基于强化学习的无人机辅助物联网抗敌意干扰算法  

Anti-jamming Algorithm with Reinforcement Learning in UAV-Aided Internet of Things

作　　者：张孟杰 赵睿[1] 王培臣 周洁 ZHANG Mengjie;ZHAO Rui;WANG Peichen;ZHOU Jie(Xiamen Mobile Multimedia Communication Lab,Huaqiao University,Xiamen,Fujian 361021,China)

机构地区：[1]华侨大学厦门市移动多媒体通信实验室,福建厦门361021

出　　处：《信号处理》2021年第1期11-18,共8页Journal of Signal Processing

基　　金：国家自然科学基金(61401165);福建省自然科学基金(2019J01055);华侨大学研究生科研创新基金(18013082036)。

摘　　要：无人机充当中继辅助物联网节点传输信号时,易遭受干扰强度动态变化的智能干扰等敌意攻击,论文建立了抗敌意干扰攻防Stackelberg博弈模型,其中物联网节点、无人机和智能干扰机为博弈的3个参与者,推导出博弈均衡点及其存在条件,揭示了参与者的信道增益、距离等参数对物联网效用等性能的影响。在未知干扰模型的条件下,论文引入了强化学习算法动态优化物联网节点的发射功率、无人机的发射功率和移动轨迹,有效的提高了系统抗干扰性能。仿真结果表明,与基于Q_learning的算法相比,基于WoLF-PHC的算法将无人机的效用提升了84.8%。The Internet of Things node can transmit messages aided by the unmanned aerial vehicle who acts as the relay node,which are vulnerable to be attacked,such as the jammer who can change the interference intensity randomly.A Stackelberg game model of attacking and defensing anti-jamming was investigated in this thesis,the Internet of Things node,the unmanned aerial vehicle and a jammer were the three players in the game,the Stackelberg equilibriums and its existence condition also were derived,which revealed the influence of channel gain,distance and other parameters on the utility of the system.The reinforcement learning algorithm were used to dynamically optimize the transmit of the Internet of Things node and the unmanned aerial vehicle and the moving trajectory about the unmanned aerial vehicle without known the interference model,which can effectively improve the anti-jamming performance of the system.The simulation results show that the algorithm-based WoLF-PHC increases the utility of the unmanned aerial vehicle by 84.8%compared with the algorithm-based Q_learning.

关 键 词：无人机 抗敌意干扰 博弈论 强化学习 物联网 

分 类 号：TN918.91[电子电信—通信与信息系统]