Research On Caching And Collaborative Offloading Strategies For Edge Networks

The rapid development of 5G communication technology(5G)and Internet of Things(Io T)technology has given rise to massive latency-sensitive tasks and computation-intensive tasks,bringing problems such as high load,high energy consumption,and high latency to mobile edge computing(MEC)system brings high load,high energy consumption,and high latency problems.How to mitigate the high latency problem,improve user experience,and reduce system energy consumption by reasonably utilizing the storage and computation resources on MEC servers is one of the current research hotspots in MEC technology.This paper integrates MEC technology and deep reinforcement learning technology,and studies the problem of cache and collaborative computing offloading in MEC.The main research contents of this paper are as follows:1.Based on the heterogeneous network scenario with multiple types of request tasks and multiple MEC servers,a collaboration computing offloading strategy based on MEC technology is proposed.A collaboration computing offloading algorithm based on Deep Q-Networks(DQN)is designed to minimize the total processing latency of the system,and the collaboration computing offloading of request tasks is performed based on the remaining computing resources of each MEC server in the MEC cluster and the corresponding unprocessed request task types.Combined with deep reinforcement learning theory,the collaborative computing offloading process is constructed as a Markov Decision Process(MDP),and the corresponding state,action,and reward are determined.Through learning and training the model,the intelligent agent can obtain the optimal action based on the corresponding state to perform collaborative computing tasks.Simulation results verify that the proposed approach can effectively reduce the system’s average processing latency.2.In the heterogeneous network scenario with multiple types of request tasks,multiple MEC servers,and a central cloud server,an asynchronous content caching and collaborative computing offloading strategy is proposed.The caching decision and collaborative computing offloading decision are decoupled in the time scale,and the content caching is updated on a long-time scale,while the request tasks’ collaborative computing offloading is performed on a small-time scale,thereby improving the system’s stability.For the two decoupled subproblems:(1)a caching decision algorithm based on Double Deep Q-Networks(DDQN)is designed,aiming to maximize the long-term average cache hit rate of the system based on the long-term arrival of request tasks to obtain the optimal cache replacement decision.(2)Combined with the caching decision,a collaborative computing offloading decision algorithm based on Deep Deterministic Policy Gradient(DDPG)is designed.Considering the cache content and computing resources on MEC servers,the collaborative computing of each server in the MEC cluster and the central cloud server is considered,aiming to minimize the system’s average total processing latency and maximize the overall computing resource utilization rate of the MEC cluster to obtain the optimal request task offloading decision.Simulation results verify that the proposed approach can effectively reduce the system’s average processing latency.