Autonomous Space Edge Computing

Run Kubernetes in Low-Earth Orbit.
Without Draining Spacecraft Batteries.

k8star transforms satellite constellations into resilient, self-orchestrating compute clusters. Powered by real-time SGP4 orbital propagation, optical laser ray-casting, and zero-ground-station autonomous workload migration.

3,161+
Satellites In Catalog
3s
Orbital Tick Loop
0
Ground Stations Needed
80 km
Atmosphere LoS Buffer
Real-time 1× Orbital Physics Mesh
SGP4 / JPL DE421
ISS (ZARYA)
NORAD 25544 • Node-0
🤖 AI HOST
SENTINEL-2A
NORAD 40697 • Node-1
☀️ SUNLIGHT
STARLINK-1007
NORAD 44713 • Node-2
☀️ SUNLIGHT
Orbital Realities

The Challenge of Edge AI in Orbit

Low-Earth Orbit is hostile to traditional Kubernetes scheduling. High-power edge inference, solar eclipses, and laser Line-of-Sight constraints require autonomous, orbital-aware middleware.

Traditional Space Architecture Bottlenecks

  • ✕
    Solar Eclipses Drain Critical Batteries: LEO satellites spend 35–40 minutes of every 90-minute orbit in Earth's shadow. Running 200W+ AI inference during eclipse empties satellite battery reserves, threatening life support and propulsion.
  • ✕
    Ground Station Communication Lag: Satellites only pass over ground stations for 5–10 minutes every 45–90 minutes. A cluster cannot wait for ground control intervention to reschedule crashed or power-starved workloads.
  • ✕
    Atmospheric Laser Attenuation: Inter-Satellite Laser links (OISL) break if optical vectors graze Earth's dense atmosphere (below ~80 km altitude) or are occluded by the planet itself.

The k8star Autonomous Solution

  • ✓
    Real-Time SGP4 Solar Propagation: Continuous 1× UTC calculations evaluate solar panel illumination against NASA JPL ephemeris (de421.bsp) every 3 seconds.
  • ✓
    Ray-Casting 3D Optical Line-of-Sight: Calculates Euclidean distances and ray-casts optical paths across all peers with an 80 km atmospheric safety buffer before failover.
  • ✓
    Zero-Downtime Autonomous Migration: Labels the closest sunlit peer with nearest-neighbor=primary and taints the dark satellite with :NoExecute, migrating AI pods over laser cross-links automatically!
Live Demonstration

Interactive Eclipse Failover Simulator

Experience how k8star autonomously orchestrates workloads when a satellite plunges into Earth's shadow. Click below to trigger the orbital failover sequence.

Simulate Orbital Transition Event
Active Host: k3d-orbital-mesh-agent-0 (ISS) | Target: k3d-orbital-mesh-agent-1 (SENTINEL-2A)
STEP 01
Eclipse Transition
SGP4 detects host satellite entering Earth's shadow. Solar panel irradiance drops to 0.0 W/m².
STEP 02
Optical LoS Ray-Casting
Calculates 3D Euclidean distances and ray-casts optical laser vectors, enforcing 80 km atmosphere clearance.
STEP 03
Neighbor Tag & Node Taint
Labels closest sunlit peer as nearest-neighbor=primary before tainting host with :NoExecute.
STEP 04
Autonomous Pod Migration
K8s evicts AI pod across laser cross-link to sunlit peer. Workload resumes with zero battery drain!
[SYSTEM READY] Orbital daemon operating at 1× realtime UTC (3s tick).
[CLUSTER] 3 Satellite nodes online: ISS, SENTINEL-2A, STARLINK-1007.
[WORKLOAD] Pod 'satellite-image-filter' healthy on ISS (ZARYA).
Core Engineering

Engineered for Space-Edge Computing

k8star bridges astrophysics, astrodynamics, and cloud-native orchestration into a single autonomous middleware stack.

SGP4 & NASA Ephemeris Engine

Propagates TLE orbits using Python Skyfield and NASA JPL de421.bsp ephemeris at 1× realtime UTC, computing geocentric 3D Euclidean coordinates and WGS84 subpoints.

Laser Line-of-Sight Ray-Casting

Simulates optical laser links (OISL) by projecting Earth's center onto the vector segment between satellites, strictly rejecting links that graze within the 80 km atmospheric boundary.

Race-Condition-Free Scheduling

Patches the sunlit optical destination node with nearest-neighbor=primary prior to applying :NoExecute taints to prevent scheduler deadlocks during eviction.

Zalando RESTful API Standard

Strictly compliant REST endpoints (/satellites, /satellite-telemetry, /cluster-nodes) with Pydantic schemas, RFC-compliant error envelopes, and kebab-case URIs.

Relational Telemetry Integrity

SQLite database engine enforcing active foreign keys (PRAGMA foreign_keys = ON) with cascading deletes across 3,100+ satellites and real-time telemetry updates.

Clean Taint Deletion Architecture

Avoids cluster lockups when satellites re-enter sunlight by executing clean JSON merge patches with empty lists [], ensuring instant workload eligibility.

Use Cases

Critical Space Mission Profiles

From high-throughput hyperspectral Earth observation to resilient defense communications.

Earth Observation & Wildfires

Real-Time Hyperspectral Image Filtering

Satellites equipped with high-resolution optical cameras process terabytes of raw imagery to detect forest fires and natural disasters in real-time, handing off inferences over laser links before eclipse entry.

Impact: Reduces ground-station downlink latency from 60 minutes to seconds.
Laser Cross-Link Constellations

OISL Inter-Satellite Compute Meshes

Mega-constellations like Starlink and Kuiper utilize optical cross-links to form an orbital internet backbone. k8star ensures compute containers follow the Sun across orbital planes.

Impact: 100% solar panel utilization across orbital planes with zero battery degradation.
Defense & SDA Transport Layer

Autonomous Threat Detection & SSA

In contested space environments where ground telemetry links may be jammed, satellite clusters execute autonomous threat detection and orbital track correlation completely offline.

Impact: Total survivability without reliance on terrestrial ground control stations.

Deploy Autonomous Intelligence to Orbit

Ready to transform your Low-Earth Orbit constellation into a self-orchestrating Kubernetes cluster? Explore the repository or start the local testbed cluster.

Try Failover Simulator Explore Source Code