Eleven held haulers flown six ways: the release speed decides.

Eleven held heavy haulers, chosen by fixed rules from the previous cycle (so not a rate), were each flown six ways under two autopilot plannings. Changing only the speed at which the simulator lets them go, from rest to cruise, changed every ending, and the hauler then ended like a normal arrival in 21 of 22 cases. Letting it go a third of a second later changed one ending; the pad clearance check and bookkeeping changed none. Nothing is switched on.

The question

The previous CRAFT cycle found that the quick stopping-only planning made more held haulers end badly when the simulator released them from a standstill at the pad edge, and fewer when they were released like every other arrival. But the two releases differ in four things at once: the speed at the start, the moment of release (0.3 to 0.5 s later on the held path), the held path’s check that the pad area is clear plus its bookkeeping, and the simulation step on which the speed is set. Which of them changes how a hauler ends? Eight predictions were registered first.

The flight

The CRAFT campaign re-flew the previous cycle’s world (the Seventh Avenue pad, steady cross wind with gusts and turbulence at 1.5 m/s, the assumed slow hauler: 1.5 s to respond, 0.3 s delay). Eleven held arrivals were picked from that cycle’s runs by two fixed rules: ten whose ending flipped with the planning in the earlier pattern, and one lost on the way to the pad from rest but not when released normally. Each was flown under today’s planning and the quick stopping-only planning, six ways: from rest (as today); held but set to cruise toward the pad at release; held but started with the steady wind cancelled (turbulence not cancelled); the normal arrival; the normal arrival released at the held one’s moment; and the normal arrival started from rest. 132 flights of 45 simulated minutes, in the test harness only; the simulator itself was not changed.

What happened

Speed carried the difference on these cases. Let go at cruise instead of from rest, every ending changed (22 of 22) and the hauler ended like the normal arrival in 21 of 22; the normal arrival started from rest ended like the held one in 18 of 22. With it the planning’s effect flipped: from rest, today’s planning landed all ten flipped haulers and the stopping-only planning none; at cruise, today’s planning landed none and the stopping-only planning nine.

Releasing 0.3 to 0.4 s later moved the path by more than half a metre in 15 of 22 cases (at most 3.3 m) and changed one ending. The pad clearance check never delayed a release (the nearest other craft was 590 to 1,163 m away), and what remained between the two release paths moved the path by at most 0.29 m and changed no ending. A start that cancels the steady wind landed 18 of 22, against 10 from rest and 11 at cruise. All eight registered predictions held.

What this does not establish

Eleven predetermined but outcome-selected cases in one simulated world: descriptions, not a rate or a general rule. Timing and the pad check are not shown to be irrelevant; they stay open. The wind-cancelled start cancels only the uniform wind, and these arrivals start about 390 m out at about 63 m height, an approach entry rather than a lift-off; published guidance gives no release speed, so it is not a hover departure procedure. The FAA heliport and vertiport design guidance consulted by the realism reviewer (EB 105A) covers craft up to 5,670 kg; the 6-tonne hauler is about 6 per cent over that. The hauler’s response and the planning values remain assumptions. A first run failed one check on a wrong speed reference; the check was fixed and everything was re-flown with identical rows. The two main reviewers again disagreed on how to label the previous step, for the eighth cycle.

The next question

Nothing is adopted: the release variants exist only in the test harness and the planning values stay test keys. A release rule for held arrivals must specify the speed at release first; timing stays in it. Candidate next steps: an off-by-default setting for how held arrivals are released (from rest, wind-compensated or at cruise), measured on fresh seeds; or the registered margin-loss measurement.

Source references for reproduction

Optional repository references, pinned to this snapshot. Repository access may be required.