This letter investigates a jamming problem in the confrontation scenario when both the opposites are intelligent. Most existing studies assumed the learning-based jammer could obtain the communication feedback as the… Click to show full abstract
This letter investigates a jamming problem in the confrontation scenario when both the opposites are intelligent. Most existing studies assumed the learning-based jammer could obtain the communication feedback as the jamming reward, which may be infeasible and limits the application. To overcome this limitation, we propose a jamming efficacy evaluation method based on the change of communication behaviors, which is oriented by the idea to reflect the trend of the real jamming effect approximatively. Considering the rival is intelligent to adjust the communication strategy dynamically, multi deep-learning classifiers are established by the jammer to predict the channel status and make the jamming channel strategy selectively, aiming to disturb the rival’s normal communication. Moreover, the jamming power strategy is optimized by multi-armed bandit (MAB) method within the energy constraint. Simulation results demonstrate the effectiveness of the jamming efficacy evaluation criteria and show that the proposed algorithm can decrease the rival’s throughput as well as saving the jamming cost.
               
Click one of the above tabs to view related content.