水下无人潜航器（Unmannd Underwater Vehicle，UUV）作为水下无人系统的关键装备，在海洋资源开发中发挥着重要作用。受到水下低通量弱通信复杂环境影响，UUV数据采集常态化缺失现象严重，特别是浅海环境中传统UUV跟踪模型未充分结合声损失特性导致UUV航行轨迹连续性差，严重制约后续UUV态势感知功能的准确性。本文提出一种融合径向基函数（Radial Basis Function，RBF）神经网络与水下射线传播理论的水下声速反演驱动UUV轨迹优化模型。首先利用RBF神经网络逼近能力，构建声传播时间（Time of Flight，ToF）与声速剖面（Sound Speed Profile, SSP）特征的时序变化关系，精准刻画水下声速场的动态特性。然后设计时序融合聚类最小残差定位算法通过梯度索引对历史数据精准分类，并将每类不同特性的轨迹数据进行粒子群连续化处理精准补全UUV电子轨迹密度。实验表明，与传统RBF模型结合最小残差定位算法相比，本文创新提出融合声速剖面UUV跟踪模型在其轨迹优化方面精确度有效提高5.41%，验证了所提方法在复杂水下环境中的有效性和鲁棒性。
Unmanned Underwater Vehicle (UUV) play a crucial role in marine resource exploration. However, low-throughput and weak communication environments often lead to severe data loss, especially in shallow waters, where traditional UUV tracking models fail to account for acoustic loss, resulting in discontinuous trajectories and reduced situational awareness accuracy.This paper proposes a UUV trajectory optimization model driven by underwater sound speed inversion, integrating a Radial Basis Function (RBF) neural network with underwater ray propagation theory. First, the RBF neural network models the temporal relationship between sound propagation time (ToF) and Sound Speed Profile (SSP) features to capture underwater sound speed dynamics. Then, a temporal fusion clustering-based minimum residual positioning algorithm classifies historical trajectory data, and a particle swarm-based method refines UUV trajectory density.Experimental results show that, compared to conventional RBF models, the proposed SSP-enhanced UUV tracking model improves trajectory optimization accuracy by 5.41%, demonstrating its effectiveness and robustness in complex underwater environments.
2026,48(2): 122-129 收稿日期：2025-5-10
DOI：10.3404/j.issn.1672-7649.2026.02.020
分类号：U666.1；TP242
基金项目：国家自然科学基金面上项目（62076096）；四川省重点研发计划面上项目（2023YFG0149）
作者简介：蔡鑫源(2001-),男,硕士研究生,研究方向为水下目标跟踪
参考文献：
[1] 宋跃才, 林海涛, 卞媛, 等. 基于移动信标的水下无线传感器网络定位算法[J]. 电子测量技术, 2023, 46(5): 44-49 .
SONG Y C, LIN H T, BIAN Y, et al. A localization algorithm for underwater wireless sensor networks based on mobile beacons[J]. Electronic Measurement Technology, 2023, 46(5): 44-49 .
[2] 田青, 赵宇, 张正, 等. 基于Enhanced Zero-DCE++的水下选通图像增强技术研究[J]. 电子测量技术, 2025, 48(1): 166-174 .
TIAN Q, ZHAO Y, ZHANG Z, et al. Enhanced Zero-DCE++-based underwater gated image enhancement[J]. Electronic Measurement Technology, 2025, 48(1): 166-174 .
[3] 胡武清, 贾云飞, 杨玉全, 等. 单平台水下定位系统设计与实现[J]. 电子测量技术, 2024, 47(14): 178-185 .
HU W Q, JIA Y F, YANG Y Q, et al. Designing and implementing a single-platform underwater positioning system[J]. Electronic Measurement Technology, 2024, 47(14): 178-185 .
[4] 谢军, 刘庆军, 边朗. PNT体系视角下卫星导航与不依赖卫星导航技术融合发展研究[J]. 中国工程科学, 2023, 25(2): 64-71.
XIE J, LIU Q J, BIAN L. Integrated development of satellite-based and satellite-independent PNT technologies: a perspective from the PNT architecture[J]. Strategic Study of CAE, 2023, 25(2): 64-71.
[5] LU J, ZHANG H, WU P, et al. Predictive modeling of future full-ocean depth SSPs utilizing hierarchical long short-term memory neural networks[J]. Journal of Marine Science and Engineering, 2024, 12(6): 943.
[6] LIU Y, CHEN Y, CHEN W, et al. Research on Large Depth Extension Method of Global Underwater Sound Speed Profile[C]//2024 OES China Ocean Acoustics (COA). IEEE, 2024: 1–6.
[7] 韩先平, 陈祥明. 折射率湿项对靶场经纬仪测高精度的影响分析[J]. 应用光学, 2024, 45 (4): 790–795.
HAN X P, CHEN X M. analysis of influence of refactivity wet term on height measurement accuracy of cinetheodlite[J]. Journal of Applied Optics, 2024, 45 (4): 790–795.
[8] BUDD M, DUCKWORTH P, HAWES N, et al. Bayesian reinforcement learning for single-episode missions in partially unknown environments[C]//Conference on Robot Learning. PMLR, 2023: 1189–1198.
[9] PELEKANAKIS K, KONDI A, RUSSO A, et al. Simulation of Climate Change Impact on Underwater Acoustic Transmission Loss[C]//OCEANS 2024-Singapore. IEEE, 2024.
[10] WU P, ZHANG H, SHI Y, et al. Real-time estimation of underwater sound speed profiles with a data fusion convolutional neural network model[J]. Applied Ocean Research, 2024, 150: 104088.
[11] LIU H, JIANG Q, QIN Y, et al. Double layer weighted unscented Kalman underwater target tracking algorithm based on sound speed profile[J]. Ocean Engineering, 2022, 266: 112982.
[12] JIANG L, LI Y, WANG B, et al. Cooperative Localization of Asynchronous AUVs with Compensation for the Acoustic Wave Bends[J]. IEEE Signal Processing Letters, 2024.
[13] YU X, XU T, WANG J. Sound velocity profile prediction method based on rbf neural network[C]//China Satellite Navigation Conference (CSNC) 2020 Proceedings: Volume III. Springer Singapore, 2020: 475–487.
[14] HUANG W, ZHOU J, GAO F, et al. Underwater sound speed profile construction: a review[J]. Arxiv Preprint Arxiv: 2310.08251, 2023.
[15] LI D, DU L. Auv trajectory tracking models and control strategies: A review[J]. Journal of Marine Science and Engineering, 2021, 9(9): 1020.
[16] ZHANG Y, HU H, FU W, et al. Particle swarm optim ization–based minimum residual algorithm for mobile robot localization in indoor environment[J]. International Journal of Advanced Robotic Systems, 2017, 14(5): 17298814-17729277.
