Computing Offloading And Resource Scheduling In Mobile Edge Computing Networks

With the development of communication technology and the prosperity of the mobile Internet,the increasing popularity of mobile devices has led to an exponential increase in the demand for high data rates and high computing capabilities.More and more existing and emerging applications,such as natural language processing,virtual reality(VR),and interactive games,require low latency,as well as intensive computations.Therefore,the future mobile communication network will face huge challenges in terms of system capacity,low latency,high density,and high reliability.To solve these problems,gradually sinking computing tasks to edge nodes close to terminal devices for processing has gradually become a trend of network architecture evolution.Computing offloading and resource scheduling problems in mobile edge computing networks have become important indicators that affect network performance.Due to the ever-increasing demands for delay-sensitive applications in 5G,multi-access edge computing networks are expected to support real-time interactive systems.This article considers the scenario of a single base station with multiple users.The base station is equipped with a server.Each user has a time-sensitive task waiting to be processed.In order to reduce the corresponding delay of the service,the task scheduler needs to make decisions based on the relevant information of the entire network.Where the computing tasks are processed,they are processed on the local device or offloaded to the edge node base station for processing? In order to minimize the weighted sum cost which includes the energy consumption and delay consumption of the entire system,this paper considers the joint computing offloading and resource allocation problem,where the resources here mainly include frequency band resources and computing resources.The problem is formulated as a mixed integer and nonlinear programming problem(MINLP),usually an NP-hard problem.In order to solve the above mentioned non-convex computational offloading and resource scheduling problems,we propose two methods to solve them,one is a learning-based method,and the other is an optimization-based method.The main contents are as follows:1.Inspired by the application of reinforcement learning in control decision-making,we adopt a novel method based on reinforcement learning and convex optimization algorithm.In the first step,the agent learn the offloading decision policy of the task scheduler through the reinforcement learning algorithm,and then use the convex optimization method to obtain the optimal resource allocation strategy based on the above offloading set.2.Based on the above model,in order to make better use of spectrum resources and reduce energy consumption,we have further studied the sub-channel allocation and transmit power control optimization under the orthogonal frequency-division multiple access(OFDMA)mechanism.With the goal of minimizing system consumption,including the energy consumption and execution delay of data transmission and task execution,we considered the joint computing offloading,sub-channels assignment,transmission power optimization,and computational resource allocation issues.Due to the nonconvexity of the objective function,we proposed an iterative algorithm based on successive convex approximation(SCA)to solve it.Related simulation results show that the proposed two methods in the article can solve the problems of computational offloading and resource scheduling more effectively.Compared with some common baseline methods,our proposed algorithm can reduce the energy consumption of the system better and obtain better performance.