Key Research On Spectrum Capacity And Spectrum Traffic Modeling In Cognitive Radio Networks

Spectrum is the basic resource in wireless communications. However, spectrum is underutilized severely due to nowadays’static regulatory policies and become a bottleneck of new wireless services’ proliferations. In Cognitive Radio Networks (CRN), cognitive users can sense wireless environments and dynamically accesses the idle spectrum unoccupied by the licensed users. Hence, CRN has gained great popularity as a promising technology to conciliate the currently inefficient spectrum utilization problem. CRN has important theoretical and practical significances to solve the spectrum underutilization problem caused by rapid wireless communication developments.Research on spectrum capacity on physical world will provide guidelines for CRN deployments, while correct understanding of spectrum traffic models is the basis of cognitive access strategy design. Hence, in this thesis we concentrate on spectrum capacity and spectrum traffic modeling based on machine learning and game theory models assisted with scientific measured spectrum data. The main innovative achievements carried out in this dissertation are as follows:(1) A new spectrum measurement method based on Fast Fourier Transform (FFT) is proposedThe FFT method can solve traditional spectrum measurement tradeoffs between time and frequency resolutions in the context of CRN, and increase spectrum sensing probabilities of burst type licensed users.(2) Digital dividend spectrum capacity in China is evaluated quantitatively Digital dividend distribution differences between urban and rural areas in china are different to those of developed countries. Through exploiting urban area’s digital dividend spectrum at first, urban areas can better play a leading and exploratory role in spectrum management reform. Field strength is proposed to be threshold in spectrum sensing algorithms, and its feasibility is also studied combined with a TV database. We also propose a set of cognitive receiver design principles to satisfy the rigid requirements of FCC.(3) A novel CRN model based on machine learning is proposedFirstly, we propose a CRN model based on machine learning. And a learning algorithm based on Artificial Neural Network is proposed to learn different spectrum traffic patterns. Future spectrum statuses are well predicted by GSM900/1800and TV signals of the first day in a week. The model has practical meanings for cognitive users to learn multiple domain spectrum dynamics from massive spectrum data.(4) Spectrum handoff is built as a game theory problem based on Radio Environment Map (REM)According to constant and predictable TV spectrum occupancy patterns in Beijing area, we propose a game theory model for spectrum handoff problem between cognitive users with the REM. The model can increase the SUs’access success rates, meanwhile decrease handoff probabilities.(5) Unified spectrum traffic model is proposed. And we first evaluate the existence of GSM white space.Unified spectrum occupancy model with Beta distribution is proposed and evaluated with empirical data. The distribution can well capture and reproduce actual spectrum statistical properties. The existence of GSM900white space is also evaluated as much as21.4MHz. We also propose an Efficient Duty Cycle (EDC) model. And its feasibility is cross-validated to decrease conflict probabilities between licensed users and cognitive users.Some issues include spectrum capacity and spectrum traffic modeling in CRNs are researched in this dissertation. And digital dividend spectrum capacity in china is also evaluated by our developed spectrum measurement platform. New ideas and methods are proposed with machine learning and game theory models in CRNs. And our platform and models will provide fundamental results in increasing spectrum utilization and promoting spectrum management reforms.