Anti-Intelligent UAV Jamming Strategy via Deep Q-Networks

The downlink communications are vulnerable to intelligent unmanned aerial vehicle (UAV) jamming attack which can learn the optimal attack strategy in complex communication environments. In this paper, we propose an anti-intelligent UAV jamming strategy, in which the mobile users can learn the optimal defense strategy to prevent jamming. Specifically, the UAV jammer acts as a leader and the users act as followers. The problem is formulated as a stackelberg dynamic game, which includes the leader sub-game and the followers sub-game. As the UAV jammer is only aware of the incomplete channel state information (CSI) of the users, we model the leader sub-game as a partially observable Markov decision process (POMDP). The optimal jamming trajectory is obtained via deep recurrent Q-networks (DRQN) in the three-dimension space. For the followers sub-game, we use the Markov decision process (MDP) to model it. Then the optimal communication trajectory can be learned via deep Q-networks (DQN) in the two-dimension space. We prove the existence of the stackelberg equilibrium. The simulations show that the proposed strategy outperforms the benchmark strategies.