海洋机器人是海洋环境观测应用的重要工具。利用海洋机器人获取的稀疏观测数据，实现区域海洋环境要素空间场高分辨率预测具有重要应用需求。面向该需求，提出一种基于海洋机器人稀疏观测数据的高分辨率海洋环境预测方法。以卷积长短期记忆神经网络ConvLSTM为基础，建立数据驱动的海洋环境空间场时间序列预测模型。应用集合卡尔曼滤波方法同化海洋机器人观测数据，实现神经网络预测模型在线学习，使模型逐步逼近海洋环境状态、预测输出逐步逼近观测数据。针对大尺度海洋环境高空间分辨率预测带来的高计算复杂性问题，引入非均匀空间分辨率预测策略，有效应对计算挑战。利用水下滑翔机南海北部观测实验数据进行了区域海洋环境二维水平面温度场预测实验，结果验证了方法的有效性。
Marine vehicles serve as pivotal platforms for oceanic environmental observations. Given the sparse nature of the observation data collected by marine vehicle networks, a significant demand arises for generating high-resolution predictions regarding spatial distributions of regional oceanic parameters. Addressing this demand, this paper proposes a novel methodology that utilizes sparse marine vehicle observations for high-resolution ocean environment prediction. Built upon the ConvLSTM neural network model, this approach establishes a data-driven model capable of predicting the spatiotemporal evolution of oceanic fields. Based on the neural network model, the ensemble Kalman filter technique is employed to effectively assimilate marine vehicle observations, enabling online learning of the neural network model that progressively aligns model and its predictions with evolving oceanic conditions and observed data. To address the computational challenges posed by high-resolution predictions over large oceanic scales, a non-uniform spatial resolution prediction strategy is introduced, effectively reducing computational burden. Experimental validation employs observations from an underwater glider network deployed in the northern South China Sea, predicting two-dimensional horizontal temperature spatial field within regional ocean environment. The results validate the effectiveness of the proposed method.
2024,46(14): 73-80 收稿日期：2023-09-19
DOI：10.3404/j.issn.1672-7649.2024.14.013
分类号：TP24
基金项目：国家自然科学基金资助项目（52271353）；辽宁省兴辽英才计划资助项目（XLYC2007035）；沈阳市自然科学基金资助项目（22-315-6-08）；机器人学国家重点实验室项目（2022-Z11）；中国科学院国际伙伴计划资助项目（173321KYSB20180011）；中国科学院沈阳自动化研究所基础研究计划资助项目（2022JC3K05）；国家重点研发计划资助项目（2022YFE0204600）
作者简介：金乾隆(1994-),男,博士研究生,研究方向为海洋机器人。
参考文献：
[1] CHEN Z, YU J, ZHANG A. Overview on observation-oriented unmanned marine vehicles with high cruising ability: Development status and prospect[J]. Journal of Ocean Technology, 2016, 35(1): 122-130.
[2] L'HÉVÉDER B, MORTIER L, TESTOR P, et al. A glider network design study for a synoptic view of the oceanic mesoscale variability[J]. Journal of Atmospheric and Oceanic Technology, 2013, 30(7): 1472-1493.
[3] GREGÓRIO R P A. Geostatistical prediction of ocean outfall plume characteristics based on an autonomous underwater vehicle[J]. International Journal of Advanced Robotic Systems, 2013, 10(7): 289.
[4] FOSSUM T O, EIDSVIK J, ELLINGSEN I, et al. Information-driven robotic sampling in the coastal ocean[J]. Journal of Field Robotics, 2018, 35(7): 1101-1121.
[5] MUNAFÒ A, SIMETTI E, TURETTA A, et al. Autonomous underwater vehicle teams for adaptive ocean sampling: A data-driven approach[J]. Ocean Dynamics, 2011, 61(11): 1981-1994.
[6] GRASSO R, BRACA P, FORTUNATI S, et al. Dynamic underwater glider network for environmental field estimation[J]. IEEE Transactions on Aerospace and Electronic Systems, 2016, 52(1): 379-395.
[7] HEUSS J P, HALEY P J, MIRABITO C, et al. Reduced order modeling for stochastic prediction onboard autonomous platforms at sea[C]//Proceedings of Global Oceans 2020: Singapore – U. S. Gulf Coast. Biloxi, MS, USA, 2020: 1–10.
[8] SHI X, CHEN Z, WANG H, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting[C]//Advances in Neural Information Processing Systems 28 (NIPS 2015). Montreal, Canada, 2015: 802–810.
[9] YANG Y, DONG J, SUN X, et al. A CFCC-LSTM model for sea surface temperature prediction[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15(2): 207-211.
[10] EVENSEN G. Data assimilation: the ensemble Kalman filter[M]. Berlin, Heidelberg, Germany: Springer-Verlag, 2009: 1–307.
[11] PERSSON P-O. Mesh generation for implicit geometries[D]. Massachusetts: Department of Mathematics, Massachusetts Institute of Technology, 2005.
[12] BEZDEK J C. A convergence theorem for the fuzzy ISODATA clustering algorithms[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1980, 2(1): 1-8.
