiments conducted on a comprehensive simulation dataset (containing 99 many-task CVRP instances) and real-world logistics scenarios demonstrate the algorithm’s superior performance across multiple metrics.

本文提出了一种基于动态相似性评价的知识转移优化框架，解决了传统进化多任务算法在求解多任务车辆路径问题（CVRPs）时计算效率低下的问题。现有方法主要依赖于基于转移矩阵的显式知识转移机制，其计算复杂度随着任务数量的增加呈指数级增长。为克服这一局限性，提出了一种三阶段优化框架：(1)通过特征空间映射技术提取多任务间的共同特征，建立可量化的相似性评价模型；(2)构建自适应知识转移反馈系统，整合转移效果监测机制和动态权重调整策略，确保知识源质量实时优化；(3)设计了一种混合交叉操作架构，将精英解传递与局部路径优化相结合，减少了计算开销。在综合模拟数据集（包含99个多任务CVRP实例）和现实物流场景上进行的比较实验证明了该算法在多个指标上的卓越性能。

{"title":"Similarity-driven knowledge transfer algorithm for many-task capacitated vehicle routing problem","authors":"Yanlin Wu , Xinyu Zhou , Hui Wang , Jia Zhao","doi":"10.1016/j.swevo.2026.102297","DOIUrl":"10.1016/j.swevo.2026.102297","url":null,"abstract":"<div><div>This study addresses the computational inefficiency of traditional evolutionary multitasking algorithms in solving many-task capacitated vehicle routing problems (CVRPs) by proposing a knowledge transfer optimization framework driven by dynamic similarity evaluation. Existing approaches predominantly rely on explicit knowledge transfer mechanisms based on transfer matrices, whose computational complexity escalates exponentially with increasing task quantities. To overcome this limitation, a three-phase optimization framework is developed: (1) Common features across multiple tasks are extracted through feature space mapping techniques, establishing a quantifiable similarity evaluation model; (2) An adaptive knowledge transfer feedback system is implemented, integrating a transfer-effect monitoring mechanism and dynamic weight adjustment strategy to ensure real-time optimization of knowledge source quality; (3) A hybrid crossover operation architecture is designed, combining elite solution transfer with local route optimization to reduce computational overhead. Comparative experiments conducted on a comprehensive simulation dataset (containing 99 many-task CVRP instances) and real-world logistics scenarios demonstrate the algorithm’s superior performance across multiple metrics.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102297"},"PeriodicalIF":8.5,"publicationDate":"2026-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146039030","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 fitness-based two roles adaptive inertia weight particle swarm optimization 一种基于适应度的双角色自适应惯性权重粒子群优化方法

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-20 DOI: 10.1016/j.swevo.2026.102291

Jun Guan , Shuanghui Ye , Wenjun Yi

Particle Swarm Optimization (PSO) has been widely applied to practical problems. Similar to other evolutionary algorithms, PSO is prone to premature convergence and local optima entrapment. To achieve a dynamic balance between exploration and exploitation, this paper proposes a fitness-based two roles adaptive inertia weight particle swarm optimization (FAIWPSO). The proposed algorithm divides the population into elite and ordinary subgroups according to individual fitness values, and adaptively reduces the number of elite particles as the iteration proceeds, thereby realizing a smooth transition from global exploration to local exploitation. The elite subgroup identifies and guides the population toward promising regions using neighborhood best information, while maintaining population diversity through non-uniform mutation. Guided by the elite subpopulation, the ordinary subpopulation adopts a weighted learning strategy to intensively exploit promising regions. Furthermore, to achieve a better balance between exploration and exploitation, a nonlinear adaptive inertia weight strategy based on both population evolution state and individual differences is introduced. Additionally, a dimension-adaptive Gaussian mutation strategy is developed, which mutates different dimensions of the global best solution depending on the evolutionary stage, enhancing the ability to escape local optima. To evaluate the effectiveness of FAIWPSO, comprehensive experiments were conducted, demonstrating that the proposed strategies substantially enhance the algorithm’s performance. On the CEC2014 test suite, FAIWPSO outperforms seven widely used PSO variants in terms of solution accuracy and stability. On the CEC2018 test suite, its overall performance surpasses that of APGSK-IMODE and MadDE.

