Caching Optimization And Performance Analysis In Wireless Mobile Networks

In recent years,with the development of wireless mobile communication technologies,wireless data traffic is undergoing the explosive increase.Wireless caching is widely regarded as a promising solution by the academia and industry to relieve the pressure of the rapid increase of network traffic.However,there are still many challenges to design and optimize the caching strategy of wireless mobile networks,as well as the performance analysis based on it,which aims to achieve the improvement of the quality of service for mobile users in the network.Based on the asymptotic analysis,convex optimization,game theory,deep reinforcement learning and other theoretical methods,this dissertation starts from the analysis of the network performance of cacheless nodes in wireless mobile networks,and further extends to the design and optimization of caching strategies under different mobile network scenarios,ultimately aiming at optimizing the performance of wireless mobile networks.The main contributions of this dissertation are summarized as follows:(1)The multicast throughput analysis of cacheless nodes based on the random walk model.By jointly considering different mobility spatial dimensions and different mobility scales of cacheless nodes,this dissertation establishes four different random walk mobility models with different application scenarios.In each mobility model,each node is equipped with the directional antenna for multicast communications.The one-hop transmission scheme and multi-hop transmission scheme are discussed based on the data transmission delay constraint.Then,the asymptotic analysis method is applied to theoretically derive the upper bound of the multicast throughput of the cacheless nodes.Numerical simulations analyze the impacts of different network size,directional antenna beam,data transmission delay and other network parameters on the upper bound of multicast throughput of cacheless nodes,which is of great significance for node communication mode and mobility mode selection in different mobile network scenarios.Moreover,it provides a theoretical basis for the subsequent performance analysis of caching nodes in wireless mobile network.(2)The caching strategy optimization and performance analysis of mobile caching nodes under the home-point regular distribution model.To further improve the network performance,we consider the impact of caching strategy on the throughput and delay performance of mobile nodes.For the realistic phenomenon that caching node's local mobility around the corresponding home-point,a home-point regular distribution network model is established.Then,we design the non-interference transmission scheduling scheme and the content mulit-hop routing scheme under this network model,and theoretical results for the throughput and delay of mobile caching nodes are derived analytically.The delay optimization problem of caching nodes is established based on the content popularity distribution and the caching capacity constraint.Combined with the characteristic of the home-point regular distribution network model,the optimal caching strategy is designed.Finally,based on the designed caching strategy,we derive the optimal throughput and optimal delay by using Lagrangian relaxation method.This study provides a theoretical basis for the caching optimization and performance analysis of mobile caching nodes under the subsequent home-point random distribution model.(3)The caching strategy optimization and performance analysis of mobile caching nodes under the home-point random distribution model.In order to extend the simple home-point regular distribution scenario to the general home-point random distribution scenario,a home-point random distribution cluster network model is established,and the network is divided into cells of equal area by theoretical calculations,thus ensuring that each caching node in the cell can find the corresponding home-point.Then,a non-interference transmission scheduling scheme and a content multi-hop routing scheme are designed based on the cell partition,and we theoretically derive the asymptotic closed-form expressions for the throughput and delay of mobile caching nodes.According to the aforementioned theoretical results,we establish the delay optimal problem of caching nodes,design the optimal caching strategy combined with the characteristic of the home-point random distribution cluster model and derive the optimal solution for delay and throughput by using Lagrangian relaxation method.Finally,according to numerical simulations and comparative analysis,it is verified that the optimal caching policy designed for different home-point distribution scenarios can optimize the throughput and delay of the mobile nodes.(4)The study of game algorithms between caching nodes and the marco base station(MBS)based on the preferences of caching nodes.On the basis of only considering mobile caching nodes in the network,a MBS with large caching capacity is introduced to establish a heterogeneous caching network architecture.In order to promote the content share between caching nodes and reduce the stress of repetitive content transmissions from the MBS,this dissertation proposes a caching incentive mechanism for caching nodes.We formulate the profit interactions between caching nodes and MBS as a Stackelberg game.In order to solve the Stackelberg Equilibrium(SE),a distributed uncoded caching algorithm and a dynamic iterative algorithm are designed,and then we can obtain the optimal joint caching strategy of the caching nodes and the optimal caching incentive reward of the MBS.Finally,the numerical simulation verifies the existence of the SE of the proposed game,and the superiority of the designed distributed uncoded caching algorithm and the dynamic iterative algorithm is verified by comparative simulations.(5)The design of the coding caching algorithm based on time-varying content popularity.Under the cooperative caching mechanism of the MBS and edge small cell base staions(SBSs),jointly considering the impact of time-varying content popularity and node mobility on the coding caching decisions of SBSs,we establish the optimization problem of the average delay cost of mobile nodes requesting contents in the double time scales.Considering the high-dimensional,continuous and time-varying caching decisions of SBSs,we transform this optimization problem into the markov dynamic decision model.An advantage actor-critic coded caching algorithm is proposed by using deep reinforcement learning methods.In addition,the caching capacity constraint problem of SBSs is solved by adding a penalty factor into the proposed algorithm.Simulation shows that the proposed coded caching algorithm has good convergence performance and can effectively reduce the average delay cost of the mobile nodes requesting contents.Finally,we make a conclusion on the caching optimization and performance analysis in wireless mobile networks.The problems that haven't been solved and the technical methods that need to be improved in this dissertation are discussed.Moreover,we point out the future research work.