Plasma State Estimation
State estimation combines a model with noisy measurements to infer the plasma's true state, including quantities no sensor reads directly.
The estimation problem
A controller needs to know the current state of the plasma, but sensors are noisy and many important quantities cannot be measured directly. State estimation fuses a dynamic model with the available measurements to produce the best estimate of the full state, along with how uncertain that estimate is.
How estimators work
- A model predicts how the state should evolve.
- Measurements arrive and are compared to the prediction.
- The estimate is corrected toward the measurements, weighted by their reliability.
- The cycle repeats each time step, tracking the state continuously.
Handling nonlinearity
Plasmas are nonlinear, so simple linear estimators are not enough. Nonlinear filters and learned estimators handle the coupling, using reduced-order models fast enough to run in real time. The estimate feeds directly into control.
Seeing the unmeasurable
The value of estimation is inferring hidden states, such as internal current and pressure profiles, from what can be measured. For a tokamak this overlaps with equilibrium reconstruction; for the tandem-mirror burner it means inferring the plug conditions that set confinement.
Uncertainty aware
A good estimator reports not just the state but its uncertainty, so control and safety systems know how much to trust it, connecting to uncertainty-driven decisions.
Shared across machines
State estimation is the common thread linking control of both machines, the digital twin, and anomaly detection.