A multi-population co-evolutionary algorithm based on dual-space division for dynamic multi-objective optimization problems

Effective optimization algorithms are urgently needed for dynamic multi-objective optimization problems (DMOPs) to efficiently track the Pareto optimal front (POF) or Pareto optimal set (POS) in response to environmental changes. This paper proposes a multi-population co-evolutionary algorithm based...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	Information sciences Ročník 728; s. 122712
Hlavní autoři:	Hu, Yaru, Xia, Jiaru, Ou, Junwei, Song, Yanjie, Li, Jun, Zheng, Jinhua, Wang, Gaige, Zhang, Yue
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	Elsevier Inc 01.02.2026
Témata:	Dynamic multi-objective optimization Multi-population co-evolutionary Multi-population diversity maintenance Prediction Subpopulation division Subpopulation division Multi-population co-evolutionary Dynamic multi-objective optimization Multi-population diversity maintenance Prediction
ISSN:	0020-0255
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	Effective optimization algorithms are urgently needed for dynamic multi-objective optimization problems (DMOPs) to efficiently track the Pareto optimal front (POF) or Pareto optimal set (POS) in response to environmental changes. This paper proposes a multi-population co-evolutionary algorithm based on dual-space division (MPDS) for solving DMOPs, which aims to achieve rapid convergence while maintaining population diversity. The algorithm divides the subpopulations within both the objective space and the decision space, integrating information from both spaces to more accurately predict the updated positions of the new population after environmental changes. The algorithm consists of three essential steps: first, it divides the subpopulations within both the objective space and the decision space after detecting an environmental change. Second, new individuals adapted to the updated POF are generated based on the evolutionary directions of the center points in each subpopulation, promoting rapid population convergence. Meanwhile, to prevent the population from getting trapped in local optima, new individuals are randomly generated within a certain region of the subpopulation to enhance diversity. Third, the original population is combined with the two populations generated in the previous steps. Finally, the solutions with good convergence and diversity are selected after performing a non-dominated sort on the combined solutions. Experiments demonstrate that our algorithm outperforms five state-of-the-art algorithms in terms of balancing convergence and diversity. •The paper proposes a dynamic multi-objective optimization algorithm based on historical collaborative strategy and interval prediction strategy.•The response strategy determines the direction of population movement by selecting optimal solutions from the historical population.•The individual-based response strategy predicts the positions of individuals in a population by considering the movement direction of special point.
ISSN:	0020-0255
DOI:	10.1016/j.ins.2025.122712