This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
A sweep found collisions at 8 m spacing, in 3 m/s turbulence, on a link sending one message a second, though any two of those settings together were clean.
What was tested
We mapped spacings of 8–15 m against two turbulence levels and three links, on two grid families over four layouts. Then the required spacing was made to grow with the gust drift: how far gusts could push craft apart during the delay plus the gap between messages.
What was found then
At 2.5–3 s old, news in that turbulence was typically 9–11 m off in relative position, more than the 8 m spacing. Collisions fell to near zero once spacing passed about one gust drift (1.5 for the lane family). Setting the spacing to at least 1.5 gust drifts took the map from 1,796 impacts to 66, for 3.9 % more work. The full test set was unchanged.
What we learned later
On 2026-09-08 the rule was restated as a rule of thumb, not a proven need: its reasoning assumes independent gusts, while the default world now shares them, and the threshold came from two families on four layouts.
README.md:1244 (scoping points); git 9d9b4a1
What changed
Required spacing = the larger of the setting and 1.5 gust drifts over the delay plus the gap between messages; no effect without a link or turbulence.
Numbers and source references
- 8 m
- separationREADME.md:885
- 3 m/s
- turbulenceREADME.md:885
- 8–15 m
- separations mappedREADME.md:919
- four seeds
- seedsREADME.md:919
- 2.5–3 s
- message ageREADME.md:885
- 9–11 m
- relative dead-reckoning error, standard deviationREADME.md:885
- 1.5
- gust drifts adoptedREADME.md:967
- 1,796 → 233
- spacing map, one driftREADME.md:967
- 66
- spacing map, 1.5 driftsREADME.md:967; git 8ee3dd7
- +3.9 %
- workREADME.md:967
- 1
- impacts left at 12 m, grid family, turbulence 3, 1 HzREADME.md:919
- 12 m
- separationREADME.md:919