This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
A third of storm-on-slow-link impacts were at ground level. Neighbours extend a craft's last heard motion in a straight line, so a grounded craft was guessed to be still descending, 12–33 m below ground.
What was tested
Two fixes were tried: cap the guessed height at the ground; and keep each craft's height command during the storm give-way rule (which fades goal steering when the wind is stronger than the craft).
What was found then
The cap changed nothing overall (six layouts: 340 to 342). It halved one case (41 to 22) but worsened the 30 m/s case (183 to 204). Keeping height was worse (173 to 190): craft holding against the downdraft hit their speed limit and lost room to dodge sideways. Neither was adopted.
What changed
Nothing.
Numbers and source references
- A third
- share of storm-link impacts at ground levelREADME.md:1186
- 12–33 m
- heard depth below groundREADME.md:1186
- 340 → 342
- six seeds with the clampREADME.md:1186; git 2b99817
- 41 → 22
- separation-8 storm rowREADME.md:1186
- 183 → 204
- speed-30 storm rowREADME.md:1186
- 173 → 190
- three seeds with altitude keptREADME.md:1186
- 30 m/s
- speed rowREADME.md:1186