AI-Driven Space Debris Collision Risk Management
Built at Women in Data Datathon 2025 by Team Datanova
Space isn’t as empty as it looks. Thousands of satellites and millions of debris fragments orbit Earth, and one wrong trajectory could trigger a chain reaction of collisions.
OrbitShield is our attempt to make space situational awareness predictive, explainable, and actionable.
We combined geospatial analytics, machine learning, and real-time alerts to identify collision risks and make orbital safety more transparent.
- Data Fusion: Pulled and consolidated 845K+ orbital records via REST APIs (Ephemeris, State Vectors, Metadata).
- Geospatial Analysis: DBSCAN clustering on orbital trajectories to detect debris-dense regions.
- ML Forecasting: Random Forest, XGBoost, and LSTM models to predict collision risks and future positions.
- Human-Centric Alerts: WhatsApp alerts with plain-English warnings instead of raw risk scores.
- Visualization: Interactive dashboard showing orbital clusters, risks, and trajectory forecasts.
- Dataset size: 845K records | 228 unique objects
- Clusters identified: 7 orbital debris clusters
- Risk classification:
- High Risk: 45 satellites
- Medium Risk: 55 satellites
- Low Risk: 4955 satellites
- Forecasting: LSTM achieved accurate position predictions (x, y, z) across epochs.
- Alerts: Generated explainable notifications with risk
- Data Pipeline: Python (Pandas, NumPy), REST APIs
- Database & GIS: PostgreSQL + PostGIS
- Clustering: DBSCAN (scikit-learn)
- Modeling: Scikit-learn (RF, XGBoost), TensorFlow/Keras (LSTM)
- Visualization: Plotly,PowerBi
- Alerts: WhatsApp API, HuggingFace Transformers
Key outputs and deliverables:
- Kamayani Rai – Applied Analytics & Machine Learning
- Kundana Rasi Tadikonda – Analytics Engineering & Visualization
- Julia Davis – Industrial Engineering & Visualization
- Kimberley Gillette – Data Science & Machine Learning
- Real-time integration with live orbital feeds.
- Reinforcement learning for automated maneuver recommendations.
- Public-facing dashboard for policy makers, educators, and global transparency.