Valhalla gives the fusion engine a repeatable test.
It generates synthetic sensor observations, records ground truth and scores EchoCore AI. The same seed produces byte-identical evidence.
- Synthetic data
- Repeatable runs
- Physics-based models
- Documented method
Valhalla models four sensor sources.
A seeded core sets the WGS84/ENU geometry and nanosecond time model for every run. Passive RF, acoustic, thermal and external radar models use that same state. Each run exports typed observations, per-field provenance, ground truth and a SHA-256 manifest.
- MODEL 01 PASSIVE RF
Passive RF
The passive radio-frequency model covers cooperative and emitting platforms. It uses the same seeded state as the other sensor models.
- MODEL 02 ACOUSTIC
Acoustic propagation
The acoustic model uses c_eff = 331.3 + 0.606·T + wind. The solver calculates delay and range from temperature and wind.
- MODEL 03 THERMAL
Thermal (pinhole)
The thermal model uses pinhole projection to keep the target position consistent with the sensor pose.
- MODEL 04 RADAR
External radar
The external radar feed supplies a fourth observation source for the fusion engine.
The scorer compares each result with ground truth.
Valhalla sends each dataset through EchoCore AI without changing it. Over an 8-seed sweep, the scorer measures track purity, cross-target contamination and clutter false tracks. These figures describe pipeline behaviour.
Plan the network before field work begins.
We want Valhalla to compare candidate node layouts against terrain, coverage overlap and confirmation time. The next step is a planning workflow that helps teams review trade-offs before deployment.