A fuzzy learning anti-jamming approach with incomplete information

This letter investigates the power control anti-jamming problem with dynamic incomplete channel state information (CSI). To cope with the challenge of existing methods in obtaining steady-state solutions with dynamic incomplete CSI, the incomplete information is mapped to a fuzzy space, and then the payoffs of the user and the malicious jammer are expressed by fuzzy numbers. Further, to formulate the continuous adjustment process of the strategies of adversarial parties, an anti-jamming fuzzy Stackelberg game model is proposed and it is proven that fuzzy Stackelberg equilibrium (FSE) exists regardless of whether the user or the jammer is the leader. Finally, a fuzzy learning algorithm based on Q-learning is introduced to obtain the proposed FSE, which evaluates the fuzzy payoffs by selecting a suitable viewpoint and utilizing a satisfaction function. The simulation results demonstrate that the proposed approach effectively addresses the impact of CSI dynamic uncertainty, gets rid of the reliance of existing methods on deterministic information, and still guides game participants to optimize strategies in dynamic uncertainty environments. Compared with traditional learning algorithms, the results of the proposed method are satisfactory.