This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
The ideal craft had no collisions from weather, crowding or speed, but nobody flies one. Which conditions make a real airframe worse?
What was tested
Every condition was rerun for four vehicle types: the ideal point craft, multirotor, eVTOL and the slow-responding test craft, on two layouts each.
What was found then
On two layouts, crowding and link quality multiplied the penalty for realistic craft; weather did not (turbulence and storm came in below calm). The eVTOL's worst case was storm strength 10 with 1 message a second (0.67 collisions per 1,000 km). It was read as three things together: weak climbing, an out-of-date picture and storm flow. Wind 9 m/s looked safe only because the slow-responding fleet had stopped: 17 trips, against 1,223 in calm.
What we learned later
With six layouts (2026-09-11) the three-way storm-and-link interaction was withdrawn, turbulence came in at about 3 times calm rather than below it, and the ranking of crowding above the link did not hold.
research/envelope.md (Six seeds instead of two); research/envelope.md@e735ce4
What changed
Nothing.
Numbers and source references
- four
- envelopesresearch/envelope.md@9d286de (conjunction map)
- two seeds
- seed rollsresearch/envelope.md@9d286de
- 0.67
- eVTOL storm 10 + 1 Hz, per 1,000 km (two seeds)research/envelope.md@9d286de
- 17 trips against 1,223
- lagged fleet at wind 9 vs calmresearch/envelope.md@9d286de
- 1 Hz
- link rateresearch/envelope.md@9d286de
- 1,000 km
- rate denominatorresearch/envelope.md@9d286de