Reinforcement Learning Based Anti-Jamming Algorithm

With the development of wireless communication technology,the number of wireless devices has also increased dramatically.The fifth generation mobile communication technology uses cognitive radio technology and new multiple access technologies to improve the efficiency of spectrum utilization in wireless systems.However,radio frequency jamming attacks can severely disrupt the availability of communication systems.The jammer can prevent legal nodes from accessing the communication network by transmitting illegal signals,and it can also interfere with ongoing communications and reduce the node's communication rate.How to prevent radio frequency jamming attacks in the new communication environment has become an urgent problem.Q-learning,which is a typical reinforcement learning algorithm,is model-free and is widely used by researchers in anti-jamming research.However,the performance of traditional Q-learning is not ideal when dealing with large-scale problems.In this paper,we introduce deep reinforcement learning and transfer learning,and redesign the algorithm model structure according to the characteristics of the anti-jamming problem.By applying deep reinforcement learning,the algorithm's ability to handle largescale problems has been significantly improved,while transfer learning can transfer the knowledge learned by agents in a jamming-free environment to anti-jamming problems.The algorithm's convergence speed is thus significantly accelerated.Experimental results show that the new algorithm has a faster convergence speed than traditional Q-learning and can be used in scenarios where there are continuous variables in the state and actions.At the same time,the jammer can also be equipped with reinforcement learning algorithms,which requires us to study the multi-agent reinforcement learning problem.Compared with the single agent problem,the system is more dynamic and unpredictable.Therefore,this paper further studies the multiagents anti-jamming problem in non-orthogonal multiple-access communication systems.Evolutionary game theory is used to study the equilibrium of the system and the strategies of each agent at the equilibrium.The simulation results are consistent with the theoretical results derived from evolutionary game theory.