粒子群算法在实际问题中得到了广泛的应用。与其他进化算法相似，粒子群算法容易出现过早收敛和局部最优困陷的问题。为了实现勘探与开发的动态平衡，提出了一种基于适应度的双角色自适应惯性权重粒子群优化算法。该算法根据个体适应度值将种群划分为精英子群和普通子群，并随着迭代的进行自适应减少精英粒子的数量，实现了从全局探索到局部开发的平稳过渡。精英亚群利用邻域最优信息识别并引导种群走向有发展前景的区域，同时通过非均匀突变保持种群多样性。在精英亚种群的指导下，普通亚种群采用加权学习策略，集中开发有潜力的区域。在此基础上，引入了一种基于种群进化状态和个体差异的非线性自适应惯性权重策略，以更好地平衡种群的探索和开发。此外，提出了一种维数自适应的高斯突变策略，根据进化阶段对全局最优解的不同维数进行突变，增强了逃避局部最优的能力。为了评估FAIWPSO的有效性，进行了全面的实验，表明所提出的策略大大提高了算法的性能。在CEC2014测试套件中，FAIWPSO在解决方案精度和稳定性方面优于7种广泛使用的PSO变体。在CEC2018测试套件上，其整体性能超过了APGSK-IMODE和MadDE。

{"title":"A fitness-based two roles adaptive inertia weight particle swarm optimization","authors":"Jun Guan , Shuanghui Ye , Wenjun Yi","doi":"10.1016/j.swevo.2026.102291","DOIUrl":"10.1016/j.swevo.2026.102291","url":null,"abstract":"<div><div>Particle Swarm Optimization (PSO) has been widely applied to practical problems. Similar to other evolutionary algorithms, PSO is prone to premature convergence and local optima entrapment. To achieve a dynamic balance between exploration and exploitation, this paper proposes a fitness-based two roles adaptive inertia weight particle swarm optimization (FAIWPSO). The proposed algorithm divides the population into elite and ordinary subgroups according to individual fitness values, and adaptively reduces the number of elite particles as the iteration proceeds, thereby realizing a smooth transition from global exploration to local exploitation. The elite subgroup identifies and guides the population toward promising regions using neighborhood best information, while maintaining population diversity through non-uniform mutation. Guided by the elite subpopulation, the ordinary subpopulation adopts a weighted learning strategy to intensively exploit promising regions. Furthermore, to achieve a better balance between exploration and exploitation, a nonlinear adaptive inertia weight strategy based on both population evolution state and individual differences is introduced. Additionally, a dimension-adaptive Gaussian mutation strategy is developed, which mutates different dimensions of the global best solution depending on the evolutionary stage, enhancing the ability to escape local optima. To evaluate the effectiveness of FAIWPSO, comprehensive experiments were conducted, demonstrating that the proposed strategies substantially enhance the algorithm’s performance. On the CEC2014 test suite, FAIWPSO outperforms seven widely used PSO variants in terms of solution accuracy and stability. On the CEC2018 test suite, its overall performance surpasses that of APGSK-IMODE and MadDE.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102291"},"PeriodicalIF":8.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146039025","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

Dual-self-learning knowledge-based local search approach for lot-streaming flexible job shop scheduling with asynchronous assembly 基于双自学习知识的异步装配批量流柔性作业车间调度方法

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-20 DOI: 10.1016/j.swevo.2026.102283

Gan Jiang , Qiuhua Tang , Lixin Cheng , Zikai Zhang

To achieve customized production of terminal products, industrial enterprises are gradually adopting a new two-stage flexible production system, where the first stage involves the flexible manufacturing of components and the second stage involves the asynchronous assembly of multi-variety products. However, the differentiation of production batches between the two stages results in a large inventory of work in progress and a long manufacturing cycle. Therefore, this study focuses on lot-streaming flexible job shop scheduling with asynchronous assembly (LS-FJSP-AA). Our approach begins with a mixed-integer linear programming model that aims to lexicographically optimize two objectives: maximum completion time and work-in-progress inventory. Subsequently, a dual-self-learning, knowledge-based local search (DK-LS) algorithm is used to solve this problem. Specifically, forward-reverse decoding based on problem-specific properties is used for objective calculation, and a multi-rule-based heuristic is then embedded into initialization to generate a better initial solution. The local search phase incorporates neighborhood structures based on the critical path and the objectives, guiding the search process towards promising solution spaces. Additionally, a dual-self-learning mechanism is proposed that considers the individual properties of cases. The first uses Q-learning to adjust the candidate solution set and the second selects operators based on historical information. Finally, comprehensive comparisons with six existing heuristics and five state-of-the-art algorithms show that the DK-LS significantly outperforms them in solving the LS-FJSP-AA.

