SatGuard

March 22, 2026 · View on GitHub

Open-source conjunction assessment pipeline for space objects — from TLE ingest to maneuver planning.

TLE ingest → SGP4 propagation → conjunction screening → collision probability → CDM output → 3D globe → fleet ops → maneuver planning

SatGuard provides a complete Python toolkit for satellite conjunction assessment, implementing industry-standard algorithms (Foster, Chan, Alfano) with rigorous validation against published references (Vallado, Alfano 2005, NASA CARA). Includes an interactive CesiumJS 3D globe, fleet batch screening, PDF reports, CW-based maneuver planning, and historical replay.

Features

Core (v0.1)

TLE Parser — Two-Line Element parsing with checksum validation
Catalog Ingest — CelesTrak (no auth) and Space-Track.org (with credentials)
SGP4 Propagation — High-fidelity orbit propagation via compiled sgp4 library
Conjunction Screening — KDTree-based spatial indexing for O(N log N) screening
Collision Probability — Three methods: Foster (2D integral), Chan (series expansion), Alfano (max Pc bounds)
Covariance Handling — Default covariance models, 3D→2D encounter plane projection
CDM Output — CCSDS Conjunction Data Message (KVN format)
CLI — One-command screening: satguard screen --norad-id 25544 --days 7

Monitoring (v0.2)

Pc Evolution Tracking — Track how collision probability changes over time (JSON history)
Trend Analysis — Detect RISING/FALLING/STABLE trends, estimate time to threshold
Webhook Alerts — Fire alerts to Slack/Discord/Teams when Pc exceeds threshold
Covariance Assessment — Evaluate matrix quality (REALISTIC/SUSPECT/DEFAULT)
CLI watch — Screen + record + alert in one command
CLI history — View Pc evolution with optional matplotlib plot

Globe 3D (v0.3)

Interactive Globe — CesiumJS with 30K+ satellites rendered via PointPrimitiveCollection + satellite.js client-side SGP4
Click-to-Inspect — Click any satellite → fly-to camera transition + detail panel (orbital elements, derived params)
Conjunction Overlay — Glowing polylines between at-risk pairs, colored by Pc severity
Orbit Filters — Toggle LEO/MEO/GEO/OTHER with color coding
Time Controls — Play/pause, speed multiplier (1x–600x), reset-to-now
FastAPI Backend — REST API serving catalog, conjunctions, and object details
CLI satguard serve — Single command to launch API + globe

Globe Enhanced (v0.4)

Siblings Detection — Filter co-orbiting objects (same launch designator) from results
3D Conjunction Arcs — Animated polylines connecting at-risk pairs on the globe
Heatmap Mode — Density overlay showing conjunction hotspots
Time Slider — Scrub through simulation time with playback
ConjunctionBrowser — Browse, sort, and filter all conjunctions in sidebar panel

Fleet Operations (v0.5)

Fleet YAML Config — Define constellations with per-object thresholds in fleet.yaml
Batch Screening — Screen entire fleet in one command (vectorized, 29s for full catalog)
PDF Reports — Conjunction summary with risk matrix (fpdf2)
CLI satguard fleet — Batch screen a constellation
CLI satguard report — Generate PDF report

Maneuver & Replay (v0.6)

CW Maneuver Planning — Clohessy-Wiltshire linearized relative motion for avoidance maneuvers
Tradespace Analysis — Grid search (delta-v × time) → minimum-fuel option meeting Pc threshold
Historical Replay — Re-propagate archived TLEs to reconstruct conjunction timeline
CLI satguard maneuver — Plan avoidance maneuver for a conjunction pair
CLI satguard replay — Replay historical conjunction from archived TLEs

Installation

pip install satguard

Or for development:

git clone https://github.com/genius-lab/satguard.git
cd satguard
pip install -e ".[dev]"

Requires Python 3.12+.

Quick Start

CLI

Screen the ISS (NORAD 25544) for conjunctions over the next 3 days:

satguard screen --norad-id 25544 --days 3

With CDM output for each conjunction:

satguard screen --norad-id 25544 --days 7 --threshold 50 --output-cdm

Python API

from satguard import (
    parse_tle, propagate_batch, screen,
    foster_pc, default_covariance, write_cdm,
)
from satguard.covariance.realism import project_to_encounter_plane

# 1. Parse TLEs
tle_primary = parse_tle(tle_string_primary)
tle_secondary = parse_tle(tle_string_secondary)

# 2. Propagate orbits (3 days, 60s steps)
states_p = propagate_batch(tle_primary, days=3.0, step_seconds=60.0)
states_s = propagate_batch(tle_secondary, days=3.0, step_seconds=60.0)

# 3. Screen for conjunctions
events = screen(states_p, states_s, threshold_km=50.0)

# 4. Compute collision probability
if events:
    event = events[0]
    cov_2d = project_to_encounter_plane(
        default_covariance("LEO"), default_covariance("LEO"),
        event.r_primary, event.v_primary,
        event.r_secondary, event.v_secondary,
    )
    pc = foster_pc(event.miss_distance_km, cov_2d, hard_body_radius=0.02)
    print(f"Pc = {pc:.2e}")

    # 5. Generate CDM
    cdm = write_cdm(event, pc)
    print(cdm)

CelesTrak Catalog

import asyncio
from satguard import fetch_catalog, fetch_tle_by_norad

# Fetch a single object
tle = asyncio.run(fetch_tle_by_norad(25544))  # ISS
print(f"{tle.name} — inclination: {tle.inclination_deg:.1f}°")

