.png){kind=icon}
.png){kind=icon}
Return
A population game-based knowledge transfer strategy for constrained multi-objective optimization
.png){kind=icon}
PRE
AI
DOI：10.1016/j.swevo.2025.102146.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In constrained multi-objective optimization problems (CMOPs), complex constraints may result in narrow feasible regions or cause the Pareto front to lie on constraint boundaries, which significantly increases the difficulty of locating feasible solutions within limited computational resources. Evolutionary multitasking optimization algorithms promote the optimization of the main task by introducing auxiliary tasks. However, even when the contributions of these auxiliary tasks diminish, they continue to consume computational resources. To address this issue, this study proposes a population game-based multitasking coevolutionary algorithm. The algorithm models the original CMOP as a multitasking optimization problem comprising two tasks. Specifically, the target task explores the feasible region of the original CMOP by evolving a population. Meanwhile, the source task is activated dynamically through a population game mechanism, aiming to explore potential feasible regions by relaxing the constraints. Through knowledge transfer, the supplementary evolutionary directions obtained from the source task provide unexplored paths for the target task, guiding the population to approach the Pareto front from both feasible and infeasible directions. Comprehensive experiments were performed on four benchmark suites. The experimental results demonstrated that the proposed algorithm exhibited competitive or superior performance compared with eight state-of-the-art algorithms.
Keywords:
constrained multi-objective optimization
multitasking evolutionary algorithm
population game
coevolutionary algorithm
feasible region exploration
Journal
IF:
8.5
Papers:
2.1K
Citations:
1.0W