A New Computing Offloading Method And Research On QoS Optimization In Edge Computing

The rise and development of 5G technology has derived a large number of services and applications,such as Internet of Vehicles,augmented reality and natural language processing.Nowadays,people's demand for intelligent applications is increasing day by day,and the generated data traffic is exploding.This trend poses a huge challenge to user equipment.Traditional cloud computing provides users with powerful services in a centralized computing model.However,cloud computing has significant shortcomings in terms of service efficiency and privacy,mobile edge computing models have emerged.Mobile edge computing moves computing and storage to devices that close to user equipment to reduce delay and energy consumption.Computing offloading is the key issue in mobile edge computing.The existing methods of computing offloading mainly have the following two limitations:(1)Most of the goals are based on the optimization of delay and energy consumption,without considering the real-time requirements of users;(2)The dynamic nature of the mobile edge computing system environment is not considered,and the interaction between the offloading decision and the system environment is ignored.To address the above two problems,this paper proposes effective solutions.A local computing model and a computing offload model are established in multi-user singlecell MEC system environment.This paper takes Quality of Service(QoS)as the optimization goal,and takes the real-time nature of the application as an important factor which influencing QoS.We propose the concept of Age of Task(AoT),and introduce it into the new definition of QoS.The edge computing system changes dynamically,and the decision result of the entity will affect the long-term benefits of the system.In this paper,the computing offloading process is modeled as a Markov decision process.Deep reinforcement learning is used to solve the computing offloading problem.We propose a new intelligent computing offloading method based on the DQN algorithm.In addition,two improved DQN algorithms are extended,and three intelligent algorithms are compared and analyzed.The experimental results show that the new computing offloading method proposed in this paper is significantly better than the full offloading method and random offloading method in many aspects such as QoS,throughput,and energy saving.It proves that the computing offloading method in this paper can significantly improve quality of service and improve the overall efficiency of the system.In addition,comparing the three DQN algorithms,it can be seen that the two improved DQN algorithms perform better than the traditional DQN algorithms,among which the Dueling-DQN algorithm is better.