Research Of Caching Technique In Wireless Networks

The 5G identified three major scenarios,namely eMBB,mMTC,and URLLC.In order to meet various demands for these three scenarios,interface capacity of 5G new radio needs to be further improved,and network architecture as well as key technologies need to evolve.Edge caching technology is one of the important research directions of 5G system.By migrating storage functions to the edge of wireless network,users and contents are closer.Accordingly,end-to-end delay and backhaul load are both reduced.Most studies assume that the popularity is an independent reference model,and focus on the theoretical analysis and optimization of wireless network after the introduction of caching function,including cache information theory,cache and wireless transmission,and random wireless network cache.Caching based on popularity dynamics can reflect the needs of users in the real world to improve the performance of the cache.The cache ecosystem contains multiple stakeholders,so practitioners need to consider economics when implementing caching technologies,namely incentive mechanism design.Social networks can be used not only to predict the probability of user encounters,but also to influence the popularity and spread of content.This thesis studies the dynamic popularity model and designs the content placement strategy in different scenarios.Then the performance optimization of caching system is extended to the design of incentive mechanism.Finally,the author further explores the popularity model with the help of the complex social contagions analysis in network science,which lays a foundation for the more efficient design of cache updating strategy in the future.The main work and innovations are summarized as follows:1.According to the evolution characteristics of the content popularity,the mobile characteristics of cellular users and the content preference characteristics,the cache updating problem in the small base station caching scenario and the content placement problem in the UAV caching scenario were studied respectively.Firstly,this thesis employs a linear regression method to predict the content popularity,which is different from the fixed content popularity assumed by most previous researches.An efficient cache updating scheme based on stochastic network model is proposed.The proposed cache updating scheme can track the network dynamics better than the existing scheme and adjust the cache placement accordingly.Further considering the UAV caching scenario,a fine-grained content preference and location model for a single user is adopted,which is different from the independent and identical request probability model assumed by most researches.Based on the proposed content preference prediction model and request location model,the UAV deployment location is designed by using the clustering algorithm in machine learning,and the content placement scheme is designed according to the corresponding UAV deployment location.The proposed scheme can improve the UAV deployment and caching efficiency more than the existing scheme,and dynamically adapt to the user state.2.For the D2D cooperative communication scenario,a prediction model of user encounter probability based on social network is proposed,and a Starkberg game model between mobile network operators and users is established to jointly design user incentive and content placement.Different from the independent reference model widely assumed in the research,the content preference model adopted in this thesis can reflect the personalization of user requirements,which is more suitable for the caching scheme design in this scenario.In addition,the joint design of user incentive and content placement can not only achieve a higher system cache hit ratio,but also facilitate the cooperation between mobile network operators and end users,compared with the caching scheme under the assumption of full cooperation.3.Aiming at the problem of information sharing and transaction credibility between content providers and mobile network operators,a caching mechanism based on block chain and smart contract was proposed,and the caching placement algorithm was designed considering the regional characteristics of content popularity.Most of the existing studies ignore the nature of distrust and competition between content providers and mobile network operators.The caching mechanism designed in this thesis can guarantee the public availability of content popularity statistics and prevent the risk of transaction fraud.By querying the relevant transaction information issued by mobile users in the block,the regional content popularity model adopted in this thesis can achieve a good compromise between the learning speed of popularity and the model accuracy.4.In order to predict the change of content popularity more accurately and fully exploit the potential of caching technology,this thesis utilizes the idea of complex social contagions in network science,and proposes an edge-weight compartmental theory to quantitatively analyze the complex contagion effect of heterogeneous adoption in weighted social networks.Most of the existing works ignore the impact of social enhancement and heterogeneous adoption,both of which are important factors in caching research.In this thesis,the model of complex contagion network is built in analogy with the attributes of caching problem.An edge-weight compartmental theory is proposed to predict the propagation process.A large number of experimental results show that the theoretical prediction is consistent with the simulation results.This study lays a theorectic foundation for further exploration of the prediction model for content popularity dynamics and the design of the corresponding cache updating strategy.