Swarm and Evolutionary Computation最新文献

{"title":"A hybrid decomposition and adaptive grid evolutionary algorithm for multi-objective optimization with irregular Pareto fronts","authors":"Wenqing Feng,&nbsp;Dunwei Gong,&nbsp;He Sun,&nbsp;Wei Cai,&nbsp;Ying Huo","doi":"10.1016/j.swevo.2026.102556","DOIUrl":"10.1016/j.swevo.2026.102556","url":null,"abstract":"<div><div>Irregular Pareto fronts (PFs) pose a critical challenge to multi-objective evolutionary algorithms (MOEAs) and particularly degrade the performance of decomposition-based MOEAs, as their search directions and neighborhood cooperation rely heavily on pre-defined uniform reference vectors. Meanwhile, conventional grid-based methods exhibit poor adaptability to irregular PFs due to their fixed uniform partitioning, and refining grid granularity for improved PF coverage leads to exponentially growing computational overhead. To address the limitations, this paper proposes a hybrid evolutionary algorithm combining decomposition and grid-based strategies, denoted as MOEA/D-AG. In the first evolutionary stage, an extrapolation-guided search mechanism is embedded into the decomposition framework to identify promising search directions, and accelerate population convergence toward the true PF. In the second stage, an adaptive grid-based mechanism is developed: grid granularity is dynamically adjusted via a two-step strategy of rapid coarse adjustment and gradual fine-tuning. This strategy adapts to the distribution characteristics of the current PF approximation and achieves flexible discretization of the objective space, thereby preserving population diversity effectively while controlling computational cost. The transition between the two stages is governed by a quantitative threshold defined by the rate of change in the population’s convergence metric, ensuring a favorable balance between convergence and diversity. Extensive experiments are conducted on 22 multi-objective benchmark problems in comparison with 11 state-of-the-art algorithms. The results demonstrate that MOEA/D-AG achieves statistically superior performance. In particular, it exhibits prominent advantages on instances with irregular PFs. Its practical effectiveness is further validated on a real-world multi-objective optimization problem arising from the carbon fiber forming process, on which the algorithm consistently yields well-converged and uniformly distributed Pareto-optimal solution sets.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102556"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157280","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}

{"title":"Constrained multi-objective optimization based on Lasso double machine learning","authors":"Yinghan Hong,&nbsp;Qiyuan Wu,&nbsp;Sirui Liang,&nbsp;Guizhen Mai,&nbsp;Yi Xiang,&nbsp;Pinghua Chen","doi":"10.1016/j.swevo.2026.102550","DOIUrl":"10.1016/j.swevo.2026.102550","url":null,"abstract":"<div><div>Constrained multi-objective optimization requires a delicate balance between objective improvement and constraint satisfaction throughout the search process. While existing stage-wise methods utilize infeasible solutions, they rely on rigid heuristic rules and fail to quantify the actual utility of these solutions. To address this limitation, we propose a novel approach leveraging double machine learning to dynamically assess the value of infeasible solutions. Specifically, we formulate a causal framework where the lagged infeasible-solution ratio serves as the treatment, and the subsequent fixed-reference online hypervolume improvement is the outcome. By employing historical-only time-block cross-fitting with Lasso regularization and double residualization, our method derives an orthogonal policy score that accurately estimates the causal effect of this treatment. This score is then converted into a bounded replacement ratio, stabilized by hypervolume feedback, and integrated into both decomposition- and dominance-based evolutionary algorithms. Extensive experiments on constrained multi-objective optimization benchmarks and synthetic temporal designs demonstrate that our approach achieves superior optimization performance. Furthermore, it yields calibrated score estimates and correct directional control, proving robust against temporal placebos and score noise.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102550"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157338","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}

