Optimal Uploading Strategy Of Mobile Edge Computing Based On Constrained Markov Decision Process

In the 5G era,with the increasing popularity of mobile devices such as smart phones,tablets and wearable devices,people are accelerating the emergence of the Internet of Things(IOT),resulting in an explosive increase in the amount of data.The storage,analysis and processing of these data bring great pressure to the computing power of mobile devices,and cannot provide a satisfactory experience quality.In order to cope with the limitation of computing power of mobile devices,computing offload is a method to alleviate the computing burden of mobile devices,and offloading computing tasks of mobile terminals to remote cloud servers is a solution of mobile cloud.However,considering the low bandwidth and service availability of mobile cloud,it will cause long-term delay of data transmission from mobile devices to remote cloud servers.For timecritical applications,sending data to remote cloud can’t meet the delay requirements.To solve these limitations,Mobile Edge Computing(MEC),as a new architecture,introduces network edge to make up for the shortcomings of remote cloud,thus shortening the propagation delay and improving the computing power of mobile devices.Compared with traditional cloud computing,edge computing has the advantages of reducing network bandwidth loss,improving the real-time response of the system and reducing the risk of data leakage.In addition,it is a promising technology to execute computing tasks on the mobile edge server,which can realize the computing agility and high computing power in MEC.Especially for computing tasks with high delay requirements,it is very important to study how to choose the optimal offload strategy to minimize the delay in the network environment with channel fading.In this paper,we consider that in a MEC system,there are many edge servers distributed near the terminal equipment,and the terminal equipment unloads the computing task to the edge servers with a certain probability at every moment,and the network bandwidth of the edge servers can dynamically change and the wireless channel is attenuated.In this paper,aiming at minimizing the unloading delay,considering the energy consumption of the terminal equipment when unloading computing tasks,the optimal unloading time optimization problem of a single terminal is defined as a CMDP,and the optimal strategy is solved by strategy iteration and Q-learning algorithm,so that the mobile terminal can achieve the optimal unloading time by selecting the optimal unloading node and transmission power in the environment with wireless channel attenuation under the constraint of minimum energy consumption.According to the above model,this paper evaluates the performance of terminal equipment in selecting unloading nodes by numerical simulation with Matlab,and mainly discusses the optimal hybrid strategy of terminal equipment when it has computing tasks.The main results are as follows: with other parameters unchanged,the larger the data uploaded by the terminal device,the longer the unloading delay.With the increase of available energy consumption of terminal equipment,the constraint in CMDP model will be smaller.The terminal equipment will choose the edge server with good network bandwidth as the unloading node,and at the same time,it will upload data with larger transmission power,so the unloading delay will be reduced.The conclusion of this paper provides a reference for time-critical applications with energy consumption constraints,and has positive significance for the sustainable use of terminal equipment.