Planned full-motion ESA selection

Version 1 sought quiet flight and found no eligible stretches. It remains unchanged. This separate version will select intervals for the now-tested full-motion calculation, which reconstructs aircraft turns and climbs from the IMU (inertial measurement unit). It will not label those intervals level or quiet.

Rules to freeze before the new inventory

Keep the original source-integrity, timing-support and receiver-quality rules: split at receiver gaps above 0.75 seconds, sequence gaps, unsupported IMU timing or configuration changes. Require every retained receiver sample to have speed at least 180 km/h, reported horizontal standard deviations at most 5 m, height standard deviation at most 10 m, PDOP at most 4 and at least eight satellites. Retain maximal supported runs lasting at least 20 minutes and covering at least 60 km. Do not interpolate observations or join unsupported gaps.

Add validation of the full reported north/east/up covariance, using the shared error policy. Invalid matrices exclude the affected observations; a working uncertainty floor cannot repair them. Coordinates must support the nonpolar receiver chart used by the fitter.

Remove the quiet-motion veto and its maneuver margins in this version only. Report recorded vertical speed, heading changes, height range and route summaries. These describe sampled motion; they do not certify an unobserved instantaneous-motion bound. Later numerical and physical-adequacy checks remain mandatory. Selection guarantees neither successful fitting nor a scientific result.

Implementation and checks

Refactor the existing lll.airquantumgrav_selection implementation around a shared selection engine. Keep the public version-1 entry point and its default output unchanged. Add an explicit full-motion entry point with its own version and policy description; do not copy the complete selector or worker.

Test version-1 compatibility against its frozen implementation, using small deterministic trajectories. Test version-2 turns and climbs, covariance failures, chart support, quality boundaries, gaps, exact duration/distance boundaries and rejection reasons. Confirm that gyro values and model outcomes are not inputs.

Only after source-bound tests pass, freeze a new finite inventory on development recording 187. Reuse its established integrity/support evidence and preserve all rejected runs. No model fitting or protected-outcome inspection belongs to this inventory. The first observed pilot, its interval choice, weighting, starts, adequacy screens, thread allocation and finite allowance need a separate freeze.

Original version-1 policy · Generated full-motion checks · Roadmap.