为了实现终端产品的定制化生产，工业企业正在逐步采用新的两阶段柔性生产体系，第一阶段是零部件的柔性制造，第二阶段是多品种产品的异步装配。然而，这两个阶段之间生产批次的差异导致大量在制品库存和较长的制造周期。因此，本研究的重点是基于异步装配的批量流柔性作业车间调度（LS-FJSP-AA）。我们的方法从一个混合整数线性规划模型开始，该模型旨在按字典顺序优化两个目标：最大完成时间和在制品库存。随后，采用双自学习、基于知识的局部搜索（DK-LS）算法来解决这一问题。具体来说，在客观计算中使用基于问题特定属性的正反解码，然后在初始化中嵌入基于多规则的启发式算法，以生成更好的初始解决方案。局部搜索阶段结合了基于关键路径和目标的邻域结构，引导搜索过程走向有希望的解空间。此外，提出了一种考虑案例个体属性的双重自学习机制。第一种方法使用Q-learning来调整候选解集，第二种方法根据历史信息选择算子。最后，与六种现有启发式算法和五种最新算法的综合比较表明，DK-LS在求解LS-FJSP-AA方面明显优于它们。

{"title":"Dual-self-learning knowledge-based local search approach for lot-streaming flexible job shop scheduling with asynchronous assembly","authors":"Gan Jiang , Qiuhua Tang , Lixin Cheng , Zikai Zhang","doi":"10.1016/j.swevo.2026.102283","DOIUrl":"10.1016/j.swevo.2026.102283","url":null,"abstract":"<div><div>To achieve customized production of terminal products, industrial enterprises are gradually adopting a new two-stage flexible production system, where the first stage involves the flexible manufacturing of components and the second stage involves the asynchronous assembly of multi-variety products. However, the differentiation of production batches between the two stages results in a large inventory of work in progress and a long manufacturing cycle. Therefore, this study focuses on lot-streaming flexible job shop scheduling with asynchronous assembly (LS-FJSP-AA). Our approach begins with a mixed-integer linear programming model that aims to lexicographically optimize two objectives: maximum completion time and work-in-progress inventory. Subsequently, a dual-self-learning, knowledge-based local search (DK-LS) algorithm is used to solve this problem. Specifically, forward-reverse decoding based on problem-specific properties is used for objective calculation, and a multi-rule-based heuristic is then embedded into initialization to generate a better initial solution. The local search phase incorporates neighborhood structures based on the critical path and the objectives, guiding the search process towards promising solution spaces. Additionally, a dual-self-learning mechanism is proposed that considers the individual properties of cases. The first uses <em>Q</em>-learning to adjust the candidate solution set and the second selects operators based on historical information. Finally, comprehensive comparisons with six existing heuristics and five state-of-the-art algorithms show that the DK-LS significantly outperforms them in solving the LS-FJSP-AA.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102283"},"PeriodicalIF":8.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146039028","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

Distributed proximal primal–dual splitting for coupled constrained optimization over directed unbalanced networks 有向不平衡网络上耦合约束优化的分布近端原始对偶分裂

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-20 DOI: 10.1016/j.swevo.2026.102282

Run Tang , Huaqing Li , Dawen Xia , Liang Ran , Jun Li , Wei Zhu

This paper investigates a distributed optimization problem over directed multi-agent networks in which each agent has access only to its own local cost function rgb]0.00,0.07,1.00subject to complex constraints, including coupled nonlinear inequality, equality, as well as a private constraint set. Departing from conventional column-stochastic matrix approaches that require explicit outdegree knowledge, a row-stochastic matrix-based method is proposed that inherently resolves graph imbalance without agent outdegree information. rgb]0.00,0.07,1.00Based on the dual decomposition framework and the prediction–correction mechanism, a novel distributed proximal primal–dual splitting method, named Dist_PPDSM, is developed for coupled constrained optimization problems. This method operates effectively over directed, unbalanced graph and establishes provable convergence across uncoordinated step sizes. The convergence and boundedness rgb]0.00,0.07,1.00of Dist_PPDSM are proved by mapping the algorithm to the framework of the maximal monotone operator. Rigorous convergence rate analysis under both general and structural convexity assumptions provides comprehensive performance guarantees across two distinct convexity paradigms. Finally, a simulation of an energy-management system demonstrates the effectiveness of the theoretical findings.

本文研究了一个有向多智能体网络上的分布式优化问题，其中每个智能体只能访问自己的局部成本函数rgb]0.00,0.07,1.00，并受到复杂约束，包括耦合非线性不等式，等式以及私有约束集。针对传统的列随机矩阵方法需要明确的出度信息，提出了一种无需agent出度信息就能固有地解决图失衡的行随机矩阵方法。rgb]0.00,0.07,1.00基于对偶分解框架和预测校正机制，提出了一种求解耦合约束优化问题的分布式近端原始对偶分裂方法Dist_PPDSM。该方法在有向非平衡图上有效地运行，并在非协调步长上建立了可证明的收敛性。通过将算法映射到极大单调算子的框架，证明了Dist_PPDSM的收敛性和有界性rgb]0.00,0.07,1.00。在一般和结构凸性假设下严格的收敛率分析提供了跨两种不同凸性范式的综合性能保证。最后，对一个能量管理系统进行了仿真，验证了理论结果的有效性。

