Louijie Compo
Menu

AeroLoop · Engineering note

AeroLoop measures friction without claiming physics convergence

Nineteen fresh PhysX flights were attempted. Eighteen recordings verified: nine mission passes and nine failures. One timestep remains unverified.

Source: https://louijiecompo.com/updates/aeroloop-contact-physics/

Nineteen fresh PhysX flights were attempted. Eighteen recordings verified: nine mission passes and nine failures. One timestep remains unverified.

Seed 301 timestep sensitivity. At 5 ms the final support error is 0.476821 m; at 2.5 ms it is 0.445225 m. Both exceed the 0.350 m limit. The 1.25 ms capture is unverified and has no mission result.
FIG 01 / Same seed, ideal feedback and 200 Hz control. The finer step is missing verified evidence, not a zero-error result. Derived from the public validation record; no new simulation.
Download timestep figure · PNG

Rejected, not reclassified. The 1.25 ms capture reached 50 s, but its normal-force record contains -0.015082 N at 2.320 s. The original reader rejected it. It has no verified friction replay, work metrics or mission claim.

Simulation-only. Small momentum residuals establish accounting consistency, not contact-physics convergence or real-airframe accuracy.

Inspect pinned validation ↗ (opens in a new tab) · Download retained numerical report

Read the contact-physics investigation →

PR #31 adds filtered ground-friction diagnostics, contact-anchor inspection and independent impulse/work analysis. Nineteen fresh PhysX flights were attempted: eighteen recordings verified, with nine mission passes and nine failures. Twelve previously recorded trajectories match all 10,001 states exactly for those cases.

Independent arithmetic covers 180,000 control intervals. Measured friction sharply reduces horizontal momentum residuals, but center-of-mass work excludes rotational and contact-point terms. The different cohort does not replace the earlier 36-flight landing-response comparison.

The 1.25 ms seed-301 capture remains unverified after a negative normal-force sample. The original reader and mission gates are preserved. Contact-physics convergence and AL-010 remain open. The portfolio renders public summaries, not a new simulation or playback of local upstream flight bundles.

Explore the topic

Follow the evidence

Explore AeroLoop

Read source evidence ↗ (opens in a new tab)

Related reading