.png){kind=icon}
.png){kind=icon}
Return
A collaborative evolution strategy for dynamic multi-objective optimization based on charged individual prediction
.png){kind=icon}
PRE
AI
DOI：10.1016/j.swevo.2026.102386.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In dynamic multi-objective evolutionary algorithms (DMOEAs), prediction-based methods have attracted substantial attention due to their ability to rapidly track the Pareto-optimal front (POF) after environmental changes. However, such methods usually rely on an implicit assumption that the difference in population distribution between adjacent environments is small, and perform prediction based on historical information under this premise. This assumption is often difficult to satisfy in practical applications; when the environment changes drastically, the population distribution may be severely disturbed, which in turn leads to significant prediction bias. To mitigate the prediction bias caused by abrupt changes in population distribution, this paper proposes a collaborative evolutionary strategy for dynamic multi-objective optimization based on charged individual prediction (denoted as CP-CEMS). By introducing a charged repulsive-force mechanism among individuals, this strategy enables individuals to maintain appropriate distances during the prediction process, thereby effectively preserving the population distribution after environmental changes. In CP-CEMS, the population is divided into a guiding population and a tracking population. The guiding population exhibits a well-structured distribution and maintains its distribution structure through the charged individual prediction mechanism, whereas the tracking population performs prediction under the synergistic guidance of the guiding population, thereby improving the convergence stability of the overall evolutionary process. Subsequently, an elite mining strategy is adopted to integrate superior individuals from the two predicted populations to construct the initial population in the new environment. Experimental results under multiple parameter settings show that CP-CEMS significantly outperforms existing methods in solving dynamic multi-objective optimization problems (DMOPs).
Keywords:
charged individual prediction
dynamic multi-objective optimization
population distribution
evolutionary algorithms
collaborative strategy
Journal
IF:
8.5
Papers:
2.1K
Citations:
1.0W