基于深度强化学习的移动边缘卸载机制

Mobile Edge Unloading Mechanism Based on Deep Reinforcement Learning

随着未来5G移动网络的发展,为了满足用户在大量数据剧增的背景下实现更高带宽、更低时延的要求,移动边缘计算(MEC)技术正逐渐引起相关学者的重视。MEC可以将计算密集型任务迁移到MEC服务器,来扩展无线网络边缘的计算能力。论文考虑了一个多用户的MEC系统,提出了深度强化学习方法来优化即时奖励和长期成本。仿真结果表明,与其他方法相比,该方案在降低总成本上有明显优势。

With the development of 5G mobile networks in the future,in order to meet the requirements of users to achieve higher bandwidth and lower latency in the context of massive data explosion,mobile edge computing(MEC)technology is gradually attracting the attention of scholars.MEC can migrate computationally intensive tasks to the Oracle MEC server to extend the comput ing power at the edge of the wireless network.The paper considers a user's MEC system and proposes a deep reinforcement learning approach to optimize instant rewards and long-term costs.The simulation results show that compared with other methods,this scheme has obvious advantages in reducing costs.