摘 要：5G 时代下数据井喷带来了网络拥堵，本地算法以及云计算的集中式处理模式不足以满足大规模物联网环境的实时性要求。边缘计算模型中大数据需要通过信道卸载到边缘服务器上，通过对传统信道选择方式的研究可知：传统基站分配方式效率低下。通过 epsilon-Greedy 算法和随机算法的比较可得出：合理设定 epsilon 值，使探索与利用相结合可实现设备自我学习选择信道。
关键词：边缘计算；自我学习；信道选择
中图分类号：TN919.2 文献标识码：A 文章编号：2096-4706（2020）06-0079-03
Channel Selection Scheme Based on epsilon-Greedy Algorithm
ZHANG Sunxuan
（North China Electric Power University，School of Electrical and Electronic Engineering，Beijing 102206，China）
Abstract：Data blowout in 5G era brings network congestion，and the centralized processing mode of local algorithm and cloud computing is not enough to meet the real-time requirements of large-scale internet of things environment. The big data in the edge computing model need to be offloaded to the edge server through the channel. By comparing the epsilon-Greedy algorithm with the stochastic algorithm，it can be concluded that the combination of exploration and utilization by setting the epsilon value reasonably can realize the self-learning channel selection of equipment.
Keywords：edge computing；self learning；channel selection
参考文献：
[1] LUAN T H，GAO L X，LI Z，et al.FOG COMPUTING：FOCUSING ON MOBILE USERS AT THE EDGE [J].COMPUTERSCIENCE，2015：1-11.
[2] KUMAR K，LIU J B，LU Y H，et al.A Survey of ComputationOffloading for Mobile Systems [J].Mobile Networks and Applications，2013，18（1）：129-140.
[3] RAHIMI M R，REN J，LIU C H，et al.Mobile CloudComputing： A Survey，State of Art and Future Directions [J].MobileNetworks and Applications，2014，19（2）：133-143.
[4] KUMAR K，LU Y H.Cloud Computing for Mobile Users：Can Offloading Computation Save Energy? [J].IEEE Computer，2010（43）：51-56.
作者简介：张孙烜（1998.09-），男，汉族，福建宁德人，本科，研究方向：电力物联网、工业物联网、机器对机器通信、无限资源分配。