{"title":"Distributed proximal primal–dual splitting for coupled constrained optimization over directed unbalanced networks","authors":"Run Tang , Huaqing Li , Dawen Xia , Liang Ran , Jun Li , Wei Zhu","doi":"10.1016/j.swevo.2026.102282","DOIUrl":"10.1016/j.swevo.2026.102282","url":null,"abstract":"<div><div>This paper investigates a distributed optimization problem over directed multi-agent networks in which each agent has access only to its own local cost function rgb]0.00,0.07,1.00subject to complex constraints, including coupled nonlinear inequality, equality, as well as a private constraint set. Departing from conventional column-stochastic matrix approaches that require explicit outdegree knowledge, a row-stochastic matrix-based method is proposed that inherently resolves graph imbalance without agent outdegree information. rgb]0.00,0.07,1.00Based on the dual decomposition framework and the prediction–correction mechanism, a novel distributed proximal primal–dual splitting method, named Dist_PPDSM, is developed for coupled constrained optimization problems. This method operates effectively over directed, unbalanced graph and establishes provable convergence across uncoordinated step sizes. The convergence and boundedness rgb]0.00,0.07,1.00of Dist_PPDSM are proved by mapping the algorithm to the framework of the maximal monotone operator. Rigorous convergence rate analysis under both general and structural convexity assumptions provides comprehensive performance guarantees across two distinct convexity paradigms. Finally, a simulation of an energy-management system demonstrates the effectiveness of the theoretical findings.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102282"},"PeriodicalIF":8.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146039026","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

Synergistic Particle Swarm Optimized Bio-inspired Artificial Neural Network for Fractional Analysis of tumor-immune competitive system with multiple time delays 多时滞肿瘤免疫竞争系统分数分析的协同粒子群优化仿生神经网络

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-20 DOI: 10.1016/j.swevo.2026.102284

Muhammad Wajahat Anjum , Noreen Sher Akbar , Muhammad Bilal Habib , Taseer Muhammad

This study introduces a novel application of machine learning based intelligent computing for developing bio-inspired neural networks and gradient backpropagation neural networks for tumor-immune interaction model. These models aim to solve nonlinear fractional cancer mathematical system described by four differential equations representing the population dynamics of cancer cells, macrophages, CD8+ T cells, and dendritic cells. Synthetic datasets were generated using the Adams-Bashforth predictor-corrector numerical method, with variations in the time delay and fractional order across each compartment. Both neural network models were trained on these datasets, divided into training and testing sets with an 80:20 ratios. Their performance was evaluated using metrics such as the R-squared score, mean squared error, and visual analyses including absolute error plots, error histograms, and loss curves. A total of six different optimizers were taken with three based on gradient based learning and three based on bio-inspired learning. The models were evaluated based on minimizing the mean squared error. The Bayesian Regularized Gradient based Neural Networks and Particle Swarm Optimized Bio-inspired Artificial Neural Network were found out to be the best performing models in the group of gradient-based and bio-inspired models respectively. However, the Particle Swarm Optimized Bio-inspired Artificial Neural Network demonstrated the highest efficiency, outperforming other gradient and bio-inspired algorithms according to statistical and graphical assessments.

本研究介绍了一种基于机器学习的智能计算的新应用，用于开发生物启发神经网络和梯度反向传播神经网络，用于肿瘤-免疫相互作用模型。这些模型旨在解决由癌细胞、巨噬细胞、CD8+ T细胞和树突状细胞的群体动力学的四个微分方程所描述的非线性分数癌症数学系统。使用Adams-Bashforth预测校正数值方法生成合成数据集，其中每个隔间的时间延迟和分数顺序有所不同。两种神经网络模型都在这些数据集上进行训练，以80:20的比例分为训练集和测试集。使用r平方分数、均方误差和视觉分析（包括绝对误差图、误差直方图和损失曲线）等指标对其性能进行评估。总共采用了6种不同的优化方法，其中3种基于梯度学习，3种基于仿生学习。基于均方误差最小化对模型进行评估。基于贝叶斯正则化梯度的神经网络和基于粒子群优化的仿生人工神经网络分别是基于梯度和仿生模型组中表现最好的模型。然而，根据统计和图形评估，粒子群优化的仿生人工神经网络显示出最高的效率，优于其他梯度和仿生算法。

