Research On Joint Optimization Of Network Resources In Mobile Edge Computing Scenario

With the advent of 5G era,the explosive growth of smart devices,especially mobile phones and Internet of Things devices,makes users'demand for low-latency and high-quality services such as Augmented Reality(AR),Virtual Reality(VR)and mobile games constantly increasing.At the same time,the demand for computing and storage resources of these emerging services is also increasing.Mobile Edge Computing(MEC),as a new category of network architecture,aims to respond to and process service requests of edge users by utilizing the computing and caching resources of base stations and nodes on the edge of the network,thereby alleviating the pressure of the core network,reducing the energy consumption of the network and the overhead of the forward link,and reducing the execution delay of tasks to make up for the shortage of mobile devices' own resources and to meet the demand of network resources for computing-intensive,ultra-low latency emerging services.Aiming at reducing the energy consumption of the system,this paper studies the optimization of MEC system in three different network scenarios.Specific research contents include the following aspects:The joint optimization of offloading strategy and wireless resource allocation is studied in the network of multi-user and multi-MEC servers.The algorithm divides the original problem into two sub-optimization problems:lower-level and upper-level.In lower-level problem,some mathematical methods and theories such as relaxation transformation,Lagrange multiplier method and gradient descent is used to optimize the subcarrier and transmission power allocation strategy,while the offloading strategy is optimized in upper-level problem by searching on the limited set based on the lower-level optimization results.This algorithm effectively reduces the overall energy consumption of the system in the multi-user and multi-MEC server complex network scenario,and the optimization effect is more obvious especially in the relatively sparse network scenario.In the multi-user multi-MEC server energy harvesting scenario,the joint optimization of offloading strategy,wireless and computing resource allocation is studied aiming at minimizing system energy consumption.Due to the fluctuation of wireless channel,the time variations of transmission and computation process are quite different.The resource allocation optimization of transmission and computation process needs to be carried out at different timescales.The whole optimization process is divided into two stages:short-term and long-term.The wireless resource allocation is optimized at short-term stage,and the offloading strategy and computing resource allocation are optimized at long-term stage.The joint optimization algorithm proposed in this study effectively reduces the energy consumption of the system,fully exploits the advantages of user diversity and diversity of MEC servers.In addition,by adjusting the energy value factor of green energy MEC servers,the utilization rate of green energy can be effectively improved.In the multi-user multi-MEC server MEC system,the joint optimization of the offloading strategy,network resource allocation and the working mechanism control strategy of base station and MEC server is studied.Considering that the resource allocation and the working state control of base station and MEC server are in different timescales,the separation optimization strategy based on mixed timescale is also adopted.Stochastic optimization algorithm optimizes the offloading strategy,wireless and computing resource allocation in the short-term stage.In the long-term stage,DQN reinforcement learning algorithm is used to optimize the working state control mechanism of communication and computing nodes.This research effectively reduces the energy consumption of the system,improves the performance of the whole network,and maintains the stability of the system state.To sum up,this paper mainly optimizes the performance of MEC system in different network scenarios from the perspectives of task offloading strategy,network resource allocation and working mechanism regulation of base station and server nodes in the network.At the same time,the fluctuation of wireless channel,mixed timescale characteristics of network resource allocation and working mechanism regulation of nodes are taken into account.Challenged by complex multi-variable and multi-constrained nonlinear programming problems,we adopt bi-level and mixed timescale stochastic optimization strategy with the specific methods including Lagrange multiplier method,gradient descent,norm transformation,relaxation transformation and DQN reinforcement learning algorithm.The simulation results show that the proposed optimization strategy and algorithm can effectively achieve the optimization goal of reducing the energy consumption of MEC system and improve the performance of MEC system.