Research on LSTM-based spatial target trajectory forecasting enhanced by attention mechanisms
Qingshan Luo, Jiahao Ji, Tao Yang, Yurui Xu, Yunsheng Yao · 2026
Read Research on LSTM-based spatial target trajectory forecasting enhanced by attention mechanisms by Qingshan Luo, Jiahao Ji, Tao Yang, Yurui Xu, Yunsheng Yao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0356376.txt.
Description
Qingshan Luo and four colleagues improve satellite orbit prediction with deep learning in this 2026 paper. Standard orbit models depend heavily on physics and starting conditions, and their errors grow over time, which matters for tracking space objects and warning about collisions. The team trained a memory-based neural network, enhanced with an attention mechanism, to learn the errors a standard model makes.
Working with a well-tracked geodetic satellite, the network used past position errors plus velocity and acceleration data to predict the next day's error, letting researchers correct the standard forecast. The attention-enhanced model beat several rival approaches, shrinking leftover errors to only a few percent of the original in each direction. It also held up on satellites at other altitudes, though longer forecasts grew harder. The authors see it as best suited to short-term corrections.
Catalog record
| CREATOR | Qingshan Luo, Jiahao Ji, Tao Yang, Yurui Xu, Yunsheng Yao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44518 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |