Computing Library › AI Plasma Control
AI Plasma Control

Layer 6: Scenario and Trajectory Supervision

The layer that schedules references through a discharge - ramp-up, flat-top, ramp-down - and adapts the plan as conditions evolve.

A discharge is a plan

Every shot follows a scenario: a planned trajectory of current, shape, density, and heating from breakdown through ramp-up, flat-top, and controlled ramp-down. The scenario layer holds that plan and issues the time-varying references the lower loops track. It is the layer that knows what the discharge is supposed to be doing at each moment.

Feedforward plus feedback

Kronos motion — reaching conditions

Scenarios are largely feedforward - a precomputed waveform of targets - but the layer also adjusts references based on the estimated state. If the plasma is denser than planned, it may delay a heating step; if a limit is approaching, it may flatten a ramp. This is supervisory control: it does not command actuators directly, it reshapes the goals the fast loops pursue.

Segments and events

Feasibility checking

Before it issues a reference, the scenario layer checks that the actuator-management layer can plausibly deliver it. Commanding an infeasible target - more shaping than the coils can provide, or a density above a stable limit - would push lower loops into saturation. The supervisor's job includes refusing to ask for the impossible.

Coordination with prediction

The scenario layer listens to the predictive digital twin. If the twin forecasts that the current plan leads toward a limit or a disruption, the supervisor selects a pre-validated alternative branch - a softer ramp, an earlier ramp-down - rather than continuing blindly. Plans are living, not scripts.