Research On Mobile Edge Computing Offloading For Multiple Servers

With the popularity and development of mobile intelligent terminals and emerging intelligent applications(such as automatic navigation,face recognition,etc.),mobile devices have encountered great challenges due to their limited computing power and battery capacities.Traditional Cloud Computing has strong computing power,but subject to long response delay and overcrowded backhaul bandwidth.In order to overcome these difficulties,experts began to study Mobile Edge Computing(MEC).At the same time,which subject to relative less transmission load and short delay.Existing MEC studies mainly target at reducing users' delay,energy consumption,but neglect economic factors.However,economic factors play a crucial role in offloading decisions.In the distributed decision-making scenario,users compete for resources,and the decisions of users influence each other.The actual computing rate on the server side is limited,and the existing literatures are idealized.And the users' information is limited,may not be able to get global information.In addition,traditional distributed non-cooperative offloading strategies all consider a small number of devices,because increasing the number of devices will increase the overhead of system control signaling,which make them very sensitive to the number of devices.This thesis first investigates the research status of MEC,and analyzes the current research concerns and shortcomings,which is the reason and entry point of this thesis.Secondly,it introduces the theoretical tools used in this study,including matching theory,queuing theory and evolutionary game theory,which lay a foundation for the research.Next,the task offloading strategy based on the matching theory is studied to optimize the allocation strategy among users,computing and communication resources,so as to maximize the users' satisfaction in the system.Considering MEC and D2D communications,the concept of cost performance ratio is proposed to measure the users'satisfaction.The budget scheme is set according to the task characteristics of users,and the charging scheme of communication resources is designed and calculated according to the principle of "better quality,higher price".The Improved Kuhn-Munkras algorithm and Matching Based Offloading Strategy(MBOS)algorithm are introduced to optimize the resource/OFDMA subcarriers allocation among users and helpers.Simulation results show that the MBOS algorithm can achieve efficient allocation of users,communication resources and two kinds of computing resources,significantly improve users' satisfaction in the system.Finally,the Population Game Based Offloading Strategy(PGBOS)algorithm is designed to minimize the cost of each user under distributed decision-making.Some assumptions in traditional game theory are not realistic and only applicable to a small number of users.Therefore,the evolutionary game model is introduced to transform the evolution of individual decision into the evolution of population state space.The dynamic evolution strategy is used to carry out the population evolution,considering the limited processing speed of the server,and the corresponding algorithm is designed to reduce the queuing delay.Simulation results show that this algorithm can significantly reduce the average cost of users in the system.At the same time,the algorithm can save signaling cost,is not sensitive to the number of users,and has more practical significance.