On-Line Learning-Based Allocation Of Base Stations And Channels In Cognitive Radio Networks

With the rapid application of 5G networks and wireless access terminals,the wireless network environment is more complicated with higher requirements which are placed on the throughput and transmission delay of the wireless transmission system.At the same time,existing spectrum allocation methods are weak on resource utilization and adaptability in complex environments,this make it difficult for them to meet new growth demands.In this thesis,cognitive radio networks allow secondary users to dynamically identify and access free channels,which is conducive to improving spectrum utilization efficiency in complex environments and speeding up the overall transmission throughput of the system.We research the problem of dynamic spectrum scheduling in cognitive radio networks.In this thesis,while the base station dynamically selects channel,the secondary user can also dynamically select the base station.Due to the wide distribution of base stations and secondary users,the communication link from the secondary users to the base station may be close to different primary users,so that the quality of the channel will dynamically change with different communication links.However,due to the ubiquitous deployment of the network and the unknown environment,it will be difficult to accurately measure this dynamic change in advance.On the other hand,the base station usually has limited resources and can only provide communication services that meet the requirements for a limited number of secondary users.Therefore,the goal of this thesis is to solve the joint optimization problem of base station and channel allocation when the communication quality of all links is unknown.That is,to allocate a suitable channel and a reasonable number of secondary users to each base station to maximize the secondary user's performance and data communication throughput.This thesis first studies the offline allocation problem when the link quality is known.We proved that the problem is NP-hard,and proposes a greedy algorithm based on the maximum load capacity of the base station.The algorithm has an approximate ratio of M-1(M is the number of base stations)while ensuring the polynomial time complexity.On the basis of this offline algorithm,this thesis proposes an online algorithm based on Multi-armed Bandits method.On the one hand,the algorithm guarantees the polynomial space complexity by making full use of the overlap between different allocation schemes.On the other hand,through the Upper Confidence Bound(UCB)mechanism,it achieves a balance between exploration and utilization.The distribution of quality in different transmission links is effectively learned.Through theoretical analysis,the maximum performance gap between the online algorithm and the offline algorithm is polynomial related to the number of base stations,channels,and users,and is related to the logarithm of time.Finally,this thesis compares the above-mentioned allocation algorithm based on Multi-arm Bandits online learning with existing algorithms such as E-greedy method and MLPS through a large number of simulation experiments.The experimental results show that the algorithm has better performance,and it is consistent with the analysis of this thesis.