Research On Resource Allocation Algorithms In Mobile Edge Computing System

The popularization of the fifth generation(5G),Internet of Things(IoT)and smart mobile terminals(MTs)have given rise to a wide variety of intelligent applications,such as augmented reality,automatic driving,video calls and voice chat.However,the mobile terminals are usually resource-constrained,which fails to satisfy the increasing demands from the resource-hungry applications.Mobile edge computing(MEC)is considered as a promising technology,which can bridge the gap between the demands of applications and the restricted capabilities of MTs.MEC has attracted wide attention because it can reduce latency,ensure efficient network operation and service delivery,and provide improved users'experience.This thesis focuses on resource allocation for MEC,and a joint resource allocation algorithm based on parallel auction and a hierarchical game based resource allocation algorithm are proposed.The main work of this thesis is as follows.(1)The research status,key technologies and existing problems of resource allocation algorithms in MEC system are introduced.In addition,the application of game theory and auction theory in resource management is introduced briefly.(2)A joint resource allocation algorithm based on parallel auction(JRAPA)is proposed for MEC.The joint allocation of wireless and cloud resources is modeled as an auction process,aiming at maximizing the utilities of service providers(SPs)and meeting the delay requirements of MTs.The auction process is composed of bidding submission,winner determination and pricing stage.At bidding submission stage,the MTs take available resources of and distance factors into account to decide the bidding priority,thereby reducing the processing delay and improving the successful trade rate.A resource constrained utility ranking(RCUR)algorithm is put forward at winner determination stage to determine the winners so as to maximize the utilities of SPs.The wireless resource auction and cloud resource auction are processed in parallel to achieve the fast matching between the buyers and the sellers,and guarantee the convergence speed.At pricing stage,the sealed second-price rule is adopted to ensure the independence between the price paid by the buyer and its own bid.The simulation results show that JRAPA outperforms other existing algorithms in terms of the convergence rate and the number of successful trade rate.It can not only achieve larger average utility of SPs but also significantly reduce the average delay of MTs.(3)A hierarchical game based resource allocation algorithm(HGRAA)is put forward,which includes the lower-level evolutionary game(LEG)and the upper-level exact potential game(UEPG),aiming at minimizing the cost of mobile terminals(MTs)and maximizing the utility of MEC servers.For MTs with different service requirements,an evolutionary game is used to model the selection of service providers,and dynamic replicator is used to obtain a stable population state so as to minimize the cost of MTs.Exact potential game model is established to solve the problem of resource sharing among MEC servers,aiming at maximizing the revenues of servers under users' quality of service(QoS)constraints.The existence of Nash equilibrium(NE)is proved.The simulation results show that the average revenue and resource utilization of MEC servers in HGRAA are higher than existing algorithms,and the cost of MTs is significantly reduced.(4)The current research work and the prospects of future research are given.