A dynamic constrained multi-objective evolutionary algorithm with a dual prediction mechanism

•Biased sampling using predicted similarity across multiple environments.•Improved prediction method applied as an initialization and search strategy.•Strategy to reuse evaluation budget from environmental detection.•Dual-population static optimizer for dynamic constrained multi-objective problems....

Full description

Saved in:
Bibliographic Details
Published in:Expert systems with applications Vol. 297; p. 129304
Main Authors: Yang, Zhe, Deng, Libao, Li, Chunlei, Qin, Yifan, Zhang, Lili
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.02.2026
Subjects:
DCMOA
DCMOPs
DMOA
DMOPs
Dynamic response mechanism
Evolutionary algorithm
DCMOPs
DMOA
DMOPs
Dynamic response mechanism
DCMOA
Evolutionary algorithm
ISSN:0957-4174
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•Biased sampling using predicted similarity across multiple environments.•Improved prediction method applied as an initialization and search strategy.•Strategy to reuse evaluation budget from environmental detection.•Dual-population static optimizer for dynamic constrained multi-objective problems. The design of evolutionary algorithms for dynamic constrained multi-objective optimization problems is a significant research area, driven by the need for real-time adaptive solutions in complex real-world systems. To address this, this paper proposes DP-DCMOA, an algorithm that introduces a novel hybrid dynamic response by synergistically combining memory-based sampling and prediction-based generation. The algorithm’s novelty lies in its dual prediction approach. For the memory-based component, it leverages the predictable time-series properties of environmental similarity to guide a robust biased sampling from historical archives. For the prediction-based component, it employs an enhanced trajectory predictor to generate new, high-quality solutions. Furthermore, a key innovation is the integration of this trajectory predictor as a continuous search operator within the static optimizer, moving beyond its traditional role as a one-shot re-initialization tool. This dual-response system is supported by a specialized dual-population static optimizer. The effectiveness of the framework is validated through extensive experiments, demonstrating significant performance advantages over state-of-the-art algorithms and a high ranking in academic competitions.
ISSN:0957-4174
DOI:10.1016/j.eswa.2025.129304