Mobile Edge Computation Offloading And Caching Based On Deep Reinforcement Learning

With the increase of the number of mobile devices and intelligent devices,the computational complexity of applications is becoming higher and higher.Mobile edge computing is a new computing paradigm in the current Internet.It is deployed at the edge of the Internet and can serve users more efficiently.This paper will focus on two aspects of computation offloading and edge caching in current mobile edge computing.For computation offloading,the current computing intensive tasks are generally composed of multiple subtasks with computing dependencies.In order to solve the problem of computation offloading between multiple subtasks with computing dependencies,this paper proposes a Deep Reinforcement Learning based Task Mapping Algorithm.First,the computation intensive task is modeled as a Directed Acyclic Graph,and then a Graph Sequence Algorithm is proposed to transform the modeled task into a specific topological sequence.The Markov Decision Process is established to process the computation offloading,and the optimal offload decision of all subtasks is determined according to the sequence sequence.The goal of the algorithm is to maximize the comprehensive benefits of latency and energy consumption.Experimental results show that the algorithm has strong decision-making ability and can obtain approximately optimal comprehensive benefits,which proves the effectiveness of the algorithm.For edge caching,in the current Internet environment,popular data will be accessed by many users.Once the number of requests increases,it will increase the burden of the network and server.The content caching of edge nodes is a very effective technology to reduce data traffic.This paper proposes a deep reinforcement learning framework for base station content caching: KNN-DDPG.The purpose of this framework is to maximize the weighted cache hit rate without understanding the content popularity distribution.By combining KNN algorithm to reduce the size of action space and improve the computational performance of the model,the experimental results show that KNN-DDPG has a higher and more stable weighted cache hit rate than other baseline cache algorithms.