Joint Optimization Of Compting,caching And Energy Harvesting In Mobile Edge Computing Based On Deep Reinforcement Learning

With the rapid development of the big data and 5G communications,how to reduce time and energy consumption to meet new applications has become one of the hot topics in recent years.As a promising computing paradigm,mobile edge computing(MEC)transfers computation and storage capabilities from the cloud to the edge by collecting and collecting a large number of idle devices distributed at the edge of the network,thereby improving the quality of service(Qo S)for computation-intensive and sensitive-latency tasks.However,how to reasonably design the service offload strategy is the challenge faced by MEC-enabled systems.Since virtualization has begun to converge the functions of many services including communication,computing,caching,and harvesting,the joint optimization scheme between them has also become a challenge for MEC-enabled systems.In this paper,a joint optimization problem among computation offloading,content caching,and energy harvesting in MEC-enable system is studied to minimize time and energy consumption.Combined with the popular deep reinforcement learning(DRL)method,the optimal dynamic decision of resource allocation is designed by using distributed edge information.Simulation experiments demonstrate the learning ability and performance of the proposed algorithm.The major contributions of this paper can be summarized as follows:We are the first to propose the joint optimization problem among computation offloading,content caching and energy harvesting in MEC-enabled system with objective of minimizing time and energy consumption.In order to make full use of various information about MEC-enabled system,we define more comprehensive states,actions and rewards.Considering the dynamics of the communication network cannot be accurately known,such as the random arrival of tasks,time-varying channels,etc.,we cannot make accurate assumptions.So we use model-free method without prior knowledge the MEC-enabled system.We propose a novel DRL-based algorithm to make dynamic decisions on resource allocation in the MEC-enabled systems Among them,the spatial attention module is for more targeted optimization,and the multi-agent deep deterministic policy gradient(MADDPG)baseline is to use the information of different edges for cooperation and competition between the edges.Numerical results show that the proposed algorithm can improve the performance in different scenarios effectively.