.png){kind=icon}
.png){kind=icon}
Return
Graph based multi-agent reinforcement learning with evolutionary population for cooperation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neunet.2025.108437.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Proposes a staged evolutionary MARL framework optimizing joint team policies through evolutionary population-based gradient-free exploration. • Integrates evolutionary algorithms and graph neural networks within a MARL framework, enabling efficient global exploration and local policy refinement. • Employs spectral normalization to stabilize Critic network training, significantly reducing evaluation instability in complex dynamic environments without restricting policy exploration.