AstraGen: Autonomous Spacecraft Flight Computer

• 

  Mission-Control HUD with 3D Earth Globe, orbital inclination rings, and live collision risk metrics.

• 

  Volumetric 3D Conjunction Covariance Ellipsoid rendering dynamic danger zones between assets.

• 

  Autonomous Google Gemini Flight Dynamics Officer (FDO) multi-axis Delta-V burn calculation.

• 

  Official NASA/ESA-compliant Conjunction Data Message (CDM) telemetry report generation.

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud o## Inspiration

With tens of thousands of active satellites and derelict debris orbiting Earth, the catastrophic risk of Kessler Syndrome—where a single collision triggers an unstoppable cascade of debris—threatens global space infrastructure. Traditional ground-based orbital tracking relies on delayed human analysis. We built AstraGen to pioneer an autonomous, real-time, mission-control-grade flight computer that fuses deterministic orbital mechanics with Google Gemini to evaluate conjunction hazards and execute automated collision avoidance maneuvers.

What it does

• Deterministic SGP4 & $J_2$ Perturbation Engine: Ingests live NORAD Two-Line Elements (TLE) to model exact orbital propagation, Earth oblateness precession ($\dot{\Omega}, \dot{\omega}$), and relative kinetic impact energy ($E_k = \frac{1}{2}mv^2$).
• 3D Volumetric Covariance Ellipsoids: Computes dynamic Mahalanobis distance matrices across Radial ($\sigma_r$), In-Track ($\sigma_t$), and Cross-Track ($\sigma_c$) planes to determine true collision probability ($P_c$).
• Gemini Multi-Agent Decision Hub: Autonomous Flight Dynamics Officer (FDO) agents calculate optimal, minimum-fuel multi-axis Delta-V burn vectors (Prograde/Retrograde, Out-of-Plane) and Hohmann transfer windows.
• NASA/ESA Conjunction Data Message (CDM) Standard: Generates cryptographically signed, standard-compliant orbital safety reports with one-click PDF export.
• 60 FPS Live WebSocket Stream & 3D Celestial Visualizer: Real-time WebGL/Three.js interactive globe rendering space assets, debris clouds, and danger zones.

How we built it

• Backend: FastAPI, Python, Pydantic, SGP4 orbital mechanics library, and WebSockets.
• Autonomous AI Agent: Google Gemini 1.5 decision engine for trajectory burn vector formulation and safety triage.
• Frontend & Visuals: Interactive Three.js (WebGL) 3D celestial engine, Tailwind CSS, JetBrains Mono typography, and Web Audio API for cockpit telemetry pings.
• Testing & Quality Assurance: Comprehensive Pytest test suites covering physics accuracy, $J_2$ drift equations, and API endpoints.

Challenges we ran into

Balancing computationally intensive 3D covariance matrix math and real-time SGP4 orbit propagation with sub-second latency was our biggest hurdle. We solved this by decoupling the physics pipeline into a high-performance deterministic calculation engine and streaming telemetry vectors via WebSockets at 60 FPS directly into Three.js.

Accomplishments that we're proud of

• Achieving 100% test pass rates across aerospace trajectory algorithms.
• Building a full-stack, deterministic space collision engine that operates autonomously without ground-station human delay.
• Designing a mission-control tier cyber-aerospace HUD that simplifies complex orbital dynamics for space operators.

What's next for AstraGen

• Live integration with Space-Track.org API for real-time tracking of 25,000+ cataloged orbital objects.
• Reinforcement learning fine-tuning on top of Gemini agents for autonomous multi-satellite constellation constellation-wide fleet avoidance coordination.f

What we learned

What's next for AstraGen: Autonomous Spacecraft Flight Computer

Built With

• fastapi
• google-gemini
• javascript
• pydantic
• pytest
• python
• sgp4
• tailwindcss
• three.js
• web-audio-api
• websockets