基于DQN的低轨卫星网络多目标智能路由算法  

Multi-Objective Intelligent Routing Algorithm for LEO Satellite Networks Based on DQN

作　　者：罗宗屹 金世超[1,2] 董涛 殷杰[1,2] LUO Zongyi;JIN Shichao;DONG Tao;YIN Jie(State Key Laboratory of Space-Ground Integrated Information Technology,Space Star Technology Co.,Ltd.,Beijing 100095,China;Beijing Institute of Satellite Information Engineering,Beijing 100095,China)

机构地区：[1]航天恒星科技有限公司天地一体化信息技术国家重点实验室,北京100095 [2]北京卫星信息工程研究所,北京100095

出　　处：《天地一体化信息网络》2025年第1期16-23,共8页Space-Integrated-Ground Information Networks

基　　金：国家重点研发计划资助项目(No.2022YFB2902504)。

摘　　要：近年来,低轨卫星网络发展迅速,被广泛应用于全球通信、互联网接入等领域。首先,对当前低轨卫星网络发展现状与主流的低轨网络路由算法进行总结。然后,针对星上路由算法设计面临的优化目标单一、收敛速度慢以及时延和丢包率等指标优化难度大等挑战,提出基于深度Q网络(DQN)的多目标智能路由算法,通过设计多目标奖励函数,实现了时延、丢包率、负载均衡多目标的性能优化。在时延方面,本算法优于最短通路优先算法;在流量分布指数和丢包率方面,本算法相较于最短通路优先算法与Dijkstra算法分别提高14%和10%以上。最后,阐述所提智能路由算法在多种应用场景中的使用方法。In recent years,LEO satellite networks have developed rapidly and are widely used in global communication and internet access.This paper first summarized the current development status of LEO networks and the main LEO network routing algorithms.At the same time,aimed at the challenges of single optimization objectives,slow convergence speed,and difficulty in optimizing indicators such as delay and packet loss,this paper proposed a multi-objective intelligent routing algorithm based on deep Q-network(DQN).With the designed DQN multi-objective reward function,it realized the optimization of delay,packet loss,load balancing.Regarding delay,the proposed algorithm is superior to the Shortest Path First(SPF)algorithm.In terms of traffic distribution index and packet loss rate,the performances of the proposed algorithm are improved by more than 14%and 10%compared with the SPF and Dijkstra algorithms respectively.In addition,this paper illustrated the application methods of the proposed intelligent routing algorithm in various scenarios.

关 键 词：低轨卫星网络 路由算法 深度Q网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]