This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
With still air clean, the pad families were run in wind alone.
What was tested
Twenty wind variants, eight layouts each. Six pad behaviours changed, including spread the hover heights, holding at the spot where craft join the queue, clearing only when the gap is closing, steering to the pad relative to the air, and a spacing margin for turbulence.
What was found then
The wind tests went from 15 collisions to 0. Crosswinds that had landed almost nothing (0–5 craft) now landed 40–80 over eight layouts. Strong turbulence cost about a tenth of landings. At storm strength 10 over sixteen layouts, collisions went from 5 to 1. The survivor met a 17 m/s gust difference, beyond what the craft can fly. The full test set (every scenario family and variant, about 400 cases) went from 2 to 1, with the work score (trips and cargo delivered) down 0.44 %.
What changed
Spread hover heights, holding at the joining spot, clearance only when the gap is closing, pad steering relative to the air, and a turbulence spacing margin.
Corrections retained in the record
Earlier wording: wind 9 and wind 13 crosswind cases, which landed nothing, land forty to eighty craft per run
Record correction: They landed 0–5 before and 40–80 over 8 seeds after (the record does not say per run).
README.md:421
Numbers and source references
- Twenty
- pad wind variantsREADME.md:421
- eight seeds
- per variantREADME.md:421
- 15 collisions to 0
- wind setREADME.md:421; git b229b69
- 0–5
- craft landed in the crosswind rows beforeREADME.md:421
- 40–80
- craft landed in those rows after, over eight seedsREADME.md:421
- a tenth
- landings lost in strong turbulenceREADME.md:421
- 5 → 1
- storm 10 over sixteen seedsREADME.md:421
- 17 m/s
- gust difference in the surviving caseREADME.md:421
- 2 → 1
- full-matrix collisions (run 17)README.md:421
- −0.44 %
- workREADME.md:421