| G magnitude | |
| colour BP-RP | |
| distance from Sun (now) | |
| guiding radius | |
| orbital period | |
| radial swing | |
| vertical swing | |
| vφ − vᶜ(Rℓ) | |
| fit error at 500 Myr |
An orrery of the Milky Way: real stars, their orbits computed on a supercomputer, replayed in your browser across half a billion years.
Safeer Ali Mirani
GPU / XR engineer and computational neuroscientist.
safeer.ali.mirani@gmail.com · LinkedIn · github.com/SafeerAliMirani
Every point is a real star from ESA's Gaia mission, Data Release 3 (June 2022). Orrery shows of them, a uniform random sample of the stars with full six-dimensional phase space (3D position and 3D velocity) that pass quality cuts: parallax signal-to-noise ≥ 10, RUWE < 1.4, radial-velocity error < 5 km/s. Distances come from inverted parallaxes; positions and velocities are transformed into a Galactocentric frame (R₀ = 8.122 kpc). This is the bright subset Gaia can fully measure, mostly within a few thousand light-years of the Sun.
Each orbit was integrated ±1 billion years as a test particle in a fixed, smooth model of the Galaxy's gravity (gala's MilkyWayPotential2022) on the CINECA Galileo100 supercomputer, as a 34-task compute array. Each orbit was then compressed to a short quasi-periodic (NAFF-style) series of about a dozen terms — roughly 112 bytes per star — and that series is evaluated live on your GPU every frame. Clicking a star re-evaluates the same series at 512 epochs to draw its orbit ribbon.
Reconstruction error is a median of parsecs at 500 Myr; of orbits (hot or near-resonant stars) are not well described by a short series and are flagged in their dossier. The model has no bar, spiral arms, gas clouds, or self-gravity, so the motion is planetarium-grade, not a live simulation. Ensemble motion is trustworthy; any single star's deep-time position is an estimate. This shows the Sun's neighbourhood shearing and phase-mixing, not tidal streams forming.
Raw WebGPU / WGSL, no libraries or frameworks. Orbit integration and fitting in Python (gala, astropy, numpy).
Gaia data: ESA / Gaia / DPAC, pulled via the Leibniz-Institut (AIP) TAP mirror. Compute: CINECA (project EIRI_E_UNISA2). Gaia DR3 · gala
This piece needs a browser with WebGPU (Chrome or Edge 113+, recent Firefox or Safari).