A deterministic core, four sensor models
A deterministic core (seeded PRNG, WGS84/ENU geometry, nanosecond time model) drives four sensor-physics models: passive RF, acoustic with a real propagation-delay solver (c_eff = 331.3 + 0.606·T + wind), thermal via pinhole projection, and external radar. Each run exports typed observations, per-field provenance, ground truth, and a SHA-256 manifest.
Passive RF
A passive radio-frequency model for cooperative and emitting platforms, driven by the same seeded core as every other sensor.
Acoustic propagation
Acoustic detection with a real propagation-delay solver: c_eff = 331.3 + 0.606·T + wind. Delay and range follow from the physics, not from a table.
Thermal (pinhole)
Thermal detections via pinhole projection, so a target's apparent position is geometrically consistent with the sensor pose.
External radar
An external radar feed, modeled as a fourth independent observation source the fusion engine has to reconcile.
The engine's evaluation oracle
Datasets run through EchoCore AI unmodified, and the scorer grades fusion output for track purity, cross-target and RID-spoof contamination, and clutter false tracks, over an 8-seed sweep against ground truth. Every number is synthetic pipeline behaviour, never detector performance.
Bifrost: a published DND counter-UAS challenge, modeled in our own simulator
DND publishes the rules for its counter-UAS challenges, so we built a representative one as an operational scenario inside Valhalla. Bifrost emits geometry- and threat-faithful synthetic test sets for that scenario and reproduces the structure of its red-team protocol as a Monte-Carlo simulation over synthetic trajectories. Modeling a published challenge in our own software lets us exercise the exact geometry and threats ahead of the field, and de-risk preparation against a known scenario instead of a guessed one. We model the scenario faithfully. We do not claim to meet any challenge outcomes, and every detection range is a modeled estimate, never measured hardware performance.
Scenario requirements modeled in Valhalla. Synthetic test sets only, not a claim of challenge compliance.
Seed of a deployment planner
Valhalla is a TRL 4-5 software testbed. The visualization layer is in active development toward the September 2026 demonstration at CFB Suffield.
Everything Valhalla produces is synthetic and labelled synthetic in field provenance. It never produces real sensor data, validated detector performance, or field evidence. Its claim is exact: physically consistent simulated detector outputs, for pipeline validation, doctrine testing, and demonstration.