This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
Switching off speed-matching (craft copying neighbours' speed and direction) brings wings back to their lanes but costs crowded craft. Could partial strength give most of both?
What was tested
The rule at half and quarter strength on both wing types (twelve layouts), and on ideal craft (they respond instantly) on a lane grid with about 2.7 times normal traffic. We also counted close-call time.
What was found then
Half strength gives back about 70 % of the lane-keeping gain, a quarter over 90 %. But losses in dense traffic rise at each step: 19, 47 and 72 more craft per run at half, quarter and off. Close-call time rises too, so the rule is not just turning close calls into collisions. The 3 m/s wing makes the same trade. Ideal craft are safer at every weaker setting (36 collisions to 1 on the crowded grid). Nothing adopted.
What we learned later
The lane grid labelled 'four times traffic' was relabelled 2.67 times (capped at 1 600 craft).
research/envelope.md (label note, F-DENSE-LABEL)
The project's flight log was rebuilt to count every run; totals quoted in cycles 48–68 used the older, lower count.
research/flight_ledger.md (header)
What changed
Nothing.
Corrections retained in the record
Earlier wording: about 19, 47 and 72 more per twelve runs
Record correction: per run
research/data/oos/aligndose.md:34
Earlier wording: about 22 and 40 more aircraft lost per twelve runs
Record correction: per run
research/data/oos/dose3.md:26
Earlier wording: (the scenario is labelled on a lane cell four times as crowdedldquo;…rdquo; …)
Record correction: garbled HTML entities; intended: labelled ×4 but capped at 1 600 craft, 2.67× the 600 base
research/envelope.md (label note)
Numbers and source references
- 12
- seeds per settingresearch/reviews/brief-c49.md
- 1 600 / 600
- craft on the crowded vs base grid (2.7×)research/envelope.md (label note, F-DENSE-LABEL)
- ~70 % / >90 %
- lane gain recovered at half / quarter (67–71 % / 92–93 %)research/reviews/brief-c49.md
- +19 / +47 / +72
- extra wing craft lost per run, dense, half / quarter / offresearch/data/oos/aligndose.md:34
- 63 / 72 / 75 / 82
- near-miss pair-seconds per craft-hour, dense wing, full / half / quarter / offresearch/reviews/brief-c50.md
- +22 / +40
- extra 3 m/s wing craft lost per run, half / quarterresearch/data/oos/dose3.md:26
- 36 → 1
- reference craft collisions, crowded grid, rule on vs offresearch/reviews/brief-c49.md