{"title":"Synergistic Particle Swarm Optimized Bio-inspired Artificial Neural Network for Fractional Analysis of tumor-immune competitive system with multiple time delays","authors":"Muhammad Wajahat Anjum , Noreen Sher Akbar , Muhammad Bilal Habib , Taseer Muhammad","doi":"10.1016/j.swevo.2026.102284","DOIUrl":"10.1016/j.swevo.2026.102284","url":null,"abstract":"<div><div>This study introduces a novel application of machine learning based intelligent computing for developing bio-inspired neural networks and gradient backpropagation neural networks for tumor-immune interaction model. These models aim to solve nonlinear fractional cancer mathematical system described by four differential equations representing the population dynamics of cancer cells, macrophages, CD8+ T cells, and dendritic cells. Synthetic datasets were generated using the Adams-Bashforth predictor-corrector numerical method, with variations in the time delay and fractional order across each compartment. Both neural network models were trained on these datasets, divided into training and testing sets with an 80:20 ratios. Their performance was evaluated using metrics such as the R-squared score, mean squared error, and visual analyses including absolute error plots, error histograms, and loss curves. A total of six different optimizers were taken with three based on gradient based learning and three based on bio-inspired learning. The models were evaluated based on minimizing the mean squared error. The Bayesian Regularized Gradient based Neural Networks and Particle Swarm Optimized Bio-inspired Artificial Neural Network were found out to be the best performing models in the group of gradient-based and bio-inspired models respectively. However, the Particle Swarm Optimized Bio-inspired Artificial Neural Network demonstrated the highest efficiency, outperforming other gradient and bio-inspired algorithms according to statistical and graphical assessments.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102284"},"PeriodicalIF":8.5,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146039027","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 multi-objective optimization algorithm via historical collaborative strategy and interval prediction strategy 基于历史协同策略和区间预测策略的动态多目标优化算法

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-19 DOI: 10.1016/j.swevo.2026.102281

Junwei Ou , Yana Li , Yaru Hu , Jiankang Peng , Jinhua Zheng , Juan Zou , Shengxiang Yang

Dynamic multi-objective optimization problems (DMOPs) involve scenarios where objective functions, decision variables, parameters, or other elements vary over time. An effective approach to address DMOPs is to integrate algorithms designed for static multi-objective optimization problems with dynamic response strategies. To improve the performance of these strategies in terms of both population diversity and convergence, this paper proposes a novel dynamic response strategy, the historical collaborative and interval prediction strategy (HCIPS). When confronted with environmental changes, we conduct a three-level population analysis: overall, historical, and individual. Firstly, the interval-based response strategy identifies interval partitioning of a population at time

t

, enabling global localization of the predicted population and effectively preserving diverse population information. Secondly, the history-based response strategy guides the population movement by selecting optimal solutions from historical populations. Thirdly, the individual-based response strategy predicts individual positions by tracking the movement of key points. This serves as a crucial complement to the history-based response strategy, compensating for its primary drawback: a lack of sufficient historical data in the early stages of evolution. Experimental results indicate that the HCIPS offers advantages in solving DMOPs compared to past state-of-the-art algorithms.

动态多目标优化问题（dops）涉及目标函数、决策变量、参数或其他元素随时间变化的场景。将静态多目标优化算法与动态响应策略相结合是解决多目标优化问题的有效途径。为了提高这些策略在种群多样性和收敛性方面的性能，本文提出了一种新的动态响应策略——历史协同和区间预测策略（HCIPS）。当面临环境变化时，我们进行了三个层次的人口分析：总体、历史和个体。首先，基于区间的响应策略识别了种群在时刻t的区间划分，实现了预测种群的全局定位，有效地保留了种群的多样性信息。其次，基于历史的响应策略通过从历史人口中选择最优解来引导人口运动。第三，基于个体的响应策略通过跟踪关键点的移动来预测个体的位置。这是对基于历史的应对策略的重要补充，弥补了其主要缺点：在进化的早期阶段缺乏足够的历史数据。实验结果表明，与现有算法相比，HCIPS在求解dops方面具有优势。

{"title":"Dynamic multi-objective optimization algorithm via historical collaborative strategy and interval prediction strategy","authors":"Junwei Ou , Yana Li , Yaru Hu , Jiankang Peng , Jinhua Zheng , Juan Zou , Shengxiang Yang","doi":"10.1016/j.swevo.2026.102281","DOIUrl":"10.1016/j.swevo.2026.102281","url":null,"abstract":"<div><div>Dynamic multi-objective optimization problems (DMOPs) involve scenarios where objective functions, decision variables, parameters, or other elements vary over time. An effective approach to address DMOPs is to integrate algorithms designed for static multi-objective optimization problems with dynamic response strategies. To improve the performance of these strategies in terms of both population diversity and convergence, this paper proposes a novel dynamic response strategy, the historical collaborative and interval prediction strategy (HCIPS). When confronted with environmental changes, we conduct a three-level population analysis: overall, historical, and individual. Firstly, the interval-based response strategy identifies interval partitioning of a population at time <span><math><mi>t</mi></math></span>, enabling global localization of the predicted population and effectively preserving diverse population information. Secondly, the history-based response strategy guides the population movement by selecting optimal solutions from historical populations. Thirdly, the individual-based response strategy predicts individual positions by tracking the movement of key points. This serves as a crucial complement to the history-based response strategy, compensating for its primary drawback: a lack of sufficient historical data in the early stages of evolution. Experimental results indicate that the HCIPS offers advantages in solving DMOPs compared to past state-of-the-art algorithms.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102281"},"PeriodicalIF":8.5,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146039024","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

