Robust Satellite-Orbit Prediction Using Artificial Neural Network Based on Levenberg-Marquardt Algorithm
Document Type
Conference Proceeding
Publication Date
1-1-2022
Abstract
High-accuracy satellite-orbit prediction is perceived to be very important for future sixth-generation (6G)) communication networks. It is crucial to acquire the precise satellites' instantaneous location information in order to facilitate the future satellite-aided communication networks. Because of the nonlinear characteristics of satellite orbits, we propose a new advanced artificial neural network (ANN) which is built upon the Levenberg-Marquardt algorithm for robust satellite-orbit prediction. Since the Levenberg-Marquardt algorithm (LMA) involves a second-order derivative of the cost function, our proposed novel LMA-based ANN approach can achieve a better performance compared to the conventional first-order derivative methods, including stochastic gradient and conjugate gradient methods. A standard satellite-orbit dataset, namely Two-Line Element (TLE) Catalog, is employed to validate our proposed new LMA-based ANN approach. Numerical results are presented to demonstrate the effectiveness of our proposed novel LMA-based ANN approach for satellite-orbit prediction.
Publication Source (Journal or Book title)
2022 International Wireless Communications and Mobile Computing, IWCMC 2022
First Page
1329
Last Page
1334
Recommended Citation
Chang, S., Wu, H., Sotiropoulos, F., & Goni, U. (2022). Robust Satellite-Orbit Prediction Using Artificial Neural Network Based on Levenberg-Marquardt Algorithm. 2022 International Wireless Communications and Mobile Computing, IWCMC 2022, 1329-1334. https://doi.org/10.1109/IWCMC55113.2022.9824105