Dynamic Spectrum Anti-Jamming for Cognitive UAV Networks Against Reactive Jamming

This paper investigates the anti-jamming (AJ) spectrum allocation problem in a cognitive unmanned aerial vehicle (UAV) network scenario, and the confrontational interaction between communication and jamming is modeled as a Stack-elberg game. Different from most studies which assume that radio jamming has fixed spectrum patterns (such as swept jamming and comb jamming) without sensing ability, we propose a deep reinforcement learning (DRL) based dynamic spectrum AJ method to cope with a reactive jammer and the unintended interference caused by a primary user. To improve the online learning ability, a DRL approach called DeepMellow is used. Simulation results demonstrate that the proposed algorithm can find an effective communication strategy to increase the communication utility in the complex spectrum environment.