Edge Caching For Cellular Systems With Device-to-Device Communications

As cellular data demand experiences an unprecedented growth in recent years,it brings tremen-dous traffic pressure to cellular networks.Due to the scarcity of wireless network resources,the traditional methods to increase the capacity of cellular networks,such as more spectrum and small-er cell,are approaching its limitation.Traffic offloading in cellular networks becomes the focus of research and industry.D2D communication makes direct communication links between UEs enabled,which can alleviate the traffic load of BSs compared with cellular communication.Be-sides,edge caching is considered as a promising approach to realize traffic localization and traffic offloading.The combination of D2D communication and caching makes the data cached at UEs can not only satisfy the demand of UEs' themselves,but also be shared with adjacent UEs through D2D communication,which reduce the traffic load of cellular systems effectively.Thus the s-tudy of D2D assisted on-device caching strategy with the aim to maximize the traffic offloading in cellular systems is of great theoretical significance and practical value.Firstly,this paper analyses the performance of D2D assisted on-device passive caching in cel-lular systems.As UEs caches some files passively,the file requests of UEs seek to be satisfied by self-caching or sharing between UEs via D2D communication,which alleviate the system traffic load.This paper studies the performance of on-device passive caching in two different UEs' mo-bility models.In static network,the topology of UEs remains stable,which means the scheduling for D2D links is enabled.Considering that the transmission demand of D2D links is changing at different times,this paper proposes a heuristic algorithm based on MWIS model to obtain the sys-tem traffic offloading ratio.In high dynamic network,the locations of UEs change so frequently that the topology of UEs is unstable.This paper models the UEs' distribution as Poisson point process,and derives the mathematical expression of the system traffic offloading ratio by utilizing the stochastic geometry.Secondly,this paper studies the D2D assisted on-device proactive caching in cellular systems.The UEs can adjust their caching contents proactively to maximize the amount of system traffic offloading.Considering that UEs have different mobility and limited cache capacity and files have different popularity and content size,the problem is modeled as a combinational optimization problem and is proved to be NP-hard,which means the optimal algorithms for the problem have exponential computational complexity and are unrealistic for the large number of UEs and files.A low-complexity algorithm for the caching adjustment between UEs is proposed to adjust the UEs' caching content distributionally when the contacts between UEs occur,and the performance bound of the algorithm is derived theoretically.Based on the caching adjustment algorithm,this paper proposes an infrastructure-assisted data offloading algorithm in which the BS should provide necessary information to UEs.In order to reduce the control information load of the BS,a fully-distributed data offloading algorithm is proposed accordingly in which UEs exchange information via D2D communication and make parameter estimation by themselves.Lastly,the joint optimization of D2D assisted cooperative transmission strategy and on-device caching strategy is explored.The D2D assisted cooperative transmission,in which multiple UEs transmit the same file to the same UE independently,can reduce the effect of wireless channel fading and improve the quality of communication,which can further improve the performance of traffic offloading combined with on-device caching.This paper models the joint optimization problem of cooperative transmission strategy and on-device caching strategy,and decomposes the problem into a subproblem about transmission strategy and a master problem about caching s-trategy.The subproblem is transformed to a finite-horizon MDP problem,and the corresponding Bellman optimality equation is derived.Thus a link selection algorithm is proposed to obtain the system transmission strategy by solving the Bellman optimality equation.Then the system caching strategy is proposed by solving the master problem with the gradient projection method.