基于深度强化学习的移动边缘计算卸载研究

Research on Mobile Edge Computing Unloading Based on Deep Reinforcement Learning

由于边缘服务器的带宽资源和计算能力相对较差,基于此提出了端边缘云三层网络架构,并设计了以系统能耗最小为目标的任务卸载和资源分配综合优化模型,该模型明确定义了马尔可夫随机过程的状态空间、动作空间和奖惩函数。采用了双深度Q网络(Double DQN)和对偶深度Q网络(Dueling DQN)算法进行仿真实验,结果表明:所提出的算法比多种基线算法和原始DQN算法的效果要好。

Due to the relatively poor bandwidth resources and computing capacity of edge servers,a three-layer network architecture of end-edge-cloud is proposed.A comprehensive optimization model for task unloading and resource allocation is designed to minimize the energy consumption of the system.This model clearly defines the state space,action space and reward and punishment function of Markov stochastic process.Simulation experiments are conducted using double deep Q-network(Double DQN)and dueling deep Q-network(Dueling DQN)algorithms.The experimental results show that the proposed algorithm performs better than various baseline algorithms and the original DQN algorithm.