Real-Time State Estimation

A twin is only useful if it tracks the real machine — state estimation fuses noisy sensors into a best current estimate, continuously.

In brief

• Combines diagnostics (magnetics, interferometry, spectroscopy) into a coherent plasma state.
• Uses Kalman/particle filters and ML observers.
• Runs inside the control loop's latency budget.
• Feeds both the controller and the predictive twin.

The detail

Dozens of diagnostics each see the plasma partially and noisily. State estimation is the mathematics — Kalman filters, particle filters, and learned observers — that fuses them into a single best estimate of the plasma's shape, current, density, and temperature, fast enough to act on. It is the bridge between raw sensors and a twin that can be trusted.