Abstract: We propose a constraint-driven control architecture for plasma confinement that enforces forward invariance of a survivable state set ("confinement corridor") in real time. The approach replaces prediction-driven disruption avoidance with admissibility-based control, where control actions are selected to ensure the plasma state remains within a predefined constraint set under actuator and physical limits. The framework integrates constraint evaluation, precursor-sensitive instability detection, admissible route selection under reachability constraints, and constraint-consistent actuation in a closed loop. A local collapse horizon estimator provides a conservative first-order lower bound on time-to-violation under smooth dynamics. Unlike classical control barrier function (CBF) approaches, the architecture incorporates a route-selection layer under actuator-constrained reachability and a non-predictive precursor detection mechanism. This enables constraint enforcement under practical diagnostic and latency limitations. The contribution is conceptual and control-theoretic: plasma confinement is formalized as a forward-invariance problem in observable space under constrained dynamics. No claim is made of experimental validation or solved fusion. The framework is falsifiable and intended as a foundation for real-time constraint-consistent plasma control. Within this framework, admissibility is determined directly from Index Data and constraint definitions, eliminating the need for forward trajectory simulation in decision-making and real-time control.
Jorge Vasconcelos (2026) studied this question.