Research On Learning And Knowledge Sharing Strategy Of Cognitive Base Station In Railway Cognitive Radio

In recent years,the Internet has been widely used in all walks of life.The demand for wireless communications has begun to show an explosive growth,and the radio spectrum resources on which it depends have suffered a serious shortage.The traditional fixed spectrum allocation strategy is unable to meet the needs of the current stage.Therefore,dynamic spectrum access(DSA)strategy is considered as an efficient solution to the shortage of spectrum resources.Hence,Cognitive Radio(CR)has become a key technology to solve the lack of radio spectrum resources by allowing the unlicensed users opportunistically access the licensed frequency bands which are already assigned to the licensed users to achieve efficient use of spectrum resources.On the other hand,with the rapid development of rail transit,the wireless communications issues in rail transit such as low degree of spectrum reuse,high call drop rate,low connection rate and low success rate of spectrum switching caused by the use of dedicated frequencies,Doppler shift,penetration loss,frequent cellular handover draw attention of the communications industry.Against this background,several spectrum strategies based on learning and knowledge sharing are proposed in this paper to solve some issues of spectral resource utilization in rail transit environment.This paper presents the concept of Rail Cognitive Base Station(RCBS).As the spectrum manager and assigner within its coverage,the RCBS deals with the communication needs of trains.RCBS has been given a variety of spectrum allocation strategies.The first is two-step decision and epsilon greedy reinforcement learning spectrum allocation strategy.By establishing new state and action pairs,the RCBS makes two-step decision which decides channel switching and allocation,and we also apply the epsilon greedy exploration to solve the exploration and exploitation dilemma to optimize of decision making process and improve the performance of spectrum management.The simulation results show that this strategy is obviously superior to some existing spectrum allocation strategies in improving the throughput of unlicensed users and reducing the spectrum handover.The second strategy is channel accessibility inference based on Bayesian network.Taking into account that the reinforcement learning strategy is only based on the existence of licensed users to determine the channel accessibility.The Bayesian inference strategy aims to calculate the channel accessibility based on a variety of priori information collected during the interaction between RCBS and the environment and after accessing the selected channel,RCBS use the obtained performance to strengthen the inference.The simulation results show that this strategy can reduce the number of spectrum switching in RCBS coverage area of unlicensed users.The two spectrum allocation strategies above only consider the spectrum allocation within a single RCBS coverage.The third strategy we proposed is that the RCBS predicts the remaining available time of the licensed channel by recording licensed users' arrival and departure parameters,and shares the result of prediction with the pre-RCBS to realize the optimal channel selection of trains when they perform spectrum handover.Through the simulation of comparing several strategies,it is proved that the times of spectrum switching within the coverage of the subsequent RCBS is effectively controlled.