Research On Dynamic Spectrum Access Technology Of Cognitive Radio

With the rapid development of the information and communication technology industry,especially the large-scale commercial use of 5G,people's demand for spectrum resources is rapidly increasing.However,the existing spectrum resource allocation mainly adopts a fixed authorized method,which is hard to meet current and future wireless communication systems.In order to alleviate the scarcity of spectrum resources,the dynamic spectrum access technology based on cognitive radio has emerged.For the complex and changeable cognitive radio network environment,dynamic spectrum access technology can allow secondary users to occupy idle spectrum of the surrounding space for communication and transmission without affecting the communication of the primary user as much as possible,which provides the possibility for "secondary utilization" of the spectrum and satisfy the increasing demand for wireless transmission.Therefore,further research on dynamic spectrum access technology has long-term significance and prospects for the development of information and communication technology industry.This paper focus on the dynamic spectrum access strategy,the specific content is as follow:(1)For the dynamic spectrum access model of a single secondary user,a dynamic spectrum access strategy based on the sample-efficient reinforcement learning is proposed.According to the perceived spectrum state information,this strategy combined with the feature that the primary user has priority to use the authorized channel in cognitive radio network,constructs an reinforcement learning model suitable for dynamic spectrum access,and optimizes the value function update formula based on state-action pairs.The simulation results show that this strategy can reduce the interference of the secondary users to the communication transmission of the primary users,and increase the spectrum access opportunities of the secondary users.(2)For the dynamic spectrum access model of multiple secondary users,a dynamic spectrum access strategy based on double deep loop reinforcement learning is proposed.This strategy uses centralized spectrum sharing to avoid interference between secondary users,employs fully connected layer to directly estimate the channel allocation scheme to aviod the value function look-up table,and adopts a long short-term memory network to improve the performance of sequential channel allocation decisions.The simulation results show that this strategy can quickly adapt to the dynamic spectrum access scenarios of multiplye secondary users,so that the secondary users have more spectrum access opportunities,and it has a certain robustness.