{"title":"Dual-space clustering-guided particle swarm optimization for multi-objective optimization","authors":"Meng Yan,&nbsp;Shuai Wang,&nbsp;Xiangzhou Gao,&nbsp;Yi Zhang,&nbsp;Guohong Dai","doi":"10.1016/j.swevo.2026.102557","DOIUrl":"10.1016/j.swevo.2026.102557","url":null,"abstract":"<div><div>In multi-objective particle swarm optimization (MOPSO) algorithms, the guiding solutions (i.e., <em>pbest</em> and <em>gbest</em>) are usually selected based on neighborhood relationships defined in either the decision space or the objective space. However, the information from these two spaces is complementary, and relying on just one of them might limit optimization performance. This study proposes a dual-space clustering enhanced MOPSO algorithm, called DCPSO. To be specific, in the proposed DCPSO, the density clustering method is performed in the decision space to identify the distribution of particles, and some outliers within the current swarm are distinguished which located in sparse regions. For particles identified as outliers, <span><math><mi>K</mi></math></span>-means clustering is performed in the objective space, and elite solutions are selected from the corresponding clusters to guide their search. Moreover, for these particles in normal clusters in the decision space, density-reachable information is used to identify neighboring particles for selecting <em>pbest</em> and <em>gbest</em>. In addition, outliers in sparse regions are allowed to serve as <em>gbest</em> candidates which helps maintain population diversity. The developed DCPSO is compared with six representative algorithms on 34 benchmark problems from 4 test suites and 9 real-world problems. The results show that DCPSO obtains the best HV and <span><math><msup><mrow><mtext>IGD</mtext></mrow><mrow><mo>+</mo></mrow></msup></math></span> values on 26 out of 34 benchmark problems and achieves the highest overall ranking among the compared algorithms.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102557"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157273","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}

{"title":"Efficient incremental non-dominated sorting for bi-objective problems via linear-space IDG representation","authors":"D. Lahoz,&nbsp;I. Alberto,&nbsp;P.M. Mateo","doi":"10.1016/j.swevo.2026.102558","DOIUrl":"10.1016/j.swevo.2026.102558","url":null,"abstract":"<div><div>This work presents an efficient incremental non-dominated sorting algorithm for bi-objective vectors. Non-dominated sorting algorithms organize solutions into layers, or Pareto fronts, based on dominance relations, and play a key role in the selection process of many well-known population-based algorithms for multi-objective optimization. Since the non-dominated sorting process is computationally expensive, its efficient design is of critical importance.</div><div>The proposed algorithm is incremental, that is, given a population whose solutions are already assigned to layers, a new solution is incorporated and the dominance structure is updated as efficiently as possible, reconstructing the layers with minimal computational effort. The method is based on irreducible domination graphs of order 1, <em>IDG</em><span><math><msup><mrow></mrow><mrow><mn>1</mn></mrow></msup></math></span>, for which a storage structure requiring linear space in the number of solutions is introduced, instead of the quadratic requirements of the general case. Based on this representation, an incremental non-dominated sorting algorithm with <span><math><mrow><mi>O</mi><mrow><mo>(</mo><mi>n</mi><mo>)</mo></mrow></mrow></math></span> complexity is developed, matching the complexity of well-known algorithms in the literature while relying on simpler data structures. Key components of the approach include locating the appropriate front for the new solution, managing displaced nodes across layers, and efficiently updating dominance relations. This linear complexity is achieved thanks to the structure of the <em>IDG</em><span><math><msup><mrow></mrow><mrow><mn>1</mn></mrow></msup></math></span>, which stores immediate dominance relations among solutions, i.e., the set of nodes in the previous Pareto layer that directly dominate each node.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102558"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157278","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A dynamic programming-based memetic algorithm for parallel berth allocation problem at multipurpose inland waterway terminals","authors":"Adnan Pasha,&nbsp;Jiyin Liu,&nbsp;Rajat Rastogi","doi":"10.1016/j.swevo.2026.102542","DOIUrl":"10.1016/j.swevo.2026.102542","url":null,"abstract":"<div><div>The authors investigate the Berth Allocation Problem (BAP) at multipurpose Inland Water Transport (IWT) terminals, which are commonly found in developing economies such as India. Extensive research on BAP is available for specialized seaport terminals in developed countries. But limited attention is given to multipurpose shallow-draft IWT terminals in developing countries where multiple cargo types are handled at the same terminal using varied equipment and barge configurations. A key feature of this study is the integration of ‘parallel berthing’, allowing two barges to berth side-by-side on the same berth. Presently, the first-come-first-served (FCFS) strategy employed at IWT terminals in India results in high completion and waiting times. To overcome these inefficiencies, a Mixed Integer Linear Programming (MILP) model is formulated which incorporates parallel berthing and minimizes both completion and waiting times. For large instances, a Dynamic Programming-based Memetic Algorithm (DPMA) is developed which uses Memetic Algorithm for berth allocation, and Dynamic Programming for scheduling and parallel berthing. The performance of the proposed approach is evaluated using operational data from an Indian IWT terminal. Computational results show that DPMA outperforms CPLEX and other (meta)heuristics in terms of solution quality, robustness, and consistency. Parallel berthing shows significant improvements over single berthing. Enabling compatibility of cargos and feasibility of berths for parallel handling further reduces completion and waiting times. The findings offer valuable managerial insights and a robust decision support framework to assist IWT terminal operators in effectively managing multiple cargo types and optimizing operations under parallel berthing conditions.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102542"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149212542","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}

