Local Level Lab

Airborne measurements. Open methods.

Checked local steps for training-compatible alternatives

Qualified first-step controls.

Usable steps were found in 12/12 directions. Each direction tests three fixed step sizes, keeping the original 27 calibration coordinates and bounds.

Saved work: 36/36 native batches, 72/72 linear programs and 144/288 exact proposal checks. Reporting performs no new integrations.

Model Processing case North direction Radius .001 Radius .0005 Radius .00025
Globe + field Zero removal Negative Usable Usable Usable
Globe + field Zero removal Positive Usable Usable Usable
Globe + field Profiled removal Negative Usable Usable Usable
Globe + field Profiled removal Positive Usable Usable Usable
Globe, zero field Zero removal Negative Usable Usable Usable
Globe, zero field Zero removal Positive Usable Usable Usable
Globe, zero field Profiled removal Negative Usable Usable Usable
Globe, zero field Profiled removal Positive Usable Usable Usable
Stationary disc Zero removal Negative Rejected Usable Usable
Stationary disc Zero removal Positive Usable Usable Usable
Stationary disc Profiled removal Negative Rejected Usable Usable
Stationary disc Profiled removal Positive Usable Usable Usable

What this establishes

These are generated trajectories, not aircraft results. A usable step must satisfy the actual earlier position limits and make the predicted directional progress. A rejected step remains in the record.

Earlier accepted probes remain valid: their north spreads are 13.33 m and 5.90 m. All six accepted and six rejected original labels were rechecked without a new native evaluation.

Passing this trial qualifies only the first-step component. It does not establish a complete optimizer, outer prediction bounds, a calibrated error model or adequate aircraft predictions. V01 remains unqualified. The 53-stretch numerical findings are unchanged.

Frozen method and limits · Prediction requirements