Research On Resource Allocation In Next Generation Wireless Network Based On Economic Theory

The next-generation wireless communication system(5G)makes multiple wireless access technologies work together to improve network access capacity and provide users the stable access service with high data rate.Meanwhile,the 5G system also realizes the separation of control plane and bearer plane in the traditional network structure,and uses the network slicing technology to efficiently allocate and flexibly schedule various types of virtualized resources in the network,which can effectively adjust network topology and transport route according to the requirements of different users on network performance.Hence,the distributed features of 5G systems become more prominent than any previous generation communication system.First,the wireless access points in the 5G system are often base stations using different radio access technologies,and each type of base station has different resource usage modes.Heterogeneous access scenarios can enable base stations with different access technologies to achieve resource complementarity.Second,the virtual network resources used and scheduled in the network slice often come from devices in different locations on the network.The flexible network structure can easily achieve resource sharing among distributed devices.It can be seen that the distributed characteristic of the 5G system raise the level of network resource sharing to a new level.However,the continuous improvement of the level of network resource sharing has led to new changes in the network operation mode of 5G system.Emerging network operations roles will participate in or affect the allocation of network resources,and the resource allocation pattern shifts from centralized type to distributed type.Therefore,under the distributed architecture,achieving efficient resource allocation among independent individuals is an important challenge for 5G systems to give full play to the advantages of resource sharing technology.In addition,due to the individual's selfishness and competition,the complete network state information may not be obtained during the resource allocation process.It is unreasonable to use the common optimization theory to solve the distributed resource allocation problem.Since game theory,auction theory and intelligent learning algorithm in economic theory can effectively solve the complex distributed resource allocation problem,this dissertation relies on relevant means to study the resources allocation under the heterogeneous scenario and network slice service mode of 5G system.The main researches of this dissertation include the following four parts:(1)Multi-stage game based resource allocation of femtocell in heterogenous scenario;(2)Paired-bid based double auction mechanism for network slice resource;(3)Multi-armed bandit based mechanism for network slice reselection and resource allocation;(4)Reinforcement learning based dynamic allocation mechanism for network slice resource.The first research studies the competition between different network operators for the resource of Femtocell Access Point(FAP)in the heterogeneous secenario of 5G systems.Due to the explosive growth of indoor data traffic in 5G systems,Mobile Network Operators(MNOs)and Mobile Virtual Network Operators(MVNOs)often provide service to users by renting Femtocells with low cost.Different Femtocell Holders(FHs)are also willing to lease access resources to obtain revenue when resources are idle.Therefore,this work proposes a multi-stage game framework to ensure that the FAP belonging to FH can select the optimal resource allocation scheme to provide access services for macro-cell users of different network operators,thereby improving the overall resource utilization efficiency of the network.In the first stage,the framework selects the winning FAPs of FH through the reverse auction model,and then in the second stage,this work models the access resource allocation for the winning FAPs of different network operators as a Stackelberg game,and finally solves the model to obtain a Nash Equilibrium(NE)as the optimal resource allocation strategy.The theoretical analysis proves the unique existence of NE in the multi-stage game framework.The simulation results verify the effectiveness of the framework in term of the overall resource utilization efficiency of the network.The second research investigates the resource allocation problem between different mobile virtual network operators(MVNO)when deploying network slices under the full resource sharing of 5G network architecture.In 5G system,an Infrastructure Provider(InP)abstracts a physical network into multiple isolated network slices by virtualization technology.Each network slice can be operated by a different MVNO as a virtual network.In order to make full use of network resources,this research constructs a network slice market model,which allows different MVNOs to lease various network slices with specific Quality of Service(QoS)guarantee from different InPs,thus providing users with different service requirements.Meawhile,this research proposes a paired bid-based double-auction mechanism for allocating resources during network slice deployment.The proposed double auction mechanism first matches and pairs the MVNO who submits highest bid with the Infrastructure Provider(InP)who submits lowest asked,and then employes reverse auction to determine the winning pair.The MVNO and InP who belong to the winning pair adjust their bid and asked according to the transcation price issued by the system,in order to maximize the social welfare of allocation scheme.The theoretical analysis verifies the economic properties of paired bid-based double-auction mechanism.Numerical results show that the proposed mechanism can significantly improve the overall network resource allocation efficiency without collecting full information on the competition strategy and utility function of the MVNOs and InPs.The third research studies the network slice reselection and corresponding resource alloction caused by user mobility in the network slice service mode.After the network slices are deployed in 5G system,users access to the corresponding network slice to obtain required service according to the network slice ID assigned by the system.However,with the mobility of users,the probability of service failure will increase because the alternate target base station which would provide access service may not have deployed the same type of user's serving network slice.Therefore,this work proposes a multi-armed bandit based mechanism for network slice reselection and resource allocation,which can perform appropriate slice selection and resource allocation for handover users.First,from the perspective of economic efficiency,the mechanism designs a reward function for each handover in the system,then the mechanism tries to maximize the long-term reward by responding to user's slice reselection,and guarantee the service continuity.To this end,the mechanism utilizes an intelligent multi-armed bandit algorithm to make the system achieve the optimization of the reselection strategy and resource allocation,so as to obtain the optimal long-term handover reward.Finally,this research compares the performance of the proposed mechanism and typical greedy scheme by conducting simulation experiments,and evaluates the influence of various parameters of the system model on the handover performance.The last research investigates the dynamic allocation of network slice resource caused by the failure of slice reselection for handover user.In the slice-based network of 5G system,network slice can be flexibly adjusted according to the changes of user requirements.Considering that the handover users could not find a suitable serving slice which can meet its own QoS requirements from the neighboring base stations during the slice reselection process,this research work proposes a dynamic resource allocation mechanism that guarantees the maximum long-term resource scheduling reward.According to the user's handover request and the service condition of network slice in the network,the mechanism allocates the reserved dynamic resources to the neighboring base stations to construct a temporary slice for the handover users,and releases the slice after serving the users.In this work,the method of reinforcement learning is used to model the dynamic resource allocation,and based on Actor-Critic(AC)algorithm,this work propose an effective resource allocation mechansim.Finally,the simulation results verify the effectiveness of proposed dymanic resource allocation mechanism for handover users,and also evaluate the impact of various parameters of the system model on the proposed mechanism.