{"title":"Models and approximate algorithm for multi-objective job shop scheduling problem with vehicles dynamic recharging","authors":"Bohan Zhang,&nbsp;Zhaohe Wang,&nbsp;Mengxing Gao,&nbsp;Ying Chen","doi":"10.1016/j.swevo.2026.102560","DOIUrl":"10.1016/j.swevo.2026.102560","url":null,"abstract":"<div><div>In practice, automated guided vehicles (AGVs) must operate continuously in manufacturing systems; however, frequent recharging inevitably increases vehicle downtime and extends overall job completion times. This study investigates a multi-objective job shop scheduling problem with vehicles dynamic recharging (MJSP-VDR), and develops exact and approximate solution approaches to solve the problem. For the exact approach, we formulate mixed-integer nonlinear programming and mixed-integer linear programming models that characterize the dynamic charging features of AGVs by defining the relationship between the initial and remaining battery levels for each recharging round. The two models are capable of generating optimal Pareto solutions for small-sized instances using the <span><math><mi>ϵ</mi></math></span>-constraint method. Furthermore, an approximate algorithm, named the competitive populations-based evolutionary algorithm (CPEA), is developed to efficiently solve medium- and large-scale instances. The method decomposes MJSP-VDR into two coupled subproblems and introduces a dynamic recharging strategy to prevent excessive charging at each AGV recharging round. We then design an individual evaluation metric that integrates crowding and Tschebyscheff values, and a competitive evolutionary mechanism that executes scenario-specific evolution based on pairwise comparisons between the best and worst individuals across populations. The computational results demonstrate that CPEA obtains the same Pareto solutions as the exact models on small-scale instances. For medium- and large-scale instances, it significantly outperforms several state-of-the-art algorithms, not only exhibiting strong robustness but also achieving higher-quality solutions, while reducing the average computational time by 41.11% and 29.96%, respectively.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102560"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157279","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}

{"title":"A two-stage learning hyper-heuristic for integrated open-field layout and fuzzy flexible job shop scheduling","authors":"Yongqin Liu,&nbsp;Nanfeng Song,&nbsp;Chao Wang,&nbsp;Min Zuo,&nbsp;Lei Wang","doi":"10.1016/j.swevo.2026.102547","DOIUrl":"10.1016/j.swevo.2026.102547","url":null,"abstract":"<div><div>The coordinated optimization of open-field layout and scheduling is critical for enhancing production efficiency in flexible job shops. However, it is difficult to obtain satisfactory solutions owing to the sparse feasible solution space in the spatiotemporally coupled layout and scheduling problem. To address this challenge, this paper develops a two-stage framework that mitigates the coupling complexity by transitioning from global exploration under relaxed constraints to strictly constrained optimization. Building on this framework, a two-stage learning hyper-heuristic algorithm (TSLHH) is proposed to solve the integrated optimization model aiming at minimizing fuzzy makespan and energy consumption. In the first stage, a double deep Q-network (DDQN)-based high-level strategy is developed for macro-exploration, locating promising solution regions under relaxed layout constraints. Then, a role-oriented DDQN is proposed for the second stage, which applies differentiated search pressures according to evolutionary roles to simultaneously enforce physical feasibility and refine solution quality. Furthermore, distinct sets of problem-specific knowledge rules based on layout geometry are embedded to correct sub-optimal spatial structures and accelerate convergence. Experimental results on 38 benchmark instances, 27 scenario-based instances and an industrial case study demonstrate that TSLHH outperforms the eight compared algorithms.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102547"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157329","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}

