Skip to content

HOPE is a hybrid orbit prediction engine that simulates and forecasts the motion of Low Earth Orbit (LEO) satellites under perturbative forces. The initial version focuses on the J2 zonal harmonic and combines traditional physics-based propagation with machine learning models such as LSTMs. The goal is to evaluate the effectiveness of ML in orbital

Notifications You must be signed in to change notification settings

RajrupaDas/HOPE

Repository files navigation

HOPE: Hybrid Orbit Prediction Engine (with plans for Explainability)

Project Type: Hybrid orbit prediction system combining classical astrodynamics and machine learning
Author: Rajrupa Das
Status: Simulation and dataset generation complete; machine learning model development in progress


Overview

HOPE is a hybrid orbit prediction engine that simulates and forecasts the motion of Low Earth Orbit (LEO) satellites under perturbative forces. The initial version focuses on the J2 zonal harmonic and combines traditional physics-based propagation with machine learning models such as LSTMs. The goal is to evaluate the effectiveness of ML in orbital prediction while maintaining physical consistency and enabling future explainability.


Motivation

Classical numerical propagators (e.g., RK4) provide accurate orbital predictions but are often computationally intensive, especially for long-duration or multi-object simulations. In contrast, machine learning offers a faster alternative but lacks inherent physical interpretability. This project explores a middle ground by using classical propagation to generate training data for ML models, aiming to create a hybrid system that is both efficient and physically grounded.


Features and Progress

Step Description Status
Step 1 Orbit simulation using Poliastro with J2 perturbation Completed
Step 2 Time-series dataset preparation for ML (sliding window, normalization) Completed
Step 3 Build and train LSTM model to predict future orbital positions Completed
Step 4 Compare ML predictions with RK4 propagation (baseline) In Progress
Step 5 Evaluate hybrid model (average RK4 + LSTM) and visualize errors Pending

Repository Structure

├── data/ # Raw and preprocessed data │ ├── orbit_j2.csv # Simulated J2 orbit data │ ├── X_train.npy # LSTM training input │ ├── y_train.npy # LSTM training labels │ ├── lstm_predictions.npy # LSTM model predictions │ └── position_scaler.save # Scaler for normalizing input ├── src/ # Source code │ ├── simulate_orbit.py # Orbit simulation script │ ├── prepare_dataset.py # Time-series dataset generation │ ├── lstm_model.py # LSTM model definition and training │ └── plot_predictions.py # Visualization and comparison plots ├── results/ # Output plots and evaluation metrics ├── research_summary.md # Technical notes and research documentation ├── requirements.txt # Python dependencies └── README.md # Project overview


Setup and Usage

Install dependencies:

pip install -r requirements.txt
python3 src/prepare_dataset.py
python3 src/lstm_model.py
python3 src/plot_predictions.py

About

HOPE is a hybrid orbit prediction engine that simulates and forecasts the motion of Low Earth Orbit (LEO) satellites under perturbative forces. The initial version focuses on the J2 zonal harmonic and combines traditional physics-based propagation with machine learning models such as LSTMs. The goal is to evaluate the effectiveness of ML in orbital

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages