Research On Collaborative Edge Caching Based On Channel Quality

Mobile data traffic in wireless network is experiencing explosive growth due to the development of mobile services,social networking,and resource-intensive applications.The explosion of data traffic has led to unprecedented pressure on wireless network.New applications have prompted content providers to adopt new technologies to improve the quality of user experience,especially for applications which are sensitively for network latency and network throughput,e.g.,Online Video,Internet of Vehicles,Virtual Reality and etc.To alleviate the ever-increasing data traffic pressure and providing lower service latency for new applications,the concept of edge caching is proposed.By caching contents in edge server,edge caching can shorten the physical distance between users and contents and avoid the waste of communication resources caused by repeated transmission of contents.Edge caching has been shown to be effective in alleviating the data traffic pressure of the backhaul link.Due to the limited capacity of edge server,how to utilize the limited cache space for efficient caching becomes the core problem of edge caching.However,the wireless environment is complex and dynamic.On one hand,due to the strong spatial-temporal dynamics of wireless service,the information of content popularity is difficult to obtain;On the other hand,wireless communication environments have great differences and strong uncertainty which makes edge caching more challenging,due to factors of user mobility,multipath effect,shadow fading,communication errors and etc.Therefore,this thesis studies collaborative edge caching policies based on channel qualities in case of unknown content popularity.This thesis investigates cooperative content caching policy in a scenario with heterogeneous channel qualities and unknown content popularity.The thesis adopts average transmission delay as performance metrics,derives the total transmission delay of serving user requests,and establishes an optimization problem for this scenario.To solve this optimization problem,a hierarchical caching policy based on channel qualities is designed.This thesis adopts a Bayes-based content popularity estimation algorithm to learn content popularity,which can achieve the balance between exploration and exploitation.After obtaining the estimated content popularity,this thesis implements hierarchical cooperative content caching based on the heterogeneity of channel qualities and the content popularity.Compared with the existing caching policy,the proposed policy achieves lower average transmission delay and higher cache hit rate.Simultaneously,edge caching policy over unreliable channel is investigated in this thesis when the channel reliability and the content popularity are unknown.In view of the limitation of the nearest base station serving policy,this thesis proposes a new user service policy based on channel reliability.By implementing this policy,user requests can always be served by the base station with highest reliability.Then,this thesis establishes an edge caching framework based on deep reinforcement learning,and utilizes a deep reinforcement learning algorithm to achieve content replacement.Compared with existing user service policy and content caching policy,the proposed policy performs better,and it is more robust.