Low-cost safe path planning and exit scheduling of multi-UAV aerial refueling based on swarm intelligence 基于群体智能的多无人机空中加油低成本安全路径规划与出口调度

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-19 DOI: 10.1016/j.swevo.2026.102293

Bin Hang , Pengjun Guo

Aerial refueling technology is a crucial means of extending unmanned aerial vehicles (UAVs) mission duration and expanding operational range, garnering extensive attention. However, planning safe and cost-effective refueling routes for multiple UAVs in complex three-dimensional airspace, and achieving efficient and orderly egress after mission completion, still face technical challenges such as inadequate path safety and low egress scheduling efficiency. To address these challenges, this paper proposes a multi-agent hierarchical collaborative optimization framework that simulates group competition and cooperation to achieve task allocation and path coordination. By integrating factors such as path length, threat sources, air turbulence, altitude-dependent energy consumption, and turning loss, a multi-dimensional cost function is constructed, forming a comprehensive trajectory optimization model for UAV aerial refueling missions. Based on flight landing scheduling (FLS) theory, a dynamic time window allocation and conflict resolution mechanism is introduced, establishing a two-stage optimization architecture of ”path planning-safe egress.” Simulation results indicate that, compared to several mainstream meta-heuristic algorithms, the proposed method achieves superior path quality and higher scheduling efficiency under complex conditions, reliably accomplishing low-cost, coordinated multi-UAV refueling and safe egress operations.

空中加油技术是延长无人机任务时间、扩大作战范围的重要手段，受到了广泛关注。然而，在复杂的三维空域中规划安全、经济的多架无人机加油路线，并在任务完成后实现高效有序的出口，仍然面临路径安全性不足、出口调度效率低等技术挑战。为了解决这些问题，本文提出了一个模拟群体竞争与合作的多智能体分层协同优化框架，以实现任务分配和路径协调。通过整合路径长度、威胁源、空气湍流、高度相关能耗、转向损失等因素，构建多维代价函数，形成无人机空中加油任务综合轨迹优化模型。基于飞机着陆调度理论，引入了一种动态时间窗分配与冲突解决机制，建立了“路径规划-安全出口”两阶段优化体系结构。仿真结果表明，与几种主流的元启发式算法相比，该方法在复杂条件下具有更好的路径质量和更高的调度效率，可靠地完成了低成本、协同的多无人机加油和安全撤离操作。

{"title":"Low-cost safe path planning and exit scheduling of multi-UAV aerial refueling based on swarm intelligence","authors":"Bin Hang , Pengjun Guo","doi":"10.1016/j.swevo.2026.102293","DOIUrl":"10.1016/j.swevo.2026.102293","url":null,"abstract":"<div><div>Aerial refueling technology is a crucial means of extending unmanned aerial vehicles (UAVs) mission duration and expanding operational range, garnering extensive attention. However, planning safe and cost-effective refueling routes for multiple UAVs in complex three-dimensional airspace, and achieving efficient and orderly egress after mission completion, still face technical challenges such as inadequate path safety and low egress scheduling efficiency. To address these challenges, this paper proposes a multi-agent hierarchical collaborative optimization framework that simulates group competition and cooperation to achieve task allocation and path coordination. By integrating factors such as path length, threat sources, air turbulence, altitude-dependent energy consumption, and turning loss, a multi-dimensional cost function is constructed, forming a comprehensive trajectory optimization model for UAV aerial refueling missions. Based on flight landing scheduling (FLS) theory, a dynamic time window allocation and conflict resolution mechanism is introduced, establishing a two-stage optimization architecture of ”path planning-safe egress.” Simulation results indicate that, compared to several mainstream meta-heuristic algorithms, the proposed method achieves superior path quality and higher scheduling efficiency under complex conditions, reliably accomplishing low-cost, coordinated multi-UAV refueling and safe egress operations.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102293"},"PeriodicalIF":8.5,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146039023","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 novel bio-inspired encoding for evolving cryptographic Boolean functions 一种用于演化密码学布尔函数的新型仿生编码

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-17 DOI: 10.1016/j.swevo.2026.102287

