摘要
随着物联网(IoT, internet of things)基站的部署愈发密集,网络干扰管控的重要性愈发凸显。物联网中,设备常采用随机接入,以分布式的方式接入信道。在海量设备的物联网场景中,节点之间可能会出现严重的干扰,导致网络的吞吐量性能严重下降。为了解决随机接入网络中的干扰管控问题,考虑基于协作接收的多基站时隙Aloha网络,利用强化学习工具,设计自适应传输算法,实现干扰管控,优化网络的吞吐量性能,并提高网络的公平性。首先,设计了基于Q-学习的自适应传输算法,通过仿真验证了该算法面对不同网络流量时均能保障较高的网络吞吐量性能。其次,为了提高网络的公平性,采用惩罚函数法改进自适应传输算法,并通过仿真验证了面向公平性优化后的算法能够大幅提高网络的公平性,并保障网络的吞吐性能。
With the increasingly dense deployment of base stations in the internet of things(IoT),the importance of inter‐ference management becomes ever more pronounced.In IoT environments,devices often employ random access,connect‐ing to channels in a distributed manner.In scenarios involving massive numbers of devices,severe interference may arise between nodes,leading to significant degradation in the throughput performance of the network.To address interference control issues in networks with random access,a multi-base station slotted Aloha network based on cooperative reception was considered,the reinforcement learning techniques was leveraged to design adaptive transmission algorithms that effectively managed interference,optimized network throughput performance,and enhanced network fairness.Firstly,an adaptive transmission algorithm were devised based on Q-learning,which was verified to maintain high network throughput performance under varying traffic conditions through simulation.Secondly,to improve network fairness,the penalty function method was employed to refine the adaptive transmission algorithm.Simulations confirm that the fairness-optimized algorithm significantly enhances network fairness while preserving satisfactory network throughput performance.
作者
黄元康
詹文
孙兴华
HUANG Yuankang;ZHAN Wen;SUN Xinghua(Sun Yat-sen University Shenzhen Campus,Shenzhen 518107,China;Sun Yat-sen University,Guangzhou 510275,China)
出处
《物联网学报》
2024年第2期26-35,共10页
Chinese Journal on Internet of Things
基金
国家重点研发计划(No.2023YFB2904100)
深圳市科技计划资助项目(No.RCBS20210706092408010)。
关键词
强化学习
物联网
随机接入
多基站网络
时隙ALOHA
reinforcement learning
internet of things
random access
multi-base station network
slotted Aloha
