Warning Console · reachable-set impact-point prediction

A maneuvering threat does not have an impact point. It has a footprint that shrinks.

Classical impact-point prediction assumes a single ballistic arc. A glide vehicle with a cross-range budget breaks that assumption on its first turn, and a solver that answers anyway returns a confident, precise, wrong point. This console propagates the whole reachable set instead, and shows the one thing that actually buys decision space: how fast the set collapses as the track is observed.

Modeled · 3-DOF Synthetic tracks Notional assets Seeded & reproducible
Concurrent tracks
Outside envelope
MC samples / track
Footprint collapse

Kill chain position

Find · Fix · Track · Predict · Assess

FindCue ingest
FixGeolocate
TrackCustody
PredictFootprint & IPP
TargetOut of scope
EngageOut of scope
AssessPost-event score
USE BOUNDARY. This surface predicts where a threat may land so that a defended asset can be warned. It does not generate aimpoints, firing solutions, weapon pairings, or engagement recommendations. Target and Engage are struck from the chain above deliberately, and nothing on this page will produce them.

Track picture

Concurrent tracks

Synthetic

Observation conditions

Degrade the picture

Every control here widens or tightens the footprint through the physics, not through a display multiplier. Degraded conditions should make the answer visibly less useful — if they don't, the uncertainty was decorative.

Nominal conditions.

Reachable set · impact-point prediction

Select a track

The shaded cloud is where this object can still reach. The rings are 50% and 90% containment. The marker is the modal cell.

Modeled
T+0:00press Play — watch the rule fire too late
50% area km²
50% semi-major km
Time to impact
Altitude km
Speed m/s
Modal support

ICD 203 / 206 assessment record

The judgment, not just the number

Decision rules

What this triggers, and who owns it

A probability with no bound action is a demo. Each rule below names the asset, the trigger, the action, the authority, and how long the action takes — so the console can check whether the decision window is long enough for the action it is recommending. When it isn't, the rule turns red instead of quietly recommending the impossible.

End-to-end timeline

Inference is not the long pole. It is barely a pole.

Sensor detection through to a unit actually beginning to move. Drag the operator-tempo control to see what changes the total and what doesn't.

Notional durations
Machine & network Model inference Human
Total chain
Model share
Human share
SegmentKindSeconds
Read this before optimising anything. Making inference ten times faster moves the total by a fraction of a percent. Making the operator handoff ten seconds shorter moves it more. A capability pitch that leads with model latency is optimising a term that does not control the outcome.

Known limits

What this console still does not do

Stated here rather than discovered by an evaluator. Anything below is a real gap, not a roadmap flourish.

No 6-DOF. Point-mass only. No attitude dynamics, control-surface model, structural or thermal model. Adequate for footprints, inadequate for anything about the vehicle itself.
No Earth rotation. Coriolis and transport terms are omitted. Over long-range trajectories this displaces impact points by a non-trivial amount in a known direction.
No powered phase. Boost and sustained cruise are outside an unpowered 3-DOF model. Cruise archetypes are approximated post-burnout only.
No discrimination. Decoys and debris are propagated, not discriminated. Telling a threat object from a penetration aid is the harder problem and is not solved here.
No real sensor feed. Tracks are synthetic and seeded. Nothing on this page is connected to a live sensor, and the latency model is notional.
No aimpoint motion. Defended assets are static points. A maneuvering ship or a relocatable node is not represented.