OrbitOps — Intelligent Satellite Mission Planning & Operations Platform

A state-of-the-art, AI-powered mission control platform for satellite fleet management, constraint-based scheduling, resource intelligence, and ground station optimization — built for real-world space operations.

---

Table of Contents

1. Overview
2. Problem Statement
3. Proposed Solution
4. Features
5. System Architecture
6. Folder Structure
7. File-by-File Explanation
8. Tech Stack
9. Algorithms Used
10. Data Structures Used
11. Design Patterns
12. Research Papers and References
13. Installation
14. Configuration
15. How to Use
16. API Documentation
17. Database Schema
18. ML Model Training Pipeline
19. Security
20. Performance Analysis
21. Testing
22. Deployment
23. Limitations
24. Future Improvements
25. Troubleshooting
26. License
27. Contributors
28. Acknowledgements
29. Conclusion

---

Overview

OrbitOps is a full-stack, AI-assisted satellite mission control platform. It provides space operations teams with a unified interface to:

• Plan and schedule satellite missions with priority-based conflict resolution
• Monitor satellite telemetry (battery, temperature, signal strength, CPU, memory) in real time
• Optimize ground station selection using multi-criteria scoring strategies
• Predict battery depletion and resource risk using trained ML models (GradientBoosting + RandomForest)
• Generate AI recommendations that identify critical fleet issues proactively
• Export PDF and Excel operational reports

The platform follows a clean React + FastAPI + PostgreSQL architecture with a dedicated ML training pipeline (/training) and pre-trained model artifacts (/trained_models). It seeds a fully operational demo dataset on first run — 20 satellites, 15 global ground stations, 14 payloads, and 100 missions — making it immediately usable without external data.

---

Problem Statement

What Real-World Problem Does This Solve?

Modern satellite operations involve managing dozens to hundreds of satellites simultaneously, each with:

• Strict resource budgets: battery (solar + eclipse cycles), onboard memory, CPU, payload power
• Time-constrained visibility windows: a satellite in LEO (~400 km) has windows of only 5–10 minutes over any single ground station
• Overlapping mission demands: multiple missions competing for the same satellite or ground station
• Silent failure risk: a battery critically draining or thermal threshold being breached during a mission can cause permanent hardware damage or mission failure

Why Is This Important?

Without intelligent tooling, operators must manually cross-reference telemetry across multiple dashboards, resolve scheduling conflicts by hand, and react to battery/thermal emergencies only after they occur. This is operationally unsafe and does not scale beyond small fleets.

Who Faces This Problem?

• Commercial satellite operators (fleet management companies)
• Government space agencies (NASA, ESA, ISRO)
• Academic and research satellite teams (CubeSat operators)
• Ground segment software engineers building mission control systems

Existing Limitations

• No predictive intelligence: traditional systems alert only after a threshold is breached, not before
• Manual conflict resolution: time-consuming and error-prone for large fleets
• Siloed tooling: separate systems for telemetry, scheduling, and planning without a unified workflow
• No data continuity: CRUD actions are not persisted back to datasets for ML retraining

---

Proposed Solution

OrbitOps addresses these limitations through a three-layer architecture:

1. Intelligent Scheduling Layer

A Branch and Bound Scheduler (BranchAndBoundScheduler.py) backed by a DynamicPriorityEngine sorts and schedules missions based on three weighted factors: base mission criticality (40%), time urgency (40%), and real-time resource availability (20%). A background job re-runs this every 15 minutes, and priority recalculation runs every 5 minutes via APScheduler.

2. ML-Powered Resource Intelligence Layer

Two trained models power real-time predictions:

• Battery Prediction (GradientBoostingRegressor, MAE ≈ 0.85 pp, R² = 0.9967): Given current satellite telemetry, predicts the remaining battery percentage at the end of the mission window and generates a multi-hour forecast trajectory.
• Resource Risk Classifier (RandomForestClassifier, 96% accuracy, macro F1 = 0.957): Classifies each satellite's current telemetry profile as LOW, MEDIUM, or HIGH risk, enabling proactive operator intervention.

3. Ground Station Optimization Layer

A GroundStationOptimizer evaluates every available ground station using multi-criteria scoring across six strategies (Minimum Latency, Maximum Coverage, Load Balancing, Minimum Cost, Energy Efficient, Balanced) and returns a ranked recommendation with rejected stations and their reasons.

End-to-End Workflow

Operator opens UI → AppContext loads live data from FastAPI →
ML models predict battery/risk → Rule engine generates recommendations →
Scheduler resolves conflicts → Operator approves optimized schedule →
Dataset sync writes back to CSV for future model retraining

