MARL-based Cooperative Jamming and Target Protection in UAV Swarms

UAVs are increasingly employed in complex electromagnetic environments to suppress unauthorized signals and protect critical assets. However, achieving efficient coordination among multiple UAVs remains a key challenge in such dynamic scenarios. Multi-Agent Reinforcement Learning (MARL) provides an effective framework for developing adaptive collaborative strategies. This paper presents a cooperative mission scenario where multiple UAVs implement coordinated electromagnetic disruption and protection protocols. Under this framework, a Protection Scenario Curriculum-based staged LearningMADDPG (PCS-MADDPG) algorithm is proposed to optimize motion synchronization between the interfering UAV swarm and the protected asset. Experimental evaluations demonstrate that the PCS-MADDPG algorithm achieves improved training efficiency and enhanced cooperative performance in dynamic electromagnetic regulation tasks.