Research On Dynamic Spectrum Allocation Methods Based On Deep Reinforcement Learning

With the rapid development of wireless communication technology,people's demand for spectrum resources is increasing day by day.The scarcity of spectrum resources is also becoming more and more serious.Therefore,it is urgent to propose new intelligent methods to improve spectrum utilization.Constructing a cognitive wireless network is an effective solution.In the face of the dynamic nature of complex environments,cognitive users can adjust behavior and control strategies accordingly,thereby making more efficient use of spectrum resources and improving the cognitive frequency efficiency of network systems.Spectrum resource management is one of the basic tasks of cognitive wireless networks,covering two core issues of power control and channel allocation.The power control refers to that the cognitive user in the network can adjust the transmit power to access the licensed frequency band in an opportunistic manner without causing interference to the authorized user.Therefore,the cognitive user can share the spectrum resource with the authorized user.The channel allocation is to allocate available channels of a certain period to the cognitive user reasonably,which improves the utilization rate of the idle spectrum resource.Due to the widespread application of cognitive wireless networks,the network structure is becoming more and more complex.It is difficult to establish a corresponding mathematical model to simulate a highly complex network environment.The model-free reinforcement learning algorithm can effectively solve this problem.As the rise of deep learning in recent years,deep reinforcement learning has shown excellent ability in dealing with complex problems and data operations.Therefore,this paper focuses on the application of intelligent algorithms for deep reinforcement learning in spectrum resour ce management in cognitive wireless networks,especially the optimization of power control and channel allocation to improve the robustness and adaptability of cognitive users in dynamic and complex wireless environments.The main contents of this paper including as follows:1.The problem of spectrum resource allocation in cognitive wireless networks is introduced.Besides,two main research problems of power control and channel allocation in dynamic environment are analyzed.2.The application of deep reinforcement learning in dynamic power control is studied and analyzed.The dynamic power control strategy of dueling DQN with prioritized experience replay is proposed,which enables cognitive users to quickly adjust control strategies in complex dynamic environments.The results show that the spectrum utilization and network energy efficiency are improved.The simulation experiments are carried out on the TensorFlow deep learning framework to verify the feasibility and superiority of the proposed method.3.Considering the joint problem of power control and channel allocation based on point 2 above,a spectrum resource allocation algorithm for long short term memory deep Q network is proposed.Cognitive users can adaptively adjust the transmit power while successfully accessing the channel and the utilization of spectrum resources is improved.The simulation experiments are carried out on the TensorFlow deep learning framework to verify the feasibility and superiority of the proposed method.