astroz

Astronomical and Spacecraft Toolkit Written in Zig

astroz is an astrodynamics and spacecraft toolkit with a fast SGP4/SDP4 propagator at its core.

Orbital Mechanics	Spacecraft Ops	Astronomy
SGP4/SDP4 propagation	CCSDS packets	FITS parsing
TLE & OMM parsing	VITA49 packets	WCS coordinates
Orbital maneuvers	Attitude determination	Star precession
Force models (J2-J4, drag, SRP, third-body)	CSPICE ephemeris	Celestial bodies
Dormand-Prince 8(7) integrator	Monte Carlo sims

Performance

SGP4 (near-earth) or SDP4 (deep-space) is picked per satellite from its orbital period. Satellites are propagated 8 at a time with SIMD (AVX512, or AVX2/SSE at 4 wide), and constellations are split across threads in time-major order. ASTROZ_THREADS sets the thread count; the default is every available core.

Single-Threaded (1.2M propagations, single satellite)

Implementation	Props/sec	Speedup vs python-sgp4
astroz	30.8M	11x
Rust sgp4	5.1M	1.8x
heyoka	3.8M	1.3x
satkit	3.5M	1.2x
python-sgp4	2.8M	1x

Multi-Threaded Constellation (13,478 satellites × 1,440 times)

Implementation	1 Thread	16 Threads
astroz	37.7M/s	303M/s
heyoka	15.7M/s	155.6M/s
Rust sgp4 (rayon)	4.4M/s	47.9M/s
satkit	3.5M/s	3.5M/s
python-sgp4	2.7M/s	2.7M/s

Measured on an AMD Ryzen 7 7840U (16 threads), each implementation in its fastest configuration (SIMD, pre-allocated outputs, batch mode). The scripts for astroz, Rust sgp4 and python-sgp4 are in benchmarks/ (see the Justfile).

The Cesium visualization example propagates the whole active catalog (about 13,000 satellites) at interactive rates. There is a live demo.

Accuracy

What the test suite checks (zig build test):

Component	Reference	Tolerance
SGP4	Vallado AIAA 2006-6753 vectors	10 m position, 1 mm/s velocity
SDP4 (GPS, GEO, HEO)	python-sgp4	10 m position
SGP4 over one day	python-sgp4	100 m position, 0.1 m/s velocity
J2 acceleration	Analytical	1%
Two-body propagation	Analytical Kepler	1 m after one orbit
Two-body energy	Conservation	1e-10 relative over 100 orbits

CI also checks the Python bindings against python-sgp4 to within 1 m.

With CSPICE enabled: zig build test -Denable-cspice=true

Python

Wheels are published for Linux (x86_64, arm64) and macOS (arm64), Python 3.10 to 3.14:

pip install astroz

astroz.api follows the python-sgp4 API and also handles deep-space (SDP4) satellites:

from astroz.api import Satrec, jday, WGS72

line1 = "1 25544U 98067A   24127.82853009  .00015698  00000+0  27310-3 0  9995"
line2 = "2 25544  51.6393 160.4574 0003580 140.6673 205.7250 15.50957674452123"
sat = Satrec.twoline2rv(line1, line2, WGS72)

jd, fr = jday(2024, 5, 6, 12, 0, 0.0)
error, position, velocity = sat.sgp4(jd, fr)

The Python README covers batch propagation, the high-level propagate/screen API, orbital mechanics helpers and numbers for migrating from python-sgp4.

Usage

Requires Zig 0.17.0.

• Fetch the package so it gets recorded in your build.zig.zon. Pin a release tag or commit for reproducible builds, or follow the latest main:

# pinned to a tag or commit
zig fetch --save https://github.com/ATTron/astroz/archive/<tag-or-commit>.tar.gz

# latest main
zig fetch --save git+https://github.com/ATTron/astroz

• Wire the astroz module into your compile step in build.zig:

const astroz = b.dependency("astroz", .{ .target = target, .optimize = optimize });
exe.root_module.addImport("astroz", astroz.module("astroz"));

• Propagate a satellite. Satellite picks SGP4 or SDP4 for you:

const astroz = @import("astroz");

var tle = try astroz.Tle.parse(tle_string, allocator);
defer tle.deinit();

const sat = try astroz.Satellite.init(tle, astroz.constants.wgs84);
const result = try sat.propagate(60.0); // minutes from epoch
const pos = result[0]; // [x, y, z] km
const vel = result[1]; // [vx, vy, vz] km/s

Examples

Run them all with zig build example, or only compile them with zig build build-examples.

Zig

• sgp4.zig: TLE and OMM parsing (including six-digit catalog numbers), SGP4 one time and eight at once with SIMD, Satellite picking SGP4 or SDP4, and the Earth-fixed position.

• constellation.zig: a mixed near-earth and deep-space constellation propagated in batches, then screened for close approaches with Screening.

• numerical_propagation.zig: one orbit propagated with two-body (RK4 and Dormand-Prince 8(7)), with J2, and with J2, solar radiation pressure and Sun/Moon gravity. Uses SPICE ephemeris when built with -Denable-cspice=true after zig build fetch-kernels.

• spacecraft.zig: impulse, plane change and phase change maneuvers and how they change the orbit, then attitude propagation.

• mission_design.zig: Earth to Mars by Hohmann and bi-elliptic transfer, the transfer propagated, and Monte Carlo statistics on it.

• packets.zig: building and parsing CCSDS packets and parsing a VITA 49 stream.

• astronomy.zig: star precession and reading a FITS image into a PNG.

  

Python

• python_sgp4.py: the python-sgp4 compatible API.
• constellation.py: loading a constellation from CelesTrak or a TLE/OMM file, propagating to Earth-fixed and geodetic coordinates, and screening it for close approaches.
• orbital_mechanics.py: circular-orbit numbers, Hohmann and bi-elliptic transfers, Lambert’s problem and numerical propagation with J2.
• Cesium visualization (live demo): the active catalog propagated and rendered in Cesium.