摘要
本文提出了一种基站休眠控制框架,首先使用一种基于时空图神经网络的移动流量预测技术,利用历史数据对基站未来一段时间的负载情况进行预测。然后设计一种基于深度强化学习的基站休眠控制算法,该算法综合考虑多种实际约束,基于预测结果优化资源分配,在提高网络能效的同时保证稳定的用户体验。真实数据集上的广泛实验证实了该框架的优越性。
This paper proposes a base station sleep control framework.Firstly,a mobile traffic prediction technology based on neural network of spatio-temporal graph was used to predict the loads of base stations in future time by using historical data.Then,a base station sleep control algorithm based on deep reinforcement learning was proposed,which took a variety of practical constraints into consideration and optimized resource allocation based on the prediction results,so as to improve the network energy efficiency and ensure stable user experience.Extensive experiments with a real-world dataset prove the advantage of this framework.
作者
杨馥瑜
赵东
YANG Fuyu;ZHAO Dong(School of Computer Science,Beijing University of Posts and Telecommunications,Beijing 100876,China)
出处
《中国科技论文在线精品论文》
2023年第2期170-178,共9页
Highlights of Sciencepaper Online
关键词
人工智能
时空预测
基站休眠
深度强化学习
artificial intelligence
spatio-temporal prediction
base station sleep
deep reinforcement learning
作者简介
杨馥瑜(1998—),女,硕士研究生,主要研究方向:时空预测;通信联系人:赵东,教授,主要研究方向:物联网、移动群智感知.E-mail:dzhao@bupt.edu.cn
