A micro dynamic multi-objective evolutionary algorithm with flexible response strategy

doi:10.1016/j.swevo.2025.102243

A micro dynamic multi-objective evolutionary algorithm with flexible response strategy

2025-12-04

0

PRE

AI

DOI：10.1016/j.swevo.2025.102243

Original

Original request for help

Save

Abstract

En 中文

Resource constraints exist in solving realistic dynamic multi-objective optimization problems (DMOPs), such as those based on low-power microprocessors. However, traditional dynamic multi-objective evolutionary algorithms (DMOEAs) often rely on large populations, and running these algorithms directly on low-power microprocessors will result in program interruptions. In contrast, the micro population can effectively balance limited computing resources and computational efficiency, making it an effective approach for running DMOEAs on low-power microprocessors. In light of this analysis, a micro dynamic multi-objective evolutionary algorithm with flexible response strategy ( μDMOEA-FRS) is proposed. This approach incorporates a flexible niche strategy to maximize information exchange among niches, identifying the current environment based on niche performance and updating relevant information accordingly. Subsequently, the static optimization phase determines the most suitable environmental selection method based on insights obtained from the dynamic response phase. This strategy significantly enhances the adaptability of the micro population in dynamic environments. Additionally, a flexible scaling mechanism is introduced, which improves the diversity of the algorithm, facilitates the exploration of new regions within the solution space, and balances convergence with diversity. The performance of μDMOEA-FRS is compared against eight state-of-the-art DMOEAs across 35 test instances, demonstrating superior results in most cases. Furthermore, for the application to real-world problems, the algorithm was simulated within a small-scale smart greenhouse equipped with a low-power microprocessor. The results confirm the feasibility of μDMOEA-FRS for optimization within low-power microprocessor environments.

Keywords:

Dynamic multi-objective optimization problem

Micro dynamic multi-objective evolutionary algorithm

Flexible response strategy

Low-power microprocessors

A micro dynamic multi-objective evolutionary algorithm with flexible response strategy

Abstract

Journal

Organization