FAIRWAY orbital clearance, in public

Methodology

How the screening works

FAIRWAY takes public orbital data, propagates it forward, and finds where any of your objects passes close to something else. It reports the geometry of that pass honestly, and it is explicit about the one thing it cannot tell you.

The workflow

You pick a fleet : your own objects, the primaries. FAIRWAY screens that fleet against the full public catalog over a forward window , ranks every close approach it finds, and shows the result as a triage queue. You open an event to see its geometry and decide whether it is worth acting on. The console is the decision surface; the map is a situational view of the same data.

How a close approach is found

Each object's orbit is propagated with SGP4, the standard analytic model, run in a background thread so the interface stays responsive. From the elements FAIRWAY derives every object's perigee and apogee, then runs a three-stage search for each primary:

  1. Shell sieve. Discard any catalog object whose altitude band cannot overlap the primary's, expanded by the screening distance. This removes the large majority of the catalog cheaply.
  2. Coarse sweep. Step the surviving pairs forward across the window and track each pair's running minimum separation.
  3. Fine refine. Around each candidate minimum, step at a finer interval to pin the true closest approach.

For every approach it keeps, FAIRWAY records the geometric miss distance , the time to closest approach , the relative velocity at that moment , and the age of each object's elements . Pairs with effectively zero relative velocity (a station and its own docked modules, or formation-flying members) are excluded: they are co-located, not conjunctions.

What it computes

Geometric miss distance and time to closest approach over the forward window, relative velocity at closest approach, and how stale each object's orbital data is. Risk state is the ratio of miss distance to the threshold you set, shown by colour, icon, label, and sort position together so it never depends on colour alone.

What it does not compute, and why

FAIRWAY does not report a probability of collision . A rigorous probability needs each object's covariance, the uncertainty volume around its predicted position, and that covariance is not published in the open catalog. Presenting a probability from public elements alone would be false precision, so FAIRWAY refuses to. It frames its output the same way CelesTrak's SOCRATES frames its own: advance notice so a human can plan, not an automated go or no-go.

Where the numbers come from

Orbital elements come from CelesTrak's public General Perturbations data, read in the modern OMM (GP) JSON format rather than legacy TLE text. That choice matters: the public catalog runs out of 5-digit object numbers in mid-2026, after which new objects get 6-digit numbers that the fixed-width TLE format cannot hold. Reading OMM keeps FAIRWAY on the right side of that change.

The data is cached and screened in your browser rather than re-fetched, both to stay responsive and to respect CelesTrak's published bandwidth limits. The catalog updates only a few times a day, so there is nothing to gain from polling it harder.

Validation

The screening is checked against CelesTrak's SOCRATES service, which performs the same kind of catalog-wide screening. Where FAIRWAY's results diverge from SOCRATES, that divergence is surfaced rather than hidden. The aim is an honest screening instrument, not a number that looks more certain than it is.

Limits to keep in mind

Public elements carry their own uncertainty, typically on the order of a kilometre and growing with the age of the data, which is why element age sits next to every result. SGP4 positions are in the TEME frame and are treated here as inertial for display; the difference is well under a degree and immaterial for a screening view. FAIRWAY does not model maneuvers, and a screened approach is a geometry to assess, not a prediction that a collision will occur. For an operational decision, an operator pairs this kind of screening with covariance data they hold privately.

Open the console to screen a fleet, or read what this is built on.