This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
The merge at the landing pad used up most of the safety margin. Would spacing out the arrivals help?
What was tested
We repeated the 150-craft pad test, releasing craft one per second, and compared it with sending them all at once.
What was found then
Close calls fell from 6,931 to 2,334. The closest approach rose from 0.30 to 0.49 of the required spacing (below 1 means a safety bubble was entered); higher is safer. Landing everyone took 37 % longer. Safety at the merge can be bought with time.
What we learned later
The all-at-once baseline here (6,931 close calls, closest approach 0.30) is the run with a ground floor; the first pad step quotes the earlier logged run without one (2,498 and 0.47).
results/run.log; results/results.json e5_funnel
What changed
Releasing craft one at a time became a test scenario (and later the one-at-a-time pad family in the JavaScript simulator).
Corrections retained in the record
Earlier wording: Slotted release cuts violations 2,498 → 1,184 and lifts minimum separation 0.47 → 0.61, for a 42% longer makespan (295 → 418 s). Figure caption: half the violations for 42% more time.
Record correction: The committed results are 6,931 → 2,334 breaches (about a third), minimum separation 0.30 → 0.49, and a 37 % longer makespan (243 → 332 s). The Flight Plan numbers match neither the committed JSON nor the README of the same commit.
results/results_phase4.json e10_metered_funnel; README.md@a2cdb52 (Findings so far)
Numbers and source references
- 150
- craft sent to the padresults/results_phase4.json e10_metered_funnel
- one per second
- release intervalexperiments/run_phase4.py@a2cdb52 (release_interval=1.0)
- 6,931
- breaches, free-for-allresults/results_phase4.json e10_metered_funnel.free
- 2,334
- breaches, meteredresults/results_phase4.json e10_metered_funnel.metered
- 0.30
- closest approach, freeresults/results_phase4.json e10_metered_funnel.free
- 0.49
- closest approach, metered (0.488)results/results_phase4.json e10_metered_funnel.metered
- 37 %
- longer time to land all craftREADME.md@a2cdb52 (Findings so far)
- 243 to 332 s
- time to land all craft, free vs meteredresults/results_phase4.json e10_metered_funnel makespan