Research On Resource Management For Multilevel Edge Computing In Wireless Fronthaul Network

With the high integration of cloud computing and the rapid development of 5G,the surging data volume is driving higher requirements for resource transmission rate,latency,and energy consumption.The C-RAN architecture uses a central wireless cloud network to manage network resources.Combined with the proposal of edge computing,it allows edge servers to be deployed at the edge of wireless fronthaul network to provide users with powerful computing services,which is an effective solution to improve the quality of users’ experience of new applications.At the same time,the huge amount of data has caused the strain of spectrum resources.In traditional orthogonal multiple access,a single time-frequency resource can only be allocated to one user,which cannot meet the current demand.Therefore,the introduction of non-orthogonal multiple access(NOMA)improves the spectrum efficiency by allowing multiple users to transmit simultaneously in the same resource block.Based on the above techniques,this thesis applies the idea of multilevel to edge computing in wireless fronthaul networks and studies the resource management of NOMA transmissions.The specific work is as follows.This thesis combines the idea of multilevel with the number of edge servers,and proposes a heterogeneous network system with m MTC and u RLLC devices coexisting in different 5G applications,where edge servers are deployed near macro base stations and each small base station,relying on a CU/DU resource pool rich in computational resources.NOMA transmission protocol is introduced to allow the small cell users to transmit and reuse the channel resources of the macro cell,thus improving the spectrum utilization and reducing the energy consumption of users.The combination of NOMA with multi-level edge servers in heterogeneous networks introduces both same-layer and cross-layer interference,making it more complicated to model the mixed integer nonlinear programming problem with joint sub-channel allocation,power control,computational resources and offloading ratios.For this reason,this thesis first splits it into three subproblems,optimizes the subproblems separately by Gale-Shapley matching algorithm,SCA algorithm and interior point method,and then obtains the optimal solution of the proposed model by alternate optimization.Finally,it is verified by simulation that the proposed scheme is better to reduce the energy consumption of target users.To address the problem that the interference brought by NOMA transmission affects the transmission rate of users when the number of shared users is large,this thesis proposes a hybrid NOMA-OMA transmission in multiple time periods in a hierarchical manner using the idea of multilevel,designs an edge computing system consisting of a cellular user with a large task and multiple edge users with a small task in the context of wireless fronthaul network,and establishes a cellular user time delay A non-convex optimization model of joint computational and communication resources is developed with the objective of cellular user delay.Then,the original problem is split into three subproblems by decoupling variables,and the Dinkelbach algorithm is used to complete the transformation of the fractional programming problem.Then the non-convex constraint is approximated by the SCA algorithm to convert it into a convex constraint so as to solve it,and the joint optimization algorithm with alternating optimization is designed to obtain the suboptimal solution of the proposed scheme model.Finally,it is numerically verified by simulation that the hybrid NOMA transmission scheme can reduce the task delay better than the OMA transmission scheme.