Dynamic threshold spectrum sensing method based on DQN combined with clustered cooperative sensing architecture

In order to meet the needs of some scenes with unknown and rapidly changing noise power, a spectrum sensing method based on deep reinforcement learning is proposed in this paper to improve the traditional energy detection method. A complex reward function is designed in the deep Q network (DQN) algorithm, which can make the agent adjust the decision threshold of energy detection more intelligently. In addition, combined with the clustered cooperative spectrum sensing architecture, the performance of spectrum sensing is further improved through twice decision fusion based on cumulative accuracy. In this paper, five other common spectrum sensing methods are compared. The simulation results show that the proposed method converges on datasets, and its sensing performance is superior to the other methods. Its detection accuracy rates are 87.88%, 90.18%, 88.13%, 84.74%, 82.43% respectively. Thus, limitations of traditional energy detection methods are broken through this method and the whole system is more intelligent and stable.