NOMA-Based Task Offloading Delay Optimization In Edge Computing

With the further acceleration of the development speed of the current mobile communication network,the business volume of the mobile communication network has experienced a rapid growth,and at the same time,users have higher and higher requirements for network experience.More users and Io T devices begin to handle computationally intensive tasks that require complex computing,but the task processing capabilities of devices are limited by insufficient processor performance and insufficient battery capacity to meet user needs.Mobile edge computing emphasizes proximity to mobile users and devices,focusing on deploying MEC computing service platforms with computing,storage and communication capabilities at the edge of the network to provide edge services as close to users as possible.As an important development direction in mobile edge computing,edge computing offloading can meet the requirements of 5G for low latency,and is gradually becoming an increasingly important technology in the 5G era.Although edge computing offloading can meet the high computing requirements of users,it also brings an increase in energy consumption and transmission delay of offloading tasks to edge servers.Therefore,in order to better utilize the advantages of edge computing offload,non-orthogonal multiple access(NOMA)is introduced into edge computing.Non-orthogonal multiple access is also considered as one of the key technologies of the fifth-generation mobile communication system.This technology can simultaneously serve multiple users on the same transmission frequency band,thereby reducing the transmission delay.But how to effectively combine the two technologies of NOMA and MEC will be a key issue.This paper discusses the delay optimization problem of edge computing offloading based on non-orthogonal multiple access,which is mainly composed of the following two parts:The first part of this paper proposes a system model for the overall offloading of edge computing using non-orthogonal multiple access in a single cell.And put forward the optimization objective and formula to minimize the total delay,and solve the problem one by one by decomposing the problem.Firstly,by discussing the situation of the power of two users in a channel under different transmission completion orders,the approximate optimal power allocation method is obtained through a series of mathematical analysis.The user and channel are then paired by polling to try the pairing method to obtain the final stable optimal pairing method.Finally,the optimal MEC server computing resource allocation scheme is obtained by the Lagrange multiplier method and the KKT method.Finally,the effectiveness and superiority of the scheme are proved through experiments,and the effect of approximate optimal solution can be achieved.The second part of this paper proposes a system model for partial offloading of edge computing using non-orthogonal multiple access in a single cell.The optimization objective and formula to minimize the total delay are proposed.Firstly,a DQN-based NOMA-MEC system delay optimization method is proposed.The decision is made by splitting the percentage offloading decision,power allocation decision and offloading decision into three DQN units.Then,based on the fact that DQN is not suitable for continuous value decisions such as percentage unloading decision and power allocation decision,the two DQN units are replaced with DDPG units for decision-making.Finally,the experiment proves the effectiveness and superiority of the scheme.