A Dynamic Distributed Spectrum Allocation Mechanism Based on Game Model in Fog Radio Access Networks

Yao Yu,Shumei Liu＊,Zhongshi Tian,Siyu Wang

School of Computer Science and Engineering,Northeastern University,Shenyang 110819,China

Abstract:With the explosive growth of highspeed wireless data demand and the number of mobile devices,fog radio access networks (F-RAN) with multi-layer network structure becomes a hot topic in recent research.Meanwhile,due to the rapid growth of mobile communication traffic,high cost and the scarcity of wireless resources,it is especially important to develop an efficient radio resource management mechanism.In this paper,we focus on the shortcomings of resource waste,and we consider the actual situation of base station dynamic coverage and user requirements.We propose a spectrum pricing and allocation scheme based on Stackelberg game model under F-RAN framework,realizing the allocation of resource on demand.This scheme studies the double game between the users and the operators,as well as between the traditional operators and the virtual operators,maximizing the profits of the operators.At the same time,spectrum reuse technology is adopted to improve the utilization of network resource.By analyzing the simulation results,it is verified that our proposed scheme can not only avoid resource waste,but also effectively improve the operator's revenue efficiency and overall network resource utilization.

Keywords:fog radio access networks (F-RAN); game theory; spectrum reuse technology; base station dynamic coverage; spectrum pricing and allocation

I.INTRODUCTION

With the rapid development of ubiquitous networks and smart cities,the high efficiency communication of Internet of Everything (IoE) have drawn great attentions [1-5],such as smart grid networks,vehicular networks,cellular networks and so on.In addition,because of the rapid increase in mobile communication services,high deployment management costs and scarce wireless resources,it is very important to use efficient wireless resource management mechanism [6,7].F-RAN is a multi-layer network structure including small sites with low transmission power,such as relay stations,picocell sites,and home base stations.It has the advantages of low communication latency,high spectrum utilization,flexible networking,strong coverage and etc.As an evolution form of heterogeneous cloud radio access networks (H-CRAN),F-RAN combines the advantages of H-CRAN and takes full use of processing and storage capabilities in edge devices [8,9].F-RAN has drawn universal attention from academia and industry in recent years [10,11].It has been considered as an ideal strategy for high efficiency and low latency that can meet the growing demand of mobile communication services.

In the study of solving the problem of network resource allocation,the relevant scholars focus on the research of access control,congestion control,packet schedule and other technical means.The study object is usually network protocol and implements resource control on the network nodes.In recent years,some scholars have tried to apply the pricing strategy of supply and demand of microeconomics to resources allocation.They envisaged guiding the resource demands of network users through the potential impact of prices,and thus have a positive effect on the rational allocation of the overall network resources.

For the diversification of operators and service diversified network environment,some issues have attracted increasing academic attention.

1) How to plan the relationship between different individuals on the operator's platform.

2) How to achieve maximum communication benefits with limited resources.

3) How to allocate spectrum resources of different operators reasonably.

In this paper,we use F-RAN as an efficient network architecture.On this basis,we focus on the application of game theory model to design spectrum pricing and allocation scheme,maximizing the revenue of operators and increasing the efficiency of network resource usage.

1.1 Related work

F-RAN wireless resource management problem has become a research focus.Zhanget al.proposed an architecture of NOMA-based F-RANs and discussed the problems of resource allocation [12].Lianget al.proposed a joint resource allocation and coordinated computation offloading algorithm for the fog RAN (F-RAN),taking the advantages of cloud radio access network (C-RAN) and fog computing [13].[14] summarized the recent advances of the performance analysis and radio resource allocation in F-RANs comprehensively.In particular,the author proposed the radio resource allocation strategies for optimizing SE and EE in F-RANs,respectively.In order to solve the delay problem in resource allocation,[15] proposed a joint distributed computing scheme and a distributed content sharing scheme with the greedy algorithm.[16] proposed an adaptive resource balancing (ARB) scheme for serviceability maximization in fog radio access networks.The overcapacity black hole problem and the unfairness of association possibility are both addressed.

