Research On Stochastic Geometry Theory Based Performance Analysis And Resource Allocation In Ultra Dense Networks

In order to realize the vision of future 5G networks--ultra high speed,ultra high capacity,ultra low latency and high energy efficiency,increasing the type and number of access points(APs)is widely considered as one of most efficient methods.As a result,the future communication networks are becoming denser and more heterogeneous,which gradually evolves into ultra dense networks(UDNs).Compared with traditional communication networks,there are several new characteristics for UDNs,such as the number of APs increases exponentially,the distances between transmitters and receivers are greatly shortened,the concept of cell-edge is weakened,and the networks transit from "base station-centric" to "user-centric",thus the network performance such as system capacity would be significantly improved.Based on the new characteristics of UDNs,traditional methods for theorectical analysis and resource allocation are no longer applicable.On one hand,for the modeling of node spatial distribution,most traditional theoretical methods for performance analysis utilize deterministic models,which relies heavily on Monte Carlo simulation or idealistic assumptions.However,when faced with the heterogeneity,randomization and densification of APs in UDNs,traditional modeling methods are too realistic,so there is an urgent need for a new-type method to describe the random topology of UDNs and further conduct accurate performance analysis.On the other hand,in UDNs,Channel State Information(CSI)explodes due to a sudden increase of the number of APs and users.Meanwhile,in addition to the traditional radio resources such as space,time and spectrum,computing,caching and other multi-dimensional network resources are involved,which considerably increase the dimension,complexity and feedback of resource allocation problems.As a result,it is difficult for point-by-point measurement and point-by-point feedback based traditioanl resource allocation methods to meet the requirements.To solve the key problems of performance analysis and resource allocation in UDNs,this paper utilizes stochastic geometry theory,stochastich process,probability theory and other methemetical methods to analyze network performance for different scenarios in UDNs,build a new framework for performance analysis in UDNs,and design low-complexity and low-overhead resource allocation schemes,aiming to explore the evolution,development and implementation of UDNs.The main contributions are as follows:1.Stochastic geometry based performance analysis for static scenarios in UDNsIn heterogeneous networks,from the perspective of the cell,we derive the average area-weighted load,average ergodic user rate,area-weighted spectrum efficiency(SE)and energy efficiency(EE).From the perspective of the network layer,we derive the average SE and EE for each layer.From the perspective of the whole network,we derive the network SE and EE,and study the impact of bias factor on the tradeoff between them.In addition,for UDNs adopting coordinated multipoint(CoMP)technology,we propose new performance metrics such as successful service probability(SSP)and effective ergodic capacity(EEC),and compare the performance of users,cells and the network when adopting joint transmission(JT)and coordinated scheduling/coordinated beamforming(CS/CB),respectively.Both of these two research points are validated by comparing the theoretical results and simulation results.Relevant research results of the performance analysis in static scenarios can provide useful theoretical guidance for the issues of UDNs such as the setting of bias factor in heterogeneous networks with cell range expansion,the selection of CoMP schemes and the setting of cluster size in CoMP networks.2.Stochastic geometry based performance analysis for mobile scenarios in UDNsFirst,in mobile communication networks,for the five basic handover criteria,we use stochastic geometry theory to analyze the relationship of the relative locations between base stations and mobile users,derive the closed-form expressions of cluster handover probability for different handover criteria,and study the effects of different handover schemes on handover performance.Second,considering the impact of outdated CSI on handover performance,two indicators,i.e.,false handover probability and miss handover probability,are proposed to measure handover failure probability brought by outdated CSI.Third,for the UDNs adopting CoMP technology,considering the effect of user mobility,this paper studies the relationship between outdated and perfect CSI,derives network coverage using stochastic geometry,and explores the effects of outdated CSI on the mobile network performance.All of these three research points are also validated by comparing the theoretical results and simulation results.Relevant research results of the performance analysis in mobile scenarios(such as handover probability,handover failure probability,network coverage)can provide useful insights for issues of mobile UDNs such as the selection of handover schemes in CoMP networks,the impact of mobility on handover failure probability,and the deployment of base station density and cluster size.3.Research on stochastic geometry based resource allocation for UDNsTo deal with the high dimension,high comlexity and large feedback issues of the resouce allocation problems in UDNs,two resource allocation schemes are proposed,i.e.,stochastic geometry theory based hierarchical power allocation scheme,and computation offloading and content caching scheme.For the uplink scenario in UDNs,considering a discrete power set,so the interfering users can be divided into several groups according to their transmit powers.Then relevant performance metrics can be derived using stochastic geometry methods,which can guide the design of alternative genetic algorithm based hierarchical power allocation scheme.For the UDNs with computation-intensive services,the users can choose to offload the computing task to a nearby AP or a group of D2D(Device to Device)users,and decide whether to cache the computing results or not.Then the design of the optimal computing offloading and caching schemes to maximize the net revenue of offloading and caching is written as an optimization problem,which is solved by the proposed distributed ADMM(Alternating Direction Method of Multipliers)based algorithm.Simulations results show that with the help of stochastic geometry theory,only a small part of CSI(useful signals of each user)is required for these two schemes,which can efficiently reduce the computing complexity and signaling overhead for the resource allocation problems while guranteeing relevant user performance.Related research results can efficiently solve the issues of high complexity and large feedback overhead for traditional wireless resource management and control,which can also provide an important theoretical basis for the engineering practice of wireless and network reource allocation in UDNs.