用于计算和传输的动态星间路由策略  

Dynamic Inter-satellite Routing Strategy forComputation and Transmission

作　　者：许柳飞 罗志勇 XU Liufei;LUO Zhiyong(School of Electronics and Communication Engineering,Sun Yat-sen University Shenzhen Campus,Shenzhen 518107,China)

机构地区：[1]中山大学深圳校区电子与通信工程学院,广东深圳518107

出　　处：《无线电通信技术》2024年第6期1153-1159,共7页Radio Communications Technology

基　　金：国家重点研发计划(2023YFB2904701);广东省重点研发计划(2024B0101020006);广东省区域联合基金重点项目(2023B1515120093);深圳市自然科学基金重点项目(JCYJ20220818102209020)。

摘　　要：针对低地球轨道(Low Earth Orbit,LEO)卫星网络具有拓扑变化快、网络节点多和节点资源状态变化等特点,提出了一种用于计算和传输的星间路由策略。该策略使用改进的图卷积网络(Enhanced-Graph Convolutional Network,EGCN)提取卫星网络的时空特征并生成节点的隐藏状态。将其作为深度强化学习(Deep Reinforcement Learning,DRL)模型的输入,DRL智能体感知下一跳节点的关键信息,从而更好地做出决策。仿真结果表明,与以前的方法相比,提出的方法提高了网络的总吞吐量,降低了端到端传输延迟。Low Earth Orbit(LEO)satellite networks have the characteristics of rapid topology changes,varying network nodes and fluctuations in node resource availability.An inter-satellite routing strategy for computation and transmission is proposed,which uses the Enhanced-Graph Convolutional Network(EGCN)to extract the spatiotemporal features from the satellite network and generates the hidden states for each network node.As input to the Deep Reinforcement Learning(DRL)model,the DRL agent senses the key information of the next-hop node to make better decisions.Simulation results show that,compared to previous methods,the proposed method not only improves the overall throughput of the network,but also reduces the end-to-end transmission delay.

关 键 词：星间路由策略 动态卫星网络 深度强化学习 图卷积神经网络 

分 类 号：TN919.23[电子电信—通信与信息系统]