This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
Wings that cannot stall sat far outside their lanes whatever the switches did. Why?
What was tested
We traced craft on one layout; then, over twelve layouts, switched off the rule that makes craft match their neighbours' speed and direction, or strengthened the pull back toward the lane, in calm and dense lanes; also with ideal craft (they respond instantly).
What was found then
Wings start 42 m from the lane centre and drift out to 110–230 m, beyond the lane edge 80 m from its centre. Speed-matching is the cause: without it, the drift near lane ends drops from 153 to 63 m in calm. A stronger pull barely helps, and costs craft. Ideal craft drift for the same reason and get safer without the rule. But for crowded wings the rule protects: without it, 72 more craft are lost per run (+49 %).
What we learned later
The stage-1 summary keeps speed-matching as the named cause of wing drift off the lane (about 200 m in calm at the later default).
research/envelope.md (per-axis table, lane keeping row)
What changed
Nothing.
Corrections retained in the record
Earlier wording: the wing fleet in these runs is outside its 80 m lane box
Record correction: 80 m is the half-width (160 m full width)
research/reviews/brief-c47.md
Earlier wording: 9–21 more lost per twelve runs on the calm cell, 62–88 on the dense one
Record correction: these are per run
research/data/oos/wingband.md:33
Earlier wording: costs 72 more aircraft lost per twelve runs
Record correction: +72 per run (1 784 → 2 650 over twelve runs)
research/data/oos/wingband.md:35
Numbers and source references
- 5 s; 1 seed; 12 seeds
- trace interval, trace seeds, seeds per variantresearch/reviews/brief-c46.md
- ×2 / ×4
- pull-back cap multiplesresearch/reviews/brief-c47.md
- 42 m
- wing offset at launchresearch/reviews/brief-c46.md
- 110–230 m
- median offset later in the runresearch/reviews/brief-c46.md
- 80 m
- lane half-widthresearch/reviews/brief-c47.md
- 153 → 63 m; 192 → 80 m
- mean offset near lane ends, rule on vs off, calm / denseresearch/reviews/brief-c47.md
- +9 to +88
- extra wing craft lost per run with a stronger pull-backresearch/data/oos/wingband.md:33
- +72 (+49 %)
- extra wing craft lost per run, dense, rule offresearch/data/oos/wingband.md:35
- 84 → 43 m
- reference craft offset, rule full vs off, calmresearch/reviews/brief-c48.md
- 16 → 0
- reference craft collisions, dense, rule full vs offresearch/reviews/brief-c48.md
- 1.84 / 2.08
- wing vs reference off-lane ratio, calm / denseresearch/reviews/brief-c48.md