Researches Of Dynamic Spectrum Accessing In Cognitive Radio Network

Ubiquitous wireless communication has become the trend of communication network, inwhich the varieties of wireless communication technologies coexist in a networkenvironment. An important prerequisite to achieve ubiquitous is enough spectrum. With therapidly growth of wireless devices, the wireless spectrum resources are very scarce. Thisphenomenon becomes the new bottleneck which restricts the development of wirelesscommunication technologies. In recent years, cognitive radio (CR) technology has drawnextensive attention. CR is regarded as the key to solve the current low utilization ofspectrum resources. The Dynamic Spectrum Allocation is one of the importanttechnologies to realize cognitive radio.The thesis first briefly introduces the existing spectrum allocation methods and focuseson the theory of partially observable Markov decision process (POMDP). With POMDPtheoretical framework, we give the theoretical analysis and simulation of the performanceof spectrum allocation. On this basis, we put forward the channel correlation factor in theoverall design of the algorithm to further improve the performance. Then we give therelated definition and proof.Secondly, this thesis studies the multi-channel access problem in an unknown channelstates environment. In the existing researches, authors usually assume that the channel statetransition probability is given in advance. But in the actual network environment, the idlespectrum is time-varying and secondary users cannot respond quickly. So the transitionprobability cannot be obtained in advance and need estimate them iteratively. We study thesecondary users how to effectively establish the primary users’ spectrum usage model inthe unknown network environment and give the theoretical analysis and proof. In themodeling process, the physical layer spectrum sensing errors are considered. For the sakeof the limitations of the present studies that focus only on the single-channel access, themulti-channel method is proposed. The simulation figures out the compromise betweensensing time and access time. We also study the influence of the primary user servicearrival rate over the performance of the algorithm. Instead of maximize the throughput ofsecondary users in the previous optimization goal, we study the maximization of the spectral efficiency of entire network. The simulation shows that the primary users’collision probabilities have an optimal point to the spectral efficiency of entire network.Finally, for the current POMDP algorithms are limited to the area of single userperformance, we explore the MAC layer protocol in multi-user mode. The thesis points outthat the Greedy algorithm is only optimal for single-user and inappropriate for multi-user.Compare with the Random algorithm we find that the performance of the Greedyalgorithm degrade dramatically in the case of multi-user. For this reason, two optimizationalgorithms are proposed. Simulation results show that both of the two optimizationalgorithms improve the performance better than the Greedy one does.