Evolutionary Enhanced Multi-Agent Reinforcement Learning for Cooperative Air Combat
The paper presents a new framework for multi-agent reinforcement learning in cooperative air combat scenarios. It introduces the Adversarial Curriculum and Evolutionary-enhanced Multi-agent Proximal Policy Optimization (ACE-MAPPO) to address challenges faced by unmanned combat aerial vehicles. Experimental results indicate that ACE-MAPPO outperforms existing methods in training stability and convergence speed.
- ▪The proposed ACE-MAPPO framework integrates evolutionary algorithms with multi-agent reinforcement learning.
- ▪A genetic soft update mechanism is introduced to enhance population diversity.
- ▪The method demonstrates improved training stability and win rates in multi-aircraft cooperative scenarios.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.25091 (cs) [Submitted on 24 May 2026] Title:Evolutionary Enhanced Multi-Agent Reinforcement Learning for Cooperative Air Combat Authors:Chengwei Li, Junlin Liu, Yang Gao View a PDF of the paper titled Evolutionary Enhanced Multi-Agent Reinforcement Learning for Cooperative Air Combat, by Chengwei Li and 2 other authors View PDF HTML (experimental) Abstract:As modern air combat evolves toward beyond-visual-range (BVR) multi-aircraft cooperative engagements, autonomous decision-making for unmanned combat aerial vehicles (UCAVs) faces significant challenges due to high-dimensional state spaces, discrete action commands, and strongly adversarial dynamic environments.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
