Deep Reinforcement Learning Based Offloading Scheme for Mobile Edge Computing

Abstract

In the period of rapid development of mobile communication, mobile edge computing is an effective method to meet the increasing computing demands. When local computing capacity is insufficient, we can offload a enormous number of computing resources to the edge server for computing. Nevertheless, the design of computation offloading policies for a virtual MEC system remains challenging. In particular, when decisions are made to offload tasks to the edge server for execution. Specifically, under a time-varying condition, it is decided whether to execute the task locally or offload the task to the edge server for calculation. This article considers data forms with different data sizes, and the data produced by each user has different priority degree of execution. In order to solve the scheduling problem of tasks in the queue, we use Deep Q Network (DQN) to solve the problem. The simulation results are given at the end of the paper and compared with the existing excellent algorithms.