---

Features

Core Features

Feature	Description	Key Source File(s)
Mission Operations Dashboard	Real-time fleet overview: active satellites, missions, alerts, resource gauges	pages/Dashboard.tsx
Mission Planning	Create, edit, duplicate, delete missions with full metadata	pages/MissionPlanning.tsx, services/mission.py
Mission Scheduler	Visual time-slot scheduler, today's schedule, calendar view	pages/MissionScheduler.tsx
Constraint-Based Optimization	Branch and Bound conflict resolution + ground station optimizer	optimization/BranchAndBoundScheduler.py, optimization/GroundStationOptimizer.py
Battery Prediction	GradientBoosting model predicts end-of-mission battery %, multi-hour forecast	services/resource_intelligence/battery_prediction.py
Resource Risk Classification	RandomForest classifies satellite telemetry as LOW/MEDIUM/HIGH risk	services/resource_intelligence/inference.py
AI Recommendations	Rule + model pipeline generates prioritized, actionable fleet recommendations	services/resource_intelligence/recommendation_engine.py
Satellite Operations	Telemetry visualization: battery, temperature, signal strength, power, CPU	pages/SatelliteOperations.tsx
Ground Station Planner	Multi-criteria ground station selection and optimization	pages/GroundStationPlanner.tsx, optimization/GroundStationOptimizer.py
Payload Planner	Assign, schedule, and monitor satellite payloads	pages/PayloadPlanner.tsx
Resources Dashboard	Fleet-wide utilization table with AI risk badges and battery forecasts	pages/Resources.tsx
Analytics	Charts, trends, and performance analytics across the fleet	pages/Analytics.tsx
Reports	Generate PDF and Excel operational reports for missions and alerts	pages/Reports.tsx, services/reportGenerator.ts
Settings	UI theme, constraint configuration, system preferences	pages/Settings.tsx
Auto-seeded Demo Data	20 satellites, 15 ground stations, 14 payloads, 100 missions on first boot	database/database.py
Dataset Sync	Mission and telemetry actions persist back to CSV for future ML retraining	services/dataset_manager.py
Background Scheduler	APScheduler jobs: priority recalc every 5 min, optimization every 15 min	jobs/scheduler.py
Health Check Endpoint	GET /health returns service status	main.py
OpenAPI Docs	Auto-generated Swagger UI at /docs	FastAPI built-in
Optional LLM Assistant	Anthropic Claude integration for natural-language recommendation explanations	services/ai/llm_service.py

---

System Architecture

High-Level Architecture

flowchart TD
    subgraph Frontend["Frontend — React / TypeScript / Vite"]
        UI[Pages & Components]
        CTX[AppContext — Global State]
        SVC[API Service Layer — Axios]
        RI[Resource Intelligence API]
        RPT[Report Generator — jsPDF / XLSX]
    end

    subgraph Backend["Backend — FastAPI / Python"]
        MAIN[main.py — FastAPI App]
        API[API Router /api]

        subgraph Endpoints["Endpoints"]
            E1[/missions]
            E2[/infrastructure]
            E3[/telemetry]
            E4[/alerts]
            E5[/recommendations]
            E6[/resources]
            E7[/optimization/ground-stations]
            E8[/maintenance/request]
        end

        subgraph Services["Service Layer"]
            MS[MissionService]
            IS[InfrastructureService]
            DS[DatasetManager]
            AI_SVC[LLM Service — Anthropic]
        end

        subgraph MLLayer["ML / Intelligence Layer"]
            REC[RecommendationEngine]
            INF[Inference — ResourceIntelligenceModels]
            BAT[BatteryPrediction]
            RU[ResourceUtilization]
            FE[FeatureEngineering]
        end

        subgraph Optimization["Optimization Layer"]
            BB[BranchAndBoundScheduler]
            PE[DynamicPriorityEngine]
            GSO[GroundStationOptimizer]
            TS[TaskSplitter]
            CS[ConstraintSolver]
        end

        JOBS[APScheduler Jobs — 5min / 15min]
    end

    subgraph Data["Data Layer"]
        DB[(PostgreSQL — Supabase / SQLite)]
        CSV[CSV Datasets — datasets/raw/]
        MODELS[Trained Models — .pkl]
    end

    UI --> CTX
    CTX --> SVC
    SVC --> API

    MAIN --> API
    API --> Endpoints
    Endpoints --> Services
    Endpoints --> MLLayer
    Endpoints --> Optimization

    Services --> DB
    Services --> CSV
    MLLayer --> MODELS
    Optimization --> DB

    JOBS --> BB
    JOBS --> PE

Loading

---

Tech Stack

Category	Technology	Version	Usage
Frontend Language	TypeScript	5.6	Strongly typed React components
Frontend Framework	React	19.0	SPA with hooks and context
Build Tool	Vite	5.4	Fast dev server and production bundler
CSS Framework	Tailwind CSS	3.4	Utility-first styling
Backend Language	Python	3.11+	Backend application logic
Backend Framework	FastAPI	0.109	Async REST API
ORM	SQLAlchemy (AsyncIO)	2.0	Async database access
Background Jobs	APScheduler	3.10	Async interval-based background tasks
ML — Regression	scikit-learn GradientBoostingRegressor	1.4	Battery prediction
ML — Classification	scikit-learn RandomForestClassifier	1.4	Resource risk classification
Database	PostgreSQL / SQLite	14+	Primary relational data store

---

Installation & Getting Started

Prerequisites

• Python 3.11+
• Node.js 18+ & npm 9+
• PostgreSQL (or local SQLite fallback)

Backend Setup

cd backend

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Run the development server
python -m uvicorn app.main:app --port 8000 --reload

Access API Swagger documentation at http://localhost:8000/docs.

Frontend Setup

cd frontend

# Install dependencies
npm install

# Start development server
npm run dev

Open application at http://localhost:5173.

---

API Documentation Overview

Base URL: http://localhost:8000/api

Method	Endpoint	Description
GET	/health	Liveness check
GET	/missions	List satellite missions
POST	/missions	Create a new mission
PUT	/missions/{id}	Update existing mission
DELETE	/missions/{id}	Delete mission
GET	/infrastructure/satellites	Get satellite fleet telemetry
GET	/infrastructure/ground-stations	Get ground station telemetry
POST	/optimization/ground-stations	Run multi-criteria ground station optimization
GET	/resources	Get fleet resource metrics & predictions
GET	/recommendations	Get AI recommendations
POST	/maintenance/request	Submit equipment maintenance request

---

Testing

Run unit & integration test suite:

python -m pytest tests/ -o pythonpath=backend