Research On Technologies Of Mobile Edge Computing And Utility Optimization In Ultra-dense Relay Network

Driven by the 5G communications vision,the demand for mobile communications services has surged.Ultra-dense relay network(UDRN)is a key technology of 5G.It can cope with future exponential traffic from indoor/congested areas,and ensure that spectrum efficiency and energy efficiency are improved due to shortened transmission distance.However,UDRN deployment brings new technical challenges,such as unbalanced load,severe interference,unfair radio resource sharing,unnecessary handover,low radio resource utilization efficiency,high energy consumption,and reduced service quality.In order to overcome these challenges and meet the performance requirements of 5G,UDRN needs to be combined with other 5G support technologies,such as wireless caching and mobile edge computing(MEC),and adopts a new generation of distributed optimization strategies such as stable matching.Caching technology can meet users' demands under peak traffic load and reduce the load of wireless and backhaul resources.The combination of MEC and UDRN can not only deal with access from public communication devices,but also provide powerful computing capabilities for users at the edge of the wireless network.MEC-based UDRN can effectively handle computationintensive and data-intensive tasks.This thesis discusses the strategy design of UDRN systems that combine with different key technologies.The main contributions are as follows:Firstly,we study the utility optimization and stable matching technology of UDRN.Considering the utility based on rate and energy consumption,a scheme that jointly optimizes power allocation and user-relay matching is proposed to maximize the system utility.The users and relays are divided into several clusters,and performs optimization strategy within a cluster.Users and relays are modeled as energy buyers and sellers,respectively.The optimal power purchased by the user from the relay is obtained through the golden section method for maximizing the user utility.The mutual preference matriices of users and relays are obtained.An improved stable matching algorithm based on Gale-Shapley(GS)algorithm is proposed to obtain the user-relay stable matching result.Simulation results show that the proposed scheme has lower complexity,higher energy efficiency,extremely low communication overhead compared with conventional algorithms,and the performance is close to optimal.Secondly,we study the relay cache technology of UDRN.A user preference-awera optimization algorithm is proposed for maximizing system utility.The users and relays are divided into several clusters,and then performs optimization strategy within a cluster.The file caching scheme is designed by taking both user preference and file popularity into account,Then,the optimal power purchased from the relay can be obtained by Karush-Kuhn-Tucker(KKT)conditions according to the caching result,and the mutual preference lists of the user and the relay are obtained,Then,the improved stable matching algorithm is applied to obtain the user-relay stable matching results.Simulation results show that caching can make better system performance,and considering user preference can bring considerable system performance gains.Finally,we study the relay MEC technology of UDRN.Considering the cost based on delay and energy consumption,a scheme that jointly optimizes task caching,offloading and stable matching is proposed to minimize the system cost.Under the condition that the matching scheme is unknown,an improved branch and bound algorithm and a low complexity suboptimal algorithm are proposed to obtain the optimal and suboptimal solutions,respectively.Then,according to the principle of minimizing individual costs,the mutual preference lists of users and relays are obtained,and the improved stable matching algorithm is applied to obtain the user-relay stable matching results.Simulation results show that the system can effectively reduce the system cost compared with the system without relay/caching/MEC;the algorithm can bring considerable performance gains compared with other conventional algorithms,and low complexity suboptimal algorithm achieves the performance close to the performance of the improved branch and bound algorithm with less time.