Bhargav M. Joshi
This repository contains the codebase developed for the MIT ARCLab STORM-AI Challenge 2025, focused on forecasting thermospheric orbit-averaged neutral mass density using AI/ML methods. Accurate density prediction is critical for satellite drag modeling, orbit propagation, and space situational awareness—especially during geomagnetically active periods. The approach integrates physics-informed feature engineering with state-of-the-art machine learning techniques, including: 🌡️ XGBoost-based Conditional Models for initial density estimation 🔄 TSLANet (Time Series Lightweight Adaptive Network) for 72-hour density trend forecasting ⚙️ Tools for data preprocessing, scaling, and feature extraction based on OMNI2 and ESA SWARM datasets 📊 Benchmarking against truth data using custom evaluation metrics such as OD-RMSE This repository is organized for transparency and reproducibility, and includes: Training scripts for decision tree and neural models Inference pipelines for both Phase 1 and Phase 1.2 datasets Environment setup (.yml) and dependencies Example visualizations and evaluation outputs We hope this project helps advance data-driven techniques in thermospheric modeling and supports broader applications in space weather forecasting and satellite operations.