Rocco Ascone , Giulia Bernardini , Luca Manzoni , Gloria Pietropolli

Discovering Boolean functions that satisfy properties such as balancedness and nonlinearity is a complex optimization problem, which is crucial to important cryptographic constructions like block and stream ciphers. The difficulty of this problem lies in the search space growing super-exponentially in the number of variables. Evolutionary approaches, including Genetic Algorithms (GAs) and Genetic Programming (GP), have been successfully applied to overcome this difficulty. The major drawback of these methods is that they evolve functions through encodings that are either exponential in the input size or hard to interpret. We address this problem as follows. (i) We propose a new encoding for Boolean functions as reaction systems, a bio-inspired computational model which can be directly translated into the compact and easily interpretable Disjunctive Normal Form (DNF). (ii) We design EvoBRS, an evolutionary optimization framework that exploits this new representation to discover Boolean functions with maximum nonlinearity (bent functions), possibly under the balancedness constraint. (iii) We back up our novel paradigm with a refined theoretical analysis of independent interest. (iv) We conduct a rigorous experimental study, demonstrating that EvoBRS consistently discovers diverse, highly nonlinear Boolean functions with and without the balancedness constraint. EvoBRS proves particularly effective on balanced functions, successfully identifying balanced maximally nonlinear instances and outperforming both GP and state-of-the-art GAs. All the discovered functions are returned in a compact and easily interpretable DNF. A preliminary version of this work appeared in Ascone et al., GECCO 2025.

发现满足平衡性和非线性等性质的布尔函数是一个复杂的优化问题，这对于块密码和流密码等重要的密码结构至关重要。这个问题的难点在于搜索空间的变量数量呈指数级增长。包括遗传算法（GAs）和遗传规划（GP）在内的进化方法已经成功地应用于克服这一困难。这些方法的主要缺点是，它们通过输入大小呈指数级增长或难以解释的编码来演化函数。我们解决这个问题的方法如下。(i)我们提出了一种新的布尔函数编码作为反应系统，这是一种生物启发的计算模型，可以直接转换为紧凑且易于解释的析取范式（DNF）。（ii）我们设计了EvoBRS，这是一个进化优化框架，利用这种新的表示来发现可能在平衡约束下具有最大非线性（弯曲函数）的布尔函数。（iii）我们用对独立利益的精细化理论分析来支持我们的新范式。（iv）我们进行了严格的实验研究，证明EvoBRS在有和没有平衡约束的情况下始终能够发现多种高度非线性的布尔函数。EvoBRS被证明在平衡函数上特别有效，成功地识别了平衡的最大非线性实例，并且优于GP和最先进的GAs。所有发现的函数都以紧凑且易于解释的DNF形式返回。这项工作的初步版本出现在Ascone等人的GECCO 2025中。

{"title":"A novel bio-inspired encoding for evolving cryptographic Boolean functions","authors":"Rocco Ascone , Giulia Bernardini , Luca Manzoni , Gloria Pietropolli","doi":"10.1016/j.swevo.2026.102287","DOIUrl":"10.1016/j.swevo.2026.102287","url":null,"abstract":"<div><div>Discovering Boolean functions that satisfy properties such as balancedness and nonlinearity is a complex optimization problem, which is crucial to important cryptographic constructions like block and stream ciphers. The difficulty of this problem lies in the search space growing super-exponentially in the number of variables. Evolutionary approaches, including Genetic Algorithms (GAs) and Genetic Programming (GP), have been successfully applied to overcome this difficulty. The major drawback of these methods is that they evolve functions through encodings that are either exponential in the input size or hard to interpret. We address this problem as follows. (i) We propose a new encoding for Boolean functions as reaction systems, a bio-inspired computational model which can be directly translated into the compact and easily interpretable Disjunctive Normal Form (DNF). (ii) We design EvoBRS, an evolutionary optimization framework that exploits this new representation to discover Boolean functions with maximum nonlinearity (bent functions), possibly under the balancedness constraint. (iii) We back up our novel paradigm with a refined theoretical analysis of independent interest. (iv) We conduct a rigorous experimental study, demonstrating that EvoBRS consistently discovers diverse, highly nonlinear Boolean functions with and without the balancedness constraint. EvoBRS proves particularly effective on balanced functions, successfully identifying balanced maximally nonlinear instances and outperforming both GP and state-of-the-art GAs. All the discovered functions are returned in a compact and easily interpretable DNF. A preliminary version of this work appeared in Ascone et al., GECCO 2025.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102287"},"PeriodicalIF":8.5,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980785","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

Multi-objective optimization in autonomous foraging using swarm robots 群机器人自主觅食的多目标优化

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

Swarm and Evolutionary Computation

Pub Date : 2026-01-16 DOI: 10.1016/j.swevo.2026.102294

Erick J. Ordáz-Rivas, Angel Rodríguez-Liñan, Luis M. Torres-Treviño

Swarm robotics is an innovative field focused on developing collective behaviors through local interactions among simple robots, enabling scalability and flexibility across a wide range of tasks. This study presents a behavioral model for collective foraging based on RAOI (repulsion, attraction, orientation, and influence) parameters, and investigates how their tuning affects multi-objective performance in robot swarms. Our approach explores the relationship between RAOI parameter configurations and task-level performance metrics, allowing systematic analysis of emergent swarm behaviors in dynamic environments.

In this work, the tuning of RAOI parameters is formulated as a multi-objective optimization problem guided by established evolutionary algorithms (MOEA/D and NSGA-III), yielding Pareto-optimal trade-offs among competing objectives. The obtained solutions illustrate improvements across multiple criteria, including task completion time, energy consumption, workload distribution, and swarm size efficiency, highlighting inherent trade-offs rather than a single optimal configuration.

The results provide insights into how RAOI-based interaction parameters influence collective foraging dynamics and overall swarm performance. The study focuses on simulation-based evaluation, offering a structured framework for analyzing and tuning swarm behaviors in foraging tasks and related collective robotics scenarios.

群机器人是一个创新领域，专注于通过简单机器人之间的局部交互来发展集体行为，从而在广泛的任务范围内实现可扩展性和灵活性。本文提出了一种基于RAOI（斥力、吸引力、方向和影响）参数的集体觅食行为模型，并研究了它们的调整如何影响机器人群体的多目标性能。我们的方法探索了RAOI参数配置和任务级性能指标之间的关系，允许系统分析动态环境中的突发群体行为。在这项工作中，RAOI参数的调整被制定为由已建立的进化算法（MOEA/D和NSGA-III）指导的多目标优化问题，在竞争目标之间产生帕累托最优权衡。获得的解决方案说明了跨多个标准的改进，包括任务完成时间、能耗、工作负载分布和群大小效率，突出了固有的权衡，而不是单一的最优配置。研究结果揭示了基于raoi的交互参数如何影响群体觅食动态和整体群体表现。该研究侧重于基于仿真的评估，为分析和调整觅食任务和相关集体机器人场景中的群体行为提供了结构化框架。

{"title":"Multi-objective optimization in autonomous foraging using swarm robots","authors":"Erick J. Ordáz-Rivas, Angel Rodríguez-Liñan, Luis M. Torres-Treviño","doi":"10.1016/j.swevo.2026.102294","DOIUrl":"10.1016/j.swevo.2026.102294","url":null,"abstract":"<div><div>Swarm robotics is an innovative field focused on developing collective behaviors through local interactions among simple robots, enabling scalability and flexibility across a wide range of tasks. This study presents a behavioral model for collective foraging based on <em>RAOI</em> (repulsion, attraction, orientation, and influence) parameters, and investigates how their tuning affects multi-objective performance in robot swarms. Our approach explores the relationship between <em>RAOI</em> parameter configurations and task-level performance metrics, allowing systematic analysis of emergent swarm behaviors in dynamic environments.</div><div>In this work, the tuning of <em>RAOI</em> parameters is formulated as a multi-objective optimization problem guided by established evolutionary algorithms (MOEA/D and NSGA-III), yielding Pareto-optimal trade-offs among competing objectives. The obtained solutions illustrate improvements across multiple criteria, including task completion time, energy consumption, workload distribution, and swarm size efficiency, highlighting inherent trade-offs rather than a single optimal configuration.</div><div>The results provide insights into how <em>RAOI</em>-based interaction parameters influence collective foraging dynamics and overall swarm performance. The study focuses on simulation-based evaluation, offering a structured framework for analyzing and tuning swarm behaviors in foraging tasks and related collective robotics scenarios.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"101 ","pages":"Article 102294"},"PeriodicalIF":8.5,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980783","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 survey of features used for representing black-box single-objective continuous optimization 黑箱单目标连续优化的特征描述

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.2026.102288

Gjorgjina Cenikj , Ana Nikolikj , Gašper Petelin , Niki van Stein , Carola Doerr , Tome Eftimov

This survey examines key advancements in designing features to represent optimization problem instances, algorithm instances, and their interactions within the context of single-objective continuous black-box optimization. These features support machine learning tasks such as algorithm selection, algorithm configuration, and problem classification, and they are also used to evaluate the complementarity of benchmark problem sets. We provide a comprehensive overview of problem landscape features, algorithm features, high-level problem-algorithm interaction features, and trajectory features, including the latest works from the past five years. We also point out limitations of the current state-of-the-art and suggest directions for future research.

本调查研究了在单目标连续黑盒优化环境中，设计特征以表示优化问题实例、算法实例及其相互作用方面的关键进展。这些特征支持机器学习任务，如算法选择、算法配置和问题分类，它们也用于评估基准问题集的互补性。我们提供了问题景观特征、算法特征、高级问题-算法交互特征和轨迹特征的全面概述，包括过去五年的最新研究成果。本文还指出了目前研究水平的局限性，并提出了未来研究的方向。

引用次数: 0