中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
首页 机构 成果 学者
  
  1. CAS IR Grid
  2. 云南天文台
  3. 中国科学院云南天文台
  4. 应用天文研究组
Perturbation Orbit Prediction Method Based on Physics-Informed ResNet

文献类型：会议论文

作者Zhao, Meng2,3; Shu P(舒鹏)1; Yang, Zhen2,3; Luo, Yazhong2,3
出版日期2025-08-01
会议日期2025-08-02
会议地点Harbin, China
关键词Orbit Prediction Two Body Orbital Dynamics J2 Perturbation Physics Informed Neural Networks (PINNs) Residual Neural Network (ResNet) Automatic Differentiation
卷号59
期号20
DOI10.1016/j.ifacol.2025.11.304
页码1111-1116
英文摘要In order to address the trade-off between prediction accuracy and computational efficiency in traditional orbit prediction methods, this paper proposes a physics-informed residual neural network for fast and accurate orbit prediction under J2 perturbation. Multiple residual blocks are employed in series to alleviate the gradient vanishing problem in deep neural networks. A physics-informed loss term is constructed by combining the differences between the J2 perturbation equations and the Keplerian equations with the automatic differentiation results of the neural network. This term is integrated with the prediction loss to jointly form the training objective. Simulation results show that the physics-informed model generally achieves a lower prediction loss compared to its non-physics-informed counterpart. Moreover, the method offers significant improvements in computational speed over traditional numerical integration. Sensitivity analysis further verifies the generalization capability of the trained model, with no notable bias or failure observed in the prediction results. Copyright © 2025 The Authors.
产权排序第3完成单位
资助机构N/A
会议录IFAC-PapersOnLine
文献子类Conference article (CA) -
学科主题天文学 ; 天体测量学
语种英语
URL标识查看原文
资助项目N/A
ISSN号2405-8971
源URL[http://ir.ynao.ac.cn/handle/114a53/28816]  
专题云南天文台_应用天文研究组
作者单位1.Yunnan Observatories, Chinese Academy of Sciences, Yunnan, Kunming, 650216, China
2.College of Aerospace Science and Engineering, National University of Defense Technology, Hunan, Changsha, 410073, China;
3.State Key Laboratory of Space System Operation and Control, Hunan, Changsha, 410073, China;
推荐引用方式
GB/T 7714		 Zhao, Meng,Shu P,Yang, Zhen,et al. Perturbation Orbit Prediction Method Based on Physics-Informed ResNet[C]. 见:. Harbin, China. 2025-08-02.

入库方式： OAI收割

来源：云南天文台

浏览0
下载0
收藏0
其他版本

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。