Hierarchical relationship modeling in multi-agent reinforcement learning for mixed cooperative-competitive environments

Shaorong Xie, Yang Li, Xinzhi Wang, Han Zhang, Zhenyu Zhang, Xiangfeng Luo, Hang Yu

Abstract: Highlights•Importance analysis of heterogeneous relationships in mixed environments.•Modeling heterogeneous relationships using hierarchical relationship graph.•Local and global information fusion using local and global graph attention networks.•Relationship modeling combined with MARL to train heterogeneous strategies•Validation of the method in typical environments (Predator–Prey and Soccer Game).