AeroLoop · Engineering note

AeroLoop separates flight diagnosis from a new flight result

A ten-tool workspace inspects 24 retained flights without changing their gates. Horizontal feedback helps mission completion, while paired tracking and recovery still expose failures.

Source: https://louijiecompo.com/updates/aeroloop-flight-diagnosis/

Horizontal feedback quality · accepted: false

Quality	2 s missions	Pairs	Recovery
ideal	3/3	0/3	3/6
noise	3/3	0/3	3/6
delay	3/3	0/3	3/6
noise-delay	2/3	0/3	3/6

24 retained flights · 0 new flights in the diagnosis slice. Seeds 0/1/2 were previously tested. Mission, pair and recovery gates remain distinct.

Combined noise and delay, seed 0: -1.192 mm headroom against the original final support-position gate. Negative means failed.

Ideal seed 1, 2 s outages: altitude-feedback error in [40,42) s
Phase	Samples	Peak error
other	0	Unmeasured
airborne-descent	246	0.0694 m
contact	11	0.0742 m
motors-off	143	2.2082 m

Phase separation explains when error was measured. It does not establish cause or change any acceptance gate. Motors-off means zero requested thrust after landing; residual thrust can decay.

Download diagnosis JSON · Download quality study

PR #22 isolates which synthetic feedback channel stays fresh during two-second main-channel outages. Fresh horizontal feedback passes 6/6 outage missions across six previously tested seeds; fresh altitude and the retained baseline pass none. Only 1/6 horizontal pairs and 7/12 recovery windows pass. Mission success is not overall acceptance.

PR #23 tests ideal, noisy, delayed and combined horizontal feedback across seeds 0/1/2. Combined noise and delay fails one of three outage missions at the unchanged final support-position gate: 0.3511916 m against 0.350 m. Every quality fails its three paired checks. The matrix preserves those separate outcomes.

PR #24, merged on 6 October in Australia/Sydney, adds a case matrix, paired 3D, full-rate plots, event navigation, phase errors, feedback layers, gate headroom, recovery inspection, provenance and CSV export. Its 24 retained flights and 240,024 samples are the same AL-022 cohort, with zero new flights. Overall stress acceptance remains false.

Ideal seed 1 illustrates why phase matters: altitude-feedback error peaks near 0.0694 m during airborne descent and 2.2082 m with motors off in the fixed [40,42) s window. Contact can be intermittent. This split describes the retained trace; it does not prove cause or change a gate. The full 3D diagnosis bundle remains local to the source project. This portfolio reads public reports and does not rerun PhysX.

Follow the evidence

Explore AeroLoop

Read source evidence ↗ (opens in a new tab)

Follow the evidence

Related reading