Raid Engagement
A multi-aircraft raid against a defended area, simulated end to end: sensors produce detections, the fusion layer maintains a live track picture, and command-and-control tasks the defence — all from one deterministic run. The same log that drives the after-action replay below drives every analysis of this engagement, byte-identically.
What you are watching: truth (the aircraft as they actually fly), what the sensors report, and what the fused picture believes — three different things, drawn from the same run. The gap between them, and how quickly it closes, is the behaviour under evaluation.
The replay is not a recording of a screen — it is the run log itself, re-rendered. Re-running the scenario with the same seed reproduces this engagement exactly.