基于深度强化学习的星地一体化网络频谱共存技术研究

打开文本图片集

关键词：星地一体化；频谱感知；强化学习

doi：10.3969/J.ISSN.1672-7274.2025.11.002

中图分类号：TN927+.2；TN929.5；TN98 文献标志码：B 文章编码：1672-7274（2025）11-0005-04

Research on Spectrum Coexistence Technology for Satellite-Terrestrial Integrated Networks Based on Deep Reinforcement Learning

WENBoxuan1，YANGYan1，ZHANGKaiqiang²，HAOJiahui²，LIBin² （1. School of Electronic and Information Engineering，Beijing Jiaotong University，Beijing 10oo44， China; 2.Beijing Xuanyu Space Technology Co.，Ltd.，Beijing101399，China）

Abstract： With the sixth-generation mobile communication technology （6G） undergoing a expansion transformation，the integration of terrestrial architectureand satelites hasbecome oneof thecore features of 6G.The scarcity of spectrum resources has emergedas asignificant chalenge hindering the development ofsatelite-terrestrial integrated networks （STINs），making it crucial to promote the research and development of effective strategies for optimizing spectrum utilization.This paper proposes a flexible spectrum sensing technique for the downlink system of low-arthorbit （LEO）satellte-terestrialintegrated networks（STINs）.Confronted with thecomplexscenariosof satelite-terrestrial integration，this paper derives the sensing model for STINsand incorporates deep reinforcement learning （DRL） into spectrum sensing，aiming to explore the available spectrum space and mitigate thefluctuation of falsealarms.Notably，the numericalresults demonstrate that the proposed technique outperforms theenergydetection benchmark scheme， achieving a throughput gain in dynamic communication environments.

Keywords： satelite-terrestrial integration; spectrum sensing; reinforcement learning

研究背景

为实现全球全域的“泛在连接”，信息服务网络将向山区、沙漠、海洋、深地、天空、太空等更广阔的区域推进，空天地一体化作为未来6G网络的一个重要特征已经获得了广泛的共识。（剩余4566字）