Why recursion breaks “normal” safety
When outputs feed back into inputs, small misestimates compound. A system can appear stable for many iterations and then cross a threshold where the state becomes unrecoverable.
- Amplification: error can grow superlinearly under feedback.
- Delayed detection: monitors often trigger after the unsafe state is already reached.
- Overshoot: transient excursions can be catastrophic in safety-critical systems.
What “deterministic” means here
Deterministic safety is a structural guarantee: if an update would violate a bound, it is rejected or modified before it affects system state. This is different from probabilistic alarms that are only likely to catch drift.
Best next step
Start with the demo to get intuition, then read the paper abstracts for the formal framing.
Run the demo with your industry preloaded
Launch the demo with a domain-shaped synthetic signal model and presets. Filter math is unchanged.