This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
A storm test found impacts rising with storm strength: none at strength 10, twenty at 22.
What was tested
Fourteen families, six layouts; we read the traces. Then we added a give-way rule. When wind drift exceeds the craft's top airspeed and a neighbour is heard within two required spacings, steering toward the goal fades, reaching zero at 1.5 times top airspeed. All effort then goes to keeping apart.
What was found then
Drift reached 15–36 m/s against a 9.6 m/s top airspeed, so both craft flew flat out on their own headings. With the give-way rule, the six layouts went from 25 collisions to 13 (storm 22: from 20 to 5), with work unchanged. Without the neighbour condition it cost 7–12 % of work. What remains is a reshuffle at the limit set by gust differences between neighbours. The full test set stayed at 0.
What changed
A give-way rule in storms stronger than the craft's top airspeed, only with a neighbour close.
Numbers and source references
- none at 10, two at 14, three at 18, twenty at 22
- impacts by storm strengthREADME.md:688
- Fourteen
- familiesREADME.md:688
- six seeds
- seedsREADME.md:688
- 15–36 m/s
- measured driftREADME.md:688
- 9.6 m/s
- airspeed capREADME.md:688
- 25 → 13
- six-seed storm sweepREADME.md:688; git 1643287
- 20 → 5
- storm 22 rowREADME.md:688
- 7–12 %
- work lost by the ungated yieldREADME.md:688
- −0.2 %
- work on the full matrix (run 22)README.md:688
- one envelope
- drift at which the yield startsREADME.md:688
- 1.5 envelopes
- drift at which goal steering reaches zeroREADME.md:688
- two standards
- neighbour distance that gates the yieldREADME.md:688