This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
If the wing drifts far from its lane, perpendicular lanes overlap widely. Is drift why it loses so many craft in calm air?
What was tested
48 new calm runs held the wing nearer its lane: heading copying (the rule that copies neighbours' heading) switched off, or a narrower corridor to be pulled back into. Drift was also read from existing runs.
What was found then
In calm the wing flies about 200 m off its lane centre on average; the ideal craft (instant response) flies 82 m off. With heading copying off the wing drifts only 64 m, yet still loses 532 per 12 runs; half-strength copying gives 614. The narrower corridor barely narrows it (199 → 177 m), and losses stay near 560. The ideal craft loses nothing in similar geometry. The conclusion then: not a lane-keeping problem, but the slow airframe failing to dodge at crossings.
What we learned later
Withdrawn: 'colliding craft sit at ordinary distances from their lane'; the pooled comparison used could not show that.
research/envelope.md:2642
Narrowed: 'slow response' was what was left after the drift test failed, not a demonstrated cause; drift as a contributing cause was not ruled out.
research/envelope.md:2697
What changed
Nothing.
Numbers and source references
- 200 m (199)
- calm wing's mean offset from its lane centrelineresearch/envelope.md:2642
- 82 m
- ideal craft's mean offset on the same runsresearch/envelope.md:2642
- 85 %
- perpendicular-lane collisions whose nearest drawn crossing is the pair's own (85.5 %)research/envelope.md:2642
- 48
- new calm runs in the narrowing testresearch/envelope.md:2697
- 64 m
- wing's mean offset with alignment off (63.6)research/envelope.md:2697
- 532
- wing craft lost per 12 runs, with and without alignmentresearch/envelope.md:2697
- 614
- wing craft lost per 12 runs at half alignmentresearch/envelope.md:2697
- 199 → 177 m
- wing's offset as the lane box is narrowedresearch/envelope.md:2697
- 560
- losses per 12 runs with the narrowed box (570 and 560)research/envelope.md:2697