{"title":"Dual-Population Hybrid Genetic Programming for Decision Forest Optimization in Adaptive Classification Rule Construction","authors":"Xingsi Xue,&nbsp;Syed Muhammad Waqas,&nbsp;Guodong He,&nbsp;Wenxi Liu,&nbsp;Jianhui Lv","doi":"10.1016/j.swevo.2026.102552","DOIUrl":"10.1016/j.swevo.2026.102552","url":null,"abstract":"<div><div>Complex data fusion requires identifying consistent associations across heterogeneous data sources with diverse semantics, structures and feature representations, posing significant challenges for reliable classification. Decision Tree (DT)-based methods provide interpretability but often lack robustness and adaptability in highly heterogeneous environments, while ensemble approaches such as Decision Forests (DFs) still face challenges in balancing diversity, efficiency and optimization complexity. This paper aims to develop an adaptive classification framework for complex data fusion by enhancing structural diversity and evolutionary optimization. To this end, a Dual-Population-based Hybrid Genetic Programming (DPHGP) framework is proposed to automatically construct and optimize DF-based classification rules. Specifically, a dual-population co-evolutionary mechanism is designed to evolve decision trees for similarity feature construction and ensemble structures for aggregation, thereby improving diversity and robustness. Meanwhile, adaptive breeding operators are introduced to balance global exploration and local exploitation during evolution. Furthermore, a Compact Genetic Algorithm (CGA) is incorporated to refine numerical constants within elite classifiers, thereby supporting more effective numerical parameter optimization. The experimental results on OAEI benchmark datasets show that DPHGP achieves competitive matching accuracy and robustness across the evaluated datasets, compared with state-of-the-art evolutionary and machine learning approaches. These results indicate that the proposed framework provides an effective and interpretable solution for adaptive rule construction in complex data fusion scenarios.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102552"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157271","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}

{"title":"Enhancing search efficiency in large-scale constrained multi-objective optimization using a decision variable importance matrix","authors":"Guangyu Qian,&nbsp;Mingcheng Zuo,&nbsp;Yong Zhang,&nbsp;Dunwei Gong,&nbsp;Ali Wagdy Mohamed","doi":"10.1016/j.swevo.2026.102553","DOIUrl":"10.1016/j.swevo.2026.102553","url":null,"abstract":"<div><div>Existing constrained multi-objective evolutionary algorithms (CMOEAs) encounter significant challenges in solving large-scale constrained multi-objective optimization problems (LSCMOPs), as the exponentially increasing search space severely restricts their optimization efficiency. To address this issue, this paper proposes a variable importance matrix-assisted large-scale constrained multi-objective evolutionary optimization method. First, multiple decision trees are constructed at the beginning of each generation based on historical population data. The resulting variable importance scores with respect to objectives and constraints are integrated to form a variable importance matrix (VIM). Second, the objective space is uniformly divided using reference vectors, and the VIM is employed to estimate the variable importance of individuals in different search subregions, resulting in the subregion-based variable importance matrix (SVIM). Finally, differentiated search directions are developed for different individuals according to the SVIM, thereby enhancing the search efficiency of the population. The proposed method can be seamlessly incorporated into existing CMOEAs and improve their performance to varying degrees. Extensive comparative experiments are conducted against nine state-of-the-art algorithms on 162 benchmark problems and an integrated coal mine energy system scheduling optimization problem. The results demonstrate the effectiveness and superiority of the proposed method.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102553"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157339","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}

{"title":"Bi-objective chance-constrained evolutionary optimization for large-scale open-pit mine scheduling under geological uncertainty","authors":"Ishara Hewa Pathiranage,&nbsp;Aneta Neumann","doi":"10.1016/j.swevo.2026.102543","DOIUrl":"10.1016/j.swevo.2026.102543","url":null,"abstract":"<div><div>The open-pit mine scheduling problem (OPMSP) is a complex optimization problem in long-term mine planning that involves numerous operational and geological constraints. Traditional deterministic approaches often ignore geological uncertainty, leading to suboptimal or unreliable production schedules. Chance constraints provide a framework for handling uncertainty by ensuring that probabilistic constraints are satisfied with a predefined confidence level. In this paper, we consider the OPMSP under geological grade uncertainty and propose a bi-objective chance-constrained formulation that simultaneously maximizes the expected discounted net present value and minimizes scheduling risk. Unlike traditional chance-constrained approaches, the proposed formulation does not require a predefined confidence level during optimization. Instead, it generates a set of Pareto-optimal solutions representing different trade-offs between profitability and risk within a single optimization run. To solve the resulting large-scale stochastic optimization problem, we employ multi-objective evolutionary algorithms and compare their performance against a single-objective chance-constrained evolutionary approach and a deterministic MILP benchmark. We further evaluate the contribution of the problem-specific initialization and mutation components through an ablation study. Experimental results on MineLib benchmark instances containing up to <span><math><mrow><mn>112</mn><mspace></mspace><mn>687</mn></mrow></math></span> blocks demonstrate that the proposed formulation effectively captures the trade-off between profitability and risk under geological uncertainty while providing greater flexibility than confidence-level-dependent approaches.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"109 ","pages":"Article 102543"},"PeriodicalIF":9.6,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149157276","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}