Research On Resource Allocation Technology Based On Game Theory In Mobile Edge Computing

With the rapid development of wireless communication,mobile devices(MDs)and compute-intensive applications are growing exponentially.As a key technology of 5G,mobile edge computing(MEC)handles tasks migrated from MDs at network edge,which extends battery life of MDs and reduces delay.As many MDs are connected to network,how to realize the efficient resource allocation in MEC becomes a problem to be solved.Based on the theoretical framework of game optimization,this paper effectively solves the problem of resource competition equilibrium and benefit maximization of various network entities.The following results are obtained.(1)Consider that operator departments(ODs)collaborate to invest in MEC resources to provide MEC service,we optimize resource ownership allocation to maximize OD's utility.Shapley value is used to model benefit distribution.A resource allocation algorithm based on incomplete contract is designed.In addition,effect of resource ownership on OD's investment is analyzed.Simulation results show that the algorithm can motivate ODs to choose optimal investment,which maximizes OD's utility and guarantees quality of service.(2)Consider that the task migration from the user to the base stations(BSs)and BSs collaborate to handle tasks,we optimize BS resource allocation to maximize system utility.The problem of user and BS association and BS cooperation are designed.We propose association and cooperation algorithm based on matching theory,which realize task allocation and efficient usage of resources.We also design a bidding mechanism to realize resource's optimal price.In addition,the algorithm is proved to achieve stable matching and competitive equilibrium.Simulation results show that the proposed algorithm can achieve higher system utility and lower transmission delay than benchmark algorithms.(3)Consider that computing service providers(CSPs)rent edge computing nodes(ECNs)to allocate to users with dynamic tasks,we optimize ECN resource allocation to maximize system utility and user's offloaded tasks.The optimization problem of ECN rent and auction are proposed.The problem of ECN auction is modeled as markov decision process.We design ECN rent algorithm based on matching theory.What's more,auction algorithms based on Q learning and deep Q network are proposed.Simulation results show that the proposed algorithms can improve system utility and task migration compared with traditional algorithm,which allocate resources efficiently to different types of users.Finally,we summarize the work and shortcomings,then the follow-up study are discussed.