This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
Every craft had been an ideal point able to accelerate at 4 m/s² instantly in any direction. That was now the weakest assumption.
What was tested
We built a vehicle model: separate up, down and sideways limits, climb and descent rates, a turn limit, response lag, command delay, and climbing costing sideways push. Then we changed one limit at a time around a multirotor, on 145 test cases and on the storm-on-slow-link cases.
What was found then
The ideal craft had 0 collisions on the test cases; a realistic multirotor had 13. Halving acceleration gave 65; quartering it, 565. Response lag was worse: 1 s gave 627 and 2 s gave 2,583. Going from 0.4 to 0.5 s of lag alone took collisions from 13 to 78. The dodge is right but arrives too late. Whole airframes: eVTOL 304, slow-responding test craft 1,262.
What we learned later
The slow-responding test craft was not a fixed wing. A real wing with a minimum airspeed was built later and stalled much of the fleet on a calm grid.
research/envelope.md (What would falsify this)
What changed
Vehicle model and airframe types added; the default stays the ideal point craft (results exactly unchanged).
Numbers and source references
- 4 m/s²
- assumed accelerationREADME.md:1388
- 145
- matrix rows per cellREADME.md:1388
- 0 / 52
- point mass, matrix / storm-linkREADME.md:1377, :1388
- 13 / 166
- multirotor, matrix / storm-linkREADME.md:1388
- 65
- half accelerationREADME.md:1388
- 565
- quarter accelerationREADME.md:1388
- 627
- 1 s lagREADME.md:1388
- 2,583
- 2 s lagREADME.md:1388
- 536
- 0.5 s delayREADME.md:1388
- 0.4 to 0.5 s
- lag stepREADME.md:1388
- 13 to 78
- collisions across that stepREADME.md:1388
- 494
- storm-link, climb acceleration 0.5README.md:1388
- 304
- eVTOL preset, matrixREADME.md:1388
- 1,262
- lagged sideways-limited preset, matrixREADME.md:1388