Research On Spectrum Analysis And Decision Technology Of Cognitive Wireless Communication

With the rapid development of wireless communication technology,the contradiction between the shortage of spectrum resources for fixed allocation and the low spectrum uti-lization of some allocated frequency bands has become increasingly prominent.Thanks to its better flexibility,cognitive radio technology is considered to be able to resolve this con-tradiction,that is,under the condition of occupying limited exclusive spectrum resources,it reasonably and timely utilizes the idle spectrum to meet the growing frequency requirements and brings as little interference as possible to the primary users.Generally speaking,cognitive radio technology includes spectrum sensing,spectrum analysis and decision,spectrum sharing,and spectrum mobility management.Based on the spectrum sensing results(data),the spectrum analysis and decision technology first analyzes and studies the multi-dimensional availability of time,space,and frequency of spectrum resources.Then by using the analysis conclusions,it formulates a corresponding cognitive access strategy with a certain optimization goal to provide the strategy basis for the access process of the cognitive system.This thesis focuses on the spectrum analysis and decision-making of cognitive radio,and analyzes the availability of the space-time-frequency dimen-sion of the shared spectrum by prediction.Besides,it also conducts a resource allocation decision research based on the results of spectrum sensing and analysis to address the key technical problems encountered in the process of cognitive wireless communication.The main contributions of this article are summarized as follows:First,this thesis studies and addresses the prediction problem of spectrum with bursti-ness in the time-domain.Specifically,by using the spectrum analyzer and other hardware devices,we first conduct time-domain data collection for frequency bands which are oc-cupied by competing ways,and preprocess the data accordingly.Then,the time-domain burstiness of the frequency band is characterized through the analysis of various channel characteristics such as occupancy,self-similarity,channel entropy,space time variance,and autocorrelation.Finally,a multi-layer perceptron method based on reinforcement learning is proposed to address the matching problem between different spectrum prediction models and different states corresponding to the different time periods of the spectrum.Experiment results show that the proposed method has a significant accuracy improvement(over 7%)compared to the commonly used spectrum prediction methods,and the method could meet the prediction needs of frequency bands with high burstiness in the time-domain.Second,this thesis studies the indirect construction technique of airspace spectrum map based on the localization of multiple directional primary users using received signal strength(RSS)measurements.In order to satisfy the need of constructing a spectrum map of a cognitive access space area where multiple directional primary users exist,we consider using the indirect method.Specifically,we first establish the model of multiple directional source positioning and parameter estimation based on the RSS measurements,and derives the Cramer-Rao bound which shows the unbiased errors of estimating the positions,trans-mission directions,transmission powers,beam widths of the primary users,and the path loss exponent of the environment.Then,an expectation maximization method on the basis of maximum likelihood is proposed to achieve the goal of multiple parameter estimation.Finally,the construction of the spectrum map of the area of interest is completed based on the estimation results of the parameters.Simulation results show that the proposed method can achieve lower positioning and parameter estimation errors.Compared with the com-monly utilized direct spatial interpolation methods,the indirect method used in this thesis has better performance.Third,this thesis studies the power intensity prediction method based on time slot and multi-channel selection strategy in frequency domain.First,aiming at the problem of spec-trum prediction with a certain historical dependence,a method based on long short-term memory network is proposed to predict the channel power intensity.The method could ef-fectively perform extraction,transfer and mapping of the important features in historical data.Next,the initial occupancy of each channel is modeled as a Markov process,and its optimal spectrum sensing time interval is calculated as the predicted length of the cor-responding channel.Finally,a channel preferred usage scheme is proposed.The scheme takes the potential throughput of each channel as a performance index,ranks the channels and recommends them to the cognitive system,and realizes the use of the optimal channel.Simulation results show that the proposed prediction method has better power intensity ac-curacy than the multilayer perceptron method,and the proposed scheme has a significant performance improvement in terms of throughput and energy consumption compared to the commonly used channel selection scheme.Fourth,this thesis studies and addresses the multi-dimensional resource allocation prob-lem of downlink communication for cognitive satellites based on the frequency recommen-dation of ground terminals.Specifically,a new beam hopping scheme is first proposed to address the problem of diverse recommended channels caused by the plentiful recommended frequency channels of many ground terminals.Then,the four-dimensional resource alloca-tion problems including frequency,time,power,and spot beam are respectively summarized with the goals of maximizing the throughput and minimizing the variance of demand and supply.The former goal considers merely the overall performance of the cognitive system,while the later goal takes into account both the performance and fairness.Finally,by di-viding the original problem into multiple sequential sub-problems,i.e.frequency selection,spot beam division,and time power allocation,and separately applying Lagrangian duality and heuristic methods,the reasonable allocations of the resources with the two goals are completed.Simulation results show that the proposed methods can well complete the re-source allocation goals and significantly improve the performance of the cognitive satellite communication system.