Expandable Orbit Decay Prediction Using Continual Learning
Very low earth orbit (VLEO) spacecraft have become an attractive proposition with obvious advantages in various missions, including communication and ground observation. Higher requirements for precise orbit decay prediction (PODP) technology are requested, which provides accurate state references f...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2024-01-01
|
| Series: | International Journal of Aerospace Engineering |
| Online Access: | http://dx.doi.org/10.1155/2024/8887634 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832544793123618816 |
|---|---|
| author | Junhua He Hua Wang Haitao Wang Xuankun Fang Chengyi Huo |
| author_facet | Junhua He Hua Wang Haitao Wang Xuankun Fang Chengyi Huo |
| author_sort | Junhua He |
| collection | DOAJ |
| description | Very low earth orbit (VLEO) spacecraft have become an attractive proposition with obvious advantages in various missions, including communication and ground observation. Higher requirements for precise orbit decay prediction (PODP) technology are requested, which provides accurate state references for orbit maintenance and necessary database for space situational awareness. The effectiveness of the traditional orbital prediction method for PODP is limited by inaccurate estimation of the spacecraft’s physical parameters and space environments. Generalization performance of machine learning techniques (MLTs) is blocked by the universal challenge known as catastrophic forgetting, resulting in limited improvement on PODP. In this study, a method of expandable orbit decay propagator (EODP) for spacecraft PODP in VLEO, based on model-agnostic MLTs, is proposed. The plasticity of the proposed model is discussed, which originates from the uncertainty of neural network (NN) parameters. The proposed method overcomes the negative effects of uncertain physical parameters and complex space environments. The model achieves at least a 70% improvement in accuracy compared to the high-precision orbital propagator (HPOP) and presents a novel approach for the future implementation of machine learning–based methods in the field of orbit prediction. |
| format | Article |
| id | doaj-art-f4371debbb2b4b2180b0733e41afd959 |
| institution | Kabale University |
| issn | 1687-5974 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | International Journal of Aerospace Engineering |
| spelling | doaj-art-f4371debbb2b4b2180b0733e41afd9592025-02-03T09:56:52ZengWileyInternational Journal of Aerospace Engineering1687-59742024-01-01202410.1155/2024/8887634Expandable Orbit Decay Prediction Using Continual LearningJunhua He0Hua Wang1Haitao Wang2Xuankun Fang3Chengyi Huo4College of Aerospace Science and EngineeringCollege of Aerospace Science and EngineeringCollege of Aerospace Science and EngineeringCollege of Aerospace Science and EngineeringCollege of Aerospace Science and EngineeringVery low earth orbit (VLEO) spacecraft have become an attractive proposition with obvious advantages in various missions, including communication and ground observation. Higher requirements for precise orbit decay prediction (PODP) technology are requested, which provides accurate state references for orbit maintenance and necessary database for space situational awareness. The effectiveness of the traditional orbital prediction method for PODP is limited by inaccurate estimation of the spacecraft’s physical parameters and space environments. Generalization performance of machine learning techniques (MLTs) is blocked by the universal challenge known as catastrophic forgetting, resulting in limited improvement on PODP. In this study, a method of expandable orbit decay propagator (EODP) for spacecraft PODP in VLEO, based on model-agnostic MLTs, is proposed. The plasticity of the proposed model is discussed, which originates from the uncertainty of neural network (NN) parameters. The proposed method overcomes the negative effects of uncertain physical parameters and complex space environments. The model achieves at least a 70% improvement in accuracy compared to the high-precision orbital propagator (HPOP) and presents a novel approach for the future implementation of machine learning–based methods in the field of orbit prediction.http://dx.doi.org/10.1155/2024/8887634 |
| spellingShingle | Junhua He Hua Wang Haitao Wang Xuankun Fang Chengyi Huo Expandable Orbit Decay Prediction Using Continual Learning International Journal of Aerospace Engineering |
| title | Expandable Orbit Decay Prediction Using Continual Learning |
| title_full | Expandable Orbit Decay Prediction Using Continual Learning |
| title_fullStr | Expandable Orbit Decay Prediction Using Continual Learning |
| title_full_unstemmed | Expandable Orbit Decay Prediction Using Continual Learning |
| title_short | Expandable Orbit Decay Prediction Using Continual Learning |
| title_sort | expandable orbit decay prediction using continual learning |
| url | http://dx.doi.org/10.1155/2024/8887634 |
| work_keys_str_mv | AT junhuahe expandableorbitdecaypredictionusingcontinuallearning AT huawang expandableorbitdecaypredictionusingcontinuallearning AT haitaowang expandableorbitdecaypredictionusingcontinuallearning AT xuankunfang expandableorbitdecaypredictionusingcontinuallearning AT chengyihuo expandableorbitdecaypredictionusingcontinuallearning |