This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
If the rules assume more agility than the craft has, spacing could be widened to compensate.
What was tested
Two additions to the required spacing: one grows it as the craft's acceleration falls short of what the rules assume; one adds cruise speed times the craft's response time. Two vertical versions were also tried.
What was found then
The acceleration term cut collisions from 65 to 11 at half acceleration, for 18 % fewer trips. The response-time term at 1 s of lag cut pad-area cases from 375 to 17; doubling it gave 20 and lost more throughput, so once is enough. Both vertical versions made things worse. Together, each craft type won back 58–86 % of its extra collisions. The cost is steep: about a fifth of throughput for a multirotor, 42 % at 1 s of lag.
What we learned later
The response-time part became set per airframe: 0.75 of the full amount was adopted for craft with lag in cycle 148, then multirotor, slow-responding test craft and wing went back to the full amount (cycles 181, 182, 185), while eVTOL kept 0.75.
equilibrium/viz/sim_core.js:175-208; research/design/tau_bubble_stage_close.md
From 2026-09-14 the rules also issue each command early by the craft's known lag (cycles 77–84).
README.md (Planner lead for lagged airframes)
What changed
Required spacing widened for craft weaker than assumed (by the square root of the acceleration ratio) and by cruise speed times response lag plus delay; the ideal craft is exactly unchanged (adopted 2026-09-09).
Numbers and source references
- 65 → 11
- matrix, half accelerationREADME.md:1435; git 50f6ccb
- 565 → 45
- matrix, quarter accelerationREADME.md:1435
- 18 % and 37 %
- fewer legsREADME.md:1435
- 1 s
- lagREADME.md:1435
- 375 → 17
- terminal rows at 1 s lagREADME.md:1435
- 20
- with the time term doubledREADME.md:1435
- 45 %
- further throughput lost with it doubledREADME.md:1435
- 48 → 56, 48 → 66
- the two vertical versionsREADME.md:1435
- 58–86 %
- collisions recovered per envelopeREADME.md:1508; git 9af5e18
- 71–82
- storm-link collisions for multirotor, half and quarter accelerationREADME.md:1508
- 52
- point-mass storm-link floorREADME.md:1508
- 9 / 78
- half acceleration with termsREADME.md:1508
- 13 / 166
- full acceleration withoutREADME.md:1508
- a fifth
- throughput cost at the multirotor defaultREADME.md:1435
- 42 %
- throughput cost at 1 s lagREADME.md:1435
- 130
- storm-link, 1 s lag, with termsREADME.md:1508
- 116
- storm-link, eVTOL, with termsREADME.md:1508
- 212
- storm-link, lagged sideways-limited, with termsREADME.md:1508