This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
Every airframe so far could stop in the air. A fixed wing cannot, and the map had nothing on it.
What was tested
New wing versions: the slow-responding airframe plus a minimum speed (5 or 3 m/s, against 8 m/s cruise). The simulator either holds the wing there or crashes it (a stall) if it drops well below. Six layouts, calm crossing grid.
What was found then
A wing held at minimum speed collides about as often as the slow-responding airframe (3.8 against 3.9 per 1000 km). But it sits at its 5 m/s minimum a quarter of the time. If stalling crashes the craft, the rules lose 83 % of the fleet in ten minutes at 5 m/s and 51 % at 3 m/s, from flying too slowly, not from collisions. We read this then as rules made for hovering craft braking a plane.
What we learned later
The 'braking for traffic' reading was not supported: most slow moments had no one nearby.
research/reviews/brief-c26.md
The rule that makes craft match their neighbours' speed and direction was named as a main cause of the slowing; later (cycle 34) the rule that turns craft onto their lane heading was added.
research/reviews/brief-c29.md
What changed
Four wing airframe versions added (checks reproduced the existing airframes' results exactly).
Numbers and source references
- 5 / 3 / 8 m/s
- minimum airspeeds and cruiseresearch/reviews/brief-c25.md
- 0.8 ×
- stall crash threshold as a share of minimum speedresearch/reviews/brief-c25.md
- 6 × 600 s
- seeds × run lengthresearch/reviews/brief-c25.md
- 3.80 vs 3.93
- collisions per 1000 km, wing held at floor vs slow airframeresearch/reviews/brief-c25.md
- 25 %
- share of craft-steps the 5 m/s wing is held at its floorresearch/reviews/brief-c25.md
- 83 % / 51 %
- fleet lost to stalls in ten minutes at the 5 / 3 m/s floor (2 974 / 1 853 of 3 600)research/reviews/brief-c25.md