Anti-Jamming Strategy Based on Reinforcement Learning with Sequence Information

In this paper, we present an anti-jamming strategy based on an actor-critic model of reinforcement learning, which uses sequence information as model input to improve learning performance compared to the conventional structure. Conse-quently, in scenarios with enemy's partial-band jamming, we demonstrate that frequency hopping for evasion can be effectively performed by determining the frequency channels based on the actor-critic model, thereby efficiently mitigating the effects of enemy's jamming.