Edge ML Space Weather Architecture for Ionospheric Research
- Winner: Out of the Box Solution Award - NASA Space Apps Challenge Kochi 2024
- Winner: Dr. Siby Mathew Endowment Intercollege Project Presentation 2025
ESWAR is an award-winning, mobile-first edge AI architecture designed to predict and visualize space weather anomalies, specifically Total Electron Content (TEC) and ionospheric disturbances. By processing 7 years of historical ISRO GPS receiver data and live hardware sensor feeds, this system delivers infrastructure-level predictive intelligence directly to a mobile endpoint.
- Powered by a custom Multi-Layer Neural Network trained on a massive, 7-year ISRO GPS receiver dataset.
- The model was converted and quantized into TensorFlow Lite (
tflite_flutter) to allow for high-speed, offline-capable, low-latency TEC predictions directly on the mobile edge, completely bypassing the need for heavy cloud inference.
- Integrates live magnetic field strength data captured via a Proton Precession Magnetometer stationed at the local observatory.
- Data is routed through a Firebase Realtime Database and instantly synced to the Flutter client, ensuring sub-minute updates.
- Interfaces with NASA OMNIWeb utilizing asynchronous REST APIs (
dio) to fetch surrounding interplanetary parameters (IMF, IMF-Bz, By, Bx, SymH).
The application is built around two primary analytical dashboards powered by fl_chart for high-performance rendering:
-
Panel 1: Magnetic Field Parameter Comparison
- Visualizes the hourly averaged geomagnetic field data from the local station alongside NASA OMNIWeb data (IMF Magnitude Avg, Bz GSE).
- Allows users to dynamically select dates and parameter types to generate custom, real-time comparison reports.
-
Panel 2: TEC Prediction & Disturbance Mapping
- Displays both 1-minute and hourly average predictions for Adjusted TEC values using the onboard TFLite model.
- Maps the Equatorial Electrojet (ΔH) change in the local geomagnetic field.
- Plots the Sym-H values, providing a complete, synchronized view of ionospheric disturbances.
- Framework: Flutter (
sdk: flutter) - State Management & UI:
provider,fl_chart,shimmer,flutter_spinkit,lottie - Edge ML:
tflite_flutter - Backend & Auth:
firebase_core,firebase_database,cloud_firestore,firebase_auth - Networking & Local Storage:
http,dio,shared_preferences,flutter_config
- Clone the repository:
git clone https://github.com/Harigovind04/Equatorial-Space-Weather-Application-for-Ionospheric-Research.git