Research On High-speed Railway Channel Optimization Strategy Based On Cognitive Base Station

As a new research object in the communication industry,cognitive radio is increasingly recognized by people.Facing the low spectrum utilization rate caused by the fixed spectrum allocation method and the shortage of spectrum resources caused by the growing railway communication services in the high-speed railway environment,the introduction of cognitive radio into the high-speed railway environment is both a major innovation and a major difficulty.How to quickly and accurately connect secondary users(referred to as various passenger communication terminals in this thesis)in cognitive radio to the private network spectrum of the high-speed railway environment,it is needed to start by analyzing the state of spectrum occupied by primary users(referred to as train control equipment in this thesis),according to the occupancy status of the primary users,the available channels are analyzed and the secondary users are connected.Therefore,this thesis proposes a spectrum prediction method based on the occupancy status of the primary users.Through predicting and combing with an effective channel allocation method,the secondary users can have better selection when they are facing the current channel allocation.(1)After introducing the research background and significance,then,high-speed railway communication environment is introduced.Firstly,the basic architecture and performance characteristics of the LTE-R(Long Term Evolution-Railway)network are introduced.Compared with GSM-R(Global System for Mobile Communications-Railway),LTE-R has the advantages of high speed and large bandwidth.Secondly,it also introduces the OFDM(Orthogonal Frequency Division Multiplexing)technology for data transmission in the LTE-R network.The use of OFDM can realize the simultaneous transmission of data of multiple users in a high-speed railway environment.Then,the formation process of a cognitive cloud base station is introduced.Finally,the main user state model and research objects are proposed,which lays a solid foundation for the subsequent main user state prediction and channel optimal allocation.(2)By comparing the two prediction methods applicable to the prediction of the state of the main users—improved GM(1,1)Markov and AOS-ELM(Adaptive Online Sequential-Extreme Learning Machine),this thesis chooses a better prediction method to predict the occupancy of primary users.In the improved GM(1,1)Markov algorithm,first of all,GM(1,1)is used to predict the history datas of the primary users' status of three different lengths,the best predicted value is chosen.Treat the sequence length corresponding to the predicted value as the fixed length of metabolism.Then,the value is corrected by the second-weighted Markov model.Finally,whenever a new data is added,an oldest historicaldata is eliminated according to the fixed length,and the main user state prediction is performed according to the new sequence.In AOS-ELM,first of all,phase space reconstruction is performed on the historical sequence to form a usable sample of ELM(Extreme Learning Machine).Secondly,the prediction accuracy comparison of some samples are used to select the optimal number of hidden layer nodes as the initial value;and on the basis of basic ELM,it is onlined and Cholesky factorized,achieving real-time prediction and simplifying the calculation process.Finally,whenever a new data is added,the importance of the new hidden layer nodes added at the same time to optimize the network structure.By comparing the accuracy and the time-consuming of prediction,an improved GM(1,1)Markov model with high prediction accuracy and short time consumption is selected as the main user state prediction method.(3)In the process of optimally assigning channels,three methods of polling algorithm,heuristic-breadth-first search and heuristic-breadth-first search with main user state prediction results are used in the field of cognitive radio in this thesis.The flexibility and accuracy of the heuristic-breadth first search for assigning channels to secondary users are verified using the number of handovers and cumulative packet loss in the transmission process obtained from three different secondary user arrival rates,and verifying the advantages of the primary user state prediction results of channel optimal allocation in further.The results show that the heuristic-breadth first search method added to the main users' state prediction results has the best effect on channel optimal allocation,reducing the number of channel switching and the number of accumulated packet losses,improving communication quality in high-speed railway environments.