Proof and documentation
SPOCK: Spacecraft Proximity and Orbital Conjunction Kinematics
What the SPOCK toy demonstration shows, the open-literature mathematics behind it, the independently checked reference values, and the clear statement that it is illustrative, not operational.
What it is
SPOCK is an interactive, in-browser companion to a published satellite collision-avoidance paper. It makes one point concrete: a satellite conjunction probability and a lattice decoding-failure probability are the same mathematics, because both are the same integral over the same 2x2 quadratic form, the same Gaussian, and the same disc. The demonstration has two tabs driven by a single computation engine:
- B-plane tab: the conjunction geometry at closest approach, with a maneuver slider.
- Toy-lattice tab: a good versus bad basis for the same lattice, closest-vector decoding, and the matching decoding-failure integral.
Both tabs call one routine (probInDisc) that asks how much Gaussian probability mass sits inside a disc. In the B-plane the quadratic form is a projected state covariance; in the lattice it is the basis Gram matrix. One piece of linear algebra, two sets of units.
Attribution: the demonstration packages an interactive companion to "Mathematical Foundations of Automated Satellite Collision Avoidance" by CAPT (USN Ret) William Edwards, PhD. The demo page credits the paper author publicly; packaging is by Qrypto Federal.
What this proves to you (plain English)
- The conjunction probability you would use to decide whether two spacecraft need a maneuver and the decoding-failure probability that underpins lattice cryptography are literally the same calculation. "One core, many markets" is made concrete here.
- Everything you see is computed live in your browser from the paper's own equations. Nothing is pre-rendered and nothing is fetched from a server, so there is no hidden lookup table to doubt.
- A collision-avoidance maneuver has a sign: within a commandable burn range there is a burn that RAISES collision probability rather than lowering it, because the along-track offset sweeps the relative position through the miss vector on its way out. The demo lets you watch that happen, which is a safety-of-flight reminder that a candidate burn must be screened, not assumed safe.
- This is a teaching and illustration tool. It is not an operational conjunction-assessment service and must not be used as one.
The public mathematics (illustrative)
All of the following is open-literature mathematics, reproduced for illustration. No proprietary operational algorithm is involved.
- B-plane (encounter plane): the plane normal to the relative velocity at closest approach. Projecting the relative position and covariance into this plane reduces the 3D conjunction to a 2D problem.
- Conjunction probability:
P_collision = P(z in disc(0, R))withz ~ N(m, C), wheremis the projected miss vector,Cthe projected covariance, andRthe combined hard-body radius. - Mahalanobis distance:
m^T C^-1 m, the squared statistical distance of the miss in the metric of the covariance. - Small-circle approximation and its closed-form relative error, obtained by expanding the density about the circle center:
P_full = f(0) * pi * R^2 * [ 1 + (R^2/8)(|C^-1 m|^2 - tr(C^-1)) ] + O(R^4). This closed form reproduces the paper's measured 0.039 percent relative error without evaluating the integral, and it yields a usable maximum-safe-radius bound. - Full circular integral: tensor-product Gauss-Legendre quadrature in polar coordinates, refined until successive orders agree to 1e-14 relative.
- Maneuver relations: a tangential burn produces an along-track drift
Delta s = -3 * Delta v_t * Tover coast timeT;Delta a / a = 2 * Delta v / vfrom vis-viva. - Toy lattice:
P_decode_fail = 1 - P(e in disc(0, lambda_1 / 2))withe ~ N(0, C), the same integral form as the conjunction probability. - Good versus bad basis: the two presets are related by a unimodular matrix (determinant +1), so they generate the IDENTICAL lattice; only the description changes, which is the asymmetry lattice cryptography rests on.
- Closest-vector problem: Babai rounding,
round(B^-1 t), decodes correctly with a short, near-orthogonal basis and returns the wrong point with the bad basis for the same target on the same lattice. The true-nearest answer is enumerated from the reduced basis, never the displayed one, so the demonstration cannot fake the effect it is showing.
Lattice point coloring uses a fixed 16-band spectrum for visualization only; it carries no proprietary content.
Correctness and behavior evidence
Graded FACT where verified; the input values driving the figures are illustrative SAMPLE values.
- The paper's computed quantities were independently checked. An automated verifier (
verify_leo.py) reports 33 checks, 0 mismatches. Geometry, the encounter-plane basis, the Mahalanobis distance, and the burn relations are verified in EXACT rational arithmetic; the two probabilities are confirmed by Gauss-Legendre quadrature AND by an independent second-order expansion that agree to 1e-6. Re-run on 2026-10-27: 33/33, 0 mismatches. [FACT, our bench, reproduced] - Paper worked example (ILLUSTRATIVE reference values, confirmed by the verifier):
| Quantity | Value |
|---|---|
| Time to closest approach t_CA | 100 s |
| Miss distance d_miss | 500 m |
| Mahalanobis distance squared d_M^2 | 6.25 |
| P_c, small-circle approximation | 1.09842e-4 |
| P_c, full integral | 1.09885e-4 |
| Relative difference | 0.039 percent |
| Along-track drift (worked maneuver) | about -1296 m |
[FACT that these match the verifier; values are an illustrative worked example, not an operational result]
- The true-nearest lattice oracle is basis-independent (enumerated from the reduced basis), so the good-basis and bad-basis answers are asserted to agree in the engine's own test, confirming the demo is not faking the Babai-failure it demonstrates. [FACT, design property]
- Every figure is computed in-browser with no network calls, no telemetry, and no external libraries. [FACT, from the demo source]
Measured performance
Not applicable in the throughput sense. The only performance-relevant facts are numerical: probabilities converge to 1e-14 relative between successive quadrature orders, and the 3D view is a compact in-page renderer rather than a bundled external engine. No latency or scale claim is made.
Client FAQ
1. Is this an operational collision-avoidance system? No. It is a toy and illustrative demonstration of the mathematics. It uses no real ephemeris and carries no proprietary operational algorithm, and it must not be used for operational decisions.
2. Are the numbers made up? No. The figures are computed live from the paper's own equations, and the paper's worked-example quantities were independently verified (33 checks, 0 mismatches) in exact rational arithmetic and by two independent quadratures. The input values you can change are illustrative samples.
3. What is the actual point of the two tabs? To show that a satellite conjunction probability and a lattice decoding-failure probability are the same integral over the same quadratic form. One engine computes both.
4. Why does the maneuver slider sometimes make things worse? Because a maneuver has a sign. Within the commandable range there is a burn that raises collision probability, since the along-track offset sweeps the relative position through the miss vector. It is a reminder that a candidate burn must be screened.
5. Does anything leave my browser? No. There are no network calls, no telemetry, and no external libraries; all computation is local to the page.
6. Is the cryptography side real? The lattice mathematics shown (unimodular basis equivalence, Babai rounding versus true nearest, the decoding-failure integral) is standard, open lattice theory presented as a teaching toy. It is not our production cryptographic implementation.
Qrypto Federal. DEMONSTRATION artifact; toy and illustrative, not operational; no real ephemeris. US English, ASCII-clean.
See it run