Skip to content

Repository files navigation

OrbitalOps

Satellite network operations platform with ML-driven link prediction and 5G-style dynamic routing.


Problem

Satellite links degrade gradually — signal drops, latency climbs, packet loss fluctuates. Operators notice too late. OrbitalOps detects degradation early using ML prediction and reroutes traffic automatically.


How It Works

Telemetry ingestion — Streams latency, packet loss, bandwidth, jitter, signal strength from orbital nodes through Kafka into TimescaleDB.

OrbitalOps Predicts (ML) — XGBoost classifier trained on telemetry features:

features = [latency_avg, packet_loss, jitter, signal_strength, bandwidth_util, congestion_score]

link_degradation_probability = model.predict(features)  →  0.82

if probability > 0.75 → WARNING
if probability > 0.90 → CRITICAL

The model learns patterns like high latency + rising packet loss → likely degradation.

Anomaly detection — Isolation Forest flags outliers in the feature space. Sigma-threshold detection in C catches spikes in real time.

5G-style dynamic routing — When a link degrades, the system computes route scores and switches traffic:

route_score = 0.35 * signal + 0.25 * bandwidth + 0.20 * (1/latency) + 0.20 * (1/congestion)

Picks the highest-scoring route. Mirrors 5G network slicing and traffic management.

Edge processing — Ground station nodes filter noise, compress data, and detect anomalies locally before forwarding to the cloud. Reduces bandwidth, mirrors 5G edge architecture.


Architecture

graph TD
    SAT["Satellites"] --> EDGE["Edge Nodes"]
    EDGE --> API["API Gateway · Java 21"]
    API --> K["Kafka"]
    K --> TP["Telemetry Pipeline"]
    TP --> DB["TimescaleDB"]
    TP --> R["Redis"]
    K --> ML["OrbitalOps Predicts · XGBoost"]
    ML --> AD["Anomaly Detection"]
    AD --> AE["Alert Engine"]
    AE --> RT["Dynamic Rerouting · Dijkstra"]
    R --> UI["Dashboard · React"]
    API --> SP["Signal Processing · C/C++"]
Loading

Tech Stack

Layer Tech
API Java 21, Spring Boot 3.2
Streaming Kafka
Database PostgreSQL, TimescaleDB
ML Python, XGBoost, Isolation Forest
Signal C, C++ (FFT, link budget, Doppler)
Dashboard React, Leaflet, Recharts
Infra Docker, Terraform, GitHub Actions

Quick Start

git clone https://github.com/aryanputta/orbitlink.git
cd orbitlink
cp .env.example .env

# Install Python Requirements
pip install -r packages/ml-service/requirements.txt
pip install -r packages/simulation/requirements.txt

docker-compose up --build

Dashboard: http://localhost:3000


License

MIT

About

OrbitLink: space-link network emulation testbed with tc-netem link profiles, controller, and Clopper-Pearson reliability stats

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages