SimOS
Deterministic, multi-fidelity, multi-agent simulation — the evaluation environment the rest of the platform stands on. Vehicles, sensors, effectors, and command-and-control run against each other in one scenario, at the fidelity each question deserves.
Deterministic to the byte
Same scenario, same seed — identical output, every time. A finished run replays byte-identically from its log: evaluation results that can be re-derived rather than re-argued, regression tests that mean something, and after-action review from the record itself instead of a screen recording.
Determinism is what turns a simulation from a demo into an experiment — change one variable, re-run, and every difference in the outcome is attributable.
Sensing is modeled, not assumed
Sensors report what statistical models let them see — detection, error, and degradation included — and a fusion layer maintains the track picture those reports support. The gap between truth, what the sensors report, and what the fused picture believes is a first-class output: it is the behaviour under evaluation, and it is watchable live in the raid engagement demonstration.
On the wire
The simulation speaks the standards the domain already uses: SAPIENT counter-UAS messaging, DIS for federation with existing simulators and range tools, Cursor-on-Target for TAK situational-awareness clients. Supplier models integrate as sealed plugins or standard FMI FMUs — no source hand-over.
From one run to a study
The same scenario machinery sweeps: hundreds of reproducible runs asking force-structure questions directly — how many stations, what spacing, how many interceptors — on the same models that later drive design and evaluation. The answer comes back as a trade curve, not an opinion. The perimeter sensing study is that machinery swept: 2,239 runs deriving a perimeter's sensor requirements from published performance bands.