This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
How can a fast emergency craft get through traffic? The first idea was plain right of way: when two craft meet, the ordinary craft dodges harder and the emergency craft less.
What was tested
100 half-hour runs in two test skies (a grid of crossing lanes and a layered lane network), with the emergency craft at 1.5× and 2× traffic speed, with and without right of way.
What was found then
The emergency craft was slow: with right of way the wing took about twice the straight-flight time on the grid. Only 1 of 8 test cases met the target of 1.2 times, and right of way made it faster on none. One fast craft cost the ordinary traffic little (5–9 % more losses on the grid, none measurable on the layered network). Right of way qualified for adoption nowhere.
What we learned later
Scored again with the statistic fixed in advance, right of way was not faster on 6 of 8 test cases; on the other 2 it was 0.01–0.06 faster, within the run-to-run scatter.
research/design/prio_delay_record.md (F-PRIO-DELAY-ESTIMAND)
The suspects for the delay were narrowed. These test cases have no radio link, so a rule that holds speed when position reports are old could not be involved; the craft matching its neighbours' speed was named as the likely cause.
research/design/prio_family_record.md (Where the delay comes from)
What changed
A right-of-way switch and separate loss counters for the emergency craft and ordinary craft were added, off by default; nothing adopted.
Corrections retained in the record
Earlier wording: The reservation test fired on four cells of eight, by margins of one to five percent.
Record correction: The 1–5 % excesses are the grid-wing cells only; the other two triggering cells had excesses of +13.5 % and ×1.39, on very small counts.
research/design/prio_family_record.md (corridor question)
Numbers and source references
- 100
- half-hour runs in the familyresearch/design/prio_family_record.md
- five
- seeds per test caseresearch/design/prio_family_record.md
- 2.009 / 2.006
- grid wing lap time vs straight-line time with right of way (1.5× / 2×)research/design/prio_family_record.md
- 1.464 / 1.515
- airspace wing lap ratio with right of wayresearch/design/prio_family_record.md
- 1 of 8
- cells meeting the 1.2× bar (airspace point 2×, 1.113)research/design/prio_family_record.md
- 5–9 %
- extra ordinary losses on the grid wing from inserting the fast craftresearch/design/prio_family_record.md
- 0 of 8
- cells where right of way was adoption-eligibleresearch/design/prio_stage_close.md §3