Research On Key Techniques Of HF Wideband Channel Selection Based On Cognitive Radio

Compared with the communications such as satellite communication,ground microwave and cable communication,high frequency(HF)has advantages of low maintenance cost,short construction period,simple equipment,convenient and flexible network,strong anti-destruction,interference and eavesdropping abilities,which show a huge potential of HF communication development in the future.Meanwhile,in HF communication the channel transmission quality is not stable,and heavy interference,limited frequency resources and other shortcomings limit the development of HF communication.Selecting a good quality and stable channel provides an effective way to overcome shortcomings of the HF communication,improving communication performance.However,how to select a reasonable channel from channels with unstable quality,heavy interference and limited frequency resources becomes an important challenge for the development of current HF communication.HF channel selection mainly involves two aspects,selections of link channel and traffic channel,respectively.The traditional HF link and traffic channel bandwidths are 3kHz,and channel selections only need to select the corresponding frequency from given frequencies with experience;Compared with the traditional HF channel selection,wideband link channel selection requires real-time spectrum sensing of channel occupancy state and channel propagation quality,and intelligently selects the link frequency through learning algorithms.However,the wideband traffic channel not only needs to consider the intelligent selection of channel frequency but also the channel bandwidth.In this dissertation,the cognitive theory is introduced into the HF wideband channel selection research to intelligently select the channel improving the HF communication performance.At present,there are some related researches on the intelligent channel selection with cognitive technology,including the channel selections based on Markov decision,reinforcement learning and optimal stopping principle.However,the research on HF channel selection is relatively few.The HF channel selection mainly uses the existing US military HF communication standard,and some researches have begun to introduce cognitive and wideband communication theory,but these researches are in the initial stage compared to other communication technologies.For the present research,there are some unresolved problems in HF channel selection,including: 1)how to quickly and accurately sense the HF spectrum;2)how to select the link channel under the condition of missing or unknown link channel quality information;3)how to select reasonable traffic channels for a variety of different services.In order to solve the above problems,the dissertation first researches the HF wideband spectrum sensing,and then researches the selection of wideband link channel for the cases of lack and unknown link channel quality information,respectively.Finally,the selection of wideband traffic channel is researched.The main research results are as follows:1)Aiming at the compressed spectrum sensing under the dynamic variation of HF spectrum sparsity,an adaptive compressed spectrum sensing algorithm is proposed based on cross validation.First,by the Johnson-Lindenstrauss(JL)theorem and the cross-validation theory,it is proved that the recovery error of the wideband signal can be estimated by the recovery error of testing measurement when the row number of the compressed sensing matrix is greater than a lower bound.Secondly,an adaptive compressed spectrum sensing algorithm based on cross validation is proposed by using the signal recovery error as the stop rule of adaptive compressed spectrum sensing sampling rate changing.Finally,parameters in the algorithm are optimized to maximize the system throughput.Simulation results show that the adaptive compressed spectrum sensing algorithm not only effectively saves sampling resources and sensing time but improves the system throughput compared with the traditional compressed spectrum sensing and two-step compressed spectrum sensing.2)Aiming at the link channel selection under with missing channel quality information,a link channel selection algorithm is proposed based on matrix completion.Firstly,the automatic link establishment(ALE)process is formulated into the Markov chain with discrete time,and the method of selecting the optimal link channel is obtained by analyzing the Markov chain.Secondly,the matrix completion method is introduced to complete missing elements in the link quality analysis(LQA)matrix to recover the channel quality missed in LQA matrix by the channel detection failure.Using the alternating least squares,the near optimal solution of matrix completion is obtained and the convergence of solution is proved.Finally,an algorithm for HF asynchronous ALE link channel selection is proposed based on the completed LQA matrix.Simulation results show that compared with the algorithm used in the current HF ALE standard,the proposed algorithm can select the channel with better channel quality,reduce the link establishment time and increase the throughput of data transmission.The performance is close to that of link channel selection algorithm based on the ideal LQA matrix.3)Aiming at the link channel selection with unknown channel quality information,a link channel selection algorithm is proposed based on multi-armed bandit(MAB).Firstly,the time and energy cost in the link establishment process is regarded as the switching cost of link channel selection,and with the minimum outage probability as the target reward function,the selection of link channel is modeled as MAB problem with switching cost.Secondly,a block MAB(BMAB)link channel selection algorithm is proposed to reduce the switching cost and outage probability.In order to reduce the influence of deep fading on HF communication,link channel selection algorithm based on RTDMAB is proposed with the introduction of repetition time diversity(RTD)into BMAB algorithm.The regret functions of BMAB and RTDMAB algorithms are proved to be logarithmical with time.Finally,the performances of these two algorithms and slot MAB(SMAB)algorithm are analyzed theoretically.The results show that the proposed algorithm outperforms SMAB in the average system outage probability,switching cost and throughput performance.And the performance of the RTDMAB algorithm outperforms BMAB algorithm when the selected channel is in deep fading.The simulation results further validate the conclusion of the algorithm performance analysis.4)Aiming at the traffic channel selection with different requirements of quality of service(QoS)in HF wideband service,traffic channel selection algorithms are proposed.The objective function of HF traffic channel selection is designed,and the traffic channel selection oriented with HF service quality is modeled as the joint selection of frequency and bandwidth.With known channel statistical information,the optimal frequency and bandwidth selection of maximizing objective function is given by greedy algorithm.When the statistical information of channels is unknown,a wideband traffic channel selection algorithm is proposed based on Q-learning.In this algorithm,the objective function of HF service is taken as the reward,and the joint frequency and bandwidth selection is taken as the learning strategy.The optimal strategy is selected by the exploration and exploitation of channel occupancy and propagation quality information.The simulation results show that there are different optimal bandwidth and frequency selections for different services with known channel statistical information,which verifies the necessity of wideband traffic channel selection research.When the statistical information is unknown,the traffic channel selection algorithm based on Q-learning can select the optimal channel to meet the required quality of service with relative huge probability.