Research On Distributed Edge Caching In Fog Radio Access Network

The increasing number of mobile devices and the surging user demand for multi-media service have triggered massive amounts of data in wireless networks,which brings about huge pressure on fronthaul links.Fog radio access network(FRAN)has been proposed to address the fronthaul bottleneck and has attracted significant research interests from academia.Edge caching technology in FRAN enables contents being repeatedly requested to be stored in fog access points(F-APs)which are close to users,thereby the fronthaul pressure can be alleviated and the user request delay can be reduced.To efficiently utilize the limited caching resource of F-APs and avoid the extra communication with the control center,it is of great importance to study the distributed edge caching placement in FRAN.However,most existing works on distributed caching either require extra information exchange among F-APs,or fail to adapt to time-variant user requests.To address these problems,this thesis aims to investigate the distributed caching placement in FRAN.Firstly,by considering the request probability that varies gently with time,a distributed caching scheme based on mean field game without capacity constraint is studied.To characterize the time-variant user requests and F-AP states,a dynamic system model is established through differential equations.To jointly minimize the request service delay and fronthaul traffic load,a stochastic differential game(SDG)is formulated.Furthermore,by exploiting the ultra density of the F-AP deployment in FRAN,the SDG that is difficult to solve directly is converted to a mean field game(MFG),in which the individual states of F-APs can be expressed by a distribution function of the mean field state.Therefore,each F-AP is able to obtain the nash equilibrium of the MFG and the corresponding caching control through an iterative algorithm that based solely on the local state and the state distribution,without observing the states of other F-APs.Simulation results show that the proposed distributed caching scheme can effectively reduce the request service delay and fronthaul traffic load.Secondly,based on the above scheme and by considering the caching capacity constraint,a distributed caching scheme based on mean field game is studied.Firstly,the fractional knapsack problem in each time slot is formulated based on the caching control obtained in MFG nash equilibrium by considering caching capacity limitation,and a greedy algorithm is proposed to solve the fractional knapsack problem.Secondly,a distributed caching scheme is proposed by combining the algorithm in the MFG and the greedy algorithm in the knapsack problem.Finally,the existence and uniqueness of the mean field equilibrium is firstly analyzed,and then the overall optimality of the proposed scheme is analyzed based on the mean field equilibrium analysis by considering the greedy algorithm to solve the fractional knapsack problem.Simulation results show that the proposed distributed caching scheme is able to reduce the request service delay and fronthaul traffic load,and achieves better performance under smaller caching capacity and more uniformly distributed request.Finally,by considering unknown request probability dynamics and request arrival rate,a distributed caching scheme based on Lyapunov optimization that satisfies caching capacity constraint is studied.To capture the dynamics of request arrival,a dynamic request queue model is established,and the queue length can indirectly depict the request service delay.Then,the optimization problem in the queuing system that containing queue stability constraint is formulated.By employing the drift plus penalty(DPP)method in Lyapunov optimization theory,the original problem is reformulated as the minimization of the queue drift and fronthaul traffic load,thus the request service delay and fronthaul traffic load is jointly optimized indirectly.Furthermore,the optimization problem is decomposed with regard to each F-AP.Therefore,each F-AP is able to obtain the caching policy in a distributive manner.Simulation results show that the proposed distributed caching scheme can reduce both the queue length and the fronthaul load,and the tradeoff between request service delay and fronthaul traffic load can be obtained by adjusting the penalty weight parameter.