# Fetch full active catalog
catalog = asyncio.run(fetch_catalog("active"))
print(f"{len(catalog)} active objects")

CLI Reference

satguard screen [OPTIONS]
  --norad-id INTEGER     NORAD catalog number [required]
  --days FLOAT           Screening window in days (default: 3)
  --threshold FLOAT      Distance threshold in km (default: 50)
  --step FLOAT           Propagation step in seconds (default: 60)
  --output-cdm           Output CDM for each conjunction
  --record               Save results to Pc history (v0.2)
  --assess-covariance    Show covariance quality metrics (v0.2)

satguard watch [OPTIONS]           (v0.2)
  --norad-id INTEGER     NORAD catalog number [required]
  --days FLOAT           Screening window in days (default: 3)
  --config PATH          Alert config TOML file
  --history-dir PATH     History storage directory

satguard history [OPTIONS]         (v0.2)
  --norad-ids TEXT        Comma-separated NORAD IDs (e.g., 25544,41335) [required]
  --history-dir PATH     History storage directory
  --plot                 Save Pc evolution plot as PNG

satguard alert-test [OPTIONS]      (v0.2)
  --config PATH          Alert config TOML file

satguard serve [OPTIONS]           (v0.3)
  --port INTEGER         API server port (default: 8000)

satguard fleet [OPTIONS]           (v0.5)
  --config PATH          Fleet YAML config file [required]
  --output-dir PATH      Output directory for results

satguard report [OPTIONS]          (v0.5)
  --config PATH          Fleet YAML config file [required]
  --output PATH          Output PDF path

satguard maneuver [OPTIONS]        (v0.6)
  --norad-id INTEGER     Primary NORAD ID [required]
  --norad-id-secondary INTEGER  Secondary NORAD ID [required]
  --max-dv FLOAT         Maximum delta-v in m/s (default: 1.0)
  --pc-threshold FLOAT   Target Pc threshold (default: 1e-5)

satguard replay [OPTIONS]          (v0.6)
  --norad-ids TEXT        Comma-separated NORAD IDs [required]
  --history-dir PATH     History storage directory

Example Output

SatGuard Conjunction Screening
========================================
Primary object:   NORAD 25544
Window:           3.0 days
Threshold:        50.0 km
Step:             60 s

Fetching primary TLE from CelesTrak...
  Object: ISS (ZARYA)
  Epoch:  2024-02-14T12:25:11

Propagating primary orbit...
  Generated 4320 state vectors

Fetching active catalog from CelesTrak...
  Catalog: 8432 objects

Screening...

Results: 5 conjunction(s) found
============================================================

  [1] NORAD 25544 vs 41335
      TCA:           2024-02-15T08:12:33
      Miss distance: 12.847 km
      Rel. velocity: 14.221 km/s
      Pc (Foster):   2.31e-06

Collision Probability Methods

Method	Reference	Best For
Foster	Foster & Estes 1992	Standard 2D Pc, isotropic covariance
Chan	Chan 2008, AIAA J.	Anisotropic covariance, series expansion
Alfano	Alfano 2005, AAS	Upper/lower bounds, quick screening

Validation

SatGuard is validated at 5 levels with 254 tests:

Level	Description	Count
L1 Unit	Core function behavior	~158
L2 Domain	Values from external sources (Vallado, Alfano, NASA CARA)	~25
L3 Property	Hypothesis-based invariant testing (Hypothesis + CW)	18
L4 Snapshot	Golden output approved by human review	1
L5 External	Cross-validation vs Chan 2008, NASA CARA, real CSMs	15

M4 two-tool verification: CW analytical vs Hill numerical integration (7/7 PASS, <0.001% error).

Run all checks:

make check-all    # mypy + ruff + pytest + smoke test
pytest tests/ -v  # tests only

Project Structure

satguard/
├── src/satguard/
│   ├── catalog/       # TLE parsing, CelesTrak & Space-Track ingest
│   ├── propagate/     # SGP4 orbit propagation
│   ├── screen/        # KDTree + vectorized SatrecArray screening
│   ├── assess/        # Collision probability (Foster, Chan, Alfano)
│   ├── covariance/    # Covariance models, projection & assessment
│   ├── history/       # Pc evolution tracking + historical replay
│   ├── alert/         # Webhook alerts (Slack/Discord/Teams)
│   ├── fleet/         # Fleet YAML config + batch screening
│   ├── report/        # PDF report generation (fpdf2)
│   ├── maneuver/      # CW equations + tradespace planner
│   ├── api/           # FastAPI backend (catalog, conjunctions, maneuver, replay)
│   ├── cdm/           # CCSDS CDM writer
│   └── cli/           # Click CLI (screen, watch, fleet, maneuver, replay, serve)
├── tests/             # 254 tests across 5 validation levels
├── verify/            # Two-tool cross-verification (CW vs Hill)
└── web/               # CesiumJS 3D globe (React + Resium)

References

Vallado, D.A. Fundamentals of Astrodynamics and Applications, 5th ed.
Alfano, S. (2005). "A Numerical Implementation of Spherical Object Collision Probability." AAS/AIAA.
Chan, F.K. (2008). Spacecraft Collision Probability. AIAA.
Foster, J.L. & Estes, H.S. (1992). "A Parametric Analysis of Orbital Debris Collision Probability." NASA JSC-25898.
CCSDS 508.0-B-1: Conjunction Data Message standard.
Clohessy, W.H. & Wiltshire, R.S. (1960). "Terminal Guidance System for Satellite Rendezvous." J. Aerospace Sciences.

Acknowledgments

Development assisted by Claude Code (Anthropic).

License

MIT — see LICENSE.