Game theory analyzed the effect of competition or cooperation between network operators on resource allocation to satisfy the effectiveness of resource allocation [17].Mehbodniya and Aıssa studied the problem of interference management in heterogeneous multimedia wireless personal area network by game theory [18].[19] used game theory to study the spectrum allocation between authorized users and cognitive users in cognitive radio networks,analyzed the economic negotiation process of cognitive users renting free spectrum resources from authorized users.[20] dealt with the spectrum allocation in the distributed cognitive radio network by the non-cooperative game theory,the proposed algorithm achieved a stable convergence at the expense of a certain system benefit.[21] analyzed the radio resource management and control between the real user and the “virtual user” of the RRC layer using the game theory for the wireless resource control layer.The proposed scheme improved the overall performance of the RRC layer in LET-A.[22] used the classical Cournot model in the game model to analyze the leased spectrum behavior of major users in cognitive radio,the simulations showed that compared with the static spectrum allocation algorithm,the proposed algorithm increased the total number of spectrum.[23] proposed cognitive radio system model and studied the spectrum allocation algorithm in the time domain of cognitive radio by the non-cooperative game theory.[24] proposed a channel selection algorithm based on DS (Dempster Shafer) theory and game theory,realizing dynamic spectrum allocation.

1.2 Our contribution

This paper studies the complex network environment of F-RAN and the spectrum pricing and allocation scheme (SPAS) based on the game model.We enable the operators allocate spectrum on demand and improve the efficiency of network resources.The detail of our contribution as follows:

1) We chose the Stackelberg game model to establish a reasonable spectrum pricing and allocation algorithm.

2) Considering the characteristics of macro-cell and small-cell,the whole network is divided into three main areas according to the regional characteristics.We study the dynamic coverage characteristics of the base station according to the coverage of the base station.

3) For dynamic coverage of base stations,we use multiple spectrum reuse techniques to improve and optimize the spectrum pricing and distribution scheme.

1.3 Outline

The remaining of this paper is structured as follows.In Section II,we present spectrum pricing and allocation algorithm based on game model under the F-RAN framework.In Section III,the algorithm based on multiple spectrum reuse technique is given.In Section IV,we implement the proposed scheme to demonstrate its practicability.The conclusion is given in Section V.

Fig.1.System network model.

II.GM-SPAS

In order to consider the utility of the proposed scheme in the actual system,we first propose a universally applicable spectrum pricing and allocation scheme based on game model to maximize operator revenue and improve overall network resource utilization efficiency.

2.1 System network model

Base station types of heterogeneous cellular networks include macro-cells,micro-cells,pico-cells,and home-cells.In this paper,the entire network presents a dual-layer network structure covered by macro-cells and smallcells,rather than the multi-layer network architecture.The small-cells layer includes all small base stations except the macro-cells such as micro-cells,pico-cells and home-cells.The system network model studied in this paper is shown in Fig.1.

In this paper,all the small base stations can be processed uniformly when establishing the mathematical models,which does not affect the validity and rationality of the algorithm.Therefore,the services that operators can provide for users are also divided into macro-cells services and small-cells services in the above-mentioned network system model.

2.2 Distributed resource supply model

Traditional network operators (TNOs) and virtual network operators (VNOs),as network operators in the HCN,can provide different types of services to users.Considering the network model in Fig.1,we assume that TNO provides macro-cell services to users,and VNO provides small-cell services to users.The model in which different operators manage different service types is called a distributed network system.The corresponding resource supply system is called distributed resource supply model (DRSM).Since VNO does not have network and authorized network bandwidth,it provides services to customers through leasing and use of the network.The VNO provides users with convenient communication services by renting network resources from the TNO.

2.3 Construction

In recent years,microeconomics has been widely applied to the network resource management.Optimize wireless resource management from the perspective of market mechanism and economics,improving the efficiency of network resource utilization.In this paper,we study the SPAS based on game model (GM-SPAS),and we use the Stackelberg to simulate and solve the equilibrium benefits.

