Research On Spectrum Prediction And Access Technology In Cognitive Radio

With the rapid development of wireless communication,more and more devices are connected to wireless networks.As a result,wireless spectrum is becoming an increasingly important and scarce resource,but current static spectrum allocation strategy can not make full use of the wireless spectrum.Cognitive radio technology is proposed under this background,which can solve the problem of spectrum resource shortage.The spectrum prediction and access technology are the key technologies in cognitive radio.Spectrum prediction technology is to predict the future spectrum information by learning the historical spectrum information.Spectrum access technology is that unauthorized users complete spectrum access efficiently according to the channel historical spectrum data and the predicted information,and make full use of the idle spectrum resources.Based on this background,this paper first gives a detailed introduction to spectrum prediction technology and access technology,and then investigates the spectrum prediction problem in bursty frequency bands and the spectrum access problem in the single-user multi-channel scene,finally the main contents and innovations of the paper are respectively shown below:For the spectrum prediction problem of bursty frequency bands,this paper innovatively designs a multi-classifier spectrum prediction algorithm based on deep reinforcement learning.Firstly,we collect ISM frequency band data with high burstiness for analysis,characterize the data burstiness by using the features of spectrum data idleness,self-similarity,predictability,and self-correlation.Secondly,according to the characteristics of the spectrum data classification training to get a multi-category prediction model.Then the deep reinforcement learning algorithm is used to complete the matching between different states and different prediction models,and finally the corresponding predictors are selected in different states to complete the prediction.At the same time,the algorithm is further optimized in two aspects.On the one hand,the classification process of spectrum data is optimized based on the prediction results.On the other hand,aiming at the problem of fixed grouping,a window sliding idea is proposed to carry out dynamic grouping.The proposed algorithm is implemented and validated on the collected spectrum data set.The simulation results demonstrate the superior performance of the proposed algorithm compared with several common spectrum prediction methods,and the prediction accuracy of the proposed algorithm is improved more than 6%,which achieves good prediction results for bursty bands and can meet the spectrum prediction requirements of ISM bands.For the multiple channel spectrum access problem of unauthorized users,a multi-channel sense access algorithm based on spectrum prediction is proposed in this thesis.Firstly,the channel priority coefficients are calculated by combining the historical spectrum data and spectrum prediction results of multiple channels.Secondly,according to the descending order of channel priority coefficient,multiple channels are perceived in order.Finally,the optimal stopping theory is used to make efficient decisions on when to stop sensing to obtain the maximum reward.The simulation results show that the proposed algorithm has significant performance advantages over traditional algorithms in terms of throughput,sensing energy consumption and energy efficiency,and proves that the proposed algorithm can maximize the use of authorized users' idle time to complete spectrum access to improve spectrum utilization.