adityagandhamal / satellite-position-estimation

Predicting and forecasting the position/trajectory of a satellite orbiting Earth

About

• The increase in the number of RSOs (Resident Space Objects) indirectly increases the risk of collision of LEO Satellites.

• Here's a snippet depicting the scale of the problem (Credits to LeoLabs):

leolabs-google-chrome-2022-01-10-13-54-53_kLfdO0PD_Trim.mp4

• The important point to be addressed here is the reliable and accurate orbit tracking of satellites to prevent a catastrophic event like the Kessler Syndrome.

• This project is an experiment on predicting and forecasting the position of a satellite orbiting earth using Deep Learning (LSTM).

• The LSTM model is trained on the data recorded over 18days and forecasts the trajectory for the next seven days.

The Satellite Trajectory

Figure.1.2022-01-14.14-16-10_Trim.mp4

Forecasted Trajectory

Figure.1.2022-01-14.15-43-23_Trim.mp4

Libraries Used:

• Tensorflow 2.x
• Keras
• Scikit Learn
• Python >= 3.7
• Numpy
• Statsmodels
• Matplotlib
• Pandas
• Pmdarima

Data

• Here's the link to the dataset https://www.kaggle.com/idawoodjee/predict-the-positions-and-speeds-of-600-satellites

• The dataset from Kaggle contains information of 600 satellites.

This experiment is performed using data of a single satellite.