Evolutionary Prediction Calibration
Compare predicted route performance with actual simulator execution. Clean repetitions reveal systematic time, distance and battery-model error before wind, LiDAR uncertainty or moving targets are introduced.
Insufficient evidence
3 more completed runs are needed before any correction is treated as provisional.
Prediction error
Mean absolute error shows typical miss size. P90 shows a conservative error band covering roughly nine out of ten recorded runs.
| Measure | Mean absolute | Mean bias | P90 absolute |
|---|---|---|---|
| Time | 0.0 s | +0.0 s | 0.0 s |
| Distance | 0.0 m | +0.0 m | 0.0 m |
| Battery reserve | 0.0 pp | +0.0 pp | 0.0 pp |
Advisory correction
These factors use 0 clean, completed, within-tolerance executions. Failed or interrupted records still affect reliability rates but do not distort normal execution bias.
- Time scale
- × 1.000
- Distance scale
- × 1.000
- Battery bias
- +0.0 pp
- Automatic use
- Disabled
Calibration evidence
Showing all historical development and campaign records. Records stay in this browser.
Safety interpretation
A calibrated browser model only means prediction error is understood in this frozen simulator. It does not establish real-world airworthiness, legal permission, GPS accuracy, obstacle detection, battery endurance or safe wildlife distance. Wind and LiDAR uncertainty remain deliberately disabled until this baseline is repeatable.