We build a two-level game between TNO and VNO,users and operators in F-RAN.The game relation of distribute resource supply system is shown in Fig.2,pMrepresents the TNO's macro-cell service price,pSrepresents the VNO's small-cell service price.The two layers of the game are played at the same time in our proposed system.By analyzing the demand of user resources,TNO allocates spectrum resources on demand for its own macro-cell service and VNO's small-cell service,so as to realize the efficient use of spectrum resources.Through the game between TNO and VNO,determining the optimal price and achieving their maximum benefits.

In the game of spectrum rental between operators,TNO determines the rental price and announces it to the VNO during the first phase of the game.Let∈represents the lease price coefficient and the lease unit price is expressed as∈pM.The game process is as follows:

1) TNO first determines the price of the macro cellular servicepMand∈;

2) VNO determines the price according to its principle of profit maximization;

3) The end user chooses the service according to the price set by the operators and its own channel condition.The selection principle of end user is se- lecting VNO's small base station service whenpM＞pS.Then,end user determines the optimal resources according to theQMandQs.QMrepresents the end user bandwidth requirements of TNO's macro-cell services,Qsrepresents the bandwidth requirements of VNO's small-cell services.We assume that all users behave in strict accordance with the above principles.The specific step of the GM-SPAS is shown in Algorithm 1.

Algorithm 1.GM-SPAS.1:Set the initial state of the price pi M(i)=0 and pi S(i)=0,set ε and other relevant parameters 2: Set the price traversal interval and the initial value Bj( j)=0 of bandwidth resource B 3:Set the current initial optimal state and the initial value of the operator's revenue Ri M(i)=0 4:if PANB (j)）（Total resource requirements 0≤=j 5:Ri M(i)=1 ← the value of the current revenue 6:if RR i i M M (i)(i)=＞=0 1 7:if pi S(i)=≤0 1 8:Record prices for services and operator's revenue at this time 9:end 10:end 11:end 12:Change prices by interval,i.e.,replace pi M(i)=1 with pi M(i)=0 .Repeat the above loop and determine the revenue again 13:Output the equilibrium price and the operator's revenue at this time 14:Change the total resource of the system,i.e.,replace B jj(=1) with B jj(=0).Finding the current new network resources again under the equilibrium price,as well as the operator's revenue

Fig.2.Game relation of distribute resource supply system.

III.MSRT-SPAS

In the study of section II,we assume that all users can choose macro-cell services or smallcell services,which means that all users are geographically indistinguishable.However,the coverage area of the base station is dynamic in the actual scenario.Therefore,this section designs the spectrum pricing and allocation scheme for the Base Station Dynamic Coverage (BSDC-SPAS) scenario.Considering the importance of spectrum reuse technology in cellular mobile communications,this section proposes a spectrum pricing and allocation scheme based on Multiple Spectrum Reuse Technologies (MSRT-SPAS).

3.1 Base station dynamic coverage analysis

The actual small-cell coverage may be within the coverage of the macro-cell or located in the area independent of the macro-cell.And the user's purchase intention will be different for different regions.Therefore,we define the Base Station Dynamic Coverage (BSDC) model as shown in figure 3.

The entire network contains three areas,including area I covered only by the macro-cell,the area II covered only by the small-cell,and the area III covered by the macro-cell and the small-cell.The proportions of the coverage areas areφ1,φ2,φ3(φ1+φ2+φ3=1),respectively,and the purchase factors of the users in the corresponding area areλ1,λ2andλ3.We refer to area I and area II as non-overlapping areas,and area III is called the overlapping area.The proportion values of the three areas change dynamically according to the actual environment.

Considering the BSDC model,the corresponding operator's revenue model changes together.The revenue principle of TNO is similar to Section II,the difference is that we need to consider the needs of users in different areas.Letandrepresent the resource requirements of the users in area I and area II,respectively.andrepresent the resource requirements of the users in the area III for the services of macro-cell and the smallcell,respectively.

In order to maximize the revenue of operators,our BSDC-SPAS adopts the regressive induction method to solve the equilibrium according to the double-layer game established in Section II.First,the TNO determines the equilibrium pricepMand finds the equilibrium pricepSof VNO.Then,we come back to solve the equilibrium pricepMof the traditional operators.We demonstrate the advantages of BSDC through simulation.In non-overlapping area,as the purchase intention of user increases,the equilibrium price increases.In addition,the greater the proportion of non-overlapping area,the greater the impact on price.Considering that the VNO uses spectrum reuse technology to provide services for small base stations in practical applications,we further propose MSRT-SPA for practical applications.

Fig.3.Base station dynamic coverage analysis model.

3.2 Construction

Spectrum reuse is the basic technology of cellular mobile communications.The basic idea is to reuse the frequencies of several cells by reuse-frequency,increasing the efficiency of spectrum utilization and increasing operator revenue.For the spectrum reuse technology,the spectrum usage rules for each area in figure 3 are set as follows:

1) Area II and area III use the same spectral resource,i.e.,the spectral allocation of these two areas is the maximum of the both;

2) When the VNO deploys small-cell services in the area II and area III,the spectrum reuse technology is applied and set its spectrum reuse factor FRF=3;

Therefore,the VNO needs to compare the resource requirements of area II and area III that can provide small base station services.Then,the VNO selects the maximum value as the total number of resources that it needs to rent from TNO ultimately.

MSRT-SPAS scheme is divided into two parts.First,we obtain the subgame equilibrium of the virtual operator,determining the equilibrium price(j=0) of the small base station service.Then,we determine the equilibrium priceof traditional operator under the total system bandwidthB jj(=0).

Set the initial price values(j=0) and(i=0,j=0) for the different services in the network,and the initial values(j=0) and(i=0,j=0) for the operator's revenue.The algorithm 2 is run for obtaining the equilibrium price(j=0).

The final equilibrium price of the virtual operator is related to the equilibrium price of the traditional operator directly.The equilibrium price of traditional operator in MSRT-SPAS can be obtained through Algorithm 3.Traditional operators allocate spectrum on demand according to their grasp of the entire network resource.

IV.SIMULATION-BASED ANALYSIS

4.1 Analysis of GM-SPAS

Figure 4 shows the trend of the equilibrium price of TNO and VNO with the total bandwidth of the system.When network resources are scarce,operators increase the service price in order to maximize revenue.When the network resources are relatively abundant,the decrease of the price can increase the user's demand for resources.In this paper,we use the game model to design the spectrum pricing and allocation algorithm.The operator maximizes its own revenue and finds the game equilibrium point finally.

Algorithm 2.Virtual operators sub-game equilibrium.1:if pp i jj S M,(0,0) ( 0)ijj==＞=2:Calculate the current size of the resource requirements for the different regions 3:if QQ S S φ2( ,i j)(i=0,j=0) (i=0,j=0)≥ φ3(,)i j 4:PAN ←QMMS φ1 + +QQφ3 φ2 FNRF/5:if PAN B k≤=k( 0)6:Ri,j S(i,j)==1 0 ← the current revenue of VNO 7:if RR i,ji,j S S (i,j)(i,j)==＞==1 0 0 0 Record prices for services and operator's revenue at this time 8:else if QQ S S φ2( ,i j)(i=0,j=0)＜ (i=0,j=0)φ3(,)i j 9:PAN ←QMMS φ1 + +QQφ3 φ3FNRF/10:if PAN Bk k≤=( 0)11:Ri,j S(i,j)==1 0 ← the current revenue of VNO 12:if RR i,ji,j S S (i,j)(i,j)==＞==1 0 0 0 Record prices for services and operator's revenue at this time 13:Change the service price of small cell according to the interval,i.e.,replace pi j S,(1,0)ij==with pi j S,(0,0)ij==.Repeat the above loop and determine the VNO's revenue again 14:Get and record the best small cell price pS j opt- (j)=0 ;

As can be seen from Figure 5,when the bandwidth allocated to macro-cell services increase from 30% to 75%,the operator's revenue drops significantly until the revenue becomes less.Therefore,compared with the non-on-demand resource allocation scheme,the on-demand spectrum pricing and resource allocation scheme based on the game model has great advantages for operators,especially when the network resources are scarce.Allocating resources on demand can not only effectively increase operator's revenue,but also greatly increase resource utilization.

