Research Of Spectrum Access Based On Machine Learning In Cognitive Radio Networks

With the rapid development of wireless communication technology and the increasing demand of people for communication, the spectrum resource is in short supply. By finding the spectrum, cognitive radio technology can make full use of idle spectrum, and fundamentally solve the problem of low spectrum utilization. The research of cognitive radio has important theoretical and practical significance.This paper focuses on the dynamic spectrum access technology based on machine learning in cognitive radio. It mainly discusses the method of spectrum prediction and the scheme of channel selection in the process of spectrum access. Details are as followsFirstly, this paper presents a method of cooperative spectrum prediction based on artificial neural network. In the current study, there are two kinds of commonly used method of spectrum prediction, one is the method of spectrum prediction based on markov process, this method needs prior knowledge of primary user, which is difficult to get in dynamic network. Another one is the method of spectrum prediction based on artificial neural network, this method can use the observation information of channel's history state to train the structure of artificial neural network. And then the artificial neural network can establish the corresponding relationship between the history state and future state of the channel. Considering the secondary user's sensing results of channel and the information in the network, this paper use cooperative spectrum prediction method based on the artificial neural network to predict the state of channel. Simulation results show that this method improves the accuracy of spectrum prediction.Secondly, by studying the model of restaurant dinning, this paper establish the model of channel selection in cognitive radio. In this model,There is a file shared among secondary users, which record each secondary usesr's channel belief and their channel selection results. Based on this model, we can make more sense of factors that affect the communication performance of secondary users on the process of making channel selection.Thirdly, this paper proposes a channel selection strategy to maximize the long-term cumulative return in a scenario with multiple alternative channels. In the dynamic network, the data transmission rate of secondary user is not only related to the channel capacity, but also affected by the behavior of the other secondary users. In order to improve the communication performance in the whole data transmission process, the user takes full account of the channel access behavior of other users when selecting the channel. Simulation results show that this scheme makes positive effect on improving the throughput of the whole system.