Research On Multi-domain Cognition Techniques In Cognitive Radio Networks

As the kernel study of next generation networks, Cognitive Radio Network(CRN) is proposed as an active network with the characteristic of cognitive with the inspiration of the technology of cognitive radio. A cognitive radio network is a network with cognitive process that can perceive current network conditions. Then it can make planning, decision and act on those conditions. Since that, it is considered to be the inevitable trend of future communication network. Environment cognition of cognitive radio networks extends from wireless environment to network environment and user environment. Therefore, cognitive radio networks need to solve the awareness of multi-domain environment which is called as technique for multi-domain cognition. Compared with network and user environment, wireless environment cognition is more complicated and difficult, especially the spectrum environment cognition. The multi-domain cognition usually obtains and processes the awareness information of multi-domain environment from local multi–domain cognition, cooperative multi-domain cognition and initiative multi-domain cognition. Researching into techniques for multi-domain cognition and putting forward the cognitive approach for each layer have significant importance for the improvement and development of cognitive radio networks.Based on the three-layered theoretical framework of multi-domain cognition, this thesis focuses on some key multi-domain techniques, including the spectrum sensing in local multi–domain cognition and learning algorithms in initiative multi-domain cognition, and presents the corresponding cognitive approaches. Then the cognitive process that integrates spectrum sensing, environment learning and statistical reasoning can be completed. The main contributions of this thesis are as follows:Firstly, a new spectrum sensing algorithm based on the correlation among the sampled signals is proposed. For the spectrum sensing methods, the strengths and weaknesses of different conventional detection algorithms are discussed and analyzed in detail. On the basis of the classical sensing algorithms, the detailed implementation processes of the new proposed algorithm is provided to deal with the detection problems in low SNR. By considering multi-path fading channel model, high auto-correlation among the samples obtained is introduced by over-sampling at the receiver. Then two different test statistics are obtained to make local decision separately according to the distribution of correlation information under binary hypothesis. Thereby, two efficient sensing schemes are adopted in the proposed algorithm. Then the sensing performance of the algorithm is verified by the simulation result in different conditions.Secondly, a learning algorithm in initiative multi-domain cognition based on Bayesian Network(BN) is proposed. By establishing the corresponding cognitive Bayesian network model, the algorithm introduces Bayesian structure learning methods into cognitive radio networks. Then by learning and reasoning the sensing results through the knowledge of probability theory, the statistical relationship among primary users is studied. The conditional probability table established in the proposed algorithm can overcome the drawbacks of the current learning methods which cannot be adaptively adjusted to the network when the number of variables changes. Moreover, the proposed algorithm simplifies the conditional mutual information function and gives the simplified expression form of the conditional probability based dependence. Simulation results indicated that the proposed method has much less computational overhead than the conditional method. The dependence values and conditional probability table are also given in this thesis, from which the statistical pattern of the primary networks with low complexity can be established. And these results as the prior knowledge are quite valuable to forecast the future network activities and make decision in cognitive radio networks.