通信学报2025，Vol.46Issue(11)：73-86,14.DOI:10.11959/j.issn.1000-436x.2025220
基于ShapeNet的非合作无人机自组织网络通联拓扑推理技术
Communication topology inference technology for non-cooperative UAV self-organizing networks based on ShapeNet
摘要
Abstract
To solve the problem of topology inference in unmanned aerial vehicle(UAV)self-organizing communication networks under non-cooperative scenarios,a topology inference method based on the interpretable neural network named ShapeNet was proposed.Firstly,a topology inference system model of the non-cooperative UAV self-organizing network was established to describe the topology inference mechanism.Then,a time-series classification algorithm for communi-cation states based on the shape features(Shapelets)of subsequences in time-series was developed,and the network to-pology structure was reconstructed according to the classification results.Finally,the ShapeNet model was developed to further enhance the efficiency of the topology inference process.The experimental results show that the ShapeNet model can distinguish the"pseudo-causality relationships"among communication-state time-series and leverage the parallel processing mechanism of the GPU to accelerate topology inference.Compared with baseline methods,the proposed ap-proach achieves the highest inference accuracy and the least inference time.关键词
自组织通信网络/ShapeNet/拓扑推理/形状特征/ShapeletsKey words
self-organizing communication network/ShapeNet/topology inference/shape feature/Shapelets分类
信息技术与安全科学引用本文复制引用
侯长波,艾琦迅,周志超,钮振宇,宋振..基于ShapeNet的非合作无人机自组织网络通联拓扑推理技术[J].通信学报,2025,46(11):73-86,14.基金项目
国家自然科学基金资助项目(No.U23A20271) （No.U23A20271）
中央高校基本科研业务费专项资金资助项目(No.3072025ZN0801)The National Natural Science Foundation of China(No.U23A20271),The Fundamental Research Funds for the Central Universities(No.3072025ZN0801) （No.3072025ZN0801）
