Research On Efficient Task Offloading Strategy In The Mobile Edge Computing

Mobile Edge Computing(MEC)is a computing mode in the mobile edge network environment.MEC migrates computing tasks generated by user equipment in the network to computing nodes located at the edge of the network for calculation,and returns the calculation results to the equipment.This process is called Task Offloading.While task offloading greatly improves the running speed of programs in user equipment,it can also reduce communication delays,save equipment energy consumption,and improve the Quality of Service(QoS)of user equipment in the network.In practical applications,due to the limited computing power and network bandwidth of edge servers,resource competition will occur when users in the same area send a large number of computing offloading requests at the same time,which affects MEC network performance.Therefore,while applying the MEC technology,a corresponding task offloading algorithm must be designed to provide an offloading strategy for the user equipment.How to formulate an offloading strategy to minimize the energy consumption of user equipment while taking into account the constraints of computing capacity,network bandwidth and other factors,as well as ensuring low-latency service quality requirements,is a very challenging topic.In this article,we first model the actual application scenarios of MEC,and formulate offline mobile task offloading problems and location-aware online mobile task offloading problems.After that,for the offline task offloading problem,this article gives an approximate optimization algorithm that can give an accurate solution;for the location-aware online task offloading problem,this article gives a reinforcement learning based method for user equipment to submit an offloading task,and a deep reinforcement learning algorithm used to allocate computing resources for MEC.Finally,for the above three algorithms,this paper verifies their performance through simulation experiments,and compares their performance with other benchmark algorithms.The experimental results show that the algorithm proposed in this paper not only meets the feasibility,but also has better performance than the benchmark algorithm.