Study On Dynamic Spectrum Access Based On Decision Theory For Cognitive Radio Networks

The rapid development of wireless radio communicaiton technologies makesthe spectrum resources becoming more and more scarce. Especially that peopleput forward higher requirements for wireless applications with the improvementof the living level. On the one hand，in the process of the investigation theresearchers find that the spectrum utilization of the licensed bands is very low，especially in the spectrum band from several hundred megahertz to3GHz. On theother hand，business capacity on a relatively small number of non-authorizedspectrum resources is quite large. Limited available spectrum and low spectrumresource utilization ask for a whole new wireless communication mode tooptimize the usage of the spectrum. Cognitive Radio technology changes theoriginal spectrum of fixed allocation. Since Cognitive Radio allocates thespectrum dynamically and make use of the spectrum resource in and morereasonable way. It has greatly improved the utilization of the spectrum. As one ofthe key technologies， dynamic spectrum access technology shows the mostessential requirements of Cognitive Radio.This paper talks about the concept and research status of Cognitive Radio atfirst, and summarize the existing dynamic spectrum use model which is dividedinto three categories. This paper also describes the cognitive radio networkarchitecture which mainly includes the centralized network and distributednetwork architecture. Then focus on dynamic spectrum access systems in adistributed network model. Primary user channel model is modeled as a discreteMarkov model and cognitive users need to select the access policy on a discretetime point. The paper also describes three dynamic spectrum access strategybased on decision theory.Cognitive user needs to make a decision that whether to access immediatelyafter detecting the channels, that is one option and access policy. Themulti-armed bandits model is suit for estiblish the model of choosing one frommany available options. Therefore, in the third chapter combines the multi-armedbandits model with the once accessing process, then use the upper confidencebound (Upper Confidence Bound) algorithm to solve this model. Thereafter, onthe one hand, due to the cognitive user decisions for spectrum access is asequential process, on the other hand, aware of the limited accuracy the cognitiveuser perception of the state of the channel is not completely accurate, so paper combines partially observable Markov decision process for the discrete-timepoint cognitive the user spectrum access process modeling, and to find theoptimal strategy with suboptimal strategy.