船舶通信网络在复杂海洋环境中面临多径效应、多普勒频移与信道时变等多重挑战，其关键节点的运行状态直接影响网络的稳定性。鉴于船舶航行环境的高度动态性和拓扑结构的时变性，难以在拓扑频繁变化与数据部分缺失的复杂场景下实现精准估计。为此，本文提出一种动态图神经网络的关键节点状态估计方法。通过构建边权重可变的动态图模型，量化链路质量并刻画拓扑演化；设计融合节点自身多源监测特征与邻域交互信息的图卷积输入层，结合注意力机制强化关键节点特征表达，并引入长短期记忆网络以建模节点状态的时序演化规律，构建端到端的状态估计模型。实验结果表明，所提方法在多种数据完整性与拓扑动态性场景下，均显著优于传统方法与现有动态图神经网络，具备更高的估计精度、更强的鲁棒性与更低的估计延迟，适用于高动态船舶通信网络中的实时状态感知与决策支持。
In view of the highly dynamic navigation environment of ships and the time-varying topological structure, it is difficult to achieve accurate estimation in complex scenes with frequent topology changes and missing data. Therefore, this paper proposes a state estimation method for key nodes of dynamic graph neural network. By constructing a dynamic graph model with variable edge weights, the link quality is quantified and the topological evolution is described. The graph product input layer is designed to integrate the multi-source monitoring features of nodes and the neighborhood interactive information, and the expression of key node features is strengthened by combining attention mechanism. The long-term and short-term memory networks are introduced to model the time series evolution law of node States, and an end-to-end state estimation model is constructed. The experimental results show that the proposed method is significantly superior to the traditional method and the existing dynamic graph neural network in a variety of data integrity and topological dynamic scenarios, with higher estimation accuracy, stronger robustness and lower estimation delay, and is suitable for real-time state perception and decision support in highly dynamic ship communication networks.
2025,47(22): 171-174 收稿日期：2025-5-29
DOI：10.3404/j.issn.1672-7649.2025.22.025
分类号：U665.26；TP391
作者简介：王峰（1981 – ），男，博士，副教授，研究方向为人工智能技术
参考文献：
[1] 徐新林, 邓异. 舰基无线通信组网节点优化部署模型仿真[J]. 舰船电子工程, 2024, 44(2): 91-95
XU X L, DENG Y. Simulation of optimal deployment model of ship-based wireless communication networking nodes[J]. Ship Electronic Engineering, 2024, 44(2): 91-95
[2] 许镭. 船舶状态监控网络的节点信任度模型研究[J]. 舰船科学技术, 2021, 43(10): 175-177.
XU L. Research on node trust model of ship state monitoring network[J]. Ship Science and Technology, 2021, 43(10): 175-177.
[3] 刘丽萍, 操刘生, 陈梦. 基于AR模型和卡尔曼滤波的UWSNs节点分层预测定位[J]. 天津大学学报(自然科学与工程技术版), 2019, 52(1): 98-104
LIU L P, CAO L S, CHEN M. Hierarchical prediction localization method based on AR model and kalman for underwater wireless sensor networks[J]. Journal of Tianjin University(Science and Technology), 2019, 52(1): 98-104
[4] 张陶, 于炯, 廖彬, 等. 基于图嵌入与支持向量机的社交网络节点分类方法[J]. 计算机应用研究, 2021, 38(9): 2646-2650+2661.
ZHANG T, YU J, LIAO B, et al. Node classification method in social network based on graph embedding and support vector machine[J]. Application Research of Computers, 2021, 38(9): 2646-2650+2661.
[5] 柏园超, 刘文昌, 江俊君, 等. 深度神经网络图像压缩方法进展综述[J]. 电子与信息学报, 2025, 47(12): 1-17
BAI Y C, LIU W C, JIANG JJ, et al. Advances in deep neural network based image compression: a survey[J]. Journal of Electronics & Information Technology, 2025, 47(12): 1-17
[6] 高智, 胡傲涵, 陈泊安, 等. 多层级几何—语义融合的图神经网络地表异常检测框架[J]. 遥感学报, 2024, 28(7): 1760-1770
GAO Z, HU A H, CHEN B A, et al. A hierarchical geometry-to-semantic fusion GNN framework for earth surface anomalies detection[J]. National Remote Sensing Bulletin, 2024, 28(7): 1760-1770