Figure 6 shows the resource requirements of different operators in the game equilibrium state.The sum of resource requirements of the two services is almost equal to the corresponding total system bandwidth.It is indicated that the scheme proposed by Section II makes full use of network resources,improves the efficiency of resource use and the benefits of TNO.

The revenue simulation of TNO and VNO is shown in figure 7.The figure 7 shows that as the system bandwidth resources increase,operators gradually increase and eventually reach a stable state.When the total bandwidth resources are less,the operator's revenue grows rapidly.When the spectrum resources of the network are abundant,the revenue growth is relatively slow.The algorithm proposed in this paper can improve the operator's revenue,but operators should apply for spectrum resources appropriately to avoid wasting spectrum resources.

Algorithm 3.Traditional carrier equilibrium price.1:if p (j ) p-=＞0 (j)=0 2:Calculate the current size of the resource requirements for the different regions 3:if QQ j optj S M φ2( ,i j)(i=0,j=0) (i=0,j=0)S ≥ φ3(,)i j S φ1 + +QQφ3 φ2 FNRF/5:if PAN B k≤=k( 0)6:R M 4:PAN ←QMMS j(j)=1 ← the current revenue of TNO 7:if RR j j 0 Record prices for services and operator's revenue at this time 8:else if QQ M (j)(j)=＞=1 M φ2( ,i j)(i=0,j=0)＜ (i=0,j=0)S φ3(,)i j S φ1 + +QQφ3 φ2 FNRF/10 :if PAN B k≤=k( 0)11:R M 9:PAN ←QMMS j(j)=1 ← the current revenue of TNO 12:if RR j j 0 Record prices for services and operator's revenue at this time 13:Change the service price of macro-cell according to the interval,i.e.,replace pM j (1)j=with pMj ( 0)j=.Repeat the above loop and determine the VNO's M (j)(j)=＞=1 M revenue again 14:Get and record the best macro-cell price,i.e.,the equilibrium price pMopt.And the final small cell service equilibrium price pSopt 15: Change the total resource of the system, i.e., replace Bk(k)=1 with Bk(k)=0 .Repeat the above loop and find the game equilibrium point under different network state Bk(k,,,....)=1 2 3 4

Fig.4.Equilibrium price changes with the total resource of the system.

Fig.5.Comparison of traditional operations' revenue.

4.2 Analysis of MSRT-SPAS

In the context of dynamic coverage of the base station,we consider the effects of the algorithm and perform simulation analysis from two aspects.

1) The impact on the operator's revenue

Figure 8 analyzes the changes in the operator's revenue by using our proposed MSRT-SPAS and BSDC-SPAS schemes.It can be seen that the operator equilibrium has increased significantly after we consider MSRT into the algorithm.Compared with TNO,this technology is more conducive to improving the benefits of VNO.

2) The impact on different services equilibrium prices

Fig.6.Resource quantity of demand (QOD) changes with the total resource of the system.

Fig.7.TNO and VNO's revenue changes with the total resource of the system.

Fig.8.Influence of multiple spectrum reuse technologies on the equilibrium revenue.

Fig.9.Influence of multiple spectrum reuse technologies on the equilibrium price.

Figure 9 shows the price comparison between the application of MSRT and the dynamic coverage of direct base stations without MSRT.It can be seen that as the system resources increase,the prices of both TNO and VNO are reduced.Therefore,MSRT-SPAS not only improves the resource utilization,but also has a positive impact on the efficiency of operators and users,and it can improve the overall efficiency of the network.

V.CONCLUSION

In this paper,we focus on the problem of wireless resource management mechanism in cellular networks under the F-RAN framework.For the social background of operator carrier diversification,we use the game theory in the economic model to explore the spectrum pricing and distribution algorithm,which according to the system model of double-layer network structure and the distributed resource supply model proposed in this paper.The simulation results and analysis show that the solution can improve the operator's revenue,achieve the on-demand spectrum allocation of operator's and enhance the efficiency of network resources under the premise of meeting the needs of multiple parties.

ACKNOWLEDGEMENT

This work was supported in part by the National Natural Science Foundation of China (61771120),and the Fundamental Research Funds for the Central Universities (N171602002).