Typical recursion points
- State estimation influencing the controller, which changes the next observation distribution
- Online adaptation / tuning loops for learned policies
- Optimization loops that trade off objectives without explicit safety bounds
What boundedness buys you
Bounding the magnitude of updates reduces the chance of sudden excursions, and makes worst-case behavior more legible for validation.
Try it
Switch the demo to noisy sine and compare EMA vs GSRF for stability under noise and repeated feedback-like updates.
Run the demo with your industry preloaded
Launch a robotics-shaped synthetic signal model and presets. Filter math is unchanged.